GLOREAM - dmu.dk · e-mail: [email protected] Scientific Secretary Annette Münzenberg Institute for...

GLOREAM

Global and Regional Atmospheric Modelling

Annual Report 1998

Names of steering committee

Sub-project co-ordinator Deputies

Peter J.H. BuiltjesTNO-MEP, Department of Environmental QualityP.O. Box 3427300 AH Apeldoorn, the Netherlandstel.: 31 55 549 3038fax: 31 55 549 3252e-mail: [email protected]

Adolf EbelUniversity of CologneInst. for Geophysics and Meteorology EURADAachenerstrasse 201-209D-50923 Cologne, Germanytel.: 49 221 400 2258fax: 49 221 400 2320e-mail: [email protected]

Hans FeichterMax Planck Institute for MeteorologyDept. Theoretical Climate ModellingBundesstrasse 55D-20146 Hamburg, Germanytel.: 49 40 41 17 33 17fax: 49 40 41 17 32 98e-mail: [email protected]

Members

Erik BergeThe Norwegian Meteorological Institute (DNMI)P.O. Box 43BlindernN-0313 Oslo, Norwaytel.: 47 22963000fax: 47 22963250e-mail: [email protected]

Carlos BorregoUniversity of AveiroDepartment of Environment and PlanningP-3810 Aveiro, Portugaltel.: 351 34 370617fax: 351 34 428600e-mail: [email protected]

Rainer FriedrichInstitute of Energy Economics and the RationalUse of EnergyHessbrühlstr. 49aD-70565 Stuttgart, Germanytel.: 49 711 78061-12fax: 49 711 7803953e-mail: [email protected]

Heinz HassFord Research Center AachenSüsterfeldstr. 200D-52072 Aachen, Germanytel.: 49 24 1942 1203fax: 49 24 1942 1301e-mail: [email protected]

Anne LindskogSwedish Environmental Research Institute - IVLP.O. Box 47086S-40258 Göteborg, Swedentel.: 46 31460080fax: 46 31482180e-mail: [email protected]

Nicolas MoussiopoulosAristolie University of ThessalonikiLab. of Heat Transfer & Environmental Eng.Box 483Thessaloniki 54006, Macedonia, Greecetel.: 30 31 99 60 11 / 30 94 34 36 00fax: 30 31 9960 12e-mail: [email protected]

Eberhard SchallerBrandenburgische Technische UniversitätHaus 215, Burger Chaussee 2D-03044 Cottbus, Germanytel.: 49 35 55 78 13 105fax: 49 35 55 78 13 128e-mail: [email protected]

Zahari ZlatevNational Environmental Research Institute, (NERI)Department of Atmospheric EnvironmentP.O. Box 358DK-4000 Roskilde, Denmarktel.: 45 45 30 1149fax: 45 46 20 1214e-mail: [email protected]

Scientific Secretary

Annette MünzenbergInstitute for Tropospheric ResearchPermoserstr. 15D-04318 Leipzig, Germanytel.: 49 341 235 2176fax: 49 341 235 2139e-mail: [email protected]

International Scientific Secretariat (ISS), München 1999

GLOREAM-report

2 of 216

Contents

I Summary

II The aims of the period’s work

III Model investigation and improvement

IV Global modelling

V Model application and assessment studies

VI Computational aspects

VII Model evaluation and validation

VIII Overview of policy relevant results

IX General aims for the coming year

X Names and details of current principal investigators

XI Authors and titles of theses

XII Publications in refereed literature

XIII Reports from the principal investigators

GLOREAM-report

3 of 216

I Summary

The most interesting new developments in the GLOREAM subproject in 1998were:

- The capabilities of a number of models to perform long term averagecalculations, f.e. focussed on AOT-40 and AOT-60 over Europe, haveincreased substantially,

- the capabilities of some model systems to perform real-time ozone forecastinghave increased substantially,

- there is an increased knowledge concerning the modelling of aerosols,

- progress has been made in the area of data-assimilation for reactive species,

- work has been started to come to a general model evaluation and validationactivity.

The second GLOREAM workshop was held in september 1998 in Madrid. It wasattended by nearly all PI's. Proceedings of this workshop are available.

At the beginning of 1998 the project had 41 individual contributions. In the coarseof 1998, 3 new contributions were added and 4 were withdrawn, so the totalnumber of contributions by the end of 1998 was 40, with 38 different PI's.

Over 1998, most cooperation with other EUROTRAC-subprojects was withSATURN and GENEMIS, and some cooperation with CMD, BIATEX and TOR-II.

Nearly all the funding for the different projects in 1998 was by national funding.Only very little funding was by the 4th frame work program of the EU, and also insome countries the national funding was very marginal.

GLOREAM-report

4 of 216

II The aims of the period's work.

The most important aim over this period was to create an active and enthusiasticgroup of people, including young scientists, who want to work and interact witheach other in the frame work of GLOREAM. In view of the fact that theGLOREAM-workshop in Madrid was well attended, and that (nearly) all PI's havesend their annual report, it can be concluded that this aim has been reached.

The general aim of GLOREAM is to investigate by means of advanced andintegrated modelling the processes and phenomena which determine the chemicalcomposition of the troposphere over Europe and on a global scale.

The associated tasks are:

- the development and improvement of three-dimensional regional and globalscale atmospheric transport-chemistry models,

- the application of complex and simplified models for specific environmentalpolicy issues and the assistance of other Eurotrac-II subprojects.

In principle all these tasks have been adressed over this period, but the assistanceof other projects has been rather limited.

GLOREAM has been split up in 5 working groups:

- Model investigation and improvement

- Global modelling

- Model application and assessment

- Computational aspects

- Model evaluation and validation

The first three working groups address different aspects and foci of GLOREAM,the last two working groups are more of a generic nature and address in principleall projects. The results over 1998 are described according to the 5 workinggroups.

GLOREAM-report

5 of 216

III Model investigation and improvement

III.1 Activities during the year

With respect to the aims of task 1, work has been done in the improvement of inputdata, parameterization of chemical and dynamical mechanisms and theinvestigations in terms of scale interaction, operational use and real-time ozoneforecasting and development of data assimilation.

Incorporation of improved input data

Emission data used in the modeling systems participating in GLOREAM are in apermanent process of improvement by intensive interaction with the EUROTRACsubproject GENEMIS and activities in the topic centre air emissions of the EEA.Specific improvements of input data has been undertaken with respect to land useand topographic information. A Geographic Information System (GIS) has beenused for this purpose together with a Relational Databank Management System(RDMBS). The GIS/RDMBS has been used to improve the interface to airpollution models (Smiatek and Stockwell, 1998).

Improvement and investigation of process parameterization

The research has been focussed on the importance of exchange processes in theconvective boundary layer for modeling of atmospheric compounds using LargeEddy Simulation (LES; Petersen; Petersen and Holtslag, 1999), the effect ofimproved chemical mechanisms on the prediction of regional and globalatmospheric chemistry models using box models(Stockwell; Seefeld andStockwell, 1999) and the development of a new particulate model to study theimpact of traffic emissions on air quality (Hass/Ackermann; Ackermann et al.,1998). Processes studies have been undertaken with respect to the impact of UVfluctuations inferred from total ozone and tropospheric aerosol (EMEP model,Jonson et al., 1999), the sensitivity of the model results on clouds (REM3, Reimer;Fleming et al., 1999), budget calculations and analysis of dynamical and chemicalprocesses on the basis of episodic model simulations (EURAD, Memmesheimer etal., 1998).

Model hierarchy, linking of different scales; operational use

Models in GLOREAM now usually use the nesting technique to consider regionaland local scales (Europe → urban). Application of the nesting technique has beenused to simulate episodes where field campaigns have been done (e.g. EURADwith respect to BERLIOZ, PIPAPO and VOTALP).

GLOREAM-report

6 of 216

Some models have been applied to seasonal or annual time scale, e.g. the EMEPEulerian photochemistry model (HIRLAM-CTM, Jonson), the Dutch LOTOSmodel (Builtjes), the Danish Eulerian model (Zlatev), the Swedish Match model(Langner), and the models which are planned to be used operationally or alreadyhave been used in an operational way, e.g. the DMU-ATMI-THOR air pollutionforecast system (Brandt; Brandt et al., 1999), REM3 (Reimer) and the modelsystem of the German Weather Service (LM coupled with EURAD-CTM and theemission model ECM; Rissmann/Jacobsen).

Data assimilation and numerical techniques

Several advection schemes have been investigated within the framework of theBulgarian Air Pollution modeling system (EMAP, NIMH; Syrakov, 1998).Considerable work has been undertaken in the further development of dataassimilation techniques including adjoint modeling (Elbern and Schmitt, 1999).Data assimilation techniques have been used to improve the initial values ofmodels. Also the extended Kalman filter data assimilation method has been furtherimproved with main emphasis on noise values for specific key parameters andinput data like emissions (Segers et al.)

III.2 Principal results

Incorporation of improved input data

Two methods for compiling landuse data and plant species have been developedbased on GIS/RDMBS technologies. A plant-specific database has been compiledfor Europe together with an interface which allows for the aggregation of the data(landuse, emission factors for plant species, biogenic emissions) to grid cells of airpollution modeling systems. The GIS allows for different map projections and thusimproves the link to different air pollution modeling systems considerably.

Improvement and investigation of process parameterization

Based on mass flux characteristics calculated by large eddy simulations (LES) inthe convective boundary layer, a simple covariance parameterization has beendeveloped which can be used in atmospheric chemistry models. Based on runswith a single column version of the model it has been found that the covarianceclosure compares well with LES (Petersen).

Box model studies using the RACM chemical mechanism and a simple depositionscheme show the importance of land use for ozone formation and the nitrogenbalance. Due to lower deposition over water, ozone concentrations over water arehigher and show only a minor decrease after sunset (Stockwell).

GLOREAM-report

7 of 216

The EURAD-MADE system (Hass/Ackermann) has been applied to a summerepisode using nesting (horizontal grid sizes 27-9-3-1 km). The aerosol model hasbeen extended for consideration of the formation of secondary organic particles ofbiogenic and anthropogenic origin. Anthropogenic and biogenic secondaryorganics show approximately the same maximum concentration but differentspatial distribution.

The REM3 model has been used to investigate the impact of diagnostic andprognostic cloud coverage data on photolysis frequencies. Based on data fromMETEOSAT it has been shown that the use of diagnostic data leads to apronounced reduction of cloud coverage compared to prognostic values from theEuropa model of the German Weather Service.

Model hierarchy, linking of different scales; operational use

Nesting techniques have been used to couple different scales from Europe to localscales. Applications involve the BERLIOZ field campaign to investigate ozoneformation in the plume of Berlin (EURAD, Ebel) and particulate matter inNordrhein-Westfalen. Typical grid sizes are 1 to 2 km. Process analysis on thebasis of modeling show the impact of dynamical and chemical processes for ozoneformation. Ozone formation in urban areas seem to be most effective in an altituderegion of several hundred meters. It has been shown that the contribution ofdifferent processes to the tendency terms in the mass continuity equation can differconsiderably, even if the concentrations are similar.

Parts of the models have been used in operational application for trace gasforecast. Operational use clearly shows the need for improved input data(emissions, landuse) for the actual situation, the incorporation of the actualmeteorological situation in the emission models and the need to improve theparametrization of cloud processes.

Data assimilation and numerical techniques

An episode in August 1997 has been selected to test a 4D variational dataassimilation scheme with the aim to optimize the initial concentrations used inCTM simulations (EURAD, Elbern and Schmitt, 1999). It has been shown that theinitial values obtained by the data assimilation scheme lead to an improvedsubsequent forecast. However, due to insufficiently known emissions, depositionrates and model bias, the influence of the optimized initial values usually decreaseswith time.

New advection schemes have been created on the basis of the TRAP approach forestimating the mass fluxes at the edges of a grid cell using different approximatingpolynomials and fitting pattern (Syrakov and Galperin, 1998). This scheme isfaster compared to the original Bott scheme.

GLOREAM-report

8 of 216

III.3 Main Conclusions

Within 1998 considerable progress has been achieved in the handling and qualityof input data, development, implementation and investigation of parametrizationsfor atmospheric models, in the extension of episodic modeling to seasonal orannual applications and the use of complex atmospheric models for operationalforecast.

An interface between air pollution modeling systems and biogenic emissionsgenerated within a Geographical Information System has been developed andtested for complex regional model applications. This will contribute to the furtherimprovement of the quality of the results obtained by chemistry-transport models.

Based on box model calculations, the importance of chemical conversion anddeposition for the concentration of nitrogen species has been performed.

Aerosols have been included in episodic regional modeling including applicationsto a summer episode. This offers the opportunity to study the source contributionsto aerosol load in the atmosphere which is an important and valuable extension ofpure photochemical modeling - also with respect to the planned EU directives.

Numerical techniques (advection schemes, data assimilation) have been improvedconsiderably. In particular, advanced data assimilation techniques based on adjointmodeling of CTMs offer the possiblity to merge observations and model results.The use of satellite data is of specific interest in the application of dataassimilation.

Operational use of CTMs and long term runs are new fields of application ofatmospheric models. This offers new possibilities for science and environmentalplanning due to the large amount of model results generated. But it also pointstowards the need that new methods have to be developed to analyse the data and toevaluate the results of long-term runs of models with measurements taken on aroutine basis for an seasonal or annual scale.

III.4 Policy relevant results

In general, the considerable development achieved with respect to input data andprogress in process modeling (e.g. aerosols) leads to further improvement ofcomplex models as a tool for the planning of air pollution abatement strategies. Inparticular, the possibility to perform long term runs provides the possibility toestimate AOT40 or AOT60 values and their response on emission reductionscenarios. The recently developed modules for the simulation of atmosphericparticulates can be used with respect to the planned EU directives byenvironmental agencies and industry.

GLOREAM-report

9 of 216

III.5 Aims for the following year

Aims for the future are a further improvement of input data (landuse, plant speciescategories, emission factors; anthropogenic emissions). This will be achieved bystronger interaction with GENEMIS and BIATEX. In particular, more informationis needed with respect to particle emissions (size distribution, composition).

Chemical process modeling will be continued and supported by sensitivity studies.Interaction with CMD is planned to improve gas phase chemistry as well asheterogenous chemistry.

Dynamical processes in the planetary boundary layer are planned to beincorporated in a more sophisticated way into the air quality models used inGLOREAM.

Emission data sets that distinguish between source categories will be used toevaluate source type contributions to the tropospheric aerosol loading and toquantify the contributions of primary and secondary particles to this load.

Process analysis and budget studies will be performed to investigate theimportance of different terms in the mass continuity equation for the concentrationfields over Europe and polluted subregions.

Some models will be extended from regional to local scales and even coupled withstreet canyon modeling. Interaction with SATURN will be enhanced in the nextyears with respect to these activities.

Chemistry-transport models will be optimized for operational use to perform tracegas forecasts. Longterm runs are planned with respect to EU-directives.

GLOREAM-report

10 of 216

IV Global modelling

IV.1 Activities during the year and principal results

Eight groups presented papers dealing with global aspects. Most of them focus onozone chemistry in the troposphere.

Law et al. simulated an episode in 1997 using the chemistry transport model(CTM) TOMCAT driven by ECMWF analyses. The model included the oxidationchemistry of methane and two NMHCs. Calculated concentrations were comparedto aircraft measurements and it was shown that the model is capable to simulatethe seasonal cycle of ozone in the free troposphere. Toupance et al. studied theinfluence of the chemistry of anthropogenic NMHCs on the distributions of ozone,nitrogen oxides and nitrogen reservoir species by using the CTM IMAGES. Theyhave shown that it is important to represent NMHC in chemistry models tocorrectly simulate the NOX and the nitrogen reservoir species particularly in thecontinental boundary layer. Zimmermann estimated the contribution of differentsources to the boundary layer ozone concentrations over Europe using the CTMMOGUNTIA. Industry and traffic are the most efficient sources of ozoneprecursors contributing by more than 50% to the ozone concentrations. Nextimportant is the downward transport of ozone from the troposphere contributing tomore than 20 % of the ozone concentrations in the boundary layer.

Transport of ozone from the stratosphere to the troposphere controls the ozoneconcentrations particularly in the upper troposphere but due to the small scale ofthe relevant processes, it is not very well represented in global CTMs. E.g. Law etal. reported an overestimate of calculated ozone concentrations in the uppertroposphere due to the smearing out of sharp gradients at the tropopause. Wernliand Bourqui studied the transport processes in the tropopause region over Europeby analysing a one year sample of ECMWF analyses and by applying a Lagrangianapproach. A pronounced winter maximum was found for deep stratosphericintrusions descending lower than 700 hPa and it was shown that the ozonedownward flux is about 50% larger in spring than in autumn.

Kelder et al. and Feichter and Lammel calculated global distributions of theconcentrations of sulfur dioxide and sulfate and compared the results to surfacemeasurements over Europe and the US. Both found relatively good agreementbetween calculated and observed sulfate concentrations but an overprediction ofthe SO2-concentrations in winter.

Results of a climate response experiment with a coupled ocean-atmosphere-chemistry model studying the effects of increasing greenhouse gas and sulfurdioxide emissions was presented by Feichter and Lammel. The climate response is

GLOREAM-report

11 of 216

similar, but weaker, if aerosol effects are included in addition to greenhouse gases.One notable difference to previous experiments is that the intensity of the globalhydrological cycle becomes weaker in a warmer climate if both direct and indirectaerosol effects are included in addition to the greenhouse gases. Borrego et al.interprete such climate model predictions on a smaller scale, the Great LisbonArea, by using the mesoscale model MEMO coupled to a photochemical model.The authors show that changes in the vertical temperature profiles in a warmerclimate has an impact on the ozone distribution. A similar downscaling strategywas pursued by Langmann et al. who developed a hierarchy of models of differentscales (ECHAM)-REMO-GESIMA and applied a one-way nesting technique. Firstsimulations showed that the model hierachy can realistically reproduceobservations of meteorological and chemical variables. The results also indicatethat the parameterization of cloud convective processes as well as theironline/offline determination plays an important role in the redistribution of tracegases.Such model hierarchies can be useful for the interpretation of climate studieson a smaller scale and for the analyses of measurement campaigns.

IV.2 Main conclusions

A number of global atmospheric transport-chemistry models focussing on thetroposphere are available and operational in GLOREAM. These models areprimarily used for scenario studies and budget calculations. Stratosphere-troposphere exchange and vice versa can be analysed using ECMWF data. Modelresults are sensitive to a proper description of cloud convective processes. Forglobal tropospheric O3 and NOX budget studies, next to CO and CH4 also thehigher NMHC should be taken into account. Climate change can have an impacton air quality and should be taken into account in scenario studies.

IV.3 Policy relevant results

The scenario studies, showing for the future (up to about 2050) an increase intropospheric ozone and on the global scale also an increase of the sulphateconcentrations and nitrate concentrations, are of direct policy relevance. The sameholds for the ozone budget studies over Europe which give an indication of the byEuropean emissions produced ozone over Europe. The fact that there are clearindications that climate change has an impact on air quality situations is ofrelevance for the analysis of future air quality.

IV.4 Aims for the coming year

There is obviously a need to improve the temporal resolution of the emissioninventories and to take into account more species. In particular, estimates of thehydrocarbon emissions are higly uncertain. Simulations of specific periods andevaluation of the models by comparing the results to observations will be a maintask in the next year.

GLOREAM-report

12 of 216

V Model application and assessment studies

V.1 Activities during the year

Trajectory calculations and nesting techniques were introduced in the Eta modelfor air pollutants long-range transport over Europe. Accuracy and sensitivityanalyses were done, namely on precipitation forecasts for further refinement ofmodel's physical package (Lazic). Various sensitivity studies to the model weredone, including effects of mountains, boundary layer and above BL verticalexchanges and moist processes effects in long range transport of air pollutants.

The Danish Eulerian Hemispheric Model has been improved considerably by thecoupling of a state-of-the-art weather prediction model to the model system(Christensen). A new hemispheric model with a photochemical scheme consistingof 55 species, more than 94 chemical reactions and 17 photolysis reactions are inprogress of development.

Long term and long range ozone formation and transport were calculated with theLOTOS model on a continental scale, making use of data assimilation techniquesto integrate ozone measurements both ground level and remote. The target area isall Europe, the simulated period of time refers to 1994 and the objective is todetermine and quantify the processes of ozone formation (Builtjes a)).

In order to set up an air pollution modelling system to be used on different spaceand time scales, the research unit headed by Calori has developed a new version ofthe STEM-II model and a turbulence and deposition pre-processor. Preliminarytests of the long-term version of the STEM-II were done over the Italian Peninsula.The model was also applied to a mesoscale domain covering the LombardiaRegion, simulating an ozone episode.

In the core of model testing methodology and testing studies AOT40 LOTOSresults were compared with other model results and with measured data(Builtjes b)).

On the mesoscale, simulations were done with the KAMM/DRAIS model. Themodel was evaluated for the 16th of September 1994, included in the TRACTexperiment, and the 26th July 1994, a day of a summer smog episode inNorthrhine-Westfalia (NRW). The resulting data was compared with aircraftmeasurements. The considered evaluation parameter was the magnitude of thedifference between measured and simulated values. (Fiedler et al.)

Assessment studies on future and present emission scenarios were done byMensink et al. and Keller et al.. Mensink et al. have carried out an integrated

GLOREAM-report

13 of 216

assessment of acidification, eutrophisation and ozone formation in Flandres,Belgium. The EUROS chemistry module was tested aiming towards thedevelopment of a compact chemistry module that allows fast computations ofscenarios for policy support. The aircraft campaign of air pollutant measurement“Transboundary Flux Measurements for Photochemical Model Validation inFlanders” - TFLUX - was carried out with the objective to support policy oncontrol high ozone episodes.

Keller et al. have tested the UAM (Urban Airshed Model) with a field campaign,the POLLUMET experiment. The assessment studies were focussed on the impactof future scenarios of Swiss energy systems. Four energy/emission scenarios weredefined and the emission inventory was extrapolated for the year 2030. Scenario 1is based on the replacement of power plants by oil driven combined cycles;scenarios 2 and 4 consider the reduction of 30 % on traffic emissions in the wholecountry, cities and highways, separately. The meteorology used on the calculationsof the impacts of these energy scenarios has resulted from a statistical analysis, byclassification into a few categories with similar pollution categories.

In the domain of public use of internet resources, a JAVA user interface wasincorporated to the OPANA software tool (an air quality modelling system), bySan José et al.. This development aims to allow users to perform all Internetprocesses automatically so that the ARL web server requirements are transparentfor the OPANA user.

V.2 Main results

Improvements were achieved in the simulation of the local Bora wind, over theDynaric Alps, by the application of a nesting technique (Lazic).

In the frame of the workshop A Comparison Of The Performance Of Large ScaleModels In Simulating Atmospheric Sulphate Aerosols(Cosam), operated underWCRP/WGNE and IGAC/GIM, 10 different large scale modelling groups fromboth North America and Europe have compared model results. The results fromthe Danish Eulerian Hemispheric Model show a good performance compared withother model groups' results and with observations (Christensen).

The main results from the ozone budget calculations between the free troposphereand the boundary layer (BL) with the LOTOS model show an inhomogenity in theozone formation over Europe. On the Iberian Peninsula there is a small inflowfrom the free troposphere into the BL, whereas in central and southern Europe anoutflow of O3 from the BL into the free troposphere can be observed. Budgetcalculations between April and September of 1994 show an inflow of O3 from thefree troposphere into the BL in the North and over the Mediterranean Sea.

GLOREAM-report

14 of 216

The application of the Karman filter data assimilation technique indicates VOCanthropogenic emissions underestimation over northwestern Europe of 30 %(Builtjes a)).

The Eulerian 3D model based on the STEM-II has been rewritten to allow thechoice between algorithms for simulation processes. The use of differentcoordinate systems is now available as well different advection algorithms, RAMSmodel interface and direct treatment of large point sources. Two chemicalmechanisms are considered, the simpler EMEP-II and a more complete one (withgas phase module and explicitly isoprene chemistry). The model output dataincludes hourly gas and aqueous phase concentrations, domain balances for userselected species, wet and dry deposition fluxes (Calori et al.). It was demonstratedthat testing atmospheric chemistry-transport models is useful to make a distinctionbetween diagnostic and integrated model testing.

It is necessary to know and establish dry deposition processes and AOT40measurement heights before comparing AOT40 model results and measured data(Builtjes b)).

Results from the integrated assessment studies on emission scenarios in Flandersshow that a more cost effective SO2 emission scenario will lead to a dominant roleof ammonia in acidification and eutrofization processes and emission reductionson this compound will be necessary. In terms of AOT60 and AOT40, the studyshows that a 50 % reduction on NOX and VOC emissions is needed to get the sameamount of reduction on the AOT60 values and for the AOT40 the effect is lesspronounced. Comparison between the costs of the ozone and acidificationabatement strategies shows a relatively high value for the acidification strategy.

Concerning the results from the KAMM/DRAIS mesoscale model (Fiedler et al.),evaluation of the 16th of September 1994 with aircraft measurements shows thatCO concentrations are underpredicted, which can indicate an underestimation ofCO emissions as input into the model. This is also observed in a scatter diagram ofNOX concentrations and agrees with the scatter diagram between NOX

concentrations simulated and measured at the ground level. However, ozone levelscomparison provides a better agreement of the mean level, also at ground level.The evaluation of KAMM/DRAIS performance on simulating ozone concentrationlevels over the NRW area shows that in areas with low emissions of ozoneprecursors simulated ozone concentrations are in better agreement with groundlevel measurements than in the other areas, where the model underestimates thepeak ozone values. A sensitivity analysis done on NOX and VOC emissions, wherethe NOX emissions input decreases and the input model VOC emissions increase,enhance the model performance on ozone concentration calculations.

GLOREAM-report

15 of 216

The evaluation of the EUROS chemical mechanism indicates that the compactscheme deviates significantly from the EMEP/MSC-W scheme. Optimizationschemes are being implemented in the EUROS mechanism.

A database of the data measured during the TFLUX was build (Mensink).

The UAM testing with data collected on the POLLUMET experiment shows goodagreement on rural stations and the time of the ozone peak simulation isreproduced on urban and suburban measurement stations. A sensitivity analysis ofthe grid cells to NOx and VOC emissions shows that ozone formation is sensitiveto NOX or to both emissions. The VOC-sensitive areas appear to be depending onwind regime, generally with low wind speeds. The UAM assessment studies on theimpact of future scenarios of Swiss energy systems show an increase of the NO2

and SO2 levels, as well as the short term deposition of SO2 and NO2 and thedeposition of HNO3 in the case of the replacement of nuclear power plants by oil-driven CCP facilities. NOX traffic emission reductions as initial conditions lead toan O3 level decrease in the countryside and an increase near the cities. In thisscenario there is a decrease in the deposition of SO2, HNO3 and NO2. For the VOCemission reduction scenario, ozone deposition will increase near the big cities suchGeneva and Zurich (Keller).

V.3 Main conclusions

The performance of the meteorological synoptic/mesoscale Eta model is now at ahigh state-of-the-art level, particularly regarding precipitation forecast. The efforton adding to this model trajectory calculation and nesting techniques has allowedobtaining reasonable results on the application of the system on the Saharan dusttransport to Europe. A sensitivity study of the treatment of the flow over complexterrain has shown to be very important to accurate transport and diffusion tracersimulations and also for long range transport of constituents (Lazic).

The performance of the large scale Danish Eulerian Hemispheric Model wasimproved with the developments done in the weather prediction module(Christensen).

In order to analyse regional differences in tropospheric ozone budget calculations,LOTOS model results were found to be a suitable method. The improvementsobtained with the implementation of data assimilation techniques for non-linearchemistry into the model are a powerful tool to combine knowledge andinformation contained in measurements and model formulations (Builtjes a)).These kinds of developments are very important for the exchange of informationon air pollutant emissions and transboundary concentration fluxes betweencountries for limited area models. Applications turn out to be more realistic(Calori et al.).

GLOREAM-report

16 of 216

Integrated model testing directed towards AOT40 requires further knowledge ofthe representativeness of the observations and the proper description of theprocesses close to the surface as dry deposition (Builtjes b)).

Again in the field of model evaluation, it was found that the KAMM/DRAISmesoscale model provides realistic results. Nevertheless, focus should be done ona reliable emission inventory, especially in areas where a smog episode may beexpected (Fiedler et al.).

Assessment studies on emission scenarios for policy abatement strategy on airpollutants applied to Belgium, although having high uncertainties, shows that thecosts for the reductions needed for the ozone strategy are relatively low whencompared to the costs needed to realise the proposed acidification strategy(Mensink). Keller et al. also have done assessment studies on energy/emissionscenarios in Switzerland with a mesoscale photochemical model. Calculationsmade on the imposed scenarios lead to the following conclusions: the replacementof the existing nuclear power plant by oil-driven combined cycle plants can havesignificant local effects, especially on SO2 concentrations and deposition andshort-term critical levels for plants; reducing the traffic emissions in the wholecountry seems to be more effective than measures limited to the cities or thehighways.

Friendly graphical interfaces developed on the basis of the JAVA programminglanguage make the access to air quality operational model systems easier for thegovernmental institutions, as well as to general public (San José et al.).

V.4 Policy relevant results

The assessment studies performed by Keller et al. and Mensink et al. will be ofinterest for Swiss and Belgium decision-makers, respectively. In the Swiss case,the results cover economical and ecological aspects, health risk and other impacts,for Belgium to support policy on the control of high ozone episodes, acidificationand eutrophisation.

V.5 Aims for the following year

Assessment of the effects of various modelling options will be performed with theEta model on the Saharan dust transport case. Sensitivity analysis of the initialsubgrid scale diffusion also will be done (Lazic).

The LOTOS model will be extended to contain an aerosol module. Modelcalculations including data-assimilation will be used to study the regionaldifferences in tropospheric ozone and aerosols over Europe. (Builtjes a))

GLOREAM-report

17 of 216

The long version of STEM-II will be tested with the two chemical mechanisms intwo different time scales: for whole 1994, with the EMEP mechanism, and forselected episodes, using the more detailed chemical mechanism. The inclusion ofan aerosol module will be attempted. This STEM-II long version will be alsoapplied on the analysis of future sustainable emission abatement scenarios overItaly (Calori et al.).

Because integrated model testing requires the use of proper statistics, research todetermine the suited statistical parameters will be performed. The research inintegrated atmospheric chemistry transport model testing of ozone and AOT40 willbe continued (Builtjes b)).

Validation of the KAMM/DRAIS mesoscale model will be continued with themeasured data of the FLuMOB Experiment. Concerning the NRW case, betteremission data will be applied (Fiedler et al.).

The new version of the UAM-V model will be tested. In order to investigate thetransboundary air pollutant transport and its impact on Switzerland, LOTOS andEURAD results will be applied as boundary conditions. Wet deposition, and laterparticulate matter, will be taken into account in the simulation of air qualitycategories (Keller et al.).

Development and implementation of the EUROS model may apply for policysupport with respect to tropospheric ozone in Belgium. Also, the analysis of thedata collected in the field experiment TFLUX will be programmed (Mensinket al.).

GLOREAM-report

18 of 216

VI Computational aspects

Large mathematical models, in which all important physical and chemicalprocesses are adequately described, can successfully be used to resolve many tasksconnected to the preservation of the concentrations and/or the depositions ofharmful air pollutants under certain prescribed critical levels. However, the use oflarge mathematical models in which all important physical and chemical processesare adequately described leads, after the application of appropriate discretizationand splitting procedures, to the treatment of huge computational tasks: in a typicalsimulation one has to perform several hundred runs, in each of these runs one hasto carry out several thousand timesteps and at each timestep one has to solvenumerically systems of ODE's containing up to several million equations.Therefore, it is difficult to treat numerically such large mathematical models evenwhen modern computers are available. One has to select appropriate numericalalgorithms which are both fast and sufficiently accurate. Moreover, the selectedalgorithms must be tuned for efficient runs on modern high-speed computers. Ifthese two tasks are successfully solved, then modellers will be able to treat muchmore efficiently at least the following three important tasks: (i) to run the model onlonger time-periods, (ii) to carry out more verifications of the model and (iii) toimplement more advanced mechanisms in the description of the physical andchemical processes. This short discussion of the importance of the computationalaspects in air pollution modelling explains why these issues were treated in manyof the GLOREAM annual reports. In some reports, Barone, Knoth, Langner,Maker, Seibert and Zlatev, the computational aspects were the major part of theexposition.

The mathematical description, the software tools and the parallel performance ofthe PNAM model are discussed in the report of G. Barone and his co-workers.Special techniques for parallel runs on an IBM SP computer are used in thecomputer code of this model. This enables the modellers to run a very complicatedthree dimensional code in reasonable CPU timings.

The MATCH model is discussed in the report of J. Langner (this model is fullydescribed in a report of the Swedish Meteorological and Hydrological Institute).The use of a fast numerical algorithm allows the modeller to run the model over aspace domain covering the whole of Europe on rather long time-periods as well asto carry out sensitivity studies (the variation of model results due to variation ofinput data and physical mechanisms). The studied period is a six months period in1994 (from April to September).

The chemical part of a large-scale air pollution model is normally the most time-consuming part when the model is run on computers. Therefore, the task of findingfast and sufficiently accurate chemical mechanisms is very important. This issue is

GLOREAM-report

19 of 216

treated in the report of P. A. Makar. The solution of this task by reduction of thenumber of variables is discussed there. Moreover, the task of creating simplifiedemission databases from highly detailed, specialised emissions is also treated inthis report.

Sometimes it is important to be able to treat different inverse problems. Such aproblem, using inverse dispersion modelling as a tool to derive emission data frommeasurements, is discussed in the report of P. Seibert and her co-workers. Theinverse problems are often very huge and very ill-conditioned (in the sense thatsmall perturbation of the input data used in these models leads to big differences inthe output results). Therefore, the search for efficient and fast algorithms also hereis crucial. An algorithm based on a simple regularisation technique proved to beefficient.

The use of fast and parallel algorithms was an important tool in the attempts toimprove the performance of the Danish Eulerian Model (the report of Z. Zlatevand his co-workers). Standard software tools (MPI: Message Passing Interface)were used in the preparation of a highly parallelized version of this model, whichafter that has been run over a time period of 10 years (1989-1998) and to calculateresults concerning exceedances of several critical levels. Also an attempt toperform different tests, aiming the validation of the model result by comparingthem with measurements, was carried out.

In summary, the computational issues are well represented in the individual annualreports (not only in the reports listed above). It was shown that after resolvingsome of the most important computational problems,

- one can try to improve the description of the physical and chemical processesin the models,

- one can solve more tasks and/or bigger tasks,

- one can carry out long simulations with different scenarios in order to studythe response of the models to variations of key parameters,

- one can start to run operationally some of the models in an attempt to predictexceedance of critical levels (as, for example, ozone critical levels) in the nexttwo to three days

and

- one can start developing advanced control strategies for keeping theconcentrations and/or the depositions of harmful pollutants under the criticallevels.

GLOREAM-report

20 of 216

VII Model evaluation and validation

Model evaluation and validation, the performance of quality assurance/qualitycontrol, is essential for all models developed and applied in GLOREAM. Beforemodels can be used in a reliable way to investigate processes and to performscenario studies, it has to be known what the models are intended for, and what themodel performance is relative to observations. All models in GLOREAM havebeen compared to some extent to observations to obtain a first idea of the modelbehavior. However, the effort spent on model evaluation differs strongly from oneproject to another, and also the methods used differ. There is a need forharmonisation.

Model evaluation (or validation, testing, verification) of atmospheric transport-chemistry models has a number of essential elements. One element is that not onlythe model itself, but also the model input as emissions and meteorology has to beevaluated, and also the observations have to be evaluated. Another element is thefact that non-linear chemistry is used, which means that there can be an agreementbetween observations and model results, so the model can be 'right', but for thewrong reasons. Again another element is the problem of comparing volumeaverage calculations with point measurements.

During the Eurotrac symposium in March 1998 a workshop was held to discussmodel evaluation, which was attended by about 30 participants who exchangedinformation and experience.

There have been several interactions between SATURN and GLOREAM to cometo a general philosophy concerning model evaluation, and with the final aim tocome to a coherent and generally accepted model evaluation methodology andprotocol.

In several GLOREAM-projects model evaluation has been adressed. In discussingthe set-up of a model evaluation methodology, it is proposed to distinguishbetween diagnostic model testing which is process oriented, and integrated modeltesting , which is directed to the overall model. As an example of integrated modeltesting the problems with model-evaluation concerning AOT-40 is discussed(Builtjes).

Model evaluation of the KAMM/DRAIS model using the TRACT-campaign aspart of the German TFS program showed that it is important to define a 'qualitymeasure' of a model for a certain variable as the percentage of cases which fallwithin a certain difference range. As an example for the TRACT-case, a range of5 ppb for ozone was set, resulting in a value of 40 % of the cases that fall withinthis 5 ppb difference for KAMM/DRAIS (Nester).

GLOREAM-report

21 of 216

A major problem in model evaluation is the determination of the spatialrepresentativeness of measuring stations. In two related studies (Tilmes,Zimmermann) work has been done to develop methods for the quantification of therepresentativeness of air quality data. This has been done by analysing theobservations by performing principle component analysis on the basis of the timevariation of the available ozone data. By applying filtering, a seasonal and anepisodic and an averaged diurnal cycle can be separated. Empirical orthogonalfunctions (EOF), which are the eigenvectors of the system, can also be calculated.In this way the amount of variance which is caused by the first, second etc. EOFcan be determined. By comparing the results of different stations, the spatialinfluence and representativeness of the EOFs can be obtained.

Model evaluation and validation will remain a key topic within GLOREAM. Incooperation with SATURN an attempt will be made to come to a general protocol.Ideas will be formulated in GLOREAM to come to a general test data base fortesting GLOREAM models.

GLOREAM-report

22 of 216

VIII Overview of policy relevant results

The modeling activity within GLOREAM covers a wide range of scales rangingfrom global to province scale, down to about 100x100 km2. Furthermore, a widerange of topics is covered in the project. Several of the contributions areaddressing topics that are of direct political relevance. Other contributions focuson subjects that are not of immediate political relevance, but rather on scientificquestions aiming to improve our understanding of the atmosphere, that at a laterstage will enable the scientific community to provide more accurate answers to thepoliticians. In the following we have summed up the most policy relevant studiesin GLOREAM 1998 and sorted them into the categories global, regional andprovince scale.

In global models the effects of several source categories and mechanisms havebeen studied with emphasis on tropospheric ozone. Topics of direct politicalrelevance include apportionment of the contribution of different sources to surfaceozone in Europe and the contribution of European sources to free troposphericozone. Studies of stratosphere-troposphere exchange are also included. The mostconcrete policy relevant study is on the impact of greenhouse warming on the airpollution levels in Lisbon, Portugal, as examined C. Borrego et al..

The largest number of policy related studies in GLOREAM are found on theregional scale. Long-term modelling and evaluation by use of measurementsshould be directly applicable to the on-going work under the Convention on LongRange Transboundary Air Pollution (LRTAP). P. Builtjes gives an approach tomodel validation. Establishing a general accepted methodology for model testingand validation is of large importance in order to assess the quality of the modelsemployed for policy making. Long-term regional studies of ozone distributions inEurope have been conducted by P. Builtjes (TNO), J.E. Jonson et al.(EMEP/MSC-W) and Z. Zlatev et al. (DMU). Model studies of ozone andacidifying compounds also applicable to regional scale policy making are given byJ. Langner (SMHI). Preliminary studies on particulate matter show that a reductionof inorganic aerosol precursor substances does not necessarily lead to a similarreduction of inorganic particular matter (J. Ackermann et al.). C. Mensink et al.have used integrated assessment studies to analyse the future depositions ofacidifying compounds in Belgium based on the emission scenarios of the LRTAP-protocols. A possible powerful policy tool in the future is the inverse modelling ofemissions as presented by P. Seibert et al..

A number of the contributions address the forecasting of surface ozone. Accordingto the so called "ozone directive", public warnings should be given when highozone events are expected. Several approaches are chosen for this purpose. Severalcontributions apply nesting of local models with horizontal resolution of the order

GLOREAM-report

23 of 216

of one km into (in this context) coarse models with a much larger model domain.New improved methods for operational forcasts of ozone are found in S. Kiilsholmet al. (Danish Meteorological Institute), J. Rissmann et al. (German WeatherService) and J. Brandt (NERI, Denmark).

Several contributions focus on the modeling of province scale air pollutionincluding industrial and urbanised regions. Such information may serve as a guideto political decision making and the results are also valuable to analyse costeffective abatement strategies in the urban area. Results from a 3-Dmeteorological/photochemical forecast system for ozone has been presented forthe Naples region by G. Barone et al.. Such data could in the future be valuable forfuture air quality regulations. Furthermore, the studies of the interaction betweenregional and province scale pollution has been carried out for the Berlin region byM. Memmesheimer et al.. Assessments of the impact of different future emissionscenarios from Swiss Energy Systems on the air pollution levels over the Swissplane have been conducted by J. Keller.

It should be noted that these policy relevant results, like the calculations made forAOT40 and AOT60, are relevant both for national and international environmentalagencies as well as for the industry. It is remarkable that the very relevant topic ofthe evaluation of proposed abatement strategies by performing scenario runs ishardly addressed within GLOREAM at the moment.

GLOREAM-report

24 of 216

IX General aims for the coming year

The more detailed aims over 1999 are already described in the work packagesmodel investigation and improvement, global modelling and model application andassessment. Main activities are expected in the areas of long term modelling,aerosol modelling, real-time ozone forecasting and data assimilation.

Concerning computational aspects, the development into parallel computing andhigh performance computing and networking will continue.

Further developments in the area of model evaluation and validation will be on themodel evaluation protocol, about proper statistical indices and about simply doingit and document it properly.

Finally, the already existing cooperation with SATURN and GENEMIS will befurther strengthened, and special attention will be given to the cooperation withCMD, BIATEX and TOR-II.

Closing remark

This general overview of GLOREAM has been made by the steering committee.The editing of the complete annual report has been done by GLOREAM'sscientific secretary, Annette Münzenberg.

Michael Memmesheimer and Adolf Ebel were especially responsible for thechapter on model investigation and improvement. Ana Cristina Carvalho andCarlos Borrego were especially responsible for the chapter on model applicationsand assessment studies. The part about global modelling has been prepared byHans Feichter. Zahari Zlatev wrote the chapter on computational aspects and thechapter of the overview of policy relevant results was made by Jan Eiof Jonsonand Erik Berge. The remaining text has been written by Peter Builtjes.

GLOREAM-report

25 of 216

X Names and details of current principal investigators

Yvonne Andersson-Skold

Melica, Fjaelly.3E

S-41317 Gothenburg, Sweden

tel. 46 31 14 22 70

fax: 46 31 77 56 950

e-mail: [email protected]

Guido Barone

Dept. of Chemistry

University of Naples & Interdept.

Center for Research on the Environment

(CIRAM)

Via Mezzocannone 4

80133 Naples, Italy

tel.: 39 81 54 76 502

fax: 39 81 55 27 771


Annemarie Bastrup-Birk

National Environment Research Institute

Dept. of Atmospheric Environment

Frederiksborgvej 399

P.O. Box 358

DK-4000 Roskilde, Denmark

tel.: 45 46 30 1178

fax: 45 46 30 12 14


Erik Berge

The Norwegian Meteorological Institute

(DNMI)

P.O. Box 43

Blindern

N-0313 Oslo, Norway

tel.: 47 22963000

fax: 47 22963250

e-mail [email protected]

Carlos Borrego

University of Aveiro

Department of Environment and Planning

P-3810 Aveiro, Portugal

tel.: 351 34 370617

fax: 351 34 428600

e-mail [email protected]

Jørgen Brandt

National Environmental Research Institute



P.O. Box 358


tel.: 45 4630 1157

fax: 45 4630 1214


Peter J.H. Builtjes

TNO-MEP, Department of Environmental

Quality

P.O. Box 342

7300 AH Apeldoorn, The Netherlands

tel.: 31 55 549 3038

fax: 31 55 549 3252


Giuseppe Calori

Politecnico di Milano

Departimento di Elettronica e Informazione

Via Ponzio 34/5

2133 Milano, Italy

tel.: 39 2239 93555

fax: 39 2239 93587


GLOREAM-report

26 of 216

Jesper Christensen

National Environmental Research Institute



P.O. Box 358

Dk-4000 Roskilde, Denmark

tel.: 45 46 30 1175

fax: 45 46 30 12 14


H.C. Davies

Atmosphärenphysik ETH

Hönggerberg HPP

CH-8093 Zürich, Switzerland

tel.: 41 1 63 33 506

fax: 41 1 63 31 058


Adolf Ebel

University of Cologne

Inst. for Geophysics and Meteorology EURAD

Aachener Strasse 201-209

D-50923 Cologne, Germany

tel.: 49 221 400 2258

fax: 49 221 400 2320


Georg A. Grell

Fraunhofer Institute for Atmospheric

Environmental Research (IFU)

Kreuzeckbahnstr. 19

D-82467 Garmisch-Partenkirchen, Germany

tel.: 49 8821 183 208

fax: 49 8821 183 243


Ingmar Ackermann

Ford Research Center Aachen

Süsterfeldstr. 200

D-52072 Aachen, Germany

tel.: 49 24 1942 1203

fax: 49 24 1942 1301


Hans Feichter

Max Planck Institute for Meteorology

Dept. Theoretical Climate Modelling

Bundesstrasse 55

D-20146 Hamburg, Germany

tel.: 49 40 41 17 33 17

fax: 49 40 41 17 32 98


Henny Kelder

KNMI (Royal Netherlands Meteorological

Institute)

P.O. Box 201

3730 AE De Bilt, The Netherlands

tel.: 31 30 2206 472

fax: 31 30 2210 407


Johannes Keller

Paul Scherrer Institute (PSJ) Air Pollution

CH-5232 Villigen, Switzerland

tel.: 41 56 3102 065

fax: 41 56 3104 525


Sissi Kiilsholm

Danish Meteorological Institute

Meteor. & Oceanographic Research Division

Lyngbyvej 100

DK-2100 Copenhagen, Denmark

tel.: 45 39 15 7452

fax: 45 39 15 7460


Oswald Knoth

Institute for Tropospheric Research

Permoserstrasse 15

D-04318 Leipzig, Germany

tel.: 49 341 235 2147

fax: 49 341 235 2139


GLOREAM-report

27 of 216

Vladimir K. Kouznetsov

Karbisheva str. 7

194021 St. Petersburg, Russia

tel.home: 812 311 85 84

tel./fax: 812 247 86 62


Bernd C. Krüger

Ecole Polytechnique Fédérale de Lausanne

EPFL - DGR - LPAS

CH-1015 Lausanne, Switzerland

tel.: 41 21 69 35 701

fax: 41 21 69 33 626


Bärbel Langmann

Max-Planck-Institut für Meteorologie

Bundesstrasse 55

D-20146 Hamburg, Germany

tel.: 49 40 4117 3239

fax: 49 40 4417 87


Joakim Langner

Swedish Meteorological & Hydrological

lnstitute

S-601 76 Norrköping, Sweden

tel.: 46 11 15 84 50

fax: 46 11 17 02 07


Katharina S. Law

Centre for Atmospheric Science

Department of Chemistry

University of Cambridge

Lensfield Road

Cambridge CB2 1EW, United Kingdom

tel.:

fax: 44 12 23 33 63 62


Lazar Lazic´

Inst. for Meteorology, Faculty of Physics

University of Belgrado

YU11001 Belgrado, Yugoslavia

tel.: 38 111 625 981/625 831

fax: 38 111 328 82 619


[email protected]

Paul Makar

Atmospheric Environment Service

4905 Dufferin Street

Downsview, Ontario M3H 5T4, Canada

tel.: 1 416 739 4692

fax: 1 416 739 5708


Clemens Mensink

VITO

Centre for Remote Sensing and Atmospheric

Processes (CETAP)

Boeretang 200

B-2400 Mol, Belgium

tel.: 32 14 33 68 41

fax: 32 14 32 27 95


Klaus Nester

Institut für Meteorologie und Klimaforschung

Forschungszentrum Karlsruhe

Postfach 3640

D-76021 Karlsruhe, Germany

tel.: 49 72 47 82 32 77

fax: 49 72 47 82 47 42


Arthur Petersen

IMAU

Utrecht University

P.O. Box 80005

3507 TA Utrecht, The Netherlands

tel.: 3130 25 33 275

fax: 3130 25 43 163


GLOREAM-report

28 of 216

Eberhard Reimer

Free University of Berlin

Institute of Meteorology

Carl-Heinrich Becker-Weg 6-10

D-12165 Berlin, Germany

tel.: 49 30 8387 1190

fax: 49 30 7931 785


Jürgen Rißmann

Deutscher Wetterdienst

Section FF25

Kaiserleistraße 42

D-63067 Offenbach, Germany

tel.: 49 69 80 62 27 46

fax: 49 69 82 36 14 93


Roberto San José

Environmental Software and Modelling Group

Technical University of Madrid

Campus de Montegrancedo

Boadilla del Monte

28660 Madrid, Spain

tel.: 34 1336 7465

fax: 34 1336 7465


Eberhard Schaller

Brandenburgische Technische Universität

Haus 215, Burger Chaussee 2

D-03044 Cottbus, Germany

tel.: 49 35 55 78 13 105

fax: 49 35 55 78 13 128


Petra Seibert

Institute of Meteorology and Physics (IMP)

University of Agricultural Sciences (BOKU)

Türkenschanzstr. 18

A-1180 Wien, Austria

tel.: 43 14 70 58 20 20

fax: 43 14 70 58 20 60


G. Smiatek

Fraunhofer-Institute for Atmospheric

Environmental Research (IFU)

Kreuzeckbahnstr. 19

82467 Garmisch-Partenkirchen, Germany

tel.: 49 8821 183 282

fax: 49 8821 183 243


Bill Stockwell

Desert Research Institute

Division of Atmospheric Sciences

2215 Raggio Parkway

Reno, Nevada 89512-1095, United States

tel.:

fax:


Dimiter Syrakov

National Institute of Meteorology and

Hydrology (NIMH)

66 Tzarigradsko shaussee

Sofia 1484, Bulgaria

tel.: 35 92 72 22 71

fax: 35 95 88 03 80 / 88 44 94


[email protected]

Stefan Tilmes

Deutscher Wetterdienst

Section PE 25

Kaiserleistr. 42


tel.: 49 69 80 62 27 46

fax: 49 69 82 36 14 93


GLOREAM-report

29 of 216

Gerard Toupance

LISA (Laboratoire Interuniversitaire des

Systèmes Atmosphériques)

University Paris 12 - CNRS

61 Avenue du Général de Gaulle

94010 Créteil, France

tel.: 33 14 51 71 603

fax: 33 14 51 71 564


Jan Verwer

Center for Mathematics and Computer Science

(CWI)

Kruislaan 413

1098 SJ Amsterdam, The Netherlands

tel.: 31 20 59 24 095

fax: 31 20 59 24 199


Jörg Zimmermann

Deutscher Wetterdienst,

FE 25

Frankfurter Strasse 135


tel.: 49 69 80 622 750

fax.: 49 69 82 361 493


Peter Zimmermann

Moguntia Global Modelling

Ludwigstr. 10

D-65479 Raunheim, Germany

tel.: 49 614 222 777

fax: 49 614 223 432


Zahari Zlatev

National Environmental Research Institute,

(NERI)

Department of Atmospheric Environment

P.O. Box 358


tel.: 45 45 30 1149

fax: 45 46 20 12 14


GLOREAM-report

30 of 216

XI Authors and titles of theses

Jørgen Brandt Modelling Transport, Dispersion andDeposition of Passive Tracers from AccidentalReleases.PhD thesis. National Environmental ResearchInstitute, Roskilde, Denmark, 1998

D. Brunner One-year climatology of nitrogen oxides andozone in the tropopause region. Results fromB-747 aircraft measurementsPh. D. thesis. ETH Zürich, Switzerland, 1998

GLOREAM-report

31 of 216

XII Publications in refereed literature

In the following list the double refereed papers which have appeared ininternational journals from the beginning of GLOREAM until the end of 1998 aregiven Other relevant literature as conference proceedings and internal reports canbe found in the reports from the principal investigators.

Ackermann, I.J., H. Hass, M. Memmesheimer, A. Ebel, F.S. Binkowski and U.Shankar, Modal aerosol dynamics model for Europe: Development and firstapplications, Atmos. Environ. 32, (1998), 2981-2999.

Alexandrov, V., A. Sameh, Y. Siddique and Z. Zlatev, Numerical integration ofchemical ODE problems arising in air pollution models, EnvironmentalModelling and Assessment, 2, (1997), 365-377.

Andreani-Aksoyoglu S. and J.Keller, Short-term impacts of air pollutants inSwitzerland: Model evaluation and preliminary scenario calculations forselectet Swiss energy systems, in: C.A.Brebbia, C.F.Ratto, H.Power (eds),Air Pollution VI, WIT Press/Computational Mechanics Publications,Southampton, (1998), 799-808.

Bastrup-Birk, A., J. Brandt, I. Uria and Z. Zlatev, Studying cumulative ozoneexposures in Europe during a seven-year period, J. Geophys. Res. 102,(1997), 23917-23935.

Bastrup-Birk, A., J. Brandt and Z. Zlatev, Using partitioned ODE solvers in largeair pollution models, Systems Analysis Modelling Simulation 32, (1998)3-17.

Bottenheim, J.W., A. Guenther, P.B. Shepson, R. Steinbrecher and W.R.Stockwell, Special section: biogenic hydrocarbons in the atmosphericboundary layer, preface, J. Geophys. Res. 103, (1998), 25463-25465.

Brandt, J., A. Bastrup-Birk, J.H. Christensen, T. Mikkelsen, S. Thykier-Nielsenand Z. Zlatev, Testing the importance of accurate meteorological input fieldsand parameterizations in atmospheric transport modelling, using DREAM -validation against ETEX-1. Atmos. Environ. 32, (1998), 4167-4186.

Elbern, H., H. Schmidt and A. Ebel, Variational Data Assimilation forTropospheric Chemistry Modeling for Tropospheric Chemistry Modeling, J.Geophys. Res. 102, (1997), 15967-15985.

Feichter, J., U. Lohmann and I. Schult, The atmospheric sulfur cycle in ECHAM-4and its impact on the shortwave radiation, Clim. Dyn. 13, (1997), 235-246.

Flemming, J. and E. Reimer, The impact of special features of numericallypredicted and analysed meteorological data on the results of ozone forecastby a PBL-CTM, in: Air Pollution Modelling and its Application XIII,Plenum Press, New York, (1998).

GLOREAM-report

32 of 216

Galperin, M. and D. Syrakov, A multi-layer parametrization of aerosol specificprocesses for use in air pollution modelling, Bulgarian Journal ofMeteorology and Hydrology 8, (1997) 62-73.

Jacob D. and R. Podzun, Sensitivity studies with the regional model REMO,Meteorol. Atmos. Phys. 63, (1997), 119-128.

Kuhn M., P.J.H. Builtjes, D. Poppe, D. Simpson, W.R. Stockwell, Y. Andersson-Sköld, A. Baart, M. Das, F. Fiedler, Ø. Hov, F. Kirchner, P.A. Makar, J.B.Milford, M.G.M. Roemer, R. Ruhnke, A. Strand, B. Vogel and H. Vogel,Intercomparison of the gas-phase chemistry in several chemistry andtransport models, Atmos. Environ. 32, (1998), 693-709.

Langmann, B. and H.-F. Graf, The Chemistry of the Polluted Atmosphere overEurope: Simulations and Sensitivity Studies with a Regional ChemistryTransport Model, Atmos. Environ. 31, (1997), 3239-3257.

Law, K.S., P.-H. Plantevin, D.E. Shallcross, H. Rogers, C. Grouhel, V. Thouret,A. Marenco and J.A. Pyle, Evaluation of modelled O3 using MOZAIC data,J. Geophys. Res., 103, (1998), 25721-25740.

Lazic, L., and I. Tosic, A real data simulation of the Adriatic bora and the impactof mountain height on bora trajectories, Meteorol. Atmos. Phys. 66, No. 3-4,(1998), 143-155.

Makar, P.A. and S.M. Polavarapu, Analytic Solutions for Gas-Phase ChemicalMechanism Compression, Atmos. Environ. 31, (1997), 1025-1039.

Memmesheimer, M., M. Roemer and A. Ebel, Budget calculations for ozone andits precursors: seasonal and episodic features based on model simulations, J.Atmos. Chem. 28, (1997), 283-317.

Mensink, C. and W. Debruyn, Transboundary flux measurements forphotochemical model validation in Flanders, Annales Geophysicae,Supplement IV to Vol. 16, (1998) C 1158.

Ostromsky, T., P.C. Hansen and Z. Zlatev, A coarse-grain parallel QR-factorization algorithm for sparse matrices. Parallel Computing 24, (1998)937-964.

Ostromsky, T. and Z. Zlatev, Application of Sparse Matrix Techniques in theChemical Part of Large Air Pollution Models, in: “Large-Scale ScientificComputations of Engineering and Environmental Problems”. (M. Griebel,O.P. Iliev, S.D. Margenov and P.S. Vassilevski, eds.) Vieweg, Wiesbaden(1998) 189-197.

Roelofs, G.J. and J. Lelieveld, Model study of the influence of a cross tropopauseO3 transports on tropospheric O3 levels, Tellus 49B, (1997), 38-55.

San José, R.; Prieto J.F., Castellanos N. and Arranz J.M., Sensitivity study of drydeposition fluxes in ANA air quality model over Madrid mesoscale area,Measurements and Modelling Environmental Pollution, Ed: R. San José andC. Brebbia, CM Publications, ISBN 1-85312-461-3, (1997), 119-130.

Stockwell, W.R., F. Kirchner, M. Kuhn, and S. Seefeld, A new mechanism forregional atmospheric chemistry modeling, J. Geophys. Res. 102, (1997),25847-25879.

GLOREAM-report

33 of 216

Stohl, A., M. Hittenberger and G. Wotawa, Validation of the Lagrangian particlemodel FLEXPART against large-scale tracer experiment data, Atmos.Environ. 32, (1998), 4225-4264.

Syrakov, D., An effective advection scheme for tracer transport - description,properties and tests, Bulgarian Journal of Meteorology and Hydrology 8,(1997) 14-22.

Syrakov, D., A PC-oriented multi-level Eulerian dispersion model - modeldescription Bulgarian Journal of Meteorology and Hydrology 8, (1997)41-49.

Syrakov, D., Aerosol version of the multi-layer Eulerian dispersion model EMAP -sensitivity analysis and real data test run, Bulgarian Journal of Meteorologyand Hydrology 8, (1997) 82-88.

Syrakov, D. and M. Prodanova, Sensitivity analysis of the dispersion model EMAPand real data run for estimating sulphur deposition over South-easternEurope, Bulgarian Journal of Meteorology and Hydrology 8, (1997) 50-61.

Syrakov, D. and M. Prodanova, Bulgarian emergency response models - validationagainst ETEX first release, Atmos. Environ. 32, (1998) 4367-4375.

Syrakov D. and D. Yordanov, On the Surface Layer Parametrization in an EulerianMulti-Level Dispersion Model, Bulgarian Journal of Meteorology andHydrology 8, (1997) 74-81.

Tilmes, S. and J. Zimmermann, Investigation on the spatial scales of the variabilityin measured nearground ozone mixing ratios, Geophys. Res. Let. 25, (1998)3827-3830.

Tosic, I. and L. Lazic, Improved bora wind simulation using a nested Eta Model,Meteorol. Atmos. Phys. 66, No. 1-2, (1998) 1-10.

Verver, G.H.L., Mixing of Reactive Gases in the Convective Boundary Layer,Phys. and Chem. of the Earth 23, (1998), 673-677.

Verver, G.H.L., H. van Dop and A.A.M. Holtslag, Turbulent mixing of reactivegases in the convective boundary layer, Bound. Layer Meteorol. 85, (1997),197-222.

Villenave, E., R. Lesclaux, S. Seefeld and W.R. Stockwell, Kinetics andatmospheric implications of peroxy radical cross reactions involvingCH3C(O)O2 radical, J. Geophys. Res. 103, (1998), 25273-25285.

Wauben, W.M.F., J.P.F. Fortuin, P.F.J. van Velthoven and H.M. Kelder,Comparison of modelled ozone distributions with sonde and satelliteobservations, J. Geophys. Res. 103, (1998), 3511-3530.

Wauben, W.M.F., P.F.J. van Velthoven and H.M. Kelder, A 3D chemistrytransport model study of changes in Atmospheric ozone due to aircraftemissions, Atmos. Environ. 31, (1997), 1819-1836.

Wernli, H., A Lagrangian-based analysis of extratropical cyclones. II: A detailedcase study, Quart. J. Roy. Meteor. Soc. 123, (1997), 1677-1706.

Wernli, H. and H. C. Davies, A Lagrangian-based analysis of extratropicalcyclones. I: The method and some applications, Quart. J. Roy. Meteor. Soc.123, (1997), 467-489.

GLOREAM-report

34 of 216

Wotawa, G., A. Stohl and B. Neininger, The urban plume of Vienna: Comparisonbetween aircraft measurements and photochemical model results, Atmos.Environ. 32, (1998), 2479-2489.

GLOREAM-report

35 of 216

XIII Reports from the principal investigators

Introduction

GLOREAM-projects are divided into the following 5 tasks:

1. Model investigation and improvement2. Global modelling3. Computational aspects4. Model evaluation and validation5. Model application and assessment studies.

The different projects constitute in the following way to the specific tasks:

Project PI Task: 1. 2. 3. 4. 5.

Andersson-Skold * *

Barone * *

Bastrup-Birk *

Berge * *

Borrego * *

Brandt *

Builtjes *

Builtjes *

Calori *

Christensen * *

Davies * *

Ebel * * *

Feichter *

Grell * *

Hass * *

Kelder *

Keller *

Kiilsholm

Knoth * *

Kouznetsov *

GLOREAM-report

36 of 216

Project PI Task: 1. 2. 3. 4. 5.

Krüger * * *

Langmann * * *

Langner * *

Law *

Lazic *

Makar *

Mensink * *

Nester *

Nester *

Petersen * *

Reimer *

Rissmann *

San Jose * *

Schaller *

Seibert *

Smiatek *

Stockwell *

Syrakov *

Tilmes * *

Toupance * *

Verwer *

Zimmermann, J. *

Zimmermann, P. *

Zlatev * * *

The following remarks have to be made:

− The project: “Model development with application to the South of Sweden” byYvonne Andersson-Skold had to be cancelled due to lack of funding. Noannual 1998 report could be made, the project is withdrawn from GLOREAM.

− The project “Danish Assessment Modelling” by Annemarie Bastrup-Birk hadto be cancelled because of change of position of the PI. No annual 1998 reportcould be made, the project is withdrawn from GLOREAM.

GLOREAM-report

37 of 216

− The project: “Application, testing and further development of a multiscalecoupled meteorological and atmospheric chemistry model” by Georg Grell hadto be cancelled due to change of position of the PI. No annual 1998 reportcould be made, the project is withdrawn from GLOREAM.

− The project “Numerical algorithms in Atmospheric Air Quality Modelling” byJan Verwer had no activities in 1998, so no annual report has been made. Theproject became active again in 1999 under a new PI, Patrick Berkvens.

− The project “Improved meso-scale modelling” by Bernd Krüger had to becancelled due to change of position of the PI. No annual 1998 report could bemade, the project is withdrawn from GLOREAM.

GLOREAM-report

38 of 216

Formulation, application and evaluation of the PNAM model

A contribution to subproject GLOREAM

Guido Barone, Pasqua D’Ambra, Daniela di Serafino,Giulio Giunta, Almerico Murli and Angelo Riccio

Dep. of Chemistry, University “Federico II” of NaplesVia Mezzocannone 4, 80134 – Napoli, Italy

1. Summary

This report provides an overview of the current status of the PNAM (ParallelNaples Airshed) model. The progress made during 1997 and 1998 mainlyconcerned the evaluation of PNAM to simulate a photosmog episode in the Naplesurban area (in Southern Italy) and its parallel performance using an IBM SPcomputer.

2. Aim of the research

The main goal of this project is the design of an operational software systemfor air quality simulations in the Campania Region. Campania is a denselypopulated area located in Southern Italy, with Naples as main urban area.The study of atmospheric pollution phenomena and the assessment of theirimpact on the health of inhabitants and overall ecosystem is particularlyrelevant in this coastal area where heavy emission loads and intense summerinsolation give raise to long exposure to high ozone levels.

The development of the software system is an ongoing activity carried out atthe Center for Research on Parallel Computing and Supercomputers (CPS-CNR), in Naples, by an interdisciplinary team of atmospheric chemists andcomputational mathematicians.

3. Activities during the year

During 1997 and 1998, work has been devoted to the development of thecomputational approach used in PNAM.

This model is based on an Eulerian approach, i.e. air pollution dynamics isdescribed by means of the mass balance equations. PNAM is an ‘off-line’model, that is atmospheric data (eddy diffusivity tensor, temperature and theparameters for the calculation of dry deposition velocities) are derived froman external source, and fed into PNAM at a regular time interval, generally

GLOREAM-report

39 of 216

every one hour. In our simulation, these values are derived from theapplication of the the Topographic Vorticity-mode Mesoscale Model(TVM). A 1.5 order closure (based on a prognostic equation for turbulentkinetic energy) is used to calculate the vertical turbulent coefficient. Thevertical wind component is internally calculated to assure incompressibility.The model is written in σ -coordinates to take into account surfaceirregularities.

The computational approach used to solve the mass balance equations isbased on a symmetric time-splitting technique that decouples advectionfrom vertical diffusion and chemistry in the following form:

ti

tT

tDC

tT

tti CLLLC 2/2/ ∆∆∆∆+ =

where LT is the advection operator and LDC is the vertical diffusion and chemistryoperator. The vertical turbulent diffusion and the chemical kinetics are coupledbecause they can have similar time scales on urban scale domains, especially forhighly turbulent wind fields.

A three-dimensional rectangular grid is considered, which is uniformly spaced inthe horizontal direction and has variable grid spacing in the vertical direction. Thesemi-discretization of the above operators is performed using cell-centered finite-difference schemes.

In PNAM, the coupled implicit solution of diffusion and chemistry isimplemented, using a modified version of the general-purpose VODE package.The solver, which has been modified to take advantage of the sparsity of theJacobian matrices arising in the application of the BDF methods used in VODE,has been successfully compared, in the solution of the vertical diffusion-reactionequations, with other special-purpose and general-purpose stiff ODE solvers ondifferent atmospheric chemical models (Barone et al., 1998a).

The parallelism has been introduced in PNAM using a ‘grid partitioning’technique, i.e. dividing the computational grid into subgrids and assigning asubgrid to each of the processors available for computation. The domain isdecomposed into vertical air columns, and is mapped onto a corresponding logicalgrid of processors in natural way. This choice is mainly driven by the problemfeatures. The coupled solution of vertical diffusion and chemistry, which is a verytime-consuming task, introduces a global coupling among points in the verticaldirection, while requires no interaction between points that lie in different verticalcolumns. Therefore, no data communication must be performed during thediffusion-chemistry steps. Moreover, this data decomposition allows to useavailable efficient and reliable sequential software, such as VODE, to solve the

GLOREAM-report

40 of 216

stiff ODE systems arising from the semi-discretization of the vertical diffusion andchemistry operators. On the other hand, a nearest-neighbour communication isrequired in the advection steps, to update the boundary points of each subgrid.Furthermore, a global communication is required each time the meteorologicaldata are read, to compute the maximum advection step allowed by the CFLcondition.

We applied load balancing techniques to address the problem of load imbalance.The main source of load imbalance is the stiffness of the chemical kinetics, whichvaries in time and space, and the different time steps used by the variable step stiffODE solver in different vertical columns. An additional source of load imbalancecan be the use of a non-homogeneous parallel machine, i.e. with processors ofdifferent computational power.

As a first experience with load balancing, the strategy used in the MM90 parallelregional weather model has been implemented in PNAM. This algorithm has aglobal decision base, i.e. the decision of remapping the computational grid ontothe processor grid is based on a measure of the workload of all the processors, anda local migration base, i.e. units of computational work are moved amongneighbouring processors.

PNAM has been written in Fortran 90, using dynamic memory allocation, pointers,modules, and other features enhancing the software flexibility.

The parallel implementation is based on the Runtime System Library (RSL),developed at Argonne National Laboratory, that is specifically tailored forefficient and straightforward implementations of finite-difference regular-gridapplications on distributed-memory parallel computers. RSL is based on the MPIstandard message-passing library, which ensures the portability of the package andof the application software based on it. For mode details about the parallelimplementation see (Barone et al., 1999).

4. Principal results

In this preliminary phase, PNAM has been used to simulate a photosmog episodeoccurred between July 25 and 27, 1995. During this period a high ozoneconcentration (about 150 ppb) was observed for the Naples basin.

From a meteorological point of view, the period under study was characterized bya high pressure weather system and stagnant conditions. The geostrophic winddirection, estimated from synoptic maps, was North-West and its intensity wasabout 4 m/s. The intense solar radiation determined an intensive photochemicalactivity with a high production of ozone and other photochemical oxidants. Forthese reasons, during July 25-27, 1995 the Naples urban area experienced a

GLOREAM-report

41 of 216

photosmog episode, and ozone concentration exceeded 150 ppb, i.e. was wellabove the attention level of 180 µg/m3 defined by the Italian legislation.

Meteorological data (wind velocity, diffusion tensor, Monin-Obukhov length,friction velocity, mixing height and temperature) were generated by theTopographic Vorticity-mode Mesoscale-β (TVM) model. The roughness lengthswere estimated from landuse data, both for the meteorological and the air qualitymodel. The landuse field was extracted from the Italian Army Geographic Institute(IGM) archive at an original resolution of 250 × 250 m2 and mapped onto thecorresponding CIT categories. Total solar radiation fields, as well as initialconditions, were estimated from the database of the monitoring network of theCampania Region (MARC). Presently, MARC automatically detects theconcentrations of various atmospheric pollutants, e.g. SOx, NOx, O3, PAN, COand VOCs, in the main urban areas of the Campania Region, in order to alert localauthorities to dangerous exceedances. Air quality data from 17 stations wereavailable. These data were interpolated to derive initial conditions for NOx, CO,NMHC and O3 concentrations at ground level. For MEK, ALD2 and HCHO thebackground value of 3 ppb was used. Unfortunately, initial conditions at upperlevels were not available; for this reason, the initial concentrations at upper levelswere calculated by linear interpolation from the ground level concentrations andthe natural background values at the simulated maximum mixing height. Tominimize the impact of initial conditions, a startup time of 24 hours was used, i.e.the simulation started on July 25, at 0:00 LST, and only the results for July 26were analyzed.

The results from the meteorological model have been compared with availablemeasured data, to assess the reliability of the computed wind field beforeproviding it in input to PNAM. TVM ws able to simulate the main patterns ofatmospheric circulation over the area under study (Barone et al., 1998b). We alsocompared results from PNAM with available measured air quality data. Amongmodeled species are CO, O3, NO and NO2. The computed results appearedpromising, since they showed the capability of PNAM of reproducing the temporaland spatial patterns of measured air quality data, although some discrepancieswere evident. For example, during the night, the O3 concentration was lower thanthat observed in Naples urban area. The phase and the amplitude of the NO2 andCO predicted concentration values generally were not correctly reproduced(Barone et al., 1998b). These inaccuracies deserved a more careful tuning of themodel.

Numerical experiments have been carried out using an IBM SP machine at CPS-CNR. This machine has 12 Power2 Super Chip thin nodes (160 MHz), each with512 MBytes of memory, which are connected via an SP switch, with a peak bi-directional bandwidth of 110 MBytes/sec. The available Fortan 90 compiler is XLFortran, version 4.1. The RSL parallel environment runs on the top of the IBMproprietary version 2.3 of MPI. First experiments have been performed without

GLOREAM-report

42 of 216

using any dynamic load balancing strategy. The maximum parallel efficiency wasabout 60% on twelve processors. Preliminary experiments have been carried outusing the dynamic load balancing strategy. There were no sensible performanceimprovements over the execution with a static data decomposition (the increase inefficiency is at most 4%). We think that some reasons for the above results shouldbe found in the choice of the workload measure, which does not take into accountidle times arising when different processors perform the advection steps onsubgrids of different sizes (for more details see Barone et al., 1999).

5. Main conclusions

The analysis of results from this preliminary experiment seems promising. Thecomparison of simulated vs. computed results from the air quality models showedthe capability of PNAM in reproducing the main temporal and spatial patterns ofair pollutant concentrations, though some dicrepancies were evident. The parallelperformance showed a parallel efficiency of about 60% on a real test case.

6. Aim for the coming years

In the coming years the main objective will be devoted to provide the CampaniaRegional Board with an operational version of PNAM. The first version of themodel, that is expected to be released at the end of 2000, will be intended for shortterm simulations (2 ÷ 3 days) over a limited area domain (250 × 250 km2) anddesigned for high performance parallel computers. The main goal of the systemwill be to provide, in a reasonable response time, a large number of simulationswith different scenarios, in order to study and to assess the influence of variationsof emission sources and land use to the related levels of concentrations ofpollutants.

More work will also be devoted to the study of load balance and nestingtechniques on parallel computers.

7. References

Barone G., P. D’Ambra, D. di Serafino, G. Giunta and A. Riccio; A Comparison ofNumerical Methods for Solving Diffusion-Reaction Equations in AirQuality Models. Technical report n. TR98-3 of the “Center for Research onParallel Computing and Supercomputers (CPS-CNR), Napoli, (1998a). Toappear on “Computing and Visualization in Science”.

Barone G., P. D’Ambra, D. di Serafino, G. Giunta and A. Riccio; Application of aparallel Air Quality model to the Campania region, Proc. “Air pollutionModelling and Simulation APMS98”, Champs-Sur-Marne, (1998b) 57-70.

Barone G., P. D’Ambra, D. di Serafino, G. Giunta, A. Riccio; PNAM: Parallelsoftware for air quality simulations in Naples area, J. Environ. Health andManag., (1999) in press.

GLOREAM-report

43 of 216

Status on the development of the EMEP regional scale Eulerianphotochemistry model


J.E. Jonson, E. Berge, D. Simpson and L. TarrasonThe Norwegian Meteorological Institute (DNMI)

P.O. Box 43, Blindern, N-0313 Oslo, Norway

1. Introduction

The EMEP Eulerian photochemistry model is virtually identical to the EMEP aciddeposition model, but the number of chemical components has been increased. Themodel has been developed in cooperation with the department of Geophysics,University of Oslo and NILU (Norwegian Institute for Air Research). At DNMI weare shifting from single layer Lagrangian models to Eulerian models. Themeteorological fields driving the chemical tracer model (CTM) will be generatedby the HIRLAM model from 1997 and onwards. Meteorological fields for theLagrangian model will not be generated by the HIRLAM model.

2. Progress in 1998

An extensive revision of the program code has been made. The code now runsmore than a factor of two faster than previous versions of the model. This processalso led to the detection of several errors in the coding that are now corrected. Thechemistry scheme in the EMEP Eulerian photochemistry model has been extendedto also include sulphur and ammonium components. Thus all components in theEMEP Eulerian acid deposition model are included in both model versions. Modelcalculations have been made for the six summer months April - September 1996,and model calculated concentrations of O3, NO2, total nitrate (HNO3 andammonium nitrate), SO2 and sulphate have been compared to measurements. Thishas been documented in Jonson et al. 1998a, 1998b. In Simpson and Jonson (1998)ozone, NO2, total nitrate, SO2 and sulphate calculated by the EMEP Eulerian andLagrangian models have been compared. Model calculations have also been madeas part of the EU project POLINAT, studying ozone and ozone precursors in theNorth Atlantic flight corridor in the upper troposphere (Jonson et al. 1998c). Aspart of the EU project PAUR model calculations of ozone, NO, NO2 and PANhave been compared to measurements at the Aegean Island of Agios Efstratios andTatoi, close to Athens. Furthermore, the effects of UV fluctuations inferred fromtotal ozone and tropospheric aerosol variations have been studied (Jonson et al.1999).

GLOREAM-report

44 of 216

3. References

Jonson, J.E., L. Tarrason and J.K. Sundet, The Eulerian 3-D oxidant model: Statusand evaluation for summer 1996 results and case studies. In: Transboundaryphoto-oxidant air pollution in Europe. EMEP/MSC-W Status Report 2/98,The Norwegian Meteorological Institute, Oslo, Norway, (1998a).

Jonson, J.E., L. Tarrason and J.K. Sundet, Calculation of ozone and otherpollutants for the summer 1996, Environ. Management and Health, (1998b),in press.

Jonson, J.E., I.S.A. Isaksen and J.K. Sundet, Calculated effects of aircraftemissions in the North Atlantic flight corridor, In: Pollution from AircraftEmissions in the North Atlantic Flight Corridor, edited by UlrichSchumann,. Air Pollution Research Report 68, European Commission,(1998).

Jonson, J.E., A. Kylling, T. Berntsen, I.S.A. Isaksen, C.S. Zerefos andK. Kourtidis, Chemical effects of UV fluctuations inferred from total ozoneand tropospheric aerosol variations. Submitted to J. Geophys. Res., (1999).

Simpson, D. and J.E. Jonson, Comparison of Lagrangian and Eulerian models forthe summer of 1996. In: Transboundary photo-oxidant air pollution inEurope. EMEP/MSC-W Status Report 2/98, The Norwegian MeteorologicalInstitute, Oslo, Norway, (1998).

GLOREAM-report

45 of 216

Impact of a doubling CO2 concentration scenario in the air qualityover Lisbon airshed


Carlos Borrego,Ana Miranda, Ana Carvalho, Carlos Fernandez and José Souto

Department of Environment and PlanningUniversity of Aveiro, 3810 Aveiro, Portugal

1. Summary

This work presents a first approach to estimate the effects of climate change in theGreat Lisbon Area (GLA) airshed air quality. To attain this goal three models wereused: (i) a GCM (the NCAR Community Climate Model - CCM3); (ii) a meso-meteorological model (MEMO); and (iii) a photochemical mesoscale model(MARS).


The GLA, because of its industrial and urban importance and high emissionslevels, is one of the Portuguese regions where the knowledge concerning impact ofclimate change in the atmospheric environment is fundamental. The contributionof the University of Aveiro to the GLOREAM project intends to assess climatechange impact on the air quality as well as mesoscale atmospheric circulationsover GLA.


During the last year contacts were established with the Supercomputational Centreof Galicia, in Spain, in order to get Global Climate Models results. Two GCMsimulations were made: a control one with an actual CO2 concentration value(355 ppmv) and a double CO2 concentration (710 ppmv) with the CCM3 model(Kiehl et al., 1998). The integration period of the GCM simulations was 36 years,with a 20 minutes time-step, for a T42 resolution. An average of the verticaltemperature and wind components variation considering 3 grid GCM cells wasmade for each simulation (between the 25 of July and the 13 of September) locatedat: (8.4W, 40.5 N); (8.4 W, 37.7 N) and (5.6 W, 37.7 N). They have been chosenin order to obtain a better characterisation of the GCM vertical results above themesoscale domain of interest. Hence, the cells climatic data were introduced asvertical meteorological information in the mesoscale model. Averaged values ofsurface water temperature estimated with the GCM were also introduced in the

GLOREAM-report

46 of 216

mesoscale model, the MEMO model (Moussiopoulos, 1990), as initial and toupdate the boundary conditions.

The mesoscale meteorological results calculated with the boundary conditionsdriven by the GCM were compared with a mesoscale simulation initialised withdata from measurements acquired during a typical summer day. The radiosondetaken for this run is from a typical summer day, which was selected for previouswork on mesoscale (Coutinho et al., 1994).

Emission rates introduced in the photochemical model, the MARS model(Moussiopoulos, 1992), were calculated by a methodology applied to theCORINAIR90 emission inventory for the GCM control scenario and for the typicalsummer (Borrego et al., 1998). For the emissions scenario concerning the doubleCO2 simulation some assumptions were considered based on the Kyoto Protocolon Climate Change. Portugal, within the global commitment made by the EuropeanUnion, assumed as a national objective a 40% limit (related to 1990 data) on theincrease of its greenhouse gases emissions until 2010. In practice, to calculate theemissions rate from point, line and area sources (other than biogenic) all theemission rates from CORINAIR90 inventory were enhanced by a “factor” of 40%.Concerning biogenic emissions no changes on the CORINE land cover data wereassumed. However, emissions rates were re-calculated for the new drivenmesoscale conditions. Background ozone concentration was not considered in anydone simulation.


From figure 1 it can be seen that for (8.4W; 40.5N) location temperature verticalprofiles obtained with both control and double CO2 simulations present similarvalues, under approximately 1500 m height. Above this height, slopes for the twotemperature curves present a ~ 5º C lag, i.e., at the same height, temperature isgreater for the double CO2 run. In the point (8.4W; 37.7N), vertical temperatureprofiles also have similar values, but in this case only until few meters abovesurface. This similarity on the lower troposphere may be due to the oceanproximity (in the (8.4W; 40.5N) location), or because the GCM considers (8.4W;37.7N) cell as water surface. In the last case, surface values of both temperatureprofiles estimated for the last location (5.6W; 37.7N), in the Southwest of Spain,are not similar as described above. The vertical temperature profiles measured inLisbon at 4th of August 1992 are characterised by an inversion in altitude at 00H00and 12H00. In the (8.4W; 40.5N) location, under 3000 m height (approximately),the measured temperature profile has higher temperatures than the two simulatedGCM scenarios. Since vertical temperature was measured in Lisbon and the cell(8.4W; 40.5N) covers the North part of Portugal the described differences could beexpected. On the contrary, measured temperature values above 4500 m are near thevertical temperature results of the GCM output control run, which can also beverified in the other two locations. Observing the graphics for the two other

GLOREAM-report

47 of 216

locations for the lower troposphere, the measured temperature is between thetemperatures simulated for both global climatic scenarios.

+HLJK

W#+P

,3

4333

5333

6333

7333

8333

9333

:333

7HPSHUDWXUH#+�&,

053 043 3 43 53 63

+HLJK

W#+P

,

3

4333

5333

6333

7333

8333

9333

:333

/RFDWLRQ#+;17:>73181,

+HLJK

W#+P

,

3

4333

5333

6333

7333

8333

9333

:333

7HPSHUDWXUH#+�&,

053 043 3 43 53 63 73

+HLJK

W#+P

,

0

1000

2000

3000

4000

5000

6000

7000

/RFDWLRQ#+819:>6:1:1,

+HLJK

W#+P

,

3

4333

5333

6333

7333

8333

9333

:333

7HPSHUDWXUH#+�#&,

053 043 3 43 53 63 73

+HLJK

W#+P

,

0

1000

2000

3000

4000

5000

6000

7000

/RFDWLRQ#+;17:>6:1:1,

Figure 1 Temperature vertical profiles for the 3 considered GCM results locations, at00H00 and 12H00.

Figure 2 shows ozone concentrations and wind fields, for the 4th of August 1992 at12H00 (figure 2 a) and 17H00 (figure 2 b). The reason for the choice of these twoinstants is due to the fact that, generally over Lisbon, at 12H00 the ozoneconcentration is increasing and at 17H00 is near the maximum value (Lopes,1997). At 12H00 the ozone plume is very thin. It starts over Lisbon city and istransported along the coastline. At this time the air mass responsible for thetransportation of this pollutant behaves as a recirculation zone with low windspeeds. At 17H00 the ozone plume is larger. The sea breeze is formed, the windsare stronger, and the combination of these factors is responsible for the plumeentering into land southwest. Ozone concentrations are found greater than80 µg m-3 at both analysed hours, which is higher than the guide value ofPortuguese legislation concerning daily average values, 65 µg m-3.

GLOREAM-report

48 of 216

20000 40000 60000 80000 100000 120000 140000 160000 180000

:HVW2(DVW#+P,

20000

40000

60000

80000

100000

120000

140000

160000

180000

6RXWK21R

UWK#+P,

0

10

20

30

40

50

60

70

80

g/m3

9 m/s

20000 40000 60000 80000 100000 120000 140000 160000 180000

20000

40000

60000

80000

100000

120000

140000

160000

180000

a) Ozone concentration field at 12H00 b) Ozone concentration field at 17H00

Figure 2 Ozone concentration and wind fields for the Lisbon airshed region,simulation of the 4th of August 1992.

20000 40000 60000 80000 100000 120000 140000 160000 180000

:HVW2(DVW#+P,

20000

40000

60000

80000

100000

120000

140000

160000

180000

6RX

WK21

RUWK

#+P,

20000 40000 60000 80000 100000 120000 140000 160000 180000

:HVW2(DVW#+P,

20000

40000

60000

80000

100000

120000

140000

160000

180000

6RXWK2

1RU

WK#+P

,

0

10

20

30

40

50

60

70

80

g/m3

9 m /s

a) control b) 2 x CO2

Figure 3 Ozone concentration and wind fields for the Lisbon airshed region, at 12H00;control and 2 x CO2 GCM simulations.

Figure 3 presents the simulation results at 12H00, both for control and double CO2

runs. Ozone concentration reaches the 50 µg m-3 in both presented situations andthe plume is transported, in the control run case, by the land breeze and the estuaryflow (oriented NE -SW, in the upper left side of the domain). At the double CO2

simulation, the plume is transported along the coastline, due to the breakdown ofthe land breeze. At 17H00 (see figure 4) ozone plume estimated with the controlscenario remains near the coast and extends in a thong shape into the AtlanticOcean, in the SW part of the domain. Ozone concentration values obtained withthe double CO2 simulations are always under 10 µg m-3. The reason for these twodifferent concentration fields relays on different vertical temperature structures.

GLOREAM-report

49 of 216

For the double CO2 scenario the vertical mixture is enhanced by the greatervertical unstability.

The “comparison” between the typical summer day simulations and the mesoscaleresults produced by the driven GCM runs, shows that the GCM model was unableto produce the particularity of the inversion in altitude at 12H00 at the Lisbonregion.

20000 40000 60000 80000 100000 120000 140000 160000 180000

:HVW2(DVW#+P,

20000

40000

60000

80000

100000

120000

140000

160000

180000

6RX

WK21

RUWK

#+P,

20000 40000 60000 80000 100000 120000 140000 160000 180000

:HVW2(DVW#+P,

20000

40000

60000

80000

100000

120000

140000

160000

180000

6RXW

K21RUWK

#+P,

10

20

30

40

50

60

70

80

g/m3

9 m /s

a) control b) 2 x CO2

Figure 4 Ozone concentration and wind fields for the Lisbon airshed region, at 17H00; control and2 x CO2 GCM simulations.

5. Main conclusions

The major and most important conclusion reached in this work is that more studiesare needed to understand the differences between the control and the typicalsimulations day. These studies should be focused on getting a better knowledge ofthe vertical atmosphere structure over the Lisbon region based on the existentradiossonde measurements. Also, regional climatic model results must be tried asmesoscale initial boundary conditions in order to get more consistencies on thephysical processes in the atmosphere. Concerning the photochemical part of themesoscale system, it is necessary to obtain information on future emissionsscenarios.

6. Possible policy relevance

The recognising from the scientific community about the evidence of GlobalClimate Change due to the greenhouse gases concentration rising, and the alertsgiving to governments throughout the last 20 years, has lead to several Protocolsbetween countries concerning reduction of greenhouse gases. The last signed

GLOREAM-report

50 of 216

Protocol was in Kyoto, at December 1997, in which Portugal, as a member of theEuropean Union, has agreed on sustainable increase their gases emissions upon40%, referent to the emission inventory of 1990. The results obtained on thisstudies can be useful both for the analysis of the regional impact of global climatechange and the identification of impacts as well as an important tool for thedevelopment of sustainable socio-economic strategies at national level (concerningroad planning, industry locations, urban and regional planning).

7. Aim for the coming year

In the coming year the forseen activities are focused on driven regional climatechange data and also on the refinement on the emissions rates prognosis forPortugal.

8. Acknowledgements

The authors gratefully acknowledge to the Fundação para a Ciência e Tecnologiafor Ana Carvalho scientific grant, under the project “The Impact of Global ClimateChange on the Atmosp. Env. of the North Atlantic and of the Iberian Peninsula”(PRAXIS/3/3.2/EMG/1949/95), and also to the Supercomputational Centre ofGalicia, Spain.

9. References

Kiehl, J.T., J.J. Hack, G. Bonan, B.A. Boville, D. Willianson, and P. Rasch; TheNational Center for Atmospheric Research Community Climate Model:CCM3. J. Climate 11, (1998) 1131-1149.

Moussiopoulos, N.; Introduction to mesoscale models & their use. Modelling theAtmospheric Flow Field, College on Atmospheric Boundary Layer Physics,Trieste, Italy, (1990).

Moussipoulos, N.; MARS - Model for the Atmospheric Dispersion os ReactiveSpecies: Technical Reference. Aristotle University, Thessaloniki, (1992).

Coutinho, M., A. Rocha and C. Borrego; Simulation of typical meso-meteorological circulations in the Lisbon region. The EUMAC ZoomingModel - Model Structure and Applications. Ed: Moussiopoulos,EUROTRAC, Garmisch-Partenkirchen, (1994), 105-116.

Borrego, C.; N. Barros, M. Lopes; M. Conceição; M.J. Valinhas, O. Tchepel; C.Ferreira, M. Coutinho and S. Lemos, Emission Inventory for Simulation andValidation of Mesoscale Models. Symp98, EUROTRAC-2, Transport andChemical Transformation in the Troposphere. Garmisch-Partenkirchen,March 1998.

GLOREAM-report

51 of 216

Modelling of transport, dispersion and deposition

A contribution to subproject GLOREAM and TOR-II

Jørgen BrandtNational Environmental Research Institute,Department of Atmospheric Environment,

Frederiksborgvej 399, P.O. Box 358,DK-4000 Roskilde, Denmark

Progress in 1998

The development of a new system for operational forecast modelling of transport,dispersion deposition and chemical transformation was started in 1998. The systemis called the DMU-ATMI THOR air pollution forecast system. It is an integratedoperational forecast system on regional and urban scale.

The goal is the development of a low-cost operational system that integrates urbanand regional models in a system for air pollution forecast, monitoring, scenarios,control and management, in support of decision makers and various environmentaland energy policy actions. The system will support the accomplishment of the EUdirectives relating to air pollution limit values for human health and give thefoundation for improving the quality of urban life. The system will provide thenecessary tool for the authorities to inform and/or warn the public and, in thefuture, to carry out the needed action (as e.g. restrictions on traffic) duringepisodes where the air pollution levels are exceeding the critical limit values.Furthermore, the system will be a part of the national monitoring programs atDMU-ATMI, both in urban and rural areas.

Currently, the system consists of a coupling of a numerical weather forecastmodel, ETA and a long range air pollution transport model, DEM, covering thewhole of Europe. The system produces operational 3 days air pollution forecasts,four times every day, on European scale, for the most important air pollutionspecies. The weather forecast model is initialized with global data from theNational Center for Environmental Prediction, NCEP, USA. The single model hasbeen optimized to run on a parallel powerful workstation with 4 processors.

The models create huge amounts of output data from a single run. These data areimpossible to comprehend without fast and advanced visualization and animationtechniques. Four times a day, nearly 900 visualizations are produced automatically

GLOREAM-report

52 of 216

and systemized so they can be seen with the use of an Internet browser. Ademonstration of the system can be seen at the web-address:

http://www.dmu.dk/AtmosphericEnvironment/thor

The DREAM (the Danish Rimpuff and Eulerian Accidental release Model) is acomprehensive, high-resolution three-dimensional tracer model, which has beendeveloped for studying short and large scale atmospheric transport, dispersion, anddeposition (wet and dry) of radioactive air pollution caused by a single but strongsource, as e.g., the Chernobyl accident. This model will be implemented in thesystem in 1999. Furthermore, will models describing the air pollution on urbanbackground and street scale be implemented.

Co-workers

Jesper H. Christensen, Lise M. Frohn, Ruwim Berkowicz, Finn Palmgren andZahari Zlatev, all at NERI-ATMI, Denmark.

Publications in 1998

Bastrup-Birk, A., J. Brandt and Z. Zlatev, Using partitioned ODE solvers in largeair pollution models, Systems Analysis Modelling Simulation (SAMS) 32,(1998), 3-17.

Brandt, J., A. Bastrup-Birk, J. H. Christensen, T. Mikkelsen, S. Thykier-Nielsenand Z. Zlatev, Testing the importance of accurate meteorological input fieldsand parameterizations in atmospheric transport modelling, using DREAM -validation against ETEX-1, Atmos. Environ. 32, No. 24, (1998), 4167-4186.

Ambelas Skjøth, C., A. Bastrup-Birk, J. Brandt and Z. Zlatev, Studying variationsof pollution levels in a given region of Europe during a long time-period.Systems Analysis Modelling Simulation, (1998), pp. 15. To appear.

Brandt, J., J. H. Christensen and Z. Zlatev, Numerical Modelling of Transport,Dispersion, and Deposition - Validation Against ETEX-1, ETEX-2, andChernobyl, Environmental Modelling and Software, (1998), pp. 17. Toappear.

Ambelas Skjøth, C., A. Bastrup-Birk, J. Brandt, and Z. Zlatev, Studying airpollution problems in France by using the Danish Eulerian Model,Environmental Modelling and Software, (1998), pp. 17. To appear.

Brandt, J., J. H. Christensen and Z. Zlatev, Real time predictions of transport,dispersion and deposition from a nuclear accident, EnvironmentalManagement and Health, February 1999, pp. 8. To appear.

Brandt, J., J. H. Christensen and Z. Zlatev, Operational air pollution forecastmodelling by using the THOR system, Physics and Chemistry of the Earth,(1999), pp. 6. To appear.

Brandt, J., J. Christensen, A. Ebel, H. Elbern, H. Jakobs, M. Memmesheimer, T.Mikkelsen, S. Thykier-Nielsen and Z. Zlatev, ETEX-1, 2nd phase:

GLOREAM-report

53 of 216

Calculations performed by NERI/Risø (Denmark) and the University ofCologne (Germany)". In: ATMES-II - evaluation of long-range dispersionmodels using data of the 1st ETEX release. EUR17756 EN, Joint ResearchCentre, European commission, Office for Official Publications of theEuropean Communities, Luxembourg, Eds.: Mosca, S., Bianconi, R.,Bellasio, R., Graziani, G. and Klug, W., (1998), pp. 6.

Brandt, J. and Z. Zlatev, Efficient Algorithms for the Chemical Part of Large AirPollution Models". In: Large-Scale Scientific Computations of Engineeringand Environmental Problems. Proceedings of the First Workshop on"Large-Scale Scientific Computations", Varna, Bulgaria, June 7-11, 1997.M. Gribel, O. P. Iliev, S. D. Margenov and P. S. Vassilevski (Eds.). Notes onNumerical Fluid Mechanics, Vol. 62. Printed by Friedr. Vieweg & SohnVerlagsgesellschaft, Germany, (1998), 145-154.

Brandt, J. and Z. Zlatev, Studying Long-Range Transport from Accidental NuclearReleases by Mathematical Models, In: Large-Scale Scientific Computationsof Engineering and Environmental Problems. Proceedings of the FirstWorkshop on "Large-Scale Scientific Computations", Varna, Bulgaria, June7-11, 1997. M. Gribel, O. P. Iliev, S. D. Margenov and P. S. Vassilevski(Eds.). Notes on Numerical Fluid Mechanics, Vol. 62. Printed by Friedr.Vieweg & Sohn Verlagsgesellschaft, Germany, (1998), 136-144.

Brandt, J. and T. Knudsen, Integrating GIS and external tools for spatio-temporalanalysis of time series of remote sensing data, In: P. Gudmandsen (Ed.),Future Trends in Remote Sensing. Proceedings from the 17th EARSeLsymposium, Lyngby, Denmark, June 17-19, 1997. Published by A. A.Balkema, Rotterdam, Netherlands, (1998), 117-123.

Knudsen, T. and J. Brandt, GIS and high-resolution atmospheric modelling forcorrections of sea surface observations from the TOPEX/POSEIDONsatellite altimeter, In: P. Gudmandsen (Ed.), Future Trends in RemoteSensing. Proceedings from the 17th EARSeL symposium, Lyngby, Denmark,June 17-19, 1997. Published by A. A. Balkema, Rotterdam, Netherlands,(1998), 373-378.

Bastrup-Birk, A., J. Brandt and Z. Zlatev, Modelling transport and dispersion fromaccidental releases". In: H. Hass and I. Ackermann (Eds.), Global andRegional Atmospheric Modelling, Proceedings of the first GLOREAMworkshop, Ford Research Centre, Aachen, Germany (September 10-12,1997), February 1998, 167-176.

Bastrup-Birk, A., J. Brandt and Z. Zlatev, Modelling the impact of ozoneconcentrations on human health and vegetation at selected sites in Europe,In: H. Hass and I. Ackermann (Eds.), Global and Regional AtmosphericModelling, Proceedings of the first GLOREAM workshop, Ford ResearchCentre, Aachen, Germany (September 10-12, 1997), February 1998, 157-166.

Bastrup-Birk, A., J. Brandt and Z. Zlatev, Long term calculations with the DanishEulerian Model, In: H. Hass and I. Ackermann (Eds.), Global and RegionalAtmospheric Modelling, Proceedings of the first GLOREAM workshop,

GLOREAM-report

54 of 216

Ford Research Centre, Aachen, Germany (September 10-12, 1997),February 1998, 147-156.

Bastrup-Birk, A., J. Brandt and Z. Zlatev, Modelling Atmospheric Transport,Dispersion, and Deposition on Short and Long Range - Validation againstETEX-1, ETEX-2, and Chernobyl - a contribution to subprojectGLOREAM, In: Proceedings from the EUROTRAC-2 Symposium ’98, VolII , March 23-27, Garmisch-Partenkirchen, Germany. Transport andChemical Transformation of Pollutants in the Troposphere. Eds. Patricia.Borrell and Peter Borrell. WIT Press, Computational MechanicsPublications, (1998), 570-574.

Bastrup-Birk, A., J. Brandt and Z. Zlatev, Studying Air Pollution Problems inEurope by Using the Danish Eulerian Model and TreGro - A contribution tosubproject GLOREAM, In: Proceedings from the EUROTRAC-2Symposium ’98, Vol II , March 23-27, Garmisch-Partenkirchen, Germany.Transport and Chemical Transformation of Pollutants in the Troposphere.Eds. Patricia. Borrell and Peter Borrell. WIT Press, ComputationalMechanics Publications, (1998), 565-569.

Skjøth, C. A., A. Bastrup-Birk, J. Brandt and Z. Zlatev, Studying ozone episodes inEurope with the Danish Eulerian Model, Proceedings of the 23stNATO/CCMS International Technical Meeting on Air Pollution Modellingand its Application, September 28 - October 2, Varna, Bulgaria, (1998), pp.8. To appear.

Brandt, J., Bastrup-Birk, A., J. H. Christensen and Z. Zlatev, Numerical Modellingof Transport, Dispersion, and Deposition - Validation Against ETEX-1,ETEX-2, and Chernobyl, In: Proceedings from the International Conferenceon Air Modelling and Simulation (APMS98), Vol 2, Champs-sur-Marne,Paris, France, October 26-29, (1998), 351-376.

Ambelas Skjøth, C., A. Bastrup-Birk, J. Brandt, and Z. Zlatev, Studying airpollution problems in France by using the Danish Eulerian Model, In:Proceedings from the International Conference on Air Modelling andSimulation (APMS98), Vol 1. Champs-sur-Marne, Paris, France, October26-29, (1998), 45-55.

Bastrup-Birk, A., J. Brandt and Z. Zlatev, Real time predictions of transport anddispersion from a nuclear accident, In: Large Scale Computations in AirPollution Modelling, Kluwer Academic Publishers, NATO Science Series, 2,Environmental Security Vol. 57, Eds.: Z. Zlatev, J. Brandt, P.J.H. Builtjes, IDimov, J. Dongarra, H. van Dop, K. Georgiev, H. Hass and R. San Jose,(1999), 53-62.

Ambelas Skjøth, C., Bastrup-Birk, A., J. Brandt and Z. Zlatev, Long-termcalculations with large air pollution models, In: Large Scale Computationsin Air Pollution Modelling, Kluwer Academic Publishers, NATO ScienceSeries, 2, Environmental Security Vol. 57, Eds.: Z. Zlatev, J. Brandt, P.J.H.Builtjes, I Dimov, J. Dongarra, H. van Dop, K. Georgiev, H. Hass and R.San Jose, (1999), 25-38.

GLOREAM-report

55 of 216

Brandt, J, Modelling Transport, Dispersion and Deposition of Passive Tracersfrom Accidental Releases, PhD thesis, National Environmental ResearchInstitute, Roskilde, Denmark, (1998), 307 pp.

GLOREAM-report

56 of 216

Regional differences in tropospheric ozone

A contribution to subproject GLOREAM and TOR-II

Peter J.H. BuiltjesTNO-MEP,

P.O. Box 342, 7300 AH Apeldoorn, the Netherlands

1. Summary

Ozone concentrations in the lower part of the troposphere show considerablespatial gradients over Europe. And even in case these concentrations are of asimilar magnitude, they might have been created by processes of differentstrengths. Using the 3-d eulerian grid model LOTOS- long term ozone simulation-budget calculations have been performed over the period April-September 1994.These budgets calculations show-among other things- that there is a largesimilarity in ozone behaviour between northern Italy and central Europe. They alsoshow the large differences between the ozone patterns over the Iberian peninsula,over northern Italy and over central Greece. A first attempt has been made tointegrate ground level measurements of ozone, radio-sonde data and troposphericozone data from the satellite instrument GOME with LOTOS-model calculationsfor the month of August 1997 using data-assimilation techniques. This approachwill result in more reliable determination of the differences in tropospheric ozone.


The aim of the research is to determine the processes, and their strength, whichdetermine the differences in ozone patterns over Europe. This will be done bymodel calculations and by making optimal use of existing observations fromground level, from radio-sondes and from satellites. The overall analysis shouldlead to a real quantitative understanding of the role that different processes play inthe ozone production over Europe.


An analysis has been made concerning regional differences in tropospheric ozoneby performing model calculations with the TNO-LOTOS model. With the LOTOS-model calculations have been made for ozone and related species over the wholeof Europe on an hour by hour basis over the whole year 1994. The horizontal gridsare 0.5 lat x 1.0 long with 4 vertical layers up to about 3 km. The model results arein reasonable agreement with the measurements, they do show similar magnitudes

GLOREAM-report

57 of 216

and similar O3-time and spatial gradients. (see also the description of theGLOREAM-project Model validation).

Using the LOTOS-model results budget calculations have been performed overselected areas and up to 2 km in the vertical by determining the ozone budgetconsisting of the net horizontal advection, the net vertical advection, the netchemical production and the dry deposition. Both marine areas and continentalareas have been considered.

Data-assimilation techniques have been developed which can handle non-linearchemistry. The method is an extended Kalman filter approach.

Data have been gathered concerning ground level ozone measurements, radio-sonde data and tropospheric ozone data derived from the GOME instrument.LOTOS-model calculations taking the available measurements and applying data-assimilation have been made for August 1997.


The budget calculations over the period April-September 1994 show that overcontinental areas there is an influx of ozone from the free troposphere to theboundary layer. This holds both for the North and the Mediterranean Sea.

The model results also indicate the a-typical behaviour over the Iberian peninsula,with contrary to other continental areas over central and southern Europe show asmall influx of ozone from the free troposphere. The other areas show in generalan outflow of ozone to the free troposphere.The results clearly show the largeinhomogenity in ozone formation processes over southern Europe.

Although the application of the extended Kalman filter data-assimilation techniquerequires the use of super-computers, the method has shown very promising results.Using this method for August 1997 indicates an underestimation of anthropogenicVOC-emissions over north-western Europe of about 20%.

5. Main conclusions

Model budget calculations are a useful method to analyse the regional differencesin tropospheric ozone.

Data-assimilation techniques for non-linear chemistry are a powerful tool tocombine knowledge and information contained in measurements and in modelformulations.

GLOREAM-report

58 of 216


The LOTOS model will be extended to contain an aerosol module. Modelcalculations including data-assimilation will be used to study the regionaldifferences in tropospheric ozone and aerosols over Europe.

7. Acknowledgement

The project has partly been funded by the EU-DG XII in the fourth frameworkprogramme under the RIFTOZ-project, and by the Dutch Remote Sensing Board.

8. References

P.J.H. Builtjes e.a., An analysis of regional differences in tropospheric ozone overEurope, 23rd ITM on air pollution modelling and its application, ClermontFerrand, France, June 1997.

P.J.H. Builtjes e.a., Regional differences in tropospheric ozone, Final report of theEU-RIFTOZ-project, December 1998.

M. van Loon e.a.., Data assimilation applied to LOTOS: first experiences, APMSConference, Paris, France, October 1998.

GLOREAM-report

59 of 216

An approach to model validation


Peter J.H. BuiltjesTNO-MEP,

P.O. Box 342, 7300 AH Apeldoorn, the Netherlands

1. Summary

It is essential that the adequacy and reliability of models which are used to studyspecific scientific questions or to adress policy issues is determined in an objectiveand general accepted way. Attempts have been made-by, apart from specificresearch, holding a workshop at the occasion of the Eurotrac-98 symposium, tocome to a kind of general methodology.

A useful distinction has been made between diagnostic model testing andintegrated model testing. Diagnostic model testing is focussed on testing thedifferent processes which are described in atmospheric chemistry-transportmodels. This testing is science oriented and often makes use of specific fieldcampaigns. Integrated model testing is focussed on testing the overall results ofatmospheric chemistry-transport models. The testing is policy oriented and oftenmakes use of the data of monitoring networks.

Integrated model testing has been used to establish the possibilities of atmosphericchemistry-transport models to determine AOT 40 values. The results show thatboth on the measurement side and the model side further developments andresearch should take place before reliable AOT 40 fields over Europe can beestablished.


The aim of the research is to contribute to the establishment of a general acceptedmethodology for model testing, and to perform model testing studies with aspecific focus.


A workshop has been organised at the occasion of the Eurotrac-98 symposiumwhich has been attended by about 30 participants. The results of the workshophave been incorporated in a presentation at the 23th ITM at Varna, Bulgaria.

GLOREAM-report

60 of 216

LOTOS-model results of AOT 40 have been analysed and compared to othermodel results, and have been compared with observations to the extent possible.


It has been shown that in testing atmospheric chemistry-transport models it isuseful to make a distinction between diagnostic and integrated model testing. It hasbeen shown that before AOT 40 measurements can be compared in a useful waywith AOT 40 model results the situation close to the surface-the height above thesurface and the dry deposition process- have to be known and established.

5. Main conclusions

Integrated model testing directed to AOT 40 requires further knowledge of therepresentativeness of the observations and the proper description of the processesclose to the surface as dry deposition.


Because integrated model testing requires the use of proper statistics, research todetermine the suited statistical parameters will be performed. The research inintegrated atmospheric chemistry-transport model testing of ozone and AOT 40will be continued.

7. Acknowledgement

The many colleagues who contributed to the many discussions on model validationare thanked for their useful contributions.

8. References

P.J.H. Builtjes, Policy development requires verified models: an impossible task?Eurotrac symposium’98, Garmisch Partenkirchen, Germany, March 1998.

P.J.H. Builtjes and A. Flossmann, Model validation, science and application23th ITM on air pollution modelling and its application, Varna, Bulgaria,Sept/Oct 1998.

P.J.H. Builtjes, Model validation of AOT 40, 2nd GLOREAM-workshop, Madrid,Sept. 1998.

GLOREAM-report

61 of 216

Episodic and long-term modelling over the Italian region


Giuseppe Calori*,Giuseppe Brusasca#, Giovanni Manzi# and Camillo Silibello#

* Dipartimento di Elettronica e Informazione, Politecnico di Milano# Enel SpA, Struttura Ricerca - Area Ambiente, Segrate (Milano)

1. Summary

A comprehensive air pollution modelling system, to be used at scales ranging fromthe mesoscale to the regional one, has been developed. The system is based on 3Dmeteorological and STEM-II Eulerian air quality model, adapted to use twodifferent chemical mechanisms: a simple one for long term simulations and a moredetailed one, including gas and aqueous phase reactions, to be used for episodes.

The more complete chemical mechanism has been tested on an intense ozoneepisode occurred during summer 1996 over the Po Valley, while the simplermechanism is actually being tested for a yearly period over the Italian region.


The final goal of the project is to set up a comprehensive modelling system to beused to analyse emission abatement strategies and critical loads/levels complianceover the Italian region.


The activities performed by the research operational unit are the following:

- development of the new version of STEM-II model;- development of a turbulence and deposition pre-processor;- application of the comprehensive model on a 240 x 232 km2 domain covering

all Lombardia Region, to simulate an intense ozone episode revealed duringJune 1996 (Calori et al., 1998);

- preliminary test of the long-term version of STEM-II for acidifying speciesover the Italian paeninsula and its comparison with ARES Lagrangian back-trajectory acid model (Calori et al., 1997).

GLOREAM-report

62 of 216

The first three activities have been performed in close conjunction with aSATURN subproject (“A Comprehensive Modelling System for PhotochemicalPollution Control in metropolitan areas” - P.I. Giovanna Finzi).


The comprehensive Eulerian 3D model is based on STEM-II (Carmichael et al.,1986; 1990). The original code has been partially rewritten in order to be easilyapplied to different areas, to use input data provided by various models/pre-processors, and to switch between alternative algorithms for simulated processes.Particular attention has been also paid to the internal documentation of the code, tofacilitate its maintenance.

Major features implemented in the model are the use of different coordinatesystems (e.g. UTM, latitude-longitude and polar stereographic), different advectionalgorithms, the complete interface with RAMS model (Pielke et al., 1992) anddirect treatment of large point sources.

Two chemical mechanisms are now available:

- a more complete mechanism, including a gas phase module also consideringexplicitly the isoprene chemistry (Lurman et al., 1986) and an aqueous phasemodule (mainly based on Chameides, 1984; Jacob, 1986) as well as someheterogeneous reactions;

- the simpler EMEP-II acid mechanism (Hov et al., 1988).

Input data needed by the model and provided by the related pre-processors are thefollowing:

- hourly values of diffuse emissions and major point sources, disaggregatedfrom CORINAIR emission data by a specific pre-processor;

- hourly meteorological fields provided by diagnostic (MINERVE - Geai, 1985 -and CALMET - Scire et al., 1990) and prognostic (RAMS) codes; cloud, rainand snow water contents, used in the aqueous chemistry module, can bedirectly assigned by the prognostic model or can be internally computed by theAdvanced Scavenging Module (ASM - Easter and Hales, 1984);

- hourly deposition velocities for each modelled species, calculated by aninferential dry deposition module taking into account land-use information(based on EUROTRAC-1 BIATEX subproject results as well as onmeasurements in the Po Valley; Cavicchioli et al., 1996);

- hourly horizontal and vertical diffusivities fields, computed by a diagnosticturbulence pre-processor (TURBANTE) or provided by RAMS model.

Output data obtained by the model are the following:

GLOREAM-report

63 of 216

- hourly gas and aqueous phase concentrations;- domain balances for user selected species;- wet and dry deposition fluxes.

The new version of STEM-II model has been also included into European EUModel Documentation System (http://aix.meng.auth.gr/lhtee/database.html).

5. Main conclusions

Since pollutant concentrations and deposition fluxes observed in a country are notsimply correlated to its emissions but also with emissions of surroundingcountries, a strong need arises for coordination among the research made bydifferent national and international agencies working on the problem. For examplein view of the use of comprehensive models over limited areas, the exchange ofdata (as emissions and concentrations computed by models over larger areas)should be made easier, in order to correctly feed the simulations with emissionsfrom neighbouring countries and pollutant fluxes through boundaries.


Activities planned for 1999 can be summarised as follows:

- complete test of the long-term version of STEM-II (based on EMEP-II acidchemical mechanism) over the Italian region for whole 1994, and itscomparison with the existing Lagrangian back-trajectory acid model (ARES);

- simulation of a few selected episodes (i.e. intense ozone, fog, precipitation)using the more detailed chemical mechanism;

- preliminary inclusion of an aerosol module and its sensitivity analysis;- analysis of future sustainable emission abatement scenarios over the Italian

area.

7. Acknowledgements

This research was supported by MURST and ENEL-SRI. The collaboration ofProf. G. Carmichael during the development and application of the modellingsystem is also greatly acknowledged.

8. References

Calori, G., G. Brusasca, G. Manzi and G. Finzi, Sensitivity analysis of aciddeposition modelling on the Italian area. In Proc. of 1st GLOREAMWorkshop, 10-12 Sept. 1997, Aachen (D), (1997).

Calori, G., C. Silibello, M. Volta and G. Brusasca, Application of a photochemicalmodelling system to an intense ozone episode over Northern Italy, Proc.

GLOREAM-report

64 of 216

International Conference on Air Pollution Modelling and Simulation APMS‘98, Champes sur Marne, (1998).

Carmichael, G.R., L.K. Peters and T. Kitada, A second generation model forregional scale transport / chemistry / deposition. Atmos. Environ. 20, (1986),173-188.

Carmichael, G.R., L.K. Peters and R.D. Saylor, The STEM-II regional scale aciddeposition and photochemical oxidant model - I. An overview of modeldevelopment and applications. Atmos. Environ. 25A, (1990), 2077-2090.

Cavicchioli, C., G. Manzi, G. Brusasca and G. Catenacci, Total acid depositionestimation in the Po Valley - Italy. Proc. of EUROTRAC Symposium ’96,Garmisch-Partenkirchen (D), 25-29/3/96, Computational Mechanics Publ.,(1996).

Chameides, W.L., The photochemistry of a remote marine stratiform cloud. J.Geophys. Res. 89, (1984), 4739-4755.

Easter, R.C. and J.M. Hales, PLUVIUS: A generalized one-dimensional model ofreactive pollutant behavior, including dry deposition, precipitationformation and wet removal, 2nd ed., Pacific Northwest Laboratory, PNL-4046 ED2, (1984).

Geai, P., Methode d’interpolation et de reconstitution tridimensionelle d’un champde vent: le code d’analyse objective MINERVE. Rep. DER/HE/34-87.03.,Electricité de France, Chatou (France), (1987).

Hov, Ø., A. Eliassen and D. Simpson, Calculation of the distribution of NOx

compounds in Europe. In: Tropospheric ozone (ed. by I.S.A. Isaksen),D. Reidel Publ. Co., (1988), 239-261.

Jacob, D.J., Chemistry of OH in remote clouds and its role in the production offormic acid and peroxymonosulfate. J. Geophys. Res. 91, (1986), 9807-9826.

Lurmann, F.W., A.C. Lloyd and R. Atkinson, A chemical mechanism for use inlong-range transport/acid deposition computer modeling. J. Geophys. Res.91, (1986), 10905-10936.

Pielke, R.A., W.R. Cotton, R.L. Walko, C.J. Tremback, W.A. Lyons, L.D. Grasso,M.E. Nicholls, M.D. Moran, D.A. Wesley, T.J. Lee and J.H. Copeland,A comprehensive meteorological modeling system - RAMS. Meteor. Atmos.Phys. 49, (1992), 69-91.

Scire, J.S., E.M. Insley and R.J. Yamartino, Model formulation and User’s Guide forthe CALMET meteorological model. Report A025-1, California Air ResourcesBoard, Sacramento (CA), (1990).

GLOREAM-report

65 of 216

Model studies of the air pollution in the Arctic by using the DanishEulerian Hemispheric Model


Jesper ChristensenNational Environmental Research Institute,Department of Atmospheric Environment,

Frederiksborgvej 399, P.O. Box 358,DK-4000 Roskilde, Denmark.

Progress in 1998

In 1998 the hemispheric model has been improved considerably by the coupling ofa state-of-the-art weather prediction model to the model system. This has improvedtreatment of the physical parametrizations. It has also made it possible to do moredetailed model calculations around Greenland with a higher spatial resolution, e.g.taking into account the influence of the orography of Greenland on both the windand precipitation pattern.

In addition a large effort has been invested in participating in a workshop inHalifax in October 1998. In the workshop 10 different large scale modellinggroups from both North America and Europe have prepared model results. Theworkshop was part of the international programme A Comparison Of ThePerformance Of Large Scale Models In Simulating Atmospheric SulphateAerosols(Cosam) operated under WCRP/WGNE and IGAC/GIM. The objective ofthis program is to compare large scale (>5000 km) models of atmospheric sulphateaerosols. The results from the Danish Eulerian Hemispheric Model show a goodperformance compared with other model groups results and with observations.

A new hemispheric model with a photochemical scheme consisting of 55 species,more than 94 chemical reactions and 17 photolyse reactions are in progress ofdevelopment. In 1998 the numerical part of the model was developed, consisting ofAccurate Space Derivative method with non-periodic boundary conditions for thehorizontal advection part, finite elements for the vertical transport and horizontaldiffusion, and finally the two-step method with local timestep and error check forthe solution of the chemical equations.

Publications in 1998

Heidam, N. Z., P. Wåhlin and J. H. Christensen, Tropospheric Gases and Aerosolsin Northeast Greenland, J. Atmos. Sci. 56, (1999), 261-278.

GLOREAM-report

66 of 216

Christensen, J., The Danish Eulerian Hemispheric Model, In: Proceedings from “AComparison of the Performance of Large Scale Models in SimulatingAtmospheric Sulphate Aerosols (COSAM), Halifax, Canada, October 19-21,(1998).

Kämäri, J., P. Joki—Heiskala, J. Christensen, E. Degerman, J. Derome, R. Hoffand A.-M. Kähkönen, Acidifying Pollutants, Arctic Haze, and Acidificationsin the Arctic, Chapter 9 in: AMAP Assessment Report: Arctic PollutionIssues. Arctic Monitoring and Assessment Programme (AMAP), S. Wilson,J. Murray and H. Huntington, Ed., (1998).

GLOREAM-report

67 of 216

A 1-Year Climatology of Stratosphere-Troposhere Exchange onthe Northern Hemisphere - Quantification and Associated

Mesoscale Processes


Heini Wernli, Michel Bourqui and H.C. DaviesInstitute for Atmospheric Science, ETH Zürich, CH-8093 Zürich, Switzerland

1. Introduction

Quantification of stratosphere-troposphere transport and the investigation of theassociated dynamical and physical processes are fundamental to the study andmodeling of tropospheric and stratospheric chemistry. Case studies of particularevents of stratosphere-troposphere exchange in mid-latitudes have indicated thatthey are associated with distinct mesoscale structures and processes in the near-tropopause/lower-stratosphere region, like the formation of tropopause folds, theirreversible break-up of narrow streamers of stratospheric air that protruded onsloping isentropes into the troposphere (or vice versa; cf. Appenzeller et al.,1996a), the diabatic decay of cut-off systems (through condensational heatingand/or radiation; e.g. Wirth 1995), and mixing induced by gravity-wave breaking.The mesoscale nature of these features indicates the delicate issue of theirappropriate representation for instance in global general circulation and chemistrymodels, and prompts the analysis of trans-tropopause exchange processes withcomparatively high-resolution NWP data sets.

This study is based upon a 1-year sample of operational analysis data from theECMWF for the northern hemisphere, with a temporal and spatial resolution of6 hours and 1°, respectively. The aims are to identify the exchange eventsrepresented by the data set during the time period May 1995 till April 1996 (fromthe stratosphere to the troposphere (STE) and vice versa (TSE)), to estimate theassociated mass flux across the tropopause, and to identify and analyze themesoscale structures in the vicinity of the individual exchange events.

2. Methodology

In order to identify the exchange events a Lagrangian approach has been chosen incombination with a PV-based definition of the tropopause (cf. Wernli and Davies1997). Trajectories are started every 24 hours on every model grid point between80 and 600 hPa during the 1-year period. STE is only attributed to air parcelswhose 9-day trajectories reside at least during a “threshold residence time” of twodays in the stratosphere (troposphere) before (after) crossing the 2 pvu-isosurface

GLOREAM-report

68 of 216

(and vice versa for TSE). This serves to eliminate air parcels that move transientlyto and for across the interface on short time scales. The Lagrangian approach alsopermits the investigation of the origin and destination of exchange air parcels onthe time scale of several days, which is of relevance for the chemical impact of theexchange event. The method has been applied previously to a selected case study(Wernli and Davies, 1997), where mass exchange estimates comparable to earlierstudies (e.g. Lamarque and Hess, 1994) have been found. A significant part of theexchange air parcels experienced a rapid downward motion to the lowertroposphere (~900 hPa).

Independently a contour search algorithm is used to objectively identify PVstreamers and cut-offs on 26 isentropic surfaces (from 300 to 350 K) every6 hours. Monthly mean fields of these mesoscale features are qualitativelycompared with the distribution of exchange events, and for every exchange event itis examined whether it occurs in the close vicinity of a streamer or cut-off.

Figure 1 Stratosphere-to-troposphere mass flux estimates for the four seasons duringMay 1995 - April 1996 based upon exchange trajectories with a thresholdresidence time of 48 hours. Values are in 1013 kg per season per 3° × 3° gridsquare.

GLOREAM-report

69 of 216

Figure 2 As Fig. 1 but for troposphere-to-stratosphere exchange.

3. Results

From a total of ~1.6.108 calculated trajectories for the 1-year period, 620’000(670’000) fulfilled the criterion for STE (TSE) particles (with a thresholdresidence time of 48 hours). Figures 1 and 2 show the geographical distribution ofthe STE and TSE mass flux for the four seasons (spring corresponds to March,April and May, etc.). STE occurs mainly in the mid-latitudes, whereas TSE has astrong maximum in the subtropics near 20°N and a weaker one near 50°N (see alsoFig. 3). Preferred geographical locations for STE during the investigated timeperiod are the region north of the Himalaya mountains, southern Europe, and theAtlantic and Pacific storm-tracks (see Fig. 1). STE activity is large in spring andwinter, whereas TSE occurs frequently in autumn and winter. This leads to adistinct seasonal cycle with a net mass flux out of (into) the stratosphere in spring(autumn), in agreement with the study of Appenzeller et al. (1996b).

Next we consider the subset of “deep exchange events” where air parcels descendto (rise from) the lower troposphere (i.e. below 700 hPa) in the case of STE (TSE).For STE they amount to 15% of the total number of events in winter, but are

GLOREAM-report

70 of 216

almost absent during summer (Fig. 4). This indicates, for example, that thefrequency of vertical transport of stratospheric ozone down to the lowertroposphere has a pronounced annual cycle with a clear maximum in the coldseason. For TSE deep exchange events are identified only in the mid-latitudes. Theseasonal variability is smaller than for deep STE but still there is a maximum(minimum) in winter (summer).

Figure 3 Zonal mean of the number of exchange events per season and per 3° × 3°grid square (weighted with cos φ); solid (dashed) lines are for STE (TSE).

Figure 4 As Fig. 3 but for the number of deep exchange events.

Figure 5 shows all trajectory points below 700 hPa which are associated with deepexchange events during the winter season. (The patterns are qualitatively similarfor the other seasons.) In the case of deep STE the low tropospheric destinations oforiginally stratospheric air are distributed over the Pacific and Atlantic storm-trackregions and over North Africa, and are almost absent over the Asian continent. Fordeep TSE the low tropospheric source regions are well confined to the entranceregions of the two dominant northern hemispheric storm-tracks, indicating thatdeep TSE events are associated with coherent air motions that accompany thedevelopment of extratropical cyclones (Wernli and Davies, 1997; Wernli, 1997).

A detailed analysis of the environment of the exchange events is performed on thetwo isentropic surfaces (separated by 2 K) which are closest to the θ-level of theexchange, and at the two enclosing data times. For example for an exchange eventoccurring between 12 and 18 UTC of a certain day at 322.3 K, it is checked at bothtime instances whether a mesoscale feature (tropospheric and stratospheric cut-offsand streamers) can be found within a distance of 2° from the exchange location onthe 322 and 324 K surfaces.

GLOREAM-report

71 of 216

Figure 5 Total number of trajectory points below 700 hPa associated with deepexchange events for the winter season (the left (right) panel is for STE(TSE)). Values are per season and per 3° × 3° grid square.

Figure 6 Ozone flux estimates associated with STE and TSE during spring. Values arein 106 kg per season per 3° × 3° grid square.

Table 1 presents the probabilities from a preliminary investigation based upon alimited number of exchange events (~20’000).

Table 1 The probabilities to find a mesoscale feature in the vicinity of an STE/TSEexchange event (see text for details).

STE TSE

strat. cut-offs 51% 17%strat. streamers 36% 27%trop. cut-offs 23% 52%trop. streamers 9% 23%

It reveals a strong preference for trans-tropopause exchange to occur in the vicinityof and associated with the dynamical evolution of mesoscale PV-streamers andcut-offs of stratospheric and tropospheric air. STE (TSE) is encountered frequentlynear stratospheric (tropospheric) cut-offs. (Note, that the probabilities are not

GLOREAM-report

72 of 216

additive in the sense that a single event is sometimes associated with differentfeatures at the tow time instances and isentropic surfaces).

Finally consideration is given to estimating the annual distribution of upward anddownward ozone flux across the tropopause in the 35-65°N latitude band coveringthe Atlantic and European sectors for the same time period. To this end theLagrangian approach is combined with aircraft ozone measurements from theprojects NOXAR (Brunner, 1998) and MOSAIC (Marenco et al., 1998). Estimatesfor the ozone flux are shown in Fig. 6 for the spring season. The upward ozoneflux (i.e. associated with TSE) has a similar amplitude in all four seasons, but thedownward flux is ~50% larger in spring than in autumn. This leads to apronounced maximum of the net ozone flux out of the stratosphere in the springseason of about 60 Tg (3months)1 within the considered region.

4. Conclusions

The preliminary evaluation of the 1-year Lagrangian data base for trans-tropopauseexchange leads to the following conclusions:

− The Lagrangian approach serves to identify STE and TSE events and theirspatial and seasonal variability, and offers insight into the origin anddestination of the exchange particles.

− A pronounced winter maximum exists for the frequency of deep stratosphericintrusions (STE events that descend lower than 700 hPa), and for TSE eventscharacterized by rapid ascent of low-tropospheric air parcels. The lattercategory occurs predominantly in the entrance regions of the main mid-latitudestorm-tracks.

− A significant portion of the identified exchange events are in the close vicinityof a PV-streamer or cut-off, underlining the importance of the evolution ofthese mesoscale features for the exchange of mass and chemical constituentsacross the mid-latitude tropopause.

− A combination of in-situ aircraft measurements and Lagrangian calculationsyields estimates for the net ozone flux out of the stratosphere in the NorthAtlantic-European region of 12-60 Tg (3months)1 with the maximum(minimum) occurring in spring (autumn).

Note that there are limitations of the present approach due to the strong month-to-month variability of STE/TSE (a one year sample is too small to reveal the typicalclimatological patterns) and to the probably inappropriate representation forinstance of deep convection and its associated rapid vertical mixing in theoperational analysis data set, especially in the tropical regions and duringextratropical summer.

GLOREAM-report

73 of 216

5. References

Appenzeller, C., H.C. Davies and W. A. Norton, Fragmentation of stratosphericintrusions, J. Geophys. Res. 101, (1996a), 1435-1456.

Appenzeller, C., J. R. Holton and K. H. Rosenlof, Seasonal variation of masstransport across the tropopause, J. Geophys. Res. 101, (1996b), 15071-15078.

Brunner, D., One-year climatology of nitrogen oxides and ozone in the tropopauseregion. Results from B-747 aircraft measurements, Ph. D. thesis,Dissertation Nr. 12556, ETH Zürich, (1998), 181pp.

Lamarque, J. F. and P. G. Hess, Cross-tropopause mass exchange and potentialvorticity budget in a simulated tropopause folding, J. Atmos. Sci. 51, (1994),2246-2269.

Marenco, A. et al., Measurement of ozone and water vapour by Airbus in-serviceaircraft: The MOZAIC airborne program, an overview, J. Geophys. Res.,(1998), in press.

Wernli, H., A Lagrangian-based analysis of extratropical cyclones. II: A detailedcase study, Quart. J. Roy. Meteor. Soc. 123, (1997), 1677-1706.

Wernli, H. and H. C. Davies, A Lagrangian-based analysis of extratropicalcyclones. I: The method and some applications, Quart. J. Roy. Meteor. Soc.123, (1997), 467-489.

Wirth, V., Diabatic heating in an axisymmetric cut-off cyclone and relatedstratosphere-troposphere exchange, Quart. J. Roy. Meteor. Soc. 121, (1995),127-147.

GLOREAM-report

74 of 216

Regional air quality studies using EURAD


M. Memmesheimer,H. Elbern, H.J. Jakobs, C. Kessler, J. Tippke, H. Feldmann, G. Piekorz,

H. Schmidt, E. Friese, A. Ebel and M. KerschgensUniversität zu Köln, Institut für Geophysik und Meteorologie, EURAD-Projekt,

Aachener Strasse 201 - 209, D-50931 Köln, Germany

1. Summary

The EURAD modeling system (European Air Pollution Dispersion Model) hasbeen further developed in collaboration with other groups within GLOREAM andother subprojects in EUROTRAC-2, in particular GENEMIS, TOR-2, AEROSOL,CMD, and LOOP. Emphasis in the model system development has been laid on theinterface with emission data, application of nesting to episode simulations, dataassimilation methods, including of aerosols and clouds, budget calculations,updates of chemical mechanisms and improvement of numerical schemes. Thedevelopment of strategies for the evaluation of air quality was an important issue.EURAD participated in a model evaluation carried out for one day during theTRACT field experiment. Applications of EURAD aims on the support of fieldexperiments (BERLIOZ, VOTALP, PIPAPO) and forecast of ozoneconcentrations. These activities have been performed within the framework ofresearch programmes funded by the BMBF (Federal Republic of Germany),DG XII, European Commission, MWF of Nordrhein-Westfalen and in closecooperation with the Ford Research Centre, Aachen.


The aims of the research are the further development and application of theEURAD model. Special emphasis is on the support of field experiments and theevaluation of the model. Budget calculations with respect to the dynamical andchemical processes which govern the temporal and spatial patterns of theconcentration fields are used as a tool to improve the knowledge on the interactionof the different processes and the exchange of air pollutants between differentregions in Europe. Further model development includes the improvement of theaerosol and cloud modules, the development of advanced data assimilationmethods using adjoint modeling techniques, the improvement and application ofnesting techniques, deposition, gas phase chemistry and numerical methods.

GLOREAM-report

75 of 216


The EURAD modeling system has been evaluated on the basis of observationsobtained during one day within the TRACT field campaign (Sept. 16, 1992;activity within the tropospheric research programme (TFS)). Other episodes havebeen prepared for evaluation within the evaluation group of the (TFS). Selected isa summersmog episode in July 1994. The areas considered for evaluation areNordrhein-Westfalen and Berlin (FLUMOB). The episode of the BERLIOZ fieldexperiment (July/August 1998) also has been selected for evaluation within TFS.Application to the Milano area and the nearby Alps has been undertaken within theEC-project VOTALP (valley experiment, foehn episode) or is planned for PIPAPOin close cooperation with LOOP. The nesting capabilities of the modeling systemhave been extended to consider regional and local effects with horizontalresolutions from 54 km down to 2 km (e.g. Berlin).

Budget calculations (Memmesheimer et al., 1997) have been performed on theEuropean scale for a summersmog episode in July/August 1990 (EUMAC-TORepisode) on the European scale, for the Berlin region on the basis of a photosmogepisode in July 1994 and for the VOTALP campaigns (Feldmann et al., 1998) Theeffect of the nesting on the processes involved in the budget calculations have beenstudied for the area of Berlin.

Advanced data assimilation methods have been developed on the basis of adjointmodeling (Elbern et al., 1997) within the TFS and the EC-project RIFTOZ. Thesemethods have been applied to an episode in July/August 1997. They are used toimprove initialization of air quality models, chemical state analysis, for sensitivitystudies and for parameter optimization such as emission rates. It allows for a betterunderstanding of chemical and dynamical processes governing the chemical stateof the atmosphere on the basis of observations.

Concerning model developement major efforts have been undertaken to improvethe parameterization of aerosols and clouds (Ackermann et al., 1998). Relationswith the subproject AEROSOL have been established to use the EURAD modelingsystem for the analysis of measurements and to get further ideas for modeldevelopment by a close cooperation between modelers and experimentalists. Thework performed with respect to the aerosols has been done in close cooperationwith the Ford Forschungszentrum, Aachen. The summersmog episode 1990 withthe well-defined frontal passage at its end has been selected as a test case for theimprovement of cloud modeling within EURAD. It also serves as a test episode toinclude the chemical mechanisms developed within the subproject CMD (gasphasechemistry and heterogeneous chemistry).

The interface to the emission data bases available within GENEMIS has beenimproved considerably. GENEMIS data provided by the IER, University ofStuttart, has been used to simulate the oxidant formation from the European scale

GLOREAM-report

76 of 216

to the urban scale of Berlin; landuse data genererated for Europe by the IFU,Garmisch-Partenkirchen, has been used to calculate biogenic emissions (VOCs andNOX) for the same episode (Memmesheimer et al., 1998).

The effect of stratosphere-troposphere exchange has been studied for severalepisodes and with different methods within the framework of the EC-projectsTOASTE-C and VOTALP.


The nesting capabilities of the EURAD model has been applied to zoom from theEuropean Scale into highly populated, industrialized regions with high emissionrates. Areas of specific interest are located in Nordrhein-Westfalen (Kessler et al.,1998), the Milano area (Feldmann et al., 1998) and in particular Berlin(Memmesheimer et al., 1998; see also fig. 1). Budget analysis has been carried outfor the urban area of Berlin to investigate the effect of nesting on the processeswhich control temporal development and spatial variations of photo-oxidants andtheir precursors in the Berline plume.

It could be shown that the contribution of different processes vary considerably fordifferent nest levels.

Aerosol dynamics and chemistry has been included into the model. First resultshave been obtained for a summersmog episode in July 1994 (fig. 2).

Data assimilation based on adjoint modeling has been tested for an episode inAugust 1997. It could be shown that the initial values can be improvedconsiderably (see fig. 3).

5. Main conclusions

The EURAD modeling system has been applied successfully to various scalesusing its nesting capabilities. The tools available within the EURAD modelingsystem has been developed further to improve the understanding of dynamical andchemical processes which control the atmospheric concentration fields (graphicaltools, budget analysis). Highly advanced data assimilation using adjoint modelinghas been developed and successfully applied. The implementation of aerosoldynamics and chemistry into a 3D Eulerian modeling system allows forapplications including the analysis of field experiments aiming on thecharacterization of aerosol patterns in different regions of Europe. EURAD isunder a permanent process of evaluation which improves the knowledge of therange of uncertainty in the model results and points to possible improvements inthe modeling system. The interface to GENEMIS, which has been established nowallows for the calculation of emission scenarios for Europe and smaller areas ofparticular interest on the basis of sophisticated and permanently improved

GLOREAM-report

77 of 216

emission data. The range of applicability of EURAD has been extended and allowsfor the preparation and analysis of field experiments (including aerosols) as well asfor the calculation of emission scenarios for air quality regulation policy ondifferent scales.


The EURAD modeling system will be applied to analyse the results of theBERLIOZ episode (July/August 1998). Analysis include budget calculations, dataassimilation and process oriented evaluation on the basis of observations.Evaluation of the summersmog episode in July 1994 will be continued forNordrhein-Westfalen and Berlin (FLUMOB). EURAD will support the modelingactivities for the PIPAPO measurement campaign in close cooperation with LOOP.It is planned to use budget analysis and data assimilation within the TOR-2.

7. Acknowledgements

EURAD is funded by the BMBF, Germany, within the tropospheric researchprogramme (TFS) and the aerosol research programme (AFS), DG XII of theEuropean Commission and the MWF, Nordrhein-Westfalen. Close cooperationand permanent support by the Ford Research Center is gratefully acknowledged.The numerical simulations have been supported by the RRZK, University ofCologne and the Research Center Jülich (ZAM, ICG2, ICG3). Support also camefrom the ECMWF, DWD, UBA, EMEP and a lot of environmental agencies allover Europe.

8. References


Elbern, H., H. Schmidt and A. Ebel, Variational Data Assimilation forTropospheric Chemistry Modeling for Tropospheric Chemistry Modeling, J.Geophys. Res. 102, (1997), 15967-15985.

Feldmann, H., M. Memmesheimer, A. Ebel, P. Seibert, G. Wotawa, H. Kromp-Kolb, T. Trickl, A. Prevot, Evaluation of a regional scale model for theAlpine region with data from the VOTALP project, EUROTRAC symposium1998, Garmisch-Partenkirchen, in press.

Kessler, Ch., A. Ebel, M. Memmesheimer, H.J. Jakobs, Small scale photo-smogsimulation with EURAD: A case study, In: Global and RegionalAtmospheric Modeling (eds.: H. Hass, I.J. Ackermann), Proceedings of thefirst GLOREAM workshop, Aachen, (1998).

Memmesheimer, M., M. Roemer and A. Ebel, Budget calculations for ozone andits precursors: seasonal and episodic features based on model simulations, J.Atmos. Chem. 28, (1997), 283-317.

GLOREAM-report

78 of 216

Memmesheimer, M., H.J. Jakobs, J. Tippke, A. Ebel, G. Piekorz, M. Weber, H.Geiss, S. Jansen, B. Wickert, R. Friedrich, U. Schwarz and G. Smiatek,Simulation of a summer-smog episode in July 1994 on the European andurban scale with special emphasis on the photo-oxidant plume of Berlin. In:EUROTRAC Symposium 1998, Garmisch-Partenkirchen, in press.

GLOREAM-report

79 of 216

BERLINHAMBURG

PARIS

ATHEN

LONDON

ROMA

WIEN

WARSZAWA

LISBOA

BUCURESTI

STOCKHOLM

MADRID BARCELONAISTANBUL ANKARA

IZMIR

MOSKAU

ST.PETERSBURG

KIEWCHARKOV

MINSK

BUDAPEST

CASABLANCA

ALGIER

BERLINHAMBURG

PARIS

ATHEN

LONDON

ROMA

WIEN

WARSZAWA

LISBOA

BUCURESTI

STOCKHOLM

MADRID BARCELONAISTANBUL ANKARA

IZMIR

MOSKAU

ST.PETERSBURG

KIEWCHARKOV

MINSK

BUDAPEST

CASABLANCA

ALGIER

Ozon [ppbv] July 27, 1994, 14 UTC Ozon [ppbv] July 27, 1994, 14 UTC

altitude 1500 maltitude 40 m

BERLIN

POTSDAM

COTTBUS

DESSAU

BERLIN

POTSDAM

COTTBUS

DESSAU

altitude 40 m altitude 40 m

Ozon [ppbv] July 27, 1994, 14 UTCNOx [ppbv] July 27, 1994, 14 UTC

Figure 1a : Ozone concentration as simulated in the near-surface layer with EURAD. High ozone values are found along the mediterranean coasts of Italy and France and over Central Europe.

Figure 1b : as 1a, but for layer 7 (about 1500 m) of the model. The transport towards Scandinavia in this altitude range can clearly be seen. Horizontal resolution for 1a,b is 54 km.

Figure 1c : Application of the nesting capabilities of the EURAD system. Shown is the NOx concentration for Berlin for Nest 3. The highways near Berlin can be identified by enhanced NOx. Horizontal resolution for 1c,d is 2 km.

Figure 1d : as 1d, but for ozone. The ozone plume of Berlin is located southeastward from the city due to northwesterly wind. The background of ozone is quite high during that day due to transport of photochemically aged air with enhanced concentrations of photo-oxidants.

APPLICATION TO A SUMMER SMOG EPISODE WITH NESTING

GLOREAM-report

80 of 216

O3 (40,35,4) Schmuecke

6 9 12 15 18 21 00

20

40

60

80

100 O3 (40,35,4) Schmuecke

6 9 12 15 18 21 00

20

40

60

80

100 O3 (40,35,4) Schmuecke

6 9 12 15 18 21 00

20

40

60

80

100

UTC

mix

ing

ratio

[ppb

v]

x x x : measured

: first guess

: result

Figure 3 shows the beneficial impact of optimizing initial concentrations with a 4D variational data assimilation scheme on chemistry transport modeling with the EURAD-CTM2. Crosses denote measurements of ozone at the station of Schmuecke for August 5, 1997, the dotted line is the model simulation without data assimilation and the solid line is the simulation after application of data assimilation. Measurements from 6:00 to 12:00 GMT are provided for the assimilation scheme, it can be seen that the simulation after assimilation is very close to the measurements during this time window. Measurements from 12:00 to midnight are not taken for assimilation but presented for showing the impact of data assimilation on a subsequent forecast. The influence of optimized initial values usually decreases with time due to insufficiently well known emissions, deposition rates and model bias.

Aerosol number concentration [1/cm**3]

July 24, 1994, 12 UTCNEAR SURFACE LAYER

B

Aerosol number concentration [1/cm**3]

1000

900

800

700

600

500

400

300

200

P

RE

SS

UR

E (

hPA

)

A=( 1, 26) ( 59, 26)=

100VERTICAL CROSS SECTION July 24, 1994, 12 UTC

APPLICATION OF THE EURAD-MADE SYSTEM

4DVAR DATA ASSIMILATION

OZONE

AUGUST 5, 1997

Figure 2 : Aerosol number concentration as calculated with the complex 3D Eulerian modeling system EURAD. The left panel shows the number concentration over Europe in the near surface layer which is about 40 m thick for this application for July 24, 1994, 12 UTC (14 CEST). The strongly polluted areas in Europe can clearly be identified. The right panel (2b) shows a vertical cross-section from West to East as indicated by the black line in the left panel (2a). The vertical cross-section extends from the surface up to 100 hPa which is about 15 km illustrate the polluted atmospheric boundary layer over the continent.

2a2b

GLOREAM-report

81 of 216

Study of the Feedback-Mechanisms between Climate Change andthe Chemical Composition of the Atmosphere


Johann Feichter and Gerhard LammelMax-Planck-Institute for Meteorology (MPI-HH)

Bundesstr. 55, D-20146 Hamburg, Germany

1. Summary

We performed an episodic simulation of the year 1993 with a global circulationmodel which treats the transport of sulfur species, chemical transformation andaerosol physics on-line with the meteorology. The aim is to evaluate the modelresults by comparison with measurements. This model version has also beenapplied in climate response experiments. Three experiments have been carried outwith a coupled atmosphere-sulfur chemistry-ocean model: one regarding only theeffect of the atmospheric greenhouse gases, one regarding additionally the directeffect due to anthropogenic sulfate particles and one regarding additionally theindirect sulfate effect. All numerical experiments have been integrated from 1860-2050. For the period 1860-1980 observed data for the anthropogenic SO2

emissions and for the greenhouse gas concentrations of CO2, CH4, N2O and someCFC's were used. For the period 1980-2050 estimated future emissions providedby IPCC were used. Changes in the ozone load of the troposphere and in the aciddeposition fluxes will be discussed as well. Additionally, we performed anexperiment to studying the effect of the reduction of ozone in the stratosphere.


The impact of changing concentrations of radiatively active gases and of aerosolparticles on the climate system is studied by the mean of a coupled atmosphere-ocean-chemistry model. Based on projections of the anthropogenic emissions(IPCC) the time-dependent climate response is studied. Additionally, we give anestimate of the tropospheric ozone budget and the acid deposition as expected foremissions projected into the future.


3.1 Evaluation

Newtonian relaxation has been applied in the climate model ECHAM4 to simulatea specific meteorological episode, the month of September 1993 (Feichter and

GLOREAM-report

82 of 216

Lohmann, 1999). For this purpose, surface pressure and the three-dimensionalwind and temperature fields were adjusted to ECMWF analyses. This methodallows the evaluation of climate model performance on smaller temporal scalesand comparison with short-term observational data. Since ECHAM4 will be usedto calculate short-lived trace gas and aerosol particle distributions, it is particularlyimportant that the model behaves well not only on climatological time scales -defined as 30-year averages - but also on time-scales on the order of days. Hence,the aim of this study was to demonstrate the ability of the model to reproduce thehydrological cycle and the distribution of some sulphur species on different spatialand temporal scales.

Three experiments have been performed: one with prescribed monthly mean sea-surface temperature (SST) but without any further adjustment (AMIP) and twoexperiments with relaxation of wind, surface pressure and temperature toobservations, one with the operational cloud physics (OPER) and one with newcloud microphysics which relates the calculated sulphate mass to the number ofcloud droplets (COUP). These model results have been compared to globalmonthly mean horizontal distributions derived from satellite data and ground basedobservations, to daily averages taken from the European EMEP and the CanadianCAPMON data base and to vertical profiles taken from aircraft measurementsperformed during the Northern Atlantic Regional Experiment (NARE).

First, we compared horizontal monthly mean distributions of liquid water path andprecipitation with observations. The liquid water path of the assimilated and of thenon-assimilated model versions looked very similar. This may be due to the factthat even in the non-assimilated version the observed SST is prescribed andsatellite data are only retrieved over the sea. The liquid water paths are mostlybetween the two observational estimates. Precipitation fluxes agreed very wellbetween observations and all model versions except for the ITCZ where all modelversions shift the maximum about 5° further north.

Comparing the three model versions we find that the distribution of precipitationflux of the assimilated versions (OPER and COUP) looks very similar. The sameholds for relative humidity and total cloud cover. The assimilation methodweakens the subgrid-scale vertical exchange and dampens slightly the hydrologicalcycle resulting in lower precipitation. The liquid water path in COUP issignificantly higher than in OPER due to aerosol-cloud interactions. This indicatesthat variables like relative humidity, cloud cover (which is parametrized dependingonly on relative humidity) and precipitation flux are mainly determined by thelarge-scale dynamics, whereas the cloud water content depends crucially on thechoice of the cloud microphysics parameterisation.

Next, we compared simulated precipitation fluxes, SO2 and sulphateconcentrations to daily mean averages over Europe and Canada. The agreement isgood, considering that a model using such a coarse grid(2.8°x2.8°) cannot

GLOREAM-report

83 of 216

reproduce extreme events. Events with no precipitation over a day are less likely tobe simulated within a model grid-box than they are to be observed at a specificmeasurement site. Analysis of the daily precipitation fluxes at 190 sites yieldedthat 86% and 82% of the sites using assimilated model versions OPER and COUP,respectively, are within a factor of two of the observations whereas in the non-assimilated version, AMIP, that is only the case in 62% of all sites. The temporalcorrelation between observations and calculations averaged over 67 sites yields acorrelation coefficient of 0.37, 0.33 and 0.17 for the experiments OPER, COUPand AMIP respectively. In contrast to the monthly mean distributions, theagreement of the daily averages of the assimilated model versions withobservations is significantly better than that of the non-assimilated model.

The agreement between sulphur dioxide and sulphate surface concentrations withobservations over Europa and Canada was worse than that for precipitation.Predicted sulphur dioxide concentrations at most sites were more than a factor oftwo higher than the observations. Sulphate is calculated to within a factor of twoof the observed values at 76% (OPER) and 38% (COUP) of the EMEP stations andat all CAPMON stations. Due to a lack of an actual emission data base we usedemission data valid for the fall of the year 1985. Emissions are known to havedeclined between 1985 and 1993, particularly in Germany where 22 of 34 sitesused reported a factor of two reduction in sulphur emissions. We clearly need abetter emission data base which is not only up to date but also reflects the spatialand temporal variability of the sulphur emissions at least over the region where theevaluation with observations is performed. In contrast to the absoluteconcentrations, the day-to-day variability is captured much better by the model.The observed and calculated time series consisting of 30 daily mean mixing ratiosof SO2 are significantly correlated at the 95% confidence level at 49% and 46% ofthe EMEP sites and at 40% and 60% at the CAPMON sites for OPER and COUP,respectively. Sulfate mixing ratios correlates at the 95% confidence level at 40%and 79% at the EMEP sites and at 50% and 60% at the CAPMON sites for theOPER and COUP experiment, respectively. To summarize, the assimilated modelversion is able to reproduce daily mean precipitation fluxes quite well.Additionally, it is able to predict the temporal variability of transport and wetremoval processes connected to the cyclonic activity in mid-latitudes. Thecorrelation for SO4

2- is significantly improved if interactions between cloudphysics and aerosol physics as in COUP are considered. We can only conclude thatthe first results of this assimilated model version are encouraging and thetechnique will be applied further to evaluate different model aspects in moredetail. In addition such model results provide useful informations for chemical off-line transport models, which are not included in other data bases, like liquid watercontent or precipitation formation rates.

GLOREAM-report

84 of 216

3.2 Climate response experiments

The experiments reported here are part of a comprehensive climate change studycarried out over the last two years at the Max Planck Institute for Meteorology inHamburg (Roeckner et al., 1998). The concentrations of the well-mixedgreenhouse gases like CO2, CH4, N2O and several industrial gases like the CFCsare prescribed as observed (1860-1980) and according to scenario IPCC-IS92athereafter. The space-time distribution of tropospheric ozone is prescribed, basedon pre-calculated fields from simulations with an atmospheric chemistry modelcoupled to the same AGCM employed in this study (Roelofs et al., 1998). Thetropospheric sulfur cycle is calculated within the coupled model using prescribedanthropogenic sulfur emissions of the past and projected until 2050 (IPCC-IS92a).The radiative impact of the aerosols is considered via both the direct and theindirect effect. The climate response is similar, but weaker, if aerosol effects areincluded in addition to greenhouse gases. One notable difference to previousexperiments is that the intensity of the global hydrological cycle becomes weakerin a warmer climate if both direct and indirect aerosol effects are included inaddition to the greenhouse gases.

We have then replaced the representation of stratospheric ozone in the controlexperiment with the observed ozone values (Bengtsson et al., 1998). We have heremade use of a compiled data set produced month by month for the time periodNovember 1978 to April 1993. We have linearized the trend for each month andfor each latitude band and extended it for the whole period 1979-97. This data setincludes thus the geographical variability, so the effect of the Antarctic ozone holehas been properly accounted for. The ongoing reduction of stratospheric ozoneresults in a clear cooling effect not only in the lower stratosphere but also in thetroposphere. The result is strongly dependent on the vertical distribution of theozone reduction, so more studies are required. It seems nevertheless that thestratospheric ozone change is highly important to include in climate changeexperiments. Due to the fact that the relative cooling effect of stratospheric ozoneis larger in the upper troposphere than in the lower part, it consequently reducesthe strong upper air warming by the greenhouse gases in better agreement withobservations.

3.3 Changing distributions of oropospheric ozone

Tropospheric ozone distributions with pre-industrial, present-day and futureemission scenarios has been calculated with a background tropospheric chemistrymodel coupled to the AGCM ECHAM4 (Roelofs and Lelieveld, 1997; Roelofs etal., 1998). In the pre-industrial atmosphere, the simulated annual troposphericozone content is 190 Tg O3. In the present-day simulation the ozone content isabout 80 Tg larger, mainly due to O3 precursor emissions from industrial processesin the NH and from biomass burning in the tropics. In the next decades, industrial

GLOREAM-report

85 of 216

growth is expected to occur mainly at NH subtropical latitudes, leading to anadditional increase of the tropospheric ozone budget by 60 Tg in 2045.

3.4 Changes in acid deposition

SOy (= SO2 + nss-sulfate) and NOy (= NO2 + HNO3 + nitrate) atmospherictransport, chemistry and deposition are simulated using the AGCM, ECHAM4(Roeckner et al., 1998; Roelofs et al., 1998). Oxidized nitrogen speciesconcentrations and dry and wet removal fluxes were taken from a modelexperiments for 1985 and 2045 (each 3 years of simulation, 19 vertical layers,3.75° * 3.75° horizontal resolutions, 30 min time step). An adequate set ofphotochemical reactions describe the O3 - CH4 - NOX gas-phase chemistry(Roelofs et al., 1998). NHy (= ammonia + ammonium) deposition fields are takenfrom a run of the global tracer transport model MOGUNTIA (10 layers, 10° * 10° ;Zimmermann et al., 1989) which uses monthly average temperature, winds andprecipitation and appropriate characteristics of the NHy cycle (Dentener andCrutzen, 1994).

Global atmospheric inputs to the continents and oceans amount today (1980 -1990)to 4.2, 4.7 and 11.7 mequiv m-2 a-1 for NHy-N, NOy-N and SOy-S, respectively,with dry and wet depositional fractions contributing almost equally (Lammel et al.,1998). These numbers are expected to increase by 70, 60 and 65 %, respectively,within 60 years. Large forested regions with the nitrogen input yet today exceeding15 kg N ha-1 a-1 are identified in Europe incl. parts of Russia, East and SoutheastAsia (China, Japan, India, Myanmar) and North America. On the regional scale itbecomes evident that Asian countries like China, Myanmar, Malaysia andIndonesia are increasingly affected, but also areas in Brazil, Ecuador, Venezuela,Perœ and Colombia. The model results show that nitrogen deposition in thecountries of the European Union (EU) is rather stable on a high level whereascountries like India, China or Brazil are facing considerable increases even beyondthe level of European deposition rates. Maximum and mean deposition ratesincrease more than twofold for China or threefold for India, respectively. Themaximum nitrogen deposition reaches levels of 50 kg N ha-1 a-1. Nearly the samepicture can be drawn for acid loads. Mean deposition rates in the EU countriesremain almost constant whereas maximum deposition significantly decreases. Aciddeposition in China raises three times and leads to acid inputs which double therates over Europe or northeastern America. The results show that acid depositionand nitrogen input is generally increasing over the continents within the next 60years. According to dynamics of population and industrial development south andsoutheast Asian regions are facing highest increases in deposition rates. In spite ofreductions in the sulfur emissions in Europe and North America a considerableenhancement of the anthropogenic sulfur emissions and of the acid deposition isexpected for the next decades in the developing countries. According to our modelcalculations based on the IPCC emission scenario the deposition of non-seasaltsulfate will increase in southeast Asia from 0.5 to 1.5 g S m-2 a-1. Globally it is

GLOREAM-report

86 of 216

projected that the emission and hence the acid deposition increases from71 Tg S a-1 in 1980 to 151 Tg S a-1 in 2050.

Yet today and even more in the near future (under a moderate emission scenario)acidification and nitrogen saturation of forest ecosystems will spread from aregional of the northern hemisphere (Europe, eastern North America, easternChina) to a global phenomenon affecting many emerging densely populated areas(in South America, south Asia, southeastern Asia) and remote but vulnerable areas(South America, central Africa, southeastern Asia, western Siberia).


We will perform more episodic studies to evaluate the model by comparing to in-situ measurements and to satellite data. Results of such experiments will beprovided as input for limited area models. Studies to explore the impact ofchanging emissions on the chemical composition of the atmosphere will becontinued.

5. Acknowledgements

The work carried out was funded by the EC project SINDICATE, ENV4-CT97-0483, and by the German BMBF project 07 AF312A.

6. References

Bengtsson, L., E. Roeckner and M. Stendel, Why is the global warming proceedingmuch slower than expected?, Report Max-Planck-Institut für MeteorologieNr. 256, Hamburg: Max-Planck-Institut für Meteorologie (1998) (acceptedfor publ. J. Geophys. Res.-Atmosphere).

Dentener, F.J. and P.J. Crutzen, A three-dimensional model of the global ammoniacycle, J. Atmos. Chem. 19, (1994), 331-369.

Feichter, J., U. Lohmann and I. Schult, The atmospheric sulfur cycle in ECHAM-4and its impact on the shortwave radiation, Clim. Dyn. 13, (1997), 235-246.

Feichter, J. and U. Lohmann, Can a relaxation technique be used to validate cloudsand sulphur species in a GCM?, Quart. J. Roy. Met. Soc. 125, (1999), 1277-1294.

IPCC - International Panel on Climate Change: Climate change 1992 - thesupplementary report to the IPCC scientific assessment (Houghton J.T.,B.A. Callander, S.K Varney., eds.), Cambridge University Press,Cambridge, (1992), 200 pp.

Lammel, G., G. Busch, F.J. Dentener, J. Feichter and G.J. Roelofs, Trends inglobal acids burden and deposition and vulnerability of forest ecosystems,Proceedings GLOREAM workshop, Sept. 1998, Madrid, (1998).

GLOREAM-report

87 of 216

Örn, G., U. Hansson and H. Rodhe, Historical emissions of anthropogenic sulfur:1860-1985, Report CM-91, Institute of Meteorology, Stockholm University,Stockholm, (1996).

Roeckner, E., K. Arpe, L. Bengtsson, M. Christoph, M. Claussen, L. Dümenil, M.Esch, M. Giorgetta, U. Schlese and U. Schulzweida, The atmosphericgeneral circulation model ECHAM-4: Model description and simulation ofpresent-day climate, Report Max-Planck-Institut für Meteorologie Nr. 218,Hamburg, Max-Planck-Institut für Meteorologie, (1996).

Roeckner, E., L. Bengtsson, J. Feichter, J. Lelieveld and H. Rodhe, Transientclimate change simulations with a coupled atmosphere-ocean GCMincluding the tropospheric sulfur cycle, Report Max-Planck-Institut fürMeteorologie Nr. 266, Hamburg, Max-Planck-Institut für Meteorologie(1998) (accepted for publ. J. Climate).

Roelofs, G.J. and J. Lelieveld, Model study of the influence of a cross tropopauseO3 transports on tropospheric O3 levels, Tellus 49B, (1997), 38-55.

Roelofs, G.J., J. Lelieveld and J. Feichter, Model simulations of the changingdistribution of ozone and its radiative forcing of climate: Past, present andfuture; Report Max-Planck-Institut für Meteorologie Nr. 283, Hamburg,Max-Planck-Institut für Meteorologie, (1998).

Zimmermann, P.H., J. Feichter, H.K Rath., P.J. Crutzen and W. Weiss, A globalthree-dimensional source-receptor model investigation using 85Kr, Atmos.Environ. 23, (1989), 25-35.

GLOREAM-report

88 of 216

Development of a new particulate model to study the impact oftraffic emissions on air quality


Ingmar J. Ackermann,Heinz Hass and Benedikt Schell

Ford Forschungszentrum Aachen GmbH, Süsterfeldstr. 200,D-52072 Aachen, Germany

1. Summary

Although quite a variety of air quality models for gaseous compounds are availablefor Europe, there is a lack of models that also provide information on particulatematter in the troposphere, although a great variety of relevant processes aresignificantly affected by these particles. Therefore the aerosol model MADE hasbeen developed and coupled to the EURAD model system. The model system issuitable for episode calculations from the regional to the local scale and theapplications will be focused on, however not limited to, the impact of vehicleemissions.


In order to achieve their scientific tasks state-of-the-art air quality models shouldbe capable of predicting particulate matter in addition to the gas-phaseconcentrations. A suitable aerosol model for the application in complex regionaltransport models has to:

− provide sufficient information on the chemical composition as well as on thesize distribution of the atmospheric particles.

− be coupled to a photochemical model to be able to represent the interactionsbetween the gas phase and the particle phase.

− cover the size range of atmospheric particles, i.e. several orders of magnitude.− be computationally efficient to keep the combined model system applicable.

The Modal Aerosol Dynamics Model for Europe (MADE) has been developed assuch an aerosol model based on the Regional Particulate Model (RPM; Binkowskiand Shankar, 1995) and successfully applied within the EURAD model system tothe simulation of tropospheric aerosols over Europe (Ackermann et al., 1998).However this system was limited to a certain part of atmospheric particulates andhas to be extended to cover the full chemical composition and size spectrum oftropospheric aerosols. Subsequently the model has to be applied - which requires

GLOREAM-report

89 of 216

the availability of suitable emission data - and tested by comparison to fieldmeasurements. Sensitivity studies then can provide insight into source typecontributions and the effectiveness of emission reduction measures.


Over the reporting period the aerosol model has been expanded to provide a morecomplete representation of the chemical composition of tropospheric particles anda more general representation of the particle size distribution.

In particular three primary particle components that are given by elemental carbon(EC), primary organic particles and primary particles smaller than 2.5 µm indiameter (PM2.5) have been added to the model formulation for the Aitken- and theaccumulation mode of the size distribution.

In addition to the inorganic ions given by Sulphate, Nitrate and Ammonium twoclasses of secondary organic aerosols (SOA anthropogenic and biogenic) arecalculated in the aerosol phase. These are formed by condensation of oxidationproducts of six different VOC classes of the RADM gas-phase mechanism.Additionally the model now treats sulphuric acid vapour explicitly in the gasphase. The chemical representation of the sub-micrometer particles is shownschematically in Figure 1.

Another extension of the model formulation is the introduction of an additionalmode that represents the coarse fraction of atmospheric particles. This mode isformed by the three model compounds soil derived, marine and anthropogenicaerosol. All of these are of primary origin. Due to their -relatively - large sizecoagulation can be neglected on these particles, however - in contrast to smallerparticles - sedimentational settling has to be considered. This is realised by theintroduction of an additional advection process into the model. Furthermore a newalternative nucleation scheme has been implemented into the model formulation.


− First test simulations with the extended model system of a coupled aerosol andgas-phase air quality model have been performed for a 10 day episode in July1994.

− A one way nesting scheme has been applied to the system with resolutions of27 km, 9 km, 3 km and 1km respectively.

− The aerosol model has been extended - among others - by the formation ofsecondary organic particles of biogenic and anthropogenic origin. In the testepisode anthropogenic and biogenic secondary organics showed approximatelythe same maximum concentrations but a different spatial distribution (Schell etal., 1999).

GLOREAM-report

90 of 216

Figure 1 Schematical representation of the fine particle (Aitken and accumulationmode) chemical composition with gas-phase precursors in MADE.

− Coarse mode particles from primary sources have been added as a new modeleading to a coverage of the complete size range of tropospheric particles bythe model.

− First sensitivity studies show that a reduction in inorganic aerosol precursorsubstances does not necessarily lead to an according reduction in inorganicparticulate matter. For example a reduced formation of .particles caused byreduced SO2 emissions can be (over)-compensated in some regions byadditional formation of nitrate aerosol (Ackermann et al., 1998; see alsoFigure 2).

5. Main conclusions

The modal aerosol dynamics model MADE has been extended and coupled to theEURAD model system, providing a model system that is capable of predictingtropospheric aerosol behaviour as well as interactions between the gas-phase andthe aerosol phase.

This model system provides the opportunity to study the source contributions toaerosol loads in the atmosphere and the response of concentrations to emissionreductions. First simulations show a very non-linear behaviour of the particulatematter that has to be taken into account in predicting the impact of any measurestaken to reduce particle loads.

NH3 NH4+

H+

H O2H O2

HNO3 NO3- H+

-2H+

SOAbiog.

SOAanthropog.

SO42-

H SO42

Vapor EC

PrimaryOrganics

PM2.5

Emis-sions

HC8

XYL

TOL

OLT

OLI

CSL

Products

Products

OHO3

NO3 SO2

CloudsOH

Emis-sions

Pre-cursors

API LIM

OH O3 NO3

GLOREAM-report

91 of 216


During the next period the completed model system will be applied to a summersmog episode from a regional (Europe) to an urban (Greater Cologne) scale with aone-way nesting procedure and an appropriate emission data set for primaryparticles. Results from a global circulation model will be used for initialisation andboundary values of the regional simulation.

Emission data sets that differentiate between different source categories will beused to evaluate source type contributions to the tropospheric aerosol loading andto quantify the contributions of primary and secondary particles to this load. Themodel results will be compared to field measurements of atmospheric particles asthese data become available.

7. Acknowledgements

This work was partly supported by the German state of NorthRhine-Westfaliaunder contract number IV-A4-2655.16.

8. References


Ackermann, I.J., H. Hass, B. Schell and F.S. Binkowski, Regional modelling ofparticulate matter with MADE, submitted to Env. Man. & Health Perspec.(1998).

Binkowski, F.S. and U. Shankar, The regional particulate matter model 1. Modeldescription and preliminary results, J. Geophys. Res. 100, (1995), 26191-26209.

Schell, B., I.J. Ackermann, H. Hass and A. Ebel, Secondary organic aerosolmodelling with MADE: biogenic and anthropogenic contributions,Proceedings of EUROTRAC Symposium ’98, Editors: P.M. Borell and P.Borell, WITpress, Southampton, (1999).

GLOREAM-report

92 of 216

GM

GMGM

Figure 2 Response of accumulation mode sulphate aerosol (top) and nitrate aerosol(bottom) concentrations to a 50% reduction in SO2 emissions as differencebetween reduction and base case simulations in µg/m3 (surface layer, July 231994 0 GMT).

GLOREAM-report

93 of 216

Regional Modelling with the Chemistry Transport Model TM3-K


H. Kelder,P.F.J. van Velthoven, J.W.M. Cuijpers, A. Jeuken, E. Meijer, and G. Verver

Royal Netherlands Meteorological InstituteP.O. Box 201, NL 3730 AE De Bilt, The Netherlands

1. Introduction

The global Chemistry Transport Model (TM3) has been developed duringEurotrac-1’s Glomac project and is now used by several groups for a number ofyears. It originates from the TM2 model described by Heimann (1998). Theversion of TM3 used at KNMI (TM3-K) has the option of being used in a regionalor hemispheric mode which is suitable for air quality simulations. Asmeteorological input 6-hourly ECMWF analyses on 31 or 19 vertical hybridsigma-pressure levels are used. The finest horizontal resolution used is 2.5 degreesfor global and hemispheric coverage and can be down to 1 degree when coveringsmaller regions. The model contains parametrisations of the boundary layer andconvective transports. A chemistry module including free tropospheric ozonechemistry has been extensively used and validated in the free troposphere(Wauben et al., 1998).

The regional version has been developed in GLOREAM in order to facilitatecomparisons with instantaneous trace gas and pollutant measurements. In view ofthe realistic description of the meteorology in TM3 the model is very well suitedfor application in measurement campaigns or for studies of pollution episodes, butit can also be used for simulation periods up to several years including scenariostudies. An example is the study of the effects of past, present and future air trafficupon the composition of the troposphere (e.g. Wauben et al., 1997). The model iscoupled to a sophisticated radiative transfer scheme, originally developed for theECHAM model.

In GLOREAM we are focusing, as far as the dynamics are concerned, on theparametrisations of the exchange of pollutants between the boundary layer and thefree troposphere, the interaction between chemistry and turbulence in the boundarylayer, and the (convective) vertical transports into the free troposphere. Anotherfocus is the validation of the budgets of chemical constituents, such as ozone,nitrogen oxides and sulphate, by using observations available over Europe.Furthermore, data assimilation techniques are developed to make optimal use of

GLOREAM-report

94 of 216

observations for model validation. An overview of the results until now is givenbelow.

2. Improvements in the dynamical description of the boundary layer

Until now the parametrisation for turbulent diffusion in TM3 has been the oneoriginally used in the ECMWF model, as described by Louis (1979). As theECMWF input is only available at 6-hourly intervals, this scheme does not give arealistic description of the diurnal evolution of the boundary layer (BL) height. Wehave included in TM3 a new parametrisation, which was originally developed byHoltslag and Boville (1993), and further developed within the regional climatemodel, RACMO, of KNMI. Main differences with the Louis scheme are that theBL-height is explicitly calculated and that separate formulations for the diffusioncoefficients are applied for the BL and free troposphere. It allows for the inclusionof a function that determines the diurnal variation of the BL height at every timestep in TM3.

3. Interaction between chemistry and turbulence in the boundary layer

The conventional approach in atmospheric transport/chemistry modelling is todescribe turbulent transport of reactive tracers similar to that of inert tracers, suchas water vapour or sensible heat. It is also generally assumed that concentrationfluctuations of reactive tracers are not correlated, and thus that concentration co-variances can be neglected. These assumptions were tested for a single irreversiblebimolecular reaction using a second order closure model in Verver et al. (1997).They show that, depending on reaction rates, the turbulent exchange coefficientwill be affected significantly by the chemical reaction. Furthermore they find forthis simple case that anti-correlated concentration fluctuations may alter effectivetransformation rates by more than 80%. A slightly more realistic case was studiedby Verver (1998) describing NO-NO2-O3-hydrocarbon chemistry that includes 3chemical reactions. A small effect was found on vertical concentration gradients,while boundary-layer averaged concentrations remained nearly unchanged. Thechemistry scheme was extended with a comprehensive set of chemical reactionsthat describes the oxidation of isoprene (Verver, 1998). With this set of chemicalreactions, the model is able to reproduce fairly well the observed concentrations ofO3, NOx and isoprene and its reaction products during the Amazon BoundaryLayer Experiment (ABLE-2a). For the fully developed boundary layer in theafternoon, the turbulent flux of NO and NOx is altered by nearly 30% when thecovariance of concentration fluctuations as well as chemistry effects on the fluxare explicitly taken into account. In that case we find a change of the meanconcentrations of some short-living radicals of roughly 10%. However, theconcentration profiles NO, NO2 and most other stable reaction products remainunchanged. Since most global and regional scale models aim to correctly representthe distribution of mean concentrations of the longer-living species, we conclude

GLOREAM-report

95 of 216

that neglecting covariance effects and chemistry effects on the flux is justified forthe cases that are studied.

4. Modelling the budgets of pollutants over Europe and the US

In TM3 a module describing the life-cycle and budget of sulphur compounds wasincluded. The module includes emissions, gas- and liquid phase chemistry, and dryand wet deposition for sulphur dioxide, sulphate, and DMS. In the description ofcloud processes, such as wet deposition, use is made of newly available archived3-dimensional cloud parameter distributions from the ECMWF model.Comparison to in situ aircraft measurements in the free troposphere hasdemonstrated the good performance of the model in simulating the 3-dimensionalcloud distribution (Ovarlez et al., 1999). The model simulated distribution ofsulphur compounds is found to correspond well with available surfacemeasurements over Europe and the US, as well as with GOME and ATSR-2satellite observations. The absolute values of the SO2 concentrations appear to betoo high over the European continent in winter. This is a problem that occurs inmany models as was found during the 1998 sulphate model inter-comparisonexercise, organised by the World Climate Research Programme (WCRP) and theInternational Global Atmospheric Chemistry project (IGAC), in which weparticipated. Work is continuing to investigate this problem.

5. Use of data assimilation techniques

An optimal-interpolation data-assimilation technique has been implemented(Jeuken et al., 1999). It has been tested with TOVS total-ozone columns in aversion of the TM3 model with simplified chemistry. Starting from a separableform of the forecast covariance matrix, the optimal interpolation equation wasrewritten into a horizontal and vertical analysis step. First, the measurements areanalysed using estimates for the horizontal error co-variances. In the second stepof the assimilation procedure the analysis increment for the column is distributedover the vertical model layers. For ozone columns this step depends only on anormalised vertical weight function which is equal to the vertical covariance.Three different estimates for this weight function were introduced, using either theozone error covariance, or the ozone time variance, or the actual ozone mass todistribute column corrections in the vertical. A comparison with independentobservations showed a considerable improvement of the total ozone field due toassimilation. The model error growth is small, making it suitable for assimilatingsparse measurements. Ozone profiles from the assimilation appear realistic andclose to the ones observed by sondes. The model is quite performing in describingozone profile structures tropopause region. Due to the absence of verticalinformation in the observations used until now, the assimilation had only littleimpact on the shape of the vertical ozone profile, which is mainly determined bythe model transport. Further research will concentrate on the assimilation of heightresolved constituent observations.

GLOREAM-report

96 of 216

6. Conclusions and outlook

We have improved the description of the diurnal evolution of the boundary layer inthe model. Tests with tracers with simple sources and sinks will be used toevaluate the difference between the old and new parametrisations.

We have studied the importance of chemistry-turbulence interactions in theboundary layer with an extensive set of chemical reactions. In contrast to studieswith highly idealised chemical systems, we found no large impact of taking intoaccount higher order chemistry terms. Hence, for the moment, there are not enougharguments to include a parametrisation of chemistry-turbulence interaction in thechemistry transport model simulations.

The simulated distributions of sulphur dioxide and sulphate were found tocorrespond generally well with available surface measurements over Europe andthe US, as well as with globally available GOME and ATSR-2 satelliteobservations. There are still some problems with wintertime SO2 concentrationsover Europe that need to be further investigated. Possibly there is a need for anupdate of the used emissions over Europe which were those from GEIA.

The use of data assimilation techniques that allow an efficient and consistentintercomparison between model values and observations is a new development thatwill be continued in the coming years.

7. References

Heimann, M., The global atmospheric tracer model TM2: DKRZ TM2 modeldocumentation, Tech. Rep. 10, Max Planck Institute fur Meteorlogie,Hamburg, (1995).

Holtslag, A.A.M. and B.A. Boville, Local versus nonlocal boundary-layerdiffusion in a global climate model, J. Climate 6, (1993), 1825-1842.

Jeuken, A.B.M., H.J. Eskes, P.F.J. van Velthoven, H.M. Kelder and E.V. Holm,Assimilation of total ozone satellite measurements in a three-dimensionaltracer transport model, J. Geophys. Res., (1999), in press.

Louis, J.F., A parametric model of vertical eddy fluxes in the atmosphere, Bound.Layer Meteorol. 17, (1979), 187-202.

Ovarlez, J., H. Ovarlez, P. van Velthoven, G. Sachse, S. Vay and H. Schlager,Comparison of water vapor measurements from POLINAT2 with ECMWFanalyses in high humidity conditions, to be submitted to J. Geophys. Res.,(1999).

Verver, G.H.L., H. van Dop and A.A.M. Holtslag, Turbulent mixing of reactivegases in the convective boundary layer, Bound. Layer Meteorol. 85, (1997),197-222.

Verver, G.H.L., Mixing of Reactive Gases in the Convective Boundary Layer,Phys. and Chem. of the Earth 23, (1998), 673-677.

GLOREAM-report

97 of 216

Wauben, W.M.F., P.F.J. van Velthoven and H.M. Kelder, A 3D chemistrytransport model study of changes in Atmospheric ozone due to aircraftemissions, Atmos. Environ. 31, (1997), 1819-1836.

Wauben, W.M.F., J.P.F. Fortuin, P.F.J. van Velthoven and H.M. Kelder,Comparison of modelled ozone distributions with sonde and satelliteobservations, J. Geophys. Res. 103, (1998), 3511-3530.

GLOREAM-report

98 of 216

Environmental Impacts of Air Pollutants fromSwiss Energy Systems


Johannes Keller,Sebnem Andreani-Aksoyoglu and Daniel Bürki

Paul Scherrer Institute (PSI), Air Pollution, CH-5232 Villigen PSI, Switzerland

1. Summary

Within the frame of the comprehensive assessment of Swiss energy systems, airquality simulations were performed by using the 3-dimensional photo-chemicaldispersion model UAM (Urban Airshed Model). The objective is to investigate theimpacts of pollutants in Switzerland for future scenarios of Swiss energy systems.We investigated four options: base case: simulations with the projected emissionsfor the year 2030; scenario 1: all nuclear power plants were replaced by oil-drivencombined cycle plants (CCP); scenarios 2 to 4: traffic emissions were reduced inthe whole country as well as in the cities and on the highways, separately. Inaddition, we proposed an approach to calculate seasonal impacts on the basis oftypical air pollution episodes. Currently, we are replacing the air quality model bythe latest version UAM-V with variable grid .


The objective is to evaluate and to assess the impacts of air pollutants on theenvironment due to conversion processes in energy systems (e.g. power plants,vehicles, etc.) operating in Switzerland. These impacts will primarily cover thedeterioration of ecosystems due to exposure to ozone, acid deposition and nitrogenaccumulation.


We tested the air quality model UAM on the basis of measured data of the Swissfield experiment POLLUMET taken on July 28-30, 1993 (Andreani and Keller,1998). The areas where the production of photochemical pollutants is either NOx

or VOC sensitive were derived by varying the NOx and VOC emissions separately.

Subsequently, four energy/emission scenarios were defined: in the base case, theemission inventory is extrapolated to the year 2030 according to officialprojections; in scenario 1), all nuclear power plants were assumed to be replacedby oil-driven combined cycle plants (CCP); in scenarios 2-4), traffic emissions

GLOREAM-report

99 of 216

were reduced by 30% in the whole country as well as in the cities and on thehighways, separately. Concentrations and deposition rates of the most importantpollutants were derived on the basis of the summer smog meteorology during thefield experiment mentioned above. Regions were identified where the peakconcentrations exceed the critical levels for plants or the limits to protect humanhealth.

A PhD student has been working with us since April 1998. Currently she isimplementing and testing the new version of UAM (UAM-V with variable grid).Later, she will investigate the influence of the boundary levels on the Swiss airquality. In this context the use of the European models EURAD and LOTOS,which are applied within the GLOREAM subproject, has been initiated.

The estimation of long-term impacts on ecosystems requires concentration anddeposition data of air pollutants for the growing season or the whole vegetationperiod. As an alternative to air quality simulations on a day by day basis, weclassified the meteorological conditions in Switzerland into a few categories withsimilar air pollution characteristics. We will simulate the air quality only for thesecharacteristic conditions and weight them according to their frequency ofoccurrence.


Model testing

In Figure 1 the monitored and the simulated diurnal ozone profiles at ground levelfor the last day of the POLLUMET period are compared at four different locations.Zurich is an urban station whereas Duebendorf is a sub-urban one. Laegern andChaumont are both rural, the first below 1000 m, the latter above. The goodagreement at the rural stations is remarkable. In the urban and sub-urban areas, theagreement at the time of peak ozone is very good. However, at night and early inthe morning, measured concentrations are very low due to the local effect of O3-NO titration which cannot be resolved in the 5 km x 5 km grid cells.

The grid cells sensitive to NOx or VOC emissions were identified by reducingthese emissions by 35% separately. Ozone formation was predicted to be sensitiveto NOx or to both emissions in most of the model region. The VOC-sensitive areaswere found to be dependent on the wind regime. More grid cells were identified asVOC-sensitive on July 29 due to the weaker wind, compared to the previous day.

GLOREAM-report

100 of 216

Figure 1 Calculated (solid line) and measured (dotted line) concentrations of ozone asa function of time on 30.7.1993 at 4 stations. Zurich: urban, Duebendorf:suburban, Laegern: rural < 1000 m. Chaumont: rural > 1000 m.

Scenario calculations

Base case: The hourly concentrations of CO, PAN, SO2, NO2, and O3, werecompared with the critical levels for short-term impacts on plants and with thelimits given to protect the public health. The predicted CO levels (daily average~400 ppb) are far below the limits for health effects. The PAN concentrationswhich can cause acute damages on plants (bi-hourly average10-20 ppb) are notreached. The maximum of the concentrations over the domain was 4.4 ppb.Similarly, SO2 levels are too low to cause any damage to plants or public. Theconcentrations of NO2 are close to the limits. On the other hand, the ozoneconcentrations exceed both the critical level for plants (75 ppb) and the limit toprotect the public health (60 ppb).

Scenario 1: The cover of the electricity deficit in 2030 (which is mainly caused bythe closure of nuclear power plants) by four oil-driven CCP is an extreme case.The simulation showed local effects in the areas where the power plants areplaced. Compared to the base case, NO2 levels increase and approach the limits.There is no significant change in CO concentrations. However, SO2 concentrationsincrease dramatically and exceed the critical levels for plants and approach thelimits given to protect the human health. Ozone levels increased by up to 10% inthe north-east of the power plants no. 3 and 4 located close together (see Figure 2).On the other hand, they decreased by the same amount in the east of the powerplants. Short-term depositions of SO2 and NO2 increased locally 15 and 8 times,respectively. Deposition of HNO3 was about 30% higher.

GLOREAM-report

101 of 216

Scenario 2: There was a decrease in daily average concentrations of NO2 and CO.SO2 concentrations were practically not affected because SO2 emissions from thetraffic in Switzerland are quite low compared to other sources. O3 levels decreasedin the countryside by 1-5% (see Figure 2). However, they increased by 10-15%around the towns. This is due to the well known fact that reduction of NOx

emissions may lead to an increase of O3 if the formation is mainly VOC limited.Dry depositions of SO2, HNO3 and NO2 decreased by 7, 20, and 40%, respectively.

Scenario 3: CO levels decreased by about 12%. This scenario predicted almost nodecrease in O3 levels. The concentrations even increased around the big cities. Theeffects on deposition are found only around the cities. HNO3 and SO2 depositionsdecreased by 6%, and about 23% decrease was predicted for NO2 depositions. Onthe other hand, ozone deposition increased by about 10% in big cities such asGeneva and Zurich.

Scenario 4: The pattern is similar to that of scenario 2, but the changes aresmaller.

Figure 2 Changes (%) in peak O3 concentrations in scenario 1 (left) and scenario 2 (right) with respect tothe base case. 1: Goesgen, 2: Muehleberg, 3: Beznau, 4: Leibstadt. Contour lines: -5,-3,-1,5,10,15. Darker areas: increased concentrations, lighter areas: decreased concentrations.

Meteorology

The diurnal variations of O3 and NO2 levels monitored at three sites in Switzerland(Payerne, Tänikon, Magadino) were taken as a measure for the air quality. Levelthresholds at specified hours of the day were set to separate high and low O3 andNO2 conditions. For each class a typical pattern of average windspeed, airtemperature and global irradiance (Figure 3) corresponding to characteristicseasons was found. For the estimation of the impact of ozone on ecosystems on thebasis of the AOT40 concept, only the classes 1, 2 and 6 are relevant.

GLOREAM-report

102 of 216

5. Main conclusions

The evaluation of the Urban Airshed Model performance in Switzerland for thePOLLUMET field experiment indicated reasonably good predictions of most ofthe pollutant concentrations. Critical levels for the short-term impacts on the plantswere predicted to be exceeded for NO2 and ozone on days with weak wind. Ozoneconcentrations exceeded the limits given to protect the public health. Scenariocalculations lead to the following conclusion: Replacing the existing nuclearpower plants with oil-driven combined cycle plants can have significant localeffects especially on the SO2 concentrations and depositions. Short-term criticallevels for plants would be exceeded in that case. An increase in NO2

concentrations can also be expected around the power plants. Reducing the trafficemissions in the whole country seems to be more effective than measures limitedto the cities or to the highways. NO2, CO and O3 concentrations decrease byreduction of traffic emissions, although O3 levels might increase around the high-emission areas. Short-term changes in pollutant depositions indicate possibility ofsubstantial variations in loads over longer periods. The calculations give usefulestimates about the severeness of air pollution concerning short-term plantdamages and public health today, and also for some future scenarios.

Air quality can be classified into 6 categories, 3 of them being relevant for summersmog episodes. Within one single category, however, there is still a significantvariability. We are optimistic that the characteristic air quality of each class can besimulated by parameterizing the dependence of the UAM results on themeteorology and by computing seasonal levels and loads by weighting thecategories according to their frequency of occurence.

Figure 3 Diurnal variation of O3 and NO2 concentrations for air quality classes1 to 6. Payerne 1993.

GLOREAM-report

103 of 216

6. Possible policy relevance

The results are intended to be used as an important element for the comprehensiveassessment of Swiss energy systems. This assessment covers economical andecological aspects, health risks and global impacts (PSI Project GaBE). Theoutcome of that project will be of interest for Swiss decision makers, utilities, etc.


We will test the new version UAM-V and compare the results with those obtainedwith the previous version. Preliminary boundary conditions will be replaced by theLOTOS and the EURAD data to investigate the transboundary influence on airquality and impacts in Switzerland. We plan to simulate typical patterns of therelevant pollutants for each air quality category to estimate the seasonal impacts.As a first case, we will model a winter smog episode taking wet deposition intoaccount. Finally, we intend to include particulate matter.

As a contribution to the EUROTRAC subproject LOOP, we will perform airquality simulations with UAM-V for the Milano area.

8. Acknowledgements

This study was supported by the “Kommission für Technologie und Innovation(KTI)”.

9. References

Andreani-Aksoyoglu S. and J.Keller, Short-term impacts of air pollutants inSwitzerland: Model evaluation and preliminary scenario calculations forselectet Swiss energy systems, in: C.A.Brebbia, C.F.Ratto, H.Power (eds),Air Pollution VI, WIT Press/Computational Mechanics Publications,Southampton, (1998), 799-808.

Keller J., D.Buerki, S.Andreani-Aksoyoglu, An approach for the estimation ofseasonal impacts of air pollutants in Switzerland on the basis of an airquality classification, Proc. EUROTRAC-2 Symposium 98, March 23-27,Garmisch-Partenkirchen, (1998).

GLOREAM-report

104 of 216

Operational Air Pollution Forecast


Sissi Kiilsholm,Jerome Chenevez, Allan Gross, Alix Rasmussen, Jens H. Sørensen

Danish Meteorological InstituteLyngbyvej 100, DK-2100 Copenhagen, Denmark

1. Summary

In an operational air pollution forecast model a statistical after-treatment byKalman filtering has given improved ozone forecasts. A verification methodsuggested by the EEA - European Air Quality Topic Centre has been useful incomparative verifications of the ozone forecasts for 1995-96.


The main purpose of these studies is to evaluate and optimize the performance ofreceptor point models for operational air pollution forecasting, with special focuson model verification, improvement of model results by statistical after-treatment,refinement of chemical solvers and development of presentations of air pollutionforecasts to the public.


The model used in these studies are DACFOS (Danish Atmospheric ChemistryForecasting System) (Jensen et.al., 1996). The system is based on a coupling of thechemical routine of the EMEP MSC-W oxidant model and DMI’s 3-D Lagrangiantransport model utilizing analysis and forecast data from DMI’s numerical weatherprediction model DMI-HIRLAM (High Resolution Limited Area Model).

Kalman filtering of DACFOS

A statistical after-treatment of the DACFOS’ results with the Kalman filterprovides a better description of the local variations of ozone concentration. Theozone concentration depends on the meteorological conditions. Given ozoneconcentration measurements, the Kalman filter combines DACFOS’ ozoneforecast with HIRLAM’s meteorological data to adjust the ozone forecast. Thevariations of parameters such as temperature, wind speed, surface heat flux andAtmospheric Boundary Layer- height are compared with DACFOS’ ozone forecastand measured ozone concentrations during four days. It is then possible to filter

GLOREAM-report

105 of 216

DACFOS’ ozone forecast to predict the following two days ozone concentration.This statistical after-treatment is performed for stations where real time ozoneobservations are available.

Several tests of different combinations of meteorological parameters have beencarried out using ozone observations during the Summer 1997 in two stations inNorth Zealand (Denmark) and three stations in England. A statistical analysiscomparing observations and 48-hour forecast presents better results than DACFOSforecast for a few Kalman filters using specific combinations of the meteorologicalparameters. The correlation coefficients for observed concentrations and Kalmanfilters forecasts are between 0.50 and 0.70.

Figure 1 shows the daily maximum ozone concentrations in Jægersborg (Denmark)during the Summer 1997 for observations, DACFOS and a Kalman filtercombining DACFOS’ ozone forecast with HIRLAM’s surface temperature and10 m wind speed. The observed [O3] in Jægersborg exceeded the threshold of60 ppb eight days in the Summer 1997. The correlation between daily maximumKalman filter’s [O3] and daily maximum observed [O3] is 0,63 while DACFOS’daily maximum values are slightly anti-correlated with the observations. One cancalculate the correlation factors between observed ozone concentration andKalman filter and with DACFOS ozone concentrations respectively at differenttime after forecast origin. The evolution of these correlation factors versus forecasttime is displayed in figure 2. The corresponding correlation factors compile thefour daily forecasts of DACFOS and the Kalman filter during the whole periodbetween 03/06/97 and 23/08/97.

0 10 20 30 40 50 60 70Number of days

0

20

40

60

80

Daily maximum [O3] in JeagersborgPeriod between 03/06/97 and 23/08/97

ObservationsDACFOS (corr = −0.08)Kalman filter (corr = 0.63)

June July August0

0,1

0,2

0,3

0,4

0,5

0,6

0,7

0,8

0,9

1

0 h 6 h 12 h 18 h 24 h 30 h 36 h 42 h 48 h

KF

Forecast length (hours)

Figure 1 Daily maximum [O3] for the Summer1997 in Jægersborg ; comparisonbetween DACFOS, KF and observations.

Figure 2 Correlation coefficients for DACFOS andKF in Jægersborg for the period between03/06/97 and 23/08/97.

GLOREAM-report

106 of 216

Figure 3 compares the correlation coefficients (Fig. 3a) and the rms values(Fig. 3b) of three Kalman filters (KF1 is the same as before but without the windspeed, KF2 is exactly the same as before and KF3 is KF2 including HIRLAM’sAtmospheric Boundary Layer height) with DACFOS for the five monitoringstations used in the present study. These diagrams show that the performances ofthe Kalman filters are comparable for all the five monitoring stations and alwaysbetter than DACFOS’.

Correlation coefficients

0

0,1

0,2

0,3

0,4

0,5

0,6

0,7

0,8

0,9

1

Jægersborg(DK)

Lille Valby (DK)

Harwell (GB) Ladybower(GB)

Strath Vaich(GB)

KF 1

KF 2

KF 3

DACFOS

Root Mean Square

0

2

4

6

8

10

12

14

16

18

20

Jægersborg(DK)

Lille Valby (DK)

Harwell (GB) Ladybower(GB)

Strath Vaich(GB)

KF 1

KF 2

KF 3

DACFOS

Figure 3a Correlation coefficients of 3 differentKalman filters and DACFOS for fivemonitoring stations during the Summer1997.

Figure 3b RMS values of 3 different Kalman filtersand DACFOS for five monitoring stationsduring the Summer 1997

Since May 1998 the Kalman filter is operational for the DMI’s station inJægersborg with four different combinations of meteorological data, including thesurface heat flux, which was not available in the Summer 1997. The same set ofKalman filters runs in a semi-operational setup for two other stations in Denmark(Lille Valby and Keldsnor) and the three English stations. As an example, thefigure 4 compares 48-hour forecast results and observations for the 14/06/98 00H(UTC) at the DMU’s station in Keldsnor. The two first Kalman filters have giventhe best statistical performances in the Summer 1997. The two other, by takingaccount of other meteorological data, complete the two former and enlarge thefield of possibilities to forecast the ozone concentration in different meteorologicalconditions.

GLOREAM-report

107 of 216

0 10 20 30 40 50time after 98061400 (hours)

0

20

40

60[O

3] (p

pb)

Kalman filter in Keldsnor using:Dacfos’ mean and Hirlam’s Temperature

Kalman FilterDACFOSObservations

0 10 20 30 40 50time after 98061400 (hours)

0

20

40

60

[O3]

(ppb

)

Kalman filter in Keldsnor using:Dacfos’ mean, Hirlam’s Temperature and Wind


0 10 20 30 40 50time after 98061400 (hours)

0

10

20

30

40

50

60

[O3]

(ppb

)

Kalman filter in Keldsnor using:Dacfos’ mean, Hirlam’s Temperature, Wind and Heat


0 10 20 30 40 50time after 98061400 (hours)

0

10

20

30

40

50

60

[O3]

(ppb

)

Kalman filter in Keldsnor using:Dacfos’ 5 levels, Hirlam’s Temperature, Wind and ABL


Figure 4 Results of the four operational Kalman filters in Keldsnor for a forecast origin onthe 14/06/98 at 00H (UTC).

Verification using contingency table and scores

The EEA European Air Quality Topic Center (EAQ-TC) has proposed a methodfor verification of the skill performance of ozone forecast models, which givesuseful and easy comparable results. The method can be used when the chosen‘event threshold’ is low enough to give some events both for measurements and forforecasts, otherwise undefined values for some of the parameters will beperformed. Due to very few exceedances of the EAQ-TC proposed ozoneexceedance threshold in Denmark, an alteration of the exceedance threshold from180 µg to 120 µg has been made. Unfortunately, this modification makes itimpossible to compare with other model verifications, based on the EAQ-TCproposed verification principles, but intercomparisons in time and space at onemodel are still possible.

Six calculated statistical parameters, three derived from the contingency table: SR(Frequency of hits), SP (Probability of detection) and SI (Success Index), twoscores: H, hit score, and S, skill score, and the fractional bias FB*, are presented ina rose, Fig. 5. The statistical parameters are calculated for the model forecast andthe persistency case.

GLOREAM-report

108 of 216

A verification for the summer seasons 1995 and 1996, in Jægersborg, showsdifferent results, partly caused by the change of the weather conditions, and partlycaused by the few numbers of events. In general, the numbers of forecastedexceedances are lower than the number of observed exceedances, giving a SPlower than SR for the forecast skills. The observed exceedances occur aroundtwice as often as the forecasted exceedances. The best forecasted year was 1995,with ‘skill scores’, S, higher than persistency. Persistency are doing well in 1996for ‘hit and skill score’, H and S, caused by exceedances on consecutive days.

Figure 5 Statistical parameters for Jægersborg calculated for the years 1995-1996,based on DACFOS 24-29 hour forecasts. Model and persistency values areshown in the same piece, where the lowest are overlaying the highest. FB*differs from the other parameters by having an ideal value at 50%.

4. Main conclusion

The Kalman filtered ozone forecasts gives generally smaller biases and rms valuesthan the raw ozone forecasts (DACFOS), and the common slight underestimationof DACFOS is generally well compensated by the Kalman filters. The forecastsobtained by the different Kalman filters are not always alike and the best forecastis not necessary given by the same filter. The performances of the Kalman filtersare very dependent on the accuracy of the meteorological data. The average of thefour different Kalman filters is then used in the operational system in order to givethe best possible forecast according to the meteorological conditions.

The comparative verification for the year 1995 to 1996 shows big differences.Factors like weather and number of threshold exceedances seem to have anoticeable effect on the statistical parameters. In cases when the statisticalparameters should be used for a comparison between different forecast models,long periods of verification must be used in order to avoid influences fromirrelevant factors.

GLOREAM-report

109 of 216


The objectives for the coming year are development of a new solver for thechemical scheme, verification for long periods on several stations over Europe onboth raw and KALMAN-filtered forecasts in order to analyze model performancewith model-comparison in mind, and further development of public warningsystems.

6. Acknowledgement

Hereby a thank to AEA Technology and NERI for providing real-timeobservations of ozone concentrations on the WWW.

7. References

Jensen, M.H., A. Rasmussen, H. Svensmark and J.H. Sorensen, DanishAtmospheric Chemistry FOrecasting System (DACFOS), DMI Technicalreport 96-3, (1996).

GLOREAM-report

110 of 216

Development od a block-structured multi-scale transport code


Oswald KnothInstitute for Tropospheric Research

Permoserstr. 15, D-04318 Leipzig, Germany

No report received

GLOREAM-report

111 of 216

Mesoscale model estimations of sulphur, nitrogen deposition andcritical loads of soil acidity


Viatcheslav Kisselev and Vladimir KouznetsovKarbisheva str. 7, 194021 St. Petersburg, Russia

The Scientific Research Institute "Atmosphere" develops and applies to practicalproblems climatological stochastic model of regional pollutant transport. Thecharacteristic feature of this model is the use as meteorological input data of only asmall set of climatological characteristics of a considered locality thus avoidingthe necessity to deal with huge arrays of routine meteorological observationsduring long time periods. The "weather" in this model is generated by a speciallydeveloped stochastic procedure providing the coincidence of long-term statistics ofgenerated and real meteorological variables.

In 1999 the climatological stochastic model was improved and includes now theblock for the estimation of aerosol transport, i.e. transport of pollutant taking intoaccount possible gravitational sedimentation. The gamma-distribution of aerosolsize is assumed, concentration and mean radius being dependent on the aerosolhistory after emission. The presence of this block allows to treat e.g. the problemof soil pollution by heavy metals using as the input data the concentrations of ametal in the aerosol of different sizes.

The earlier developed version of this model for the treatment of acidifyingpollutant transport (sulfur and nitrogen compounds) were applied for theestimation of ecological situation in several regions of Russia. The correspondingloads from both internal and external sources were estimated for Kaliningrad,Belgorod and Novgorod regions, and Republic Karelia. On the basis ofcomparisons of these estimates with critical loads the recommendations on theoptimisation of system of measures for improvement of ecological situation in theconsidered locations were prepared.

GLOREAM-report

112 of 216

Development and Application of a Model Hierarchy for theDetermination of Chemical Processes in the Troposphere under

Present and Future Environmental Conditions


Bärbel Langmann,Susanne Bauer and Daniela Jacob

Max-Planck-Institut für MeteorologieBundesstr. 55, 20146 Hamburg, Germany

1. Summary

The on-line implementation of transport modules and chemical modules into themesoscale models REMO and GESIMA has been finished in a first version. Themodel hierarchy was applied for the simulation of a short episode in July 1994.The TRACT episode in September 1992 was another model evaluation period forthe off-line system REMO-CTM.

2. Aims of the research

The aim of the project is the development of the meteorological models REMOand GESIMA to Atmosphere-Chemistry-Models (on-line determination ofchemical and transport processes) and the construction of the model hierarchyREMO-GESIMA (coupling by one-way-nesting of the meteorological andchemical variables) to improve the understanding of formation processes ofphotooxidants. Additionally the regional model should be one-way-nested intoglobal model simulations to get the information of realistic initial and boundaryconcentrations for the limited area of interest. Evaluation studies are plannedbefore scenario simulations will be carried out.


a) REMO-CTM (off-line mode):A Chemistry-Transport-Model (CTM) (Langmann and Graf, 1997) was coupled tothe meteorological model REMO (Jacob and Podzun, 1997) and a one-way-nestingprocedure for trace species was implemented. Simulations, analysis and sensitivitystudies (concerning vertical resolution, ozone initial and boundary concentrations,vertical diffusion and dry deposition) of the TRACT episode in September 1992and a summersmog episode in July 1994 were carried out with the model systemREMO-CTM in 1/2° horizontal resolution over Europe and in 1/6° horizontalresolution over Germany.

GLOREAM-report

113 of 216

b) REMO (on-line mode):Transport, chemical transformation and deposition routines have beenimplemented in a first version. A simulation of the summersmog episode in July1994 in 1/2° horizontal resolution over Europe has been carried out forcomparisons with the off-line system REMO-CTM results. At the moment it isanalyzed.

c) GESIMAThe meteorological and chemical nesting of the non-hydrostatic model GESIMA(Kapitza and Eppel, 1992; Eppel et al., 1995) in REMO has been completed.Transport of trace species, the determination of gas-phase transformation rateswith the RADM II mechanism, dry deposition and the release of trace species fromarea and point sources has been implemented. A first simulation with the modelwas carried out in 4 km horizontal resolution for the area Berlin-Brandenburgduring characteristic periods of one day of the summersmog episode in July 1994.The results show that GESIMA is able to simulate ozone destruction during thenight, the beginning of ozone formation early in the morning and the transport ofozone to areas which are located on the lee side.


In Langmann et al. (1999) main results of the sensitivity studies carried out for theTRACT episode in September 1992 are described shortly. Additionally apreliminary simulation of several days in September 1992 with concentrationsfrom global model simulations for the initialization and at the lateral boundaries ofREMO-CTM was carried out and a significant influence on the regional ozonedistribution near the surface and even stronger in the free troposphere wasrecognized. Last but not least, results of a detailed evaluation (TRACT,16.09.1992, 13-17 local time) of REMO-CTM and EURAD, KAMM/DRAIS,MCCM and METRAS are summarized in Schaller and Wenzel (1998).

Figure 1 Near surface ozone concentrations as determined by REMO-CTM andREMO on-line in 0.5° horizontal resolution and as measured at UBAstations.

GLOREAM-report

114 of 216

Here we focus on the summersmog episode in July 1994. In large parts ofGermany sunny weather with near surface temperatures exceeding 30°C wasfavourable for regional ozone production. In Figure 1 measured and simulated (byREMO-on-line and REMO-CTM) near surface ozone concentrations averagedover 24 UBA stations are shown. During the first four days of the episodemeasured near surface ozone concentrations are reproduced with REMO-CTM andREMO on-line without significant differences between the two simulations. ThenREMO on-line is able to determine measured concentrations much more realistic.The differences between on-line and off-line simulation (up to 20 ppbv inmaximum ozone values) are mainly caused by subgrid scale convective processes.The convective cloud parameterization of the CTM (Langmann and Graf, 1997)results in much weaker trace species mixing than the parameterization followingTiedtke (1989) which is applied for the on-line simulation. Ozone decrease in theplanatary boundary layer and ozone increase in the free troposphere over Germany(Figure 2, on-line minus off-line results) is mainly an indirect effect. It is caused toa great extent by convective mixing of short living precursor substances such asNOx and VOC’s and the resulting changing oxidizing conditions throughout thetroposphere. The accumulated impact from previous cloud exchanges occuringupwind also contributes to the different ozone distributions.

Figure 2 a) Vertical ozone distribution as function of time determined by REMO on-line averaged over Germany. b) Vertical ozone differences: on-line minusoff-line results also as average over Germany. The black line indicates thePBL height.

GLOREAM-report

115 of 216

5. Main conclusions

First simulations with the model hierarchy REMO-GESIMA can realisticallyreproduce observations of meteorological and chemical variables. From thesesimulations we conclude that− modelled near surface concentrations are significantly dependent on horizontal

model resolution,− the parameterization of cloud convective processes as well as their on-line/off-

line determination plays an important role for the redistribution of tracespecies and a realistic reproduction of photooxidants concentrations,

− initial and boundary concentrations of longer living species such as ozoneshould be improved for simulations over limited areas (e.g. Europe) duringtransport dominated episodes.

6. Aims for the coming year

A detailed comparison of REMO on-line with REMO-CTM off-line results will becarried out. Another model run with REMO on-line for the period 01.06.1994-31.08.1994 is in preparation to check the model during longer time scales.Additionally a simulation of a complete period during the summersmog episode inJuly 1994 with GESIMA is in preparation. Comparisons and evaluation studieswith observation data of the FLUMOB episode are planned for REMO andGESIMA.

7. Acknowledgements

This work has been supported by the Ministry for Education, Science, Researchand Technology (BMBF) within the Tropospheric Research Program (TFS).Emission data have been provided by the EURAD group in Cologne and IER,Stuttgart.

8. References

Eppel, D. P., H. Kapitza, M. Claussen, D. Jacob, W. Koch, L. Levkov,H.-T. Mengelkamp and N. Werrmann, The Non-Hydrostatic MesoscaleModel GESIMA. Part II: Parameterizations and Applications, Contr. Atmos.Phys. 68, (1995), 15-41.

Jacob D. and R. Podzun, Sensitivity studies with the regional model REMO,Meteorol. Atmos. Phys. 63, (1997), 119-128.

Kapitza, H. and D. P. Eppel, The Non-Hydrostatic Mesoscale Model GESIMA.Part I: Dynamical Equations and Tests, Contr. Atmos. Phys. 65, (1992), 129-146.

Langmann, B. and H.-F. Graf, The Chemistry of the Polluted Atmosphere overEurope: Simulations and Sensitivity Studies with a Regional ChemistryTransport Model, Atmos. Environ. 31, (1997), 3239-3257.

GLOREAM-report

116 of 216

Langmann, B., S. Bauer and D. Jacob, Model hierarchy for tropospheric chemistrysimulations, Proceedings of EUROTRAC Symposium ’98, Editors: P.M.Borell and P. Borell, WIT Press, Southampton, (1999), in press.

Schaller, E. and A. Wenzel, TFS-Modellevaluierung, Fall 1: TRACT, 16.9.92,13-17 Uhr MESZ, internal article in German, (1998).

Tiedtke, M., A comprehensive mass flux scheme for cumulus parameterization inlarge scale models, Mon. Wea. Rev. 117, (1989), 1779-1800.

GLOREAM-report

117 of 216

Annual report of the contribution from the SwedishMeteorological and Hydrological Institute to GLOREAM


Joakim LangnerSwedish Meteorological and Hydrological Institute

SE-601 76 Norrköping, Sweden

1. Summary

Model simulations from the MATCH model for a six-month period during 1994have been evaluated using available observations, mainly from the EMEP network.The combination of meteorological data from HIRLAM, the MATCH model and amodified chemical scheme from the EMEP model gives good results consideringknown sources of error. Future work to improve and extend the capabilities of themodel system include: Improved treatment of dry deposition using more detailedinformation about landcover. Improved description of cloud and precipitationprocesses including aqueous phase chemistry and improved treatment of radiationand calculation of photolysis rates.


− Develop models for real time and seasonal simulation of tracer transport,chemistry and deposition over Europe,

− Analyse the sensitivity of such models to input data and model formulation,− Evaluate and validate models using available observed data.


The main activity during 1998 has been to evaluate model simulations from theMATCH model (Robertson et al., 1999) for a six-month period during 1994 usingavailable observations. The majority of the observations were taken from theEMEP network. As many as possible of the observed chemical components werecompared to the model. Some highlights are shown below. A report describing theresults has been published (Langner et al., 1998).

Work on updating the descriptions of dry deposition and wet scavenging, includingaqueous phase chemistry of sulfur, has started.

The work was presented at the second GLOREAM workshop in Madrid 16-18September 1998.

GLOREAM-report

118 of 216


Model simulations from the MATCH model for a six-month period during 1994was evaluated using available observations, mainly from the EMEP network.Meteorological input data to MATCH was taken from archived output from theoperational version of HIRLAM at SMHI.

The results for primary components show the largest deviations between modelcalculations and observations. Apart from difficulties with the representativity ofthe observations this can to a large extent be attributed to uncertainties in emissiondata and to the limited horizontal and vertical resolution of the model. Thecorrelation coefficient between daily observed and calculated values is above 0.5at more than a third of the stations for SO2. For NO2 the correlation between theobserved and calculated concentration is generally lower than for SO2.

Apr−2

Apr−30

May−28

Jun−

25

Jul−23

Aug−20

Sep−17

0

5

10

15

20

25

300.0

2.0

4.0

6.0

8.0

NO

2 pp

b(v)

0.0

2.0

4.0

6.0

8.0

0.0

1.0

2.0

3.0 ObservationModel

Apr−2

Apr−30

May−28

Jun−

25

Jul−23

Aug−20

Sep−17

FI4 AhtariSE5 Bredkalen

SE2 Rorvik NO1 Birkenes

IE1 Valentina Obs. DE2 Waldhof

NL10 Vreedepeel CS3 Kosetice

Figure 1 Observed and model calculated time series of diurnal average concentrationof NO2 at Bredkälen, Ähtäri, Rörvik, Birkenes, Valentina Observatory,Waldhof, Vreedepeel and Kostice in 1994. Units: ppb(v).

GLOREAM-report

119 of 216

A limited amount of observational data is available for primary hydrocarbons. Asurprisingly good correspondence between model calculations and observationswas found for several of the hydrocarbons of anthropogenic origin. The results arebest for the hydrocarbons with long residence times like ethane and n-butane. Forthese components the average calculated concentrations are within a factor of twofrom the corresponding observed hydrocarbon concentrations for most stations andthe correlation coefficient is above 0.8 at four stations for ethane. The agreementfor hydrocarbons with shorter residence times is worse but still rather goodconsidering the uncertainties related to emission data and formulation of thechemical scheme.

For isoprene, the only biogenic hydrocarbon included, the results are not so good,with low and sometimes negative correlation and large deviations betweenobserved and calculated average concentrations, exceeding a factor of 10 at severallocations. This indicates that the isoprene emissions used are probably quite farfrom being realistic.

The results for secondary components are generally better than for primarycomponents. Average concentrations of HNO3 + NO3

- are with few exceptionswithin a factor of two from the observations. The correlation coefficients areabove 0.5 for more than half of the stations. For sulfate the model has a tendencyfor underprediction but the correlation is generally higher than for HNO3 + NO3

-,with r-values above 0.8 for several stations.

The agreement between calculated and observed concentrations for formaldehydeis reasonably good, indicating that the model may be of some use for studying thedistribution of this compound.

Comparison of model calculations with over 80 stations with hourly measurementsof ozone shows that the model is capable of predicting average surfaceconcentrations of ozone within plus/minus 40%, with a correlation coefficientabove 0.5 for more than half of the stations. The general geographical variation ofsurface ozone is well described by the model. When looking at AOT40 andAOT60 statistics it is evident, however, that the model underestimates high ozoneconcentrations. This is probably due to a combination of several factors, includingmodel resolution and formulation and incomplete knowledge about emissions andsurface characteristics.

The calculated wet deposition of nitrate and sulfate show a tendency foroverprediction compared to the observed deposition. Accumulated deposition forthe six-month period is within a factor of 2 for most of the stations. Given thatmodel derived, and not observed, precipitation fields were used, this result isencouraging, considering also that the precipitation fields were not taken from theoptimal forecast lengths and that a fairly simple description of wet scavenging wasused.

GLOREAM-report

120 of 216

Ozone diurnal mean concentration April - Sept. 1994 (ppb(v))

SE13

ES5

FI22

NO43

NO30

NL9

SE35FI4

NO47

AT3

NO41

GB32GB34

NL10

CH31

NL2GB6

SK6 TR1

CS1

DE3SI32

AT4

CH5

CS3

ES1

SI31

AT2DE4

DE5

DK31

SI33

DE31

DE8SE32SK4

GB38

ES4IT4

NO51LT15PT4

DE26

ES3DK32DE17CH2

DE7

DE2

SE2

SE11

GB35NO44

FI9

DE35

GB41NO48

NO15

GB37

FI17NO39

GB36

GB2

IE31

ES2

SE12

DE9CH3

DE12GB15

DE1

NO45

GB31GB14

GB39

LV10

GB13

DE11

GB33

NO1 CH4

20

25

30

35

40

45

50

55

20 25 30 35 40 45 50 55

Observed

Mod

el

Figure 2 Scatterplot of observed and model calculated six-month (April-September1994) average surface concentrations of ozone. Units: ppb(v).

5. Main conclusions

In summary the combination of meteorological data from HIRLAM, the MATCHmodel and the modified chemical scheme from the EMEP model gives goodresults considering known sources of error. Future work to improve and extend thecapabilities of the model system include:

− Improved treatment of radiation and calculation of photolysis rates.− Improved description of dry deposition using more detailed information about

landcover.− Improved description of cloud and precipitation processes including aqueous

phase chemistry.− Use of observed precipitation and improved wet scavenging scheme.− Higher resolution and nesting.

GLOREAM-report

121 of 216

Current schemes for estimating isoprene emissions appear to be far from realisticjudging from comparisons between model calculations and the limited amount ofavailable isoprene concentration data.


− Compare model calculations from several models for 1996 and evaluatedifferences.

− Update parameterizations of dry deposition, wet scavenging, aqueous-phasechemistry and radiation in the MATCH model.

− Evaluate the sensitivity of the MATCH model calculations to changes inparameterizations and chemical boundary data.

7. Acknowledgement

This research has been supported by the Nordic Council of Ministers within theproject “Sensitivity analysis of regional-scale atmospheric models”.

8. References

Langner, J., R. Bergström and K. Pleijel, European scale modeling of sulfur,oxidized nitrogen and photochemical oxidants. Model development andevaluation for the 1994 growing season, Swedish Meteorological andHydrological Institute, RMK No. 82, (1998), 71 pp.

Robertson, L., J. Langner and M. Engardt, An Eulerian Limited Area AtmosphericTransport Model, J. Appl. Met. 38, (1999), 190-210.

GLOREAM-report

122 of 216

Study of photochemical oxidant budget variability in relation todynamics, chemistry and climate change


Kathy Law,Paul-Henri Plantevin, Maurette Cahill and Mathew EvansCentre for Atmospheric Science, Department of Chemistry

University of Cambridge, Lensfield Road, Cambridge, CB2 1EW, UK

The Cambridge global off-line chemistry and transport model, TOMCAT was runfor the period July 1994 to December 1995 at ~5.6 x 5.6 degree (T21) horizontalresolution and compared extensively to ozone measurements from the EU-fundedMOZAIC (Measurement of Ozone and Water Vapour by Airbus in-serviceaircraft) programme on a flight by flight basis. Comparisons with the data showthat the TOMCAT is capable of reproducing the seasonal cycle of O3 in the freetroposphere but O3 concentrations in the tropopause region are overestimated dueto the smearing out of sharp gradients.

These results have been published in Law et al. (1998). A stratospheric ozonetracer is also being used to diagnose the amount of ozone entering the troposphere.This will be used in the determination of the tropospheric ozone budget. Shorterruns, corresponding to periods when measurements were made as part of the EUTOASTE-C (Transport of Ozone and Stratosphere Troposphere Exchange) projecthave also been performed and comparisons made with ozone aircraft (MOZAIC)and lidar data. The results show that the model is capable of reproducingtropospheric fold events even at fairly coarse horizontal resolution. The model wasrun with 31 levels and forced using ECMWF meteorological analyses.

The TOMCAT model was also run for 3 months from 1st June 1997 to 31st

September 1997 again using analysed meteorological data from ECMWF and atT21 horizontal resolution and with 31 vertical levels. The model included CH4,C2H6 and C3H8 oxidation chemistry, emissions of CH4,CO, NOx, HCHO, C2H6 andC3H8, and deposition schemes for wet and dry removal of trace gases. Modelresults were compared with data collected on board the UK Met. Research FlightC-130 aircraft as part of the EU TACIA (Testing Atmospheric Chemistry inAnticyclones) and the UK NERC ACSOE (Atmospheric Chemistry Studies in theOceanic Environment) projects between 15th August and 23rd September 1997 inthe English Channel, North Sea and North Atlantic (flights based in the Azores).

Model results for O3, H2O2, CO, NO, NOy, C2H6, C3H8 and water vapour wereinterpolated onto the flight location and time every 20 seconds during each flight

GLOREAM-report

123 of 216

enabling direct comparison along the flight tracks. Many general features seen inthe data are reproduced by TOMCAT which is very encouraging given that themodel was run at fairly coarse horizontal resolution and with a rather limitedhydrocarbon scheme. More detailed examination of discrepancies, particularlyrelated to the modelled levels of source gases (e.g. CO, VOCs) will be carried outto investigate whether the emissions used in the model over Europe are too high.

The TOMCAT model run was also run with European emissions (NOx, CO, ethaneand propane etc.) switched off in order to assess the impact these emissions on thephotochemical production of ozone and other oxidants. The data was collectedduring an anticyclonic period over Europe. The model simulations suggest that upto 50 ppbv of additional ozone was produced due to European emissions on anygiven day throughout this period in August 1997. A detailed budget code todiagnose the contribution of chemical and dynamical processes governing theozone budget is being developed for the TOMCAT model.

High levels of ozone observed during a flight on the 19th August 1997 were alsostudied using the Cambridge photochemical trajectory model, CiTTy-CAT as partof the TACIA project. This study concentrated on the role of surface processes oncontrolling this level of ozone. Likely uncertainties in the emissions ofanthropogenic hydrocarbons (between 0.5 and 2 times the emissions estimates)leads to a variation in the ozone of 20 ppbv. The uncertainties in the emissions ofbiogenic hydrocarbons (between 0.1 and 10 times the emissions estimates) leads toa variation of 50 ppbv. Variations in deposition process are likely to contribute anuncertainty of roughly 10 ppbv. If improvements are to be made in the accuracy ofthese models better estimates of emissions are required and in particular, moreaccurate emission estimates for biogenic hydrocarbons which can have a largeimpact on the amount of O3 produced over Europe.

Funding Acknowledgements

Funding was provided by the UK Natural Environmental Research Council(NERC), the UK Universities Global Atmospheric Modelling Programme(UGAMP) and by the EU-funded projects MOZAIC and TACIA.

Relevant Publications

Law, K.S., P.-H. Plantevin, D.E. Shallcross, H. Rogers, C. Grouhel, V. Thouret,A. Marenco and J.A. Pyle, Evaluation of modelled O3 using MOZAIC data,J. Geophys. Res., 103, (1998), 25721-25740.

GLOREAM-report

124 of 216

Application of synoptic/meso scale Eta Modelin long range transport processes


Lazar LazicInstitute of Meteorology, University of Belgrade

P.O.Box 368, 11001 Belgrade, Yugoslavia

1. Summary

The work of the Institute of Meteorology, University of Belgrade (IMUB) group in1998 consisted of efforts in three main directions. One was addressed to the modeltrajectory calculations. It had included work on the implementation and testingtrajectory calculations based on the Eta Model in case of complex terrain.

The second main direction was development of a nesting technique, based on theEta Model, suitable for application to problems of the long-range transport of airpollutants.

The third main direction was application of the resulting model on the case ofSaharan dust transport to Europe.

2. Aims of the research

Aims of the research were, in the first place, development of the meteorologicalpart of the transport model in terms of the implementation of trajectory calculationand nesting technique based on the Eta Model. Studies of the requirements needed,in terms of sophistication of the representation of various atmospheric processes(mountain representation, boundary layer, land-surface processes, moist processes)needed for high-quality results in impact study situations (transport simulationswith time-scale of 2-3 days).

The second principal aim was the application and verification of the resulting codeon the various sensitivity studies as well as to studies of chosen European long-range transport of air pollutants problems.


Regarding the IMUB meteorological module (Eta Model) (Mesinger et al., 1988),work has initiated on analyses of the accuracy of its precipitation forecasts(Mesinger, 1998 a, b, c) and on further refinement of model’s physical package. A

GLOREAM-report

125 of 216

number of specific features of this package were considered as candidates forimportant contributions to realistic simulation of constituent transport. For asummary of the overall design of this package the reader is referred to the paper byJanjic (1990, 1994).

Considerable attention was given to add to this model trajectory calculations(Lazic and Tosic, 1998) and nesting technique (Tosic and Lazic, 1998), and toapply the model for various sensitivity studies as well as to studies of chosenEuropean long-range transport of air pollutants problems (Vukmirovic et al., 1998;Vukmirovic et al., 1999). Space scale IMUB primarily has in mind is severalthousand kilometers, with the mesh-size of the order of 30 km. Dynamical topicswas studied include effects of mountains (Lazic and Tosic, 1998; Tosic and Lazic,1998); vertical exchange between the boundary layer and the troposphere abovethe boundary layer; and effect of moist processes in long-range transport of airpollutants from the point of view of the resolution and dynamical treatmentrequired for realistic results.


Trajectory calculations are implemented in the Eta Model and applied to flow overcomplex terrain (Lazic and Tosic, 1998). The Eta Model has a vertical coordinatewhich permits a step-like representation of mountains and quasi-horizontalcoordinate surfaces, the so-called eta coordinate. A realistic real data simulation ofa local bora wind case over Dynaric Alps is achieved using the model with a 28 kmhorizontal resolution and 16 layers in the vertical. Numerical experiments withdifferent mountain heights and shapes in the bora wind region are performed.Three-dimensional trajectories over various mountains mimicking real mountainsbut differing primarily in elevation are calculated and analysed. The maximumbora wind speed is predicted as expected through three-dimensional channels inthe step mountain representations.

Improvements are obtained in simulations of the local bora wind using a nestingtechnique (Tosic and Lazic, 1998). A version of the Eta Model is used for both thelower resolution as well as for the higher resolution nested run. For nesting, atechnique has been developed whereby for the boundary conditions “underground”model values are used. They are obtained by horizontal interpolation of modelvariables inside the model’s eta layers. Simulations are initialized using real data.This is done in order to assess the model’s performance in simulating the localbora wind, an important mesoscale phenomenon caused by orography.Improvements achieved by nesting are illustrated by examples of precipitation aswell as a trajectories forecast.

The Balkan Peninsula is situated in the impact zone of Saharan dust storms. Thecase of Saharan dust transport to Belgrade in the period of 14-17 April 1994 isanalyzed using the Eta Model (Vukmirovic et al., 1998). To improve visualization

GLOREAM-report

126 of 216

of the vertical structure for the simulated wind, air back trajectories are calculatedfrom the lowest six model levels from 347 m up to 1423 m with horizontal gridresolution of 1° x 1° and high grid resolution of 10' x 10'. Calculated trajectoriesshow that convective storms along particle pathways provided the required upliftin the complex orographic conditions. Despite being less reliable in the simulationof lower level trajectories, the step-mountain eta coordinate model is very suitablefor illustrating air parcel crossing of the emission field from anthropogenicpollutant sources. According to the analysis of the three lowest trajectories,transport of trace metals from Macedonia and south Serbia might be dominant inthe observed episode. Turbulent flow enhanced the coagulation process of initiallyclean dust particles with particles containing Pb and Cd. The coagulation andscavenging processes below and in clouds increased deposition rates of Pb and Cdin Belgrade in the course of wet removal, and consequently trough resuspensionprocesses. Dry deposition samples contained characteristic particles up to 30µm indiameter with Fe content and significant ratio Si/Fe of 3 to 5, determined forselected single particles by the SEM/EDX method. Following dry and wetdeposition of Cd and Pb, a residual effect of dustfall is noticed throughout thevegetational period.

5. Main conclusions

− The performance of the meteorological synoptic/mesoscale Eta Model used bythe IMUB group, in particular with regard to the realism of precipitationforecasts, is now at a high state-of-the-art level.

− A considerable attention was given to add to this model trajectory calculationsand nesting technique.

− The preliminary version of the system including constituent transport alongtrajectories, diffusion and wet and dry deposition code, has been shown toperform satisfactorily. The application of this system on the Saharan dusttransport to Europe gave reasonable results.

− A sensitivity study of the treatment of the flow over complex terrain hasshown it to be very important for correct simulations of the transport anddiffusion of tracers, and possibly also for long range transport of constituents.


− Sensitivity tests will be performed on the Saharan dust transport case in orderto assess the effects of various modeling options on the transport/depositionresults. Examples of options to be considered are the choice of horizontalresolution; representation of mountains (terrain-following vs. the step-mountain system); choice of the constituent advection scheme; effects oftransports by convection; and parameterization of horizontal and verticaldiffusion.

GLOREAM-report

127 of 216

− The effects of the initial subgrid scale diffusion on the overall quality of thelong range constituent transport simulations will be studied. This sensitivitystudy will be based on results for the Saharan dust transport case.

− Work on other cases will be initiated.− Further development of trajectory model, nested model and continued

improvement of the Eta Model and model verification.

7. Acknowledgements

This contribution is not being financed by any scientific foundation. However, it isbased on and/or is benefiting from efforts supported by a number of institutions orfunding agencies. These are, in chronological order, the FederalHydrometeorological Institute, Belgrade, Yugoslavia; Science Association ofSerbia, Belgrade, Yugoslavia; NCEP, Washington, DC, U.S.A. and SerbianAcademy of Sciences and Arts, Belgrade, Yugoslavia. Work on the contribution isperformed in collaboration with the Institute of Physics/Energy Conversion Dept.,Zemun, Yugoslavia.

8. References

Janjic, Z. I., Physical package for the step-mountain, eta coordinate model, Mon.Wea. Rev. 118, (1990), 1429-1443.

Janjic, Z. I., The step-mountain eta coordinate model: Further developments of theconvection, viscous sublayer and turbulence closure schemes, Mon. Wea.Rev. 122, (1994), 927-945.

Lazic, L., and I. Tosic, A real data simulation of the Adriatic bora and the impactof mountain height on bora trajectories, Meteorol. Atmos. Phys. 66, No. 3-4,(1998), 143-155.

Mesinger, F., Z.I. Janjic, S. Nickovic, D. Gavrilov and D.G. Deaven, The step-mountain coordinate: Model description and performance for cases ofAlpine lee cyclogenesis and for a case of an Appalachian redevelopment,Mon Wea Rev. 116, (1988), 1493-1518.

Mesinger, F., Quantitative precipitation forecasts of the “early” Eta Model: Anupdate. Preprints, Sixteenth Conference on Weather Analysis andForecasting, Symposium on the Research Foci of the U.S. WeatherResearch Program, Phoenix, AZ, 11-16 January 1998; AmericanMeteorological Society, Boston, MA, (1998), 184-186.

Mesinger, F., Performance of the 48-km Eta in forecasting tracks of majorlandfalling Atlantic hurricanes of the 1966 season, Bertha and Fran.Preprints, Sixteenth Conference on Weather Analysis and Forecasting,Symposium on the Research Foci of the U.S. Weather Research Program,Phoenix, AZ, 11-16 January 1998; American Meteorological Society,Boston, MA, (1998), 526-528.

Mesinger, F., Comparison of quantitative precipitation forecasts by the 48- and bythe 29-km Eta Model: An update and possible implications. Preprints,

GLOREAM-report

128 of 216

Twelfth Conference on Numerical Weather Prediction, Phoenix, AZ,11-16 January 1998; American Meteorological Society, Boston, MA,(1998), J22-J23.

Tosic, I. and L. Lazic, Improved bora wind simulation using a nested Eta Model.Meteorol. Atmos. Phys. 66, No. 1-2, (1998), 1-10.

Vukmirovic, Z., M. Unkasevic, L. Lazic, I, Tosic, S. Rajsic and M. Tasic, Analysisof an exceptional Saharan dust event in Serbia using the Eta Model.Submitted to Atmos. Environ., (1998).

Vukmirovic, Z., M. Unkasevic, L. Lazic, I. Tosic and M. Tasic, Analysis ofSaharan dust transport using the Eta Model. Proc. of EUROTRACSymposium ’98. Eds: P.M. Borrell and P. Borrell. WITPRESS,Southampton, (1999), 9-37 (1-4).

A brief paragraph explaining results from the EUROTRAC-2 work:

(Application of synoptic/meso scale Eta Model in long range transportprocesses

Principal investigator: Lazar Lazic, Institute of Meteorology, University ofBelgrade)

Initial attention of this contribution was given to add to the Eta Model trajectorycalculations (Lazic and Tosic, 1998) and nesting technique (Tosic and Lazic,1998), and to apply the model for various sensitivity studies as well as to studies ofchosen European long-range transport of air pollutants problems (Vukmirovic etal., 1998; Vukmirovic et al., 1999). Space scale which primarily has in mind isseveral thousand kilometers, with the mesh-size of the order of 30 km. Dynamicaltopics studied include effects of mountains (Lazic and Tosic, 1998; Tosic andLazic, 1998); vertical exchange between the boundary layer and the troposphereabove the boundary layer; and effect of moist processes in long-range transport ofair pollutants from the point of view of the resolution and dynamical treatmentrequired for realistic results.

GLOREAM-report

129 of 216

Speciation in Chemical Mechanisms and Emissions Databases


P.A. MakarEnvironment Canada

4905 Dufferin Street, Downsview, Ontario, Canada, M3H 5T4

1. Summary

The first part of the subproject (Part A: Tropospheric Mechanism Numerics) hasbeen completed, with the development and testing of a method to compress thedifferential equations of gas-phase reaction mechanisms. The processing timerequirements for gas-phase chemistry simulations have been reduced by up to afactor of 3.7, by combining Bernoulli’s integrating factors and the assumption oflinear time dependence within each substep of a multistep integration. No formalloss in accuracy relative to the original system was observed, with average errorsfor a suite of 2511 test cases being the same magnitude as the iteration errorallowed within the numerical solver. The method thus allows a significantreduction in the processing time required by reaction-transport models. Themethod also provides a means by which highly detailed, speciated reactionmechanisms could be used in large scale reaction transport models. The secondpart of the project (Part B: Lumping of Emissions for Pollution Models) willcommence in mid-1999, using the mathematics developed in Part A.


The study has two components, both dealing with ways of improving the accuracyor computation time of portions of regional and global pollution models. The firstcomponent (Part A, completed) is an examination of the system of differentialequations used for tropospheric chemistry modelling, to determine methods for thereduction of the number of variables used in numerical integration. Simulationsand forecasts of atmospheric pollution are hampered by the large amount ofcomputational time required to integrate the system of equations describing gas-phase chemical reactions. One means of reducing the computational burden is toreduce the number of variables being integrated. Previous published work by thePI showed that the use of integrating factors can reduce the number of variableswhich must be retained in a numerical integration, greatly decreasing thecomputational requirements of a chemistry simulation. In Part A, this work wasextended to provide a more general framework for mechanism compression, andwas tested on a tropospheric reaction mechanism of realistic complexity. Thesecond component of the study (Part B) will examine a related issue: the methods

GLOREAM-report

130 of 216

used to create simplified emissions databases from highly detailed, speciatedemissions.


A general mathematical framework for the compression of non-self-reactingchemical species was developed in the spring of 1998. By the summer of 1998,this had been used to develop accurate methods to temporarily compress a reactionmechanism for numerical integration. The details of these techniques are describedbriefly below. The results were presented at the INRIA/ENPC InternationalConference on Air Pollution Modelling and Simulation, in Paris in October 26-29of 1998 (Makar, 1998). The paper appearing in the conference proceedings iscurrently under review for journal publication.

4. Principal Results

The mathematical development for the methods can be found in Makar (1998).Here, only the final formulae for the different species will be presented. The stiffsystem of differential equations formed by atmospheric chemical reactions is givenby:

∂∂ c (t)

t = P [c(t)] - L [c(t)] c (t)i

i i i

* *(1)

The terms Pi and Li describe the chemical production rates (concentration/unittime) and loss frequencies (time-1) of the i’th chemical species of the set ofchemical reaction employed in the model. The use of Bernoulli’s integrating factorand the assumption of linearity across an integration substep was used to createanalytic formulae for some of the species in a chemical reaction mechanism. Theseformulae could be used in the integration code in place of the original species,with no formal loss of solution accuracy and great reductions in processing time.Five compression formulae were identified, depending on the type of reactions inwhich the variable to be compressed is involved. These are:

Non-Zero Production from One Variable, Constant Loss Frequency:

Here, Pi is non-zero and Li is a constant. Dropping the subscript i for simplicity,and introducing the superscripts 0 and δt for values at the start and end of anintegration substep, the concentration of the variable to be compressed as afunction of time becomes:

c = P - P exp(-L t)

L +

P - P (1 - exp(-L t) )

t L + c exp(-L t) t

t 0 0 t

20δ

δ δδ δδ

δ . (2)

GLOREAM-report

131 of 216

An example for which (2) is applicable is the singlet-D excited state oxygen atom,O(1D), in which the production term is the photolysis of ozone and the loss term isthe sum of the loss frequencies of O(1D) by water and the total number density.O(1D) is a very fast reacting radical (usually having the largest loss frequency ofany chemical species during the daytime).

Zero Chemical Production (Loss Only) Variables:

These include most unoxygenated hydrocarbons, and variables such as SO2, whichare not produced by chemical reactions. For these cases the equation forconcentration as a function of time can be written as:

c (t) = c (0)

exp L (t ) dti

i

i 0 00

t

∫

, (3)

An example is given by the case of a hydrocarbon, HYD, which reacts with OH,O3, and NO3, with reaction rate constants kOH, kO3, and kNO3, respectively. Thelatter are constant within the integration step, and (3) becomes:

[HYD(t)] = [HYD(0)] exp - k [OH(t )] dt - k [O (t )] dt - k [NO (t ) dtOH o o O3 3 o o NO3 3 o o

0

t

0

t

0

t

∫∫∫

(4)

Here, the integrated quantities are approximated by the trapezoid integration of thespecies within the integrand for each substep, summed over all substeps to thecurrent time. These variables must be stored during the integration, but are notexplicitly part of the integration themselves (i.e. the rate of change of the integralof OH is not calculated). This particular solution is described in earlier work(Makar and Polavarapu, 1997) - here it is noted that the integrated OH, O3 andNO3 quantities need not be considered as additional variables in the system whichmust be solved.

Production from Two Variables, Constant Loss Frequency:

This example results from two chemicals reacting to produce the one which is tobe compressed, the latter undergoing a first order loss:

A + B D, k

D E, k

D

t = k [A][B] - k [D]

1

21 2

→→

∂∂

(5)

An example is the formation and destruction of peroxyacetylnitrate. Here, theconcept of substep linearity is applied to the reactant species A and B. Definingthe slopes of these variables across the interval (0, δt) as mA and mB, we have

GLOREAM-report

132 of 216

A(t) = A(0) + t (mA/δt); B(t) = B(0) + t (mB/δt) (6)

Substitution and integration yields the following formula for the value of D at theend of a model substep:

D(δt) = k1 { mA mB/k2 + 2 mA mB /(k23 δt2) + mA B(0) /k2 - 2 mA mB /(k2

2 δt)

- mA B(0) /(k22 δt) + A(0) mB /k2 - A(0) mB /(k2

2 δt) + A(0) B(0) /k2

+ [ mA B(0)/(k22 δt) - 2 mA mB /(k2

3 δt2) - A(0) B(0) /k2

+ A(0) mB /(k22 δt) + D(0) /k1] exp(- k2 δt) (7)

Production from Two Variables, Time varying Loss Frequencies:

The general case arises from a single species which is coupled to the rest of thesystem by two or more reactions of the form:

A + B D, k

D + E F, k

D

t = k [A][B] - k [E][D]

1

21 2

→→

∂∂

(8)

There are many species which have this form of interaction with the rest of thesystem, examples being H2O2, and oxygenated volatile organic compounds such ascarbonyls and alcohols.Using the subscript j to represent the j’th integration substep, and denoting SPj,i as:

SP = t

2 P (0)exp L (t ) dt P ( t )exp L (t ) dt j,i

j

i i n n i j i n

+ t

n

jδδ

0 0

t t

∫ ∫

+

, (9)

The concentration of the compressed species as a function of time can beapproximated by:

c (t) =

c (0) + SP

exp L (t ) dti

i j,ij

i 0 0

∑

∫

0

t (10)

Large Production Rates and Loss Frequencies:

Species with large production and large non-constant loss rates may presentdifficulties, due to the need to evaluate the exponential of the integrated loss term.For the fastest reacting species (e.g. O(3P)), if the loss rate during a substep is

GLOREAM-report

133 of 216

approximated by the average of its values at the start and end of the substep, then(2) may be used to a high degree of accuracy.

Numerical Tests

The formulae were tested using a standard Gear-type solver and a detailed gas-phase mechanism, for 2511 test case half-hour integrations. The tests were done instages, each stage adding additional compressed variables. The CPU time resultsare shown in Fig. 1. In the figure, “All” refers to the original system of equations,“O(1D)” refers to the singlet oxygen radical being compressed using (2), “O(3P)”refers to the additional compression of triplet-P oxygen using (2) and an averagedloss frequency, “14 Species” refers to the additional compression of 14 speciesusing (4), “7 Species” refers to the compression of 7 additional species using (10),and “3 Species” refers to the compression of 3 additional species using (7).

Figure 1 Processing time (including breakdown by process) for different levels ofcompression, SGI R10000, 195 MHz cpu, average/per integration, 2511 testcases.

Processing time could thus be reduced by a factor of 3.7x over the original systemof equations. The accuracy of the solution for each test case was determined bycalculating the percent differences from the uncompressed solutions for each ofthe 75 time-varying species in the reaction mechanism. The average errors wereinsignificant, with magnitudes similar to the iteration epsilon of the numericalintegration.

5. Main conclusions

The numerical experiments conducted here indicate that large, complex reactionmechanisms may be accurately compressed for numerical integration, with littleloss in accuracy and a substantial reduction in processing requirements.

GLOREAM-report

134 of 216

6. Policy relevant aspect of “Speciation in Chemical Mechanism andEmissions Databases”

The work conducted to date suggests that more complex reaction mechanisms maybe used in regional and global models than have been used in the past, and/or thatthe computer processing time required for current mechanisms may be greatlyreduced. The use of the new techniques may therefore allow for more detailedmodel simulations, aimed at very specific emitted hydrocarbon species, rather thanthe generic “lumped” variables currently in use. Species specific emissionsscenario runs may therefore be possible or improved with the use of thesemethods. Alternately, a larger number of model runs may be performed in the sameamount of processing time. The work improves the level of detail possible in amodel simulation, thus improving the realism of the simulation and hence theaccuracy of policies derived from those simulations.


Certain aspects of the mathematics for chemical compression can be applied to thecompression of emissions as is typically used in regional and global models. Themathematics of emissions compression will be examined in the upcoming year.

8. Acknowledgements

This research was supported by the Air Quality Research Branch of theAtmospheric Environment Service, Environment Canada.

9. References

Makar, P.A., Integrating Factors and the Compression of Gas-Phase ReactionMechanisms, APMS98 Proceedings Vol. 2, Institut National de Rechercheen Informatique et en Automatique, (1998), 435-444.

Makar, P.A. and S.M. Polavarapu, Analytic Solutions for Gas-Phase ChemicalMechanism Compression, Atmos. Environ. 31, (1997), 1025-1039.

GLOREAM-report

135 of 216

Integrated assessment of acidification, eutrophicationand ozone formation in Flanders


Clemens Mensink,Peter Viaene and Jan Duerinck

VITO, Boeretang 200, 2400 Mol, Belgium

1. Summary

In order to reduce acidification, eutrophication and ozone formation in Flanders,Belgium, policy goals are formulated and possible abatement strategies arediscussed and optimised with respect to cost effectiveness. In a European contextthis methodology is followed for the realisation of the EU abatement strategy onacidification and tropospheric ozone (Amman et al., 1998). An integratedassessment of acidification, eutrophication and ozone formation was carried outthis year. The study was based on the results of various atmospheric transportmodels (OPS, LOTOS, EMEP) and included an evaluation of the costeffectiveness of the emission control measures and an evaluation of the effects. Foracidification and eutrophication the effects were evaluated in terms of critical loadexceedances. For ozone formation, the impacts have been evaluated using theAOT60 as an indicator for the protection of human health and the AOT40 as anindicator for damage on vegetation.

Integrated environmental assessment techniques are applied to describe multipleeffects of multiple pollutants. In particular, the study focused on acidification,eutrophication and ozone formation (multiple effects) caused by SO2, NOx, NH3

and VOC (multiple pollutants). The assessment included a cost-benefit analysis toestimate both costs and benefits of the proposed emission reduction measures.Regional scale air quality models are essential tools in obtaining the relationbetween the pollutants and their effects. Another useful tool is the concept ofcritical loads (for acidification and eutrophication) and critical levels (for ozoneformation). Both indicators estimate the impact on ecosystems and human healthin terms of exceedances. Economical techniques, like the derivation andapplication of cost-curves and external cost evaluations of potential benefits, arerequired for an explicit monetarisation of both costs and benefits. Air qualitymodels and economical cost functions are very sensitive to their input data and aretherefore in some cases facing high uncertainties.

At the moment two integrated assessment studies are carried out at a Europeanlevel. First of all there is the work of the Convention on Long Range

GLOREAM-report

136 of 216

Transboundary Air Pollution focusing on the preparation of a multi-pollutantprotocol that will address NOx and VOC emissions. Secondly, the possibilities forcost-effective emission reductions in Europe are explored to support the EUabatement strategies on acidification and ozone. The preparatory studies for thesetasks are based on calculations using the RAINS and EMEP models.

The outcome of the policy related integrated assessment studies at a Europeanlevel has important implications for policies on a national or regional level as well.The computations with the RAINS and EMEP models are carried out on a gridwith a spatial resolution of 150 km x 150 km, providing emission reductionproposals on a country to country base. Therefore it is especially important forsmall countries and regions to verify in more detail the model results as well as theregional impact stemming from the proposed emission reductions. This was alsorecognised by the Flemish government and resulted in a project addressing theseissues. The project focussed on the following objectives:

− obtain insight in input data used in the RAINS and EMEP models and toupdate or correct them when needed,

− find out what are the implications for the Flanders region of the proposedemission reduction strategies, and

− adapt or tune the environmental policy in Flanders towards the emissionreduction objectives using a cost-benefit analysis.

The study focused on three scenarios reflecting three different abatementstrategies. A first scenario can be characterised as a reference scenario, based onthe current national reduction plans in Europe and the current national andEuropean legislative actions. In a second scenario the abatement policy asproposed in Flemish environmental policy strategy (MINA-plan 2, 1997) isembedded in the reference scenario. A third scenario aims at a reduction of thearea of unprotected ecosystems in Europe by at least 50% in 2010 compared to1990. This scenario is known as the 50% gap closure scenario.

The impact of the three emission reduction scenarios on acidification andeutrophication was evaluated using the Operational Priority Substances (OPS)model (van Jaarsveld, 1990). Results for yearly averaged SOx-, NOy-, NHx- andtotal N-depositions have been obtained on a grid with a resolution of 5 x 5 km2

covering Flanders. The impact of the emission reduction scenarios on ozoneformation was evaluated on the basis of the model calculations carried out by theLOTOS model for Belgium for the summer period of 1994 (Builtjes et al.,1996).On the benefit side, exceedances of critical loads for nitrogen and sulphurdeposition were determined for 652 forest ecosystems in Flanders. On the costside, the cost-curve for each of the pollutants was used to estimate the total costper scenario. Expected uncertainties in the results were addressed and discussed.

GLOREAM-report

137 of 216

In a parallel study, a compact chemistry mechanism developed for use in theEUROS model was evaluated. EUROS is an Eulerian grid model that originatesfrom the Dutch National Institute of Public Health (RIVM). The model is intendedfor long term simulations and should be applicable on a midrange workstation. Thechemical mechanism included in EUROS is therefore limited to 17 reactions and16 species. The 16 species considered in the mechanism are: SO2, sulphate aerosol(SO4), NO, NO2, nitrate aerosol (NO3a), O3, OH, CH4, C2H4, C2H6, C3H6, C4H10,

xylene (XYL), isoprene (ISO), CO and HNO3. The concentration of methane(CH4) is assumed to be constant and in the current calculations it is fixed at1700 ppb. To evaluate the mechanism it was compared to the larger EMEP/MSC-W scheme (Barrett and Berge, 1996). For the comparison, four of the box testsused in the Chemical Mechanism Working Group (CMWG) modelintercomparison (Poppe et al., 1996) were repeated: LAND, BIO, PLUME/1 andPLUME/2.


One of the most actual topics in environmental policy in Belgium deals withphotochemical smog episodes during summertime. During the summer of 1995, 32days were reported with hourly concentrations exceeding 180 µg/m3. In order toprovide support for environmental policy decisions, regional scale photochemicaltransport models as well as statistical prediction models are being implemented bydifferent research groups in Belgium. The aim is to provide a hierarchy of modelsto represent the relevant processes governing photochemical smog formation aswell as to represent the time scales and length scales that are relevant forphotochemical smog formation and its transport over Belgium and WesternEurope. In a joint effort, the interaction between the different scales and the needfor high resolution modelling will be addressed in particular.

As a principle investigator, VITO will further contribute to GLOREAM by furtherdevelopment, validation and application of the regional scale model EUROS.Proposed research activities for this model are:

− further development of adaptive or local grid refinement techniques (to beapplied near strong solution gradients, near strong sources, in specific userdefined areas); sensitivity studies on grid refinement;

− further development and extension of the existing chemistry modellingroutines;

− model validation (episodic modelling) by means of comparisons of modelresults with measurements obtained from a measurement network as well asfrom aircraft measurement campaigns;

− model intercomparisons from which model sensitivities and further modelimprovements can be evaluated, and

− evaluation of abatement strategies from emission scenario computations.

GLOREAM-report

138 of 216


In co-operation with RIVM in the Netherlands the EUROS chemistry module wastested and compared with other chemistry modules. The aim is to develop a verycompact chemistry module that allows fast computations of scenarios for policysupport.

On the 28th August 1998, an aircraft measurement campaign was carried out in theframework of STAAARTE (Scientific Training and Access to Aircraft forAtmospheric Research Throughout Europe). The campaign was titled“Transboundary Flux Measurements for Photochemical Model Validation inFlanders”. With the results of this project VITO wants to extend its contributiontowards the evaluation and support of policy measures to control episodes of highozone concentrations. The project focuses on validation of computer models thatare able to simulate smog episodes.

A study on the integrated assessment of acidification, eutrophication and ozoneformation in Flanders was carried out for the Flemish environmentaladministration (AMINAL). This project required a co-operative contribution fromboth atmospheric and economic scientists.


Results of the integrated assessment study showed that the dominant role ofammonia in acidification and eutrophication in Flanders will increase in the future,especially in the 50% gap closure scenario. This is due to the fact that this scenariofocuses on more cost effective SO2-emission reductions rather than on NH3-emission reductions. The gap closure scenario is not sufficient to achieve the longterm policy objectives in Flanders (1400 mol H+/ha/yr and 5-20 kg N/ha/yr in2010). It seems that a further reduction of NH3 emissions will be inevitable to meetthe long-term objectives. However, it was found that they are relatively expensiveand that there is still a potential for less expensive NOx reductions. It was alsofound that at least a 50% reduction in NOx and VOC emissions is needed to reducethe AOT60 with approximately 50% in Flanders. The effect on the AOT40 is lesspronounced. The gap closure scenario yields emission reductions slightly over50% for both NOx and VOC. The model results seem to confirm that this issufficient to reduce the AOT60 with 50% in Flanders. The costs for the reductionsneeded for the ozone reduction strategy are relatively low when compared to thecosts needed to realise the proposed acidification strategy.

Evaluation of the compact EUROS chemistry mechanism indicates that resultswith the compact scheme deviate significantly (confidence level 0.95) from resultsthat can be obtained with the EMEP/MSC-W scheme. The preliminary results ofthe optimisation that is being used in an attempt to improve the mechanism

GLOREAM-report

139 of 216

indicate that the highly condensed chemistry is able to approximate thecharacteristics of the more extended EMEP chemistry.

A database with raw flight data has been composed containing the airbornemeasurements of the STAAARTE-TFLUX campaign. The database has to beprocessed further and will be extended with ground based measurements andballoon soundings.

5. Main conclusions

In 2010 ammonia is expected to be responsible for almost 50% of the acidificationin Flanders and at least 70% of the nitrogen deposition in Flanders. The 50% gapclosure scenario has to be adopted by Flanders in order to achieve the intermediatepolicy goals in 2002. On a first sight further reductions of ammonia inside (andoutside) Flanders seem to be needed to achieve the long-term policy goals in 2010.However, they are relatively expensive and there seems to be still a potential forless expensive NOx reductions.

The expected impact of the 50% gap closure scenario could be confirmed, but theuncertainties in the results of the assessment are rather high. The costs for thereductions needed for the ozone strategy are relatively low when compared to thecosts needed to realise the proposed acidification strategy.

Results obtained with the compact EUROS chemistry scheme deviate significantlyfrom results that can be obtained with the EMEP/MSC-W scheme. A furtherparameterisation of the condensed chemistry is currently being undertaken byRIVM.


Development and implementation of the EUROS model for policy support withrespect to tropospheric ozone in Belgium, with the following specific tasks:

− biogenic and anthropogenic emission modelling,− implementation of meteorological data (ECMWF and ALADIN), and− optimisation of the chemistry and advection modules in EUROS.

Preparation of STAAARTE-TFLUX data set, which contains:

− the airborne measurements,− the ground based measurements (automatic measurement network), and− ozone soundings at the RMI Uccle.

GLOREAM-report

140 of 216

7. Acknowledgements

The authors would like to thank the Flemish environmental administration(AMINAL) and the Flemish Environmental Agency (VMM) for their financialsupport in carrying out the projects mentioned above. Furthermore we would liketo thank the MRF (Met Research Flight) at Farnborough (UK) and RIVM atBilthoven (NL) for their fruitful co-operation.

8. References

Amman, M., I. Bertok, J. Cofala, F. Gyarfas, C. Heyes, Z. Klimont, and W.Schopp, Cost-effective Control of Acidification and Ground-Level Ozone,Third Interim Report to the European Commission, DG-XI, IIASA,Laxenburg, (1997).

Barrett, K. and E. Berge (editors), Transboundary Air Pollution in Europe, Part 1:Estimated dispersion of acidifying agents and of near surface ozone, EMEPMSC-W Status Report, Oslo, (1996).

Builtjes, P.J.H. and G. Boersen, Model calculations to determine the influence ofEuropean emission reductions on ozone concentrations over Belgium, TNO-report TNO-MEP-R 96/274, Apeldoorn, (1996).

Jaarsveld, J.A. van, An Operational atmosferic transport model for PrioritySubstances; specification and instructions for use, RIVM report nr.222501002, Bilthoven, (1990).

MINA-plan 2, Het Vlaamse milieubeleidsplan 1997-2001 (Environmental PolicyStrategy 1997-2001), Ministry of the Flemish Community, Dept. ofEnvironment and Infrastructure, Brussels (in Dutch), (1997).

Poppe, D., Y. Andersson-Sköld, A. Baart, P.J.H. Builtjes, M. Das, F. Fiedler, O.Hov, F. Kirchner, M. Kuhn, P.A. Makar, J.B. Milford, M.G.M. Roemer, R.Ruhnke, D. Simpson, W.R. Stockwell, A. Strand, B. Vogel and H. Vogel,Gas-phase Reactions in Atmospheric Chemistry and Transport Models: aModel Intercomparison, EUROTRAC Int. Sci. Secr., Garmisch-Partenkirchen, Germany, (1996).

GLOREAM-report

141 of 216

Evaluation of the Model System KAMM/DRAIS


Franz Fiedler, Klaus Nester and Tianliang ZhaoInstitut für Meteorologie und Klimaforschung (IMK)

Forschungszentrum KarlsruhePostfach 3640, D-76021 Karlsruhe, Germany

1. Summary

The model evaluation for KAMM/DRAIS, a mesoscale meteorological andchemical transport model, is carried out for two episodes. The first one isSeptember 16, 1992, a day of the TRACT experiment and the second one July 26,1994, a day of a summer smog episode in Northrhine-Westfalia (NRW). Asevaluation parameter the amount of the difference between measured andsimulated values is selected. The measure of quality for a certain variable is thepercentage of values, which fall into a predefined difference range.

In the TRACT case most of the considered variables of the KAMM/DRAIS modelsystem provide „quality measures“ of 40 % or more, which is a still acceptablevalue. In the NRW case only the ozone concentrations are considered. At that daythe highest ozone concentrations are measured in the areas of high precursoremissions. Only with a modified emission inventory such high ozoneconcentrations could be simulated. The model evaluation concerning this case isnot yet finished.


The model system KAMM/DRAIS consists of the meterological model KAMMand the dispersion model DRAIS. It is designed to simulate the transport,diffusion, deposition and chemical transformation of the relevant species in amesoscale area. Comparisons with measurements in a lot of studies have shownthat the models provide realistic results, but a systematic model evaluation has notyet been done. Therefore, a systematic model evaluation is the aim of this project.It will be based on data from the TRACT experiment 1992 and from a summersmog episode in July 1994 in NRW and Berlin.

The model evaluation is carried out together with other groups in the frame of theGerman Tropospheric Research Programme. An evaluation concept has beendeveloped to guarantee that the evaluation of all models is accomplished in thesame way. The results of the different models are compared with the measured

GLOREAM-report

142 of 216

data. The evaluation is carried out with partial data collectives in order todistinguish between data from different time periods of the day and from aboveand near the ground. The results of such a model evaluation indicate how realisticthe simulations are and where there are still weak points in the model or in theinput data.

3. Activities during the year and principal results

As first case for this evaluation an episode of the TRACT experiment, the 16th

Sept. 1994, has been selected. Simulations with the KAMM/DRAIS model havebeen carried out for this day. The large scale information for the simulation isderived from the results of an EURAD simulation with a grid resolution of 60 km.In the time period between 11 UTC and 15 UTC half an hour averages of thevariables considered for the evaluation are built. These data are compared withaircraft measurements performed in the southern and northern part of theexperimental area in the mentioned time period. The comparison is done in thefollowing way. For each variable a range has been determined, which describes anacceptable difference between the measured and simulated values. The „qualitymeasure“ of a model for a certain variable is the percentage of cases, which fall inthis difference range. High percentages mean that the simulation provides realisticresults. Low percentages indicate a weakness in the model or in the input data. Thecomparison with the aircraft measurements has been performed at the Universityof Cottbus for all participating model groups. Table 1 summarises the results forthe KAMM/DRAIS model. Most of the variables have values of 40 or morepercent. 40 is a still acceptable percentage. The percentage for the wind speed islower than 40 % because the range is selected too small.

Table 1 Result of the model evaluation for the TRACT case.

variable range percentageozone 5 ppb 40

nitrogene oxydes 1.5 ppb 38carbon monoxide 75 ppb 15

potential temperature 1.5 K 78specific humidity .85 g/kg 76mixing height 200 m 43wind direction 15 degrees 61wind speed .75 m/s 32

GLOREAM-report

143 of 216

The species CO and NOX provide also percentages less than 40 %. From theresults of other simulations it is known that the model underpredicts the COconcentrations. Because CO is a less strong reactive species, it can be assumed thatthe emissions used in the model underestimate the real emissions. In the case ofNOX the scatter diagram indicates also that the emissions used in the model are toolow. But NOX is much more reactive and it may be possible that the transformationinto other reactive nitrogen species is too strong in the model or the measured NOX

includes further nitrogen species and corresponds more to NOy.

Therefore, additional evaluations with the measurements at the ground levelstations are carried out. The concentration measurements of ozone and NOX at theground level stations have been prepared and compared with the results of theKAMM/DRAIS model. Two time periods, one between 11 o’clock a.m. and4 o’clock p.m. and the other one between 6 o’clock a.m. and 11 o’clock p.m., aredistinguished. The amount of the difference between measured and simulatedconcentrations for the two species is calculated and these data are summarised in acumulative frequency distribution.

For NOX remarkable differences occur in the distributions of the two periods. Thisresult is not astonishing, because the highest concentrations of NOx are measuredin the morning and evening hours. In contrast to the lower concentrations in theafternoon, which are less locally influenced, these concentrations are simulatedless accurately. The direct comparison of the measured and simulated NOX

concentrations in a scatter diagram shows, that the model results are considerablylower than the measured values. This is true for both time periods and it is inagreement with the results of the comparison with the aircraft measurements. Theaverage value of the measured NOX concentrations from all ground level stations isabout twice as high as the corresponding simulated value. Because the groundlevel concentrations depend on the emissions, this result seems to indicate anunderestimation of the emissions used in the model. But most of the stations arelocated in urban areas and may be strongly influenced by local sources. In this casea grid size of 5 km in the model is not sufficient to resolve the influence of thesesources. In order to eliminate such effects, all stations, which indicate such astrong local influence, have been eliminated for the comparison. But the scatterdiagram for the afternoon hours shows still the same behaviour as before. If thesimulated NOy concentrations are taken into account instead of the NOX values,the agreement is much better. This is an indication that the measured NOX

concentrations may include other nitrogen species and correspond more to the NOy

concentrations.

GLOREAM-report

144 of 216

Figure 2 Cumulative frequency distribution of the difference between the measuredand simulated ozone concentration

The corresponding comparison with the ozone data provides a better agreement ofthe mean value. But the scatter diagram indicates, that the lower ozoneconcentrations are overestimated and the higher concentrations are underestimatedby the model. The cumulative frequency distributions of the difference betweenmeasured and simulated concentrations for the two periods are quite similar(Fig.2). This means that the reliability of the simulation is nearly the same for thewhole day.

As second episode for the evaluation procedure, the 26th July 1994, has beensimulated. The model area comprises 210 km * 180 km and encloses the state ofNorthrhine-Westfalia. The simulation was done with a horizontal grid resolution of3 km. The emission inventory for this area has been provided by the IER Stuttgart.The other external data have been taken from an EURAD model simulation.Comparisons between the measured and simulated ozone concentrations at groundlevel have been carried out. The measurements show, that the highest ozoneconcentrations are found in areas with high emissions of the ozone precursors. Inthe less polluted areas the ozone concentrations are considerably lower. In theseareas measured and simulated ozone concentrations are in better agreement than inthe other areas, where the model underestimates the peak ozone values.

Because the emissions are only preliminary, a new simulation with modifiedemissions has been carried out. In order to find the maximum effect on the ozoneconcentrations, the NOX emissions have been reduced and the VOC emissionsincreased, both by a factor of 2. The results are really astonishing. In the areas ofhigh precursor emissions a remarkable increase of the ozone concentrations isfound. Now, at many stations they correspond to the observed values. Fig. 3 showsthe comparison of the diurnal cycle of the ozone concentration without and withthe emission modification at two city stations.

GLOREAM-report

145 of 216

Figure 3 Comparison of the diurnal cycle of the ozone concentration at 2 city stations in NRW.

At stations in the more remote areas the effect of the emission modification on theozone concentration is less remarkable as can be seen on Fig. 4. The results provethat a reliable estimation of the emissions is a prerequisite for a reliable simulationof the ozone concentration distribution in a summer smog episode.

Figure 4 Comparison of the diurnal cycle of the ozone concentration at 2 „remote“ stations in NRW.

4. Main conclusions

The model evaluation for the TRACT episode has shown that for most of theconsidered variables a percentage of about or more than 40% for the „qualitymeasure“ is found. This means that the model provides realistic results. For thespecies CO the emissions have to be improved in order to get a better agreementwith the measurements. The reason for the weak agreement in the case of the NOX

GLOREAM-report

146 of 216

concentrations is not yet clarified. If the measured NOX is really only NOX, thenthe emissions of NOX used in the model are lower than the real emissions.Otherwise the measured NOX concentrations should be better compared with thesimulated NOy concentrations.

The simulation of the NRW case demonstrates the importance of a reliableemission inventory, especially in areas, where a strong formation of ozone can beexpected. A model evaluation for such episodes performed with preliminary lessaccurate emission data has only a limited value.


The model evaluation for the NRW case will be finished with better emission data.

The third episode, July 26/27, 1994, will be simulated in the area of the city ofBerlin. The results will be compared with measured data of the FLuMOBExperiment. The evaluation will be carried out in the same way as in the TRACTcase.

6. Acknowledgements

We thank the ministry for science and technology for the financial support of theproject. All colleagues involved in the model evaluation activities areacknowledged for the good cooperation.

GLOREAM-report

147 of 216

Model Network for the Simulation and Forecast of the AirPollution - Mesoscale Simulation


Franz Fiedler, Klaus Nester and Walburga WilmsInstitut für Meteorologie und Klimaforschung (IMK)

Forschungszentrum KarlsruhePostfach 3640, D-76021 Karlsruhe, Germany

1. Summary

The model system KAMM/DRAIS is modified, so that it can be used for real-timeforecasts of the air pollution, especially the ozone concentration distribution, inmesoscale areas. The preprocessing programs, which calculate the input data forthe model system from the external data transferred from the other groups arealtered and completed. A local network program regulates automatically therunning of the whole model system beginning with the external data transfer up tothe start of the model and all necessary consecutive runs.


The purpose of the close cooperation of four groups in a network is the predictionof the air pollution, especially the ozone concentration, from the European scaledown to the smallest mesoscale in different regions. The emission data will beprovided by the „Institut für Energiewirtschaft und Rationelle Energieanwendung“(IER) of the University Stuttgart. The „Deutsche Wetterdienst“ (DWD) inOffenbach provides the routine meteorological data and uses additionally achemical transport model based on that of the EURAD group. The „Institut fürGeophysik und Meteorologie“ (IGM) at the University of Cologne contributeswith the EURAD model, which runs on the European scale with nesting options.At the „Institut für Meteorologie und Klimaforschung“ (IMK) in Karlsruhe themodel system KAMM/DRAIS is applied in the mesoscale α range.

Aim of this project is the disposition of a special version of the mesoscale modelsystem KAMM/DRAIS in a model network in order to perform real-timesimulations of the relevant air pollutants in mesoscale areas. This new modelversion will later be applied to predict the ozone concentration distribution duringsummer smog episodes. The simulations will contribute to find the best measuresto avoid an excess of the ozone concentration limits.

GLOREAM-report

148 of 216

3. Activities of the year

The external data from the other groups have to be preprocessed, so that they canbe used as input for the KAMM/DRAIS model. The emission program reduces theprovided emission data for the VOC’s to the classes of the model and calculatesthe reaction rates for all classes depending on the averaged emissions of theoriginal species. Point sources belonging to the same grid point are also combinedprovided that the source height and the other emission conditions are the same.This partly existing program was completely revised.

From the EURAD model the meteorological data are used to determine the basicstate variables of the KAMM model. Additionally, the initial conditions for thehumidity and the temperature at the surface are calculated with the EURAD data.Earlier simulations have shown, that especially under strongly varying windconditions a nudging of the horizontal flow field is necessary. These nudging flowfields are also derived by interpolating the wind data from the EURAD model.These preprocessing programs have also been modified or completed. The initialand boundary conditions for the chemical species are calculated from thecorresponding EURAD results. Now this coupling is completed. A correspondingcoupling with the DWD data is in preparation. It will then be possible to use alsothe data from DWD as input for the preprocessing programs. The connection of theKAMM/DRAIS model with the models of the other groups in the network isshown schematically in Fig. 1.

Figure 1 Connection of the KAMM/DRAIS model with the other models.

KAMM+

DRAIS

Nudging fields

Basic statefields

Initial values forhumidity and ground

level temperature

Reduction to RADM2species and

determination of HCreaction rates

IMK

Initial and boun-dary conditions for

DRAIS

Photolysis ratesLand use data

Orography dataPrediction of meteorolo-gical variables and air

pollutant concentrationsin a subregion

IGM

IER

DWD

Localmodel

EURADmodel

Emissionmodel

GLOREAM-report

149 of 216

In order to use the model system KAMM/DRAIS as real-time model it is necessaryto regulate the running of the whole model system from the beginning of the datatransfer to the IMK node. This is now realised in a local network program, whichhas the following structure. At the beginning the different preprocessing programsare called consecutively. They calculate the emissions, the basic state variables,the nudging fields, the initial fields of the ground level temperature, the specifichumidity and the concentrations of the considered species as well as the boundaryconditions of these species on different processors. At the end of the networkprogram it is tested, whether all data for the model run are available. In that casethe start run is initiated, which again starts all necessary consecutive runs. In theother case the network program finishes. The individual preprocessing programsdo not only calculate the input data for the model system. They also control thedata transfer. The data of the other groups are stored on a workstation at the IMK.First, the preprocessing programs check, whether the necessary data used by theindividual program are present on the workstation. If the data are available, theyare transferred to the vector computer VPP300, where the program is executed.Otherwise, this test is repeated several times in given time steps. After the last testthe program ends with a special error code. In order to avoid unnecessary datatransfers in the case of several applications of the network program for the sameepisode, it is additionally tested, if the data are already available on the VPP300.The nomenclature of the data files provided by the other groups and the contentsof these files have been commonly fixed, which is necessary for a smooth-runningdata exchange in the network. A schematic representation of the local networkprogram is given in Fig. 2.


The network program has been successfully applied for the simulation of theNorthrhine-Westfalia case, a smog episode on July 26, 1994. The emission datahave been provided by IER Stuttgart. All other data for the preprocessing camefrom an EURAD simulation. The ozone concentration measurements at that dayshowed a special feature. The highest concentrations are measured in areas mostlypolluted by the precursors. In the remote areas the peak ozone concentrations areconsiderably lower. These values are quite well represented by the model results.This is demonstrated by the comparison with the measured ozone concentrations attwo stations in this region (Fig. 3). In the higher polluted areas the observed ozoneconcentrations are remarkably underestimated by the modelled values. In theframe of the model evaluation project it has been tried to find the reason for thisdiscrepancy in order to improve the model results.

GLOREAM-report

150 of 216

Figure 2 Schematic representation of the local network program.

Notice of themissing data

PutIV1=1

Are all dataavailable?

Furtherpreprocessing

programs

Are results of the2. preprocessing

program availableon the VPP?

Are input data of the1. preprocessing

program availableon the HP6?

Are input data of the1. preprocessing


Are results of the1. preprocessing


START

END

Startconsecutive runs

StartKAMM/DRAIS

PutIV1=1

PutIV1=0

Is the number ofmaximum callsIzmax reached?

Transfer input data for the1. preprocessing

program to the VPP

Start1. preprocessing

program

yes

yes

yes

yes

yes

no

no

no

no

no

GLOREAM-report

151 of 216

Figure 3 Diurnal cycle of the measured and simulated ozone concentration at two stations.

5. Main conclusions

The network version of the model system KAMM/DRAIS is realised, so that it canbe applied with the emission data from IER and the data of the EURAD model. Itruns automatically, if all external data are available. Only the model domaintogether with its characteristics has to be specified before the start of the localnetwork program.


As next step the modification of the preprocessing programs will be finished. Thenit is possible to use the results of the DWD model, which has another coordinatesystem. Additionally, it is foreseen to realise a parallel version of the model systemKAMM/DRAIS in order to reduce the run-time of the model.

7. Acknowledgements

We thank the ministry for science and technology for the financial support of theproject. All colleagues involved in the network activities are acknowledged for theexcellent cooperation.

GLOREAM-report

152 of 216

Boundary-Layer and Cloud Convection in AtmosphericChemistry Models


Arthur C. PetersenInstitute for Marine and Atmospheric Research Utrecht, Utrecht University

P.O. Box 80005, 3508 TA Utrecht, The Netherlands

1. Summary

Convection and chemistry in the atmospheric boundary layer are studied.Specifically the question is addressed whether chemical species need specialtreatment in large-scale chemistry models related to specific aspects of theturbulent transport-reaction problem. We conclude that for the clear boundarylayer this is not the case. Although originally planned, cloud convection has notbeen studied in this finished contribution.


The goal of this contribution is the improvement of large-scale (global andmesoscale) 3-D modeling of chemistry in the troposphere. This is achieved byadding a number of significant physical and dynamical processes which cannot beresolved in space and time by large-scale models. These processes includeturbulent vertical transport, short-wave radiation, and their influence on chemicalprocesses in the convective boundary-layer and convective clouds. Newparameterizations are developed and tested through process studies with large-eddy simulation (LES) and are included in a 1-D version of a climate model.


The influence of turbulence on chemistry in the convective atmospheric boundarylayer (CABL) has been studied. An important aspect of the turbulent transport-chemistry problem is the fact that reactive species are not always well-mixed dueto short chemical timescales associated with certain important reactions-shorterthan or comparable to the convective mixing timescale. A decomposition ofboundary-layer air in upwards and downwards flowing air, updrafts anddowndrafts, was used to study the segregation of chemical species (this is calledthe “mass-flux approach”). Another aspect is the influence of chemistry on fluxestimates of reactive species. In January 1999 this contribution to GLOREAM wasfinished, coinciding with the end of the Ph.D. research project of Arthur Petersen.Several papers were written in the period 1997-1998.

GLOREAM-report

153 of 216


An LES investigation of the mass-flux characteristics of both reactive andnonreactive scalars reveals that 65% of the flux is captured by an updraft-downdraft decomposition and about 25% of the covariance between two arbitraryscalars. The CABL profiles of scalars are modeled accurately by an off-line mass-flux scheme that has two major simplifying assumptions in it, as compared to theexact plume-budget equations. First, only one-way lateral exchange is considered,and second, the subplume-scale fluxes are taken into account by an increase of themass flux. The accuracy of the mass-flux estimate of the covariance is estimated tobe a factor of 2, which is accurate enough to improve modelled reaction rates bytaking the estimated covariance into account.

Based on the mass-flux characteristics of scalars a simple covarianceparameterization is developed which can be used in atmospheric chemistry modelsto assess the importance of turbulent covariances for atmospheric chemistry. Theparameterization considers both turbulent and chemical production of covariances.It makes use of a distinction between short-lived and long-lived species. The short-lived species do not have to be transported by the flux closure that is used incombination with the covariance closure. The evolution of the updraft anddowndraft concentrations of nontransported species is completely determined bychemical reactions involving also longer-lived species (that are transported).

It is not necessary to combine this covariance parameterization with a mass-fluxscheme for flux. A nonlocal first-order closure suffices to accurately model fluxand concentration profiles of reactive species in the CABL, provided that the fluxdivergence due to chemistry is also taken into account in the nonlocal contributionto the flux. Concerning the potential error related to the neglect of the higher-orderchemical terms in the flux budget it is found that this error is the largest forrelatively complex flux closures (like full second-order flux closure) and thesmallest for very simple flux closures (like the nonlocal first-order flux closure).This explains why a simple flux closure can model the fluxes in thephotochemistry case accurately while in a second-order flux closure a large fluxcorrection due to higher-order chemical terms is indicated.

The covariance closure, developed using nonreactive and simple chemistry cases,is found to compare well with LES for a more complex photochemistry case. Thisprovides support for the actual application of the covariance parameterization in afull climate-chemistry model. A study was presented where a single-columnversion of a global climate-chemistry model is used to study the impact of thecovariance parameterization. The model is run for several locations on the globe. Itis found that the impact of turbulent covariances on the modeled concentrations isvery small (less than 1%) everywhere.

GLOREAM-report

154 of 216

5. Main conclusions

The updraft-downdraft decomposition is very useful for solving the turbulenttransport-reaction problem. More specifically we found that:

1. sufficiently accurate estimates of covariances can be made on the basis of theupdraft-downdraft decomposition;

2. the tool that was used to study this decomposition, LES, showed no problemsrelating to LES subgrid-scale covariances (these subgrid-scale covarianceswere typically negligible, even near the bottom and top of the CABL).

The mass-flux approach for calculating covariances can be combined with any fluxclosure. A first-order covariance closure can be formulated that takes bothturbulent and chemical production of covariances into account.

The contribution of chemical higher-order moments to the flux in higher-orderclosures is typically larger than the error that is made by using a first-order fluxclosure. The latter error is expected to be small in atmospheric chemistry.

Turbulently and chemically produced covariances contribute at most 2% to thedaytime species budget of the NMHC species isoprene. The impact onconcentrations is typically only 0.5%. Other species are often even less affected bythe modeled covariances. One must note, however, that in reality covariances arealso produced by subgrid-scale surface heterogeneity (affecting emission anddeposition patterns). This heterogeneity is expected to lead to considerable effectsfor large-scale models with grid sizes that are significantly larger than 10 km.

6. Results of policy relevance

One of the uncertainties in computer models of air quality has been eliminated.Formerly is was not known whether the influence of turbulence on chemicalreactions in the clear atmospheric boundary layer could be neglected. Now it hasbeen shown using fine-scale computer models that this is permissible.


This contribution has finished in January 1999. All project time has been used forthe study of the clear boundary layer. Although originally planned, convectiveclouds have not been studied in this GLOREAM contribution.

8. Acknowledgements

We acknowledge support by the Netherlands Earth and Life Sciences Foundation(ALW, formerly the Geosciences Foundation, GOA) with financial aid from theNetherlands Organization for Scientific Research (NWO). The contribution was

GLOREAM-report

155 of 216

sponsored by the National Computing Facilities Foundation (NCF) for the use ofsupercomputing facilities.

GLOREAM-report

156 of 216

Application of Cloud Information and Classification fromMETEOSAT Data for CTM REM3


Andrea Oestreich, Johannes Flemming and Eberhard ReimerInstitute of Meteorology, Freie Universität Berlin

Carl-Heinrich Becker-Weg 6-10, D-12165 Berlin, Germany

1. Introduction

The photochemical transport model REM3 is used to forecast operationally ozonesince 1997. The „Europa-Modell“ of the DWD supplied the required meteorologicalinput data base. In order to investigate the influence of the meteorological forecasterror, diagnostic calculations are carried out for the years 1994-1998. Beside theforecast application, long term calculations are of importance because theenvironmental concern focusses more and more on long term threshold values likeAOT40. Long term diagnostics provide the possibility to evaluate the model not onlyin an episodical or case study context but in a seasonal time scale.

For comparison an analysis scheme for concentration measurements, based onoptimum interpolation, is used to tackle the problem of the difference in the scalebetween the model and the observations. For now only German ozone measurementscould be included in the analysis procedure.

The evaluation of the model quality revealed a rather good performance in periodswith higher ozone values. Larger biases occurred in episodes with lower ozone, whenthe model tends to simulate even lower concentrations. Cloud cover data appeared tobe the most sensitive meteorological input for the CT modelling, because the cloudcover controls photolysis frequencies and the mixing layer growth.

However, the cloud cover forecast by the Europa-Modell is rather inaccurate,because many of the processes, which lead to cloud formation are approximated insuch a model. The comparison with diagnostic, i.e. observed data showed a differentprobability density for the total cloud cover. The „Europa-Modell“ data tend to havemore total coverage then the diagnostic data. The parameterisations in REM3 lead toa pronounced reduction of the photolysis frequencies, when the total cloud coverequals 8 okta. Thus, an underestimation of the photolysis frequencies has to beassumed. In order to correct this underestimation an „effective total cloud cover“value was introduced. It reflects the different transmissivity of different cloud levelsby utilising a parameterisation by Stull (1988).

GLOREAM-report

157 of 216

Correct information about the cloud cover and its influence on the photolysisfrequencies and the development of the mixing layer is essential to improve chemicaltransport modelling. For the diagnostic application of CTM cloud information is alsoimportant to handle problems of heterogeneous chemistry. Satellite data are aprospective way to require diagnostic information about the horizontal structure ofthe cloud cover and cloud type.

The parameterisations for the influence of the clouds have to be adapted to thedifferent origin of data. Investigations have been started to improve REM3 forecastsand diagnostics by considering information about the cloud type and cloud levelfrom satellite.

2. Utilisation of satellite data

Cloud cover as well as temperature may show considerably small scale variations,which are poorly detectable by the relatively low spatial resolution of the synopticstations. With the use of METEOSAT data the spatial resolution will be enhanced.The surface temperatures can be calculated from the emitted radiance measured bythe infrared channel and from atmospheric parameters (e.g. the amount of watervapour). Especially in relation to clouds the use of the satellite data leads to muchmore information, because the spatial distribution of clouds with respect to heightand thickness can be determined with a bispectral cloud classification. For long termmodelling the first requirement was therefore to embed the classification algorithminto an operational mode.

3. Data set

The Institute for Meteorology receives the METEOSAT data directly with a PDUSstation (Primary Data Users Station). Different sectors, like the European Sector(BERMET) used here, are archived. The time resolution for the test case from23.7.94-1.8.94 is one hour for the BERMET sector. The three METEOSAT-5channels are all available:

- the visible channel (VIS): 0,4 - 1,1 µm- the infrared channel (IR): 10,5 - 12,5 µm- the water vapour channel (WV): 5,7 - 7,1 µm

The water vapour channel shows little differences between cloudy and cloud-freeareas and has only little advantage for a classification, but complicates it. Thereforethe WV-data are not used here.

The resolution at the sub-satellite point is 5 x 5 km2. The original data were rectifiedinto a polar stereographic projection (after a procedure by Koslowsky). Theprojection is orientated at 60°N and the resulting resolution can be chosen in relation

GLOREAM-report

158 of 216

to the requirements of the further analysis. For a better compability with NOAA-AVHRR-data a resolution of the stereographic projection of 1,2 x 1,2 km2 is useful.

4. Used Sectors

Six sectors of Europe were chosen:

For five sectors the spatial resolution is 1,2 x 1,2 km2. The main sector, CEUROPA,is centered in Central Europe and covers the region from ~45°N to ~60°N and from~3°W to ~25°E, rsp. If adjacent regions are needed, slightly different sectors can beused. They are more or less a shift of CEUROPA to the four main directions and arecoded with the first letter of the directions (EEUROPA, SEUROPA, WEUROPA,NEUROPA). All in all, the sectors cover an area from ~42°N to ~65°N and from~8°W to ~30°E. The sixth sector includes the PHOXA-area and is namedTEUROPA. The spatial resolution is lower and amounts to 2,0 x 2,0 km2. The imagesize is in all cases 1536 x 1536 pixels (requirement of the classification algorithm).

Because the models are working with phi/lambda-co-ordinates, the satellite datashould be transformed to this geographic projection. The sectors are coded with a “g”at the end. The size of the different sectors is oriented at the test case of July 1994, Trepresents the PHOXA-grid. Tab. 1 gives an overview:

SECTOR LATITUDINAL

COVERAGE

LONGITUDINAL

COVERAGE

RESOLUTION 0,25° x 0,5° ARE

REPRESENTED

BY... (NO. ofPIXELS)

CEUROPAG 47°N - 58°N 0°E - 19°E 1' x 1' 30 x 15

EEUROPAG 46°N - 60°N 6°E - 30°E 1' x 1' 30 x 15

SEUROPAG 43°N - 55°N 0°E - 20°E 1' x 1' 30 x 15

WEUROPAG 47°N - 59°N 8°W - 16°E 1' x 1' 30 x 15

NEUROPAG 50°N - 64°N 2°W - 22°E 1' x 1' 30 x 15

TEUROPAG 42°N - 66°N 10°W - 30°E -- 20 x 16

Table 1 Different sectors, area coverage and resolution.

GLOREAM-report

159 of 216

The last column indicates the sum of pixels being the basis for the calculations forone PHOXA-grid-cell, e.g. fractional cloud cover or mean albedo.

The image size for the transformed data with 1536 x 1536 pixels is the same as forthe images in stereographic projection, but the real image size varies from sector tosector. Only the TEUROPAG-sector fills the whole image.

There is principally a free choice of sector area and -size, but for every sector alandmask, which matches exactly the sector, is needed. While for the stereographicprojection a new landmask has to be created for every new sector, the landmasks forthe phi-/lambda-co-ordinates can easily be received by the transformation from thestereographic projection landmask into geographic projection.

5. Cloud classification

Clouds in different heights and spectral intervals have different radiative properties.In a two dimensional VIS/IR-histogram different cloud types have a distinctivesignature and can be separated with a priori defined thresholds from each other(although the transitions of the different cloud layers and thicknesses arecontinuously) as well as from the underlying surface. For example, while thin cirrushas a high transmittance in the visible spectrum and therefore a relatively similaralbedo compared to the surface in the VIS-channel, it has a much less emittance inthe infrared because of its low top-of-cloud temperature. The result is a clearly lowergrey value compared to the surface.

Because of the utilization of the VIS-channel, the used classification algorithm issuitable only for daytime. For a maximum-likelihood-classification (up to 24) testclasses have to be determined. A trained interpretor is able to decide, which area in asatellite image can be seen as to be cloud-free or filled stratus and so on. Those areasare defined as test classes, for which the mean values and the covariance matrix arecalculated. In the next step the probability to belong to the different classes iscalculated for every pixel. The pixel will be assigned to the class with the highestprobability.

Berger has developed a procedure for an objective determination of the test classes.Before running the classification, it has to be decided, which classes should beused.The next step is the elimination of erroneous pixels with the help of an offsetfor each of the both channels. The grey values of the VIS image, which represent thesolar reflectance, are corrected by the cosine of the solar zenith angle and the result isan albedo image. Then the grey values of the IR image representing the black bodyspectral radiance are inverted. Here some specific coefficients depending on thesatellite and sensor calibration are needed and have been taken from theMETEOSAT calibration reports. The result is a brightness temperature (at the top ofthe atmosphere) image. Both images are the basis for the classification, which can beapplied to every region, season and time (if the solar zenith angle is lower than 80°).

GLOREAM-report

160 of 216

Land area is separated from ocean and lakes by a landmask. The pixels with thelowest grey values in the albedo image as well as in the temperature image representthe cloud-free areas, once for water, once for land. Between the cloud-free pixels andthe pixels with the highest values albedo and brightness temperature for all cloudlayers and all optical thicknesses are interpolated. From the initial declaration, whichclasses should be used, the algorithm reads the parameters of each class in the"artificial" histogram and tries to find corresponding areas in the images. For thefound test classes, the real parameters are taken, for the others those given by thehistogram. Afterwards, albedo and temperature image are classified.

For the application of the cloud classification in models an automation of theprocedure was required. The classification algorithm itself has not been changed.Firstly the METEOSAT data are rectified and the satellite coefficients mentionedabove are saved in an extra coefficient file. If the classification of an image withphi/lambda-co-ordinates is wanted, the transformation has to be inserted. The initialdeclarations for the satellite coefficients, the first and last hour, input and outputstorage and the code of the sector are stored in a parameter file, so that thecalibration reports are not needed. A start file reads the parameter file, the coefficientfile and the co-ordinates for the sun height correction of the chosen sector, and handsover the parameter to the classification procedure. The result is a classified imagewith 24 classes (with two classes each for cloud-free land and ocean areas).Afterwards the parameter file is automatically updated for the next classification.

Actually the start file is layed out for an hourly classification for one day. A storagecapacity of at least 177 000 blocks (~175 mb) is needed.

The procedure needs only initial declarations for each day and a control of the datastorage and satisfies principally the requirements of the boundary layer model due tooperational disposal of the classification results.

The results of the classified image can be used directly in model calculations. Undercertain circumstances it may be useful, to distinguish only between cloud-free andcloudy areas or between optical thin and optical thick cloudiness, rsp. For these casesa cloud mask is produced from the classified image. All pixels not assigned to thewanted classes are set to zero, while the other pixel values are kept. The cloud maskcan now be overlayed to the albedo and temperature image.

The masked images can be classified a second time (with different class declarationsfor the classification algorithm). The advantages would be a more differentiatedclassification of the cloud-free areas (e.g. distinction between great forests and cities)and an elimination of pixels representing very thin or scattered clouds, which hadbeen classified as cloud-free at the first classification.

GLOREAM-report

161 of 216

6. Calculation of surface temperature and albedo

The albedo image is produced within the classification procedure. If the image iskept after the classification, an albedo exists for each cloud-free pixel. It has to bementioned that the albedo values are the normalized values due to vertical solarincidence.

In the temperature image produced for the classification, the pixels represent thebrightness temperature at the top of the atmosphere. The surface temperature foreach pixel is received with an atmospheric correction (LOWTRAN-7). For thecalculation the CALMET-procedure from Billing is used. Different atmosphericparameters like water vapour content, aerosol or stability are introduced. Actually astandard atmosphere is used, which differs usually from the real atmosphere.Alternatively vertical profiles from analyzed temperature and humidity fields will beused in a next step (Reimer and Scherer, 1992) for atmospheric correction of thesatellite data.

7. Look-out

The next steps are

- the geographic separation of the surface test classes,- the use of atmospheric correction of the IR-data for the classification

algorithm,- the utilization of analyzed vertical profiles instead of the standard

atmosphere,- the elaboration of the second classification algorithm.

For up to three years the cloud information from satellite data will be determined fora Central European area and will be used for validation and tests with the CTMREM3.

8. LiteratureBerger, F., Die Bestimmung des Einflusses von hohen Wolken auf das

Strahlenfeld und auf das Klima durch Analyse von NOAA-AVHRR Daten,Met. Abh. Inst. f. Met. FU Berlin, NF, Serie A Monographien Band 6 Heft 3,(1992).

Berger, F., S. Jagdhuhn, B. Rockel and R. Stuhlmann, Radiation BudgetComponents inferred from NOAA-AVHRR and Meteosat Data for theBaltic Sea, in: IRS'96: Current problems in atmospheric radiation, (1996).

Flemming, J. and E. Reimer, The impact of special features of numericallypredicted and analysed meteorological data on the results of ozone forecastby a PBL-CTM, in: Air Pollution Modelling and its Application XIII,Plenum Press, New York, (1998).

GLOREAM-report

162 of 216

Flemming, J., E. Reimer and R. Stern, Comparison of photochemical model resultsbased on diagnostic and prognostic meteorological input data, in:Proceedings of EUROTRAC Symposium 98, Editors: P.M. Borell andP. Borell, WITpress, Southampton, (1999), 472-482.

Hass H. and A. Ruggaber, Comparison of two Algorithms for CalculatingPhotolysis Frequencies Including the Effects of Clouds, Meteorol. Atmos.Phys. 57, (1995), 87–100.

Reimer, E. and B. Scherer, An operational meteorological diagnostic system forregional air pollution analysis and long-tem modelling, in: Air PollutionModelling and its Application IX ,Plenum Press, New York, (1992).

Stull, R.B., An Introduction to Boundary Layer Meteorology, AtmosphericSciences Library, Kluwer Academic Publishers, Dordrecht, (1988).

GLOREAM-report

163 of 216

Development of a model system for an operationalozone forecast at the DWD


Jürgen Rißmann,Ingo Jacobsen, Stefan Tilmes, Jörg Zimmermann

Deutscher Wetterdienst,Abteilung Klima und Umwelt

Kaiserleistraße 42, D-63067 Offenbach

1. Summary

In a close cooperation with other projects an integrated model system for anoperational ozone forecast is going to be developed at the Deutscher Wetterdienst(DWD). The model system consists of the coming non-hydrostatic mesoscaleNWP model of the DWD, the Lokal-Modell (LM), the emission calculation modelECM, and the chemistry transport model EURAD-CTM. First test forecasts inAugust, 1998, have shown the potential of the system for operational forecasts. Forhigh resolution applications a massive-parallel two-way nesting option has beendeveloped for the LM.


Within the framework of the German Tropospheric Research Programme (TFS)the DWD is part of a working group, whose aim is to develop a model system fortropospheric ozone modelling on regional scales. The partners of the DWD are theInstitute for Geophysics and Meteorology (IGM) at the University at Cologne, theInstitute for Energy Economics and the Rational Use of Energy (IER) of theUniversity of Stuttgart, and the Institut für Meteorologie und Klimaforschung(IMK) at the University of Karlsruhe. The aim of the group is to develop a modelnetwork consisting of a numerical weather prediction model (LM of DWD), anemission forecast model (ECM, new development by IER), a regional chemistrytransport model (EURAD-CTM by IGM), and a high-resolution meteorologicaland chemistry transport model (KAMM/DRAIS by IMK). The model network iscapable to work in a distributed mode on different nodes with data transferbetween the nodes via internet as well as in an integrated mode for time-criticaloperational real-time applications with all single models running on the same node.

The special interest of the DWD is the use of the model system for an operationalozone forecast (i.e. in the integrated mode), whereas the other groups are more

GLOREAM-report

164 of 216

interested in distributed model runs for hindcasting of photo smog episodes,budget investigations, model evaluations, emission reduction scenarios, etc.


The project started working in March, 1997. The cooperation with the networkgroup comprised the definition of a model domain, the horizontal and verticalstructure of the model grids, the use of a common land-use database (contributionof Fraunhofer Institute for Atmospheric Research in Garmisch-Partenkirchen), andagreements regarding the interfaces between the different models.

Another main task of the DWD project has been the development andimplementation of a two-way nesting option in the LM to allow for modellingconsistent meteorological data for different scales. This feature, which thecommon meteorological driver of the EURAD-CTM, the MM5, already exhibitssince several years, is required for chemistry transport modelling with the networkmodel on different scales.

Figure 1 LM and EURAD-CTM results for August, 12, 1998, 14:00 UTC, 38 hour forecasts.Grid with 109 × 109 grid points and a horizontal resolution of about 21 km.

The first results of LM-CTM test runs showed severe problems with the couplingof the EURAD-CTM, commonly used with the hydrostatic MM5, and the non-

GLOREAM-report

165 of 216

hydrostatic LM, resulting in unrealistically high ozone concentrations near thesurface in mountaineous regions. These problems at first led to extensiveinvestigations regarding the vertical velocity, which is a prognostic variable in theLM. These investigations could prove the LM vertical velocities as reliable. Thereason for the errors finally could be found in the neglection of the local massdivergence term (Dudhia, 1993) in the transport equations of the EURAD-CTM,which vanishes for hydrostatic calculations, but is an important term in a non-hydrostatic model above steep orography.

First successful ozone forecasts were yielded for the photosmog episode in August,1998, when operational test runs were carried through from August, 5 untilAugust, 15, with a data set of preliminary emissions for 1998 substituting thepredictions of the ECM, which was still under development during 1998.

Fig. 1 shows the LM temperature, wind, and pressure reduced to mean sea level,and the EURAD-CTM of near surface ozone mixing ratio, both 38 hour forecastsfor August, 12, 1998, 14UTC. While the day before the highest ozoneconcentrations were measured during the summer, at August, 12, trafficrestrictions have been applied in some federal states in the south-west of Germanydue to exceedings of the threshold value of 240 µg/m3.

Beside the works with respect to the model system the development of the two-way nesting option for the LM proceeded, at first in a sequential version for ashared-memory computer, which was finally parallelized for a distributed-memorymachine as an implementation of full value for the massive parallel LM.

Figure 2 LM forecast on different two-way nested grids, 15 hour forecast for August, 12, 1998,15:00 UTC, Coarse grid with 109 × 109 points and 21 km horizontal resolution,1. nest 103 × 103/7 km, 2. nest 82 × 82,/2.3 km.

GLOREAM-report

166 of 216


The operational ozone forecast test runs in August, 1998, were performed with thenetwork model system with the following setup:

− The model domain spans an area of about 2200 × 2200 km2 of central Europe.The horizontal grid has 109×109 grid points with a resolution of 0.1875°(≈ 21 km).

− The meteorological driver, the LM, is a non-hydrostatic limited area modelwith rotated geographic coordinates and a hybrid vertical coordinate, whichallows for a combination of terrain following σ-levels in the troposphere andp-levels in the stratosphere. For the ozone forecast, according to the σ-coordinate of the EURAD-CTM, the LM is running only with a pure σ-systemwith 32 layers up to 20hPa. However, the results of the investigations of thevertical velocity recommend the use of a hybrid coordinate system due tomuch lesser orographic effects in the stratosphere.

− The lateral boundary values are provided by the operational Europa-Modell ofthe DWD with a horizontal resolution of 55 km. The massive parallel LM isrun on 64 processors of the CRAY T3E of the DWD and needs about50 minutes computing time for a 48 hour forecast.

− The anthropogenic emissions used for the chemistry transport modelling, arecalculated for typical warm sunny weather conditions by the IER based on theemissions of 1994, separately for working days, saturdays and sundays.

The EURAD-CTM, adjusted to be driven by the meteorological input of a non-hydrostatic model, is run with the same horizontal resolution and 16 vertical layerswith the same vertical resolution near the surface. The sequential EURAD-CTM isrun on the CRAY YMP8 of the DWD and needs about 7 hours computing time fora 48 hour forecast due to a bad autotasking behaviour of the program.

A first version of a massive parallelized two-way nesting algorithm for the LM hasbeen completed, which allows for an arbitrary number of nests of arbitrary nestingdepth. The interpolation procedures are based on the MM5 nesting implementation(Zhang et al., 1986, 1987, Grell et al., 1994) and implemented by means of arecursive program organization. Results of a 12h forecast for August, 12, 1998,with two nested grids are shown in fig. 2.

5. Main conclusions

The ozone forecasts for August, 1998, have demonstrated, that the network groupsuccessfully has implemented the main part of the kernel of the model system,consisting of the non-hydrostatic numerical weather prediction model LM of theDWD and the regional chemistry transport model CTM of the IGM/EURADgroup. The models system brings together a (pre-)operational NWP model, whichensures an operational verification and continuous improvement of the

GLOREAM-report

167 of 216

meteorological data, and a well accepted, often used and evaluated chemistrytransport model. Up to now the computing times for a CTM forecast are a limitingfactor for real-time applications. First approaches to optimize the programstructure yielded encouraging results. Hence, two hours CPU time for a 48hforecast, which are a threshold value for an operational forecast at the DWD, areexpected not to be out of reach. The two-way interactive nesting option of the LMwill allow for chemistry transport calculations on different horizontal scales.

6. Political relevance

The main goal of the project is the development of a fully operational modelsystem. This is not only the prerequisite for the use of Eulerian models withcomplex chemistry in real time ozone forecasting but also for several other taskswhich need long term simulations, e.g. AOT40. With respect to ozone forecasting,operational CTMs can be used to simulate the effects of emission reduction priorto the enactment of legal steps.


During 1999 the forecast model for anthropogenic and biogenic emissions, ECM,developed by the IER, will be included into the model system and will be aqualitative step towards a real-time prediction of atmospheric chemistry. Real-timeforecasts with all components LM-ECM-CTM and high resolution calculationswith KAMM/DRAIS chemistry transport modelling driven by these results areplanned for the summer period of 1999.

8. Acknowledgements

The development of the network model system is a joint project of differentworking groups within the TFS framework. We thank our collegues from the IGM(Hermann Jakobs, Joachim Tippke, and Adolf Ebel), the IER (Burkhard Wickert,André Heidegger, and Rainer Friedrich), and the IMK (Walburga Wilms, KlausNester, and Franz Fiedler), and the IFU (Gerhard Smiatek) for their contributionsand the fruitful cooperation.

9. References

Dudhia, J., A nonhydrostatic version of the Penn State-NCAR mesoscale model:Validation tests and simulation of an atlantic cyclone and cold front, Mon.Wea. Rev. 121, (1993), 1493-1513.

Grell, G.A., J. Dudhia and D.R. Stauffer, A description of the fifth-generationPenn State/NCAR mesoscale model (MM5), NCAR Technical Note TN-398+ STR, (1994), 122p.

GLOREAM-report

168 of 216

Zhang, D.-L., H.-R. Chang, N.L. Seaman, Th.T. Warner and J.M. Fritsch, A two-way interactive nesting procedure with variable terrain resolution. Mon.Wea. Rev. 114, (1986), 1330-1339.

GLOREAM-report

169 of 216

On the use of INTERNET resources to improve substantially theOperational Regional and Urban model (OPANA)


R. San José1,M.A. Rodriguez,1 I. Salas1 and R.M. González2

1 Environmental Software and Modelling Group, Computer Science School,Technical University of Madrid, Boadilla del Monte 28660 Madrid (Spain)

2 Department of Meteorology, Faculty of Physics, Complutense University ofMadrid, Ciudad Universitaria, 28071 Madrid (Spain)

1. Summary

INTERNET resources have increased exponentially during the last five years. Thespeed connection has increased several hundreds and the number of servers andcomputers connected is becoming closer to 300 million. This situation is bringinga new concept and architecture on modelling: the distributed modelling systems.The Operational air quality models such as OPANA (the original academic versionis called ANA) request important computer power and data on-line bothmeteorological and air pollution network data. The nesting capabilities of themodels have powered the possibilities of zooming over regional and urban areashowever the requests of having large domains for longer forecast periods continueto be a must. This contribution is focusing on this particular topic on operationalmesoscale air quality modelling and shows how using WWW resources thetemporal horizon of regional and urban models are substantially increased. Themodelling system makes use of the MRF/AVN global spectral meteorologicalinformation located in different Internet servers to initialize the OPANA modelwith vertical meteorological numerical soundings during the period of simulation.Incorporating a JAVA interface to allow a fast and automatic communicationbetween user computer (where OPANA is located) and MRF/AVN servers carriesout this feature.

2. Aim of the research: The JAVA/ANA project

The ANA model is a Mesoscale Air Quality model for regional and urban domains(San José et al., 1997). In 1996 the laboratory was involved in the EMMA projectwhich was funded by the EU in DGXIII-Telematics for Environment program. Theobjective of this project was to provide to the users (Environmental Offices inEuropean cities and regional communities) with sophisticated environmentalsoftware tools -which were being used in the universities and researchlaboratories- in a way that they could be used operationally for forecasting and

GLOREAM-report

170 of 216

environmental impact studies. OPANA is an operational version of the ANAmodel, which fulfils these requirements. The EMMA version of OPANA is theoperational version, which was applied over Madrid Community. Operationalversion of Mesoscale Air Quality Models have to fulfil specific requirements suchas being operational which means the following items: a) a friendly graphical userinterface, b) simulation times in accordance with the forecasting objectives and c)computational platforms should be cost competitive. OPANA was developed withthe Tcl/Tk friendly graphical user interface (VISANA) (interpreter softwarelanguage) which allows managing software applications in FORTRAN, C, C++,etc. VISANA allows the user to prepare the initial meteorological and air pollutiondata sets, to run the simulation and to visualize the results. The points b) and c)were solved by designing a proper model domain architecture to allow to run theapplication under overnight conditions (12-16 hours) with a medium powercomputer workstation. The model domain and the computer power limited thequality of the OPANA forecasts. The JAVA/ANA project is focusing onimproving the quality of the results without increasing model domain size andcomputer power.


MRF/AVN is a global spectral meteorological model (Medium Range Forecast)developed by Sela (1982). MRF focus on long ranges forecasts and AVN isfocusing on medium and short-range forecasts. NOAA (National Oceanic andAtmospheric Administration) is running operationally these models since 1998.The NOAA NCEP (National Centers for Environmental Prediction) centers arerunning these models under CRAY-90 machines (in 1999 the 5 CRAY’s will besubstituted by the Class VIII Computer System (IBM) with high performancecomputing architecture which will increase on about 100 times the performance).Air Resources Laboratory (ARL) workstations access CRAY-90 databases andserves the information to the ARL Web servers which are accessed by theINTERNET users. The MRF/AVN data is provided in GRIB format and ARLworkstations generate text format for being accessed by request of an Internet user.A JAVA user interface was incorporated to the OPANA software tool to allow theOPANA user to perform all Internet processes automatically so that the ARL webserver requirements are transparent for the OPANA user.

MRF is running ever 24 hours (at 0h00) and with a temporal resolution of 12hours, the output spatial resolution is 191 km, the number of vertical levels is 13and the forecast duration is 288 hours. AVN is running with two modes: AVN111and AVN191. AVN111 is running ever 12 hours (at 0h00 and 12h00) and with atemporal resolution of 6 hours, the output spatial resolution is 111 km, the numberof vertical levels is 23 and the forecast duration is 48 hours. AVN191 is runningever 6 hours (at 0h00, 06h00, 12h00 and 18h00) and with a temporal resolution of6 hours, the output spatial resolution is 191 km, the number of vertical levels is 13and the forecast duration is 72 hours. Figure 1 shows the locations on the Madrid

GLOREAM-report

171 of 216

Community model domain (80 x 100 km) where the vertical meteorologicalnumerical prediction soundings are downloaded and incorporated at the ANAmodel as initial information for the 120 hours simulation period. Figure 2 showshow the different vertical meteorological numerical soundings are implementedduring the simulation at each spatial location in the model domain. Figure 3 showstwo ozone patterns when comparing monitoring station data -in this example wesee the ozone concentrations which have been taken at “Alcobendas” monitoringstation- with results from OPANA simulations -with and without MRF/AVNinformation-. The preliminary results are indicating that MRF/AVN is going toincrease substantially the quality of the air pollution forecasts in spite ofMRF/AVN are not incorporating an air quality chemical transport model yet inorder to download also the vertical information from the different air qualityspecies. The computer power and model domain as remain as it is during all theexercise.

Figure 1 Locations where MRF/AVN vertical numerical meteorological information (wind,temperature and humidity) is incorporated to OPANA model for improving thequality of the results.

GLOREAM-report

172 of 216

Figure 2 The different MRF/AVN versions related to different spatial and temporalresolutions are incorporated at different times steps into OPANA model.

Figure 3 Ozone patterns when comparing monitoring data with different OPANAversions (with and without MRF/AVN information).

GLOREAM-report

173 of 216

4. Main conclusions

We have improved the quality of the results of the operational version of theMesoscale Air Quality Model (ANA) without increasing the requested computerpower and model domain (which means that the same landuse and topography areused) by using operational vertical meteorological numerical information (wind,temperature and humidity) from the MRF/AVN global spectral meteorologicalmodels (NOAA-NCEP-ARL). Further investigations should be done onquantifying the improvement by running a significant number of simulations underthese new input requirements. The modelling system is capable of downloading allrequested information by using a couple JAVA/Tcl/Tk friendly graphical userinterfaces.

5. References

San José, R.; Prieto J.F., Castellanos N. and Arranz J.M., Sensitivity study of drydeposition fluxes in ANA air quality model over Madrid mesoscale area,Measurements and Modelling Environmental Pollution, Ed: R. San José andC. Brebbia, CM Publications, ISBN 1-85312-461-3, (1997), 119-130.

Sela J.G., The NMC Spectral Model, NOAA Tech. Rep. NWS-30, (1982), 36 pp.

GLOREAM-report

174 of 216

Statistical tools to quantify diagnostic and forecasting capabilitiesof air quality models


Eberhard SchallerLehrstuhl für Umweltmeteorologie, Brandenburgische Technische Universität

Haus 215, Burger Chaussee 2, D-03044 Cottbus, Germany

No report received

GLOREAM-report

175 of 216

Inverse Dispersion Modelling as a Tool to Derive Emission Datafrom Measurements


Petra Seibert,Gerhard Wotawa, Helga Kromp-Kolb

Institute of Meteorology and Physics, University of Agricultural Sciences Vienna,Türkenschanzstr. 18, A-1180 Wien, Austria

1. Summary

The inverse modelling technique has been explored in a first approach usingtrajectories instead of a full dispersion model to derive a source-receptor matrix forsulfur emissions in Europe, based on a one-year data set from the EMEPmonitoring network. A simple regularisation constraint without detailed a-prioriknowledge was found to be sufficient to recover qualitatively the major sulfuremission areas in Europe. This method can be used as an alternative to trajectorystatistics for source determination.

Some preliminary work has been done for the inverse modelling of the Chernobylaccident source term.

2. Aim of research

This contribution aims at the development of inverse modelling methods to deriveinformation on emissions from measurements in the regional scale. Such methodsshall be applied to suitable data sets and results be compared with conventionalemission estimates. In addition, recommendations on the optimum monitoringnetwork design shall be made.


The preliminary work on using trajectories as a substitute for a full dispersionmodel to derive source-receptor relationships and the development ofregularisation techniques for such data sets has been continued and completed forthe time being.

In preparation for a test of the inverse modelling approach with the Chernobylaccident, the ATMES data set of radiological measurements (Klug et al., 1992) hasbeen acquired from JRC Ispra, and the respective gridded precipitation analysisfrom KNMI. Furthermore, ECMWF reanalysis data (Gibson et al., 1997) for the

GLOREAM-report

176 of 216

period of the Chernobyl accident have been extracted from ECMWF’s MARS database in the format required for the planned application of the Lagrangian particledispersion model FLEXPART (Stohl et al., 1998).

The principal investigator participated in the "Workshop on Inverse Methods inGlobal Biogeochemical Cycles" from 16 to 20 March 1998 in Iraklio, Greece. Shepresented a poster on inverse modelling of sulfur emissions in Europe based ontrajectories. This paper has been submitted for publication in an AGU GeophysicalMonograph volume with results of this workshop (Seibert, 1999).

From 2 to 4 November 1998, the principal investigator participated in the annualworkshop of the EUROTRAC-2 subproject GENEMIS in Budapest, and gave anintroductory presentation on inverse modelling.


An inverse modelling study based on trajectories and daily measurements ofairborne sulfate at 13 background stations in Europe during one year has beencarried out. A source-receptor relationship matrix was calculated from thetrajectories which were utilised as a first approximation to a full dispersion model.The matrix was inverted with simple constraints for regularisation (Neumaier,1998). No detailed a-priori information was used, and instead the total length ofthe solution (emission vector) as well as its “roughness” (represented by theLaplacian of the emission field, weighted with a function depending on thetrajectory density) was minimised. This worked well enough to reproduce majoremission areas in Europe (black triangle, England, Northern Italy, Kola peninsula)in a qualitative way. However, the dynamic range of the emission values wasseverely underestimated. This method can serve as an alternative to conventionaltrajectory statistics (Seibert et al., 1994; Seibert and Jost, 1994; Stohl, 1996) aswell as a first step towards inverse modelling based on full dispersion models.


The FLEXPART model shall be applied to the ATMES Chernobyl data set toproduce a source-receptor matrix, where the source term is divided into time andrelease-height slots, to invert this matrix and apply it to the radioactivitymonitoring data in order to retrieve the source strengths which will be comparedwith conventional estimates. We shall also see whether we can use theFLEXPART model in backward (adjoint) mode.

The IMPO Lagrangian box chemistry and transport model (Wotawa et al., 1998),which is already designed for backward (receptor-oriented) mode shall be adaptedfor the production of source-receptor matrices, especially with respect to sulfurcompounds. Then it shall be used for inverse modelling of the sulfur emissions inEurope and possible other data sets. Alternatively, the FLEXPART model could be

GLOREAM-report

177 of 216

used in backward mode, however, this would mean that chemistry can not beconsidered.

6. Acknowledgements

The funding of the Austrian “Fonds zur Förderung der wissenschaftlichenForschung (FWF)” under grant P1295-GEO is gratefully acknowledged. We thankECMWF, ZAMG, EMEP MSC-W, FMI and A. Stohl (University of Munich) formeteorological and air quality data as well as trajectories.

7. References

Gibson, J.K., P. Kallberg, S. Uppsala, A. Hernandez, A. Nomura and E. Serrano,ERA Description. ERA Project Report Series, No. 1. European Centre forMedium-Range Weather Forecasts, Reading, UK, (1997).

Klug, W., G. Graziani, G. Grippa, D. Pierce and C. Tassone (eds.), Evaluation oflong range atmospheric transport models using environmental radioactivitydata from the Chernobyl accident. The ATMES report, Elsevier, London,(1992), 366 pp.

Neumaier, A., Solving ill-conditioned and singular linear systems: A tutorial onregularization, SIAM Review 40, (1998), 636-666.

Seibert, P. and D.T. Jost, Investigation of potential source areas by statisticaltrajectory analysis of ALPTRAC aerosol measurements, EUROTRAC-Newsletter 14, (1994), 14-17.

Seibert, P., H. Kromp-Kolb, U. Baltensperger, D.T. Jost, M. Schwikowski, A.Kasper and H. Puxbaum, Trajectory Analysis of Aerosol Measurements atHigh Alpine Sites, in: P.M. Borrell, P. Borrell, T. Cvitas, W. Seiler (eds.),Proc. EUROTRAC Symposium '94, SPB Academic Publishing bv, TheHague, (1994), 689-693.

Seibert, P., Inverse modelling of sulfur emissions in Europe based on trajectories,Submitted for Inverse Methods in Global Biogeochemical Cycles (AGUGeophysical Monograph, eds. P. Kasibhatla, M. Heimann, D. Hartley),(1999)

Stohl, A., Trajectory statistics - A new method to establish source-receptorrelationships of air pollutants and its application to the transport ofparticulate sulphate in Europe, Atmos. Environ. 30, (1996), 579-587.

Stohl, A., M. Hittenberger and G. Wotawa, Validation of the Lagrangian particlemodel FLEXPART against large-scale tracer experiment data, Atmos.Environ. 32, (1998), 4225-4264.

Wotawa, G., A. Stohl and B. Neininger, The urban plume of Vienna: Comparisonbetween aircraft measurements and photochemical model results, Atmos.Environ. 32, (1998), 2479-2489.

GLOREAM-report

178 of 216

Geographic Information System (GIS) Methods in Land UseMapping for Air Pollution Models


Gerhard Smiatek,Rainer Steinbrecher and Thomas Schoenemeyer

Fraunhofer-Institute for Atmospheric Environmental Research (IFU)Kreuzeckbahnstr. 19, 82467 Garmisch-Partenkirchen, Germany

1. Summary

Land use, topographic information and elevation data are very important inputparameters when modeling transport of air pollutants and biogenic VOCemissions. The available data, however, are often very heterogeneous. Theemployed map scales and classification schemes differ depending on the scope ofdata acquisition. To suit the needs of the modeling community these data have tobe reclassified, resampled and converted to formats required by the models. Forthis purpose a set of methods has been developed. They employ the capabilities aof the Geographical Information System (GIS) and Relational DatabaseManagement System (RDBMS) technology. There are two methods for compilingland use data available: (1) general categories and (2) more specific categoriessuch as plant species. The plant species-specific data are used for modelingbiogenic emissions.

2. Aim of the Research

The aim of the research is the development of GIS/RDBMS-based methods forcompilation land use and plant species-specific data for use with meteorology -chemistry models.


Two methods for compiling land use data at general categories and plant species-specific categories have been developed. They employ the capabilities a of theGeographical Information System (GIS) and Relational Database ManagementSystem (RDBMS) technology. A plant species-specific database has beencompiled for Germany and Europe.

GLOREAM-report

179 of 216


4.1 Land use in general categories

The GIS-based procedure for land use data compilation consists of severalmethods. They allow the compilation of land use data in geographical projectionfor a user defined area and for a set of different nomenclatures. Proportions andmajority in user specified grid cell can be calculated. The data description includesdata sources, formats, nomenclatures, status information and others and isavailable in the HTML-format.

4.2 Species-specific data

The principal goal here is to provide data on the area of the emitting plant species(i.e. trees) for every cell of a defined modeling area. The developed GIS-basedprocedure employs two levels (s. Fig. 1). At the first level the land use data givenaccording to general categories is used. These data allow a rough estimate of thebiogenic emission for those categories or regions where plant species-specific dataare not available.

At the second level the available plant species data are processed. These plantspecies-specific data replace appropriate parts of the data set created at the firstlevel. In addition, foliar biomass density data and emission factors for a set ofspecified chemical compounds is linked to both the land use and the plant speciesdata. Then appropriate output files are written and passed to the air quality model.Land use data in general categories is usually available at higher resolution thenthe grid size employed by air quality models. Therefore, the data can be aggregatedinto the grid cells of the model area. The geometrical resolution of the species datais much coarser. The available data pertains to regions, such as forest districts orother administrative units. Thus, the species specific data have to be disaggregatedinto the model grid cells. The aggregation process of the land use is simple andaccurate. The area of the individual land use categories are summarized for eachgrid cell. In the disaggregation procedure a grid coverage representation of themodel area is created for a specified modeling area described by a map projection,size of the grid cell, number of rows and columns. The user can display metainformation on data sets available for that modeling area and choose one or moredata sets for the disaggregation procedure. Then the geometrical data of the chosendata sets are projected to the specified map projection and intersected with the gridcoverage. A spatial selection process creates a table with the information on thearea of the different territorial units within each grid cell. The identificationnumber of the region relates via a relation function of the GIS to the species datain the relational database. Thus, all necessary information for calculation of theweighting factors for each region and species and the disaggregation is available inonly one virtual table. Processing this table yields model specific data for furtheruse within the meteorology-chemistry model.

GLOREAM-report

180 of 216

Figure 1 Information flow in processing land use data for modelling biogenicemissions (source: Smiatek and Stockwell, 1998).

5. Main conclusions

The developed GIS/RDBMS methods can manage and provide data for airpollution models and for modeling biogenic emissions. The major advantages arethe exact replication of the modeling area and the capability to provide land usedata in general categories and plant species-specific for user specified resolutionsand projections.

6. Aim for coming year

The major aim of the further work is the improvement of the plant speciesdatabase, especially for forests. This task is very difficult due to the heterogeneousstructures in ownership and administration found in Germany and Europe. It isalso planned to optimize the disaggregation procedure by rewriting parts of theroutines written in ARC.INFO AML in Perl.

GLOREAM-report

181 of 216

7. References

Smiatek G. and W.R. Stockwell; Application of A Geographical InformationSystem (GIS) to Land Use Mapping for Biogenic Emission Modeling inGermany and Europe. Proc. of the A&WMA Conference: EmissionInventory: Living in a Global Environment, Dec. 8-10, New Orleans, (1998),in print.

GLOREAM-report

182 of 216

Effect of Improved Chemical Mechanisms on the Predictions ofRegional and Global Atmospheric Chemistry Models


William R. Stockwell,Darko Koracin and Daniel L. Freeman

Desert Research Institute, Division of Atmospheric Sciences,2215 Raggio Parkway, Reno, Nevada 89512-1095, United States

1. Summary

The rate of conversion of NOx to HNO3 is important on all spatial scales becausethe concentration of NOx available along with CO and organic compoundsdetermines the rate of ozone formation. We have used the Regional AtmosphericChemistry Mechanism (RACM; Stockwell et al., 1997) to estimate the conversionrates for over-water and over-land conditions. We found that most of the NOx isconverted to HNO3 and deposited. Typical peak conversion rates were estimated tobe near 30% per hour but after 5 hours almost 70% of NOx may remain in the gas-phase. We found that the ozone concentration as a function of time can be higherover the water than over the land because the deposition velocities of NOx andozone are greater over land than over water.


The aim of this research is to evaluate the effect of improved chemicalmechanisms on the predictions of air quality models. This requires simulation ofrealistic scenarios. Deposition is an important loss process for many atmosphericspecies and to make our simulations more realistic we have implementeddeposition in our box model with the RACM mechanism. To explore theimportance of disposition on atmospheric chemistry we created over-land andover-water scenarios based on typical conditions for the San Diego Harbor area.We used these cases to simulate ozone formation and NOx to HNO3 conversionand compared the simulation results.


Our box model system (Seefeld and Stockwell, 1999) was used for the gas-phasechemistry calculations. This box model is based on the Regional AtmosphericChemistry Mechanism. This is a standard atmospheric chemistry mechanism thatis widely used for the modeling of ozone and other air pollutants. Because of theimportance of the deposition losses of nitrogeneous compounds it was added to the

GLOREAM-report

183 of 216

model. Deposition was implemented into the box model as a first order lossfollowing Equation (1)

kd = Vd / z (1)

In Equation (1) kd is the deposition rate parameter, Vd is the deposition velocityand z is the mixing height. The deposition velocities for ozone, NO, NO2, PANs,HNO3, HONO, N2O5, organic nitrates, SO2, SO4

=, H2O2 and organic peroxideswere taken from Hertel et al. (1994). The deposition velocities and calculateddeposition rate parameters were calculated for a mixing height of 300 m.

The base scenario’s initial concentrations were based on Kuhn et al. (1998) butadapted for San Diego coastline conditions. The mixing heights, relative humidityand temperature were also typical conditions for summertime.

Our base simulations were made for a one day period that started at 0300 LST ofthe first day. Two additional simulations were made, one that started at midnightand a second that started at 0600 LST. The photolysis rate coefficients werecalculated for June 21 according to Madronich (1987) with an actinic fluxcomputed by a radiative transfer model which is based on the delta-Eddingtontechnique (Joseph et al., 1976).


The peak value of the rates of NOx conversion to unreactive HNO3 is near 30% perhour. The fraction of NOx remaining in the gas-phase is a highly non-linearfunction of time. After 5 hours almost 70% remains in the gas-phase but after10 hours the fraction is only 21.2%. Most of the NOx is converted to HNO3 anddeposited as shown in Figure 1.

Ozone concentrations as a function of time can be higher over the water than overthe land. The production of ozone is a strong function of both the NOx and organiccompounds for both over-land and over-water cases. Figure 2 shows theconcentrations of ozone for the over-land case while Figure 3 shows theconcentrations of ozone for the over-water case. Comparison of Figures 2 and 3shows that the ozone concentrations as a function of time are higher over-waterthan over-land. The deposition velocities of NOx and ozone are greater over-landthan over-water and this is the cause of the greater ozone concentrations over-water. For NO the ratio of its deposition velocity over-land to its depositionvelocity over-water was 3; for NO2 the ratio was 6.8; for ozone the ratio was 8.7and for HNO3 the ratio was 0.61.

GLOREAM-report

184 of 216

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

0.00

0.25

0.50

0.75

Nitr

ogen

Bal

ance

(pp

b)

6 12 18 24

Time (hrs)

��

TPAN Dep

��

PAN Dep

��ONIT Dep

��NO Dep

��

NO2 Dep

��

2 N2O5 Dep

��HONO Dep

��HNO3 Dep

��

TPAN

��

PAN

��ONIT

��OLNN

��

OLND

��

NO3

NO2

��NO

��

2 N2O5

��

HONO

��HNO4

��HNO3

Figure 1 Nitrogen balance of species deposited for over-water base case simulation.The time is local solar time (LST). Organic nitrogen species names are inRACM format.

0

10

20

30

40

50

60

Ozo

ne (

ppb)

0 6 12 18 24 30Time (hrs)

High VOC, High NO xHigh VOC, Base NOx

Base VOC, High NOxBase

Figure 2 The concentration of ozone produced over-land is plotted. The time is localsolar time (LST). The amount of ozone formed from NOx is a function of thebackground VOC and time.

GLOREAM-report

185 of 216

0

10

20

30

40

50

60

Ozo

ne (

ppb)

0 6 12 18 24 30Time (hrs)

High VOC, High NOxHigh VOC, Base NOx

Base VOC, High NOxBase

Figure 3 The concentration of ozone produced over-water is plotted. The time is localsolar time (LST). The amount of ozone formed from NOx is a function of thebackground VOC and time.


We plan to generalize our box model to include a more realistic treatment ofmeteorological processes along with extensive updates to the RACM chemicalmechanism and to determine the effects of these improvements on modelpredictions. The updates to the RACM chemical mechanism will be made withinthe EUROTRAC2 subproject CMD.

6. Acknowledgements

The authors thank PG&E Generating for their support of this work.

7. References

Hertel, O., Christensen, J., and Hov, Ø., Modelling of the end products of thechemical decomposition of DMS in the marine boundary layer, Atmos.Environ. 28, (1994), 2431-2449.

Joseph, J.H. and Wiscombe, W.J., The delta-Eddington approximation forradiative flux transfer, Atmos. Sci. 33, (1976), 2452-2459.


GLOREAM-report

186 of 216

Madronich, S., Photodissociation in the atmosphere: 1. Actinic flux and the effectson ground reflections and clouds, J. Geophys. Res. 92, (1987), 9740-9752.

Seefeld, S. and W.R. Stockwell, First-Order Sensitivity Analysis of Models withTime Dependent Parameters: An Application to PAN and Ozone, Atmos.Environ., 33, (1999), 2883-3086.

Stockwell, W.R., F. Kirchner, M. Kuhn, and S. Seefeld, A new mechanism forregional atmospheric chemistry modeling, J. Geophys. Res. 102, (1997),25847-25879.

Resulting Publication List - William R. Stockwell

Bottenheim, J.W., A. Guenther, P.B. Shepson, R. Steinbrecher and W.R.Stockwell, Special section: biogenic hydrocarbons in the atmosphericboundary layer, preface, J. Geophys. Res. 103, (1998), 25463-25465.

Gross, A., K.V. Mikkelsen and W.R. Stockwell, A phase-space method for bi-molecular gas-phase reactions: application to the CH3CHO + HO andCH3OOH + HO reactions, J. Phys. Chem., (1998), submitted.


Seefeld, S. and W.R. Stockwell, First-Order Sensitivity Analysis of Models withTime Dependent Parameters: An Application to PAN and Ozone, Atmos.Environ., 33, (1999), 2883-3086.

Stockwell, W.R., H. Geiger and K. H. Becker, Estimation of incrementalreactivities for multiple day scenarios: an application to ethane, acetone anddimethyoxymethane Atmos. Environ., (1998), submitted.

Villenave, E., R. Lesclaux, S. Seefeld and W.R. Stockwell, Kinetics andatmospheric implications of peroxy radical cross reactions involvingCH3C(O)O2 radical, J. Geophys. Res. 103, (1998), 25273-25285.

GLOREAM-report

187 of 216

Bulgarian Modelling: Further Development of PC-oriented AirPollution Modelling for the Region of South-Eastern Europe -

Model Improvement and Applications


Dimiter SyrakovNational Institute of Meteorology and Hydrology (NIMH)

66 Tzarigradsko shaussee, Sofia 1484, Bulgaria

1. Summary

Two problems related with air pollution modelling are discussed in this article.The first one is adequate simulation of the transport part of the semi-empirical airdispersion equation; the second one being development of a Surface Layer (SL)parameterization to be applied for dry deposition modelling in case of tracer’s re-emission.

The most exploited advection scheme is Bott’s one (Bott, 1989). It is explicit,positively definite, conservative, and possesses limited numerical dispersion andgood transportivity. Integrated flux approach is used when calculating mass fluxesat the cell borders. Here a new approach called TRAP is proposed decreasingcomputations to a great extent. The flux area is supposed trapezoidal and the fluxis determined as a product of Courant number and a single value of theapproximating polynomial, taken in the middle of the passed distance. If the same4th order polynomial as in Bott’s scheme is used and the Bott's normalization isapplied the TRAP-scheme turns out to be several times faster displaying the sameproperties. Some faster variants of Bott and TRAP schemes are elaborated usinginterpolation polynomials of smaller order. They show almost the same quality oftransport description as the Bott and TRAP schemes. Another important advantageof the new versions is the fact that they need only two grid points at the borders ofthe model domain as boundary ones.

Presently, the resistance approach for describing dry deposition processes islargely used in air pollution numerical modelling. The dry deposition velocity Vd ispresented as reciprocal sum of 3 resistances. One of these resistances, theaerodynamical one, makes Vd to depend on height z. Here, this term is extractedfrom the scheme and a proper parameterization of the diffusion processes in thesurface layer is elaborated accounting for surface source action. As a result twotasks are solved: exact boundary condition to the vertical diffusion equation isobtained and a good estimate for the roughness level concentration is produced.The parameterization is tested under various conditions, both for single surface

GLOREAM-report

188 of 216

and high source and for a combination of surface+high sources. The tests confirmits good quality. Concentration profiles and dry deposed mass are describedadequately by a grid with practically no levels in SL.


The common tendency in air pollution modelling is the development of more andmore complicated models oriented to high performance computers and computernetworks. Usually, such models are conjugated with proper meteorological modelsystems producing wind and turbulence fields with high horizontal, vertical andtime resolution. Such modelling is not possible for many East-European countries,so far, in spite these countries also need dispersion modelling for variousapplications. The aim of this contribution to GLOREAM is to develop models andmodel systems oriented to small computational platforms (mainly PC), working inconditions of lack of precise meteorological input. The work is directed toimprovement of the Eulerian multi-level dispersion model EMAP, created inNIMH (Syrakov, 1997b). It is PC-oriented and relatively simple physical andmathematical approaches are searched for as to model the various air dispersionprocesses. In order to perform long-term integration, special attention is paid to thetime and memory optimization of the different calculation schemes.

The EMAP improvement is planned in the next points:

− improvement of the advection scheme;− development of surface layer parameterization, accounting for surface sources

(cases of re-emission, area sources etc.);− incorporation of proper photo-chemistry block for simulation of smaller scale

and urban pollution events;− validation of aerosol parameterization against measured data.


The activities of Bulgarian group during 1998 are connected with two of the abovementioned topics; namely:

a) Development of new advection schemes, andb) Elaboration of proper SL parameterization to be used in dry deposition and

surface concentration description.

The results are presented on some international forums as:

− EMEP contributions “in-kind” meeting, 17-18 February, 1998, Moscow,Russia;

− EUROTRAC Symposium’98, 23-27 March, Garmisch-Partenkirchen,Germany;

GLOREAM-report

189 of 216

− 5th International Conference on Harmonization within AtmosphericDispersion Modelling for Regulatory Purposes, 18-21 May 1998, Rhodes,Greece;

− NATO Advanced Research Workshop on Large Scale Computations in AirPollution Modelling, 6-10 July 1998, Sofia, Bistritza, Bulgaria;

− 23rd NATO/CCMS International Technical Meeting on Air PollutionModelling and its Application, 28 September - 2 October 1998, Varna,Riviera, Bulgaria.


A. Some new advection schemes are created on the base of TRAP approach(Syrakov and Galperin, 1998a, b, c) for estimating the mass fluxes at the edges of agrid cell using different approximating polynomials and fitting pattern. The Bottscheme and these new created schemes are put to the two-dimensional rotationaltest (Smolarkiewicz,1982). Two other well known schemes, the one ofMacCormack, Holmgren's variant (Holmgren, 1994), and the slope scheme(Roussel and Lerner, 1981), are added to the tests. The tested schemes are:

1. McC MacCormack scheme;2. R&L Roussel and Lerner’s SLOPE scheme;3. Bot Bott scheme;4. Tr4 TRAP scheme with 4 order Lagrange polynomial;5. IFB Integrated flux with 3 order Bessel polynomial;6. TrB TRAP scheme with 3 order Bessel polynomial;7. IF3 Integrated flux with 3 order Lagrange polynomial, shifted pattern;8. Tr3 TRAP scheme with 3 order Lagrange polynomial, shifted pattern;9. IF2 Integrated flux with 2 order Lagrange polynomial, shifted pattern;10. Tr2 TRAP scheme with 2 order Lagrange polynomial, shifted pattern.

A grid field of 101×101 points with ∆x = ∆y = 1 is the test domain. A rotationalwind field with constant angular velocity of ω≈0.1 (628 time steps per 1 rotation)and center in point (51,51) is imposed on this area. Two types of sources: point-shaped one (a limited δ-function) and cone-shaped one - with maximumconcentration of Co

max = 3.87∆x at point (76,51) and cone base of 15∆x, aresupposed. A number of criteria (not described here) are established for estimationof the schemes properties. The quantitative results of all schemes are presented onTable 1. The normalisation of the time of integration is made by the time of TrB .

GLOREAM-report

190 of 216

Table 1 Estimates [%] of different schemes’ performance.

1 rot. McC R&L Bot Tr4 IFB TrB IF3 Tr3 IF2 Tr2

p Cmax 6.40 5.79 3.78 3.81 3.48 3.53 3.48 3.53 2.40 2.40

o Cmin -1.88 -.34 .00 .00 .00 .00 .00 .00 .00 .00

i ΣΣΣΣM .00 .00 .00 .00 .00 .00 .00 .00 .00 .00

n DXc .0 -.2 23.4 22.4 39.5 39.3 39.5 39.3 21.7 21.4

t DYc .1 .1 -30.8 -26.6 -81.7 -78.3 -81.7 -78.3 -28.9 -25.6

T 155 2276 766 110 110 100 110 109 89 89

Cmax 91.56 90.67 91.16 91.19 91.04 91.09 91.04 91.09 87.76 87.75

c Cmin -1.45 -.61 .00 .00 .00 .00 .00 .00 .00 .00

o ΣΣΣΣM .000 .001 .000 .000 -.001 .000 .000 .000 .000 .000

n DXc .0 -.1 .0 .0 .0 .0 .0 .0 .0 -.1

e DYc .1 .1 -.2 .0 -.6 -.4 -.6 -.4 -.3 -.1

T 114 1361 659 120 106 100 109 108 89 90

It can be noticed that none of the tested schemes describes satisfactorily theadvection of a single disturbance in the concentration field (point source). There isno difference scheme, approximating the advection equation, capable to describeadequately this severe discontinuity in the concentration field. This can be seen inCmax values, which fall down to some percents of the initial maximum. The steepgradients are squashed down immediately after the start of movement and afterthat the new nearly-Gauss-shaped distribution is advected. The cone experimentsshow much better results.

B. Let us have a vertical grid with z0 and z1 as bottom boundary and firstcomputational levels. Let h be a height between them and r0 be the roughnessheight; C0, C1, Ch and Cro being concentrations at these heights, respectively. Thedry deposition boundary condition to the vertical diffusion equation can bepresented as:

C B C B QB K V b c

B cc K V ad

do oo

= += −=

= +1 11,

( ) /

/,

γγ

γ

and the roughness level concentration can be calculated as

C G C G QG aB b

G aB Vrd

o oo o

= += += + +

1 11 1

1,

( )

( ) / ( ),

γα α

where Q is the surface source strength [M/L2S] and

γ α α= +/ ( )Vd , α κ ςθ= u f h* / ( ) , K k h z zc= −( ) / ( )1 o ,

GLOREAM-report

191 of 216

af f

f fb

f f

f fh h= −

−= −

−θ θ

θ θ

θ θ

θ θ

ς ςς ς

ς ςς ς

( ) ( )

( ) ( ),

( ) ( )

( ) ( )1

1 1o

o

o ,

κ = 0.4 being Karman constant, Vd - roughness level dry deposition velocity, fθ(ζ) -universal SL function, ζ = z/L, L - Monin-Obukhov length, kc(z) - vertical turbulentexchange coefficient.

In Fig. 1, examples of the calculated profiles evolution are presented,demonstrating the ability of the proposed parameterization to account for theaction of surface source. A 4-level grid (computational level heights50,200,650,1450m, boundary levels at 20 and 2500 m) is used, the lowest innerlevel (z1) placed at the top of SL.

a. b. c.

0 20 40 60 80 100C oncen tration [units/m ]

0.1

1

1E +1

1E +2

1E +3

Hei

ght [

m]

T im e s teps

N =0

N =1 0

N =3 0

N =1 00

N =3 00

N =1 000

0 20 40 60 80C oncentration [units/m ]

0.1

1

1E +1

1E +2

1E +3

Hei

ght [

m]

T im e s teps

N = 10

N = 30

N = 100

N = 300

N = 1000

0 20 40 60 80C oncentration [units/m ]

0.1

1

1E +1

1E +2

1E +3

Hei

ght [

m]

T im e s teps

N = 10

N = 30

N = 100

N = 300

N = 1000

Figure 1 Concentration profiles for the cases of high source (a), surface source(b) andhigh + surface sources (c).

5. Main conclusions

A. The version TrB of the TRAP scheme shows almost the same characteristics asthe original Bott one, but it is some times faster. Another important advantage ofthis low order scheme is the fact that it needs only two boundary points at theborders of the model domain.

B. The tests of the proposed SL parameterization verify its good quality.Concentration profiles and dry deposed mass are described adequately by a grid,which practically has no levels in the surface layer. The parameterization iseffective for a great variety of situations, for airborne, surface and combinedsources. The parameterization is built in a simple re-emission and degradationscheme (Syrakov, 1998). The performed sensitivity tests show a behaviour whichis very reasonable from physical point of view.

GLOREAM-report

192 of 216

6. Acknowledgements

Deep gratitude is due to the EUROTRAC Scientific Secretariat for its efforts tosupport the participation of some Bulgarian scientists in EUROTRACSymposium’98.

7. References

Bott, A., A positive definite advection scheme obtained by nonlinearrenormalization of the advective fluxes, Mon. Wea. Rev. 117, (1989), 1006-1015.

Holmgren, P., An Advection Algorithm and an Atmospheric Airflow Application,J. Comp. Phys. 115, (1994), 27-42.

Russell, G.L. and J.A. Lerner, A New Finite-Differencing Scheme for the TracerTransport Equation, J. Appl. Meteor. 20, (1981), 1483-1498.

Smolarkiewicz, P.K., The multidimensional Crowley advection scheme, Mon.Wea. Rev. 113, (1982), 1109-1130.

Syrakov, D., On the TRAP advection scheme - description, tests and applications,in: G. Geernaert., A. Walloe-Hansen and Z. Zlatev (eds), Regional Modelingof Air Pollution in Europe, Proc. 1st REMAPE Workshop, Copenhagen,Danmark, September 1996, National Environmental Research Institute,Danmark, (1997a), 141-152.

Syrakov, D., A PC-oriented multi-level Eulerian dispersion model - modeldescription, Bulgarian Journal of Meteorology and Hydrology 8, (1997b),41-49.

Syrakov, D., Influence of re-emission on pollution distribution: one dimensionalmulti-level model, in: Bulgarian contribution to EMEP. Annual report for1997, NIMH - EMEP/MSC-E, Sofia-Moscow, January, (1998), 21-25.

Syrakov, D. and M. Galperin, On Some Effective Bott-type Advective Schemes,in: P.M. Borrell and P. Borrell (eds), Proc. EUROTRAC Symposium’98, 23-27 March, Garmisch Partenkirchen, Germany, WITpress, Southampton, UK,(1998a), 8-26(1-7).

Syrakov, D. and M. Galperin, On some explicit advection schemes for dispersionmodelling applications, Proc. 5th International Conference onHarmonization within Atmospheric Dispersion Modelling for RegulatoryPurposes, 18-21 May 1998, Rhodes, Greece, (1998b), 236-243.

Syrakov, D. and M. Galperin, On some flux-type advection schemes for dispersionmodelling application, in: Z. Zlatev (eds), Large Scale Computations in AirPollution Modelling, Proc. NATO Advanced Research Workshop, 6-10 July1998, Sofia, Bistritza, Bulgaria, (1998c), (to be published).

GLOREAM-report

193 of 216

Development of a Data Assimilation Scheme for ChemicalConstituents Suitable for Operational Application


Stefan Tilmes,Jörg Zimmermann, and Jürgen Rißmann

Deutscher WetterdienstKaiserleistr. 42, D-63067 Offenbach, Germany

1. Summary

Uncertainties and the representativeness of measurements must be well knownbefore the data can be used in a sound way. This is in particular true for the use oftrace gas measurements for model evaluation or data assimilation. Different tometeorological quantities the concentrations of chemical species to a large degreedepend on small scale fluctuations and the measurements suffer from largeobservation errors. The representativeness of such data is not only due to thespatial distance. It is also determined by the characteristics of the air mass in termsof the chemical age or the closeness to emission sources. In this sense we areengaged in the development of methods for the quantification of therepresentativeness of air quality data.

2. Aim of research

Together with the contributions of Jürgen Rißmann and Jörg Zimmermann the aimof our research is the development of a model system suitable for operationalchemistry-transport simulations. Time critical data assimilation and onlineverification of the results have to be established to reveal the reliability of the airquality forecasts. As denoted above, in advance investigations on the availablemeasurements are required.

Up to now we concentrate on ground level ozone observations. These are the datawith the best availability and spatial and temporal coverage. Time series of morethan 300 German sites for the years 1994 and 1995 are under investigation. Thepresent contribution deals with the assignment of well defined fractions of thevariance in the data to the spatial scales which they are representative for. Thequantitative description of the chemical regime of the ozone data will be touchedbriefly.

GLOREAM-report

194 of 216

3. Recent activities

Apart from very rural areas which are quite rare in Europe the course of the ozoneconcentration is dominated by a photochemically induced diurnal variation. Theshape, phase, and amplitude of this diurnal variation mainly depends on theavailability of NOx, the distance to the emission sources, and the meteorologicalconditions as well.

We examined these features by performing a principal component analysis (PCA)on the basis of diurnal variations of the available ozone data. The PCA is analgebraic tool for the investigation of the variance structure within the underlyingdata. Empirical orthogonal functions (EOFs or the eigenvectors of the system) aredetermined. The corresponding eigenvalues give the amount of variance in thecomplete dataset which is due to the single EOFs. For example the first EOF fromthe analysis of the ozone data has the shape of the typical diurnal variation and isresponsible for more than 70% of the variance. The following 5 - 7 EOFs representanother 25% of the variance. The trailing eigenvectors only contribute less to thevariance and are mainly due to short term fluctuations.

Now the original data can be expressed in terms of the EOFs. This gives timeseries for each observation site and for each eigenvector consisting of oneprincipal component (or amplitude) to the corresponding EOF per day. In a furtherstep correlations between these new time series are calculated for each componentseparately. From this information about the spatial influence andrepresentativeness of the EOFs will be obtained.

Figure 1 Left: correlation coefficients for the measured ozone time series of about 300German sites for the years 1994 and 1995 against the distance between thesites; right: relative accumulated variance as explained by the denotednumber of EOFs against the radii of influence/representativeness of theensemble of EOFs.

GLOREAM-report

195 of 216

The second aspect in association with the ozone data is the question of quantifyingthe current regime of an air mass. The main influence may be expressed in classesranging form “close to traffic” to “remote”. The criteria for the classification arebased on the individual shape of the diurnal variations. Time series which areunder a strong influence of fresh NOx emissions show a large amplitude of thetypical diurnal variation compared to the diurnal mean concentration. This resultsin close-to-zero concentrations during the nighttime hours when all the ozone istitrated to NO2. On the other hand a “remote” regime (or the absence ofmeteorological conditions which favour photochemistry) manifests itself by theabsence of marked diurnal variations and low concentration values as well.

We follow two ways to describe these properties in a more quantitative way withjust the ozone data at hand. First, the shape of the frequency distribution of thedata for a summer season shows the predominant or climatological regime of thesite. A symmetric distribution without low concentrations is typical for a remotesite. On the other hand nonsymmetric distributions with the maximum of thefrequency at low values clearly indicate the closeness to the emission sources.Second, with the results of the PCA it is possible to quantify the prevailing regimeof a single diurnal time series. As explained above, the principal component to thefirst EOF describes the amplitude of the diurnal variation in the data. Togetherwith the diurnal mean value a measure for the regime can be derived. Bothmethods have the advantage that they are also applicable for simulated ozone data.It is just necessary to evaluate and calibrate them by means of long simulatedepisodes to achieve statistical significance. Indeed the methods are not valid forstations where the pollution reaches levels too high to allow for the formation ofozone.


The left panel of Fig. 1 gives an estimate of the overall measurement error. Shownare the correlation coefficients between the ozone time series measured over a twoyear period (1994 and 1995) for the German sites against the spatial distancebetween the locations. Values extrapolated to vanishing spatial distances of about0.93 - 0.95 denote a fraction of 10 - 14% of the variance which is not spatiallycorrelated and thus due to the uncertainties of the measurements.

The PCA of the ozone measurements beyond others achieved two main results.This is on the one hand the splitting of the variance of the data into independentparts represented by the EOFs. On the other hand this is the assignment of theEOFs to spatial scales, it comes out that the representativeness decreases withincreasing number of EOF. A combination of these results is presented on the rightof Fig. 1. The correlation distances for two threshold values of the coefficient isplotted on the x-axis and the accumulated relative variance on the y-axis. Theprincipal results from this are described in more detail in Tilmes and Zimmermann(1998). The figure shows that the first EOF, which is already responsible for more

GLOREAM-report

196 of 216

than 70% of the variance correlates on the synoptic scale. About six EOFs aresufficient to explain the course of the ozone concentrations up to the designateduncertainties of the measurements. The representativeness of these eigenvectorsranges up to about 3 - 4 km and the corresponding radius of influence is about30 km.

Figure 2 Reconstruction (thick lines) of observed (thin lines) ozone time series [ppbv]by different numbers of EOFs for different sites in Berlin, Germany, timeperiod 25. - 28. July 1994.

Fig. 2 supports these findings. Presented are the reconstructions of ozonemeasurements by different numbers of EOFs. Shown are time series for severalstations in Berlin, all within a radius of below 18 km (episode 25. - 28. July 1994).Different features of these time series are due to different EOFs. For example thediurnal maximum is explained by the first 1 - 4 EOFs while the minimumconcentration is often missed before EOF 4 or 5. Effects on an even smaller scaleare nocturnal maxima or the exact time of the day-time maximum which are often

GLOREAM-report

197 of 216

fit by EOF 6. These features differ for neighbouring stations indicating their poorspatial representativeness.

An example for the application of the measure for the ozone regime is presented inFig. 3. We investigated a photosmog episode on a Central European area in July1994 by means of the ground level ozone measurements and simulated dataobtained frorn the EURAD group. We determined the mean regime for everyobservation site and every model grid-point and plotted the frequency distributionof these values. It becomes clear that the measurement networks are only oflimited representativeness for the modelled data. The observations are dominatedby urban sites while the model regimes cover only a very narrow range between“suburban” and “rural”. The latter for example may be due to an unsufficientresolution of the emission data used for the simulation under consideration.

Figure 3 Relative frequency of the ozone regimes determined for observation sites andmodel gridpoints during a sommersmog episode in July 1994.

5. Main conclusions

With the suggested methods it is possible to assign well defined parts of ozonetime series to certain spatial scales which allows for a thorough interpretation ofmeasured ozone time series. One can estimate what part of the variance inmeasurements might be simulated by a model with a certain gridsize and resolutionof the input data. The current regime of air masses can be quantified. All thesefeatures can be applied in terms of short term or climatological investigations ofmodelled or measured data. The results are useful for the quality control ofmeasurements and model evaluation/comparison and data assimilation as well.

6. Policy relevance

The methods allow for a thorough investigation on the representativeness ofground level ozone data. This is of interest in terms of discussing observation

GLOREAM-report

198 of 216

networks or stations for regulatory purposes. On the other hand it is useful formodel comparison and evaluation and the description of the model climate. In thissense it would be nice to get into contact with the responsible people in theenvironmental protection agencies.


The next focus will lie in the investigation and evaluation of longer modelledozone time series with the presented methods. A question to be addressed iswhether the EOFs and the above presented findings are transferable to themodelled data. Procedures for model evaluation and data assimilation will bedeveloped and applied in collaboration with the other participating projects. Amodel comparison guided by these methods would be a nice exercise withinGLOREAM.

8. Acknowledgements

We are grateful to the Umweltbundesamt Berlin for kindly providing the ozonedata of its stations and the networks of the federal states, the EURAD group(University of Cologne) provided the simulated data. This work was supported bythe Bundesministerium für Bildung, Wissenschaft, Forschung und Technologie ofthe Federal Republic of Germany under grant 07TFSIO/LTI-C4.

9. References

Tilmes, S. and J. Zimmermann, Investigation on the spatial scales of the variabilityin measured nearground ozone mixing ratios, Geophys. Res. Let. 25, (1998),3827-3830.

GLOREAM-report

199 of 216

Sensitivity study to the representation of the anthropogenicemissions in global CTM models:

On the importance of a detailed chemistry of the NMHC

Contribution to subproject GLOREAM

F. Brocheton(°),B. Aumont (°), G. Toupance (°) and J.F. Müller (°°)

(°) LISA, UMR CNRS 7583 – Université Paris VII & Paris XII, Créteil (France)(°°) BISA, Bruxelles (Belgique)

1. Objective of the ongoing work

Nitrogen oxides (NOx) are the limiting precursors for the ozone productionthroughout most of the troposphere. Thus precise simulations of the global ozonebudget require an accurate simulation of the NOx distribution which is one of themost difficult to achieve in the current global models. For example, Thakur et al.(1999) showed that the distribution of reactive nitrogen oxides (NOy = NOx +organic nitrates, …) simulated by three different global models are all indisagreement with observations.

In the northern hemisphere, the main NOx sources are anthropogenic emissionswhich contribute in the range of 30 to 60% of the NOx load in the middletroposphere. This long range transport of NOx occurs mainly through secondarynitrogen species which constitute the NOx reservoirs (PANs, RONO2, …). Thesesecondary species are produced close to the NOx sources and are stronglydependent on the Volatile Organic Compounds (VOC) and NOx mixing ratios.

Due to computational limitations, global CTMs use on the one hand a coarse grid-size resolution (from 1°*1° to 10°*10°) much larger than the size of the majorurban sources (typically 0.1°*0.1°). Such grid-sizes induce a dilution effect of theemissions. This may lead to significant computational errors on the total amount ofnitrogen reservoirs and then on the impact of man-made activities on the globalatmosphere. On the other hand, CTMs also use a very simplified chemical scheme,specially for the chemistry of the higher hydrocarbons which are either representedby a single model species or not. Such simplifications on the chemistry ofanthropogenic NMHC may also lead to significant errors on the simulated flux ofreactive nitrogen species exported from the Continental Boundary Layer (CBL) tothe free troposphere.

GLOREAM-report

200 of 216

2. Results

In 1998, we have then focused first on the estimate of the influence of a detailedrepresentation of the chemistry of the anthropogenic NMHC on the simulateddistributions of ozone, nitrogen oxides and nitrogen reservoir species at globalscale.

To address this question, different simulations were performed using the IMAGESmodel, in which the NMHC chemical scheme was simplified by successive steps.So first, a reference scheme with a detailed representation of the higher NMHCwas developed. On the basis of different global (EDGAR, …) and regional(GENEMIS, NAPAP, …) emission databases, a global speciation for thesecompounds has been elaborated. Nevertheless, after this step, the number of VOCdefined is close to 50. An explicit representation of these compounds in a globalchemical scheme is not possible due to the computational limitations. Acondensation representation based on the use of model species was then applied.The method used as the model species is then as much as possible close to the onesused in regional models.

In order to quantify the capacity of the reference scheme to simulate thephotochemistry of the lower troposphere, comparisons using a lagrangian modelfor different scenarios have been performed. The global reference scheme wasthen compared to a high detailed chemical mechanism representative of the lowercontinental photochemistry. These comparisons showed that this reference schemeagrees well with the detailed chemical mechanism, specially to simulate the NOx

and PANs mixing ratios. Finally, the reference scheme includes 81 species and 252reactions. The primary VOC taken into account in this study are C2H6, C3H8, C2H4,C3H6, acetone, isoprene, terpenes, and five model species for the higher COV:Aromatic (for the aromatic compounds), alkane (≥ C4 alkanes), alkene(≥ C4 alkene) and MEK (for the higher ketones) compare to one model species(OTHC) in the previous chemical mechanism of IMAGES.

Different sensitivity tests to the chemical representation of the higher VOC havebeen performed with the IMAGES model. These tests show that it is important torepresent such compounds to correctly simulate the NOx and the nitrogen reservoirspecies mixing ratios, specially in the CBL. But, the use of a single model species(OTHC) rather than five model species, such as in the reference scheme, issufficient. For example, we can see on the table below how the tropospheric globalbudget of the ozone, NOx and especially the NOx reservoirs species are influencedby the chemical representation of the higher COVs.

GLOREAM-report

201 of 216

Tableau 2.1 Annual tropospheric budget (in Teragrams per Year) of ozone, NOx, sum of PAN compounds,sum of RONO2 compounds simulated with the IMAGES model. The simulation with one modelspecies is used as the reference simulation for the calculation of the relative difference.(Source = photochemical production + emissions; Sink = photochemical loss + depositon ).

Simulation without higherCOVs

Simulation with onemodel species

Simulation with five modelspecies

O3 Burden 1457 -0.4% 1463 1463 0%

Source 7699 -2.9% 7928 7899 -0.35%

Sink 7683 -2.9% 7913 7883 -0.36%

NOx Burden 0.577 +0.34% 0.575 0.571 -0.7%

(NO+NO2+HNO3

+Source 264 -10.6% 292 293 +0.34%

HNO4+2N2O5) Sink 271 -9.7% 300 300 0%

RONO2s Burden 0.0216 -6.9% 0.0232 0.0254 +9.5%

Source 4.77 -10.5% 5.33 5.76 +8%

Sink 5.66 -9.9% 6.28 6.64 +5.7%

PANs Burden 0.280 -23.9% 0.368 0.388 +5.5%

Source 204 -12.5% 233 233 0%

Sink 204 -12.5% 233 233 0%

The next step of this study will be then to quantify how the spatial representationof the emission in CTM influence the simulated flux of reactive nitrogen exportedfrom the CBL to the free troposphere.

GLOREAM-report

202 of 216

Investigation of Trace Gas Variability for Use in ModelEvaluations


Jörg Zimmermann and Stefan TilmesDeutscher Wetterdienst

Kaiserleistr.42, 63067 Offenbach, Germany

1. Summary

Long time series of ozone and partly of other trace gases observed in Germansurface measurement networks were analysed with the Kolmogorov-Zurbenko-filter. Seasonal and episodic components and an averaged diurnal cycle could beseparated. The seasonal component of the ozone time series was found to haverelatively high concentrations in spring. A possible explanation for these higherozone concentrations are a stronger dilution of primary trace gases in spring byclean tropospheric air and subsequent a lower destruction of ozone by NO. Atrajectory box model is improved by using up-dated emissions and an adjustmentof the mixing layer height. The improved model predicts measured ozoneconcentrations in an episode in July 23 to 27, 1994 with an error below 25% and acorrelation coefficient between 0,5 to 0,8.


The evaluation of models includes the comparison of measurements sparselycovering the model domain and of model results valid for grid volumes on ahorizontal scale of kilometers. Measurements and model results cannot becompared directly, since measurements are locally influenced and are frequently atpositions not representative for a grid volume. Investigations of long time series oftrace gas measurements reveal spatial and temporal variabilities and structures, andthereby help to identify useful statistical measures for model validation. Thereforethis year long term series of measurements of ozone, nitrogene oxides, and carbonmonoxide were analysed. The aim was to separate variability on different temporaland spatial scales. A trajectory box model is used for the investigation of trace gasvariability from the model side. The goals of this year were to achieve a morerealistic simulation of processes in the boundary layer and to improve theforecasting skills of the model.

GLOREAM-report

203 of 216


Measurements of ozone, nitrogen oxides, and carbon monoxide were analysedwith the Kolmogorov-Zurbenko-filter (KZ-filter) (Rao and Zurbenko, 1994) andpartly with principal component analyses (PCA). For this purpose ground basedmeasurements from operational networks of the environmental agencies inGermany were investigated for the years 1994 to 1996. The data were available asone-hour mean values.

The Kolmogorov-Zurbenko-filter is a 15-days running average, which is applied tothe data five times. Here the KZ-filter is applied on hourly data instead of dailymaxima and the original data are used instead of the logarithms. The KZ-filter isapplied for each hour of the day independently (e.g. for all 1 hour UT values, thenfor all 2 hour UT values etc.). Therefore the time series is separated in two parts.K(t) contains short term variations (including episodes of up to about one week).The long term part S'(t) still contains an averaged diurnal cycle. This averageddiurnal cycle T(t) is separated with a second filter with a 25-hours running averageapplied five times. The remaining long term variation S(t) now contains onlysignals on a scale of two weeks and more. No attempt was made to remove ayearly trend, because the time series is too short with only up to three years ofdata. The complete ozone time series O3(t) is separated in three terms:

O3(t) = S(t) + T(t) + K(t)

The separated time series were investigated for all stations in Germany. Statisticalparameters calculated include correlation distances, autocorrelation, correlationbetween ozone, CO, and nitrogen oxides and standard deviations for the fourseasons.

Parallel to this work a trajectory-box-model was improved for the purpose toinvestigate trace gas variability from the model side. Main changes are made to theemission data base, which was shifted from CORINAIR 1990 to emissionscalculated by the Institut für Energiewirtschaft und Rationelle Energieverwendung,University of Stuttgart (IER) for 1994.

The mixing layer height for the trajectory box model is taken from the weatherprediction models of the Deutscher Wetterdienst. In case of stable boundary layersstrong deviation between measurements of ozone and model results are observed.It was suspected that in this case the diagnosed mixing layer height might beinappropriate to measure the height of the layer of air in contact with the ground.Therefore it was attempted to include a more realistic simulation of the diurnalcycle of trace gas concentrations. For this aim corrections of the mixing layerheight were tested.

GLOREAM-report

204 of 216


The yearly cycles of the ozone time series derived from the long term componentS(t) of the KZ-filtered data show similar features in the years 1994 to 1996. Asexpected, ozone concentrations are higher in summer and lower in winter.However, the shape of the yearly cycles are asymmetric compared to the yearlycycle of radiation. A first major increase of the ozone concentration happens earlyin the year, usually in March, followed by a second increase in June. At somestations two different maxima of the ozone concentration in spring and summercan be observed. The maximum in spring is more pronounced at stations at higherelevations and at stations, which are in the northern part of Germany. The decreaseof the ozone concentration starts in September and is more rapid than the increase.Obviously only a part of the yearly cycle of ozone concentration can be explainedby photochemical processes. In the literature the well known spring maximum ofozone is discussed as a possible result of an increased influx of stratospheric air inthe troposphere in spring. Here spring maxima are observed even at sea level,where a stratospheric influence is very doubtful. However, since only three yearsare taken into account, it is still not clear, in how far this conclusion can begeneralized.

Yearly cycles of NO, NO2, and CO are anticorrelated to the ozone cycles andcorrelated with each other. However, the anticorrelation is strong in winter andweak in summer. The decrease of the primary pollutants in spring proceeds slowerthan the increase starting in September. It thereby mirrors the antisymmetry of theozone yearly cycle according to the yearly cycle of solar radiation.

Year to year correlations of the yearly cycles of the ozone concentration exceed0.8 at most stations. Correlations between stations are above 0.9 in Germany. Onaverage the correlation exceeds 0.95 for a distance up to 500 km. Thus this part ofthe ozone time series represents processes on the macro scale. However, it alsoincludes constant site specific features like position with respect to the boundarylayer and average pollution level.

The two short term components, the averaged diurnal cycle T(t) and the episodicvariations K(t) were investigated with view on station specific and season specificfeatures. Station specific features of T(t) reproduced findings of others (see thecontribution of Stefan Tilmes et al.). The averaged diurnal cycle T(t) was found tobe very specific of the station type with respect to pollution level and relativeposition in the boundary layer. Its variance also depends on the season. In summerand spring the average variance of T(t) is about five times as high as in fall andwinter. K(t) is in summer and spring only 50 percent higher than in fall and winter.That means, that the largest part of the variance of the ozone time series in springand summer is in the diurnal cycle, but in fall and winter in the episodiccomponent. Therefore in fall and winter the ozone concentration changes lessregularly and is more difficult to predict. Additionally, in summer the seasonal

GLOREAM-report

205 of 216

component S(t) constitutes a large part (about 50 percent) of the ozoneconcentration at not too polluted stations. So the persistence term can be large. Inwinter, however, the seasonal component is small compared to the short termcomponents.

The trajectory-box-model LOOP is evaluated by a comparison with measurementsof ozone at several German stations for the episode of July 23 to 27, 1994.Changing the emission data from the Corinair 1990 data base to emission datafrom IER at Uni Stuttgart for 1994 did improve the ozone values, especiallybecause of a decrease of the emission. Changes were stronger at stations in EasternGermany, because here the resolution of the data was improved tremendously.However, the correlation between modelled and measured ozone concentrationsdidn’t improve much. The model still shows an overprediction of ozone at acoastal station, indicating problems with local winds and an insufficient resolutionin the coastal area. Fig. 1 shows results from a site in Freiburg in south-westGermany. Note, that the Corinair emissions are already adjusted with a factor of0.8 to account for emission reductions from 1990 to 1994.

23.7. 24.7. 25.7. 26.7. 27.7.

O3 [ppb], Freiburg, 1994

100

50

0

observed

modelled,Corinair emissions

modelled,IER emissions

Figure 1 Comparison of observed and calculated ozone mixing ratios for Corinair1990 emissions and IER 1994 emissions. Mixing layer height is set to 30 mfor stable conditions.

Methods of adjusting the mixing layer were tested in order to increase thecorrelation of measured and modelled ozone concentrations. While it was notpossible to use the vertical eddy diffusion coefficient diagnosed in the weatherprediction model, good results were achieved by using the horizontal wind speedas an indicator for slow vertical turbulent transport. As a result the mixing layerheight was fixed to 30 m above ground for a wind speed below 3 m/s and aweather prediction model derived mixing layer height below 900 m. Thecorrelation increased at most suburban and urban stations. The improved modelpredicts measured ozone concentrations in the episode with an error below 25%

GLOREAM-report

206 of 216

and a correlation coefficient between 0,5 to 0,8. The correlation is lower atmountain sites, where the variance of the ozone signal is very low, and at a coastalsite, where local effects seem to be important, which are not resolved by themodel.

5. Main conclusions

The KZ-filter is a method suitable to identify different components of time seriesof ozone. A seasonal and an episodic component and an averaged diurnal cycle canbe separated. The seasonal component shows higher concentrations in spring andlower concentrations in fall than should be assumed from a solely photochemicalorigin of ozone. The stronger anticorrelation with NO in winter and springindicates a predomination of the variance by the titration with NO at this time. Thevariation of the ozone concentration in winter and spring is irregular to a largeamount. This indicates, that the ozone variation is caused by dynamic processeswhich dilute primary pollutants in spring time to a much larger amount than inwinter or fall. A higher amount of dilution in spring compared to fall could causemean ozone concentrations to be higher in spring than in fall and to cause thespring time ozone maximum. This dilution process can be more efficient at thecoast, where a reservoir of clean air is near to the measurement sites or inmountaineous location, where vertical motions bringing more ozone reach air fromthe upper troposphere may have a large influence. Thus the spring time ozonemaximum can be explained without considering an influence of stratospheric air.However, these conclusions are based on only three years of data and need furthercorroboration by a larger data set and a thorough comparison with meteorologicaldata. The different components of the ozone time series are connected to differentspatial scales. This can be studied in more detail by using principal componentanalysis (see the report of Stefan Tilmes et al.).

The ability of the trajectory box model to predict ozone concentrations can beimproved by up-dating emission data and by an adjustment of the mixing layerheight based on wind data.


It is planned to use the separated components of the ozone time series for acomparison with different chemistry and transport models in order to validatecertain features of the models and their spatial resolution. The trajectory modelwill be further improved and long time series of trace gases will be calculated andcompared with measurements.

7. Acknowledgements

We are grateful to the Umweltbundesamt (UBA) Berlin for kindly providing thedata of the measurement networks of the states and of UBA. This work was

GLOREAM-report

207 of 216

supported by the Bundesministerium für Bildung und Forschung of the FederalRepublic of Germany under grant 07TFS10/LT1-C4.

8. References

Rao, S.T., and I.G. Zurbenko, Detecting and tracking changes in ozone air quality,J. Air & Waste Manage. Assoc. 44, (1994), 1089-1092.

GLOREAM-report

208 of 216

Ozone in the Troposphere over Europe:A Source Segregated Analysis from a Global Point of View


Peter Zimmermann,MOGUNTIA Global Modelling

Ludwigstr. 10, D-65479 Raunheim, Germany

The tropospheric ozone budget over Europe is controlled by both physical andchemical fluxes. Basically ozone is produced by photochemical reaction indifferent ways at different altitudes of the atmosphere. In the stratospheric ozonelayer, photodissociation of molecular oxygen and recombination of the O-atomswith other O2-molecules leads to ozone (O3) formation. In the troposphere, ozoneproduction is initiated by NOx-molecules in the course of methane oxidation byOH. Consequently locations of enhanced ozone production inside the troposphereare those being exposed to NOx-emissions which can have both natural andanthropogenic sources such as:

− industrial fossil fuel combustion including car traffic and aircraft,− biomass burning,− soil exhalation,− lightning.

Anthropogenic air pollution nowadays represents a major tropospheric ozonesource and it has been growing with rising civilisation. Thus, industry and trafficare at present the most efficient producers of ozone precursors in the troposphericair space over Europe. However, because the upper and lateral boundaries arepermeable, advective and turbulent air fluxes are importing ozone from remotephotochemical sources into that domain as well as exporting locally formed ozoneout. Furthermore with an estimated tropospheric lifetime of about three months,the same ozone molecules could pass the European air volume up to several timesbefore being removed from the atmosphere by deposition or photolysis.Consequently, the ozone in the tropospheric air over Europe is composed ofdifferent contributions and can be roughly classified by the categories listed abovein addition to ‘stratospheric ozone’ which has its origin in the stratosphere and ismixed in verticlly or horizontally. ‘Industrial ozone’ can even be furthersubdivided into ‘European’ and ‘other’ contributions with respect to thegeographical location of the particular NOx-emission source which it is owing itsorigin to.

GLOREAM-report

209 of 216

The budget of the ‘European ozone composition’ and its seasonality are theproduct of a complex four dimensional interactive transport/chemistry system andits evaluation requires a global model with linearised transport parameterisation.The MOGUNTIA transport model for chemical tracers in the global troposphereprovides the feasibility to trace the path of NOx-molecules emitted from variouskinds of sources as well as the path of their chemical products individually.

Preliminary results in form of monthly mean distributions of model simulatedbackground ozone mixing ratios and their percentile composition of the classifiedcontributions over Europe were presented at the EUROTRAC2 Symposium ’98.

Ozone (July means)

Figure 1 Surface layer ozone distribution zoomed in on Europe.

Figure 2 ‘Industrial NOx - ozone’ in percent of the mixing ratios in 1.

GLOREAM-report

210 of 216

Figure 3 ‘Lightning NOx - ozone’ percentage.

Figure 4 Ozone other seasons.

GLOREAM-report

211 of 216

Figure 5 ‘Biomass burning NOx - ozone’ percentage (March).

Figure 6 ‘Stratospheric ozone’ percentage in the European troposphere.

GLOREAM-report

212 of 216

Comprehensive Air Pollution Studies by the Danish EulerianModel


Z. Zlatev,C. Ambelas Skjøth and A.A. Antonov

National Environmental Research InstituteFrederiksborgvej 399, P.O. Box 358, DK-4000 Roskilde, Denmark

1. Summary

The protection of our environment is one of the most important problems in themodern society. The importance of this problem was steadily increasing during thelast two-three decades, and the environment protection will become even moreimportant in the next century. Reliable and robust control strategies for keeping thepollution caused by harmful chemical compounds under certain safe levels have tobe developed and used in a routine way. Large mathematical models, in which allimportant physical and chemical processes are adequately described, cansuccessfully be used to solve this task. The use of one such model, the DanishEulerian Model, is discussed in this report.


The Danish Eulerian Model is described in Alexandrov et al., (1997), Bastrup-Birket al., (1997), Zlatev, (1995), Zlatev et al. (1996a, 1996b). Its space domain coversthe whole of Europe together with parts of Asia, Africa and the Atlantic Ocean.The domain is discretized by a (96x96x10) grid. This means that (50 km x 50 km)grid squares are used in the horizontal planes. A variable mesh is applied in thevertical direction. The CBM IV scheme (Gery et al., 1989, Zlatev, 1995) with 35chemical species is used in the chemical part of the model. The main objective isto use the model in studying relationships between emission sources and highlevels of concentrations and/or depositions of certain chemical species in differentparts of Europe (and first and foremost in Denmark and in the region surroundingDenmark). The model is used to study both episodic and long-term variations ofthe concentrations and depositions. It should be mentioned here that the two-dimensional version of the model is still used when long-term variations (over atime-period of up to ten years) are studied.

GLOREAM-report

213 of 216


The main activities in the work on the project “Comprehensive Air PollutionStudies by the Danish Eulerian Model” were concentrated in the following sixdirections during 1998:

(a) Improving the chemical and physical mechanisms used in the model.Experiments with several chemical schemes have been carried out (this is acontinuation of the work reported in Alexandrov et al., 1997). Seasonal variationsin the deposition mechanisms have been introduced.

(b) Improving the presentation of the output results. Better and fastervisualization and animation routines were developed and used. The use of efficientgraphical tools is very important for the validation procedures and for thedissemination of model results among interested specialists.

(c) Work on better validation of the model results. Different tests have beencarried out to verify the reliability of the results computed by the numericalalgorithms applied in the model. The chemistry-advection rotation test (Hov et al.,1989), which is an extension of the well-known and commonly used advectionrotation test (Molenkampf, 1968 and Crowley, 1968), was used to verify thealgorithms used to describe the transport and the chemical reactions.Measurements, taken in EMEP stations located in different European countries,were also extensively used in the experiments.

(d) Incorporating better and faster numerical algorithms which will allow usto run the model efficiently on different high-speed computers. It is still ratherdifficult (and even impossible when the time-interval is too long) to run the threedimensional version of the Danish Eulerian Model in long-term simulations.Efforts to improve the performance of the model, when it is run on the availablehigh-speed computers, have systematically been carried out. The final aim is toprepare a flexible version of the model based on an extensive use of object-oriented software. This aim has not been achieved in 1998, but some verypromissing results have been obtained. Several easily exchangable modules havebeen developed and attached to the model. The model was run on several high-speed computers, including parallel computers with distributed memory. Astandard tool, MPI (the Message Passing Interface, Gropp et al., 1994), has beenused in the latter case. This will allow us to achieve easily good results ondifferent computers with distributed memory.

(e) Running the model over a very long time period. The Danish EulerianModel has been run over a time period of nine years (from 1989 to 1997). It waspossible to perform these simulations only with the two-dimensional version of themodel. This emphasizes the great importance of obtaining good results in theefforts described in the previous paragraph.

GLOREAM-report

214 of 216

(f) Studying ozone episodes during the summer months of 1989-1997. Ozoneepisodes, in which several critical levels were exceeded in different Europeanregions, have been studied. It has been shown that in many West European andCentral European countries the levels of 90 ppb and 120 ppb have been exceededin many summer days. Also the AOT40 (accumulated over a threshold of 40 ppb)values were exceeded in many parts of Europe; this being true both for the AOT40values for crops and for AOT40 values for forest trees.


The major results obtained in the six major directions that are discussed in theprevious section are listed below.

(a). The introduction of more advanced chemical schemes and depositionmechanisms tends to improve the results (this is especially true for the seasonalvariation of some species).

(b). The improvement of the visualization techniques allows us to perform in amore efficient way different checks of the reliability of the output results; this isdemonstrated in Fig. 1 - Fig. 4.

(c). The combined check of the accuracy of the numerical algorithms and thereliability of the output results (tested by comparisons with measurements)increased the confidence in the model, but it should be emphasized here that muchmore work is needed in this direction.

(d). The improvement of the numerical algorithms resulted in performance whichis five times better than the performance of the model in the beginning of 1998.This allows us to solve now much bigger tasks and much more tasks. However,much better performance is still needed.

(e). Runs over period of many years allowed us to study some trends concerningpollution levels in different European regions.

(f). The studies of episodes with high ozone concentrations in the summer periodsof 1989-1997 indicates that the critical levels of 90 ppb and 120 ppb are exceededin the most polluted areas in Europe. The critical levels for AOT40 values (bothfor crops and for forest trees are exceeded in nearly the whole of Europe(excluding the Northern and central parts of Scandinavia and Russia). In someparts the critical levels are exceeded by a factor greater than seven. Moreover, inmany regions this happens every year.

Results representing different aspect of this project have been reported at teninternational meetings (as invited addresses on some of these meetings). Theresearch is documented in several papers in international journals and proceedings.

GLOREAM-report

215 of 216

Averaged monthly ozone concentrations in air over Frederiksborg (DK32)

0

10

20

30

40

50

60

70

80

90

100

1989

1990

1991

1992

1993

1994

1995

Period

Con

cent

ratio

n in

ppb

.

Calculated

Observed

Averaged monthly ozone concentration in air over Rörvik (SE 2)

0

10

20

30

40

50

60

70

80

90

100

1989

1990

1991

1992

1993

1994

1995

Period

Con

cent

ratio

n in

ppb

.

Calculated

Observed

Figure 1 Comparison of averaged monthlycalculated and measured ozoneconcentrations at Frederiksborg(Denmark) over a seven year period.

Figure 2 Comparison of averaged monthlycalculated and measured ozoneconcentrations at Rörvik (Sweden) overa seven year period.

Averaged monthly ozone concentration in air over Deuselbach (DE 4)

0

10

20

30

40

50

60

70

80

90

100

1989

1990

1991

1992

1993

1994

1995

Period

Con

cent

ratio

n in

ppb

.

Calculated

Observed

Averaged monthly ozone concentrations in air over Aston Hill (GB31)

0

10

20

30

40

50

60

70

80

90

100

1989

1990

1991

1992

1993

1994

1995

Period

Con

cent

ratio

n in

ppb

.

Calculated

Observed

Figure 3 Comparison of averaged monthlycalculated and measured ozoneconcentrations at Deuselbach(Germany) over a seven year period.

Figure 4 Comparison of averaged monthlycalculated and measured ozoneconcentrations at Aston Hill (GreatBritain) over a seven year period.


There are still many open questions and many modules of the model can be furtherimproved. Increasing the performance of the model, so that the three dimensionalversion can be applied in more and more cases is one of the most important tasks.This is a difficult task (see Peters et al., 1995 and Zlatev, 1995), but the computersare becoming faster and faster, which indicates that some progress in this directioncan be achieved. An attempt to run the model on more refined space grids is also

GLOREAM-report

216 of 216

one of the important tasks. Intercomparison with other models could gives someideas for improving some parts of the model.

7. Acknowledgements

This research work on this EUROTRAC-2 project was partially supported byNATO (ENVIR.CRG.930449, OUTS.CRG.960312 and ENVIR.ARW.971731),EU (ESPRIT, #22727 and # 24618) and NMR (the Nordic Council of Ministers) ina project in which teams from Denmark, Finland, Norway and Sweden arecollaborating. A grant from the Danish Natural Sciences Research Council gave usaccess to all Danish supercomputers.

8. References

Alexandrov, V., A. Sameh, Y. Siddique and Z. Zlatev, Numerical integration ofchemical ODE problems arising in air pollution models, EnvironmentalModelling and Assessment 2, (1997), 365-377.

Bastrup-Birk, A., J. Brandt, I. Uria and Z. Zlatev, Studying cumulative ozoneexposures in Europe during a seven-year period, J. Geophys. Res. 102,(1997), 23917-23935.

Crowley, W.P., Numerical advection experiments, Mon. Wea. Rev. 96, (1968),1-11.

Gery, M.W., G.Z. Whitten, J.P. Killus and M.C. Dodge, A photochemical kineticsmechanism for urban and regional computer modeling. J. Geophys. Res. 94,(1989), 12925-12956.

Gropp, W., E. Lusk and A. Skjellum, Using MPI: Portable Programming with theMessage Passing Interface. MIT Press, Cambridge, Massachusetts, (1994).

Hov, Ø., Z. Zlatev, R. Berkowicz, A. Eliassen and L.P. Prahm, Comparison ofnumerical techniques for use in air pollution models with non-linearchemical reactions, Atmos. Environ. 23, (1988), 967-983.

Molenkampf, C.R., Accuracy of finite-difference methods applied to the advectionequation, J. Appl. Meteor. 7, (1968), 160-167.

Peters, L.K., C.M. Berkowitz, G.R. Carmichael, R.C. Easter, G. Fairweather,S.J. Ghan, J.M. Hales, L.R. Leung, W.R. Pennell, F.A. Potra, R.D. Saylorand T.T. Tsang, The current state and future direction of Eulerian models insimulating the tropospherical chemistry and transport of trace species: Areview, Atmos. Environ. 29, (1995), 189-221.

Zlatev, Z., Computer Treatment of Large Air Pollution Models, Kluwer AcademicPublishers, Dordrecht-Boston-London, (1995).

Zlatev, Z., I. Dimov and K. Georgiev, Three-dimensional version of the DanishEulerian Model, Zeitschrift für Angewandte Mathematik und Mechanik 76,(1996a), 473-476.

Zlatev, Z., J. Fenger and L. Mortensen, Relationships between emission sourcesand excess ozone concentrations, Computers and Mathematics withApplications 32, (1996b), 101-123.

GLOREAM - dmu.dk · e-mail: [email protected] Scientific Secretary Annette Münzenberg Institute for...

Documents

Transcript of GLOREAM - dmu.dk · e-mail: [email protected] Scientific Secretary Annette Münzenberg Institute for...