On utilization of OMI products in the atmospheric composition monitoring and evaluation of dispersion models

M. Sofiev, J.Vira, M.Prank,

J.Tamminen, J.Hakkarainen,

J.Soares, R.Vankevich, M.Lotjonen

Outlook

• Introduction

• Example 1: Eyjafjallajokull

� Means of the plume monitoring

� Models, satellites, and in-situ: who does what

• Example 2: air quality model verification with OMI

• Example 3: CALIPSO aerosol profiles

• Example 4: MODIS/AATSR/SEVIRI wild-land fires

• Summary

� user requirements

Eruption of Eyjafjallajokull (Iceland), 14.4→→→→c.m.

• 15.4: first runs, emergency-type setup� Relative source strength, 3, 8km height

• 16.4: first source calibration, results open at SILAM Web site� Source height up to 8km (IMO report)� Strength: left relative

• 18.4: second source calibration, first ensemble, first hemispheric forecast� Source strength: 2 tons/sec of ash� ECMWF 00 and 12 UTC, 120 hours fcst� HIRLAM 00, 06, 12, 18 UTC, 54 hours fcst

• 18.4 – c.m. operational ensemble forecast, output at http://silam.fmi.fi, daily source recalibration, periodic hemispheric forecast

• 19.5 – c.m. operational hemispheric forecast

How to monitor the plume?• Dispersion models

� YES!– immediately available via rapid-response national emergency systems

– provide detailed 4-D information about the plume dispersion– can (roughly) estimate uncertainties of the predictions

� BUT…– source term is unknown, thus absolute concentrations are unknown

… Release is strongly irregular, which will be missed by the models

– injection height is poorly known, thus dispersion pattern is unreliable

• In-situ observations: practically useless for high plumes• In-situ 3D (lidars, baloons, aircrafts, …)

� YES!– arguably the most accurate observations

� BUT…– need the plume to reach the site

– can mix it up with e.g. anthropogenic pollution– Essentially point-wise, do not show the plume as a whole

How to monitor the plume? (2)

• Satellites

� YES!

– quickly available after the satellite overpass

– can provide a good map of the plume

� BUT…

– GEO satellites have problems at high latitudes, LEO satellites provide just a couple of shots per day

…clouds obscure large areas

– cannot distinguish the plume from e.g. anthropogenic pollution

– complicated algorithms, many assumptions that can be wrong in a specific case

• Conclusion: no golden key, ALL sources of information must be used in a complementary manner

Source calibration: manual, daily

SILAM AOD→

MODIS AOD OMI AAI↓ ↓→

15.4.2010

Plume dispersion: SILAM-EC hemispheric forecast

Observations

• Lidars in Switzerland, Germany, UK

• Sun photometers network

• Balloons in Finland, aircrafts in Germany, Iceland, and Finland

• OMI AOD, SO2, and aerosol indices

• MODIS AOD

• AATSR, SEVIRI, MISR, ....

MODIS, 16.4

OMI, SO2, 19.4

OMI + SILAM, jointly

Example 2: OMI for AQ model evaluation

• Setup of the experiment

� 8 regional + 2 global AQ models compute atmospheric pollution including NO2 for 2008-2009 over Europe

� OMI NO2 in column is taken for the same period

� maps are compared for winter and summer seasons

Outcome

Individual models vseach other and OMIAugust 2008

Outcome (2): SILAM sensitivity study

• Conclusion: the trick is in chemistry. SILAM own combined gas-phase and heterogeneous chemistry seem to do better job with regard to NO2 background

� Not for ozone, though

a) b) c)

50km emission 50km emission + low diff CB4 chemistry (!)

Example 3: overdoing may cause confusion

• A model & satellite experiment of detecting the height of the wild-land fire smoke

• Preparatory task: identify the fires whose smoke was observed by CALIPSO

• Model setup

� Time and region: August 2006 and 2007, Europe, ECMWF meteo data, resolution 20 km; 26 vertical layers up to >6 km

� Observations: CALIPSO measurements AOD(location, heigth)

� Modelling analysis: source apportionment via footprint computation

� Looking for: point sources of particulate matter

August 2006

August 2007

Example 4: satellite observations of wild-land fires

On integration of Fire Assimilation System and chemical transport model for monitoring the impact of wild-land fires on atmospheric composition and air quality

M. Sofiev1, R. Vankevich2, J.Soares1, M.Lotjoinen1, J. Koskinen1, J. Kukkonen1

1 Finnish Meteorological Institute2 Russian State Hydrometeorological University

Russian State Hydrometeorological University

Summary

• A non-trivial task usually requires a combination of all tools to monitor the development

� Eyjafjallajokul: the SILAM dispersion model, lidars and satellite products from several instruments

• Source calibration became possible from AOD products and, in principle, can be performed in real-time

� data assimilation in emergency case is an expensive tool but methodologies exist that reduce its costs

Summary: user requirements

• More, more, more, better, better, better…

• By far the most widely used instrument is MODIS – who comes after ?

• User’s thoughts aloud

� For modelling purposes “mean” pictures are of little use due to short model memory. The information comes from individual time-labeled frames, via data assimilation

� No need to solve inverse problems if forward ones are in reach!

– Models are to mimic what satellites measure, NOT the other way round

� The information should come with uncertainty estimation – the real one, not just precision of devices

� If something is unknown, that should be stated. Unsupported guesses bring only noise and damage the results

Acknowledgements

• Projects

� EU-FP 7 GEMS and MACC

� ESA PROMOTE, Ozone SAF

� Academy of Finland IS4FIRES

• Observational data producers

� MODIS, OMI, and CALIPSO teams

� MeteoSwiss

� AERONET sun photometer network