73JANUARY 2005AMERICAN METEOROLOGICAL SOCIETY |

D (ESA 1989). These preparatory activities, includingtheoretical studies, technical developments, and fieldcampaigns, have been described in a “Report for mis-sion selection” (ESA 1999). This report was presentedto the European Earth Observation community at aselection meeting. Four candidate Earth observationmissions were considered, and the atmospheric dy-namics mission for wind profile measurement wasselected as the second of these Earth Explorer coremissions to be implemented. It is now an approvedmission of the European Space Agency with a targetdate for launch in 2007.

The primary aim of the Atmospheric DynamicsMission (ADM-Aeolus) of the European Space Agencyis to provide global observations of vertical wind pro-files. Presently, knowledge of the 3D wind field overlarge parts of the Tropics and major oceans is quiteincomplete. This leads to major difficulties both instudying key processes in the coupled climate systemand in further improving the numerical forecast sys-tems. Progress in climate modeling is intimatelylinked to progress in numerical weather prediction(NWP). The wind profile measurements provided byADM-Aeolus are expected to demonstrate improve-ments in such atmospheric modeling and analysis.

AFFILIATIONS: STOFFELEN—Royal Dutch Meteorological Institute(KNMI), De Bilt, Netherlands; PAILLEUX—Météo-France, Toulouse,France; KÄLLÉN—Meteorological Institute of the University ofStockholm, Stockholm, Sweden; VAUGHAN—Microwave ManagementAssociates, Marlow, Bucks, United Kingdom; ISAKSEN AND

ANDERSSON—European Centre for Medium-Range WeatherForecasts, Reading, United Kingdom; FLAMANT—LaboratoireMétéorologique Dynamique, CNRS, Palaiseau, France; WERGEN—Deutscher Wetterdienst (DWD), Offenbach, Germany; SCHYBERG—Meteorologisk Institutt, Oslo, Norway; CULOMA, MEYNART,ENDEMANN, AND INGMANN—European Space Research andTechnology Centre (ESTEC), Noordwijk, NetherlandsCORRESPONDING AUTHOR: Paul Ingmann, Directorate of EarthObservation Programmes, ESA/ESTEC, Mail Code EOP-SMA,NL-2200 AG Noordwijk, NetherlandsE-mail: paul.ingmann@esa.intDOI:10.1175/BAMS-86-1-73

In final form 24 August 2004©2005 American Meteorological Society

THE ATMOSPHERIC DYNAMICSMISSION FOR GLOBAL WIND

FIELD MEASUREMENTBY AD STOFFELEN, JEAN PAILLEUX, ERLAND KÄLLÉN, J. MICHAEL VAUGHAN, LARS ISAKSEN, PIERRE FLAMANT,

WERNER WERGEN, ERIK ANDERSSON, HARALD SCHYBERG, ALAIN CULOMA, ROLAND MEYNART,MARTIN ENDEMANN, AND PAUL INGMANN

The ADM-Aeolus mission will provide global wind profile observations

with the aim to demonstrate improvement in atmospheric wind analyses

for the benefit of numerical weather prediction and climate studies.

uring the past 15 yr the European Space Agency(ESA) has been evaluating the prospects for aspaceborne Doppler wind lidar (DWL) for mea-

surement of the global wind field. Early concepts hadbeen developed by the Doppler lidar working group

74 JANUARY 2005|

These advances will, in turn, enhance the long-termdatabases being created by NWP data assimilationsystems to serve the climate research community. Assuch, ADM-Aeolus promises to also provide data thatare needed to address some of the key concerns of cli-mate research, including climate variability, validationand improvement of climate models, and processstudies that are relevant to climate change.

The concept of the mission is explained in Fig. 1.A very high performance DWL will be accommo-dated on a satellite flying in a sun-synchronous orbitat an altitude of ~400 km, and will provide near-globalcoverage (see Fig. 2). The DWL is an active instru-ment that fires pulses of laser light toward the atmo-sphere along a line of sight (LOS). In the return sig-nal, backscattered light from molecules and particlesat different levels in the atmosphere is collected and

a sequence of Doppler frequency shifts is measured.The Doppler shift depends on the velocity in LOS di-rection of the scattering particles that move with theairflow. Thus, it allows the determination of the meanwind component velocity profiles.

Meteorological implications and utilization of sucha system are discussed in the following section. Thesection titled “ADM-Aeolus Doppler wind lidar de-scription” discusses technical aspects of a 355-nm-wavelength direct detection scheme and receiver sys-tem combining two interferometers for bothmolecules and larger particles. Conclusions are pre-sented in the last section.

METEOROLOGICAL IMPACT OF WINDPROFILE MEASUREMENTS. Impact for numeri-cal weather prediction. Reliable instantaneous global

analyses of winds are neededto improve the understand-ing of atmospheric dynamicsand climate processes, andalso to improve the quality ofNWP models. Indeed thereis a synergy between ad-vances in climate-relatedstudies and those in NWP,because climate studies areincreasingly using analyses ofatmospheric (and other)fields from data assimilationsystems that are designedoriginally to provide initialconditions for operationalweather forecasting models.These scientific applicationsare severely limited by thelack of direct three-dimen-sional wind information overthe oceans, the Tropics, andthe Southern Hemisphere,where radiosonde observa-tions are scarce.

CURRENT METEOROLOGICAL OB-SERVATIONS AND THEIR LIMITA-TIONS. A wide variety of ob-servation types are currentlyavailable routinely for assimi-lation in NWP systems. Theyconstitute the Global Ob-serving System (GOS) at thebasis of operational weatherforecasting and climate stud-

FIG. 1. Doppler wind lidar principle and measurement geometry: The lidaremits a laser pulse toward the atmosphere, then collects, samples, and retrievesthe frequency of the backscattered signal. The received signal frequency isDoppler shifted from the emitted laser light due to the spacecraft motion, Earthrotation, and wind velocity. The lidar measures the wind projection along thelaser line of sight, using a slant angle versus nadir. Also shown is the mappingof atmospheric heights to layers measured by the detector. The vertical aswell as the horizontal values can be programed, providing observation flex-ibility.

75JANUARY 2005AMERICAN METEOROLOGICAL SOCIETY |

ies. The different observation types can be classifiedin the following way:

1) Surface data: These are the synoptic reports fromland stations and ships, moored and driftingbuoys, and scatterometer winds from satellite ra-dars [such as European Remote Sensing (ERS)and Seawinds]. The surface data provide valuableinformation on the location and intensity ofweather systems and good estimates of surfacewinds over both land and the oceans. However,these data naturally do not provide any profilinginformation.

2) Single-level upper-air data: These include mea-surements obtained from en route aircraft, andthe atmospheric motion vectors (AMVs) that areobtained from geostationary or polar satellite im-agery. The aircraft data provide frequent and ac-curate profiling information at main airports, andhorizontally dense coverage along the main airtraffic routes. However, from a global perspectivethe coverage is insufficient. The wind informationin the AMVs is more uncertain than that from insitu data, because the intrinsic assumption thattracked features (mainly clouds) in the images isadvected by the atmospheric flow and is not al-ways true. The height assignment of the derivedwinds can also be problematic, leading to biasesand systematic errors. Geostationary satellite

AMVs provide good global coverage within ap-proximately 50° of the equator, while polar satel-lite AMVs are useful above 70° latitude.

3) Multilevel upper-air data: These consist of radio-sondes, pilot balloons, radar wind profilers, mea-surements obtained from aircraft during landingand takeoff, and sounding data from polar-orbit-ing satellites. The satellite sounders provide goodglobal coverage of microwave and infrared radi-ance data, which can be assimilated directly for anaccurate definition of the global temperature andhumidity fields (Andersson et al. 1994). Radio-sondes and pilots constitute the main source ofglobal wind profiling information, with goodavailability mainly from the continents in theNorthern Hemisphere.

Despite the sophistication of modern data assimi-lation methods (e.g., Rabier et al. 1998), large uncer-tainties remain in some wide areas of the globe, espe-cially for the wind field. Over the oceans the upper-airwind analysis relies mainly on spaceborne radianceobservations, which, when coupled with accurate sur-face pressure information and geostrophic adjustmenttheory, can provide some information, indirectly, alsoon the wind field. In the Tropics the geostrophic as-sumption is not valid, and direct measurements of thewind are required to produce accurate analyses of theatmospheric flow. The wind information is essential

FIG. 2. ADM-Aeolus DWL coverage in a 6-h time window (dots). The 925-hPa-level analysis wind impact of thesimulated LOS DWL data in the ECMWF observation system simulation experiment (Marseille et al. 2000) isalso shown (arrows).

76 JANUARY 2005|

in the Tropics because it governs the dynamics. In theextratropics, wind data are the primary source of in-formation for small horizontal scale features and deepvertical structures (small scales compared to theRossby radius of deformation), as depicted in Fig. 3.It is only for the large horizontal scale features andshallow vertical structures that the wind field can bederived from the mass field in the assimilation pro-cess with reasonable accuracy (large scales comparedto the Rossby radius of deformation). A space-basedDWL provides wind profile information globally,which is of value particularly in the Tropics and overthe oceans where such observations are otherwiselacking.

PROSPECT OF DWL OBSERVATIONS. One important aspectto be considered for NWP is the detectability of pre-cursor features. Atmospheric structures that are pre-cursors to the development of extratropical cyclonescan often be identified in areas where cyclogenesistakes place. In the preparation of the ADM-Aeolusmission, vertical wind shear has been studied as a rel-evant example. Midlatitude weather systems are be-lieved to have their origin in processes that are encap-sulated in the theory of baroclinic instability. Asuitable measure of the baroclinicity is provided by theEady growth rate index (e.g., Hoskins et al. 1978;Hoskins and Valdes 1990) in which the vertical windshear is the main variable.

The mean vertical wind shear varies with the geo-graphical region, height, and wind component, but

lies generally between 3 and 4 m s-1 per kilometerdepth. The distribution of the shear angle with re-spect to north is very anisotropic and dominant inthe zonal direction in most geographical regions. Ac-cording to its current design, ADM-Aeolus will mea-sure Doppler shift in the across-track direction andvertical wind shear in the zonal direction will be rela-tively well resolved.

The detectability of vertical wind shear was stud-ied by Stoffelen and Marseille (1998) and Veldmanet al. (1999) using a realistic ADM-Aeolus simulator(Marseille and Stoffelen 2003). They investigated theDWL signal return and wind retrieval quality (hori-zontal LOS wind component) for the 25% of caseswith the strongest wind shear at each vertical level inthe forecast model of the European Centre for Me-dium-Range Weather Forecasts (ECMWF). It turnedout, as illustrated in Fig. 4, that the signal strength andthe corresponding wind retrieval quality were quitesimilar in the middle and upper troposphere to caseswithout strong vertical wind shear, and, as such, theoccurrences of shear and cloud were not very well cor-related. It was furthermore noted that relatively fewreturns with very good quality appeared when strongshear was measured, because this only occurs in thecase when shear is determined from two subsequentcloud layers, which is relatively unlikely. A relativelylow detectability of wind shear was observed only inthe lower troposphere in the storm-track region, de-spite the occurrence of aerosol returns at this level.This is probably due to the correlation of shear and

boundary layer cloud at this level.Therefore, it was concluded thatADM-Aeolus will likely have theability to observe vertical windshear in broken cloud conditions,though with a reduced capabilityclose to the surface. Here, otherwind measuring systems will re-main important (e.g., surface sta-tions and scatterometer).

WIND PROFILE OBSERVING SYSTEM EX-PERIMENTS. Even at very short range,NWP models can be hit by severefailures, which, although it cannotalways be proven, are suspected tobe due to a lack of meteorologicalobservations in precursor areas,which are critical for the initial stateof these models. One example is thestorm of 24 December 1997 (nowcalled the “Christmas Eve Storm”),

FIG. 3. Rossby radius of deformation for a latitude of 45° as a function ofhorizontal scale and equivalent depth. Open area denotes the rangewithin which the wind field dominates the atmospheric dynamics, andthree-dimensional wind measurements are important.

77JANUARY 2005AMERICAN METEOROLOGICAL SOCIETY |

which deepened very quickly in the middle of the At-lantic Ocean on 23 and 24 December 1997, then hitIreland and the Irish Sea in the afternoon of the 24th,and finally the north of England and Scotland. Theforecasts of the storm from all of the operational mod-els were very poor, even at very short range as illus-trated in Fig. 5. The examination of the different op-erational analyses on the 22 and 23 December,together with the satellite imagery, indicates that theseanalyses were often unable to accurately catch the dif-ferent weather systems (cyclogenesis, developingwaves, and their precursors), possibly because ofstrong cross-Atlantic circulation, rapid baroclinicdevelopments, and insufficient data coverage (e.g.,Hello et al. 2000). Studies are ongoing to determinethe potential of ADM-Aeolus in such cases.

Detailed studies of extreme cases are just one wayto evaluate the potential impact of an observing sys-tem. Another and more general way is to run “impactstudies” (comparisons of analyses and forecasts withand without the observing system) on longer periods(typically a few weeks) and to evaluate the results byextracting signals, which are averaged over those pe-riods. These impact studies are called observing sys-

FIG. 4. Simulated performance classification of the ADM-Aeolus DWL in the storm-track region between 40° and60° latitude (a) for all profiles, and (b) only for the data points at each level with the 25% strongest wind shears.The classification categories for wind error standard deviation denoted respectively very good, good, low, andvery low are separated by the subsequent approximate thresholds of 1.4, 2.8, and 4.2 m s-----1 (for details, see Marseilleand Stoffelen 2003). Wind shear detectability is only weakly correlated with the occurrence of cloud (Stoffelenand Marseille 1998; Veldman et al. 1999).

tem experiments (OSEs) when an existing observingsystem is studied, and observing system simulationexperiments (OSSEs) when a future observing systemis studied through simulation of the new data.

Several OSSEs and OSEs have been run in the past,which are relevant for a spaceborne wind lidar (seealso Baker et al. 1995). In one particular OSE, run byCress and Wergen (2001), real wind data from NorthAmerica radiosondes and aircraft were tested in or-der to show the impact of accurate 3D wind observ-ing systems on a small portion of the earth. Figure 6shows the impact on global analyses of not using windobservations over North America. After 11 days ofassimilation, the differences, initially located on NorthAmerica, had propagated not only downstream overthe Atlantic Ocean (areas often critical for forecastsover Europe), but also along the whole tropical belt,thus highlighting the uncertainties in analyzing thewind field. Figure 6 stresses once more the need toobserve the wind directly in the Tropics, because it isimpossible to reconstruct from other observationsthrough a proper atmospheric mass/wind balance( agar 2004). Also, outside the Tropics, accurate windobservations appear to be more important than tem-(Žagar

78 JANUARY 2005|

perature observations for the quality of NWP mo-dels. This has been shown by several impact studies(see, e.g., Cress and Wergen 2001).

Marseille et al. (2000) performed an OSSE forADM-Aeolus in a state-of-the-art data assimilationsystem to verify the expectations based on earlier OSEand OSSE. Marseille and Stoffelen (2003) describe theADM-Aeolus DWL simulator that is used in thisstudy.

Importance for climate studies. Climate change issueshave received substantial attention in recent years dueto the increasing awareness that human activities maymodify the climate of the earth (Houghton et al.2001). An increased greenhouse effect due to risinglevels of carbon dioxide in the atmosphere is the mostprobable cause of the recent global warming. Thereare also other greenhouse gases in the atmosphere—the most important one being water vapor, whichdominates over carbon dioxide in terms of the totalgreenhouse warming. The water vapor feedback iscrucial in determining climate change sensitivity. Anincreased heating at the earth’s surface will lead to anincrease in atmospheric water vapor concentrationsand, thus, a further heating implying a positive feed-

back. Some authors, however,have argued that in the Tropicsfeedbacks involving water va-por, clouds, and circulationchanges may even act to de-crease climate change sensitivity(see, e.g., Lindzen et al. 2001). Avery important question is howa further increase in carbon di-oxide and other man-madegreenhouse gases may affect thetotal climate system. The mosteffective tools available to an-swer such questions are physi-cally based climate models thatclosely resemble NWP models.All of the benefits of wind datadiscussed in previous sections,relating to NWP models, are,thus, also relevant to atmo-spheric/oceanic circulationmodels that are used for climatestudies. The ocean and the at-mosphere act together as a heatengine, where the differentialheat input from the sun gener-ates turbulent motion fields thatare not adequately covered by

the present-day observing system. The ADM-Aeolusmission for the first time sets out to globally observethis motion.

Climate change simulations have been performedwith a range of atmosphere–ocean general circulationmodels (AOGCMs) and projections of future climatechange differ considerably between different models(Houghton et al. 2001). All models are based on thesame fundamental physical and numerical principles,but their representation of physical processes (param-eterizations) differ. Examples of such parameterizedprocesses are cloud representation and radiation cal-culations. A particular AOGCM can be compared tothe observed climate through control simulations ofpast climate change, and the parameterized processescan be compared with observations of clouds and ra-diation either in situ or from space. If discrepanciesare found, changes in the formulations of parameter-izations can be tested to see if the simulations improve.A fundamental difficulty with such comparisons is thenontrivial relation between changes in parameteriza-tions and the circulation response of an AOGCM. If,for example, the heating rate in a cloud parameter-ization scheme is changed this will affect the heattransfer between the ocean and the atmosphere,

FIG. 5. Analysis and forecasts of the Christmas Eve Storm—mean sea levelpressure (MSLP) maps illustrating the Christmas Eve Storm, which hit theBritish Isles on 1200 UTC 24 Dec 1997, after a rapid development fromthe middle of the Atlantic Ocean (courtesy A. Persson, ECMWF). Top left:manual analysis for 1200 UTC 24 Dec 1997; Top right: ECMWF 12-h fore-cast from 0000 UTC 24 Dec 1997 (valid 12 UTC); Bottom: equivalent 12-hforecasts from the (left) Met Office and (right) Deutscher Wetterdienstmodels.

79JANUARY 2005AMERICAN METEOROLOGICAL SOCIETY |

which in turn will change the large-scale wind diver-gence in the atmosphere. Tropical circulation systemsare very much determined by the wind field andchanges in the distribution of water vapor that to alarge extent determine cli-mate change feedback sensi-tivity (Cess et al. 1990, 1996,1997). However, the latter ismainly controlled by advec-tion (Pierrehumbert andRoca 1998). The interactionbetween cloud field heatingpatterns and the large-scaletropical circulation may thusgive rise to intricate circula-tion pattern changes whenonly convective heating pa-rameterizations are changed.Comparing a model simula-tion with observations ofclouds and radiation is notsufficient to understand howwell a model simulates an ob-served phenomenon. Thewind field must also be deter-mined in order to obtain afull understanding of the ef-fect of parameterizationchanges.

In Fig. 7 (adapted from Kistler et al. 2001), the zon-ally and temporally averaged wind field is shown fora 15-yr period in the tropical region, and results fromboth the ECMWF and the National Centers for En-

FIG. 6. Difference in the wind and geopotential analysis at 500 hPa caused by not using wind profile observationsfrom radiosondes, pilots, and aircraft over the United States and Canada during 11 days of assimilation. Thefields are valid for 30 Jan 1998; the contour interval for height is 20 gpm.

FIG. 7. Comparison of the zonal average of the zonal wind component for the(top left) NCEP–NCAR reanalyses and (top right) 15-yr ECMWF Re-Analy-ses. The fields are Dec, Jan, Feb averages between 1979 and 1994. The differ-ence in the reanalyses and the difference scaled by the total temporal vari-ance (%), are displayed in the bottom panels (after Kistler et al. 2001).

80 JANUARY 2005|

vironmental Prediction (NCEP)–National Center forAtmospheric Research (NCAR) reanalyses are used.The difference between the reanalyses is substantialin the tropical region; the shift in the wind field is ofthe same order as the wind field itself. Both reanaly-ses use the same observation database; the differencesoccurring are, thus, due to differences in the assimi-lating models. As noted above, the model differencesare mainly coupled with parameterizations of cloudsand radiation. A conclusion from these results is thatthe presently available observations are not sufficientto constrain the wind field in the Tropics. More windprofile observations are needed.

Model simulations of future climate change showa large spread, and some of this spread is due to dif-ferences in the parameterization of atmospheric pro-cesses. An intercomparison between state-of-the-artAOGCMs shows that projected changes in tempera-ture and precipitation differ considerably betweenmodels under the same greenhouse gas concentrationchange scenarios (Räisänen 2001). These differencesare even more marked when regional climate changeis considered, Timmerman et al. (1999) and Nodaet al. (1999) have investigated the change in El Niñopatterns under global warming and found contradic-tory results using different AOGCMs.

The impact of the wind observations is, however,not limited to tropical regions. It has been suggestedthat global warming can alter the statistics ofmidlatitude disturbances (Carnell and Senior 1998),and this may, in turn, lead to changes of larger-scalecirculation patterns. Regional climate change verymuch depends on the statistics of these large-scalepatterns, and more comprehensive wind data will helpin the understanding of processes that govern regionalclimate change.

ADM-Aeolus observation of the tropical circulation. TheADM-Aeolus mission will provide a much improvedwind dataset to be used for climate process studies.As noted before, it is necessary to assimilate the winddata in a NWP-based data assimilation system to pro-vide a dynamically consistent set of atmospheric cli-mate states to be used for climate model developmentand climate process understanding. Main improve-ments are expected in tropical areas where presentwind profile observations are very sparse. From dy-namical considerations it can be determined that thewind field is the governing atmospheric state variablein the Tropics, and wind data are thus most crucialin these regions. In line with this, after defining theappropriate mass–wind relationships for the effectiveassimilation of the wind data (e.g., •agar et al.

2004a,b), assimilation of ADM-Aeolus line-of-sightwinds in the tropical region shows a clear positiveimpact within an idealized framework (•agar 2004).

Important variables in the tropical circulation andthe hydrological cycle are heat and humidity trans-port. Stoffelen and Marseille (1998) and Veldmanet al. (1999) studied the detectability of tropical fluxesof heat and humidity in cloudy areas. Obviously, pro-cesses of humidity transport, condensation, and pre-cipitation are often associated with cloud. On theother hand, multiple shots in the case of cloud poros-ity (Winker and Emmitt 1998) can enhance the de-tectability of atmospheric heat and humidity fluxes incloudy conditions as discussed in ESA (1999).

As an example of the usefulness of the ADM-Aeolus DWL for studying atmospheric processes thatare associated with the energy balance, Fig. 8a showsthe wind performance classification in the Tropics forthose meridional moisture fluxes that are larger thanthe mean plus one standard deviation of the flux vari-ability at the level considered. By comparing thesetypes of histograms with the corresponding histo-grams of performance classification for all flux con-ditions (Fig. 8b), it can be concluded that many re-turns from clouds appear in moist air at allgeographical locations. At the lowest altitudes in themiddle and lower troposphere, cloud obstruction ofthe moisture flux is most substantial. Aerosol scatter-ing provides generally improved performances inmoist conditions with respect to the molecular scat-tering signal. However, extensive cloud fields in theplanetary boundary layer may prevent sampling of thelowest atmospheric layer over relatively large regions.

Thus, assuming that LOS winds can be measuredat cloud-top heights that are representative for thegeneral flow, a DWL in space provides essential in-formation on the tropical circulation in the upper tro-posphere. On the other hand, clouds obscure the de-tectability of moisture fluxes in the middletroposphere, particularly in the Tropics. Due to thepresence of aerosol scattering, the near-surface mois-ture flux is clearly visible in the majority of cases.

Observational requirements for data assimilation. Likemany other meteorological observations, the space-borne LOS wind component profiles by themselvesseem at first glance to be of limited value, but in thecontext of atmospheric data assimilation systems theywould in fact be an essential component of the GOS,just like radiosonde wind profiles today. For a mis-sion intended to demonstrate the feasibility of a full-scale spaceborne wind observing system to improveglobal atmospheric analyses, the requirements for data

(Žagar

Žagar

81JANUARY 2005AMERICAN METEOROLOGICAL SOCIETY |

quality and vertical resolution are the most stringentand most important to achieve. Under this assump-tion, the density of profile observations is of lower pri-ority among the requirements. However, the deriva-tion of the coverage specification is supported byweather forecast impact experiments. These includedthe inputs of the conventional wind profile network,which is thin and irregular but of key importance (seethe section titled “Impact for numerical weather pre-diction”). Moreover, the coverage specification iscompatible with the World Meteorological Organi-zation (WMO) threshold requirements. The WMOrecognizes the prime need for wind profile data(WMO 2000) and has defined wind profile measure-ment requirements (WMO 1996, 2001).

The full wind vector consists of two horizontal anda vertical component. The average vertical wind com-ponent is small over a typical meteorological modelgrid box (Courtier et al. 1992) and is, in general, neg-ligible. A lidar instrument can only observe the com-ponent along its LOS. Because the horizontal compo-nents are needed, the horizontal projection of the LOSwinds (HLOS) is the quantity of interest.

To provide the two horizontal wind components,the same volume needs to be viewed from two differ-

ent azimuth angles. Because this requirement imposessevere technical constraints, an alternative solution todemonstrate a spaceborne DWL is to provide onecomponent only. The provision of only one windcomponent is no technical limitation to modern dataassimilation schemes, just as it is possible to assimilatea temperature measurement without a wind measure-ment. Figure 2 depicts the simulated analysis windimpact of a few tracks of ADM-Aeolus DWL data.

Lorenc et al. (1992) performed an OSE where ei-ther no, one, or two components of a cloud motionwind (CMW) vector were assimilated, and they foundhalf of the beneficial forecast impact when only onewind component, rather than both, was used. Thus,it appears that single-component HLOS winds canprovide significant NWP improvement. Moreover, inan additional experiment, where 50% of all CMWvectors were randomly removed, 50% of the benefi-cial forecast impact of all CMW was also found in-deed. Thus, the expected analysis and forecast impactof two-wind-component measurements is the samefor two collocated orthogonal components and fortwo spatially well-separated measurements of onesingle component. These findings are further sup-ported by impact studies (Leike et al. 2001; Marseille

FIG. 8. Simulated performance classification of the ADM-Aeolus DWL in the tropical region below 20° latitude(a) for all profiles, and (b) only for the data points at each level with the 20% strongest humidity flux. Classifica-tion is as in Fig. 4. The detectability of the near-surface humidity flux is somewhat reduced but better than 60%in the Tropics (Stoffelen and Marseille 1998).

82 JANUARY 2005|

et al. 2000), which demonstrated a positive impact ofsimulated single-LOS winds. This confirms that thematching of multiple azimuth “looks” in one geo-graphical area is not required for data assimilation.However, two independent wind components may bedesirable in certain flow configurations (Riishojgaardet al. 2004).

At very small scales (e.g., the footprint of a DWL),and in extreme cases such as thunderstorms, the ver-tical component of the wind may be quite substantialand the assumption to neglect vertical motion isstrictly not valid in such situations. However, currentNWP models cannot represent these small scales, andas such the vertical motion is regarded as an unwantedcomponent of the measurement and is treated as partof the so-called spatial representativeness error(Lorenc et al. 1992).

Over data-sparse areas the 2–3 m s-1 accuracy re-quirement is expected to be sufficient to provide abeneficial impact on meteorological analyses; the re-quirement is comparable with the typical first-guesserror. This requirement is significant because expe-rience in meteorological data assimilation shows thatobservations with an accuracy that is poorer than thefirst guess often fail to have a significant beneficialimpact on NWP. In summary, Table 1 provides anoverview on the ADM-Aeolus requirements.

The expected meteorological impact in the Trop-ics is the most certain, and, from a climatological

point of view, is also the most useful. Moreover, to im-prove atmospheric analysis beyond the Tropics, andmore particularly NWP in Europe, the above require-ments have been carefully chosen to be able to dem-onstrate the beneficial impact of DWL winds atmidlatitudes (ESA 1999).

ADM-AEOLUS DOPPLER WIND LIDAR DE-SCRIPTION. Measurement principle. ADM-Aeoluswill use a single LOS perpendicular to the flight di-rection (Fig. 1). The spaceborne lidar will emit a spec-trally narrow laser pulse directed at a 35° slant angletoward the atmosphere. Laser light of a 355-nm wave-length is scattered in the atmosphere by molecules(Rayleigh–Brillouin scattering) and by small aerosoland cloud particles (Mie scattering) that move withthe wind. A very small fraction of such scattering isbackscattered toward the spacecraft. The scatteredlight has a different frequency than the light that isemitted by the laser because of the Doppler effect. Dueto the scatterer’s relative movement in the directionof observation, a frequency shift df occurs, given by

df = –2V/l, (1)

where V, the LOS velocity between the transmitterand scatterers, is due to three contributions—space-craft motion, Earth rotation, and wind flow—and lis the laser-transmitted wavelength. The Doppler fre-

quency shift due to the plat-form moving at 7 km s-1 iscancelled as much as possibleby looking at 90° from thedirection of travel and, more-over, can be corrected. Figure9 shows schematically thewavelength distribution ofthe light signal.

The wide-bandwidthRayleigh–Brillouin scatter-ing spectrum, equivalent to~600 m s-1 full width at halfmaximum (FWHM), isnearly of Gaussian form atshorter wavelengths. A Dop-pler Gaussian shape appliesstrictly at low pressureswhere correlations betweenthe motions of gas moleculescan be disregarded. At higherpressures, close to one atmo-sphere, the molecular veloc-ity correlations arise due to

Key parameters

Vertical domain (km)

Vertical resolution (km)

Horizontal domain

No. of profiles (h-1)

Profile separation (km)

Horizontal integration length (km)

Horizontal subsample length (km)

Accuracy (HLOS component) (m s-1)

Zero-wind bias (m s-1)

Wind speed slope error (%)

Data reliability (%)

Data availability (h)

Length of observational dataset (yr)

PBL

0–2

0.5

1

Troposphere

2–16

1

Global

>100

>200

50

0.7–50

2

0.1

0.5

95

3

3

Stratosphere

16–20

2

3

Observational requirements

TABLE 1. Observational requirements of the Atmospheric DynamicsMission.

83JANUARY 2005AMERICAN METEOROLOGICAL SOCIETY |

the propagation of pressure waves through the scat-tering medium, and result in the Rayleigh–Brillouinspectrum. This effect is taken into account in theevaluation of the lidar performance. The two contri-butions, from aerosols and molecules, are centred atthe same frequency and superimposed:

1) narrow-bandwidth aerosol (Mie) scattering typi-cally equivalent to a ~10 m s-1 width or less (inpractice, the line width is largely due to the laserline width), and

2) broadband molecular (Rayleigh–Brillouin) scat-tering with an equivalent width (FWHM) of~600 m s-1.

In addition, there will be a broadband, semicon-tinuous background of reflected sunlight. The rela-tive magnitude of the first two components is ex-pressed by the scattering ratio (the ratio between thetotal scattering and molecular scattering) and islargely determined by the laser wavelength accordingto a l–4 dependence for molecules and l-a for particles,where the exponent “a” varies from 0 to 3, depend-ing on the particle size and wavelength. The scatter-ing ratio is given by (1 + Cl–a+4), where C is a con-stant. Typically, in the midtroposphere at 10 µm (inthe infrared), aerosol (Mie) scattering is strongest,whereas at 355 nm (in the ultraviolet) the molecular(Rayleigh–Brillouin) scattering predominates and thescattering ratio is close to unity (except in clouds).Aerosol scattering is highly variable (several orders ofmagnitude) depending on the previous history of theair mass and is usually very strong within the plan-etary boundary layer and at high levels in the atmo-sphere, for example, thin and incipient cirrus clouds.Molecular scattering that is related to temperatureand pressure varies more slowly and predictablythrough the atmosphere (roughly a factor 3 from thesurface up to the tropopause level). At the lidar re-ceiver the light signal from the atmosphere: S(R) = KR–2 b(R) T 2 [R is the range from spacecraft, K is aninstrumental constant, b is the backscatter coefficient,and T is the atmospheric transmission] is sampled se-rially in time (for R = ct/2) in 20 successive range gates(DR), which determine the height resolution in the at-mosphere, where typically DR equals 0.25–1 km.

ADM-Aeolus design. In early proposals dedicated towind measurements from space the National Oceanicand Atmospheric Administration (NOAA) and Na-tional Aeronautics and Space Administration (NASA)considered CO2 laser technology, with heterodyne de-tection and a conical scan at ~45º from nadir to

sample the atmospheric wind field. This basic conceptwas proposed for both the WINDSAT (Huffaker1978; Osmundson 1981) and Laser AtmosphericWind Sounder (LAWS; Baker 1995) projects. TheCO2 laser technique was considered as mature at thattime, but the necessary accuracy for LOS measure-ments resulted in demanding requirements for a largeCO2 laser system, a large scanning telescope, and a lag-angle compensation unit. More recently, a multipleperspective CO2 Doppler lidar has been studied inFrance by Centre Nationale d’Etudes Spatiales (CNES;CNES 1988) in the framework of the BilanEnergetique du System Tropical (BEST; or TropicalSystem Energy Budget). In BEST, the scanning re-quirement was made simpler using four (or even onlytwo) fixed LOSs set at 90° from one another and 45°from nadir. More recently, NASA considered a sys-tem based on heterodyne detection at shorter wave-lengths with a 2.1-µm solid-state laser. None of theseprojects proceeded beyond the feasibility level (phaseA) or consolidated (phase B) study. It is worth not-ing that the CO2 technology and a conical scan arecurrently used in the airborne WIND instrument de-veloped in French–German cooperation (Werneret al. 2001). Wind infrared Doppler lidar (WIND) has

FIG. 9. Schematic spectrum (blue) of the light collectedby a lidar in the UV near-visible region showing thescattering from aerosols (Mie return) and molecules(Rayleigh–Brillouin return). The received spectrum isshifted with respect to the emitted laser light (red line).The dotted curve represents a not-shifted spectrum(zero wind speed). The molecular return signal isbroad due to thermal motion. The particular returnsignal is narrow for the particles that are heavier thanmolecules, and both are in collisional (kinetic energy)equilibrium. Also shown the position of the receiver’sspectral filters A and B used for the double-edged de-tection technique.

84 JANUARY 2005|

been involved successfully in international field cam-paigns since its completion in 1999 (Reitebuch et al.2001).

In parallel with these studies, ESA has been con-sidering the feasibility of a space-based Doppler windlidar. These activities have resulted in the selection ofa high-performance Doppler wind lidar based on di-rect detection implementing interferometric tech-niques at the receiver level. This technique had beendeveloped 30 yr ago (Benedetti-Michelangeli et al.1972) and has been used routinely since then at532 nm (Chanin et al. 1989). For eye safety reasons aUV wavelength is preferable. As a result, such a sys-tem, with pulsed laser operating at 355-nm wave-length would utilize both Rayleigh–Brillouin scatter-ing from molecules and Mie scattering from thincloud and aerosol particles.

The direct-detection UV concept at 355 nm, uti-lizing both aerosol and molecular scattering, was se-lected as the best compromise in terms of perfor-mance, especially for its robustness in terms of its lackof aerosols and the consequent impact on perfor-mance. This concept can provide compliant windmeasurements up to high altitudes (20 km and above)for an operational mission, which is basically indepen-dent of the aerosol content. The technical and pro-grammatic risks were evaluated as acceptable, and fi-nally the mission based on the present instrument

concept promises to be fully compliant with the re-quirements as discussed in the section, “Meteorologi-cal impact of wind profile measurements.”

An intercomparison and validation study was con-ducted in the framework of the ESA campaign called“VALID.” During 2 weeks in 1999, several ground-based lidars, a 72-MHz radar, and radiosondes havebeen operated simultaneously. The measurementshave been compared with the ECMWF analysis(Delaval et al. 2000). The results presented during theGranada, Spain, meeting in 1999 were convincingenough to go ahead with the present mission.

The ADM-Aeolus laser source is based on a single-mode, 150-mJ, 100-Hz pulse/repetition frequency,diode-pumped and frequency-tripled (355 nm) Nd-YAG laser (see Table 2).

A 1.5-m-diameter Cassegrain afocal telescope isproposed as transceiver (for both transmitting andreceiving). An isothermal and lightweight design willbe used for the telescopic mirrors and structure, yield-ing the required optical quality and stability withoutadditional focusing or alignment mechanisms.

The combination of two receivers—one for theMie spectrum and one for the Rayleigh–Brillouinspectrum—thus provides an optimal performanceover the whole altitude range. The Mie receiver is aFizeau spectrometer, combined with a charge-coupled device (CCD) detector. This spectrometer

has been validated in the frameworkof the ESA Technology ResearchProgramme. The Rayleigh receiver isplanned as a double Fabry–Perotetalon. Both aerosol (Mie) and mo-lecular (Rayleigh–Brillouin) scatter-ing will be examined simultaneously.The central part of the spectrum is fil-tered out and directed toward the Mieanalyzer. The Fizeau fringe is super-imposed on some Rayleigh residuallight and atmospheric radiance, whichprovide an additional noise back-ground. Computational analysis de-termines the center of the Mie fringe,and, after appropriate frequency cali-bration, the Doppler shift.

The broadband Rayleigh–Brillouin(molecular) scattering is analyzed bythe dual-filter or double-edged tech-nique (see Fig. 9). Two filters formedby the double Fabry–Perot etalon arepositioned on either side of the laserwavelength in order to perform adifferential measurement of the scat-

SatelliteAltitude 400 kmOrbit Sun synchronous, 6–18 hMass 1200 kgSingle LOS At 90° from platform direction of travelNadir slant angle 35°

InstrumentWavelength (3rd harmonic) 355 nm, single modePulse energy 150 mJ in 30 nsRepetition rate 100 Hz in burst modeBurst operation 7 s on, 5-s warm-up, 16 s off

Signal format (postprocessing)Altitude range above surface -1 to +30 km (extendable)(for Mie and Rayleigh channel)Vertical resolution 0.25–5 km (adjustable)

Table 2. Technical parameters of the spaceborne ADM-AeolusDoppler wind lidar.

Parameters Value

85JANUARY 2005AMERICAN METEOROLOGICAL SOCIETY |

tered light. Because the line shape is Gaussian it canbe described by two parameters. Two measurementswithin the Rayleigh–Brillouin spectrum are thereforesufficient to provide the line center and, thus, theDoppler frequency shift and wind speed.

In addition, because ADM-Aeolus is a high spec-tral resolution lidar the signal strengths in theRayleigh and Mie channels can be used to derive sec-ondary products like aerosol and cloud optical prop-erties.

The high-sensitivity CCD detectors accumulatebatches of 15 or 50 return signals on the chip. Theseaccumulated signals are downlinked for quality con-trol and further averaging in the processing station.The full observation of 700 shots is typically takenover 7 s, followed by a quiet period of about 21 s. Inthis 7-s-measurement period the satellite will havetraveled approximately 50 km and thus the wind fieldswill have been effectively averaged over this distancein the propagation direction. The vertical height reso-lution is determined by techniques of time gating thereturn signal. At low levels a resolution of 0.5 km isrequired, extending to 1 km in the troposphere andto 2 km above a 16-km altitude. Extensive studies ofsystem performance by both scientific laboratoriesand industry in the framework of preparatory stud-ies had been carried out (Courtier et al. 1992; Delavalet al. 2000; ESA 1989, 1999; Lorenc et al. 1992;Marseille and Stoffelen 2003; Marseille et al. 2000;Vaughan et al. 1999; Stoffelen and Marseille 1998).Overall system analysis, with budgeting for a widerange of potential errors sources, shows performancewithin the required specification.

ADM-Aeolus will provide about 3000 globally dis-tributed wind profiles per day, above thick clouds ordown to the surface in clear air, at typically 200-kmseparation along the satellite track. The 200-km sepa-ration is compliant with synoptic analysis (see thesection titled “Meteorological impact of wind profilemeasurements”). Wind information in thin clouds orat the top of thick clouds is also attainable; informa-tion on other elements such as cloud and aerosols canbe extracted as well. A near-real-time delivery of datato the main NWP centers is anticipated.

After the detailed design of the hardware, integra-tion and testing of the major building blocks, as wellas satellite and instrument integration and testing, thelaunch is planned to be in late 2007.

Ground-based and airborne measurement cam-paigns are underway to obtain realistic atmosphericsituations for use in algorithm development, and toallow the final selection of instrument parameters foroptimised in-orbit measurement performance.

SUMMARY AND CONCLUDING REMARKS.The Atmospheric Dynamics Mission (ADM-Aeolus),is currently being developed by the European SpaceAgency within its Living Planet Programme. TheADM-Aeolus will demonstrate the capability of a spa-ceborne Doppler wind lidar to accurately measurewind profiles in the troposphere and the lower strato-sphere (0–27 km). The mission thus contributes toresolving one of the main identified deficiencies of thecurrent Global Observing System. From thebackscattered frequency-shifted laser light it will bepossible to obtain about 3000 globally distributed pro-files of horizontal line-of-sight winds daily, and withgood vertical resolution. The accuracy of ADM-Aeolus winds, in most cloud-free regions and abovethick clouds, is expected to be comparable to that ofradiosonde wind measurements.

The ADM-Aeolus laser will emit a narrow line-width pulse directed at a 35° slant angle toward theatmosphere. Atmospheric scattering of light in thechosen wavelength (355 nm) is due to both molecu-lar (Rayleigh–Brillouin) and aerosol (Mie) scattering,providing a sufficient return signal emanating fromlayers throughout the troposphere and the lowerstratosphere. The strength of the return signal, and,thus, the quality of the derived wind, will depend on thecloudiness and also, in the lower troposphere, on theaerosol loading. In overcast situations high-qualitywind observations will be obtained down to the cloudtop. The mission also provides ancillary informationon the aerosol concentration and cloud-top height.

The ADM-Aeolus wind profiles will find wide ap-plication in NWP and climate studies, improvingthe accuracy of numerical weather forecasting, ad-vancing our understanding of tropical dynamics, andprocesses relevant to climate variability and climatemodeling. With a target launch date in 2007, work hasalready been instigated preparing for the future real-time assimilation of ADM-Aeolus wind data into op-erational NWP models. The midlatitude focus is onthose regions where forecast performance is knownto be particularly sensitive to the accuracy of initialconditions. In particular, a beneficial impact on theprediction of severe storm events is expected and fur-ther investigated. For the Tropics, the focus of cur-rent investigations is on defining the appropriatemass–wind relationships for effective assimilation ofthe ADM-Aeolus wind data in state-of-the-art data as-similation systems (e.g., •agar et al. 2004a,b; •agar2004).

During its projected 3-yr lifetime the ADM-Aeoluswill demonstrate the feasibility of global wind fieldmeasurement from space. Based on the results that are

Žagar Žagar

86 JANUARY 2005|

obtained with ADM-Aeolus, future operational mis-sions may be built, fully exploiting the concept of spa-ceborne Doppler wind lidars.

REFERENCESAndersson, E., J. Pailleux, J.-N. Thépaut, J. Eyre, A. P.

NcNally, G. Kelly, and P. Courtier, 1994: Use ofcloud-cleared radiance data in three/four dimen-sional variational data assimilation. Quart. J. Roy.Meteor. Soc., 120, 627–653.

Baker, W. E., and Coauthors, 1995: Lidar-measuredwinds from space: A key component for weather andclimate prediction. Bull. Amer. Meteor. Soc., 76, 869–888.

Benedetti-Michelangeli, G., F. Congedutti, and G.Fiocco, 1972: Measurement of aerosol motion andwind velocity in the lower troposphere by Doppleroptical radar. J. Atmos. Sci., 29, 906–910.

Carnell, R. E., and C. A. Senior, 1998: Changes in mid-lati-tude variability due to increasing greenhouse gases andsulfate aerosols. Climate Dyn., 14, 369–383.

Cess, R. D., and Coauthors, 1990: Intercomparison andinterpretation of climate feedback processes in 19atmospheric general circulation models. J. Geophys.Res., 95, 16 601–16 615.

——, and Coauthors, 1996: Cloud feedback in atmo-spheric general circulation models: An update. J.Geophys. Res., 101, 12 791–12 794.

——, and Coauthors, 1997: Comparison of the seasonalchange in cloud-radiative forcing from atmosphericgeneral circulation models and satellite observations.J. Geophys. Res., 102, 16 593–16 603.

Chanin, M. L., A. Garnier, A. Hauchecorne, and J.Porteneuve, 1989: A Doppler lidar for measuringwinds in the middle atmosphere. Geophys. Res. Lett.,16, 1273–1276.

CNES, 1988: BEST—Tropical system energy budget.Final Rep., 58 pp.

Courtier, P., P. Gauthier, and F. Rabier, 1992: Study ofpreparation for the use of Doppler wind lidar infor-mation in meteorological assimilation systems. ESAContract Rep. 3453, 82 pp.

Cress, A., and W. Wergen, 2001: Impact of profile ob-servations on the German Weather Service’s NWPsystem. Meteor. Z., 10, 91–101.

Delaval, A., P. H. Flamant, C. Loth, A. Garnier, C. Vialle,D. Bruneau, R. Wilson, and D. Rees, 2000: VALID-2—Performances validation of direct detection andheterodyne detection Doppler wind lidars. ESA Con-tract Final Rep., 76 pp.

ESA, 1989: ALADIN—Atmospheric Laser Doppler In-strument. Working Group Rep. ESA SP-1112, 45 pp.

——, 1999: Atmospheric dynamics mission. MissionSelection Rep. ESA SP-1233(4), 157 pp.

Hello, G., F. Lalaurette, and J. N. Thépaut, 2000: Com-bined use of sensitivity information and observationsto improve meteorological forecasts: A feasibilitystudy applied to the “Christmas Storm.” Quart. J.Roy. Meteor. Soc., 126, 621–647.

Hoskins, B. J., and P. J. Valdes, 1990: On the existenceof storm tracks. J. Atmos. Sci., 47, 1854–1864.

——, I. Draghici, and H. C. Davies, 1978: A new look atthe w-equation. Quart. J. Roy. Meteor. Soc., 104, 31–38.

Houghton, J. T., Y. Ding, D. J. Griggs, M. Noguer, P. J.van der Linden, X. Dai, K. Maskell, and C. A.Johnson, Eds., 2001: Climate Change 2001: The Sci-entific Basis. Contribution of Working Group I to theThird Assessment Report of the International Panel onClimate Change. Cambridge University Press, 572 pp.

Huffaker, R. M., 1978: Feasibility study of satellitebornelidar global wind monitoring system. NOAA Tech.Memo. ERL WPL-37, 276 pp.

Kistler, R., and Coauthors, 2001: The NCEP–NCAR 50-year reanalysis: Monthly means CD-ROM and docu-mentation. Bull. Amer. Meteor. Soc., 82, 247–267.

Leike, I., J. Streicher, C. Werner, V. Banakh, I. Smalikho,W. Wergen, and A. Cress, 2001: Virtual Doppler li-dar instrument. J. Atmos. Oceanic Technol., 18, 1447–1456.

Lindzen, R., M.-D. Chou, and A. Y. Hou, 2001: Does theearth have an adaptive infrared iris? Bull. Amer.Meteor. Soc., 82, 417–432.

Lorenc, A. C., R. J. Graham, I. Dharssi, B. Macpherson,N. B. Ingleby, and R. W. Lunnon, 1992: Preparationfor the use of Doppler wind lidar information inmeteorological assimilation systems. ESA ContractRep. 3454, 90 pp.

Marseille G. J., and A. Stoffelen, 2003: Simulation ofwind profiles from a space-borne Doppler wind li-dar. Quart. J. Roy. Meteor. Soc., 129, 3079–3098.

——, ——, F. Bouttier, C. Cardinali, S. de Haan, and D.Vasiljevic, 2000: Impact assessment of a Dopplerwind lidar in space on atmospheric analyses andnumerical weather prediction. Proc. Fifth Int. WindsWorkshop, Lorne, Australia, Coordination Group forMeteorological Satellites of WMO, EUMETSATPublication 28, 275–282.

Noda, A., K. Yoshimatsu, S. Yukimoto, K. Yamaguchi, andS. Yamaki, 1999: Relationship between natural variabil-ity and CO2-induced warming pattern: MRI AOGCMexperiment. Proc. 10th Symp. on Global Change Stud-ies, Dallas, TX, Amer. Meteor. Soc., 359–362.

Osmundson, J. S., 1981: Wind satellite/WINDSAT/experi-ment. Proc. Seminar on Shuttle Pointing of Electro-op-tical Experiments, Los Angeles, CA, SPIE, 395–398.

87JANUARY 2005AMERICAN METEOROLOGICAL SOCIETY |

Pierrehumbert, R. T., and R. Roca, 1998: Evidence forcontrol of Atlantic subtropical humidity by largescale advection. Geophys. Res. Lett., 25, 4537–4540.

Rabier, F., J. N. Thépaut, and P. Courtier, 1998: Extendedassimilation and forecast experiments with a four-dimensional variational assimilation system. Quart.J. Roy. Meteor. Soc., 124, 1861–1887.

Räisänen, J., 2001: CO2-induced climate change inCMIP2 experiments: Quantification of agreementand role of internal variability. J. Climate, 14, 2088–2104.

Reitebuch, O., C. Werner, I. Leike, P. Delville, P. H.Flamant, A. Cress, and D. Engelbart, 2001: Experi-mental validation of wind profiling performed by theairborne 10-µm heterodyne Doppler lidar WIND. J.Atmos. Oceanic Technol., 18, 1331–1344.

Riishojgaard, L. P., R. Atlas, and G. D. Emmitt, 2004: Theimpact of Doppler lidar wind observations on asingle-level meteorological analysis. J. Appl. Meteor.,43, 810–820.

Stoffelen, A., and G.-J. Marseille, 1998: Study on the util-ity of Doppler wind lidar data for numerical weatherprediction and climate. ESA Contract Rep. 4198, 67pp.

Timmerman, A., J. Oberhuber, A. Bacher, M. Esch,M. Latif, and E. Roeckner, 1999: Increased El Niñofrequency in a climate model forced by greenhousewarming. Nature, 398, 694–696.

Vaughan, J. M., N. J. Geddes, P. H. Flamant, P.Drobinski, and C. Flesia, 1999: Wind lidar: Funda-mental review of heterodyne and direct detection.ESA Contract Rep. TIDC-CR-7191, 94 pp.

Veldman, S. M., A. Stoffelen, G. J. Marseille, J. van Es,H. A. Knobhout, and E. A. Kuijpers, 1999: Lidar Per-formance Analysis Simulator (LIPAS). Contract Fi-nal Rep. ESA CR(P)-4290, 50 pp.

Werner, Cl., and Coauthors, 2001: WIND infraredDoppler lidar instrument. Opt. Eng., 40, 115–125.

Winker, G. D., and D. M. Emmitt, 1998: Relevance ofcloud statistics derived from LITE data to futureDoppler wind lidars. Proc. Ninth Conf. on CoherentLaser Radar, Linköping, Sweden, SPIE, 144–147.

WMO, 1996: Guide to Meteorological Instruments andMethods of Observation. 6th ed. Secretariat of theWorld Meteorological Organization, WMO Series 8.

——, 2000: Statement of guidance regarding how wellsatellite capabilities meet WMO user requirementsin several application aeas. WMO Satellite Rep. SAT-22, WMO/TD 992, 29 pp.

——, 2001: Statement of guidance regarding how wellsatellite capabilities meet WMO user requirementsin several applications areas. WMO Satellite Rep.SAT-26, WMO/TD 1052, 52 pp.

•agar, N., 2004: Assimilation of equatorial waves by lineof sight wind observations. J. Atmos. Sci., 61, 1877–1893.

——, N. Gustafsson, and E. Källén, 2004a: Dynamical re-sponse of equatorial waves in variational data assimi-lation. Tellus, 56A, 29–46.

——, ——, and ——, 2004b: Variational data assimila-tion in the Tropics: The impact of a background er-ror constraint. Quart. J. Roy. Meteor. Soc., 130, 103–125.

Žagar,