Irrigation and Drainage Systems (2005) 19: 223–249CSpringer 2005Review on estimation of evapotranspirationfrom remote sensing data: From empiricalto numerical modeling approachesDOMINIQUE COURAULT, BERNARD SEGUIN & ALBERT OLIOSOInstitut National de la Recherche Agronomique, unit´e Climat-Sol-Environnement,Domaine St Paul, bˆatiment Climat, site Agroparc, 84914 AVIGNON, cedex 9, France;E-mail: [email protected]. Different methods have been developed to estimate evapotranspiration from remotesensing data, from empirical approaches such as the simplified relationship to complex methodsbased on remote sensing data assimilation along with SVAT models. The simplified relationshiphas been applied from small spatial scale using airborne TIR images to continental scale withNOAA data. Assimilation procedures often require remote sensing data over different spectraldomains to retrieve input parameters which characterize surface properties such as albedo,emissivity or Leaf Area Index. A brief review of these different approaches is presented, witha discussion about the main physical bases and assumptions of various models. The paperreports also some examples and results obtained over the experimental area of the AlpillesReseda project, where various types of models have been applied to estimate surface fluxesfrom remote sensing data.Keywords: evapotranspiration, operational application, remote sensing, simplified model,SVAT modelIntroductionDetailed knowledge of land surface fluxes, especially latent and sensible com-ponents, is important for monitoring the climate of land surface, for evalu-ating parameterization schemes in weather and climate models used to pre-dict fluxes exchanges between the surface and the lower atmosphere, andfor agricultural applications such as irrigation scheduling. The main meth-ods classically used to measure evapotranspiration (ET) are available at thefield scale (Bowen ratio, eddy correlation system, soil water balance), butdo not allow estimating the fluxes when dealing with large spatial scales.For operational applications, water managers and irrigation engineers need tohave accurate estimations of surface fluxes, and especially ET. Nowadays, therecommended FAO 56 method is used in numerous countries. This method224consists of estimating crop evapotranspiration (Etc) for a crop canopy usinga reference evapotranspiration (Etr) and a crop coefficient (Kc), where Etr isretrieved using the Penman–Monteith method (PM). The latter provides Etroveragrass under optimum soil moisture conditions with a constant value ofthe surface canopy resistance considering then the grass as a single big leaf(Allen et al., 1998, FAO 56 method). However, surface resistance can varyaccording to the day, the weather conditions, especially available radiationand vapor pressure deficit (Ortega et al., 2004). The determination of crop co-efficients is also debatable because a lot of factors occur (Neale et al., 2005).The ET crop surfaces under non-standard conditions is adjusted by either awater stress coefficient or modifying the Kc coefficient. Actual evapotranspi-ration (Etact) corresponds to the real water consumption according to weatherparameters, crops factors, management and environmental conditions. How-ever, several other crop and surface characteristics have to be considered: croptype/variety/development stage, ground cover and root system development.Remote sensing data with the increasing imagery resolution is a useful toolto provide such information over various temporal and spatial scales. Differentmethods havebeen developedto use this information in surface flux estimationschemes. It is always difficult to classify these methods, since their complexitydepends on the balance between the empirical and physically based modulesused. Nevertheless, we propose in this paper four model categories which arebased on:– Empirical direct methods where remote sensing data are introduced directlyin semi-empirical models to estimate ET (for example, the simplified rela-tionship using Thermal Infra Red (TIR) remote sensing and meteorologicaldata). We will present the main assumptions of this approach in the firstsection of the paper. It allows characterizing crop water status both at thelocal scale using ground measurements and over large irrigated areas usingsatellite data using the cumulative temperature difference (Ts −Ta), alsoknown as a stress degree day (SDD).– Residual methods of the energy budget combining some empirical rela-tionships and physical modules. Most current operational models (such asSEBAL, S-SEBI described further) use remote sensing directly to estimateinput parameters and ET.– Deterministic methods generally are basedon more complexmodels such asSoil–Vegetation–Atmosphere Transfer models (SVAT), which compute thedifferent components of energy budget (ISBA, Meso-NH). Remote sensingdata are used at different modeling levels, either as the input parameters tocharacterize the different surfaces, or in assimilation procedures which aimat retrieving adequate parameters for the ET computation. Some examplesof this approach will be shown in the third section.225– And lastly vegetation index methods,orinference methods based on the useof remote sensing to compute a reduction factor (such as Kc or PriestleyTaylor-alpha parameters) for the estimation of the actual evapotranspira-tion. These approaches consider a potential or reference ET obtained fromground measurements. Differents papers deal with these approaches in thisspecial issue (Allen et al., 2005; Neale et al., 2005; Garatuza et Watt, 2005).Before presenting these different approaches, a brief review about energybudget is required for a better understanding of the relationships betweenET and the driving variables such as surface temperature (Ts). Then we willdescribe some models using remote sensing to estimate ET. We note thatthis is not an exhaustive review since we have chosen to deal with widelyused models. For more details, one can refer either to overviews on the useof remote sensing for evapotranspiration monitoring such as proposed byKustas & Norman (1996), or to web sites such as: http://www.cgiar.org/iwmi.In conclusion, we will discuss about the application of these models for cropmonitoring and water management, present potentialities and limits, and onfuture remote sensing tools.Evapotranspiration and energy budgetEvapotranspiration estimation (corresponding to the