AN INTERCOMPARISON OF TECHNIQUES TO DETERMINE THE AREA-AVERAGED LATENT-HEAT FLUX FROM INDIVIDUAL IN-SITU OBSERVATIONS - A REMOTE-SENSING APPROACH USING THE EUROPEAN FIELD EXPERIMENT IN A DESERTIFICATION-THREATENED AREA DATA

A knowledge of the area-averaged latent heat flux [lambda E] is necess ary to validate large-scale model predictions of heat fluxes over hete rogeneous land surfaces. This paper describes different procedures to obtain [lambda E] as a weighted average of ground-based observations. The weighting coefficients are obtained from remote sensing measuremen ts. The remote sensing data used in this study consist of a Landsat th ematic mapper image of the European Field Experiment in a Desertificat ion-Threatened Area (EFEDA) grid box in central Spain, acquired on Jun e 12, 1991. A newly developed remote sensing algorithm, the surface en ergy balance for land algorithm (SEBAL), solves the energy budget on a pixel-by-pixel basis, From the resulting frequency distribution of th e latent heat flux, the area-averaged latent heat flux was calculated as [lambda E] = 164 W m(-2). This method was validated with field meas urements of latent heat flux, sensible heat flux, and soil moisture. I n general, the SEBAL-derived output compared well with field measureme nts. Two other methods for retrieval of weighting coefficients were te sted against SEBAL. The second method combines satellite images of sur face temperature, surface albedo, and normalized difference vegetation index (NDVI) into an index on a pixel-by-pixel basis. After inclusion of ground-based measurements of the latent heat flux, a linear relati onship between the index and the latent heat flux was established. Thi s relationship was used to map the latent heat flux on a pixel-by-pixe l basis, resulting in [lambda E] = 194 W m(-2). The third method makes use of a supervised classification of the thematic mapper image into eight land use classes. An average latent heat flux was assigned to ea ch class by using field measurements of the latent heat flux. Accordin g to the percentage of occurrence of each class in the image, [lambda E] was calculated as 110 W m(-2). A weighting scheme was produced to m ake an estimation of [lambda E] possible from in situ observations, Th e weighting scheme contained a multiplication factor for each measurem ent site in order to compensate for the relative contribution of that site to [lambda E], It was shown that [lambda E] derived as the arithm etic mean of 13 individual in situ observations leads to a difference of 34% ([lambda E] = 104 W m(-2)), which emphasizes the need for impro ved weighting procedures.