8 Evaporation
169
authors concluded that these findings were likely due to the large differences in
T rad between soil and vegetation, rather than to the land surface energy balance.
8.3.5 Soil Vegetation Atmosphere Transfer (SVAT) models [4]
This approach was rather popular in the early years of thermal infrared remote
sensing and heat-balance studies. Use of SV AT models requires a detailed and
realistic description ofland surface - boundary layer interactions. Initially numerical models were used to establish case-specific look-up tables to estimate evaporation from a pair of day-night observations of surface radiometric temperature
(e.g. Soer, 1977 and 1980; Rosema and Bijleveld, 1977; Carlson and Boland,
1978). Camillo et aI. (1986) described the use of a SV AT model to estimate soil
hydrologic properties by calibrating the calculated soil brightness temperature
against remote measurements. Using this method, values were estimated for hydraulic conductivity, matric potential and soil moisture at saturation, and a soil
texture parameter. An inversion algorithm to infer heat fluxes at the land surface
was validated by Taconet and Vidal Madjar (1988). Their SVAT model was
driven by satellite (METEOSA T and A VHRR) thermal infrared measurements and
weather observations while ground based measurements at the same spatial resolution as the satellite data were used for validation. The performance of the inversion was improved by using a boundary layer model to characterize the surface
layer instead of ground-based weather observations.
Difficulties with the determination of all model variables led to a trade-off. Area
estimates of evaporation were obtained (e.g. Moran and Jackson, 1991) with
measurements of surface reflectance, temperature and normalized difference
vegetation index (a measure for the amount of green vegetation) and ground measurements of all the remaining land surface variables.
The SV AT used did not describe flow processes in the atmospheric boundary
layer. The near surface air layer was characterised using ground measurements of
air temperature, humidity and wind at a reference height (e.g. 2m) to solve the
system of heat transfer equations. The difficulty with this approach is that both air
temperature and humidity are strongly coupled with the surface. Vieira and Hatfield (1984) for example proposed to estimate surface temperature from air temperature.
Almost twenty years of research in this direction seem to indicate that estimation
of evaporation by model inversion is severely limited by the significant requirements for ancillary data and by the relatively poor accuracy of remote measurements of radiometric surface temperature (Feddes et aI., 1993; Bastiaanssen et aI.,
1994). Majumdar (1991) concluded that field experiments were needed to obtain
satisfactory, site-specific estimates of model variables for further use in evaporation mapping. Taconet et aI. (1995) used an advanced model to estimate evaporation from a single measurement of surface radiometric temperature at midday. Independent,measurements of soil water content, of vegetation density and of canopy
height were required. Once the accuracy of estimates is improved, however, SV AT
169
authors concluded that these findings were likely due to the large differences in
T rad between soil and vegetation, rather than to the land surface energy balance.
8.3.5 Soil Vegetation Atmosphere Transfer (SVAT) models [4]
This approach was rather popular in the early years of thermal infrared remote
sensing and heat-balance studies. Use of SV AT models requires a detailed and
realistic description ofland surface - boundary layer interactions. Initially numerical models were used to establish case-specific look-up tables to estimate evaporation from a pair of day-night observations of surface radiometric temperature
(e.g. Soer, 1977 and 1980; Rosema and Bijleveld, 1977; Carlson and Boland,
1978). Camillo et aI. (1986) described the use of a SV AT model to estimate soil
hydrologic properties by calibrating the calculated soil brightness temperature
against remote measurements. Using this method, values were estimated for hydraulic conductivity, matric potential and soil moisture at saturation, and a soil
texture parameter. An inversion algorithm to infer heat fluxes at the land surface
was validated by Taconet and Vidal Madjar (1988). Their SVAT model was
driven by satellite (METEOSA T and A VHRR) thermal infrared measurements and
weather observations while ground based measurements at the same spatial resolution as the satellite data were used for validation. The performance of the inversion was improved by using a boundary layer model to characterize the surface
layer instead of ground-based weather observations.
Difficulties with the determination of all model variables led to a trade-off. Area
estimates of evaporation were obtained (e.g. Moran and Jackson, 1991) with
measurements of surface reflectance, temperature and normalized difference
vegetation index (a measure for the amount of green vegetation) and ground measurements of all the remaining land surface variables.
The SV AT used did not describe flow processes in the atmospheric boundary
layer. The near surface air layer was characterised using ground measurements of
air temperature, humidity and wind at a reference height (e.g. 2m) to solve the
system of heat transfer equations. The difficulty with this approach is that both air
temperature and humidity are strongly coupled with the surface. Vieira and Hatfield (1984) for example proposed to estimate surface temperature from air temperature.
Almost twenty years of research in this direction seem to indicate that estimation
of evaporation by model inversion is severely limited by the significant requirements for ancillary data and by the relatively poor accuracy of remote measurements of radiometric surface temperature (Feddes et aI., 1993; Bastiaanssen et aI.,
1994). Majumdar (1991) concluded that field experiments were needed to obtain
satisfactory, site-specific estimates of model variables for further use in evaporation mapping. Taconet et aI. (1995) used an advanced model to estimate evaporation from a single measurement of surface radiometric temperature at midday. Independent,measurements of soil water content, of vegetation density and of canopy
height were required. Once the accuracy of estimates is improved, however, SV AT