In general, remote sensing techniques cannot measure evaporation or evapotranspiration (ET) directly. However, remote sensing does have two potentially very important roles in estimating evapotranspiration. First, remotely sensed measurements offer methods for extending point measurements or empirical relationships, such as the Thornthwaite [1948], Penman [1948] and Jensen-Haise [1963] methods, to much larger areas, including those areas where measured meteorological data may be sparse. Secondly, remotely sensed measurements may be used to measure variables in the energy and moisture balance models of ET. Although, there has been progress made in the direct remote sensing of the atmospheric parameters which affect ET, such as the Raman LIght Detection And Ranging (LIDAR), this is essentially a ground based, point measurement and will not be covered in this report.
The question of how to use the spatial nature of remote sensing
data to extrapolate point ET measurements to a more regional scale
has been addressed in several ways. Using the temperature sounders
on the meteorological satellites in a linear regression model,
Davis and Tarpley [1983] estimated shelter temperatures with an
error of about 2
K for clear or partly cloudy conditions
Price [1982] used thermal data from the Heat Capacity Mapping
Mission (HCMM) to estimate regional scale ET rates which were found
to be comparable to pan evaporation data. Jackson [1985],
and Gash [1987] have proposed an analytical framework for
relating the horizontal changes in evaporation to horizontal
changes in surface temperature. Kustas et al [1990]
demonstrated these concepts for an agricultural area under clear
sky conditions. Humes et al, [1994] has proposed a simple
model using remotely sensed surface temperatures and reflectances
for extrapolating energy fluxes from a point to a regional scale;
however, other than for clear sky conditions, variations in
incoming solar radiation, meteorological conditions and surface
roughness limits this approach.
Several variables related to energy balance equation can be measured by remote sensing and simple meteorological measurements. Generally the latent heat term is determined as the residual of the other terms in the energy balance. Incoming solar radiation can be estimated from satellite observations of cloud cover, primarily from geosynchronous satellites [ Brakke and Kanemasu, 1981; Tarpley, 1979]. Pinker and Laszlo [1992] have proposed a model that infers incoming short wave fluxes and surface albedos from GOES data. Pinker et al [1994] used this model to demonstrate that incoming shortwave radiation can be measured quite accurately, even under variable cloud conditions, at the basin scale.
For clear sky conditions, the surface albedo may be estimated by measurements covering the entire visible and near infrared waveband, while empirical relations using narrow spectral bands can be used to determine albedo over heterogeneous surfaces [ Jackson, 1985; Brest and Goward, 1987]. Although albedo estimated this way is not the true hemispherical albedo [ Kimes at al, 1980], the lack of directional data or simple models to make this correction have not been developed.
Surface temperature can be estimated from measurements in thermal infrared wavelengths that is, the 10.5 to 12.5 micron waveband, either assuming a surface emissivity (close to unity for natural surfaces) or having measured values of the surface emissivity [ Wan and Dozier, 1989]. Surface temperatures can be used to estimate the outgoing long wave radiation term in the net radiation equation [ Kustas et al, 1994].
The soil heat flux term can be estimated with remote sensing measurements. A simplified approach defines the ratio of soil heat flux to net radiation in terms of vegetation cover which, in turn, is determined from visible and near infrared measurements [ Clothier et al, 1986, Choudhury et al, 1987, Kustas and Daughtry, 1990]. The diurnal effects [ Owe and van de Griend, 1990] and influence of soil moisture [ Brutsaert, 1982] are assumed to be secondary for large areas [ Kustas et al, 1994]
The sensible heat flux can be estimated using several approaches including the bulk resistance approach proposed by Monteith [1973] and similarity principles for the unstable boundary layer [ Brutsaert and Sugita, 1992], where the surface temperatures are measured by remote sensing. These approaches have met with varying degrees of success [ Hall et al., [1992], Brutsaert and Sugita, 1992, Brutsaert et al., 1993, Kustas et al., 1994].
One formulation of potential ET that lends itself to remote sensing inputs is that developed by Priestley and Taylor [1972]. The Priestley-Taylor equation is the basis for the model which Kanemasu et al. [1976], used for estimating ET with satellite data. Estimates of the net radiation from geostationary satellite data are used by Heilman et al. [1977] in a Priestley-Taylor type of equation to estimate ET.
Additional approaches for estimating ET from remote sensing data are being explored. Barton [1978] and Davies and Allen [1973] have modified this formula for evaporation from an unsaturated land surface by the surface layer soil moisture. Barton used airborne microwave radiometers to sense soil moisture remotely in his study of evaporation from bare soils and grasslands. Soares et al [1988] demonstrated how thermal infrared and C-band radar could be used to estimate bare soil evaporation. Choudhury et al, [1994] have shown strong relationships between evaporation coefficients and vegetative indices. Another approach being pursued is the development of numerical models that simulate the heat and water in the soil and drive it with the energy balance at the surface [ Camillo et al., 1983, Taconet et al., 1986]. Taconet and Vidal-Madjar [1988] have used this approach with Advanced Very High Resolution Radiometer (AVHRR) and Meteosat data.