Land surface temperature derived from the Along Track Scanning Radiometer

Terrestrial radiometric surface temperatures were recorded at multiple observation angles in conjunction with satellite overpasses of the Along Track Scanning Radiometer (ATSR) over two florally distinct regions of Zimbabwe. The experimental sites were, respectively, an open grassland savannah and a medium density natural woodland, each of which extended over some 1000 square kilometers of predominantly flat terrain. The principal surface components at each location were vegetation and soil in varying proportions. Infrared brightness temperatures were corrected for atmospheric effects using solutions to the radiative transfer equation with coefficients derived from local atmospheric radiosoundings. The empirical equations of Idso (1981) and the RADGEN radiative transfer model (Zavody et ai., 1995) were used to derive atmospheric emittance and transmittance for the in-situ and ATSR data respectively. Both in situ and ATSR derived radiative temperatures exhibited a pronounced angular variation over each surface, with differences upwards of 5°C between measurements at nadir and forward (55°) zenith angles. This effect has been attributed to the temperature differentials that exist within heterogeneous canopies, which typically display variations in vegetation cover related to the observation angle. A simple two component canopy architecture was coupled with a linear mixture model to partition the ensemble surface emission. The fractional vegetation cover was estimated using multi-angle radiative temperatures and was in excellent agreement with in situ estimates. Empirical equations were derived from the in situ data which related vegetation and soil temperatures at each location. These relationships were used to constrain the surface component temperature regime so that ATSR dual-angle radiative surface temperatures were sufficient to derive the vegetation and soil temperatures and fractional cover. The difference between canopy and soil temperatures separated using the dual-angle data was greater than 30°C at certain times of the year, and the modal standard deviation for all component temperature estimates was 3.2°C.