Relations between radiative surface temperature (T(R)) and visible and near-infrared reflectances expressed as the normalized difference (ND) from a Landsat Thematic Mapper scene were analysed to study the heat balance of agriculture and native evergreen forests in southeastern Australia. The scene was at 0922 h (local time) during late spring (15 October 1986) with phenology of winter annual species between flowering and grain filling. Factors determining the slope of, and residual scatter about, the T(R)/ND relationships were analysed using a coupled two-layer soil-vegetation model of the surface heat balance. Inverse linear relationships were found between T(R) and ND for agriculture, but not for forests. This was a result of a wide range of ND and T(R) values in agricultural areas caused by wide variations in fractional vegetation cover. The relationships between ND and fractional vegetation cover was not general, thus, when forest and agricultural data were combined, the lower near-infrared reflectance of forests (16.9 per cent) compared with agricultural crops (29.0 per cent) resulted in forest data falling below the regression line for agriculture. From the agricultural relations, 60 to 70 per cent of the variation in T(R) was explained by ND, indicating that fractional vegetation cover was the dominant factor determining T(R). Residual variability of T(R) was attributed to spatial variability in ambient temperature, rates of soil evaporation and variations in stage of phenological development affecting stomatal resistance. Energy-balance analysis of the effect of soil water availability on the slope of the T(R)/ND relationship indicated opposite effects depending on whether the reduction in evaporation was from soil or from vegetation. Thus, the generality of using the slope of the T(R)/ND relationship to predict surface resistance to evaporation may be limited. It was concluded that extraction of information contained in T(R) and ND data for regional estimation of evaporation requires the separation of T(R) into soil and vegetation temperatures and an alternative to ND that relates more generally to fractional vegetation cover.
