Thermal remote sensing of the land surface for numerical weather prediction models

Soil-vegetation-atmosphere transfer (SVAT) models provide the lower boundary for numerical weather prediction (NWP) models and general circulation models (GCM). Typically, these models are not parameterized with reference to actual measured land surface behaviour, but have parameters specified according to approximate and simplified concepts of land surface type. Spatial variability of processes and fluxes coupled with uncertainty in input rainfall mean that significant uncertainty should be associated with land surface models used in NWP models. In this paper, an assimilation approach-thermal remote sensing-is advocated; whilst subject to some uncertainty, it may be usefully employed to update land surface models, addressing issues of both parameter and input rainfall uncertainty.