Numerical Weather Prediction (NWP)-models require accurate simulation of land surface fluxes. Land surface models increasingly use complex descriptions of the physical mechanisms governing land surface processes. Consequently, these models may require large numbers of soil and land surface parameters controlling the vertical fluxes. Uncertainty in land surface parameterisation and the heterogeneous nature of landscape hydrological controls may lead to considerable uncertainty in predicted land surface fluxes and state variables such as soil moisture. In this paper, the predictive uncertainty associated with possible model parameterisations, and the role of calibration data, are reviewed. It is noted that spatial variability may lead to significant uncertainty when represented by one-dimensional models. As heterogeneity is an essential property of the land surface, surface parameterisation needs to adequately describe this heterogeneity. This paper discusses the use of remotely sensed thermal imagery with surface energy budget models for prediction of land surface fluxes. Particular emphasis is given to the role of remote sensing in assessing spatial variability in land surface processes, obtaining landscape fluxes and predicting soil moisture status. Recent studies have shown that remote sensing in combination with land surface modelling has significant potential in improving the estimation of fluxes and derived variables such as surface wetness. Remotely derived fluxes and surface wetness information may be assimilated in NWP-models providing better land surface boundary conditions as well as contributing to improved land surface models. This paper concludes with a brief discussion of recent work on the assimilation of satellite derived heating rates in regional atmospheric mesoscale models.