Uncertainty modelling and regional scale estimation of evapotranspiration: improving predictions using remotely sensed surface temperatures

The characterisation of the land surface within a modelling context is a complex problem. Issues related to the physical representation of system dynamics, the scale at which processes are modelled and the specification of appropriate parameter ranges all complicate the effective simulation of land surface behaviour. The problem of model uncertainty is particularly pertinent to hydrological investigations and recent efforts have been directed towards developing techniques to both assess and improve this issue. This thesis reports on an investigation into the efficacy of incorporating both remotely sensed and ground based measurements into a parsimonious land surface scheme to reduce model uncertainty and provide enhanced predictions of the energy balance components. Remote sensing has long been lauded as a means of bridging the point to regional scale gap and much research has focused on developing approaches to utilise information from this source. Examining the diurnal changes in land surface temperatures determined from satellite platforms offers a simple physical basis from which variations in the latent and sensible heat fluxes can be assessed. Land surface schemes are routinely calibrated to measured fluxes in an effort to improve model predictions. However, detailed records of surface fluxes are difficult to obtain as they require intensive field campaigns. Furthermore, surface flux observations are restricted to their measurement scale, making regional scale prediction difficult. Identifying alternative data sources with which to calibrate surface schemes would prove extremely useful, particularly at those scales beyond which field based equipment is capable of measuring. The techniques reported in this thesis provide a means of assessing the information content of calibration data and examine their ability to constrain model predictions. The ultimate aim of such approaches is to provide improved model characterisation of surface fluxes. Successful application of these methods would prove very useful, as it offers a means of moving beyond the point scale calibration processes inherent to surface flux measurements, to the intermediate and larger scales associated with grid based modelling and GCM applications.