Snow fractional cover in land surface schemes

Land surface scheme forcing data sets are constructed using observations from the Seward Peninsula in Alaska. Non-parametric statistical techniques are successfully applied to fill data gaps in precipitation.; An extensive validation of the NCAR models, LSM and CLM, is undertaken using a variety of independent data such as turbulent fluxes measured via eddy correlation techniques. The models perform consistently compared to pervious studies, that is, CLM is a warmer dryer model than LSM. Several weaknesses regarding parameter assignment in the summer albedo, snow density and the lack of organic soil layers are identified in CLM. The additional physics in CLM are shown to perform better than the simpler LSM if the appropriate parameters are assigned.; Using synthetic data generated from observations, a spatially implicit comparison between the models and remotely sensed products is made. The aim is to assess the ability of the models to capture the albedo and fractional snow cover over a larger area than a point. The models cannot emulate the observed variability of the region, even that within vegetation classes. Part of this stems from the uniform view of vegetation as well as the ever problematic snow albedo parameterization. The MODIS and AVHRR albedo products are compared against each other and can demonstrate consistent performance suggesting they may provide a key to the fractional cover issue.; The implicit inclusion of a partial snow cover in the models is compared to explicit simulation of snow covered and snow free areas. Some differences are evident in the resulting mean surface temperature. The impact of non-linear behaviour in surface temperature with fractional cover was investigated through an upscaling exercise. Results suggest the models are largely scale invariant in off-line simulations.