Evaluation of land surface models using ground-based point-scale measurements

Numerical models developed to simulate land-surface processes for different applications have to be validated against observations. This dissertation consists of two separate model validation and evaluation studies conducted over different regions. Each of these studies focuses on the dominant land-surface processes under those particular climatic conditions.; The Project for Intercomparison of Land-surface Parameterization Schemes (PILPS) Phase 2(d) experiment at Valdai, Russia looks into the models' abilities to correctly simulate land-surface hydrology in a cold region where snow and frozen soil related processes dominate. Building on previous work, I use newly available observations to investigate the thermal and hydrological impacts of frozen soil on soil temperature, soil moisture, and runoff, both in observations and model simulations. This study shows that the thermal impact of frozen soil on soil temperature is important and models can have correct soil temperature simulations in cold regions only if they explicitly account for soil water freezing or thawing. The hydrological impact of frozen soil is inconclusive in this study. No evidence suggests that the existence of frozen soil stops infiltration in this region. I also found that hysteresis in snow cover fraction, which is absent in all current parameterizations, might have quite an impact on snow simulations, in addition to snow physics.; The evaluation of the North American Land Data Assimilation System (NLDAS) over the Southern Great Plains focuses on validating the NLDAS forcing and evaluating the land surface models during the warm season. This study has validated the high quality of NLDAS retrospective forcing and revealed the differences in the partitioning of water and energy between the models and observations. The inter-model variation is quite striking given that models are driven by the same atmospheric forcing with the same land surface specifications. Several additional experiments conducted in this study imply that models can be improved and the difference between models can be reduced with careful calibrations and some fine-tuning. This also illustrates the value of the evaluation work in that evaluation and diagnosis of models provides insights on where the sources for errors are and in which direction models can be improved.