Automatic processing of multi-resolution data for use in water management and hydrologic modeling

Water management in irrigated areas could be improved by better understanding the demand side of the problem, i.e., by forecasting the crop water requirement at the farm scale. Since the nature of this problem is spatially distributed, distributed models should be used in order to approach a solution. Physically based distributed models are infamous for being data greedy. Also, the input data for distributed models in general comes at different spatial resolutions and is available at different frequencies. Discrepancy in spatial scale hinders the use of physically based distributed models. Therefore to address the problem at hand, there is a need to embed the scale reconciliation component in the hydrologic model itself. Data-driven models provide the flexibility to be integrated with a scale-reconciliation module. This work is a step forward to address the processing of multi-resolution data to be mapped onto a desired output space through a model that could be calibrated automatically.; The present work has three main steps: First, development of a real-time or "online" calibration method to serve as the automating power of the later developed algorithms; second, a downscaling algorithm for spatial data that possess certain spatiostatistical properties and that are required as inputs to an evapotranspiration downscaling model. This model is calibrated using the first algorithm. Finally, a forecasting-downscaling algorithm is developed to use the evapotranspiration (ET) data that comes from a global Land Surface Model (LSM) as a calibration benchmark to provide short-term forecasts of ET at fine spatial resolution. The last model is also calibrated automatically using the first algorithm.