From Drought Monitoring to Forecasting: A Combined Dynamical-Statistical Modeling Framework
Drought is the most costly hazard among all natural disasters. Despite the significant improvements in drought modeling over the last decade, accurate provisions of drought conditions in a timely manner is still one of the major research challenges. In order to improve the current drought monitoring and forecasting skills, this study presents a hybrid system with a combination of remotely sensed data assimilation based on particle filtering and a probabilistic drought forecasting model. Besides the proposed drought monitoring system through land data assimilation, another novel aspect of this dissertation is to seek the use of data assimilation to quantify land initial condition uncertainty rather than relying entirely on the hydrologic model or the land surface model to generate a single deterministic initial condition. Monthly to seasonal drought forecasting products are generated using the updated initial conditions. The computational complexity of the distributed data assimilation system required a modular parallel particle filtering framework which was developed and allowed for a large ensemble size in particle filtering implementation. The application of the proposed system is demonstrated with two case studies at the regional (Columbia River Basin) and the Conterminous United States. Results from both synthetic and real case studies suggest that the land data assimilation system significantly improves drought monitoring and forecasting skills. These results also show how sensitive the seasonal drought forecasting skill is to the initial conditions, which can lead to better facilitation of the state/federal drought preparation and response actions.