Abstract:
Data assimilation offers an attractive way to merge and interpret the increasing amount of hydrologic information provided by remote sensing and ground-based data sources. But a number of formidable conceptual and operational challenges need to be confronted before the potential of hydrologic data assimilation can be realized. These include 1) development of assimilation methods which are computationally feasible while making best use of all available information, 2) identification of realistic error models, 3) development of methods for identifying systematic model and measurement biases, 4) development of robust estimation algorithms that are not overly sensitive to outliers or erroneous model inputs, 5) identification of rigorous methods for quantifying the accuracy of data assimilation products. Two of the most promising options for large-scale hydrologic data assimilation are dynamic variational methods (4DVAR) and ensemble Kalman filtering (EnKF). This talk examines the distinctive features of each, compares their advantages and disadvantages, and discusses the prospects for dealing with the challenges listed above.