Download Powerpoint presentations (Landscape and Portrait) 

Options for Hydrologic Data Assimilation: Matching the Method to the Problem

Environmental data assimilation methods facilitate the interpretation of large amounts of diverse data, including non-traditional data obtained from remote-sensing. Most of these methods use models to enhance or improve the information provided by field measurements. Model-based data assimilation algorithms have the important advantage of being able to estimate variables which are not directly observable. Such algorithms have been used to map large-scale ocean and atmospheric circulation, to derive subsurface soil moisture profiles, and to locate petroleum reservoirs. In each case, models make it possible to estimate unobservable variables (e.g. subsurface soil moisture) from related observable quantities (e.g. microwave radiance). 

New remote sensing technologies and spatially distributed models offer great potential for improving our understanding of hydrologic processes, especially at large scales. Data assimilation will play a significant role in this enterprise. There is, however, much work to be done before operational data assimilation algorithms can be used to process all of the new data that will become available over the next few decades. In particular, we need to develop models and estimation approaches that are specifically designed to be used in data assimilation applications. These applications typically involve multiple data sources that observe different processes acting at different scales. The quantity of data to be processed will be much larger than has traditionally been the case in hydrology and much of this data will be of marginal value. Data assimilation algorithms must be sufficiently sophisticated to screen and interpret a continual stream of measurements in an efficient and physically meaningful way. 

This paper discusses some of the conceptual and design issues which hydrologists need to consider when applying data assimilation techniques. In particular, we examine some of the advantages and limitations of different approaches to data assimilation and we identify the methods that are most appropriate for particular problems. Methods to be discussed include variational assimilation, Kalman filtering (both traditional and ensemble filtering), multi-scale estimation, and heuristic techniques.