Stochastic Systems Group
Home Research Group Members Programs  
Demos Calendar Publications Mission Statement Alumni

SSG Seminar Abstract

Hydrologic Data Assimilation with the Ensemble Kalman Filter
Prof. Dennis McLaughlin

Ralph Parsons Laboratory
Dept.of Civil and Environmental Engineering

joint work with Rolf Reichle and Dara Entekhabi

Soil moisture controls the partitioning of atmospheric moisture and energy fluxes at the land surface and is a key variable in weather and climate prediction models. Passive microwave remote sensing provides a useful source of information about large-scale variations in near surface soil moisture, especially when combined with ground-based micrometeorological and soil texture data. We show how an Ensemble Kalman filter (EnKF) can be used to assimilate L-band (1.4 GHz) microwave radiobrightness observations into a land surface model which relates soil moisture to available measurements. We use an optimal smoother (a dynamic variational method) as a benchmark for evaluating the filter's performance. In a series of synthetic experiments we investigate the effect of ensemble size and non-Gaussian forecast errors on the filter's estimation accuracy. With a state vector dimension of 4608 and a relatively small ensemble size of 100 (500), the top node saturation error at the final update time is reduced to 0.35 (0.23) of the value obtained from an open loop simulation with no assimilation (as compared to 0.20 for the optimal smoother).

The ensemble distribution of soil moisture values is typically symmetric except under very dry or wet conditions, when the effects of nonlinearities in the state equation become significant. A comparison of the filter solution to the smoothing solution reveals that filter deviations from the optimal solution approach a non-zero lower bound for large ensemble sizes. Moreover, the ensemble-derived error standard deviations are consistently lower than the actual errors. This behavior suggests that the analysis (update) step is suboptimal. However, the degree of suboptimality (as compared to the optimal smoother) is relatively small. Overall, our synthetic experiments indicate that the Kalman filter is a flexible and robust data assimilation option which gives satisfactory estimates even for moderate ensemble sizes.

Problems with this site should be emailed to