Skip to main content
ARS Home » Northeast Area » Beltsville, Maryland (BARC) » Beltsville Agricultural Research Center » Hydrology and Remote Sensing Laboratory » Research » Publications at this Location » Publication #312679
Title: Improving flood prediction by the assimilation of satellite soil moisture in poorly monitored catchments

Author
item ALVAREZ, C. - University Of Melbourne
item RYU, D. - University Of Melbourne
item Crow, Wade
item SU, CHUN-HSU - University Of Melbourne

Submitted to: Meeting Abstract
Publication Type: Abstract Only
Publication Acceptance Date: 12/15/2014
Publication Date: 1/1/2015
Citation: Alvarez, C., Ryu, D., Crow, W.T., Su, C.-H. 2015. Improving flood prediction by the assimilation of satellite soil moisture in poorly monitored catchments [abstract]. American Geophysical Union Fall Meeting Supplement. Abstract H13M-04.

Interpretive Summary:

Technical Abstract: Flood prediction in poorly monitored catchments is among the greatest challenges faced by hydrologists. To address this challenge, an increasing number of studies in the last decade have explored methods to integrate various existing observations from ground and satellites. One approach in particular, is the assimilation of satellite soil moisture (SM-DA) into rainfall-runoff models. The rationale is that satellite soil moisture (SSM) can be used to correct model soil water states, enabling more accurate prediction of catchment response to precipitation and thus better streamflow. However, there is still no consensus on the most effective SM-DA scheme and how this might depend on catchment scale, climate characteristics, runoff mechanisms, model and SSM products used, etc. In this work, an operational SM-DA scheme was set up in the poorly monitored, large (>40,000 km^2), semi-arid Warrego catchment situated in eastern Australia. We assimilated passive and active SSM products into the probability distributed model (PDM) using an ensemble Kalman filter. We explored factors influencing the SM-DA framework, including relatively new techniques to remove model-observation bias, estimate observation errors and represent model errors. Furthermore, we explored the advantages of accounting for the spatial distribution of forcing and channel routing processes within the catchment by implementing and comparing lumped and semi-distributed model setups. Flood prediction is improved by SM-DA (Figure), with a 30% reduction of the average root-mean-squared difference of the ensemble prediction, a 20% reduction of the false alarm ratio and a 40% increase of the ensemble mean Nash-Sutcliffe efficiency. SM-DA skill does not significantly change with different observation error assumptions, but the skill strongly depends on the observational bias correction technique used, and more importantly, on the performance of the open-loop model before assimilation. Our findings imply that proper pre-processing of SSM is important for the efficacy of the SM-DA and assimilation performance is critically affected by the quality of model calibration. We therefore recommend focusing efforts on these two factors, while further evaluating the trade-offs between model complexity and data availability.