Data Assimilation of Soil Moisture in a Distributed Hydrologic Model: A Case Study over the Russian River Basin

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Tuesday, 6 January 2015: 2:00 PM
127ABC (Phoenix Convention Center - West and North Buildings)
Andrea R. Thorstensen, University of California, Irvine, CA; and P. Nguyen, K. Hsu, R. J. Zamora, and S. Sorooshian

Soil moisture is a key element to consider when simulating and predicting hydrologic events, as it is a variable closely connected to processes including evaporation, infiltration and runoff. However, measuring it at the proper spatial-temporal resolution for catchment scale applications remains a challenge. Given the measurement gap between global space-based measurements (1-3 day return, and a spatial resolution on the order of 10's of km), and extremely localized in-situ measurements (sub-hourly, and point profiles), effective assimilation of limited measurements in process models becomes an attractive option. This study considers the ability of top layer assimilation versus profile assimilation of soil moisture in order to investigate whether assimilating surface information alone is sufficient for improving the soil moisture state within a hydrologic model.

Soil moisture assimilation takes place at the daily scale in these experiments to mimic the frequency of space-based retrievals. For hydrologic simulations, the National Weather Service's Hydrology Laboratory Research Distributed Hydrologic Model (HL-RDHM) is used. HL-RDHM includes the conceptually-based Sacramento Heat Transfer with enhanced EvapoTranspiration (SAC-HTET) for rainfall-runoff production, which includes a scheme to transform the native conceptual storages to physically meaningful soil layers. This transformation allows for assimilation of soil moisture at various layers into HL-RDHM. For this study, the model is run at single pixels collocated with observation sites within northern California's Russian River Basin. These observation sites are operated by NOAA's Earth System Research Laboratory and provide point soil moisture profile measurements as well as the air temperature and precipitation measurements needed to force the model. HL-RDHM is run at an hourly scale, and the Ensemble Kalman Filter was selected for assimilation purposes. The experiments were run for the year 2013 for top layer only assimilation as well as multi-layer assimilation and the results were compared to an open (no assimilation) model run.

Results from this work suggest that even assimilation of top layer soil moisture alone may provide an improvement on the open run in recovering the soil moisture state, particularly in the early stages of simulation when unknown initial states are assumed. As the model was run with a-priori parameters derived from soil surveys, this work also indicates that when calibration via streamflow is difficult (as in the case of ungauged basins), data assimilation of soil moisture into a hydrologic model may aid in remediating model deficiencies caused by potentially incomplete soil surveys from which default parameters for HL-RDHM are made.