Page Banner

United States Department of Agriculture

Agricultural Research Service

Title: Continental-Scale Evaluation of Assimilated Soil Moisture Retrievals From the Advanced Microwave Scanning Radiometer

item Bolten, John
item Crow, Wade
item Jackson, Thomas
item Zhan, X
item Reynolds, Curt

Submitted to: Geoscience and Remote Sensing Letters
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 6/24/2009
Publication Date: 3/1/2010
Publication URL:
Citation: Bolten, J.D., Crow, W.T., Jackson, T.J., Zhan, X., Reynolds, C.A. 2010. Continental-Scale evaluation of assimilated soil moisture retrievals from the advanced microwave scanning radiometer. Geoscience and Remote Sensing Letters. 3:57-66.

Interpretive Summary: This manuscript discusses the enhancing the U. S. Department of Agriculture (USDA) Foreign Agricultural Service (FAS) global crop assessment decision support system soil moisture forcing via the integration of spaceborne soil moisture estimates within an Ensemble Kalman Filter (EnKF) framework. USDA FAS crop yield forecasts affect decisions made by farmers, businesses, and governments by predicting fundamental conditions in global commodity markets. Regional and national crop yield forecasts are made by crop analysts based on the Crop Condition Data Retrieval and Evaluation (CADRE) Data Base Management System (DBMS). Soil moisture availability is a major factor impacting these forecasts and the CADRE DBMS system currently estimates soil moisture from a simple water balance model (the Palmer model) based on precipitation and temperature datasets operationally obtained from the World Meteorological Organization and U.S. Air Force Weather Agency. Baseline soil moisture estimates within the CADRE system are currently based on output from a modified Palmer two-layer soil moisture accounting model derived in the late-1970s. Input data required by the modified-Palmer two-layer soil moisture includes daily precipitation estimates and daily minimum/maximum temperature measurements. Precipitation and temperature data are based on ground meteorological station measurements from the World Meteorological Organization (WMO), and gridded weather data from the US Air Force Weather Agency (AFWA). Unfortunately, water balance model estimates of soil moisture suffer from errors in satellite precipitation products over land (used as model input) and difficulties with selecting appropriate model soil and vegetation parameters. These errors impact the credibility of Palmer model soil moisture predictions and reduce their ultimate value for crop yield forecasting. In order to reach the full potential of soil moisture data for crop condition assessment and agricultural yield forecasting, soil moisture estimates derived from water balance modeling must be effectively constrained with remotely-sensed soil moisture estimates. We have developed a new soil moisture analysis product based on the assimilation of Advanced Microwave Scanning Radiometer (AMSRE) soil moisture retrievals into the modified Palmer model. We evaluate the integrated soil moisture product over the conterminous United States within a data denial experiment. Our specific data denial methodology compares soil moisture results obtained from three separate two-layer Palmer model simulations: 1) a “benchmark” run based on forcing the model using reliable ground-based precipitation, 2) an “open loop” Palmer model run forced by relatively less reliable satellite-based rainfall accumulations, and 3) an “EnKF” loop derived by assimilating AMSRE soil moisture retrievals into the open loop run. Gauge-corrected AFWA precipitation is used for the benchmark run and uncorrected, real-time precipitation estimates provided by the Tropical Rainfall Measuring Mission (TRMM 3B40RT) are used for the open loop and EnKF simulations. The application of the EnKF to assimilate AMSRE soil moisture retrievals is evaluated based on how efficiently it transforms low accuracy, open loop results (generated with the least accurate rainfall product) to match benchmark results (generated using the most accurate rainfall product). We demonstrate an increase in correlation between the EnKF and benchmark soil moisture time series from the application of the EnKF in both the surface and root zone layers.

Technical Abstract: Soil moisture is a fundamental data source used in crop growth stage and crop stress models developed by the USDA Foreign Agriculture Service for global crop estimation. USDA’s International Production Assessment Division (IPAD) of the Office of Global Analysis (OGA). Currently, the PECAD DSS utilizes a modification of the Palmer two-layer soil moisture model to estimate surface soil moisture. This model uses a simplified water balance scheme from precipitation and temperature data sources to provide secondary estimates of soil moisture and may be improved by the addition of direct observations of soil moisture. A data assimilation system has been designed to integrate surface soil moisture estimates from the EOS Advanced Microwave Scanning Radiometer (AMSR-E) with the online soil moisture model used by IPAD to provide nowcasts of crop conditions and agricultural-drought. The integration of AMSR-E observations into the two-layer soil moisture model employed by IPAD can potentially enhance the reliability of the IPAD soil moisture estimates due to AMSR-E’s improved repeat time and greater spatial coverage. Assimilation of the AMSR-E soil moisture estimates is accomplished using a 1-D Ensemble Kalman filter (EnKF) at daily time steps. Assessment of the AMSR-E assimilation has been completed for a five year duration over the conterminous United States. To evaluate the ability of the filter to compensate for incorrect precipitation forcing into the model, a data denial approach is employed by comparing soil moisture results obtained from separate model simulations forced with precipitation products of varying uncertainty. An analysis of surface and root-zone anomalies is presented for each model simulation over the conterminous United States, as well as statistical assessments for each simulation over various land cover types.

Last Modified: 06/28/2017
Footer Content Back to Top of Page