Validation of Advanced Microwave Scanning Radiometer Soil Moisture Products

Validation is an important and particularly challenging task for remote sensing of soil moisture. A key issue in the validation of soil moisture products is the disparity in spatial scales between satellite and in situ observations. Conventional measurements of soil moisture are made at a point, whereas satellite sensors provide an integrated area/volume value for a much larger spatial extent. In this paper, four soil moisture networks were developed and used as part of the Advanced Microwave Scanning Radiometer-Earth Observing System (AMSR-E) validation program. Each network is located in a different climatic region of the U.S., and provides estimates of the average soil moisture over highly instrumented experimental watersheds and surrounding areas that approximate the size of the AMSR-E footprint. Soil moisture measurements have been made at these validation sites on a continuous basis since 2002, which provided a seven-year period of record for this analysis. The National Aeronautics and Space Administration (NASA) and Japan Aerospace Exploration Agency (JAXA) standard soil moisture products were compared to the network observations, along with two alternative soil moisture products developed using the single-channel algorithm (SCA) and the land parameter retrieval model (LPRM). The metric used for validation is the root-mean-square error (rmse) of the soil moisture estimate as compared to the in situ data. The mission requirement for accuracy defined by the space agencies is 0.06 m <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3</sup> /m <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3</sup> . The statistical results indicate that each algorithm performs differently at each site. Neither the NASA nor the JAXA standard products provide reliable estimates for all the conditions represented by the four watershed sites. The JAXA algorithm performs better than the NASA algorithm under light-vegetation conditions, but the NASA algorithm is more reliable for moderate vegetation. However, both algorithms have a moderate to large bias in all cases. The SCA had the lowest overall rmse with a small bias. The LPRM had a very large overestimation bias and retrieval errors. When site-specific corrections were applied, all algorithms had approximately the same error level and correlation. These results clearly show that there is much room for improvement in the algorithms currently in use by JAXA and NASA. They also illustrate the potential pitfalls in using the products without a careful evaluation.