Location: Hydrology and Remote Sensing LaboratoryTitle: Seasonal evaluation of SMAP soil moisture in the U.S. corn belt
|WALKER, V.A. - Iowa State University|
|HORNBUCKLE, B. - Iowa State University|
Submitted to: Remote Sensing
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 11/1/2019
Publication Date: 11/1/2019
Citation: Walker, V., Hornbuckle, B., Cosh, M.H., Prueger, J.H. 2019. Seasonal evaluation of SMAP soil moisture in the U.S. corn belt. Remote Sensing. 11(21):248. https://doi.org/10.3390/rs11212488.
Interpretive Summary: Soil moisture remote sensing is difficult to meet accuracy goals for regions with high amounts of vegetation, such as cropland. A study was conducted in north central Iowa to understand error characteristics and to determine what variables may be responsible for errors between an in situ network and Soil Moisture Active Passive (SMAP) satellite products. It was shown that there are seasonal patterns associated with crop growth that violate assumptions made within the remote sensing algorithm. It was also determined that soil texture was not a significant influence on the more dynamic properties of the land surface such as roughness, vegetation, and temperature. The results of this study will help to refine satellite algorithms for overall improved performance.
Technical Abstract: NASA’s Soil Moisture Active Passive (SMAP) Level 2 soil moisture products are not meeting mission goals in the U.S. Corn Belt according to our seasonal evaluation conducted at a SMAP Core Validation Site in central Iowa. The single-channel algorithm (SCA) soil moisture products are too dry in early spring and late fall before and after crops are present, and too noisy in late spring and early summer when crops begin to grow. We investigated likely contributing factors. The climatology of vegetation’s effect on soil moisture retrieval in the SCA can differ by more than 14 days from what is retrieved by SMAP’s dual-channel algorithm (DCA). Soil and vegetation temperatures, assumed to be equal by all retrieval algorithms, are not: vegetation is about 2 K colder at 6 AM and about 2 K warmer at 6 PM. The effective temperature in version 2 products is too warm as compared to in situ soil temperatures. We propose a new effective temperature model that is consistent with observations, decreases the unbiased root-mean-square-error (ubRMSE) overall, and increases the coefficient of determination (R2) of the DCA in every month. However, some monthly dry biases increase to more than 0.10m3/m-3. The single-scattering albedo, w, has a significant impact on soil moisture retrieval. While the DCA has its lowest ubRMSE and highest R2 when w is non-zero, the SCA have their lowest ubRMSE and highest R2 when w = 0, and the dry bias of all algorithms increases as w increases. Errors in soil texture are not significant, but soil surface roughness should not be static and have a higher overall value. Our findings make it clear that a new retrieval algorithm that can account for changing soil roughness and vegetation conditions is needed.