Assessing use of Vegetation Attribute from SAR to Improve Performance of the SMAP-Sentinel Active-passive High Resolution Soil Moisture Product

Abstract:

The SMAP project generates a high-resolution (3km) active-passive soil moisture product. This product is obtained by combining the SMAP L-band radiometer data and the Copernicus Sentinel-1A/1B C-band Synthetic Aperture Radar (SAR) data. Assessment of this high-resolution product was conducted and the result shows an unbiased root-mean-square-error (ubRMSE) of 0.05 m³/m³ in regions with low vegetation density (~< 3 kg m^-2). This high-resolution (3 km) soil moisture product is useful for agriculture applications. However, the assessment shows a high amount of errors in soil moisture retrievals over the agricultural sites. It is suspected that the vegetation attributes used during soil moisture retrievals were out of sync due to the use of the NDVI-based vegetation climatology. The Sentinel-1A/1B SAR data provides the cross-pol (VH) observations that carry vegetation attribute information. We hypothesize that by including the Sentinel-1A/1B cross-pol (VH) derived vegetation parameter (e.g., Vegetation Optical Density, tau) in the retrieval algorithm the accuracy of soil moisture over the agricultural domain will improve considerably. The SMAP-Sentinel active-passive algorithm does not exploit this valuable information inbuilt in the algorithm. As an alternative to NDVI, the cross-pol (VH) Sentinel-1A/1B measurements could be used as a variable that is empirically correlated to vegetation optical density (tau). We will establish the empirical relationship between the cross-pol (VH) backscatter and for different landcovers at a global extent using 3 years of Sentinel-1A/1B and the SMAP Multi-Temporal Dual-Channel Retrieval algorithm (MT-DCA) data. The developed empirical model will be used to predict the over the cropland and further used in the Tau-Omega model to retrieve soil moisture. Finally, soil moisture retrievals will be assessed against the Core Cal/Val sites to evaluate improvement in ubRMSE.