Vegetation signal crosstalk present in official SMAP surface soil moisture retrievals

Successful surface soil moisture (SM) retrieval from space has been enabled by microwave satellite measurements of Earth's upwelling brightness temperature (TB). Nevertheless, correction for the impact of vegetation on TB emission remains a challenge for SM retrieval algorithms. Such correction is often performed in a simplified manner. For example, the Single Channel Algorithm (SCA) uses ancillary climatological normalized vegetation difference index values as a proxy for vegetation optical depth (z) - resulting in SM retrievals that do not account for interannual z variability. Official NASA Soil Moisture Active/Passive (SMAP) mission SM products are all based, to varying degrees, on the SCA. Here, we utilize an instrumental variable analysis and alternative SMAP SM retrievals derived from the Multi-Temporal Dual Channel Algorithm (MTDCA) - that better account for time variations in z - as a benchmark for examining SMAP Level 3 SM retrievals for the presence of signal crosstalk associated with the neglect of interannual z variability. Results suggest that failing to account for such variability introduces a spurious vegetation-based signal into monthly climatological SMAP SM anomalies. The SMAP Dual Channel Algorithm (DCA), which serves as the current SMAP baseline algorithm, reduces - but does not eliminate - this crosstalk. Results therefore suggest the need for caution when applying SMAP SM retrievals to science applications aimed at understanding SM coupling with the terrestrial biosphere. Plain language summary: Satellite observations of natural microwave emission from Earth's land surface can be converted into estimates of both surface soil moisture and vegetation water content. Such estimates have a variety of applications. However, the separation of the vegetation signal from the soil moisture signal is challenging and often performed using only approximate methods. This paper uses a novel approach to evaluate how accurately state-of-the-art soil moisture retrieval algorithms perform such partitioning. Results suggests that spurious vegetation signals remain in existing soil moisture products - with some approaches removing it more than others. Such "crosstalk" between soil- and vegetation-based signals limits the value of existing satellite soil moisture products for agricultural and ecohydrological applications and motivates the development of improved retrieval algorithms.