Optimal model-based temperature inputs for global soil moisture and vegetation optical depth retrievals from SMAP

Abstract

The accuracy of global L-band soil moisture (SM) and vegetation optical depth (L-VOD) products retrieved through the τ-ω model is highly dependent on temperature inputs obtained from model-based temperature products. However, the performance of these temperature products in the retrieval of global-scale SM and L-VOD has not yet been evaluated. Therefore, this study aimed to evaluate four commonly used model-based temperature products as input to the SMAP-INRAE-BORDEAUX (SMAP-IB) algorithm for retrieving SM and L-VOD. Specifically, we investigated differences in SMAP-IB retrievals of SM and L-VOD using four model-based temperature sources as input, along with four configurations concerning the parameterization of effective soil (T_G) and vegetation (T_C) temperatures. Triple collocation analysis (TCA) results showed that SM retrievals based on GLDAS temperatures (SM_GLDAS), with T_C set to skin temperature and T_G calculated from shallow soil temperatures at layers 1 (0–10 cm) and 2 (10–40 cm), led to the highest global median TCA correlation (TCA-R) value of 0.780. In particular, SM_GLDAS achieved the highest TCA-R values over 34.94% of global pixels, predominantly in forested areas. Comparison with in situ measurements also showed improved regional performance of SM_GLDAS. In contrast, SM retrievals using MERRA2 temperature inputs, employing the same configurations for T_C but different soil temperature layers (1 (0–10 cm) and 4 (40–80 cm)) for T_G, yielded the lowest TCA-R value of 0.755. Overall, using the GLDAS temperature products as inputs to the retrieval algorithm resulted in the best performance for both SM and L-VOD retrievals. These new findings are valuable for selecting optimal model-based temperature datasets as inputs to the development of future satellite mission algorithms.