Validation of SMOS Soil Moisture Products over the Maqu and Twente Regions

TLDR

Validation of SMOS soil moisture products is essential to assess their inaccuracies and limitations, with factors such as RFI, inaccurate land cover, and undetected frozen soils potentially affecting retrieval accuracy. This study aimed to validate SMOS soil moisture products by comparing them with in‑situ data from Maqu, China, and Twente, Netherlands, in 2010. The authors compared SMOS soil moisture with in‑situ measurements from Maqu and Twente, while refining RFI filtering and retrieval inputs such as land surface temperature and land cover to enhance accuracy. SMOS soil moisture generally matched in‑situ seasonal patterns, but showed moderate correlation (R²≈0.55–0.51 for ascending, 0.24–0.41 for descending) and a systematic dry bias of ~0.13 m³/m³ in Maqu and ~0.17 m³/m³ in Twente for ascending passes.

Abstract

The validation of Soil Moisture and Ocean Salinity (SMOS) soil moisture products is a crucial step in the investigation of their inaccuracies and limitations, before planning further refinements of the retrieval algorithm. Therefore, this study intended to contribute to the validation of the SMOS soil moisture products, by comparing them with the data collected in situ in the Maqu (China) and Twente (The Netherlands) regions in 2010. The seasonal behavior of the SMOS soil moisture products is generally in agreement with the in situ measurements for both regions. However, the validation analysis resulted in determination coefficients of 0.55 and 0.51 over the Maqu and Twente region, respectively, for the ascending pass data, and of 0.24 and 0.41, respectively, for the descending pass data. Moreover, a systematic dry bias of the SMOS soil moisture was found of approximately 0.13 m(3)/m(3) for the Maqu region and 0.17 m(3)/m(3) for the Twente region for ascending pass data. Several factors might have affected the retrieval accuracy, such as the presence of Radio Frequency Interference (RFI), the use of inaccurate land cover information and the presence of frozen soils not correctly detected in winter. Improving the RFI filtering method and the quality of the retrieval algorithm inputs, such as land surface temperature and land cover, would certainly improve the accuracy of the retrieved soil moisture.