Validation and expansion of the soil moisture index for assessing soil moisture dynamics from AMSR2 brightness temperature

Long-term remotely sensed soil moisture (SM) data is essential for understanding the land-atmosphere hydrological and energy interactions at both local and global scales. Passive microwave SM retrieval remains challenging at the global scale, especially in areas with complex terrain conditions, due to the difficulties in acquiring accurate land surface parameters (e.g., vegetation and surface roughness) across large extents and the uncertainties caused by the assumptions associated with current retrieval algorithms. This study addresses these challenges by providing a comprehensive evaluation of satellite SM products and introducing Soil Moisture Index (SMI)-based indicators derived from Advanced Microwave Scanning Radiometer 2 (AMSR2) brightness temperature (TB). It is noteworthy that SMI, recently proposed and showing promising results in the low-frequency L-band, lacks validation for high-frequency microwave observations. This research fills this critical gap by evaluating the emissivity-based soil moisture index (ESMI) utilizing C- and X-band TB. To mitigate potential uncertainties in land surface temperature (LST) data used in emissivity-based methods, we propose an alternative brightness temperature-based soil moisture indicator (TBSMI). Simulation experiments demonstrated the effectiveness of TBSMI for capturing SM dynamics with a strong correlation coefficient of 0.95. TBSMI was then evaluated against in situ measurements from a total of 553 ground stations in 12 dense and 4 sparse SM networks worldwide under different climatic and environmental conditions from 1 April 2015 to 31 December 2017. Inter-comparisons were also made with two widely used AMSR2 SM products [i.e., the land parameter retrieval model (LPRM) product, and the Japan Aerospace Exploration Agency (JAXA) product], as well as with the ESMI. The results suggested that TBSMI exhibited the best performance with a mean R of 0.65 against in situ observations, followed by ESMI (mean R of 0.58), while LPRM and JAXA achieved lower correlation with mean R of 0.52 and 0.41 respectively. Specifically, TBSMI and ESMI retained a robust capability in densely vegetated areas, where LPRM and JAXA deteriorated sharply. Moreover, TBSMI demonstrated stable performance across diverse conditions, providing an accurate and robust alternative for monitoring SM. Our study highlights the unique advantages of the SMI approach in capturing SM dynamics under complex land surface conditions, and can be useful for diverse hydrological applications and climate change studies.