Oral or Poster

IAMAS

JMP02 - Machine Learning in atmospheric, ocean and earth-system prediction: forecasting, simulation and scientific analysis

Advancing Global Soil Moisture Estimation Using Multiple Satellite Sensors based on the Local Ensemble Transform Kalman Filter

Satellite-derived soil moisture remote sensing data provide broader spatial coverage than ground observations but are limited by spatiotemporal discontinuities due to the nature of polar-orbiting satellites. This study investigates whether integrating multiple satellite datasets can mitigate these limitations and enhance soil moisture estimation. A global soil moisture data assimilation system is developed using a physically based land surface model (LSM) and the Local Ensemble Transform Kalman Filter (LETKF). The system assimilates Level 2 soil moisture retrievals from the Soil Moisture Active Passive (SMAP), Soil Moisture and Ocean Salinity (SMOS), Advanced SCATterometer (ASCAT), and Advanced Microwave Scanning Radiometer 2 (AMSR2) satellites. The assimilation results are validated against in-situ soil moisture measurements across the U.S. from May to September for the years 2015?2020, assessing improvements before and after data assimilation. Among the single-sensor assimilation experiments, SMAP demonstrates the highest performance in improving temporal anomaly correlation for surface soil moisture compared to the LSM without data assimilation. The performance improvements achieved through data assimilation follow the order: SMAP, ASCAT, AMSR2, and SMOS. Multi-satellite data assimilation outperforms single-sensor assimilation at both the surface and root-zone levels, yielding higher anomaly temporal correlation and lower unbiased root-mean-square error (ubRMSE). A station-wise comparison of high- and low-performance improvement cases reveals that satellite data quality is the primary factor influencing assimilation performance, while the Kalman gain plays a secondary role. The findings of this study highlight the advantages of multi-sensor data assimilation in improving the spatiotemporal representation of soil moisture, demonstrating its potential for enhancing land surface modeling and hydrological applications.