GNSS-R Soil Moisture Retrieval Method Based on Transformer Fusion Network

Global Navigation Satellite System-Reflectometry (GNSS-R) remote sensing technology, with its advantages of low cost, short revisit cycle, and high-precision positioning, has been widely applied in soil moisture retrieval. To improve the accuracy of soil moisture retrieval, this paper proposes a Transformer-based fusion network (Transformer Fusion Network, TF-NET). TF-NET leverages a multi-head self-attention mechanism to extract deep features from GNSS-R observation data and auxiliary data, and enhances the feature correlation between different data sources through a cross-attention mechanism, thereby achieving high-precision soil moisture prediction. In this study, the Yellow River Delta was selected as the experimental area, and the Cyclone Global Navigation Satellite System (CYGNSS) and Tianmu Satellite were chosen as the primary data sources. Due to the complementary spatial coverage of the two satellites, soil moisture was retrieved separately from each satellite data, and the results were spatially combined. In the validation process, soil moisture observation data from the Soil Moisture Active Passive (SMAP) satellite were used as reference data. Experimental results indicate that the joint use of CYGNSS and Tianmu satellite data significantly improves the accuracy and reliability of soil moisture retrieval.