Satellite fusion improves global soil moisture mapping

A new study published July 8 in Satellite Navigation shows how combining reflected navigation-satellite signals from Tianmu-1 and Fengyun-3 can improve global soil moisture monitoring. The approach could sharpen drought, flood and irrigation decisions by filling gaps in a measurement that is hard to track continuously worldwide.

Why it matters: - Soil moisture affects crop growth, drought formation, flood risk and land-atmosphere exchanges of water and heat. - Reliable global monitoring remains difficult because ground sensors are sparse and satellite products face limits from clouds, vegetation, cost and coarse resolution. - Better soil moisture maps could improve drought early warning, flood forecasting, irrigation planning, water-resource management and climate-risk analysis.

What happened: - Researchers from Wuhan University published the study on July 8, 2026, in Satellite Navigation. - The team built a dual-branch attention-fusion Transformer to combine Level-1 GNSS-R observations from Tianmu-1 and Fengyun-3. - The study used reflected navigation signals to retrieve global soil moisture with stronger continuity and robustness across changing land-surface conditions. - The paper is available through the original source.

The details: - The model gridded TM-1 and FY-3 observations to the 36-km EASE-Grid 2.0 using surface reflectivity as the main satellite observable. - The framework added auxiliary inputs from SMAP surface roughness and temperature, MODIS NDVI, GTOPO30 DEM, and SoilGrids clay and silt content. - Two mission-specific branches learned TM-1 and FY-3 features separately before an attention module balanced the cross-mission information. - A Transformer then modeled temporal dependencies in the fused dataset. - The combined TM-1 + FY-3 record delivered 79.7% average global monthly temporal coverage. - Against SMAP references, the model reached a correlation coefficient of 0.88 and RMSE of 0.053 m³/m³. - Independent validation with International Soil Moisture Network measurements produced a correlation of 0.67 and ubRMSE of 0.041 m³/m³. - Extended Triple Collocation analysis showed a correlation of 0.75 and random error standard deviation of 0.030 m³/m³. - The model performed best in arid and sparsely vegetated regions, where reflected signals more directly capture surface moisture changes. - The work was supported by the National Natural Science Foundation of China and the Funds for Creative Research Groups of Hubei Province.

Between the lines: - The study argues that multi-mission GNSS-R fusion is not just about adding more satellite observations. - The real gain comes from teaching the model to recognize how each mission senses the land surface differently. - That matters because orbit, geometry and signal response vary across satellites, and simple fusion can blur those differences. - The authors say the attention mechanism helps the system use each mission more effectively under changing vegetation, climate and land-cover conditions.

What's next: - The researchers said future work may add more GNSS-R missions to the framework. - The team also pointed to uncertainty-aware fusion as a next step. - Validation will need to expand in tropical, Asian, African and Oceanian regions. - More GNSS-R constellations could help reduce spatial and temporal gaps in global hydrology monitoring. - The approach may complement conventional microwave missions and use relatively low-cost receivers on low Earth orbit platforms.