When Does Soil-Specific Machine Learning Make Sense for SAR-Based Soil Moisture Mapping? Part2

In this follow-up video, I address a key question raised after my previous presentation on random forest soil-specific calibration for SAR-based soil moisture estimation: Is sandy loam truly special, or are the results just due to land cover, bare soil, or vegetation differences? I walk through detailed data checks—correlation analysis, LAI and bare soil percentages, and additional random forest model splits by SAR band and land cover—to show what’s really driving the high performance in sandy loam soils. I also discuss why neither bare soil nor vegetation alone can explain the results, and highlight the importance of soil texture for remote sensing-based soil moisture retrievals. Special thanks to Prof. Sekhar Muddu, the REWARD project, and my research team for their support. Watch till the end for insights into future work and next steps! If you missed the earlier video, check it out for a full background on the machine learning setup, feature engineering, and soil moisture data collection here: https://youtu.be/pV3QS_K8X28