Soil Sight: Remote Sensing-Based Soil Fertility Dataset

Traditional soil testing in agricultural regions is often expensive, time-consuming, and lacks spatial coverage. This dataset addresses the need for a high-resolution, remote sensing-based alternative to enable faster, scalable, and data-driven soil fertility assessment. The CSV file contains 1,38,192 unique geo-reference points across a 123 km² agricultural region, with each 30 m × 30 m grid cell capturing multiple parameters relevant to soil health and fertility. These include: Latitude and Longitude (WGS 84 coordinates) to spatially locate each point; NDVI (Normalized Difference Vegetation Index), derived from Sentinel-2 bands B8 and B4, indicating vegetation health; EVI (Enhanced Vegetation Index), which improves sensitivity in high biomass regions; SAVI (Soil-Adjusted Vegetation Index), which minimizes soil brightness influence; Soil_Moisture, estimated using Sentinel-1 C-band SAR VV polarization backscatter values; Elevation, derived from SRTM 30 m Digital Elevation Model; and Fertility_Level, a categorical classification into Very High, High, Moderate, or Low fertility zones. All data was preprocessed using Google Earth Engine (2019–2025), applying radiometric calibration, atmospheric correction, and cloud filtering. This comprehensive dataset is ideal for use in precision agriculture, irrigation planning, site-specific fertilizer application, and machine learning-based agro-environmental modeling.