A Harmonized District-Level Crop Yield and Climatology Panel for India, 1997–2022

District-level crop yield estimation in India is constrained by two persistent data problems: administrative boundary instability (over 120 new districts created between 1997 and 2022) and the misalignment of climate and satellite predictors with these shifting administrative units. This dataset resolves both issues, combining Directorate of Economics and Statistics (DESAgri) crop yield records for 15 major crops with monthly ERA5-Land reanalysis, CHIRPS precipitation, and MODIS vegetation indices, aggregated to a consistent set of 666 Indian districts over a 26-year period (1997–2022). Historical district splits and renames were reconciled to stable administrative boundaries using an area-weighted cropland allocation procedure, and all climate variables were cropland-masked using MODIS land cover classification to isolate agriculturally relevant signal. The release includes: Fifteen crop-level yield panels (area, production, yield) A monthly district-level climate panel (precipitation, temperature, soil moisture, NDVI/EVI), 1997–2022 A fully documented district-name harmonization crosswalk (102 mapping entries) A complete data dictionary Validation evidence confirming key-matching integrity and baseline predictive usability Full methodology and technical validation are documented in the accompanying Data Descriptor manuscript (included in this release) and the companion modeling study, TerraYield (Chaturvedy & Gururajan, in preparation). Code: https://github.com/AadityaChaturvedy/TerraYield-ML