training_samples_zarr_v32_final

The dataset used in the crop yield prediction model (covering the contiguous United States) integrates NDVI, ERA5 meteorological data, and soil moisture data. It calculates indicators including monthly average temperature, monthly total precipitation, downward shortwave radiation, accumulated temperature, number of heat stress days, number of extreme rainfall days, length of drought periods, surface soil moisture (0–7 cm), and subsurface soil moisture (7–28 cm). For each indicator, three resolution channels are computed and stored to provide the model with a multi-scale receptive field.