Global Crop Dataset-Rice (GCD-Rice)

The map is generated using high spatial resolution Landsat and Sentinel-1 datasets, based on a phenological approach known as time-weighted dynamic time warping and a random forest model. Validation based on 231121 field survey samples indicates that the average user accuracy, producer accuracy, and overall accuracy of GCD-Rice are 89.84%, 88.55%, and 89.24%, respectively.

Based on linear regression tests of statistical area and identified area, the overall average R² value, slope, and RMAE are 0.768, 0.874, and 0.346, respectively.This dataset is provided in GeoTIFF format with the spatial resolution of 30 m. There are a total of 16 primary folders, representing various countries, and the secondary folders represent the rice of each season. In total, there are 918 GeoTIFF files.classification system:0: non-rice1: rice