Mapping annual 30-m paddy rice yield for different cropping systems in mainland Southeast Asia from 2001 to 2021

The gridded yield at 30-m resolution from 2001 to 2021 in mainland Southeast Asia (MSEA) for paddy rice of single-, double-, and triple-cropping systems is mapped (MSEA-RiceYield30m) by integrating multi-source data to address limitations of existing datasets characterized by coarse resolution and discontinuous time spans. MSEA-RiceYield30m is generated using machine learning models based on a series of features, which include soil properties, irrigation conditions, climate change, and environmental data. The model used for MSEA-RiceYield30m achieves high accuracy in estimating annual yield of paddy rice for different cropping systems, with mean values of MAE, RMSE, MAPE, and R2 of 142.43 kg/ha, 211.44 kg/ha, 4.57%, and 0.93, respectively. MSEA-RiceYield30m also performs well in validating paddy rice yield records, with an RMSE of 726.56 kg/ha in average when compared with sub-national units. Furthermore, in representative years, its accuracy surpasses other paddy rice yield datasets in MSEA, with a mean value of RMSE that is 1010.68 kg/ha. The dataset spans a relatively long time scale, and captures the full spatial extent of all paddy rice under all three different cropping systems, offering high spatial resolution and filling the gap in yield data across major rice distribution areas. MSEA-RiceYield30m had potential to sustainability with accessible link at https://doi.org/10.5281/zenodo.17149039.