This dataset is based on the annual rice map of Asia, integrating multiple predictive factors into three machine learning (ML) models to generate a seasonal rice yield dataset with high spatial resolution (4km) from 1995 to 2015 (AsiaRiceEild4km). Classify the predictive factors into four categories, consider the most comprehensive rice growth conditions, and determine the optimal ML model based on the inverse probability weighting method. The results showed that AsiaRiceYield 4km had good accuracy in estimating seasonal rice yield (single grain rice: R2=0.88, RMSE=920 kg · ha-1); Double cropping rice: R2=0.91, RMSE=554 kg · ha-1; Three grain rice: R2=0.93, RMSE=588 kg · ha-1). Compared with the SPAM model, the R2 of 4km rice yield in Asia increased by an average of 0.20, and the RMSE decreased by an average of 618 kg · ha-1. Especially, constant environmental conditions, including longitude, latitude, altitude, and soil properties, contribute the most (~45%) to the estimation of rice yield. At different growth stages of rice, the predictive factors of the reproductive stage have a greater impact on rice yield prediction than those of the nutritional stage and the entire growth stage. This dataset is a new type of long-term gridded rice yield dataset that can fill the gap of high spatial resolution seasonal yield products in major rice producing areas and promote relevant research on global agricultural sustainable development.
By collecting comprehensive rice area maps, rice yields from 1400 administrative units (the minimum administrative scale unit for each country with rice paddies), leaf area index information (from remote sensing products), and rice yield growth environmental conditions (location, time, soil, and climate), rice yields were estimated.
Based on the annual rice map of Asia, multiple predictive factors were integrated into three machine learning (ML) models to generate a seasonal rice yield dataset with high spatial resolution (4km) from 1995 to 2015 (AsiaRiceHead4km).
The data quality is good.
This work is licensed under
CC BY 4.0 (Creative Commons Attribution 4.0 International License).
CSTR:11738.11.NCDC.ZENODO.DB4057.2023
10.5281/zenodo.6901968
