Abstract
Rice is the most important staple food in Asia. However, high-spatiotemporal-resolution rice yield datasets are very limited over a large region. The lack of such products hugely hinders the studies on accurately assessing the impacts of climate change and simulating agricultural production. Based on dynamic rice maps in Asia, we incorporated four predictor categories into three machine learning (ML) models to generate a high-spatial-resolution (4 km) rice yield dataset (AsiaRiceYield4km) for main rice seasons from 1995 to 2015. Four predictor categories considered the most comprehensive rice growing conditions and the optimal ML model was determined for each rice season based on an inverse proportional weight method. The results showed that AsiaRiceYield4km has a good accuracy for seasonal rice yield prediction (single rice: R2 = 0.88, RMSE = 920 kg/ha, double rice: R2 = 0.91, RMSE = 554 kg/ha, and triple rice: R2 = 0.93, RMSE = 588 kg/ha). Compared with Spatial Production Allocation Model (SPAM), R2 of grided rice yields was improved by 0.20 and RMSE was reduced by 618 kg/ha on average for single rice. Particularly, constant environmental conditions including longitude, latitude, elevation, and soil properties contributed the most (~45 %) to rice yield prediction. As for different growing periods of rice, we found that the predictors in reproductive period had more impacts on rice yield prediction than those of the vegetative period and the whole growing period. AsiaRiceYield4km is a novel high-spatial-resolution gridded rice yield dataset that can fill the unavailability of seasonal yield products across major rice production areas and promote more relevant studies on agricultural sustainability in the world. AsiaRiceYield4km can be downloaded from an open-data repository (DOI: https://doi.org/10.5281/zenodo.6901968; Wu et al., 2022).
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have