High-resolution crop yield and water productivity dataset generated using random forest and remote sensing.

Area of Science:

• Agricultural Science
• Remote Sensing
• Environmental Science

Background:

• High-resolution crop yield and crop water productivity (CWP) datasets are essential for understanding agricultural production capacity.
• Existing datasets for key Chinese dryland crops like maize and wheat are insufficient.
• Spatiotemporal variations in crop production require accurate data for effective management.

Purpose of the Study:

• To generate and evaluate a long-term, 1-km resolution dataset of crop yield and CWP for maize and wheat across China.
• To assess the accuracy of MOD16 products for crop evapotranspiration estimation.
• To validate a novel yield estimation model at both local and regional scales.

Main Methods:

• Utilized multiple remotely sensed indicators and the random forest algorithm.
• Generated a 1-km resolution dataset for crop yield and CWP for maize and wheat.
• Evaluated MOD16 products against eddy covariance flux tower data for evapotranspiration accuracy.
• Validated the yield estimation model using local and regional data.

Main Results:

• MOD16 products demonstrated accuracy comparable to flux tower data for crop evapotranspiration (maize RMSE: 4.42 mm/8d, wheat RMSE: 3.81 mm/8d).
• The proposed yield estimation model achieved good accuracy at local (maize rRMSE: 26.81%, wheat rRMSE: 21.80%) and regional (maize rRMSE: 15.36%, wheat rRMSE: 17.17%) scales.
• Generated comprehensive spatiotemporal patterns of maize and wheat yields and CWP across China.

Conclusions:

• The developed datasets offer a valuable resource for agricultural research and management in China.
• MOD16 data serves as a reliable proxy for crop evapotranspiration in dryland conditions.
• The study provides critical insights for optimizing agricultural production strategies to ensure food security.