黑龙江大宗作物估产模型: 以水稻、大豆、玉米为例.

Establishing the yield estimation model of staple crops in Heilongjiang Province has a positive effect on grasping the agricultural situation of this province in time, and can provide a method reference for the yield estimation modeling in other regions. Firstly, the trend yield, agricultural climate indicators, remote sensing indicators, and planting fluctuation indicators were calculated. Then, yield estimation zones were established for the three crops ( soybean, corn, and rice), and modeling variables were selected using a combination of random forest and leave one method within each zone. Finally, the maximum, minimum, mean, and root mean square error of the mean absolute relative precision (MARE) of all models established based on the retention method are calculated to reflect the accuracy and stability of the final modeling. The results show that Heilongjiang Province is divided into 3, 3, and 4 yield estimating zones for soybeans, rice, and corn, respectively. The variable screening results show that the importance of trend yield > normalized difference vegetation index(NDVI) anomaly > accumulated temperature anomaly > precipitation Z-index. The MARE of the final models established for the three crops are all less than 10%, with the exception of corn yield estimation zone 1. The variance of MARE of all models is less than 0. 05% . The model established can accurately predict the crop yield under different agroclimatic conditions, and can provide a method reference for fine screening variables for yield estimation modeling. [ABSTRACT FROM AUTHOR]