CERES-Maize区域应用效果分析

作物区域模拟是利用有限的空间数据，尽可能反映作物生育期、产量等性状的时空变化规律。本文尝试利用CERES-Maize模型进行区域模拟，并探讨区域应用的效果。结果表明，经区域校准后的CERES-Maize模型用于区域模拟时，基本可以反映出产量的变化规律，网格模拟产量与农调队调查产量的相对均方根差（RMSE％）为29．9％，符合度0．78。全国2144个网格71．3％的RMSE％在30％以内，其中RMSE％〈15％的为29．4％；就各区域而言，种植面积最大的玉米生态2区（占全国总面积的34％）效果最好，该区63．2％网格模拟的RMSE％小于15％。但区域模拟过程中还存在一系列误差，主要包括：模型本身的误差、作物品种遗传参数的误差、按一定区域范围归并的品种和管理参数引起的误差、区域划分引起的误差和空间数据的误差等，今后需要进一步校准和修正。

Performance of CERES-Maize in Regional Application

The primary purpose of regional simulation is to predict the crop yield spatially and temporally, with available data set. Performance of regional simulation and its uncertainties were analyzed, in the case of CERES-Maize model. The results show that the average relative root mean square error （RMSE） between simulated and census yield across the whole of China is 29.9% , with an agreement index of 0.78. Of all the 2144 simulation grids, 71.3% grids show RMSE less than 30% , and 29.4% of which with RMSE less than 15% . The performance is different among regions. The best performance occurs in AEZ 2, the largest maize cultivation areas taken 34% of maize cultivation area of China, where 63.2% grids have RMSE less than 15%. Considerable uncertainties exist in the regional simulation, including limitations of crop model, the aggregation of coefficient parameters and management practices to AEZ system, the rationality of definition of Agro-Ecological Zone system to present climate, and errors in dataset, etc.