不同时间尺度太阳辐射数据对作物生长模型的影响(英)

逐日太阳辐射数据是作物模拟模型的重要输入参数之一。然而,在很多情况下,候、旬、月尺度的辐射信息相对容易获取。该文利用长时间序列(1961-2000)逐日太阳辐射数据,分别建立研究区候、旬、月不同时间尺度太阳辐射数据库,利用两个常用的作物生长模型(CERES-Maize和CGOPGRO-Soybean),以逐日数据(太阳辐射和模拟结果)为基准,分别探讨在雨养和灌溉条件下,不同时间尺度太阳辐射数据对作物生长模型的影响。结果表明:在不同时间尺度下,模型的输出(花期和作物产量)都接近于基准值。总体来看,两个模型模拟的花期平均误差和平均相对误差均接近于0,均方根误差为3.5d;CERES-Maize模型的模拟产量低于基准值,而CGOPGRO-Soybean的模拟结果高于基准值。在雨养和灌溉条件下,CERES-Maize的平均相对误差和均方根误差分别为-0.59%,120kg/hm2和-0.52%,129kg/hm2,CGOPGRO-Soybean的平均相对误差和均方根误差分别为5%,152kg/hm2和4.7%,165kg/hm2。短期数据误差(RMSE)是影响模型精度的主要因素。CGOPGRO-Soybean模型对不同时间尺度太阳辐射数据和水情信息比CERES-Maize模型敏感。当缺少逐日太阳辐射数据时,在雨养和灌溉条件下,候、旬、月尺度的太阳辐射数据都可以用于作物生长模型。

Sensitivity analysis of crop growth models to multi-temporal scale solar radiation

The records of daily solar radiation (Rs, MJ·m-2·d-1) are the important inputs for crop simulation models. However, for some model users, Rs at longer temporal intervals are more available than that at daily scale. The objective of this study was to analyze the sensitivity of simulated crop growth and production using CERES-Maize and GROPGRO-Soybean, two widely used crop growth models, to uncertainty in Rs at different time scales (5-day, 10-day, and monthly). Daily radiation data (1961-1990) from Vegetation/Ecosystem Modeling and Analysis Project (VEMAP) for the state of Georgia, USA were used to create 5-day, 10-day, and monthly mean daily Rs data sets. Datasets related to daily Rs were used as background baselines. The overall performance of the models was not significantly affected by Rs under the studied time scales. Within locations, the simulated days to anthesis and grain yields from 5-day, 10-day, and monthly Rs were close to that from daily Rs for maize and soybean under rainfed and irrigated conditions, respectively. Mean values of relative mean bias error (RMBE), mean bias error (MBE) and root mean square error (RMSE) of the simulated days to anthesis were 0, 0 and 3.5 d for the two crops under the studied scenarios, respectively. The simulated yields were underestimated for maize and overestimated for soybean using 5-day, 10-day, and monthly Rs for both rainfed and irrigated conditions, respectively. Under rainfed and irrigated conditions, the average RMBE and RMSE were -0.59%, 120 kg/hm2 and -0.52%, 129 kg/hm2 for maize yield, and 5%, 152 kg/hm2 and 4.7%, 165 kg/hm2 for soybean, respectively. Short-term bias in the difference between evaluated time scales and daily scale could affect the outputs of the crop models. Under the scenarios evaluated, CGOPGRO-Soybean model showed higher sensitivity to changes in multi-temporal Rs and water regimes than CERES-Maize model. Based on the results of this study, it can be concluded that 5-day, 10-day, and monthly mean daily Rs could be used as an input for crop growth simulation models when daily Rs are not available.