当前位置: X-MOL 学术Int. J. Plant Prod. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Intercomparison and Performance of Maize Crop Models and Their Ensemble for Yield Simulations in Brazil
International Journal of Plant Production ( IF 2.1 ) Pub Date : 2019-10-31 , DOI: 10.1007/s42106-019-00073-5
Yury C. N. Duarte , Paulo C. Sentelhas

Maize yield prediction is of extreme importance for both identifying those locations with high potential for this crop and determining the yield gaps of the crop where it is currently produced. The most feasible way to estimate crop yields is with the use of crop simulation models, since well calibrated and evaluated. Even though, these estimations have uncertainties once the crop models are not complete. Recent studies have shown that crop models´ uncertainties can be reduced when several models are used together, in an ensemble. Considering that, this study aimed to calibrate and evaluate three crop simulation models (AEZ-FAO; DSSAT-CERES-Maize and APSIM-Maize) to estimate maize potential and attainable yields and to assess the performance of different ensemble strategies to reduce their uncertainties for maize yield prediction. Weather, soil and maize yield data from 79 experimental sites in Brazil were used for calibrating and evaluating these models. After that, the models showed only a good performance, with mean absolute errors (MAE) between 727 and 1376 kg ha −1 , R 2 between 0.49 and 0.79, d index between 0.78 and 0.94, and C index from 0.54 to 0.84. When the ensemble was applied, using the combination of two models (DSSAT-CERES-Maize and APSIM-Maize), the results showed a better performance than each single model or even the average of them, with MAE = 799 kg ha −1 , R 2 = 0.79, d = 0.94 and C = 0.84, allowing us to conclude that the ensemble of simulated maize yields is a good strategy to reduce uncertainties on simulations.

中文翻译:

巴西产量模拟中玉米作物模型及其组合的比对和性能

玉米产量预测对于确定具有该作物高潜力的位置和确定目前生产该作物的作物的产量差距都极为重要。估计作物产量最可行的方法是使用作物模拟模型,因为经过良好的校准和评估。尽管如此,一旦作物模型不完整,这些估计就具有不确定性。最近的研究表明,当多个模型一起使用时,可以减少作物模型的不确定性。考虑到这一点,本研究旨在校准和评估三种作物模拟模型(AEZ-FAO;DSSAT-CERES-Maize 和 APSIM-Maize),以估计玉米的潜力和可实现的产量,并评估不同组合策略的性能,以减少其不确定性。玉米产量预测。天气,来自巴西 79 个试验地点的土壤和玉米产量数据用于校准和评估这些模型。之后,模型仅表现出良好的性能,平均绝对误差 (MAE) 在 727 和 1376 kg ha -1 之间,R 2 在 0.49 和 0.79 之间,d 指数在 0.78 和 0.94 之间,C 指数从 0.54 到 0.84。当集成应用时,使用两个模型(DSSAT-CERES-Maize 和 APSIM-Maize)的组合,结果显示出比单个模型甚至它们的平均值更好的性能,MAE = 799 kg ha -1 , R 2 = 0.79、d = 0.94 和 C = 0.84,使我们能够得出结论,模拟玉米产量的集合是减少模拟不确定性的好策略。这些模型仅表现出良好的性能,平均绝对误差 (MAE) 在 727 和 1376 kg ha -1 之间,R 2 在 0.49 和 0.79 之间,d 指数在 0.78 和 0.94 之间,C 指数从 0.54 到 0.84。当集成应用时,使用两个模型(DSSAT-CERES-Maize 和 APSIM-Maize)的组合,结果显示出比单个模型甚至它们的平均值更好的性能,MAE = 799 kg ha -1 , R 2 = 0.79、d = 0.94 和 C = 0.84,使我们能够得出结论,模拟玉米产量的集合是减少模拟不确定性的好策略。这些模型仅表现出良好的性能,平均绝对误差 (MAE) 在 727 和 1376 kg ha -1 之间,R 2 在 0.49 和 0.79 之间,d 指数在 0.78 和 0.94 之间,C 指数从 0.54 到 0.84。当应用集成时,使用两个模型(DSSAT-CERES-Maize 和 APSIM-Maize)的组合,结果显示出比单个模型甚至它们的平均值更好的性能,MAE = 799 kg ha -1 , R 2 = 0.79、d = 0.94 和 C = 0.84,使我们能够得出结论,模拟玉米产量的集合是减少模拟不确定性的好策略。
更新日期:2019-10-31
down
wechat
bug