CONTEXT

In agricultural activities, the decision-making process is central to agricultural system management and subsequent crop yield. As a powerful tool in field-specific decision-making processes, crop simulation models have the potential to simulate crop yields on a large scale. However, their performance is often biased by the spatial heterogeneity of environment and management factors when applied over a large scale.

OBJECTIVE

The major objectives of this study include: (1) Predicting and evaluating the annual yields of dominant crops with real rotation scenarios; (2) Locating fields with low crop yield and determining possible reasons; and (3) Evaluating the improvement for crop yield with different management strategies.

METHODS

This study proposed a crop yield simulation framework at the regional level by coupling a cropping system model (CropSyst) with a geographic information system (QGIS) to provide more reliable information for the decision-making process. In the study of a cropland concentrated USGS sub-watershed (Hydrologic Unit Code: 031402030101) in Geneva County, Alabama, we estimated the annual yields of four regionally dominant crops (i.e., corn, cotton, soybean, and peanuts) from 2016 to 2018. Low yield fields were identified in the simulation results visualization. Moreover, four management strategies were tested at a field scale to improve annual yields.

RESULTS AND CONCLUSIONS

Overall, the simulated crop yields were significantly correlated with the recorded values (Pearson's r = 0.99). However, the performance of the regional model varied for different crops. The model achieved the best performance for soybean with a high index of agreement (0.93) and modeling efficiency (0.86). For cotton, the model achieved positive model efficiency (0.23) and a good index of agreement (0.59). For peanut and maize, the model fitted records well but not sensitive enough. According to the visualization of simulation results, we located fields with low yields. The low organic matter content and high sand percentage of the soil were the potential causes of the nitrogen deficiency, which leads to the low yield subsequently. In field scale tests, four proposed management strategies could increase the cotton yields as high as 74.4%. But some strategies would also increase greenhouse gas emissions at the same time.

SIGNIFICANCE

This study bridges the gap between local cropping system models and the regional estimation of crop yields. The GIS-based crop simulation framework developed here demonstrates the potential of cropping system models to provide reliable information at a regional scale and hence significantly broadens their application in the agricultural decision-making process.

中文翻译：

---

基于 GIS 的地方农业决策和区域作物产量模拟框架

语境

在农业活动中，决策过程对农业系统管理和随后的作物产量至关重要。作为特定领域决策过程中的强大工具，作物模拟模型具有大规模模拟作物产量的潜力。然而，当大规模应用时，它们的性能往往会受到环境和管理因素的空间异质性的影响。

客观的

本研究的主要目标包括： (1) 在真实轮作情景下预测和评估优势作物的年产量；（二）定位农作物产量低的田地，确定可能的原因；(3) 评估不同管理策略对作物产量的提高。

方法

本研究通过将种植系统模型 (CropSyst) 与地理信息系统 (QGIS) 耦合，提出了区域层面的作物产量模拟框架，为决策过程提供更可靠的信息。在对阿拉巴马州日内瓦县农田集中的 USGS 子流域（水文单位代码：031402030101）的研究中，我们估算了 2016 年至 2018 年四种区域优势作物（即玉米、棉花、大豆和花生）的年产量. 在模拟结果可视化中识别出低产田。此外，在田间规模上测试了四种管理策略，以提高年产量。

结果和结论

总体而言，模拟作物产量与记录值显着相关（Pearson's r = 0.99）。然而，区域模型的性能因作物不同而异。该模型以高一致性指数 (0.93) 和建模效率 (0.86) 获得了大豆的最佳性能。对于棉花，该模型实现了正模型效率 (0.23) 和良好的一致性指数 (0.59)。对于花生和玉米，模型拟合记录很好，但不够灵敏。根据模拟结果的可视化，我们定位了低产田。土壤有机质含量低、含砂率高是造成缺氮的潜在原因，进而导致产量低下。在田间规模试验中，提出的四种管理策略可将棉花产量提高至 74.4%。但某些策略也会同时增加温室气体排放。

意义

这项研究弥合了当地种植系统模型与区域作物产量估算之间的差距。这里开发的基于 GIS 的作物模拟框架展示了种植系统模型在区域范围内提供可靠信息的潜力，从而显着拓宽了它们在农业决策过程中的应用。