Abstract Most crop models were developed and tested in homogeneous field conditions. However, these crop models are increasingly applied beyond the field scale for larger regions. Inadequate representation of the spatial variability at a larger scale introduces significant errors in the models’ predictions, yet attention to this topic is lacking. The selection of optimal crop models and their inputs when moving from the field to a regional scale must be performed carefully using strict guidelines while considering uncertainty propagation. This paper reviews crop modeling applications and their constraints in large-scale studies. The discussion focuses on the core issues that arise when applying crop models to a range of spatial scales: (i) parameterization and calibration of model inputs beyond field scale; (ii) constraints in the selection of model inputs at various scales; and (iii) retrieval and integration of remotely sensed crop variables into the crop model. Further, this review highlights cutting-edge approaches, namely scalable yield modeling, semi-empirical crop models, and global modeling initiatives, which can be used in a multi-scale assessment of agricultural systems.

中文翻译：

---

将作物模拟模型从田间规模升级到大区域的实践

摘要 大多数作物模型是在均匀的田间条件下开发和测试的。然而，这些作物模型越来越多地应用于更大区域的田间规模。在更大范围内对空间变异性的不充分表示会在模型的预测中引入重大错误，但缺乏对这一主题的关注。从田间转移到区域尺度时，必须使用严格的指导方针仔细选择最佳作物模型及其输入，同时考虑不确定性传播。本文回顾了作物建模应用及其在大规模研究中的限制。讨论的重点是将作物模型应用于一系列空间尺度时出现的核心问题：(i) 超出田间尺度的模型输入的参数化和校准；(ii) 在不同尺度上选择模型输入的限制；(iii) 遥感作物变量的检索和整合到作物模型中。此外，本综述重点介绍了可用于农业系统多尺度评估的前沿方法，即可扩展的产量建模、半经验作物模型和全球建模计划。