Cropping system models are deployed as valuable tools for informing agronomic decisions and advancing research. To meet this demand, early career scientists are increasingly tasked with building crop models to fit into these system modelling frameworks. Most, however, receive little to no guidance as to how to do this well. This paper is an introduction to building a crop model with a focus on how to avoid pitfalls, minimize uncertainty, and maximize value. We synthesized knowledge from experienced model builders and literature on various approaches to model building. We describe (1) what to look for in a model-building dataset, (2) how to overcome gaps in the dataset, (3) different approaches to fitting and testing the model, and (4) how to avoid common mistakes such as over-parameterization and over-fitting the model. The process behind building a crop model can be overwhelming, especially for a beginner, and so we propose a three-pronged approach: conceptualize the model, simplify the process, and fit the model for a purpose. We revisit these three macrothemes throughout the paper to instil the new model builder with the methodical mindset needed to maximize the performance and impact of their crop model.

作物系统模型被部署为有价值的工具，用于为农艺决策和推进研究提供信息。为了满足这一需求，早期职业科学家越来越多地承担着构建作物模型以适应这些系统建模框架的任务。然而，大多数人几乎没有收到关于如何做好这件事的指导。本文介绍了如何构建作物模型，重点是如何避免陷阱、最大限度地减少不确定性和最大化价值。我们综合了经验丰富的模型构建者的知识和关于模型构建的各种方法的文献。我们描述了 (1) 在模型构建数据集中寻找什么，(2) 如何克服数据集中的差距，(3) 拟合和测试模型的不同方法，以及 (4) 如何避免常见错误，例如过度参数化和过度拟合模型。构建裁剪模型背后的过程可能会让人不知所措，尤其是对于初学者而言，因此我们提出了一种三管齐下的方法：将模型概念化、简化过程以及使模型适合某个目的。我们在整篇论文中重新审视了这三个宏观主题，以向新模型构建者灌输有条不紊的思维方式，以最大限度地提高作物模型的性能和影响。