Genotype-by-Environment-by-Management (GxExM) interactions represent many unknowns for crop improvement programs, which hampers the development of improved varieties, especially for highly variable environments like those limited by rainfall. While breeding programs have traditionally used statistical tools to deal with these interactions, process-based crop modeling has recently become an alternative and powerful approach. Overall, while statistical methods remain the most optimal solution to deal with GxExM interactions when many production datasets across time and space are available from multi-environment trials (MET), in silico methods like crop modeling can be used if such data is lacking, or if MET data don’t cover the entire target region. Yet, despite several reviews on the potential uses of process-based modeling tools to aid such issues, their practical use in helping breeding programs is still in its infancy. After exposing the pros and cons of process-based modeling, this paper presents the step-by-step process that would allow breeding programs to harness this tool to help guide their breeding decisions. We also argue that the issue of GxExM interactions should be tackled in a co-construction process, involving breeders, agronomists, extensionists, and modelers from the beginning, and this would bring crop models one step closer to being used to help make plant breeding decisions.

Genotype-by-Environment-by-Management (GxExM) 相互作用代表了作物改良计划的许多未知因素，这阻碍了改良品种的开发，特别是对于受降雨限制的高度可变的环境。虽然育种计划传统上使用统计工具来处理这些相互作用，但基于过程的作物建模最近已成为一种替代且强大的方法。总体而言，虽然统计方法仍然是处理 GxExM 交互的最佳解决方案，但当跨时间和空间的许多生产数据集可从多环境试验 (MET)中获得时，计算机如果缺少此类数据，或者 MET 数据未覆盖整个目标区域，则可以使用作物建模等方法。然而，尽管对基于过程的建模工具在帮助此类问题方面的潜在用途进行了多次审查，但它们在帮助育种计划中的实际用途仍处于起步阶段。在揭示了基于过程的建模的优缺点之后，本文介绍了允许育种计划利用该工具来帮助指导其育种决策的分步过程。我们还认为，GxExM 相互作用的问题应该在一个共建过程中解决，从一开始就涉及育种者、农学家、推广者和建模者，这将使作物模型更接近用于帮助做出植物育种决策.