Grain yield response to consumptive water use a.k.a. evapotranspiration (ET_c) is a critical dimension of food-water nexus in irrigated agroecosystems. With the largest water footprint, irrigated corn production in the U.S. must be assessed for its ET_c and water use efficiency (WUE) at a scale meaningful for producers and water managers. Field experiments significantly inform our understanding of how corn yield responds to ET_c at the field/plot-level, but their collective synergism remains untapped. This analysis uses a literature synthesis of existing measured data from nationwide experiments that measured corn ET_c, to build models that elucidate and predict corn ET_c and WUE under diverse environment and on-farm management. The resulting synthesis (n = 1,362) captured wide ranging conditions of mean aridity, environmental variability, management, and crop outcomes, providing an unprecedented opportunity to train data-driven models. Random forest models effectively predicted seasonal total ET_c and seasonal mean WUE using a handful of environmental and management-based predictors, demonstrating RMSEs of 33 mm (5.1%) and 0.28 kg m⁻³ (10.4%), respectively. Irrigation depth was the most important followed by total precipitation received during the growing season in determining ET_c and WUE. Trained on the largest compilation of experimental data sets of corn ET_c, these data-driven models represent nuanced corn water production functions that also account for weather and agronomy besides describing economic yield-ET_c response.

中文翻译：

---

美国灌溉玉米的用水量和效率：从数百个现场管理年的观察中学习


粮食产量对用水消耗（又称蒸散_量）的响应是灌溉农业生态系统中食物与水关系的一个关键维度。美国灌溉玉米生产的水足迹最大，必须评估其 ET _c和水利用效率 (WUE)，其规模对生产者和水资源管理者有意义。田间实验极大地帮助我们了解玉米产量如何在田间/地块水平上响应 ET _c ，但它们的集体协同作用尚未开发。该分析利用对全国玉米 ET _c测量实验的现有测量数据进行文献综合，建立模型来阐明和预测不同环境和农场管理下的玉米 ET _c和 WUE。由此产生的综合结果 ( n = 1,362) 捕获了平均干旱度、环境变异性、管理和作物结果的广泛条件，为训练数据驱动模型提供了前所未有的机会。随机森林模型使用一些基于环境和管理的预测因子有效地预测了季节性总 ET _c和季节性平均 WUE，证明 RMSE 分别为 33 mm (5.1%) 和 0.28 kg m ⁻³ (10.4%)。在确定 ET _c和 WUE 时，灌溉深度最重要，其次是生长季节收到的总降水量。这些数据驱动模型接受了最大的玉米 ET _c实验数据集的训练，代表了细致入微的玉米水生产函数，除了描述经济产量-ET _c响应之外，还考虑了天气和农学。