Increasing drought and extreme rainfall are major threats to maize production in the United States. However, compared to drought impact, the impact of excessive rainfall on crop yield remains unresolved. Here, we present observational evidence from crop yield and insurance data that excessive rainfall can reduce maize yield up to −34% (−17 ± 3% on average) in the United States relative to the expected yield from the long‐term trend, comparable to the up to −37% loss by extreme drought (−32 ± 2% on average) from 1981 to 2016. Drought consistently decreases maize yield due to water deficiency and concurrent heat, with greater yield loss for rainfed maize in wetter areas. Excessive rainfall can have either negative or positive impact on crop yield, and its sign varies regionally. Excessive rainfall decreases maize yield significantly in cooler areas in conjunction with poorly drained soils, and such yield loss gets exacerbated under the condition of high preseason soil water storage. Current process‐based crop models cannot capture the yield loss from excessive rainfall and overestimate yield under wet conditions. Our results highlight the need for improved understanding and modeling of the excessive rainfall impact on crop yield.

Extreme weather events threaten food security, yet global assessments of impacts caused by crop waterlogging are rare. Here we first develop a paradigm that distils common stress patterns across environments, genotypes and climate horizons. Second, we embed improved process-based understanding into a farming systems model to discern changes in global crop waterlogging under future climates. Third, we develop avenues for adapting cropping systems to waterlogging contextualised by environment. We find that yield penalties caused by waterlogging increase from 3–11% historically to 10–20% by 2080, with penalties reflecting a trade-off between the duration of waterlogging and the timing of waterlogging relative to crop stage. We document greater potential for waterlogging-tolerant genotypes in environments with longer temperate growing seasons (e.g., UK, France, Russia, China), compared with environments with higher annualised ratios of evapotranspiration to precipitation (e.g., Australia). Under future climates, altering sowing time and adoption of waterlogging-tolerant genotypes reduces yield penalties by 18%, while earlier sowing of winter genotypes alleviates waterlogging by 8%. We highlight the serendipitous outcome wherein waterlogging stress patterns under present conditions are likely to be similar to those in the future, suggesting that adaptations for future climates could be designed using stress patterns realised today.

Accurate and cost-effective quantification of the carbon cycle for agroecosystems at decision-relevant scales is critical to mitigating climate change and ensuring sustainable food production. However, conventional process-based or data-driven modeling approaches alone have large prediction uncertainties due to the complex biogeochemical processes to model and the lack of observations to constrain many key state and flux variables. Here we propose a Knowledge-Guided Machine Learning (KGML) framework that addresses the above challenges by integrating knowledge embedded in a process-based model, high-resolution remote sensing observations, and machine learning (ML) techniques. Using the U.S. Corn Belt as a testbed, we demonstrate that KGML can outperform conventional process-based and black-box ML models in quantifying carbon cycle dynamics. Our high-resolution approach quantitatively reveals 86% more spatial detail of soil organic carbon changes than conventional coarse-resolution approaches. Moreover, we outline a protocol for improving KGML via various paths, which can be generalized to develop hybrid models to better predict complex earth system dynamics.