Observation-Constrained Agroecosystem Model Inversion Reveals Continental-Scale Variation of Winter Wheat Traits

Area of Science:

• Agricultural Science
• Climate Science
• Computational Biology

Background:

• Genotype × Environment × Management (G×E×M) interactions are crucial for predicting crop performance under climate change.
• Current continental-scale crop models lack spatial resolution to capture adaptive trait variation.

Purpose of the Study:

• To infer spatially explicit physiological controls of U.S. winter wheat using model-data fusion.
• To improve predictions of agricultural performance by accounting for trait variability.

Main Methods:

• Integrated the Ecosys mechanistic agroecosystem model with a deep learning-enabled model-data fusion inversion.
• Constrained simulations with satellite-derived photosynthesis and county-level yield records (2008-2022).

Main Results:

• Recovered coherent patterns of winter wheat maturity, reproductive capacity, harvest index, and root-shoot allocation.
• Optimized simulations accurately reproduced observed carbon uptake and yield variability.
• Inferred physiological profiles aligned with major winter wheat market classes and explained climate sensitivities.

Conclusions:

• Observation-constrained trait inversion reveals biologically meaningful crop-class variation.
• Provides a scalable, physiologically grounded framework for diagnosing adaptive diversity and climate vulnerability in agroecosystems.