Interpretable Machine Learning for Revealing Main and Interactive Effects of Ozone and Meteorological Factors on Crop Photosynthesis

Abstract Elevated ozone (O 3 ) and severe meteorological conditions could significantly influence crop photosynthesis in winter wheat and summer maize in the North China Plain. However, the complex climate‐pollution interactions make it difficult to use traditional methods to separate each factor's direct effects from their combined impact, thereby hindering our ability to understand and manage crop productivity under varying environmental conditions. Here, we quantified the main and interactive effects of O 3 and meteorological conditions on the photosynthesis of winter wheat and summer maize by integrating ground/satellite measurements with a machine learning model and an interpretability algorithm. Results showed that O 3 and meteorological factors each had nonlinear main effects on crop photosynthesis in both winter wheat and summer maize. Specifically, O 3 had negative effects on crop photosynthesis above thresholds of 111.7 ± 1.7 μg/m 3 for winter wheat and 130.8 ± 0.5 μg/m 3 for summer maize. Meanwhile, interactions between O 3 and meteorological factors accounted for 21.92% and 20.79% of the variation in photosynthesis in winter wheat and summer maize, respectively. These interactive contributions were substantially higher than the isolated main effects of O 3 alone (3.3% and 3.19%, respectively), with dominant interactions involving temperature in winter wheat, and CO 2 concentration in summer maize. Typically, interactions between O 3 and various meteorological elements exceed the main effect of at least one of them. These findings highlighted the importance of complex nonlinear interactions between air pollution and meteorological factors in regulating crop photosynthesis. They helped identify high‐risk situations in agricultural production arising from the combined effects of atmospheric pollution and meteorological conditions.