Integrating UAV-Based Phenotyping With Process-Based Crop Modeling To Evaluate Soybean Cultivar Resilience Under Climate Change

This research combines high-throughput phenotyping with process-based modeling to enhance the evaluation of soybean cultivars under both historical and projected future climate conditions in Indiana. Climate variability, characterized by elevated temperatures, modified precipitation patterns, and an increasing incidence of catastrophic events, jeopardizes soybean productivity throughout the U.S. Midwest. Conventional field trials and breeding processes frequently lack the geographical and temporal precision required to assess cultivar performance under prospective climate stressors. This research mitigates these constraints by integrating remotely sensed canopy characteristics obtained from Unmanned Aerial Vehicles (UAVs) with the VIC-CropSyst modeling framework to simulate soybean yield across various genotypes, locations, and meteorological conditions.The research initially establishes a comprehensive array of statistical models to model seasonal above-ground biomass utilizing RGB and multispectral UAV footage gathered over three growing seasons (2021–2023) at Purdue’s Agronomy Center for Research and Education (ACRE). A sigmoidal RGB-based model attained the highest prediction accuracy (R² = 0.81, RMSE = 74.1 g), surpassing both linear and multispectral models. Biomass forecasts were subsequently employed to calibrate cultivar-specific physiological parameters in the VIC-CropSyst model, facilitating spatially explicit simulations of soybean output spanning over 300 grid cells encompassing Indiana's primary production areas. Simulations were performed based on a historical baseline scenario (1915–2013) and six Localized Constructed Analogs (LOCA) downscaled Representative Concentration Pathways RCP4.5 mid-century climate projections (2040–2060). The analysis indicated that cultivar yield rankings differed markedly across climate scenarios, underscoring the necessity for stability criteria in breeding decisions. Certain genotypes, specifically Diers and Rainey, outperformed other varieties both spatially and temporally. Moreover, all varieties showed substantial variability and crop breed rank fluctuations, rank of crop breeds based on yield, under anticipated climatic conditions. This study emphasizes the constraints of absolute yield outputs and presents relative yield ranking as a probabilistic and interpretable metric of cultivar resilience. These findings provide critical insights for breeding strategies focused on enhancing genetic selection for climate resilience. The research examines the hydrologic effects of cultivar selection alongside yield analysis by simulating surface runoff and tile drainage. Employing soil and drainage information pertinent to Indiana agriculture, VIC-CropSyst simulations indicated that cultivar selection exerted negligible hydrologic influence at the landscape scale, implying that water resource management may be independent of genotype selection in this scenario. This is a crucial discovery for harmonizing agricultural output with water policy and nutrient leaching approaches.The methodology delineates a reproducible process for converting UAV imagery into inputs for crop models, verifying outputs against ground-reference yield data, and generating spatial statistics for genotype evaluation. The findings establish UAV-assisted crop modeling as a scalable, non-destructive, and economical method for digital agriculture, broadening its applicability to diverse areas, crops, and climatic conditions. This research enhances the convergence of precision agriculture, climate adaptation, and decision-support systems by offering a scientifically robust and practically feasible approach for the selection of climate-resilient soybean cultivars. The results provide insights to breeders, agronomists, and policymakers by connecting geographical phenotyping with predictive modeling to tackle emerging food security concerns posed by climate change.