Soybean simulation system for multi-environment breeding trial optimization
A hybrid crop modeling and machine learning system that predicts soybean cultivar performance across the US Midwest, enabling breeders to optimize trial allocation, reduce costs, and accelerate cultivar evaluation for multi-environment trials.
Overview
This project offers a soybean simulation system that combines mechanistic crop modeling with machine learning algorithms to improve multi-environment breeding trial (MET) efficiency. By addressing the limitations of traditional METs, such as insufficient on-site weather and soil data and the high cost of testing many cultivars, the system helps breeders predict genotype-by-environment interactions and extend varietal evaluation across new geographic areas. The goal is to expedite cultivar evaluation and optimize trial allocation in the US Midwest, reducing time and resource investment while improving decision-making for breeders.
Technical specifications
Key features:
- Hybrid modeling approach: Integrates the APSIM mechanistic crop model with learning algorithms to predict soybean performance under diverse environmental conditions
- Validated prediction capability: Linear regression and clustering analysis on five years of MET data successfully predicted time to maturity for 150 cultivars across 30 Sub-Saharan African locations within plus or minus 10 days of observed field values
- Genotype-by-environment interaction prediction: Extends varietal evaluation beyond tested trial locations, identifying how cultivars perform in untested regions
- High-resolution spatial analysis: Uses geospatial mapping to identify suitable regions for trial allocation based on expected stability of top yielding cultivars
- Multi-trait assessment: Evaluates maturity predictions alongside grain quality and yield correlations
Technology readiness level
This is a one-year research project currently in active validation. The team has completed initial validation of maturity predictions on a five-year MET dataset spanning 150 cultivars across 30 locations in Sub-Saharan Africa. Current work focuses on validating predictions outside the original MET and assessing correlations with grain quality and yield. Future validation will deploy the APSIM crop model across the US Midwest using high-resolution maps to identify optimal trial allocation regions. The team is seeking partners with expertise in geospatial tasks and large-scale model deployment to broaden the project's impact.
About University of Illinois, Urbana-Champaign
The University of Illinois Urbana‑Champaign is a flagship public research university with large‑scale research capacity and a broad academic portfolio. An on‑campus Research Park co‑locates corporate R&D teams and startups with faculty, while the National Center for Supercomputing Applications provides advanced computing and data capabilities for collaboration. Integration with a regional health system and an engineering‑based college of medicine enables clinical translation, and a long‑standing extension network links campus innovation to partners statewide. Research is supported by competitive federal funding from NSF, NIH, DOE, USDA, and DoD. A technology transfer office streamlines IP, licensing, and startups, complemented by incubators and prototyping in the Research Park.
delivered to your inbox.