A new way to model how crops grow
Researchers developed a framework to identify how environmental factors and genetics affect plant growth dynamics.
A new framework identifies reduced systems of ordinary differential equations (ODEs) for crop growth. These equations describe growth as a function of environmental variables and genotype-specific coefficients (GSCs).
The method uses Sparse Identification of Nonlinear Dynamics (SINDy) to infer model structure from data. This uses sparse regression on a physiologically informed library of candidate functions. GSCs parameterise cultivar-specific differences in response to environmental conditions.
Current process-based crop growth models (CGMs) capture environmental dynamics. However, they are difficult to scale across many genotypes because they require calibration of many parameters. Quantitative genetic models allow modelling many genotypes but often summarise time-varying environments using static covariates.
The approach was evaluated on an in silico dataset from WOFOST for five potato cultivars across twelve growing seasons. The identified ODEs generalise to held-out years. They achieve for end-of-season yield prediction when compared to full-scale WOFOST simulations.
Key facts
- The models achieve for end-of-season yield prediction.
- The framework uses SINDy to infer model structure directly from data.
- The models are linear in GSCs, making them compatible with genomic prediction approaches.