Mechanistic crop modelling and AI for ideotype optimization: Crop-scale advances to enhance yield and water use efficiency
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Advancing our understanding and optimization of crop–environment interactions is essential for enhancing food production while conserving resources. Mechanistic crop models (MCMs) simulate crop growth by integrating physiological processes such as photosynthesis, biomass allocation, and phenological development. When combined with satellite-derived environmental data, these models support large-scale assessment of yield potential and water use efficiency, particularly under drought-prone conditions.
In this study, we evaluate the CERES-Rice model to define an ideotype that maximizes grain productivity while optimizing water use. A global sensitivity analysis using the Morris method identifies the most influential genetic-based parameters affecting traits such as biomass accumulation, grain yield, tiller number, anthesis timing, and maturity. These parameters were then optimized using a Genetic Algorithm (GA), simulating 1,884 virtual cultivars across 5,692 runs, guided by an integrated Harvest Index–Water Use Efficiency (HI-WUE) metric.
The resulting ideotype exhibits favorable traits including optimized phenological timing, and improved grain and water conversion efficiencies across four contrasting environments. To evaluate its genetic feasibility, the optimized parameter profile was compared to 21 field-characterized rice cultivars (indica, japonica, and hybrids) using multidimensional similarity metrics, including Euclidean, Mahalanobis, and Cosine distances.
This integrative framework demonstrates how combining MCMs with AI-based optimization and genotypic comparison can guide ideotype design and accelerate cultivar improvement. The findings highlight actionable physiological traits and genetic targets for breeding climate-resilient, high-efficiency rice varieties.
Author Summary
Efforts to improve crop productivity and conserve water resources are a key motivation that guides modern agricultural strategies. This study investigates how AI-driven advancements and mechanistic crop modelling can be combined to enhance drought-adaptation strategies in rainfed systems, with an emphasis on capturing genotype-by-environment interactions at regional scales.
We introduce an integrated framework that couples the CERES-Rice crop model with genetic algorithm to simulate a wide range of virtual cultivars under diverse environmental conditions. Using the proposed HI-WUE metric, combining harvest index and water use efficiency, this framework identifies ideotypes optimized for resource efficiency. A multidimensional similarity analysis quantifies the genetic distance between computationally optimized ideotypes and field-characterized cultivars, revealing key trait alignments and supporting physiologically grounded, genetically feasible breeding recommendations.
Central to this approach is the model’s ability to comprehensively reproduce the behavior of individual varieties, relying on iterative calibration and validation processes. This process ensures robust ideotype predictions and demonstrates how the co-development of crop models and phenotyping platforms, through collaboration between biologists and modelers, can translate these predictions into practical breeding outcomes for climate-resilient crop varieties. While mechanistic models may not yet fully capture genetic complexity at the gene network or 3D architectural levels, their capacity to represent functional genetic diversity remains a promising asset. Future work will likely focus on enhancing these models to better integrate the complex dynamics and underlying mechanisms of plant stress responses, further strengthening their predictive power for breeding outcomes.
