A model-based decision support framework to optimize cultivar recommendation. A case study on Pisum sativum L.

Cultivar recommendation is a crucial decision-making process for successful cropping seasons. Crop models can effectively support the identification of suitable cultivars for a given context, as they allow exploration of a wide range of agro-climatic conditions that would be difficult to capture with multi-environment trials. In this study, a framework for cultivar recommendation based on the integration of crop models with ideotyping techniques is applied to pea ( Pisum sativum L.) in the Emilia-Romagna region, Northern Italy. Eighteen agro-climatic contexts were identified by considering the two main sowing windows and regional climate and soil data. Statistical distributions of seven functional traits related to phenology, canopy architecture, biomass partitioning, and photosynthetic efficiency were derived for 20 commercial pea cultivars and used to run a variance-based sensitivity analysis of the STICS crop model for each agro-climatic context. Model parameters corresponding to the seven functional traits were used as sensitivity analysis inputs, with a composite function accounting for both yield and its stability as the target output. Sensitivity analysis results were used to design pea ideotypes for each agro-climatic context, and the similarity between the phenotypic profiles of commercial cultivars and those of ideotypes was evaluated using the Euclidean distance weighted by the sensitivity index. Spatially distributed recommendations based on context-specific, similarity-based cultivar rankings were derived. A clear change in cultivar rankings was found across agro-climatic contexts, highlighting the need to explicitly analyse G×E×M interactions. The simple methodology used to derive the cultivar phenotypic profiles and the process-based modelling approach allow easily extending the framework to different sowing times, newly released cultivars, or different crops, all of which are key features for transferring the framework to operational farming contexts. While the preliminary evaluation of the recommendations provided by the system supports its potential reliability, further validation through multi-environment trials is required before full deployment at the operational level.