A model-based approach to assist variety evaluation in sunflower crop

TL;DR

This paper introduces a model-based method using crop simulation to enhance sunflower variety testing across diverse environments, aiming to improve selection accuracy and crop performance predictions.

Contribution

It presents a novel application of crop modeling and simulation to extend and improve sunflower variety evaluation beyond traditional field trials.

Findings

Model prediction accuracy was 54.4%.

Hybrid ranking correlation was Kendall's τ = 0.41.

Generated 2100 virtual trials for variety assessment.

Abstract

Assessing the performance and the characteristics (e.g. yield, quality, disease resistance, abiotic stress tolerance) of new varieties is a key component of crop performance improvement. However, the variety testing process is presently exclusively based on experimental field approaches which inherently reduces the number and the diversity of experienced combinations of varieties x environmental conditions in regard of the multiplicity of growing conditions within the cultivation area. Our aim is to make a greater and faster use of the information issuing from these trials using crop modeling and simulation to amplify the environmental and agronomic conditions in which the new varieties are tested. In this study, we present a model-based approach to assist variety testing and implement this approach on sunflower crop, using the SUNFLO simulation model and a subset of 80 trials from a large multi-environment trial (MET) conducted each year by agricultural extension services to compare newly released sunflower hybrids. After estimating parameter values (using plant phenotyping) to account for new genetic material, we independently evaluated the model prediction capacity on the MET (model accuracy was 54.4 %) and its capacity to rank commercial hybrids for performance level (Kendall's $\tau$ = 0.41, P < 0.01). We then designed a numerical experiment by combining the previously tested genetic and new cropping conditions (2100 virtual trials) to determine the best varieties and related management in representative French production regions. We suggest that this approach could find operational outcomes to recommend varieties according to environment types. Such spatial management of genetic resources could potentially improve crop performance by reducing the genotype-phenotype mismatch in farming environments.