Crop growth model predictions for environmental characterization

The INVITE project’s WP4 has investigated the possibility to use crop growth models for predicting grain yield at variety and multi-environmental trial levels. This was first applied to sunflower and Terres Inovia post-registration trials from 2003 to 2020 (1,431 different trials).

Geographical distribution of field trials conducted between 2003 and 2020 by Terres Inovia.

The importance of an accurate soil characterisation for final prediction with the SUNFLO crop growth model was highlighted through these activities. Despite uncertainties in predicting the response of a wide range of sunflower cultivars, the model could be used for the environmental characterisation of each trial when sufficient crop management, climate and soil data are available.

Terres Inovia’s experimental network provided a huge amount of data and detailed description of each trial, however, the data did not correspond to the requirements of a crop growth model as many crop management and soil input variables were missing, and the quality of the data was variable according to the data collection effort. Since the quality of the data is essential to obtain accurate simulation results, it was concluded that the results were insufficient to separate the best varieties.

Nevertheless, it was demonstrated that the SUNFLO model could simulate the environment effect alone, and it was thus decided to use these modelling results to explore the issue of envirotyping. The aim here is to create a numerical experiment that accurately describes farming conditions and define groups of environments to provide year-independent context. To this end, further activities will explore methods to cluster time series of simulated stressors by functional data analysis. The objective is to prototype a decision support system for recommending variety choice at sowing time, considering varietal characteristics as well as the cropping context. This tool is already in development using the R language and the Shiny package, and is based on three main areas of advice:

  1. Classify varieties in terms of performance: use observed performance results use simulation to access the stability of these results over time.
  2. Express the agronomic merit of a variety according to its own disease resistance characteristics and the environmental biotic risk.
  3. Envirotyping to describe growing conditions and characterise performance in this context.

For further information, contact Philippe Debaeke (philippe.debaeke@inrae.fr)

The full report about these research activities is available on the website of the European Commission here.