Utilisation d'un modèle de culture pour évaluer le comportement des génotypes: Pertinence de l'utilisation d'Azodyn pour analyser la variabilité du rendement et de la teneur en protéines du blé tendre

The change of crop management to less intensive systems requires to rapidly forecast the ability of each variety to valorise theses systems for various pedo-climatic conditions. The experiments, generally conducted in this aim, often give a partial response because, the range covered are limited comparatively to the agricultural conditions in wich these genotypes are able to be used, and too long in regard of the rapid change of varieties used by farmers. To accelerate this step of cultivar evaluation, it is possible to use new forecasting tools. The work presented here aims at studying the possibilities offered by the use of crop models to analyse and predict the production of different cultivars, in a large range of crop conditions, for existing or future genotypes. The questions raised focused on the necessary adaptation of these tools to give a good account of the genotype behaviour and we have proposed new tracks to evaluate the ability of such tools to choose the best variety adapted to particular conditions. The concepts and formalisms included in the crop models make them a priori relevant to evaluate and predict the variability of genotype behaviour, as the interactions between crop functioning and environment status are simulated. However, few studies aimed at developing methods for estimating genotypic parameters. And few of these studies evaluated the ability of various crop models adapted to cultivars to give a good account to their various production among environments. Based on the example of one crop model, Azodyn, simulating yield elaboration and grain protein content through the different processes of nitrogen absorption and remobilisation in the crop-soil system, we have proposed a method to identify and estimate the genotypic parameters of the model, and several criteria for the evaluation of this adapted model to give an account of the cultivar behaviour. In a first step, we identified, among all parameters of the model (around 80), those which varied among cultivars, from specific experiments, from the literature or the expertise. Thus we have identified that only three parameters and three inputs of the model which variations among genotypes were necessary to be taken into account to improve the predictive quality of the model. These genotypic characteristics act as behaviour markers and can be easily measured by the different variety users. These parameters have been estimated for fourteen genotypes, through a direct measurement on specific experiments. The influence of the parameters and input, identified as genotypic, on yield and grain protein content were quantified by a sensitivity analysis of the model. This study enabled us to propose a hierarchy between those six characteristics, which appeared variable among environments. This was then a good means to identify the adaptation factors of the varieties to the environments, which were mainly: the maximum grain weight, the precocity at the beginning of stem elongation and at flowering, and the ability to produce a high grain number. No parameters specific to nitrogen nutrition appeared as determinant in the model outputs. Then, we have evaluated the ability of the model to give a good account of the genotype production on 21 environments, varying through the nature and the intensity of limiting factors. We showed that, taking into account only the low number of genotypic parameters, it was possible to give a good account of the mean yield and grain protein content of each genotype, for a large range of environments. We also showed that the crop model used was a relevant tool to predict the genotype variability and the classification of the various genotypes for several crop conditions.