Recherche de carte d'idéotypes de sorgho d'après un modèle de culture : optimisation conditionnelle à l'aide d'un métamodèle de krigeage

Abstract : In the Sahel region, the irregular rainfall distribution in time and space generates variety x year and variety x location interactions. Therefore, determining variety with the best expected yield would take many years of experimentation in each location.Alternatively, the best variety could be identified by maximizing the predicted yield using a crop simulation model that describes growth and development of a crop in interaction with agro-environmental conditions.The average yield depends on the probability distribution of environmental inputs, which is location specific, while the cultivar parameters that maximizethis yield define the ideotype, i.e. the selection target.In this work, we want to draw the map of sorghum ideotypes in Sub Saharan Africa. To face the problem of optimizing a complex model, an algorithm conventionally used in this context is the Efficient Global Optimization method (EGO), based on kriging as a surrogate model. Here, the distribution of meteorological inputs follows a stochastic model whose parameters varycontinuously in space along a North-South gradient. Consequently, the optimization of varietal parameters is conditional on these climate parameters. Moreover, the function to maximize is noisy, because expectation and quantilesare merely estimated with a limited number of simulations. We aimed at adapting the EGO algorithm to the conditional optimization of a noisy function. Extensions exist either for the optimization of noisy functions or for the conditional optimization of deterministic functions, ie the search for the values of a subset of parameters that optimize the function conditionally to the values taken by another subset, which are fixed. A metaphor for conditional optimization is the search for a crest line. No method has yet been developed for the conditional optimization of noisy functions: this is what we propose in this thesis. Testing this new method on test functions shows that, in case of a high level of noise on the function, the PEQI criterion that we propose is better than the PEI criterion usually implemented in such a situation.The application of this new optimization method sorghum ideotypes parameters mapping has been tested in the area covered by Senegal, southern Mali and Burkina Faso. It consisted in maximizing the expected yield with respect to 4 parameters of Samara model: vegetative phase length, maximum root length, stem reserve potential, and leaf mortality. The results of this optimization partly coincide with the sensitivity analysis conducted on these same parameters.
Document type :
Liste complète des métadonnées
Contributor : Abes Star <>
Submitted on : Thursday, December 6, 2018 - 5:28:06 PM
Last modification on : Friday, March 29, 2019 - 9:12:15 AM
Document(s) archivé(s) le : Thursday, March 7, 2019 - 2:56:35 PM


Version validated by the jury (STAR)


  • HAL Id : tel-01947412, version 1



Diariétou Sambakhé. Recherche de carte d'idéotypes de sorgho d'après un modèle de culture : optimisation conditionnelle à l'aide d'un métamodèle de krigeage. Applications [stat.AP]. Université Montpellier, 2018. Français. ⟨NNT : 2018MONTS022⟩. ⟨tel-01947412⟩



Record views


Files downloads