[Defended thesis] Girault Gnanguenon: Modelling and viewing relations between agrienvironmental time courses and product quality using a parcimonious Bayesian approach

Modelling and viewing relations between agrienvironmental time courses and product quality using a parcimonious Bayesian approach

My name is Girault Gnanguenon Guesse and I'm a PhD student in biostatistics at the UMR MISTEA (Mathématiques, Informatique et STatistique pour l'Environnement et l'Agronomie), a joint unit of the Institut National de Recherches Agronomiques (INRA) and Montpellier SupAgro.
After completing a degree in Environmental Management at the Catholic University of West Africa, I had my first experience in Benin as a research assistant in charge of data management and analysis at the Laboratoire de Recherches en Ecologie Animale et de Zoogéographie. There, I developed my skills in business intelligence and statistics for the environment and agronomy, which reinforced my plans to become a researcher in biostatistics.
I have a Master's degree in Mathematics - Statistics and Life Data from the University of Poitiers. One of my internships involved the analysis of a multi-year dataset for the seed company Vilmorin. The aim was to understand the yields of certain seed multiplication farmers in a context of climate change. The results made it possible to propose theoretical levers to farmers to optimize their yields. It is in this context of proposing decision-support tools to professionals that my thesis is based.

• Starting date: December 2017
• University: MUSE Montpellier Université d’Excellence / Institut Agro 
• PhD school:  ED 166 I2S – Information, Structures, Systèmes
• Scientific field:  Biostatistics
• Thesis management:  Nadine Hilgert (Inrae, UMR Mistea)
• Thesis supervisors:  Bénédicte Fontez (Institut Agro – Montpellier SupAgro, UMR Mistea) et Thierry Simonneau (Inrae, Lepse)
• Funding: #DigitAg – Inrae
• Research project: INNOVINE
• #DigitAg: Cofunded PhD – Axes 5  et 6 – Challenges 1, 5 et 7

Keywords: Latent factor models, parsimonious approach, Bayesian inference, agronomy

Abstract: Traditional knowledge plays an important role in agricultural practices. For instance, in the vine and wine food chain, decisions taken in vineyards mainly rely on expert knowledge-based approaches. Confronted with new challenges, stakeholders in agricultural production chains need advanced quantitative-based decision support tools. The aims of this PhD are i) to propose a knowledge discovery method to deal with big data from time courses, ii) to explain and predict product quality. Data integration should deal with high resolution data from sensors or agronomic models, low resolution observations and expert knowledge. It requires taking into account the reliability of all sources and data uncertainties. This calls for a coupling between informatics and data analysis, and constitutes the core of the PhD. The main application concerns the vine and wine food chain, in close relation with industrial partners (consulting professionals: ITK, Fruition Sciences, technical institute IFV) and public research laboratories (Joint Units LEPSE and SPO)

Jury compound:

• Élodie BRUNEL-PICCININI, Maître de conférences, Université de Montpellier : Examinatrice
• Bénédicte FONTEZ, Maître de conférences, Institut Agro – Montpellier SupAgro : Co-Encadrante de thèse
• Romain GLÈLÈ KAKAÏ, Professeur des universités, Université d’Abomey-Calavi : Examinateur
• Nadine HILGERT, Directrice de recherche, INRAE – Montpellier : Directrice de thèse
• Patrice LOISEL, Chargé de recherche, INRAE – Montpellier : Co-Encadrant de thèse
• Tristan MARY-HUARD, Chargé de recherche, INRAE – Le Moulon : Rapporteur
• Philippe PIERI, Chargé de recherche, INRAE – Bordeaux : Examinateur
• Thierry SIMONNEAU, Directeur de recherche, INRAE – Montpellier : Co-Directeur de thèse
• Nancy TERRIER, Chargée de recherche, INRAE – Montpellier : Invitée
• Anne-Françoise YAO, Professeur des universités, Université Clermont-Auvergne : Rapporteure

Contact : benedicte.fontez [AT] supagro.fr

Social network: LinkedIn

Download the thesis manuscript

Acts of conference

GNANGUENON GUESSE G, LOISEL P, FONTEZ B, SIMONNEAU T, HILGERT N. An exploratory penalized regression to identify combined effects of functional agri-environmental variables, 6 – 31 August 2020. The 30th « virtual » International Biometric Conference, Séoul, International Biometric Society, 2020

GNANGUENON GUESSE G, LOISEL P, FONTEZ B, SIMONNEAU T, HILGERT N. Explorer l’influence conjointe de prédicteurs fonctionnels sur une réponse réelle via une régression pénalisée. Recueil des soumissions de la 52èmes Journées de Statistiques, Société Française de Statistique (SfdS), 2020, pp 375-380, sciencesconf.org :jds2020:319910