Application Domains

Plant Phenotyping

We develop information systems for data management, as well as methods for data validation and analysis. The OpenSILEX software suite enabled the development of the PHIS Information System, which has become an international reference for the structuring and exploitation of large-scale data in the field of high-throughput plant phenotyping, notably through the national phenotyping infrastructure PHENOME and the European infrastructure EMPHASIS.

Within the PHENOME-EMPHASIS phenotyping community, the resource center has, for example, helped overcome a major challenge related to the identification of genetic resources. Indeed, it is not possible to guarantee that a genetic resource (variety, accession) has a unique identifier, as the same species may be referenced under different identifiers across multiple resources (thesauri, AGROVOC, NCBI Taxonomy, NALT Thesaurus, etc.). The resource instance has made it possible to catalogue and reconcile the various identifiers associated with a genetic resource (official catalogues, genetic resource collections, accessions, etc.) using a methodology for identifier description developed through collaboration among several user groups.

Vine and Wine

Agricultural sectors require integrated approaches to address major challenges, including reducing inputs, adapting to climate change, and using digital tools to support production and processing while meeting consumer expectations. The vine and wine sector is highly representative of these challenges.

• We developed a data structuring, linking and exploration tool called Vine2Wine for the Experimental Unit (UE) Pech Rouge. This tool integrates data from soil and climate characterization, vineyard management and wine tasting. Following this project, data collection procedures became fully digital through the IADO mobile application based on OpenSILEX.
   

• We also participated in the NV2 project (a collaborative initiative involving both industrial and academic partners) focused on grapevine growth and nitrogen storage based on multi-year measurements. Some of the collected data consist of temporal observations from simultaneous processes (growth of different organs, nutrient storage, etc.), are climate-dependent, highly complex, and in some cases destructive (e.g. carbon biomass measurements). A Bayesian model was developed to estimate the parameters of these processes simultaneously.
• In collaboration with the UMR SPO, we also work on the modelling and control of alcoholic fermentation processes.

Digital Agriculture and Precision Livestock Farming

Digital agriculture and precision livestock farming aim to design and develop new digital tools and services for crop and livestock production. This requires the integration of expertise from agricultural sciences, engineering, digital technologies and the social sciences.

• Agricultural data are inherently interdisciplinary and must comply with the FAIR principles (Findable, Accessible, Interoperable and Reusable). The methods implemented in OpenSILEX are designed to make data FAIR while managing issues related to complexity (relevance, validity, multi-scale dimensions, etc.), semantic heterogeneity of sources and reproducibility.

• Faced with the rapid growth in the number of ontologies and semantic resources produced across scientific disciplines, it has become essential to provide shared platforms capable of receiving, hosting, aligning and promoting their reuse. AgroPortal is a reference platform for hosting and managing semantic resources (vocabularies, thesauri, terminologies, ontologies, etc.) in the fields of agriculture and food.
• To ensure animal welfare and health, we use machine learning methods and probabilistic models to analyse animal behaviour from a variety of sensor data.

Agroecology

Our research contributes to the agroecological transition, as illustrated by several of the unit's achievements.

• We have used constraint programming to model and solve crop allocation problems in small diversified farms. The objective is to provide alternative solutions for vegetable crop planning that integrate agroecological constraints and promote sustainable agricultural practices.
• Through the modelling, integration and semantic exploitation of variables originating from multidisciplinary domains such as plant phenotyping and agroecology, we have built a unified, interoperable and scalable knowledge graph.
• We have also investigated the modelling of agro-sylvo-pastoral systems to study the impact of fertility transfer by livestock through manure and urine on agricultural production. Our work has focused on modelling, analysing and optimising these systems, particularly in situations where no nitrogen-fixing plants are present in pastures.
• We develop theoretical results on algorithms used in participatory ecology research projects such as eBird and Pl@ntNet.

Water and Environment

The research carried out by the MISTEA unit also contributes to the preservation and management of natural resources, particularly water and soils.

• Lake ecosystem models are studied to simulate the population dynamics of microalgae (cyanobacteria) in lakes, with the long-term objective of controlling harmful algal bloom events.
• We work on irrigation modelling using reduced-order models adapted to automatic control techniques (control design, state observers and optimal control), on accounting for the specificities of treated wastewater reuse (TWW reuse) for fertigation with non-conventional water compositions (particularly nitrate-rich water), and on methodologies based on "dual modelling" approaches.