Description du projet
Intégrer des modèles spatiaux dans l’apprentissage automatique écologique
Dans le domaine de la recherche écologique, il est essentiel de produire des cartes spatiales précises qui englobent des variables essentielles telles que la biodiversité et le climat. Les méthodes actuelles d’apprentissage automatique, comme les forêts d’arbres décisionnels, sont efficaces mais négligent souvent les schémas spatiaux complexes inhérents aux processus écologiques. Cette limitation entrave notre compréhension des écosystèmes complexes et compromet la précision des prévisions. Soutenu par le programme Actions Marie Skłodowska-Curie, le projet PRISM s’attaque à ce problème en intégrant et en validant des modèles spatiaux dans des modèles d’apprentissage automatique. S’inspirant de la géographie, de l’écologie et de l’informatique, PRISM adopte une approche scientifique ouverte afin de diffuser largement ses résultats. Il favorise la collaboration entre les chercheurs et les institutions, en enrichissant les compétences et les méthodologies. En fin de compte, PRISM promet d’améliorer la recherche écologique, en offrant des prédictions plus précises et une meilleure compréhension des modèles spatiaux.
Objectif
Ecological research necessitates the production of spatial maps representing an array of critical variables such as biodiversity, climate, land cover, and soil carbon storage. While current machine learning methods, such as the well-known random forest (RF), have been effective in generating maps for these variables, they often overlook the intricate spatial patterns inherent in ecological processes. The PRISM project seeks to introduce a novel approach that addresses this limitation by integrating and validating spatial patterns within machine learning models. The project will draw upon insights from various fields, including geography, landscape ecology, statistics, and computer science. To ensure the widespread dissemination and accessibility of its findings, PRISM will adopt a comprehensive open science approach, including manuscript publications, the development of open-source software, and the sharing of repositories containing data and code, enabling others to reproduce and build upon the project's results. Through this project, an exchange of knowledge is anticipated between the researcher and the host institution, fostering a collaborative partnership. Under the supervisor's mentorship, the researcher will acquire essential skills in group organization, grant preparation, and research leadership. The researcher will enrich the host institution by creating innovative methods for spatial data analysis, implementing impactful teaching methodologies, and sharing the principles of open science. Ultimately, the PRISM project is poised to fuel the researcher's interdisciplinary growth, positioning him as a valuable asset in both academia and industry. The project's outcomes have the potential to improve how ecological research is conducted, leading to more accurate predictions and a deeper understanding of complex spatial patterns in ecological systems.
Champ scientifique
- natural sciencescomputer and information sciencesdata science
- natural sciencesbiological sciencesecologylandscape ecology
- natural sciencesearth and related environmental sciencessoil sciencesland-based treatment
- natural sciencesbiological sciencesecologyecosystems
- natural sciencescomputer and information sciencesartificial intelligencemachine learning
Programme(s)
- HORIZON.1.2 - Marie Skłodowska-Curie Actions (MSCA) Main Programme
Régime de financement
HORIZON-TMA-MSCA-PF-EF - HORIZON TMA MSCA Postdoctoral Fellowships - European FellowshipsCoordinateur
48149 MUENSTER
Allemagne