Towards non-standard structural equation models for studying interactions between agriculture and biodiversity

Context and objectives:
This internship is part of the SPATMAN project, supported by CESAB (Center for Synthesis and Analysis on Biodiversity) of the Foundation for Research on Biodiversity (FRB). SPATMAN aims to understand how the spatial organization of human activities modulates pressures on biodiversity by combining ecological, socio-economic, and spatial data at the subnational level.
In this context, the internship will contribute to the project’s methodological reflection by exploring the relevance and limitations of structural equation models (SEM) for representing the complexity of agroecological systems.
The objective is to test different linear and non-linear approaches using an existing agri-ecological dataset (500 ENIs) and other complementary, available, spatial datasets, in connection with the preparation of a scientific perspective article on non-standard SEMs.

Objectives and approach:
The student will conduct exploratory work combining data analysis and methodological reflection:
1. Explore the data (structure, variability, correlations) and build standard SEM models as a starting point.
2. Test non-linear and contextual extensions: quadratic effects, interactions, regional or spatial dependence.
3. Evaluate and compare approaches (frequentist, Bayesian, Machine Learning, possibly hybrid) and discuss their potential for modeling complex systems.
4. Propose ways to adapt to a more flexible SEM framework, in line with the collective discussions of the SPATMAN project.

Expected output:
· Exploratory analyses and documented modeling (scripts, figures, tables, comments).
· Summary note on the results and methodological avenues identified.
· Contribution to a literature review on nonlinear SEMs and their use in ecology.

Required profile:
Master’s student (ecology, biostatistics, data science, quantitative geography, environmental sciences) or on a gap year.
Required skills:
· Good knowledge of the basics of statistics, multivariate, and machine learning modeling also using a geo-spatial approach,
· Skills in GIS, R, and/or Python,
· Autonomy, rigor, and a taste for interdisciplinary thinking.

Scientific supervision: Karine Princé (MNHN) & Frédéric Gosselin (INRAE )(main supervisors), in collaboration with Cathleen Petit (FRB), Marco Vizzari (U. Perugia) & Maurizia Sigura (U. Perugia).

Duration: 4 to 6 months (with possibility of adjustment)

Desired period: from February 2026

Location: Centre d’Ecologie et des Sciences de la Conservation – Muséum National d’Histoire Naturelle, Paris

Remuneration: according to legal and institutional terms (and according to on funding availability)