Robin Zbinden

Robin Zbinden is a Researcher at the École Polytechnique Fédérale de Lausanne (EPFL), affiliated with the School of Architecture, Civil and Environmental Engineering (ENAC). He is part of the Laboratoire de science computationnelle pour l'environnement et l'observation de la Terre (ECEO), focusing on computational methods for environmental and Earth observation challenges. His work bridges machine learning, remote sensing, and ecological modeling to address climate-driven environmental questions. He is also a doctoral student in the Programme doctoral en informatique et communications at EPFL, specializing in data-driven approaches for species distribution and environmental systems.

Research interests include species distribution modeling (SDM), multi-modal data fusion, and algorithmic solutions for imbalanced ecological datasets. His work emphasizes adaptive machine learning frameworks like MaskSDM, integrating climate, satellite, and field data to predict species responses to environmental changes. Applications span biodiversity conservation, climate impact assessments, and sustainable food systems. His lab’s computational methods prioritize explainability and scalability, addressing gaps in ecological modeling robustness and generalization.

Notable contributions include foundational work on MaskSDM (2025), which enhances SDM flexibility through Shapley values, and studies on monarch butterfly migration under climate shifts. His 2024 research on pseudo-absence selection for deep learning models and 2023 exploration of neural networks in SDM highlight methodological innovations. Recent projects address food environment metrics via sales logs and carbon footprint perception studies (2019). Collaborations span environmental policy, geospatial analysis, and interdisciplinary sustainability challenges.

Lab affiliations include the ECEO team at EPFL Valais Wallis, where he develops computational tools for environmental observation. His research narrative combines technical algorithm design with ecological applications, aiming to bridge data science and real-world conservation needs.