Events

Date: 12 November 2025   Time: 14:00 - 15:00

Location: Online via the Teams link below

Ever-increasing data sizes and advancements in modelling have greatly transformed environmental sciences in recent years. Ecology is no exception to this trend, as large machine, mostly deep learning models, have become ubiquitous in ecologically-related prediction tasks, such as identifying animals in aerial image or classifying vocalisations in acoustic recordings. Yet, such models are still somewhat rarely sown beyond prediction—i.e., related to ecological inference. In this talk, I pick up a staple inference task, species distribution modelling (SDM), and describe our attempt at making deep learning models actually work for it. The outcome is a tale of initially alluring opportunities that turned into unexpected pitfalls, high expectations that became sober realities, and the realisation that, while new approaches could absolutely transform the SDM arena, this will not happen without close collaborations between disciplines and integration of complementary expert knowledge from all sides.

Please follow this Teams link