I develop and train deep Species Distribution Models (SDMs) to predict the spatial distribution of three orders of emerging aquatic insects: Ephemeroptera (mayflies), Plecoptera (stoneflies), and Trichoptera (caddisflies). The emergence of these insects contributes to the provision of essential ecosystem services such as nutrient cycling, pollination, and pest control.

My models integrate very high-resolution drone imagery and LiDAR data to capture fine-grained landscape features that drive insect presence.

A key component of my work is to explore the value of very high-resolution drone data (multispectral and LiDAR) for the development of fine-scale SDMs. This raises key questions, such as the influence of the choice of sensor, grain size, and spatial extent on DeepSDM model's predictions. Consequently, I also use satellite, super-resolution and government data for comparison purposes.

Making deep learning models interpretable is central to my research. I work on concept-based explainability methods applied to SDMs, with the goal of identifying which landscape features (concepts) drive model predictions. This contributes both to ecological understanding and to building trust in AI-assisted biodiversity assessment.

I have released a high-resolution landscape dataset designed for concept-based XAI benchmarking — see the Publications section for details.