
LIFE uses state-of-the-art remote sensing products and data on species' habitat preferences to quantify changes in Area of Habitat (AoH). Estimated contemporary AoH is compared with potential AoH in the absence of humans to estimate a species’s probability of persistence and to test the effects of different land-use actions. More details on LIFE can be found in the Cambridge Open Engage preprint: LIFE: A metric for quantitatively mapping the impact of land-cover change on global extinctions.
Anthropogenic threats:
I am applying advances in predictive modelling to tie the distribution and severity of threats to more easily observable variables (e.g., satellite data products). Broadly, I am using publicly available environmental and socioeconomic predictors along with species traits to understand where these threats occur and how populations respond in terms of local population abundance. Current spatial threat mapping efforts either aggregate species taxonomically or consider few species, and disproportionately focus on the tropics; the overall goal is to generalise results taxonomically and spatially.
For my MRes project, I focused on predicting species-specific abundance responses to hunting pressure. I applied automated machine learning to efficiently assess machine learning models, experimented with recently proposed spatial and species deep learning embeddings, and assessed model generalisability to unseen taxa and geographies. See the project repository for more details.