Dr. Sara Martino (Norwegian University of Science and Technology, Noruega): "A Spatial Tale of Two Dolphin Species and Many Data Sources".

Date and Location: Tuesday, January 20, 2026 · 12:00 h · Saló de Graus de la Facultat de Ciències Matemàtiques.

Abstract: In this talk, I present a joint modeling framework for integrating conventional survey data and social media sightings to estimate the spatial distribution of two dolphin species in the Mediterranean Sea. All data sources are treated as presence-only point patterns and modeled using log-Gaussian Cox processes, with source-specific detection functions to explicitly account for observation bias. Detection models for structured surveys exploit known sampling protocols, while several proxy-based approaches are explored to model bias in social media data.
Bayesian inference is carried out using the SPDE approach with INLA and inlabru. The results show that integrating heterogeneous data sources improves spatial coverage and enables the estimation of intensity surfaces with quantified uncertainty, while underscoring the importance of carefully modeling observation processes to ensure identifiability.