Spatiotemporal Transformers for Predicting Avian Disease Risk from Migration Trajectories

Avian - borne pathogens pose persistent risks to wildlife and, at times, to public health; the ongoing H5N1 panzootic has caused extensive mortality in wild birds and livestock with continued geographic expansion and zoonotic concern (Mostafa et al., 2025) . Migratory birds link distant ecosystems across seasons and can create spatiotemporal pathways for pathogen dispersal, with phylodynamic and migration evidence supporting long - range movements of HPAI along flyways, including documented trans - Atlantic incur sion events (Banyard et al., 2024) . Operational early - warning and risk assessment commonly rely on outbreak notifications and targeted surveillance (e.g., WOAH's WAHIS Early Warning), often paired with mathematical or statistical models. Classic global modeling integrated bird flyways, phylogenies, and trade to forecast international spread of H5N1 and identify invasion pathways (Kilpatrick et al., 2006) . Recent genomic and phylodynamic analyses further link dispersal patterns to migration timing and characterize post - introduction spread across flyways (Nguyen et al., 2025; Yang et al., 2024) .