arXiv2025
Extreme weather events are projected to intensify global migration, increase resource competition, and amplify socio-spatial phenomena, including intergroup conflicts, socioeconomic inequalities, and unplanned displacements, among others. Addressing these challenges requires consolidating heterogeneous data to identify, estimate, and predict the dynamical process behind climate-induced movements. We propose a novel hybrid approach to reconstruct hazard-induced displacements by analysing the statistical properties of a diffusion process (walks) that explores the spatial network constructed from real displacements. The likely trajectories produced by the walks inform the typical journey of individuals, identifying potential hazards that may be encountered when fleeing high-risk areas. As a proof of concept, we apply this method to Somalia's detailed displacement tracking matrix, containing 20,220 movements dating from February 8 to June 18, 2025. We reconstruct the likely routes that displaced persons could have taken when fleeing areas affected by conflict or climate hazards. We find that individuals using the most likely paths based on current flows would experience mainly droughts and conflicts, while the latter becomes less prominent at every subsequent step. We also find that the probability of conflict and drought across all trajectories is widely dispersed, meaning that there is no typical exposure. This work provides an understanding of the mechanisms underlying displacement patterns and a framework for estimating future movements in areas expected to face increasing hazards.