Deconvoluting Temporally Aggregated Migration Data Using a Bayesian Mixture Model for Insights Into Climate Induced Migration Dynamics
Abstract
Extreme climate events, such as floods and droughts, may trigger episodes of internal and trans-boundary migration. These may temporarily exacerbate migration due to economic factors, or conflict. Migration data are recorded either as population flows from source to sink regions, or as changing stocks of population at different locations. Often, these data are reported annually, or every 5 or 10 years corresponding to the frequency of census or population survey instruments. Since climate data are available at a high temporal resolution, e.g., monthly, a precise assessment of the role the climate event played in the migration dynamics is difficult, as investigators typically try to make these assessments at the coarser resolution. This impacts both the attribution of the climate effect and the ability to predict a future migration outcome for a climate exigency. We present a Multi-level Bayesian Mixture Model that can be applied to data from multiple spatial units with population stocks or flows reported every 5 to 10 years, and to monthly or seasonal data on key climate variables to deconvolute the background migration rates and those induced by the climate events in the historical record. The potential spatio-temporal migration response to climate can then be predicted probabilistically, including possibly the role of other migration drivers. Applications to data for Mexico-US migration are presented to illustrate the methods and provide insights into the role of climate.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2021
- Bibcode:
- 2021AGUFMGC43E..06C