Research
Machine Learning for Environmental Event Detection
Modeling Climate-Induced Agricultural Scarcity and Migration in Central America: Evidence from New Machine-Coded Environmental Event Data
This study examines climate adaptation responses across Central America using AI-driven analysis of 25 million local media articles (2012-2024). We developed fine-tuned Large Language Models to create the MLEED dataset, measuring 13 types of environmental adaptations at unprecedented scale and frequency. Key findings reveal El Niño reduces agricultural productivity and is linked to displacement, while La Niña increases productivity. Slow-onset disasters decrease international migration, whereas sudden-onset disasters increase it. Higher agricultural commodity prices correlate with increased migration, reflecting income effects needed for expensive migratory journeys. Climate disasters sometimes trigger resource-based violence, subsequently increasing international migration patterns.