How can we better predict the number of climate migrants?

Researchers found that most modelers tend to focus on slow-onset climate changes such as sea level rise instead of sudden-onset events such as natural disasters. They also found that although social, political, and economic factors are the main drivers of migration, they are often underrepresented in models. People rarely move for environmental reasons alone, though they might move for better education, economic opportunities, or security-related reasons that are impacted by climate change. Other limitations of current models include the neglect of potential climate-related immobility, future unknowns about how irreversible environmental changes might impact migration, and the personal motivations that people have to stay in or leave a place.  

Recommendations for improving models include improved data collection to better understand historic migration patterns, advanced multi-level modeling to capitalize on the strengths of each model, and inputs from a broad community of experts in different fields. DCID researchers found that at this stage of development, forecasting models are not yet able to provide reliable numerical estimates of future climate-related migration. Rather, models are best used as tools to consider a range of possible futures, explore systems dynamics, and test theories or potential policy effects.  

Methods and Results 

Researchers conducted an in-depth analysis of five models: gravity, radiation, systems dynamics, agent-based, and statistical extrapolation. A range of geographic regions, climate hazards, and migration types were included in each model. Most models forecast internal migration and do not specify whether their focus is on temporary or permanent migration.  

Models vary greatly and each has its limitations and benefits. For example, the gravity model forecasts climate-related migration indirectly as the difference in population distributions with and without different levels of climate change. Agent-based models estimate changes in out-migration, in-migration, and return migration but their results often do not track the trajectory that migrants follow. The statistical extrapolation model used climate-related factors such as remote sensing of flooding to determine which groups were more likely to migrate across borders in Southeast Asia. These findings point to the importance of understanding the strengths and limitations of a model before applying it and reporting results to address a model’s limitations.