Exploring interactions between socioeconomic context and natural hazards on human population displacement

Area of Science:

• Environmental science
• Social science
• Data science

Background:

• Climate change intensifies extreme weather events, causing significant human displacement.
• Understanding internal displacement patterns is crucial for effective disaster response and planning.

Purpose of the Study:

• To model and understand disaster-induced internal displacement using explainable machine learning.
• To identify key environmental, societal, and economic drivers of displacement.
• To quantify differential vulnerability to climate-related hazards.

Main Methods:

• Development of a global database of disaster-induced movements (2016-2021).
• Application of explainable machine learning (ML) techniques, including Shapley values and causality-based methods.
• Integration of environmental, societal, and economic factors for predictive modeling.

Main Results:

• Displacement is primarily linked to poor household conditions and intense precipitation.
• Explainable ML models reveal the interplay of factors contributing to displacement.
• Empirical evidence of differential vulnerability to climate-related disasters is established.

Conclusions:

• Poor living conditions and intense rainfall are significant drivers of climate-induced displacement.
• Quantifying vulnerability is essential for targeted mitigation and adaptation strategies.
• Explainable ML provides valuable insights into complex displacement dynamics.