Abstract
Migration as a result of climate related disasters is an uncertain and complicated problem to the destination countries, making forecasting systems that are predictive and informative on its behaviours necessary. Through an intelligent system that combines ensemble machine-learning prediction with an agent-based simulation, this study supports climate migration risk analysis and policy preparedness of the United Kingdom. We trained and evaluated Random Forest, XGBoost, CatBoost, and LightGBM models on secondary data of 20 migrant source countries, with dates ranging from 2009-2023, to predict the migration flows with the help of environmental, socioeconomic , governance, and conflict index (R 2 ≈ 0.61), The SHAP explain-ability analysis demonstrates that climatic extreme situations, governance preparedness, and conflict severity are all major and non-linear causes of migration pressure. These model results are then incorporated into an agent-based model where household migration choices and adjustive policy reactions are the outcomes of interactions under climate and governance scenario subject to a set scenario conditions. The outcome of the simulations depicts the relationship between migration force and institutional reaction and the way the governance capacity moderates the influence of climate stress on migration outcomes.