Analyzing deforestation drivers in Somalia using machine learning and 2SLS with economic growth, population dynamics, renewable energy, and globalization

Abstract

Deforestation is an escalating environmental concern in Somalia, driven by a complex interplay of economic, demographic, and energy-related factors. Despite its severity, limited empirical research has investigated these drivers using advanced analytical methods. This study employs three machine learning models —Random Forest, Gradient Boosting Regression, and Kernel Ridge Regression (KRR) to investigate the influence of GDP per capita, urban population, population growth, renewable energy consumption, and industry value-added on deforestation trends from 1990 to 2023. The KRLS model provides localized marginal effects, offering interpretable insights into the nonlinear relationships between predictors and changes in forest cover. Results indicate that GDP per capita, urbanization, and population growth significantly and positively influence deforestation, while renewable energy consumption exhibits a statistically significant negative effect. Industry value added shows a weak and statistically insignificant relationship. Among the models, Gradient Boosting demonstrates the highest predictive accuracy. These findings underscore the unsustainable nature of Somalia’s current development trajectory and highlight the need for integrated policies that promote renewable energy, regulate urban expansion, and address demographic pressures. This study contributes to the emerging literature on machine learning applications in environmental policy analysis, offering data-driven insights for sustainable development in fragile contexts.