Data-driven climatic zoning and future trend forecasting in Chad using artificial neural networks

The country’s potential for renewable energy and climate change adaptation is strongly influenced by its unique climate gradients, ranging from the semi-arid Sahel belt and the humid Sudan region in the south to the extremely arid Sahara in the north. The integrated data-driven framework presented in this work (i) objectively defines climatic zones using K-means clustering optimized by the Kneedle algorithm; (ii) quantifies statistical relationships between surface temperature and solar irradiance; and (iii) generates medium-term projections (2025–2050) using an Artificial Neural Network (ANN) model. The Kneedle-based optimization produced strong geographical partitioning across climatic gradients and determined that K = 4 was the statistically optimal clustering configuration (Difference Score = 0.756). Correlation analysis indicated a robust positive radiative-thermal relationship in the Sahelian and Sudanian zones (r = 0.49–0.63, p < 0.01) and a significant negative association in the Saharan zone (r = -0.37, p < 0.05). Although forecasts indicate continued warming in all regions, with average irradiance expected to reach 291.690 W/m² in the north, 262.521 W/m² in the center, and 241.686 W/m² in the south by mid-century, the ANN exhibited strong predictive accuracy (R² > 0.90 across validation datasets), highlighting its ability to effectively capture nonlinear patterns. Although the statistical accuracy is robust, the projections are derived from data-driven extrapolation rather than physically based climate simulations, and thus may not fully capture long-term regime shifts or nonlinear atmospheric feedbacks. Collectively, these results indicate that data-based climatic zoning provides a versatile analytical framework for climate-aware infrastructure development in arid and semi-arid regions, enhancing spatial precision in renewable energy planning.