Assessing Climate Efficiency with Random Forest, DEA, and SHAP in the Eastern Black Sea Region, Türkiye

Abstract

The study is based on Land Surface Temperature (LST) and Air Temperature data and Nonparametric Data Envelopment Analysis (DEA) technique to evaluate heat efficiency and detect anomalies in the thermal regime in the Eastern Black Sea Region, particularly in Hopa and Artvin, during the period 2000-2024. The regulating role of the Black Sea has resulted in Hopa having the warmest and most stable temperature patterns, with daytime temperatures 1.8 to 3.7 degrees C higher than Artvin. Previous DEA analysis of daytime temperatures has shown that the 2018-2020 period had the highest daily temperatures, while the 2001-2010 decade was characterized by the highest nighttime temperatures. A future heat map based on Monte Carlo simulation using six climate change scenarios indicates that in the most optimistic case, assuming a temperature increase of +0.8 degrees C, efficiency scores could increase as high as 0.995. On the other hand, if global warming leads to a sudden temperature increase above +7.2 degrees C, there is a 21.7% climate efficiency loss. Sensitivity analysis showed that technological innovation and good governance are the main positive factors affecting climate efficiency. Random Forest (RF) and SHapley Additive Explanations (SHAP) analyses were applied to determine the impact of climate factors on DEA scores and also indicated areas requiring risk assessment. The findings highlight the importance of considering location-specific climate adaptation strategies. Based on the observed thermal contrasts between coastal and inland environments, potential adaptation considerations may include urban heat management and agricultural water stress in coastal areas such as Hopa, and cold-climate resilience and energy-efficient infrastructure in inland locations such as Artvin.