Stochastic Weather Simulation of Türkiye’s Geographical Regions: A Monte Carlo Framework

Abstract

Understanding and predicting regional climatic variations is crucial for agricultural planning, water resource management, and disaster mitigation in the face of increasing global climate instability. This study presents a comprehensive stochastic simulation framework designed to model the diverse meteorological profiles of Türkiye’s seven geographical regions. The developed model utilizes a randomized structure that learns monthly parameters from historical daily datasets spanning the 2010–2024 period and generates daily temperature and precipitation scenarios for 2025. Methodologically, a Normal Distribution was employed for temperature modeling, while a first-order Markov Chain was utilized to determine the occurrence of precipitation. To account for the characteristically positive and right-skewed nature of rainfall intensity on wet days, the Gamma Distribution was preferred a standard approach in current literature for modeling daily precipitation amounts. To capture the inherent stochasticity of atmospheric processes, a Monte Carlo simulation approach was implemented. Each representative city was initially simulated with 300 iterations to ensure statistical robustness. To rigorously assess predictive accuracy, validation for 2024 was performed using a Monte Carlo simulation with 400 independent runs. The findings demonstrate that the model effectively captures regional climatic trends and provides a reliable synthetic dataset for environmental planning. Validation metrics indicate strong agreement between modeled and observed climatic behavior, with historical monthly temperatures, precipitation totals, and wet-day probabilities consistently falling within the simulated 95% confidence intervals. These results suggest that integrated stochastic models offer a high-fidelity and computationally efficient alternative to complex numerical weather predictors for regional climate assessment.