Modeling and Forecasting of the Local Climate of Odesa Using CNN-LSTM and the Statistical Analysis of Time Series

This study investigates the climatic dynamics of Odesa, Ukraine, by integrating over 200 years of archival meteorological records with recent observations from the Davis Vantage Pro2 weather station and advanced machine learning techniques. The results reveal a distinct warming trend since 1985, with average annual temperatures projected by a CNN–LSTM model to rise by more than 6–7 °C above the mid-20th-century baseline by 2029, indicating an exceptionally rapid regional climatic shift. Spatial analysis of the July 2024 heatwave demonstrated pronounced thermal gradients, with the strongest overheating observed inland and the moderating influence of the Black Sea reducing temperature extremes in coastal areas. Precipitation analysis (1985–2024) showed an overall statistically insignificant increase; however, the summer months exhibited drying tendencies, a trend reinforced by model forecasts. Solar radiation dynamics (2012–2024) highlighted significant local variability shaped primarily by atmospheric conditions rather than solar activity, with notable monthly increases in October, November, and February. The novelty of this research lies in combining long-term datasets with deep learning methods to produce localized climate scenarios for Odesa, offering new insights into the city’s transition toward extreme warming, shifting precipitation patterns, and evolving solar energy potential. The findings have direct implications for environmental modeling, energy efficiency, and the development of climate change adaptation strategies in urbanized coastal regions.