Akademik Veri Yönetim Sistemi
Journal of Atmospheric and Solar-Terrestrial Physics, cilt.277, 2025 (SCI-Expanded)
According to the assessments of the Intergovernmental Panel on Climate Change (IPCC), Türkiye, located within the Mediterranean basin, is among the regions most susceptible to the adverse impacts of climate change. This heightened vulnerability is largely attributed to its geographic location, climatic characteristics, and socio-economic structure, which together amplify the risks associated with rising temperatures and increasing climate variability. In the present study, monthly mean air temperature data for Türkiye, recorded by the Turkish State Meteorological Service between 1970 and 2022 (TSMS dataset), were analyzed in combination with reanalysis-based satellite observations obtained from the ERA5 (ERA5 dataset). These historical records formed the foundation for developing temperature projections extending to the year 2050. To achieve this, two complementary time-series forecasting approaches were applied: the Long Short-Term Memory (LSTM) deep-learning model, known for its ability to capture nonlinear dependencies and long-range temporal patterns, and the Seasonal Auto-Regressive Integrated Moving Average (SARIMA) model, a classical statistical method suitable for handling seasonality and trend components in climate data. The projection results revealed that Türkiye's mean temperature anomaly relative to the 1970–1980 baseline period is expected to rise by approximately 2.52 °C when based on in-situ observational data, and by about 3.48 °C when derived from ERA5 reanalysis estimates. These findings consistently indicate a significant warming trajectory, regardless of the dataset or modeling approach applied.