Akademik Veri Yönetim Sistemi
Meteorology and Atmospheric Physics, cilt.138, sa.2, 2026 (SCI-Expanded, Scopus)
Carbon dioxide (CO2) is the most significant greenhouse gas (GHG) globally because of its global warming potential and anthropogenic emissions. In light of the urgent nature of climate change, accurately forecasting CO2 becomes a critical task for developing effective adaptation strategies. Although various models have been developed to forecast CO2, these models typically focus on localized regions, cover short-term periods, or rely on a limited dataset. In this context, the study presents a seasonal autoregressive integrated moving average (SARIMA) model and a long short-term memory (LSTM) neural network to forecast global monthly CO2 concentration data. A time series analysis of monthly CO2 concentrations from 1979 to 2024 was conducted to predict future CO2 concentrations. Observed and predicted values were compared using performance metrics such as R, RMSE, MAE, and MAPE. While the results were similar, the SARIMA model generally performed better than the LSTM, achieving the lowest MAPE (1.5140%), RMSE (0.4511 ppm), and MAE (0.3493 ppm) and the highest R-value (0.9994) during testing. In addition, the results of this study highlight a consistent increase in CO2 concentrations and project an approximately 16% rise relative to current levels by 2050.