AI-Based Climate Change Forecasting Models

Abstract

Climate change forecasting demands models that capture complex nonlinear interactions across atmospheric, oceanic, and terrestrial systems at spatial and temporal scales ranging from regional weather patterns to century-scale global trends. Physics-based general circulation models (GCMs) remain the gold standard for climate projection but are computationally expensive and struggle with sub-grid-scale processes that must be parameterised rather than resolved. This study evaluates six AI-based approaches -- convolutional LSTMs, graph neural networks on climate grids, Fourier neural operators (FNO), vision transformers applied to reanalysis fields, hybrid physics-ML models (ML-corrected GCMs), and ensemble gradient-boosted models for regional downscaling -- across four climate forecasting tasks: global temperature anomaly prediction (ERA5, 1979-2021, 1-12 month horizons), regional precipitation forecasting (E-OBS European gridded data, seasonal), extreme event detection (tropical cyclone genesis from ERA5 fields), and statistical downscaling (CMIP6 coarse-resolution to 0.1-degree regional fields). A total of 2,160 experiments were conducted using standardised train-validation-test splits respecting temporal ordering. The hybrid physics-ML model achieved the lowest temperature RMSE at 6-month horizon (0.24 +- 0.03 K vs. 0.38 +- 0.04 K for the persistence baseline and 0.32 +- 0.04 K for the pure ML ConvLSTM), confirming that physics-informed architectures outperform pure data-driven approaches on climate time scales. The FNO achieved the best spatial pattern reconstruction (SSIM = 0.92 +- 0.02 for global temperature fields) at 1,000x lower computational cost than equivalent GCM resolution. For extreme events, the vision transformer achieved the highest tropical cyclone genesis detection AUC (0.924 +- 0.008) by capturing large-scale atmospheric patterns across hemispheric fields. Gradient-boosted downscaling achieved the smallest regional precipitation bias (2.4 +- 0.8 mm/month) at 50x lower cost than dynamical downscaling. A practical model selection framework mapping forecast horizon, spatial resolution requirements, and computational budget to recommended AI approaches is proposed.