Surface Temperature Prediction with Bayesian Structral Time Series
Madhavan Balasubramaniam · Preprints.org · 2025
This article develops a Bayesian Auto-Regressive forecasting model for predicting global surface temperatures and compares its performance with a frequentist approach. Using the NASA GISS Surface Temperature dataset, we first implement an AR(4) model and then incorporate trend and seasonality components. Results show that the Bayesian approach improves generalization and provides probabilistic parameter estimates, making it more robust for long-term forecasting.