Score-driven Models for Air Temperature Analysis

This study analyzes historical temperature trends using Generalized Autoregressive Score (GAS) models, which capture time-varying dynamics in time series data. Using temperature data from Czechia, we examine average temperatures as well as extremes, including maximum and minimum values, and identify a consistent upward trend across all indicators, particularly since 1950. The GAS models demonstrate superior in-sample fit and out-of-sample forecasting performance compared to traditional approaches such as ARMA and GARCH models. Forecasts based on the GAS model project a substantial increase in temperatures of 1.29 °C over the next 40 years.