Score-driven global climate zones from 1940 to 2024: A new objective climate classification method

We employ a novel score-driven climate clustering model to analyze global climate change from January 1940 to December 2024. This model dynamically reclassifies global geographic locations using climate variables from ERA5, which is the fifth-generation reanalysis from the European Centre for Medium-Range Weather Forecasts (ECMWF). The dataset includes monthly data from 930 uniformly distributed locations worldwide, covering eight variables: air temperature at 2 m, dew point temperature at 2 m, mean sea level pressure, eastward component of the 10-meter wind, northward component of the 10-meter wind, 10-meter wind gust, total precipitation, and downward short-wave solar radiation flux at the Earth’s surface. Our climate clustering model is a multivariate, score-driven, multi-regime-switching framework that analyzes the conditional mean and covariance matrix of these climate variables. It specifies a dynamic transition probability matrix for the different climate regimes, effectively representing climate clusters. We estimate the transition, filtered, and predictive probabilities for each climate regime. We focus on five distinct climate clusters. We present the evolution of the predictive probabilities for these clusters across various geographic locations experiencing observable climate change. Additionally, we explore alternative data and model specifications, guided by the feasibility of statistical estimation. Our findings include an annual analysis of 110 uniformly distributed locations worldwide for two variables: air and dew point temperatures at 2 m. The results indicate several geographic locations in both the Northern and Southern Hemispheres that transition to higher temperature clusters throughout the sample period.