We use a novel score-driven climate clustering model to study climate changefrom 1940 to 2024. The model performs a dynamic reclassification of global geographic locations using climate variables. We obtained data from ERA5, the fifth-generation European Centre for Medium-Range Weather Forecasts (ECMWF) reanalysis. Global climate and weather data are taken from 930 locations worldwide. The climate clustering model, which involves a smallnumber of parameters, is a multivariate score-driven multi-regime-switching model of the conditional mean vector and the conditional covariance matrix of climate variables. The model specifies a dynamic transition probability matrix for regimes, representing climate clusters. We consider alternative specifications of the model, influenced by model performance and statisticalestimation quality. Motivated by the literature, we consider five climate clusters and report the evolution of the predictive probability of each climate cluster for the geographic locations where climate change is observed. The results indicate several geographic locations in the Northern and Southern Hemispheres that switch to higher temperature clusters during the sample period.