Towards better characterization of global warming impacts in the environment through climate classifications with improved global models

Climate classifications are useful to synthesize the physical state of climate with a proxy that can be directly related biota. However, they present potential drawback, namely strong sensitivity because use hard thresholds (step functions). Thus, minor discrepancies in base data produce large differences type climate. such an priori limitation is also strength used better gauge model performance. Although previous attempts classifying climates world using global outputs were encouraging, applicability their impact studies has been limited by past issues. Notwithstanding persistence certain imperfections and limitations current models, high-quality simulations from phase six Coupled Intercomparison Project (CMIP6) opened new possibilities field, thanks improved representation atmospheric oceanic physics. The purpose this paper twofold: show CMIP6 sufficiently robust for studies, those identifying error sources issues deserve further attention models. 52 models evaluated three schemes, classical Köppen, extended-Köppen, modified Thornthwaite. We first assessed ability reproduce types comparing observational data. Models performed best Köppen extended-Köppen classification methods (Cohen's kappa ? = 0.7), had moderate scores Thornthwaite (? 0.4). By tracing back observed biases, we able pinpoint misrepresentation dry as major source error. Another finding was most still some difficulties representing seasonal variability precipitation over several monsoonal regions, thereby yielding wrong there. future Substantial changes projected SSP5-8.5 scenario. These include shrinkage polar/frigid (22%) increase (7%).