Basin Dependence of the Tropical Cyclone Genesis Environment and the Future Changes Revealed by Machine Learning Methods

Abstract Due to the limited data available and the lack of consensus on a tropical cyclone (TC) genesis theory, understanding the differences in TC genesis in different ocean basins is limited, and possible changes in TC genesis under the influence of future climate change are controversial. This study applies principal component analysis (PCA) to investigate the TC genesis environment in different ocean basins. The results show that the TC genesis environments are similar in all basins except the North Atlantic Ocean (NA). Compared with other basins, the TC genesis environment in the NA has a higher potential intensity (PI) and mid-level vertical velocity and a lower low and mid-level relative humidity. A maximum entropy (MaxEnt) model is established using various observational environmental variables and is subsequently transferred to the Coupled Model Intercomparison Project Phase 6 (CMIP6) model output to produce future projections of TC genesis. The MaxEnt model captures the unique TC genesis environment in the NA and predicts a more consistent decrease in the TC genesis probability in the 2100s under different future scenarios. A multivariate environmental similarity surface (MESS) analysis is applied to the CMIP6 model dataset, and it reveals that the CMIP6 model predicts the weakest change in the TC genesis environment in the NA compared to other ocean basins. Overall, our machine learning (ML) results suggest that the TC genesis environment in the NA has different characteristics compared to other ocean basins, and its TC genesis probability is projected, with high confidence, to decrease in the future using the CMIP6 models.