Tropical cyclones in the CMCC seasonal prediction system 3.5: performance and model bias

Tropical Cyclones (TCs) are powerful large-scale phenomena affecting tropical and subtropical coastal regions worldwide. For this reason, since the 1970s, research centers have been issuing seasonal forecasts to predict TC activity months in advance. In recent years, the statistical performance of dynamical seasonal forecast models has been extensively analyzed. The present study aims to systematically assess the performance of the CMCC-SPS3.5, a dynamical seasonal prediction system, in simulating TC activity and investigate the simulated physical relationship between TC climatology and large-scale drivers. In line with other studies, TC numbers are overestimated in the Southern Hemisphere, while showing regional differences in the Northern Hemisphere. Intense TCs are underestimated across all basins, which is likely due to coarse horizontal model resolution, limiting the ability to perform separate analysis for TC intensity and frequency. However, the model is skillful in predicting the interannual variability and the seasonal cycle's main features for TC numbers and pressure accumulated cyclone energy for most basins. Spatial distribution of TCs also shows significant basin-dependent differences. With respect to the coastline impact, the model has lower skill in the Southern Hemisphere and Eastern Pacific, whereas it has a better performance over the Western North Pacific and North Atlantic. The analysis indicated that dynamic variables are more helpful in explaining spatial distribution biases, supporting the idea that they are generally more skillful predictors for variations of TC activity. Analysis of a genesis potential index suggests that the simulated relation between TC and large-scale drivers is different compared to the observations. Lastly, in CMCC-SPS3.5 the impact of ENSO on large-scale drivers conditions tends to be overestimated spatially, which likely explains the amplification of ENSO-TC teleconnection.