Abstract
The tropical cyclone genesis frequency (TCGF) is an essential metric for gauging the performance of climate models. Previous evaluations on CMIP family models usually employ one realization for each model and show their diversities in performance. The single model initial condition large ensemble experiments provide a unique opportunity to quantify how internal variability may affect the model evaluation skill. Here, taking the TCGF in the Western North Pacific (WNP) as an example, we use two genesis potential indices as proxies to evaluate the performance of the FGOALS-g3 large ensemble simulation with 110 members. We show that while internal variability does not have a significant influence on the TCGF mean state evaluation, the TCGF-ENSO (El Niño–Southern Oscillation) relationship is significantly modulated by the decadal scale internal variability. For mean state simulation, the FGOALS-g3 large ensembles show reasonable performance in the simulation of TCGF spatial pattern but have differences compared with ERA5 in magnitude. Physical process analysis indicates that compared with ERA5, nearly all dynamic terms are more unfavorable for tropical cyclogenesis due to the cold sea surface temperature anomalies in the midlatitude, while the thermodynamic terms are more conducive to more TCs. For interannual variability, the ENSO-TCGF connection is significantly modulated by the tropical Pacific decadal variability (TPDV) mode by influencing the vertical wind shear in the WNP. Particularly, the model simulation skill depends on the choice of genesis potential indices. Our finding highlights the importance of considering decadal-scale internal variability in the evaluation of interannual ENSO-TCGF variability.