An event-based space–time analog approach is described for diagnosing ENSO mechanisms and variability in coupled models, specifically to evaluate whether these models capture the full range of ENSO variability through direct comparisons with observed events. In this analog approach, model states were drawn from an extensive library of states previously simulated in multi-century present-day simulations using the Community Earth System Model (CESM1) and the Energy Exascale Earth System Model (E3SM1) at both ocean eddy-parameterized (~ 1°) and ocean eddy-resolving (~ 0.1°) scales by matching their sea surface temperature (SST) and sea surface height (SSH) components to observed initial tropical Indo-Pacific SST and SSH anomalies. The subsequent space–time evolution of these states over the next several months in the simulations is evaluated for capturing selected ENSO events. The results indicate that neither model shows significant improvements in mean-state biases when the resolution is increased. In contrast to their biases, the low-resolution E3SM1-LR and CESM1-LR models consistently outperform their higher-resolution counterparts in all Eastern and Central Pacific warm events, as well as in cold events for SST and SSH-based analogs. However, RMSE values for SSH-based analogs display errors about twice as large as those of SST-based, which ramp up quickly with the analog ranking. The analog approach successfully captures the observed space–time evolution and peak characteristics of strong Eastern and Central Pacific warm and cold events. Nevertheless, all models display excessive variability in the western-central Pacific during both warm and cold events and struggle to accurately reproduce the magnitude of anomalies near the eastern boundary. Furthermore, the stronger the event, the more effective the analogs are, with only modest differences related to resolution, as low-resolution SST-based analogs consistently outperform their high-resolution counterparts.
Siqueira et al. (Fri,) studied this question.