9:30 AM - 9:45 AM
[ACG38-03] Emulating GCM Experiments with Reduced-Complexity Models: An application to Tropical Interbasin Interactions
Keywords:Tropical interbasin interaction, GCM pacemaker experiment, Reduced complexity model, Linear inverse model (LIM)
Pacemaker experiments using general circulation models (GCMs) are widely employed as a framework for understanding remotely forced response of atmosphere-ocean coupled system. Meanwhile, reduced-order models such as the linear inverse model (LIM) and the extended recharge oscillator (XRO) have been extensively used to investigate the mechanisms of climate variability and predictability, particularly in the tropics, with a focus on basin-to-basin interactions. Given this background, we conducted a perfect model experiment that integrates these two approaches, examining whether LIM and XRO can reproduce the results of GCM pacemaker experiments targeting tropical variability.
As a case study, we investigated the remote impacts of the tropical Atlantic variability, which have received increasing attention in recent studies and is a common application of pacemaker experiments. We conducted pacemaker experiments using a coupled atmosphere-ocean GCM in which the sea surface temperature (SST) of the tropical Atlantic was restored to that of a control simulation. LIM and XRO were trained on the control simulation, and a newly developed method was applied to emulate the pacemaker experiments.
Our results show that the deterministic signal extracted from the ensemble mean of the GCM pacemaker experiments was successfully reproduced by both models. Furthermore, while many GCM pacemaker experiments adopt an ensemble size of around 10 members, few studies have critically examined this choice. Our analysis suggests that such ensemble sizes may not be sufficient for obtaining robust deterministic signals, and that it is desirable to use more than 50 ensemble members to obtain robust signals at least within this particular modeling framework.
These findings suggest that reduced-order models can complement GCMs and improve the interpretation of pacemaker experiments. They also highlight the influence of ensemble size on the robustness of the remotely forced response, indicating the potential need for conducting large-ensemble GCM pacemaker experiments in future studies.
As a case study, we investigated the remote impacts of the tropical Atlantic variability, which have received increasing attention in recent studies and is a common application of pacemaker experiments. We conducted pacemaker experiments using a coupled atmosphere-ocean GCM in which the sea surface temperature (SST) of the tropical Atlantic was restored to that of a control simulation. LIM and XRO were trained on the control simulation, and a newly developed method was applied to emulate the pacemaker experiments.
Our results show that the deterministic signal extracted from the ensemble mean of the GCM pacemaker experiments was successfully reproduced by both models. Furthermore, while many GCM pacemaker experiments adopt an ensemble size of around 10 members, few studies have critically examined this choice. Our analysis suggests that such ensemble sizes may not be sufficient for obtaining robust deterministic signals, and that it is desirable to use more than 50 ensemble members to obtain robust signals at least within this particular modeling framework.
These findings suggest that reduced-order models can complement GCMs and improve the interpretation of pacemaker experiments. They also highlight the influence of ensemble size on the robustness of the remotely forced response, indicating the potential need for conducting large-ensemble GCM pacemaker experiments in future studies.