Improvement of Climate Model Simulation through Inter-Model Diversity: An ENSO Example

Abstract:

In this study, a new methodology is developed to improve the climate simulation of state-of-the-art coupled GCMs, by post-processing based on the inter-model diversity (i.e. ensemble spread from the Multi-Model Ensemble (MME)). Based on the close connection between the interannual variability and climatological state, the distinctive relation between the inter-model diversity of the interannual variability, and that of the basic state, is found. Based on this relation, the simulated interannual variabilities can be improved, by correcting their climatological bias. In order to test this methodology, the dominant inter-model difference in precipitation responses during the ENSO is investigated, and its relationship with climatological state. It is found that the dominant inter-model diversity of the ENSO precipitation in CMIP5 is associated with the zonal location shift of the positive precipitation center during El Nino. This dominant inter-model difference is significantly correlated with the difference in the basic state. The models with wetter (dryer) climatology than the climatology of the MME over the central Pacific, tend to shift positive ENSO precipitation anomalies to the east (west). Based on the model’s systematic errors in atmospheric ENSO response and bias, it is shown that the models with better climatological state tend to simulate more realistic atmospheric ENSO responses. Therefore, the statistical method to correct the ENSO response by minimizing mean bias mostly improves the ENSO response. After the statistical correction, the deficiencies in simulating the MME ENSO precipitation are improved, so that the pattern correlation of tropical atmospheric MME response is increased from 0.81 before the correction, to 0.92 after the correction. In particular, this improvement is robust in the models whose original response is far from realistic. These results provide the possibility that the methodology developed in this study can also be applied to improve climate projection and seasonal climate prediction.