A novel method to improve temperature simulations of general circulation models based on ensemble empirical mode decomposition and its application to multi-model ensembles

摘要

A novel method based on the ensemble empirical mode decomposition (EEMD) method was developed to improve model performance. This method was evaluated by applying it to global surface air temperatures, which were simulated by eight general circulation models from the Coupled Model Intercomparison Project Phase 5 (CMIP5). The temperature simulations of the eight models were separated into their different components by EEMD. The model's performance improved after the first high-frequency component was removed from the original simulations by EEMD for each model, on both the global and continental scale. Moreover, EEMD was more effective in improving the model's performance compared to the wavelet transform method. The multi-model ensembles (MMEs) were calculated based on the EEMD-improved model simulations using the Average Ensemble Mean, Multiple Linear Regression, Singular Value Decomposition and Bayesian Model Averaging methods. The results showed that the MME forecasts performed better when the calculations were based on the EEMD-improved simulations as opposed to the original simulations on both the global and continental scale. Therefore, the results of the MME were further improved by using the EEMD-improved model simulations. This new method provides a simple way to improve model performance and can be easily applied to further improve MME simulations.