Evaluation of the Pacific Decadal Oscillation from 1901 to 2014 in CMIP6 models

Abstract

The simulation of internal climate variability is key to understanding climate change. Internal fluctuations can fully obscure or amplify the underlying climate-change signal in many fields over years to decades. The Pacific Decadal Oscillation (PDO) is an important mechanism that dominates interdecadal climate variations. Here, the capacity of 36 Coupled Model Intercomparison Project Phase 6 (CMIP6) models for reproducing the PDO during the period 1901-2014 was evaluated across spatial pattern, amplitude, periodicity and phase aspects. The results suggested that approximately 71% of the models were capable of capturing the spatial pattern of the PDO. However, most of the other models underestimated the percentage of the variance explained and the pattern amplitude. CMIP6 models can capture the periodicity over shorter periods (<20 yr), while most CMIP6 models underestimate the periodicity of the PDO on the interdecadal scale (60 yr periods). In particular, the probability distribution functions (PDFs) of the raw PDO index showed that most CMIP6 models could capture the unimodal probability distribution. Some of the models were able to reproduce the bimodal probability distribution of the PDO index filtered by the ensemble empirical mode decomposition method. Moreover, the findings also showed that the multimodel ensemble (MME) for 36 CMIP6 models could effectively capture the spatial pattern and periodicity over a shorter period (<20 yr) of the PDO. Moreover, the MME exhibited higher fidelity to the observation in the PDF of the PDO index, although the positive area was larger than the observation. These findings are crucial because they may help detect simulation gaps and provide valuable references for the design and applicability of future models worldwide.