Comparison of the Pacific Decadal Oscillation in climate model simulations and observations

The ability of climate models to simulate the Pacific Decadal Oscillation (PDO) and its global teleconnections is examined using the outputs of 20 Coupled Model Intercomparison Project (CMIP5) models from historical experiments. It is shown that the global warming trend is reproduced by the models, but there is a considerable difference between the models in the PDO region. The standard deviation of detrended sea surface temperature anomalies (SSTAs) in the models is similar to that of the observations, especially in the central and eastern equatorial Pacific and the North Pacific, which are the regions closely associated with the PDO and El Nino-Southern Oscillation. The globally averaged SSTA during 1900-1940 in the CMIP5 models and observations showed large divergence, which is mainly due to the SSTA in the PDO region. The most significant cycle of the PDO indexes in most of the models had a 60-80 years period, which is different from the observed PDO period (40-60 years). However, the models perform better for the spatial pattern of the PDO, and 90% of the spatial correlations between the models and observations of the spatial pattern of the PDO are above 0.8. In addition, the models can simulate the basic correlation between the PDO and Pacific SSTA variability, but most of them overestimate the relationship in the Indian and Atlantic Oceans. The CMIP5 models are better at simulating the impact of the PDO on winter precipitation in southern North America but are less accurate in simulating summer precipitation in eastern China. This is probably because most of the models are not accurate in simulating the impacts of the PDO on the western Pacific subtropical high and the East Asian summer monsoon.