Joint statistical-dynamical approach to decadal prediction of East Asian surface air temperature

Abstract

A joint statistical-dynamical method addressing both the internal decadal variability and effect of anthropogenic forcing was developed to predict the decadal components of East Asian surface air temperature (EATs) for three decades (2010–2040). As previous studies have revealed that the internal variability of EATs (EATs_int) is influenced mainly by the ocean, we first analyzed the lead-lag connections between EATs_int and three sea surface temperature (SST) multidecadal modes using instrumental records from 1901 to 1999. Based on the lead-lag connections, a multiple linear regression was constructed with the three SST modes as predictors. The hindcast for the years from 2000 to 2005 indicated the regression model had high skill in simulating the observational EATs_int. Therefore, the prediction for EATs_int (Re_EATs_int) was obtained by the regression model based on quasi-periods of the decadal oceanic modes. External forcing from greenhouse gases is likely associated with global warming. Using monthly global land surface air temperature from historical and projection simulations under the Representative Concentration Pathway (RCP) 4.5 scenario of 19 Coupled General Circulation Models participating in the fifth phase of the Coupled Model Intercomparison Project (CMIP5), we predicted the curve of EATs (EATs_trend) relative to 1970–1999 by a second-order fit. EATs_int and EATs_trend were combined to form the reconstructed EATs (Re_EATs). It was expected that a fluctuating evolution of Re_EATs would decrease slightly from 2015 to 2030 and increase gradually thereafter. Compared with the decadal prediction in CMIP5 models, Re_EATs was qualitatively in agreement with the predictions of most of the models and the multi-model ensemble mean, indicating that the joint statistical-dynamical approach for EAT is rational.