资源环境科技发展态势分析平台

The prediction of multiyear to decadal climate variability is important for stakeholders and decision-makers who are engaged in near-term planning activities. The decadal climate prediction experiments (DPEs) by predicting near-term climate with initialized global climate models (GCMs) provide robust skill at predicting sea surface temperature variability in some ocean regions as the North Atlantic. However, the state-of-the-art DPEs, which reproduce the observed warming trend associated with forced climate change, fail at predicting land surface air temperature (SAT) interdecadal variability. Here, we develop an effective statistical-dynamical model to predict spatial and temporal evolutions in Northern-Hemisphere (NH) summer land SAT. We identify two dominant interdecadal variability modes of the NH summer land SAT, whose evolutions are synchronized with forced climate change and Atlantic multidecadal variability (AMV), respectively. Based on statistical relationships with physical interpretations, time series of the forced responses and the AMV skillfully predicted by GCMs, the land SAT over the past one hundred years is predicted retrospectively with significantly improved skill compared to that predicted by the DPEs. Our results indicate that the decadal variability of the NH land SAT is predictable, with predictability rooted in atmospheric interdecadal circumglobal teleconnection (CGT) forced by the AMV. More skillful NH climate prediction by DPEs, which would be more practical for stakeholders and decision-makers, can be achieved by improving interdecadal CGT simulations in GCMs.