资源环境科技发展态势分析平台

We assess the interactions between stratospheric water vapor (SWV) and surface temperature during the past two decades using satellite observations and the Community Earth System Model (CESM). From 1992 to 2013, to first order, the observed SWV exhibited three distinct piece-wise trends: a steady increase from 1992 to 2000, an abrupt drop from 2000 to 2004, and a gradual recovery after 2004, while the global-mean surface temperature experienced a strong increase until 2000 and a warming hiatus after 2000. The atmosphere-only CESM shows that the seasonal variation of tropical-mean (30A degrees S-30A degrees N) SWV is anticorrelated with that of the tropical-mean sea surface temperature (SST), while the correlation between the tropical SWV and SST anomalies on the interannual time scale is rather weak. By nudging the modeled SWV to prescribed profiles in coupled atmosphere-slab ocean experiments, we investigate the impact of SWV variations on surface temperature change. We find that a uniform 1 ppmv (0.5 ppmv) SWV increase (decrease) leads to an equilibrium global mean surface warming (cooling) of 0.12 +/- 0.05 A degrees C (-0.07 +/- 0.05 A degrees C). Sensitivity experiments show that the equilibrium response of global mean surface temperature to SWV perturbations over the extratropics is larger than that over the tropics. The observed sudden drop of SWV from 2000 to 2004 produces a global mean surface cooling of about -0.048 +/- 0.041 A degrees C, which suggests that a persistent change in SWV would make an imprint on long-term variations of global-mean surface temperature. A constant linear increase in SWV based on the satellite-observed rate of SWV change yields a global mean surface warming of 0.03 +/- 0.01 A degrees C/decade over a 50-year period, which accounts for about 19 % of the observed surface temperature increase prior to the warming hiatus. In the same experiment, trend analyses during different periods reveal a multi-year adjustment of surface temperature before the response to SWV forcing becomes strong relative to the internal variability in the model.