学术活动

【9.10】Influences on annual global mean surface temperature since 1891 and real-time forecasts 2000-2012

【报告人】Dr. Chris Folland

【报告人单位】UK Met Office Hadley Centre

【报告时间】2012年9月10日(星期一)10:00

【报告地点】科研楼303会议室

 

Abstract

We first discuss the character and causes of the interannual to multidecadal variations and trends in annual global mean surface temperature over 1891-2010 in observations and in transient integrations of the HadCM3 coupled climate model for 1888-2002. We prominently use the new HadCRUT4 observed data set that has been recently published. Besides net increasing anthropogenic forcing from greenhouse gases and anthropogenic aerosols, four natural factors have a strong influence. These include the Atlantic Multidecadal Oscillation (AMO), the El Nino-Southern Oscillation (ENSO) and volcanic and solar forcing. Pacific Decadal Variability appears to have little explanatory power beyond the considerable contribution of ENSO which influences decadal trends as well as interannual variations. Most of the variance of observed and modelled global mean temperature over 1891-2010 can be explained by these factors in cross-validated multiple linear regression studies. Cross validation overcomes biased skill estimates due to persistence and the overfitting of multiple explanatory factors in linear regression. As important, it allows estimates of uncertainty in the regression results. We use these results to investigate the slowing of observed recent global warming in the light of climate model results showing that periods of no warming or even cooling well over a decade in length are expected in the presence of the current rate of increasing anthropogenic forcing due to natural variability. Our most important conclusion is that the AMO may have contributed about 20% to warming of observed global mean surface temperature since 1976.

The second part of the talk discusses the performance of real-time forecasts of global mean surface temperature by statistical and dynamical methods. The forecasts have been issued each year for the year ahead in press releases by the UK Met Office since 2000. They have a fairly high skill consistent with the skill of cross validated regression analyses. Since 2008, the forecasts have included a real-time component from the Met Office’s dynamical decadal forecast model, DePreSys. DePreSys is a version of HadCM3; DePreSys hindcasts back to 2000 have a high skill comparable to that of the issued forecasts. Moreover they also have significant skill in forecasting the spatial pattern of annual mean surface temperature for the coming year.

We conclude that we have a reasonable, though incomplete, understanding of the causes of the variations and trends in observed global mean surface temperature since 1891.

附件下载:
d