a, James Ross Island (JRI) temperature reconstruction7, 9 (green) alongside continent-scale temperature reconstructions5 for South America (red) and Antarctica (blue; excludes JRI). Anomalies shown as 7 yr (thin grey lines) and 70 yr (thick lines and grey shading) moving averages, relative to AD 1961–1990 means (dashed lines). b, Location of JRI (green cross) and proxies used in the South America (red crosses) and Antarctica (blue crosses) temperature reconstructions. Shading shows spatial correlation coefficient (r; p < 0.1) of the annual SAM index1 with 2 m air temperature in the ERA-Interim reanalysis16 (January–December averages; AD 1979– 2012).
Proxy records.
We use the deuterium isotope record from the JRI ice core as a temperature proxy for the Antarctic Peninsula region8, 9. Full details of the ice core site and isotope analysis can be found in ref. 7. We use the JRI1 age model with annual layer chronology since AD 1807, as in ref. 9. Deuterium isotope measurements made at 10 cm resolution along the upper 300 m of the ice core correspond to better than annual resolution since AD 1111 and were binned to produce annual (~ January–December) averages. The 111 years between AD 1000 and 1110 comprise 85 isotope measurements and interpolation was used to generate a pseudo-annual resolution record over this interval.
We also use temperature-sensitive proxy records for the Antarctic and South America continental regions5 to capture the full mid-latitude to polar expression of the SAM across the Drake Passage transect. The annually resolved proxy data sets compiled as part of the PAGES2k database are published and publically available5. For the South American data set we restrict our use to records south of 30° S and we do not use the four shortest records that are derived from instrumental sources. Details of the individual records used here and their correlation with the SAM are given in Supplementary Table 1.
Data are available at http://www.nature.com/nclimate/journal/v4/n7/full/ftp://ftp.ncdc.noaa.gov/pub/data/paleo/icecore/antarctica/james-ross-island/ and http://www.nature.com/ngeo/journal/v6/n5/full/ngeo1797.html
SAM reconstruction.
The proxy records from the South America, Antarctic Peninsula and Antarctic continent regions (where SAM has a significant influence on temperature; Fig. 1b) were used to reconstruct an annual average SAM index since AD 1000. We employ the widely used composite plus scale (CPS) methodology5, 11, 12 with nesting to account for the varying length of proxies making up the reconstruction. For each nest the contributing proxies were normalized relative to the AD 1957–1995 calibration interval, which represents the interval of maximum overlap between the annual (January–December) Marshall–SAM index (http://www.antarctica.ac.uk/met/gjma/sam.html) and most of the proxy network (Supplementary Table 1). The normalized proxy records were then combined with a weighting12 based on their correlation coefficient (r) with the SAM during the calibration interval (Supplementary Table 1). The combined record was then scaled to match the mean and standard deviation of the instrumental SAM index during the calibration interval. Finally, nests were spliced together to provide the full 1,008-year SAM reconstruction. Alternate methods for carrying out the CPS reconstruction were explored and the primary findings discussed here are shown to be robust across different methodologies (Supplementary Figs 1 and 2).
For each proxy nest a 95% confidence interval was defined as 1.96 times the standard deviation of the residuals of the SAM reconstruction from the Marshall–SAM index during the calibration interval. The reduction of error statistic was also calculated to test the performance of the reconstruction. The brevity of Antarctic instrumental records limits the ability to cross-validate the SAM reconstruction using separate calibration and verification intervals18. Instead, we assess the significance of reduction of error values by repeating 1,000 CPS simulations where the proxy network was replaced by AR(1) time series matching the length and lag-1 autocorrelation of the proxies and we use the upper 95th percentile to determine the critical reduction of error level (REcrit) for each proxy nest. We further verify the SAM reconstruction against the extended Fogt–SAM index10 (http://polarmet.osu.edu/ACD/sam/sam_recon.html). To carry out this assessment the four seasonal reconstructions were averaged to estimate an annual (December–November) Fogt–SAM index, which was then scaled to match the variance of the Marshall–SAM index from 1961 to 1990.
Model output.
We use multimodel output from the subset of CMIP5 climate models that ran transient Last Millennium simulations since AD 85021, 22. Historical simulations from the same ensemble were used to extend the model output from AD 1850. The CMIP5 Last Millennium and Historical experiments use transient radiative forcings that include orbital, solar, volcanic, greenhouse and ozone parameters as well as land use changes21, 22. All data was accessed from the Earth System Grid Federation node (http://pcmdi9.llnl.gov/esgf-web-fe/), with the exception of the historical portion of the HadCM3 Last Millennium simulation (provided by A. Schurer, Edinburgh University) and the Commonwealth Scientific and Industrial Research Organisation (CSIRO) Mk3L simulations (http://www.nature.com/nclimate/journal/v4/n7/full/ftp://ftp.ncdc.noaa.gov/pub/data/paleo/gcmoutput/phipps2014/). To assess the importance of different radiative-forcing mechanisms, we used multiple simulations of the past 1,500 years carried out with the CSIRO Mk3L coupled climate model, as described in ref. 23. Supplementary Table 2 gives further details on the climate model data sets.
We use monthly resolution mean sea-level pressure fields to calculate the zonal mean at 40° S and 65° S. The model-generated data were averaged into January–December annuals to match the proxy data, normalized relative to the AD 1961–1990 interval and differenced to generate a SAM index1. We also use surface air temperature model output to examine SAM–temperature relationships at our proxy sites in the Last Millennium climate simulations (Supplementary Fig. 4).
Data archive.
The SAM reconstruction developed here is archived with the World Data Center for Paleoclimatology (http://hurricane.ncdc.noaa.gov/pls/paleox/f?p=519:1:0::::P1_STUDY_ID:16197).