Anthropogenic influence on the autocorrelation structure of hemispheric- mean temperatures

Science

Volumn 282, Issue 5394, 1998, Pages 1676-1679

  Wigley, T M L ^a   Smith, R L ^b   Santer, B D ^c  

^a NATIONAL CENTER FOR ATMOSPHERIC RESEARCH   (United States)

^b University of North Carolina at Chapel Hill   (United States)

^c Program for Climate Model Diagnosis and Intercomparison   (United States)

Author keywords

[No Author keywords available]

Indexed keywords

AIR TEMPERATURE; ANTHROPOGENIC EFFECT; AUTOCORRELATION; CLIMATE FORCING;

ANTHROPOLOGY; ARTICLE; DATA ANALYSIS; GLOBAL CLIMATE; PHYSICAL MODEL; PRIORITY JOURNAL; SIMULATION; STRUCTURE ANALYSIS; TEMPERATURE MEASUREMENT;

EID: 0032573380     PISSN: 00368075     EISSN: None     Source Type: Journal    
DOI: 10.1126/science.282.5394.1676     Document Type: Article

References (40)

1. R. A. Madden and V. Ramanathan, Science 209, 763 (1980); T. M. L. Wigley and P. D. Jones, Nature 292, 205 (1981).
2. R. A. Madden and V. Ramanathan, Science 209, 763 (1980); T. M. L. Wigley and P. D. Jones, Nature 292, 205 (1981).
3. T. P. Barnett and M. E. Schlesinger, J. Geophys. Res. 92, 14772 (1987); B. D. Santer, T. M. L. Wigley, P. D. Jones, Clim. Dyn. 8, 265 (1993); B. D. Santer et al., ibid. 12, 77 (1995); G. C. Hegerl et al., J. Clim. 9, 2281 (1996).
4. T. P. Barnett and M. E. Schlesinger, J. Geophys. Res. 92, 14772 (1987); B. D. Santer, T. M. L. Wigley, P. D. Jones, Clim. Dyn. 8, 265 (1993); B. D. Santer et al., ibid. 12, 77 (1995); G. C. Hegerl et al., J. Clim. 9, 2281 (1996).
5. T. P. Barnett and M. E. Schlesinger, J. Geophys. Res. 92, 14772 (1987); B. D. Santer, T. M. L. Wigley, P. D. Jones, Clim. Dyn. 8, 265 (1993); B. D. Santer et al., ibid. 12, 77 (1995); G. C. Hegerl et al., J. Clim. 9, 2281 (1996).
6. T. P. Barnett and M. E. Schlesinger, J. Geophys. Res. 92, 14772 (1987); B. D. Santer, T. M. L. Wigley, P. D. Jones, Clim. Dyn. 8, 265 (1993); B. D. Santer et al., ibid. 12, 77 (1995); G. C. Hegerl et al., J. Clim. 9, 2281 (1996).
7. B. D. Santer et al., Nature 382, 39 (1996); S. F. B. Tett, J. F. B. Mitchell, D. E. Parker, M. R. Allen, Science 274, 1170 (1996).
8. B. D. Santer et al., Nature 382, 39 (1996); S. F. B. Tett, J. F. B. Mitchell, D. E. Parker, M. R. Allen, Science 274, 1170 (1996).
9. i-n). n may be positive (Y leading) or negative (X leading). When X = Y, these lag correlations are referred to as autocorrelations, terminology that we apply here to both the X = Y and X ≠ Y cases.
10. W. A. Woodward and H. L. Gray, J. Clim. 6, 953 (1993); ibid. 8, 1929 (1995); R. S. J. Tol, Theor. Appl. Climatol. 49, 91 (1994); R. K. Kaufmann and D. I. Stem, Nature 388, 39 (1997).
11. W. A. Woodward and H. L. Gray, J. Clim. 6, 953 (1993); ibid. 8, 1929 (1995); R. S. J. Tol, Theor. Appl. Climatol. 49, 91 (1994); R. K. Kaufmann and D. I. Stem, Nature 388, 39 (1997).
12. W. A. Woodward and H. L. Gray, J. Clim. 6, 953 (1993); ibid. 8, 1929 (1995); R. S. J. Tol, Theor. Appl. Climatol. 49, 91 (1994); R. K. Kaufmann and D. I. Stem, Nature 388, 39 (1997).
13. W. A. Woodward and H. L. Gray, J. Clim. 6, 953 (1993); ibid. 8, 1929 (1995); R. S. J. Tol, Theor. Appl. Climatol. 49, 91 (1994); R. K. Kaufmann and D. I. Stem, Nature 388, 39 (1997).
14. We used the same observed data as IPCC, updated (P. D. Jones, personal communication). For sources, see: N. Nicholls et al., in Climate Change 1995: The Science of Climate Change, J. T. Houghton et al., Eds. (Cambridge Univ. Press, Cambridge, 1996), pp. 133-192. Our analyses span 1881-1996. The conclusions of this paper do not depend on the precise start or end points.
15. We used the same observed data as IPCC, updated (P. D. Jones, personal communication). For sources, see: N. Nicholls et al., in Climate Change 1995: The Science of Climate Change, J. T. Houghton et al., Eds. (Cambridge Univ. Press, Cambridge, 1996), pp. 133-192. Our analyses span 1881-1996. The conclusions of this paper do not depend on the precise start or end points.
16. We use data from two 1000-year unforced simulations with coupled ocean/atmosphere general circulation models: the Geophysical Fluid Dynamics Laboratory (GFDL) model [S. Manabe and R. J. Stouffer, J. Clim. 9, 376 (1996)]; and the U.K. Hadley Centre model (HadCM2) [S. F. B. Tett, T. C. Johns, J. F. B. Mitchell, Clim. Dyn. 13, 303 (1997)]. For the control-run results we used the full areal coverage to define the hemispheric means. For the observed data, coverage is incomplete and tends to increase with time. To test whether such coverage differences affected our results, we masked the control-run data with typical observed coverages and re-computed the correlations. The results were simitar to the full-coverage results. Standard errors associated with the sample autocorrelations are typically of the order 0.05 or smaller in the case of the model data, and in the range 0.1 to 0.15 for the observational data. These were calculated by applying standard asymptotic formulae for the variance of sample autocorrelations [for example, see. p. 342 of W. A. Fuller, Introduction to Statistical Time Series, (Wiley-Interscience, ed. 2, New York, 1996)]. In doing so, we assume that the true autocorrelations used for calculating the theoretical results are those estimated from the model data. The result that the standard errors are larger for the observed series than the model series reflects the difference in sample sizes (116 against 1000). These results were also checked from the model data using a resampling procedure, based on the empirical standard deviation of sample autocorrelations calculated from maximally overlapping 116-year subseries of the 1000-year model runs; this produced results consistent with the asymptotic formulae. The results show that the difference between sample autocorrelations for the observed and either of the control-run series are 2 to 3 times the standard errors for the observed series. On this basis we conclude that the two sets of autocorrelations are indeed significantly different.
17. We use data from two 1000-year unforced simulations with coupled ocean/atmosphere general circulation models: the Geophysical Fluid Dynamics Laboratory (GFDL) model [S. Manabe and R. J. Stouffer, J. Clim. 9, 376 (1996)]; and the U.K. Hadley Centre model (HadCM2) [S. F. B. Tett, T. C. Johns, J. F. B. Mitchell, Clim. Dyn. 13, 303 (1997)]. For the control-run results we used the full areal coverage to define the hemispheric means. For the observed data, coverage is incomplete and tends to increase with time. To test whether such coverage differences affected our results, we masked the control-run data with typical observed coverages and re-computed the correlations. The results were simitar to the full-coverage results. Standard errors associated with the sample autocorrelations are typically of the order 0.05 or smaller in the case of the model data, and in the range 0.1 to 0.15 for the observational data. These were calculated by applying standard asymptotic formulae for the variance of sample autocorrelations [for example, see. p. 342 of W. A. Fuller, Introduction to Statistical Time Series, (Wiley-Interscience, ed. 2, New York, 1996)]. In doing so, we assume that the true autocorrelations used for calculating the theoretical results are those estimated from the model data. The result that the standard errors are larger for the observed series than the model series reflects the difference in sample sizes (116 against 1000). These results were also checked from the model data using a resampling procedure, based on the empirical standard deviation of sample autocorrelations calculated from maximally overlapping 116-year subseries of the 1000-year model runs; this produced results consistent with the asymptotic formulae. The results show that the difference between sample autocorrelations for the observed and either of the control-run series are 2 to 3 times the standard errors for the observed series. On this basis we conclude that the two sets of autocorrelations are indeed significantly different.
18. We use data from two 1000-year unforced simulations with coupled ocean/atmosphere general circulation models: the Geophysical Fluid Dynamics Laboratory (GFDL) model [S. Manabe and R. J. Stouffer, J. Clim. 9, 376 (1996)]; and the U.K. Hadley Centre model (HadCM2) [S. F. B. Tett, T. C. Johns, J. F. B. Mitchell, Clim. Dyn. 13, 303 (1997)]. For the control-run results we used the full areal coverage to define the hemispheric means. For the observed data, coverage is incomplete and tends to increase with time. To test whether such coverage differences affected our results, we masked the control-run data with typical observed coverages and re-computed the correlations. The results were simitar to the full-coverage results. Standard errors associated with the sample autocorrelations are typically of the order 0.05 or smaller in the case of the model data, and in the range 0.1 to 0.15 for the observational data. These were calculated by applying standard asymptotic formulae for the variance of sample autocorrelations [for example, see. p. 342 of W. A. Fuller, Introduction to Statistical Time Series, (Wiley-Interscience, ed. 2, New York, 1996)]. In doing so, we assume that the true autocorrelations used for calculating the theoretical results are those estimated from the model data. The result that the standard errors are larger for the observed series than the model series reflects the difference in sample sizes (116 against 1000). These results were also checked from the model data using a resampling procedure, based on the empirical standard deviation of sample autocorrelations calculated from maximally overlapping 116-year subseries of the 1000-year model runs; this produced results consistent with the asymptotic formulae. The results show that the difference between sample autocorrelations for the observed and either of the control-run series are 2 to 3 times the standard errors for the observed series. On this basis we conclude that the two sets of autocorrelations are indeed significantly different.
19. 2 greenhouse gases, and sulfate aerosol effects were modeled by using changes in surface albedo.
20. 2 greenhouse gases, and sulfate aerosol effects were modeled by using changes in surface albedo.
21. T. M. L. Wigley and S. C. B. Raper, Nature 357, 293 (1992); S. C. B. Raper, T. M. L. Wigley, R. A. Warrick, in Sea-Level Rise and Coastal Subsidence: Causes, Consequences and Strategies, J. D. Milliman and B. U. Haq, Eds. (Kluwer, Dordrecht, Netherlands, 1996), pp. 11-45. Simulations run from 1765 through 1996. The model differentiates between land and ocean in each hemisphere and gives hemispherically-specific temperature change results.
22. T. M. L. Wigley and S. C. B. Raper, Nature 357, 293 (1992); S. C. B. Raper, T. M. L. Wigley, R. A. Warrick, in Sea-Level Rise and Coastal Subsidence: Causes, Consequences and Strategies, J. D. Milliman and B. U. Haq, Eds. (Kluwer, Dordrecht, Netherlands, 1996), pp. 11-45. Simulations run from 1765 through 1996. The model differentiates between land and ocean in each hemisphere and gives hemispherically-specific temperature change results.
23. P. D. Jones, Clim. Mon. 17, 80 (1988); updated (P. D. Jones, personal communication).
24. P. D. Jones, Clim. Mon. 17, 80 (1988); updated (P. D. Jones, personal communication).
25. Reviewed by C. K. Folland, T. R. Karl, K. Ya. Vinnikov, in Climate Change. The IPCC Scientific Assessment, J. T. Houghton, G. J. Jenkins, J. J. Ephraums, Eds. (Cambridge Univ. Press, Cambridge, 1990), pp. 195-238; C. K. Folland et al., in Climate Change 1992: The Supplementary Report to the IPCC Scientific Assessment, J. T. Houghton, B. A. Callander, S. K. Varney, Eds. (Cambridge Univ. Press, Cambridge, 1992), pp. 135-170; and by N. Nicholls et al. (6).
26. Reviewed by C. K. Folland, T. R. Karl, K. Ya. Vinnikov, in Climate Change. The IPCC Scientific Assessment, J. T. Houghton, G. J. Jenkins, J. J. Ephraums, Eds. (Cambridge Univ. Press, Cambridge, 1990), pp. 195-238; C. K. Folland et al., in Climate Change 1992: The Supplementary Report to the IPCC Scientific Assessment, J. T. Houghton, B. A. Callander, S. K. Varney, Eds. (Cambridge Univ. Press, Cambridge, 1992), pp. 135-170; and by N. Nicholls et al. (6).
27. Reviewed by C. K. Folland, T. R. Karl, K. Ya. Vinnikov, in Climate Change. The IPCC Scientific Assessment, J. T. Houghton, G. J. Jenkins, J. J. Ephraums, Eds. (Cambridge Univ. Press, Cambridge, 1990), pp. 195-238; C. K. Folland et al., in Climate Change 1992: The Supplementary Report to the IPCC Scientific Assessment, J. T. Houghton, B. A. Callander, S. K. Varney, Eds. (Cambridge Univ. Press, Cambridge, 1992), pp. 135-170; and by N. Nicholls et al. (6).
28. A. Kattenberg et al., in Climate Change 1995: The Science of Climate Change, J. T. Houghton et al., Eds. (Cambridge Univ. Press, Cambridge, 1996), pp. 285-357.
29. We have, nevertheless, tested this by subtracting estimated volcanic effects for all known major eruptions from the observed data. The influence on the autocorrelation structure is negligible.
30. D. V. Hoyt and K. H. Schatten, J. Geophys. Res. 98, 18895 (1994) . Other reconstructions, such as by J. L. Lean, J. Beer, and R. Bradley [Geophys. Res. Lett. 22, 3195 (1995)], are similar. The results presented here do not depend on which solar data set is used.
31. D. V. Hoyt and K. H. Schatten, J. Geophys. Res. 98, 18895 (1994) . Other reconstructions, such as by J. L. Lean, J. Beer, and R. Bradley [Geophys. Res. Lett. 22, 3195 (1995)], are similar. The results presented here do not depend on which solar data set is used.
32. 2 data. The IS92a scenario is the central "existing policies" scenario produced by IPCC in 1992 [J. A. Leggett, W. J. Pepper, R. J. Swart, in Climate Change 1992: The Supplementary Report to the IPCC Scientific Assessment, J. T. Houghton, B. A. Callander, S. K. Varney, Eds. (Cambridge Univ. Press, Cambridge, 1992), pp. 69-95]. The modifications made to this scenario are described in T. M. L. Wigley, Geophys. Res. Lett. 25, 2285 (1998).
33. 2 data. The IS92a scenario is the central "existing policies" scenario produced by IPCC in 1992 [J. A. Leggett, W. J. Pepper, R. J. Swart, in Climate Change 1992: The Supplementary Report to the IPCC Scientific Assessment, J. T. Houghton, B. A. Callander, S. K. Varney, Eds. (Cambridge Univ. Press, Cambridge, 1992), pp. 69-95]. The modifications made to this scenario are described in T. M. L. Wigley, Geophys. Res. Lett. 25, 2285 (1998).
34. -2).
35. T. M. L. Wigley, P. D. Jones, S. C. B. Raper, Proc. Natl. Acad. Sci. U.S.A. 94, 8314 (1997). Note that these climate sensitivities depend on the assumed magnitudes of anthropogenic and solar forcing. In particular, they vary considerably if the magnitude of aerosol forcing is altered within the (large) uncertainty range of this component.
36. These values differ slightly from those in (77) because we use a different optimization interval.
37. We do this by fitting the U-D model (8) results for anthropogenic-plus-solar forcing to the observations (best-fit sensitivity 3.2°C) and then disaggregating the hemispheric-mean modeled temperatures into their solar, effective sulfate aerosol (see below), and residual anthropogenic components. The "effective aerosol" response is the sum of responses to direct and indirect sulfate aerosol forcing and tropospheric ozone. HadCM2 considers only direct sulfate aerosol forcing. Because its magnitude and pattern are similar to the effective aerosol forcing used in the U-D model, we considered the two to be equivalent for the purposes of producing adjusted-observed data.
38. Solar-plus-aerosol forcing leads to cooling. Thus, removing this component gives residuals with a larger positive trend than in the raw data.
39. G. A. Meehl, G. J. Boer, C. Covey, M. Latif, R. J. Stouffer, Eos 78, 445 (1997).
40. Supported by USDOE (T.M.LW. and B.D.S.), NOAA (Award No. NA87GP0105 to T.M.L.W.) and NSF (DMS-9705166 to R.L.S.). Observed temperatures from P.D. Jones and D.E. Parker; control-run GCM data provided through the CMIP (27) project; SUL and GHG data provided by J.M. Gregory. NCAR is sponsored by the National Science Foundation.