ENERGY & ENVIRONMENT VOLUME 14 á NUMBER 6 á 2003 ISSN 0958-305X ©2003 MULTI-SCIENCE PUBLISHING CO. LTD. 5 Wates Way, Brentwood, Essex CM15 9TB, United Kingdom Energy & Environment is abstracted/indexed in: Engineering Index, Pollution Abstracts, Geographical Abstracts, Environmental Abstracts, Biomass Bulletin and International Development Abstracts. Energy & Environment is now available on-line, free to subscribers to the printed edition. For access details contact Multi-Science. Email: mscience@globalnet.co.uk Website: www.multi-science.co.uk CORRECTIONS TO THE MANN et. al. (1998) PROXY DATA BASE AND NORTHERN HEMISPHERIC AVERAGE TEMPERATURE SERIES Stephen McIntyre 512-120 Adelaide St. West, Toronto, Ontario Canada M5H 1T1; Ross McKitrick Department of Economics, University of Guelph, Guelph Ontario Canada N1G2W1. ABSTRACT The data set of proxies of past climate used in Mann, Bradley and Hughes (1998, “MBH98” hereafter) for the estimation of temperatures from 1400 to 1980 contains collation errors, unjustifiable truncation or extrapolation of source data, obsolete data, geographical location errors, incorrect calculation of principal components and other quality control defects. We detail these errors and defects. We then apply MBH98 methodology to the construction of a Northern Hemisphere average temperature index for the 1400-1980 period, using corrected and updated source data. The major finding is that the values in the early 15th century exceed any values in the 20th century. The particular “hockey stick” shape derived in the MBH98 proxy construction – a temperature index that decreases slightly between the early 15th century and early 20th century and then increases dramatically up to 1980 — is primarily an artefact of poor data handling, obsolete data and incorrect calculation of principal components. Correspondence etc. should be directed to smcintyre25@yahoo.ca ACKNOWLEDGMENTS: Helpful comments and encouragement were received from R. Carter, R. Courtney, D. Douglass, H. Erren, C. Essex, W. Kininmonth, T. Landscheit and referees. All remaining errors are ours. No funding from any source was sought or received for this research. 1. INTRODUCTION. In a widely cited paper (e.g. ref. 2 and others), Mann, Bradley and Hughes (1998, hereafter MBH98) constructed a temperature history of the Northern Hemisphere for the period 1400-1980. The result was the well-known “hockey stick”-shaped graph suggesting that the climate of the late 20th century is unusual compared to the centuries preceding it. This temperature history was extended to the period 1000-1399 752 Energy & Environment · Vol. 14, No. 6, 2003 in Mann, Bradley and Hughes (1999), who claimed that “temperatures in the latter half of the 20th century were unprecedented” and that “even the warmer intervals in the reconstruction pale in comparison with mid-to-late 20th-century temperatures”. The temperature history was given bold prominence by the Intergovernmental Panel on Climate Change (2001) where it appears in Figures 2-20 and 2-21 in Chapter 2 of the Working Group 1 Assessment Report, Figure 1b in the Working Group 1 Summary for Policymakers, Figure 5 in the Technical Summary, and Figures 2-3 and 9-1B in the Synthesis Report. Referring to this figure, the IPCC Summary for Policymakers (p. 3) claimed it is likely “that the 1990s has been the warmest decade and 1998 the warmest year of the millennium” for the Northern Hemisphere. The IPCC view of temperature history has in turn been widely disseminated by governments and used to support major policy decisions.1 MBH98 applied 112 proxies and historical temperature measurements in what they called a “multiproxy approach” to construction2 of a temperature index from 1400 to 1980. Although the “multiproxy” approach was apparently a novelty within the climatological community, the same algebraic and statistical methods are commonly used in economics, business and elsewhere in the social sciences, though the terminology differs from discipline to discipline.3 Upon request, Prof. Mann instructed an associate to supply the collated proxy set, together with applicable weights, to the first author. When attempting to replicate MBH98 principal component (PC) calculations, an extremely low (6%) explained variance for those in the Texas-Mexico dataset was noticed, leading to a close examination of the data collation. Anomalous start years (see details below) were noticed and it was verified that these occurred only in MBH98 data and were not due to collation errors on our part. Explained variance improved significantly by moving the MBH98 data one year later, confirming that an MBH98 collation error had almost certainly occurred. We then noticed copy errors in the 1980 values for these series and stretches of identical values in other places in the database. This led to a systematic comparison of MBH98 data to original data, identifying obsolete versions and undisclosed truncation of time series. Independent calculations of the proxy principal components convinced us that those in MBH98 were erroneous We updated and corrected the database and then applied MBH98 methodology, as publicly disclosed, to construct a temperature index from 1400 to 1980. The newly calculated temperature index (see Figure 7) contradicts the MBH98 assertion of late 20th century uniqueness. We find that the particular “hockey stick” shape derived by MBH98 is primarily an artefact of poor data handling and use of obsolete proxy records. 1See, for instance, the Government of Canada web site http://www.climatechange.gc.ca/english/issues/what_is/index.shtml. 2MBH98 refers to the index resulting from their calculation as a “reconstruction.” This is a misnomer since it is a novel index, rather than the recomputation of something previously observed. Therefore it will be referred to herein as “construction.” 3For a critique of applying stationary linear maps to nonstationary phenomena like climate see Essex and McKitrick (2002) chapter 5. Characteristics of Carbon Transactions 2. ERRORS AND DEFECTS IN THE MBH98 PROXY DATA BASE The term “proxy” denotes some physical data or measurement that can potentially serve as an indirect record of local temperature conditions, including tree ring widths and densities, coral dO18, dC13 and calcification rates, ice core dO18, melt percentages and so forth. Thirteen MBH98 series are based on instrumental temperature records and are not, strictly speaking, “proxies”. For consistency with MBH98, we will use the term “proxy” to include these series. We will denote a proxy series with the prefix ‘#’, i.e. proxies #1—#112. Twenty- two of the 112 proxies date back as far as 1400, while all 112 are available as of 1820. Twenty-three MBH98 proxies cease to be available in the 1970s. Thirty-one of the 112 proxies are principal components (PCs) from tree ring datasets, of which 28 were PCs calculated by MBH98 themselves from 300 tree ring datasets. Three are PCs from 14 Texas-Oklahoma sites, 9 are PCs from 20 Texas-Mexico sites, 9 are PCs from 232 International Tree Ring Data Base (ITRDB) US/Canada tree ring sites, 3 are PCs from 18 South American sites and 4 are from 16 Australian sites. Inconsistently, individual US, Canadian and Mexican tree ring sites are included separately in the list of 112 proxies rather than being incorporated into the PCs for that area (see e.g. Appendix, #49, #51-61, #106.) More information about the proxies is available at the Supplementary Information web site (see Appendix). The database used by MBH98 contains the errors and defects listed below. We detail each of these points in this section, then in Section 3 we show how correcting these errors and defects affects the calculation of the Northern Hemisphere average temperature index using MBH98 methodology. (a) unjustified truncation of 3 series; (b) copying 1980 values from one series onto other series, resulting in incorrect values in at least 13 series; (c) displacement of 18 series to one year earlier than apparently intended; (d) unjustified extrapolations or interpolations to cover missing entries in 19 series; (e) geographical mislocations and missing identifiers of location; (f) inconsistent use of seasonal temperature data where annual data are available; (g) obsolete data in at least 24 series, some of which may have been already obsolete at the time of the MBH98 calculations; (h) listing of unused proxies; (i) incorrect calculation of all 28 tree ring principal components. (a,f) Series #10 and #11 (Central England and Central Europe air temperatures respectively) use June-July-August averages. This raises three concerns: annual data were available in the primary sources; other station temperature series used by MBH98 (#21- #31), where identified, are annual; and MBH98 claims to calculate an annual temperature index. The Central England Temperature series is truncated at 1730 rather than the available 1659 in source data, which removes a major late 17th century cold period (see Supplementary Information). Series #10 has a 1987 value which is 0.43 deg C higher than in the source data though this does not appear to affect any calculations discussed herein. Central Europe (#11) is truncated at 1550 rather 754 Energy & Environment · Vol. 14, No. 6, 2003 than the available 1525, which removes the warmest temperatures in the series (compare Figure 1 Top and Bottom panels). #11, which is an exceptionally long series of direct temperature information, also shows a notable lack of 20th century uniqueness. In series #100, MBH98 also crop two very high values from the start of the series. These truncations are not justified and were not disclosed by MBH98. -3 -1 1 2-2.0 -0.5 1.0 -3 -1 1 2 JJA ANN MBH 1400 1500 1600 1700 1800 1900 2000 Figure 1. Temperature anomalies (C) by year for (Top panel) Central European historical air temperatures, MBH98 series #11; those differ from (Middle panel) Central European historical (Annual) from 1525-1979, which would be more relevant for inclusion in the calculation of an annual index; (Bottom panel) Central European historical (June-July-August) from 1525-1979, which matches, in the period of overlap, the incomplete record used by MBH98. (b,c) In the MBH98 collated data set the 1980 values for series #72-#80, which are the 9 Texas-Mexico principal components computed by MBH98, are identical to 7 decimal places, an obviously impossible result (see Table 1) and therefore an error. The 1980 values are likewise identical in the 3 Vaganov principal components (series #81-#83) and 4 of the 9 ITRDB US principal components computed by MBH98 (series #84, #90, #91 and #92); see Table 2. Interestingly, all but two of these series as collated in the MBH98 database begin in years ending in *99 or *49, rather than the apparently intended *00 and *50, and appear to have been displaced one year backward in collation. This suggests a simple clerical error, in which the series in question were copied into a file at the wrong row, then a 1980 value was filled in from an adjacent cell. Series #85-#89 commence in 1499, but lack the telltale 1980 value. The displacements will result in any extreme year in the past being one year off in these datasets, attenuating its effect in the compilation. The copy errors constitute a significant fraction of the MBH98 dataset for 1980 – the final year of the MBH98 proxy-constructed index.4 4To access the underlying data consult the supplementary information sources listed in the Appendix. Table 1. Identical 1980 values (bold) in Texas-Mexico series in MBH98 data set. Directory: TREE/STAHLE/SWM/BACKTO_1700MBH Series #: #72 #73 #74 #75 #76 #77 #78 #79 #80 Record Name: pc01.out pc02.out pc03.out pc04.out pc05.out pc06.out pc07.out pc08.out pc09.out 1976 -0.04758900 0.09825240 -0.01345320 0.01161880 0.01822490 0.03648180 0.04604640 -0.04273910 0.00526230 1977 0.02738590 -0.11581500 0.02995960 0.01370230 0.03782570 0.00327476 0.07170230 0.03729640 -0.10195200 1978 0.09249040 -0.00125138 0.08667150 0.07659540 0.02200060 0.04614070 0.03223540 0.02464170 0.02726110 1979 -0.01054950 -0.17253000 -0.00999568 -0.04078750 0.09144420 -0.00608904 -0.00508424 -0.03537360 -0.08408310 1980 0.02303040 0.02303040 0.02303040 0.02303040 0.02303040 0.02303040 0.02303040 0.02303040 0.02303040 Table 2. Identical 1980 values (bold) in Vaganov PCs and 4 ITRDB PCs in MBH98 data set. Directory: TREE/VAGANOV/BACKTO_1750 TREE/ITRDB/NOAMER/BACKTO_1750MBH Series #: #81 #82 #83 #84 #90 #91 #92 Record Name: pc01.out pc02.out pc03.out pc01.out pc07.out pc08.out pc09.out 1976 -0.03291460 0.02597240 0.01647480 0.00888690 0.10327000 0.18590800 0.03833640 1977 0.01170380 -0.09346030 -0.01559880 0.02153980 0.10721500 0.13304000 0.00842804 1978 0.05759400 -0.02107700 -0.13369700 0.02241670 0.11963800 0.12615900 0.01346540 1979 -0.11100000 -0.07345260 0.02438820 0.04898920 0.13681500 0.16866300 0.02801240 1980 -0.04063530 -0.04063530 -0.04063530 0.04345260 0.04345260 0.04345260 0.04345260 Characteristics of Carbon Transactions Table 3. Filled data series (bold) in MBH98 data set. NaN denotes a missing value, although values beyond 1980 are not relevant to the calibration interval. Directory: TREE/MANNETAL97MBH98 Series #: 45 46 51 52 54 56 58 Record Name: cpatagonia.dat npatagonia.dat treeline1.dat treeline10.dat treeline2.dat treeline4.dat treeline6.dat 1970 16.79999900 0.85000002 1.27000000 0.61400002 1.10500000 0.99900001 1.35900000 1971 15.50000000 -1.10000000 1.40900000 0.46000001 1.41199990 1.42200010 1.30300000 1972 15.90000000 0.17000000 1.25700000 0.83399999 1.38800000 1.22200000 1.38800000 1973 15.50000000 0.63999999 1.10700000 0.56199998 1.19700000 1.07100000 1.46000000 1974 16.70000100 -0.43000001 1.13300000 1.10400000 1.14400010 1.13500000 1.62899990 1975 14.90000000 -0.43000001 0.93199998 1.10400000 1.36600010 1.22400000 1.61300000 1976 14.70000000 -0.43000001 1.16100000 1.10400000 1.36600010 1.22400000 1.17600000 1977 16.10000000 -0.43000001 1.58500000 1.10400000 1.36600010 1.22400000 1.57300000 1978 15.10000000 -0.43000001 1.58500000 1.10400000 1.36600010 1.22400000 1.57300000 1979 15.10000000 -0.43000001 1.58500000 1.10400000 1.36600010 1.22400000 1.57300000 1980 15.10000000 -0.43000001 1.58500000 1.10400000 1.36600010 1.22400000 1.57300000 1981 15.10000000 NaN NaN NaN NaN NaN NaN 1982 15.10000000 NaN NaN NaN NaN NaN NaN Table 4. Filled data (bold) in MBH98 data. Directory: TREE/ITRDB/SOAMER/BACKTO_1600 TREE/ITRDB/AUSTRAL/BACKTO_1750MBH98 Series #: 93 94 95 96 97 98 99 Record Name: pc01.out pc02.out pc03.out pc01.out pc02.out pc03.out pc04.out 1970 -0.05519220 0.03191820 -0.00840994 0.07964660 -0.03334510 0.00628749 0.03972250 1971 0.04456720 0.02654390 0.06869810 0.05834090 0.03159730 0.01390980 -0.00292208 1972 -0.03087120 0.03992700 0.00302668 0.15582700 0.07014980 0.03358720 -0.07598700 1973 -0.02466770 0.11485700 -0.05301170 0.18438500 0.04514380 -0.04919020 -0.05758820 1974 0.03531060 0.07091270 0.00376018 0.11299600 0.01402680 -0.00682486 -0.10635300 1975 0.04918980 0.07842340 -0.02821910 0.16178501 0.02186560 0.02133480 0.00791537 1976 0.04792530 0.07830090 -0.02856930 0.16103400 0.08604140 0.05941720 -0.05302600 1977 0.04792530 0.07830090 -0.02856930 0.16103400 0.08604140 0.05941720 -0.05302600 1978 0.04792530 0.07830090 -0.02856930 0.16103400 0.08604140 0.05941720 -0.05302600 1979 0.04792530 0.07830090 -0.02856930 0.16103400 0.08604140 0.05941720 -0.05302600 1980 0.04792530 0.07830090 -0.02856930 0.16103400 0.08604140 0.05941720 -0.05302600 756 Energy & Environment · Vol. 14, No. 6, 2003 Characteristics of Carbon Transactions (d) MBH98 insert extrapolated, interpolated or copied values during the critical calibration period into 19 series. We refer to these as “fills” hereafter. In the data set provided to the authors, the following 17 series contain end-of-sample fills for one or more years including 1980: #6, #45, #46, #50-#52, #54-#56, #58, #93-#99. Series #53 was filled for 4 years at its beginning and series #3 for 16 years in the calibration period. In the case of #3, MBH98 inexplicably replaced available source values for 1962-64 with filled values. For examples see Tables 3, 4 and 5. Table 5. Filled series (bold) in data for MBH98 PCs. MBH98 Series #: 51 54 56 58 53 Record Name: ak031 ak032 cana157 cana153 cana036 1400 NA NA NA NA 0.723 1401 NA NA NA NA 0.723 1402 NA NA NA NA 0.723 1403 NA NA NA NA 0.723 1404 NA NA NA NA 0.723 1405 NA NA NA NA 0.874 1406 NA NA NA NA 1.026 1407 NA NA NA NA 1.029 1408 NA NA NA NA 1.203 1409 NA NA NA NA 1.055 1970 1.270 1.105 0.999 1.359 1.376 1971 1.409 1.412 1.422 1.303 1.554 1972 1.257 1.388 1.222 1.388 1.463 1973 1.107 1.197 1.071 1.460 1.618 1974 1.133 1.144 1.135 1.629 1.483 1975 0.932 1.366 1.224 1.613 1.743 1976 1.161 1.366 1.224 1.176 1.577 1977 1.585 1.366 1.224 1.573 1.583 1978 1.585 1.366 1.224 1.573 1.851 1979 1.585 1.366 1.224 1.573 1.618 1980 1.585 1.366 1.224 1.573 2.204 Series #50 is especially noteworthy. The values of series #50 for the entire period from 1962 to 1982 are copied from series #49 (see Table 6). Although MBH98 attribute both series #49 and #50 to Fritts and Shao (1992), series #49 is actually derived from Briffa et al. (1992). 758 Energy & Environment · Vol. 14, No. 6, 2003 Table 6. Final 20 values in MBH98 Series #49 and #50 are identical. Directory: TREE/MANNETAL97 MBH98 Series #: 49 50 Record Name: trd.dat trw.dat 1958 0.38000000 0.34000000 1959 -0.15000000 0.45000000 1960 0.280000000 0.020000000 1961 0.120000000 0.550000010 1962 -0.039999999 -0.039999999 1963 0.600000020 0.600000020 1964 -0.779999970 -0.779999970 1965 -0.800000010 -0.800000010 1966 0.289999990 0.289999990 1967 -0.230000000 -0.230000000 1968 -0.949999990 -0.949999990 1969 0.910000030 0.910000030 1970 0.319999990 0.319999990 1971 0.110000000 0.110000000 1972 -0.020000000 -0.020000000 1973 -0.010000000 -0.010000000 1974 -0.079999998 -0.079999998 1975 -0.680000010 -0.680000010 1976 -0.090000004 -0.090000004 1977 0.150000010 0.150000010 1978 -0.140000000 -0.140000000 1979 0.020000000 0.020000000 1980 -0.239999990 -0.239999990 1981 -0.010000000 -0.010000000 1982 0.059999999 0.059999999 These fills are neither required nor justified statistically and exceed MBH98 disclosure. There is no disclosure of the extent of data filling or its potential impact on the constructed temperature index in the text of the Nature article and, their supplementary web page (http://www.ngdc.noaa.gov/paleo/ei/data_supp.html) says only “Small gaps have been interpolated. If records terminate slightly before the end of the 1902-1980 training interval, they are extended by persistence to 1980.” Inconsistently, however, series #11, #102, #103, #104, #106 and #112 terminate prior to 1980 but were not filled in the MBH98 dataset. The fills in 1980 are pervasive: at least 30 (and up to 36) proxies in 1980 have values arising from copy errors or extrapolation. (e) Geographical mislocations and missing attributions occur in the MBH98 data. For example, MBH98 use 11 precipitation series, for which they cite Jones and Bradley (1992) (hereafter “JB92”). JB92 (Table 13.3) lists 17 precipitation series, of which 12 are digitally published at the World Data Center for Paleoclimatology Characteristics of Carbon Transactions (http://www.ngdc.noaa.gov/paleo/paleo.html, hereinafter denoted WDCP). In only two MBH98 precipitation series (#35 and #37) did the correlation between JB92 and MBH98 data exceed 0.9, permitting a reasonably secure identification of locations; other correlations were less than 0.5 excluding the possibility of identification. The JB92 series for Paris, France (48.8N, 2.5E) can be identified with MBH98 series #37 both from the high correlation and the identity of starting date (see Figure 2, which graphs both these series). However, MBH98 series #37 is located at the grid-box centred at 42.5N, 72.5W near Boston, Massachusetts. 20 MBH 0 -20 2 JB92 0 -2 1750 1800 1850 1900 1950 2000 Figure 2. (Top) Station precipitation with erroneous location given as 42.5N, 72.5W in MBH98 (series #37); (Bottom) the record corresponds instead to that of JB92 Paris, France (48.8N, 2.5E) JB92 is scaled to a 1901-1950 reference period (i.e. subtracting the 1901-50 mean and dividing by the 1901-50 standard deviation). MBH98 appear to have a scale error by a factor of 10. Two MBH98 precipitation series are in India and derive from an unreported source, since no Indian locations are listed in JB92. The other 7 MBH98 precipitation series derive either from unreported sources, from the 5 JB92 series not digitally published at WDCP or have been heavily transformed in collation. Two of the MBH98 temperature grid-box series had no locational counterparts in JB92 (Table 13.1): series #26 (52.5N, 17.5E grid-box) and series #29 (62.5N, 7.5E grid-box). In addition, MBH98 series #20 (Central Greenland ice core) is materially mislocated to the north and west. On comparison with source data, it can be seen that MBH98 have also reversed the geographical locations of series #46 and #47. (g) Digitally published versions at the World Data Center for Paleoclimatology (WDCP, http://www.ngdc.noaa.gov/paleo/paleo.html) supercede the versions used by MBH98 for the following 24 series: #1, #2, #3, #6, #7, #8, #9, #21, #23, #27, #28, #30, #35, #37, #43, #51, #52, #54, #55, #56, #58, #65, #105 and #112. A listing of FTP sources is provided in the Appendix and details for each of the above series, including comparisons of different data editions, is provided in the Supplementary Information. (Since many datasets used by MBH98 remain digitally unpublished, this listing is only from datasets where a comparand was identified.) For the purposes of this study, it is immaterial whether the MBH98 datasets were obsolete as at the time of publication of MBH98 or whether they have become obsolete subsequently. However, at least some 760 Energy & Environment · Vol. 14, No. 6, 2003 datasets used by MBH98 were already obsolete in 1998. In response to an inquiry about series #51- #61, WDCP confirmed that the updated versions for four of the series were available as early as 1991-1992. [WDCP, pers. comm., Sept. 2003]. In some cases, the differences between MBH98 and updated series were isolated; in other cases, the differences were systematic. As an example of relatively isolated differences, MBH98 series #28 corresponds closely to a Z-transformation (subtracting the mean, dividing by the standard deviation) of the JB92 Leningrad series for most of its history, but there are major and puzzling discrepancies in the 1760s, including a discrepancy of over 4 degrees C in 1764 (see Figure 3). As with the Central European temperature series (and other long temperature series), the 20th century values are not unique. -2 0 2 -3 -1 1 3 C1500 MBH 1400 1500 1600 1700 1800 1900 2000 Figure 3. (Top) Station temperature for grid box 75.5N, 32.5E from MBH98 (series #28); (Bottom) Leningrad temperature from JB92 at WDCP (anomaly from 1951-1970). As an example of systematic differences, MBH98 series #56 (Twisted Tree, Heartrot Hill, a northern treeline ring width series) used an early version of site data with values only up to 1975 and with MBH98 fills from 1976 to 1980. The updated version now at WDCP has data up to 1992 (see Figure 4) and differs quite dramatically from the MBH98 series. The MBH98 version of series #56, like MBH98 versions of many northern treeline series (#51-#58, #60-#61) shows an increased ring width index in the 1902-1980 period. However, in the WDCP series, there is a dramatic and sustained reduction in ring widths in the 1980s, with a complete reversal of the increases in the first decades of the century. This pattern occurs in other series updated into the 1990s (series #51 and #54) and was apparent by 1984 in the northerly series #59 (Hornby Cabin) (see Supplementary Information). The later edition of #56, presumably for quality control reasons, discontinued some early estimates made in the first edition. (h) Five series purportedly in the multiproxy network (fran003, ital015, ital015x, spai026 and spai047 in the MBH98 list “ITRDB –Miscellaneous”) cannot be located in either the MBH98 collated set or the proxy PC compilations. (i) Of the 112 proxies in MBH98, 28 are principal components calculated by MBH98 from International Tree Ring Data Base (ITRDB) site chronologies stored at WDCP for the sites listed in MBH98 Supplementary Information (see http://www.ngdc.noaa.gov/paleo/ei/data_supp.html) for the following five different Characteristics of Carbon Transactions regions: Texas-Oklahoma, Texas-Mexico, North America, South America and Australia (including New Zealand). The principal component calculations have two types of problems: first, MBH98 does not establish consistent rules for inclusion or exclusion of sites in regional aggregates and, second, the MBH98 principal components fail to maximize explained variances. 600 1200 0.5 1.5 2.5 NGDC MBH 1400 1500 1600 1700 1800 1900 2000 Figure 4. (Top) Twisted Tree, Heartrot Hill (northern treeline) ring width index from MBH98 (series #56) (1459-1975 plus 5 fills at end); (Bottom) Twisted Tree, Heartrot Hill ring width index from WDCP (1530-1992). MBH have divided WDCP values by 1000. Neither series is Z-transformed. As to the first problem, MBH98 do not provide justification for excluding the Texas-Oklahoma and Texas-Mexico sites from the North American compilation. Similarly puzzling are the occurrences of other sites as individual proxies rather than being incorporated into the regional PC groups. Series #106 occurs within the Texas- Mexico region; series #49-64 are all North American sites or reconstructions; series #46-47 are within the South American region and series #43 and #45 are reconstructions within the Australia-New Zealand region. The second problem was determined indirectly as the MBH98 principal component calculations are unpublished. We collated the source data from WDCP for all sites listed in MBH98 (except, immaterially, one MBH98 US site which could not be identified in the WDCP database). The collations are available in Supplementary Information. The start dates of the MBH98 PC’s are not consistent with those of available data. In 12 cases, MBH98 commenced their calculation after the date in which all records were available (e.g. Australia-New Zealand region where MBH98 commenced in 1750, although a start date of 1625 was possible.) In 16 cases, MBH98 commenced their PC index in a period prior to that available in the data (e.g. Texas-Mexico). Because standard PC algorithms fail in the presence of missing data, an important part of the methodology—namely how missing data were treated in the PC calculation—remains unexplained in MBH98. We computed all 28 PCs, together with their explained variances, using a standard principal component algorithm for the maximum period in which all records were available within each region. For comparison, weighting factors for the MBH98 PCs eigenvectors were computed which maximized the explained variance of the underlying ITRDB data, and the resulting explained variance was compared to our own computations using a standard algorithm. In all cases, explained variance for the 762 Energy & Environment · Vol. 14, No. 6, 2003 recomputed PCs exceeded that for the MBH98 PCs (see Table 7). Indeed it was the observation of the unusually poor fit between the MBH98 Texas-Mexico PCs and the underlying ITRDB data that led to the detailed audit undertaken in this paper. Table 7. 5 Regions in which MBH98 computed principal components. Each column shows the number of source sites listed by MBH98, the number found at WDCP; the number of PCs extracted; the dates spanned at WDCP and in the MBH98 PCs; the explained variance of each group. REGION: Texas-Texas-ITRDB North South Australia-. Oklahoma Mexico America America NZ # of Source Sites Listed 14 20 232 18 16 # at WDCP 14 20 231 18 16 of MBH PCs 3 9 9 3 4 WDCP Available Period 1698-1980 1760-1977 1619-1971 1568-1972 1625-1974 MBH PC Start 1-3: 1700 1: 1400 1-2: 1400 1-2: 1600 1-4: 1750 2: 1499 3-6: 1499 3: 1750 3-4: 1599 7: 1599 5-9: 1699 8-9: 1749 MBH PC End5 1980 1979 1980 1976 1976 Explained Variance: MBH 32% 6% 14% 26% 38% Recalculated 39% 76% 40% 35% 46% Figure 5 shows the MBH98 and re-calculated Australian PC1. The Australian PC1 is one of relatively few MBH98 series that shows anomalous 20th century behaviour and which closes on a dramatic “uptick”. The correct computation shows that this feature of this particular MBH98 series is entirely an artefact of incorrect calculation. Figure 5. (a) Australia PC1 in MBH98 (series #96) graphed over time (b) PC1 for the MBH98 Australia dataset calculated using standard algorithm. 5Excluding filled values Characteristics of Carbon Transactions 3. TEMPERATURE INDEX CONSTRUCTION USING CORRECTED DATA A corrected and updated proxy database has been developed, in which the measures outlined above were adopted, including the following: • the most recent editions of the MBH98 series have been used where identified and available; • arbitrary MBH98 truncations and fills have been deleted; • correct tree ring principal component calculations have been used. We replicated the methodology of MBH98 as closely as we could using publicly available documentation and such private assistance as we were able to obtain. MBH98 purports to establish relationships between the proxies and 16 temperature principal components calculated from the Climate Research Unit (CRU) instrumental temperature database, using a subset of 1,082 out of 2,592 cells and the 79-year period from 1902-1980 as a calibration period. These 16 temperature principal components are referred to as TPC1—TPC16. Prior to this calculation, the CRU data was scaled cellwise. We downloaded original temperature data from CRU and gridpoint locations from the MBH98 website and calculated scaling factors for downstream use in calculation of northern hemisphere temperature averages. Four MBH98 cells contained no observations in the CRU data and were excluded from all calculations. Following the description of MBH98 procedures in their Supplementary Information, our construction is done piecewise for each of the periods listed in Table 8, using the roster of proxies available throughout the period and the selection of TPCs for each period listed in Table 8. There are slight discrepancies between 1500 and 1750 in the number of proxies which MBH98 reported to be available and the number actually available in the MBH98 data set (see columns 2-3). The anomalous listing of TPCs 6 and 8 in the period 1750 to 1759 is assumed to be an erroneous rendering of TPCs 7 and 9, but there is little sensitivity to this assumption. Following MBH98, the number of TPCs used in the construction decreases from 11 in the latest period to 1 in the earliest period, as shown in Table 8. Following MBH98 procedures as publicly disclosed, for each combination of proxy roster and TPC selection, the proxies were first calibrated against the temperature PCs in the calibration period of 1902-1980 and then the temperature PCs were constructed in each period using the proxy and TPC rosters prescribed by MBH98 for the period, together with weighting factors supplied to the authors by an associate of Prof. Mann. From these constructed PCs, using MBH98 eigenvalues and eigenvectors, gridded temperature series for 1,082 cells were obtained. From the cells in the northern hemisphere (excluding the four cells with no observations and hence no scaling factor), a northern hemisphere average temperature index was calculated. We have posted scripts for this construction in Supplementary Information. It should be noted that each of the above steps in the MBH98 northern hemisphere temperature index construction is a linear operation on the proxies. Accordingly, given the roster of proxies and TPCs in each period, the result of these linear operations is a set of proxy weighting factors, which generates the NH average temperature construction. These weighting factors are not disclosed in MBH98. 764 Energy & Environment · Vol. 14, No. 6, 2003 Table 8. Intervals defining proxy groups and subset of temperature PCs used in coefficient fitting process. Interval No. of proxies No. found Number of Temp Temperature reported available in data set PCs fit to proxies PC Identifiers 1400-1450 22 22 1 1 1450-1500 24 24 2 1,2 1500-1600 28 34 2 1,2 1600-1700 57 54 4 1,2,11,15 1700-1730 74 73 5 1,2,5,11,15 1730-1750 79 78 5 1,2,5,11,15 1750-1760 89 89 8 1-3,5,6,8,11,15 1760-1780 93 93 9 1-5,7,9,11,15 1780-1800 97 97 11 1-5,7,9,11,14-16 1800-1820 102 102 11 1-5,7,9,11,14-16 1820-1971 112 112 11 1-5,7,9,11,14-16 1972+ 112 106-111 11 1-5,7,9,11,14-16 Figure 6. (a) MBH98 NH temperature series (C), 1400-1980, which relies heavily on (b) TPC1 from MBH98. (c) Authors’ replication of TPC1 using MBH98 methods and data. (d) Authors’ TPC1 using MBH98 methods but with data corrected as outlined in text. Characteristics of Carbon Transactions The well-known “hockey stick” shape of the MBH98 northern hemisphere temperature index is shown in Figure 6a. It depends strongly on the temperature PC1 (Figure 6b) so we will illustrate its replication, although all TPCs were calculated and used in the NH construction. Our replication of TPC1 using the MBH98 method and data is shown in Figure 6c. Our version of TPC1 in Figure 6c is clearly similar to the calculation of MBH98 in Figure 6b (correlation 0.95 in the 20th century), indicating substantial success in replicating the MBH98 methodology, but some differences remain, possibly due to undisclosed variations in their procedures and assumptions. The TPC1 construction using corrected data is in Figure 6d, showing higher 15th century values than 20th century values, unlike the MBH98 TPC1. Figures 7 and 8 show the impact of the corrections on northern hemisphere temperature construction. In Figure 7 the top line is the MBH98 construction (reproducing Figure 6a), while the bottom line shows the Northern Hemisphere multiproxy temperature index resulting from the application of MBH98 procedures on an updated and correctly collated assembly of the MBH98 library of proxy data. On the basis of corrected and updated data, 15th century values are higher than those in the 20th century, contradicting the MBH98 conclusion of a unique late 20th century climate. Figure 8 shows 20-year smoothed series for comparison. 0.8 00.2 0.4 0.6 1400 1450 1500 1550 1600 1650 1700 1750 1800 1850 1900 1950 Corrected -0.2 -0.4 -0.6 -0.8 Figure 7. Temperature anomalies index (deg C) 1400-1980 for Northern Hemisphere average temperature construction from (top) Mann et. al. (1998); and (bottom) based on this work using corrected and updated data as outlined in text. -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 MBH98 Anomalies (C) Anomalies (C) 766 Energy & Environment · Vol. 14, No. 6, 2003 -0.8 -0.4 0.0 0.4 0.8 Temperature Index (deg C) Mann et al. 1998 Corrected version 1400 1500 1600 1700 1800 1900 2000 Year Figure 8. As Figure 7, using 20-year running mean to smooth. 4. CONCLUSIONS The MBH98 hockey stick-shaped NH temperature reconstruction discussed here has been extremely influential in discussions of 20th century global warming. Together with a pre-1400 extension derived in Mann et. al. (1999) and a spliced instrumental temperature series, this index figured prominently in the IPCC Third Assessment Report (IPCC 2001) and numerous other publications. However, the dataset used to make this construction contained collation errors, unjustified truncation or extrapolation of source data, obsolete data, incorrect principal component calculations, geographical mislocations and other serious defects. These errors and defects substantially affect the temperature index. Although not all of the dataset could be audited, it was possible to prepare a data base with substantially improved quality control, by using the most recent data and collating it correctly, by avoiding arbitrary filling in or truncation of data and by computing principal components using standard algorithms. Without endorsing the MBH98 methodology or choice of source data, we were able to apply the MBH98 methodology to a database with improved quality control and found that their own method, carefully applied to their own intended source data, yielded a Northern Hemisphere temperature index in which the late 20th century is unexceptional compared to the preceding centuries, displaying neither unusually high mean values nor variability. More generally, the extent of errors and defects in the MBH98 data means that the indexes computed from it are unreliable and cannot be used for comparisons between the current climate and that of past centuries, including claims like “temperatures in the latter half of the 20th century were unprecedented,” and “even the warmer intervals in the reconstruction pale in comparison with mid-to-late 20th-century temperatures” (see press release accompanying Mann et al 1999) or that Characteristics of Carbon Transactions the 1990s was “likely the warmest decade” and 1998 the “warmest year” of the millennium (IPCC 2001). REFERENCES Briffa, K.R., P.D. Jones, and F.H. Schweingruber, 1992, Tree-Ring Density Reconstructions of Summer Temperature Patterns across Western North America since 1600, Journal of Climate, Vol. 5, No. 7. Essex, C. and R. McKitrick (2002). Taken By Storm: The Troubled Science, Policy and Politics of Global Warming. Toronto: Key Porter. Fritts, H.C. & Shao, X.-M. (1992), Mapping climate using tree-rings from western North America, Dendroclimatic evidence from the northern Soviet Union, in Climate since A.D. 1500, (eds Bradley, R.S. & Jones, P.D., 269-294, Routledge, 1992). Intergovernmental Panel on Climate Change (2001) Climate Change 2001: The Scientific Basis. Cambridge: CUP. Jones, P.D. & Bradley, R.S. (1992) Climatic variations in the longest instrumental records, in Climate Since A.D. 1500, (eds Bradley, R.S. & Jones, P.D., 246-268, Routledge, 1992). Mann, M.E., Bradley, R.S. & Hughes, M.K. (1998) Global-Scale Temperature Patterns and Climate Forcing Over the Past Six Centuries, Nature, No. 392, pp. 779-787, 1998. Mann, M.E., Bradley, R.S. and Hughes, M.K., (1999). Northern Hemisphere Temperatures During the Past Millennium: Inferences, Uncertainties, and Limitations, Geophysical Research Letters, No. 26, pp. 759-762. See also press release at http://www.umass.edu/newsoffice/archive/1999/030399warming.html. APPENDIX: SUPPLEMENTARY INFORMATION SOURCES Supplementary information for this paper, including detailed information about all 112 proxy series, the computations and data used for the Figures, are available at http://www.climate2003.com/index.html and http://www.uoguelph.ca/~rmckitri/research/trc.html. The supporting web site for the MBH98 paper is http://www.ngdc.noaa.gov/paleo/ei/data_supp.html. FTP References for Updated MBH98 Series. Column 1 is MBH98 series number. Column 2 is MBH98 series descriptor. Column 3 shows whether a digital update is referred to in the text. Column 4 shows the digital publication reference (see Supplementary Information). Column 5 is applicable line in multi-set FTP reference. NA- No digital publication located. NV- Digital publication located, but not compared. Series # MBH98 Description Ref Digital Publication Line 1 Burdekin River Coral Fluorescence Yes ftp://ftp.ngdc.noaa.gov/paleo/coral/west_pacific/great_barrier/burdekin_2001.txt 47 2 Great Barrier Reef Coral Thickness Index Yes ftp://ftp.ngdc.noaa.gov/paleo/coral/west_pacific/great_barrier/aims10coreavg.txt3 Galapagos Urvina Bay UR-86 dO18 Yes ftp://ftp.ngdc.noaa.gov/paleo/coral/east_pacific/urvcomp.txt4 Red Sea, Aqaba Core 18 dO18 No ftp://ftp.ngdc.noaa.gov/paleo/coral/red_sea/aq18-18o.txt5 Red Sea, Aqaba, Core 18 dC13 No ftp://ftp.ngdc.noaa.gov/paleo/coral/red_sea/aq18-13c.txt6 Espiritu Santu, Vanuatu dO18 Yes ftp://ftp.ngdc.noaa.gov/paleo/coral/west_pacific/vanuatu_annual.txt7 New Caledonia dO18 Yes ftp://ftp.ngdc.noaa.gov/paleo/coral/west_pacific/nc_published_1992-1657_qtr.txt8 Gulf of Chiriqui, Panama dO18 No ftp://ftp.ngdc.noaa.gov/paleo/coral/east_pacific/secas-10yr-iso.txt9 Gulf of Chiriqui, Panama dC13 No ftp://ftp.ngdc.noaa.gov/paleo/coral/east_pacific/secas-10yr-iso.txt10 Central England Historical Yes http://www.metoffice.com/research/hadleycentre/CR_data/Daily/HadCET_act.txt11 Central Europe Historical Yes ftp://ftp.ngdc.noaa.gov/paleo/climate1500ad/ch6.txt12 Quelccaya Summit dO18 No ftp://ftp.ngdc.noaa.gov/paleo/icecore/trop/quelccaya/q83cor1.txt13 Quelccaya Summit Accum. (m) No ftp://ftp.ngdc.noaa.gov/paleo/icecore/trop/quelccaya/q83cor1.txt14 Quelccaya Ice Core 2 dO18 No ftp://ftp.ngdc.noaa.gov/paleo/icecore/trop/quelccaya/q83summ.txt15 Quelccaya Ice Core 2 Accum (m) No ftp://ftp.ngdc.noaa.gov/paleo/icecore/trop/quelccaya/q83summ.txt16 Dunde Ice Core dO18 NA NA 17 West Greenland Ice Melt (pct) NA NA 18 Svalbard Ice Melt, 5-yr avg (“pct) NA see ftp://ftp.ngdc.noaa.gov/paleo/climate1500ad/ch26.txt19 Penny, Baffin Island dO18 NA NA 20 Central Greenland (Stack) dO18 NA NA 21 Station temperature 42.5N, 92.5W Yes ftp://ftp.ngdc.noaa.gov/paleo/climate1500ad/ch13.txt 2370 22 Station temperature 47.5N, 2.5E NA NA 23 Station temperature 47.5N, 7.5E Yes ftp://ftp.ngdc.noaa.gov/paleo/climate1500ad/ch13.txt 780 24 Station temperature 47.5N, 12.5 E NA NA 25 Station temperature 47.5N, 17.5E NA NA 26 Station temperature 52.5N, 17.5E NA NA 27 Station temperature 57.5N, 17.5E Yes ftp://ftp.ngdc.noaa.gov/paleo/climate1500ad/ch13.txt 272 28 Station temperature 75.5N, 32.5E Yes ftp://ftp.ngdc.noaa.gov/paleo/climate1500ad/ch13.txt 1289 768 Energy & Environment · Vol. 14, No. 6, 2003 Characteristics of Carbon Transactions 29 Station temperature 62.5N, 7.5E NA NA 30 Station temperature 62.5N, 12.5E Yes ftp://ftp.ngdc.noaa.gov/paleo/climate1500ad/ch13.txt 52 31 Station temperature 62.5N, 42.5E NA NA 32 Station Precipitation 12.5N, 62.5E NA NA 33 Station Precipitation 17.5N, 72.5E NA NA 34 Station Precipitation 37.5N, 77.5W NA NA 35 Station Precipitation 42.5N, 2.5E Yes ftp://ftp.ngdc.noaa.gov/paleo/climate1500ad/ch13.txt 3650 36 Station Precipitation 42.5N, 7.5E NA NA 37 Station Precipitation 42.5N, 72.5W Yes ftp://ftp.ngdc.noaa.gov/paleo/climate1500ad/ch13.txt 384838 Station Precipitation 47.5N, 2.5E NA NA 39 Station Precipitation 47.5N, 12.5E NA NA 40 Station Precipitation 52.5N, 12.5E NA NA 41 Station Precipitation 52.5N, 2.5W NA NA 42 Station Precipitation 57.5N, 7.5W NA NA 43 Tasmania T-reconstruction Yes ftp://ftp.ngdc.noaa.gov/paleo/treering/reconstructions/tasmania/tasmania_recon.txt44 Java NA NA 45 New Zealand T-reconstruction Yes ftp://ftp.ngdc.noaa.gov/paleo/climate1500ad/ch24.txt 65 46 cpatagonia 41S! Yes ftp://ftp.ngdc.noaa.gov/paleo/climate1500ad/ch23.txt 187 47 npatagonia 38S! No ftp://ftp.ngdc.noaa.gov/paleo/climate1500ad/ch23.txt 236 48 Upper Kolyma River, Russia T NA 49 Western North America T (MXD) No ftp://ftp.ngdc.noaa.gov/paleo/treering/reconstructions/westnamerica/briffa1992/briffa1992.txt50 Western North America T (RW) NV ftp://ftp.ngdc.noaa.gov/paleo/treering/reconstructions/westnamerica/readme_westnamerica_recons.txt51 Treeline, 412 Alaska Yes ftp://ftp.ngdc.noaa.gov/paleo/treering/chronologies/northamerica/usa/ak031.crn52 Treeline, Fort Chimo PQ Yes ftp://ftp.ngdc.noaa.gov/paleo/treering/measurements/northamerica/canada/cana002.rwl53 Treeline, Gaspe PQ Yes ftp://ftp.ngdc.noaa.gov/paleo/treering/measurements/northamerica/canada/cana036.rwl54 Treeline, Arrigetech AK Yes ftp://ftp.ngdc.noaa.gov/paleo/treering/measurements/northamerica/usa/ak032.crn55 Treeline, Sheenjek R, Alaska Yes ftp://ftp.ngdc.noaa.gov/paleo/treering/chronologies/northamerica/usa/ak033.crn56 Treeline, TTHH Canada Yes ftp://ftp.ngdc.noaa.gov/paleo/treering/chronologies/northamerica/canada/cana157.crn57 Treeline, Mackenzie Mts, Canada No ftp://ftp.ngdc.noaa.gov/paleo/treering/chronologies/northamerica/canada/cana154.crn 770 Energy & Environment · Vol. 14, No. 6, 2003 58 Treeline, Coppermine R, Canada Yes ftp://ftp.ngdc.noaa.gov/paleo/treering/chronologies/northamerica/canada/cana153.crn59 Treeline, Hornby Cabin, Canada No ftp://ftp.ngdc.noaa.gov/paleo/treering/chronologies/northamerica/canada/cana155.crn60 Treeline, Churchill, Canada No ftp://ftp.ngdc.noaa.gov/paleo/treering/chronologies/northamerica/canada/cana158.crn61 Treeline, Castle Penin, Canada No ftp://ftp.ngdc.noaa.gov/paleo/treering/chronologies/northamerica/canada/cana159.crn62 Precip-Recon- SE USA-NC NA NA 63 Precip-Recon- SE USA –SC NA NA 64 Precip-Recon – SE USA -GA NA NA 65 Mongolia, Tarvagatny Pass Yes ftp://ftp.ngdc.noaa.gov/paleo/treering/chronologies/asia/mong003.crn66 Yakutia, Russia T-reconstruction NA NA – but compare to ftp://ftp.ngdc.noaa.gov/paleo/treering/chronologies/asia/russ142w_crns.crn67 Fennoscandia T-reconstruction NA NA – but compare to ftp://ftp.ngdc.noaa.gov/paleo/treering/chronologies/europe/swed019x_crns.crn68 Northern Urals T-reconstruction NA NA 69 USA, OK PC1 NA NA 70 USA, OK PC2 NA NA 71 USA, OK PC3 NA NA 72 Mexico PC1 NA NA 73 Mexico PC2 NA NA 74 Mexico PC3 NA NA 75 Mexico PC4 NA NA 76 Mexico PC5 NA NA 77 Mexico PC6 NA NA 78 Mexico PC7 NA NA 79 Mexico PC8 NA NA 80 Mexico PC9 NA NA 81 Vaganov12 Chronologies PC1 NA NA 82 Vaganov 40 Chronologies – PC1 NA NA 83 Vaganov 58 Chronologies PC1 NA NA 84 USA PC1 NA NA 85 USA PC2 NA NA 86 USA PC3 NA NA Characteristics of Carbon Transactions 87 USA PC4 NA NA 88 USA PC5 NA NA 89 USA PC6 NA NA 90 USA PC7 NA NA 91 USA PC8 NA NA 92 USA PC9 NA NA 93 South America PC1 NA NA 94 South America PC2 NA NA 95 South America PC3 NA NA 96 Australia PC1 NA NA 97 Australia PC2 NA NA 98 Australia PC3 NA NA 99 Australia PC4 NA NA 100 CHIN04 Yes ftp://ftp.ngdc.noaa.gov/paleo/treering/chronologies/asia/chin004.crn101 CHIN04X No ftp://ftp.ngdc.noaa.gov/paleo/treering/chronologies/asia/chin004x.crn102 FRAN009 No ftp://ftp.ngdc.noaa.gov/paleo/treering/chronologies/europe/fran009.crn103 FRAN010 No ftp://ftp.ngdc.noaa.gov/paleo/treering/chronologies/europe/fran010.crn104 FRAN011 No ftp://ftp.ngdc.noaa.gov/paleo/treering/chronologies/europe/fran011.crn105 INDI008X Yes ftp://ftp.ngdc.noaa.gov/paleo/treering/chronologies/asia/indi002x.crn106 MEXI001 No ftp://ftp.ngdc.noaa.gov/paleo/treering/chronologies/northamerica/mexico/mexi001.crn107 MORO003 No ftp://ftp.ngdc.noaa.gov/paleo/treering/chronologies/africa/morc011.crn108 MORO007 No ftp://ftp.ngdc.noaa.gov/paleo/treering/chronologies/africa/morc001.crn109 MORO008 No ftp://ftp.ngdc.noaa.gov/paleo/treering/chronologies/africa/morc014.crn110 SPAI011 No ftp://ftp.ngdc.noaa.gov/paleo/treering/chronologies/europe/spai011.crn111 SPAI012 No ftp://ftp.ngdc.noaa.gov/paleo/treering/chronologies/europe/spai012.crn112 SWED002B Yes ftp://ftp.ngdc.noaa.gov/paleo/treering/chronologies/europe/swed002.crn