Attriution of polr wrming to humn influence HAN P. GILLETT 1 *, DÁITHÍ A. STONE,3, PETER A. STOTT 4, TORU NOZAWA 5, ALEXEY YU. KARPECHKO 1, GABRIELE C. HEGERL 6, MICHAEL F. WEHNER 7 AND PHILIP D. JONES 1 1 Climtic Reserch Unit, School of Environmentl Sciences, University of Est Angli, Norwich NR4 7TJ, UK Atmospheric, Ocenic nd Plnetry Physics, Deprtment of Physics, University of Oxford, Clrendon Lortory, Prks Rod, Oxford OX1 3PU, UK 3 Tyndll Centre for Climte Chnge Reserch, Environmentl Chnge Institute, University of Oxford, South Prks Rod, Oxford OX1 3QY, UK 4 Met Office Hdley Centre, Fitzroy Rod, Exeter EX1 3PB, UK 5 Ntionl Institute for Environmentl Studies, Tsuku, Irki 35-856, Jpn 6 School of Geosciences, The University of Edinurgh, Grnt Institute, The King s Building, West Mins Rod, Edinurgh EH9 3JW, UK 7 Lwrence Berkeley Ntionl Lortory, 1 Cyclotron Rod, Berkeley, Cliforni 947, USA *e-mil: n.gillett@ue.c.uk Pulished online: 3 Octoer 8; doi:1.138/ngeo338 The polr regions hve long een expected to wrm strongly s result of nthropogenic climte chnge, ecuse of the positive feedcks ssocited with melting ice nd snow 1,. Severl studies hve noted rise in Arctic tempertures over recent decdes 4, ut hve not formlly ttriuted the chnges to humn influence, owing to sprse oservtions nd lrge nturl vriility 5,6. Both wrming nd cooling trends hve een oserved in Antrctic 7, which the Intergovernmentl Pnel on Climte Chnge Fourth Assessment Report concludes is the only continent where nthropogenic temperture chnges hve not een detected so fr, possily s result of insufficient oservtionl coverge 8. Here we use n up-to-dte gridded dt set of lnd surfce tempertures 9,1 nd simultions from four coupled climte models to ssess the cuses of the oserved polr temperture chnges. We find tht the oserved chnges in Arctic nd Antrctic tempertures re not consistent with internl climte vriility or nturl climte drivers lone, nd re directly ttriutle to humn influence. Our results demonstrte tht humn ctivities hve lredy cused significnt wrming in oth polr regions, with likely impcts on polr iology, indigenous communities, ice-sheet mss lnce nd glol se level 11. We use the CRUTEM3 (refs 9,1) ner-surfce temperture dt set, which consists of lnd sttion temperture oservtions gridded on 5 5 grid, with no infilling of missing dt. We use dt from Jnury 19 July 8 nd tke 5-yr mens t loctions where t lest 5% of the monthly dt re present. Supplementry Informtion, Fig. S1 shows the numer of grid cells with dt for the Arctic (65 N 9 N) nd the Antrctic (65 S 9 S) in ech 5-yr period. Most 5-yr men grid cell tempertures in the polr regions re sed on dt from single sttion, nd in these cses gridded dt re single-sttion temperture nomlies. In the Antrctic there re no sttion dt efore 1945 nd we therefore restrict our nlysis to the period 195 1999. Expedition records, minly locted in the Antrctic Peninsul nd Ross Se sectors, suggest tht Antrctic ws 1 C colder in the first hlf of the twentieth century compred with the 1957 1975 period 1. In the Arctic, there re t lest fifteen cells with dt throughout the twentieth century; therefore, we use the whole century for our nlysis of Arctic temperture. Figure 1 shows men oserved Arctic nd Antrctic temperture in lck, sed on the 19 8 nd 195 8 periods respectively 13. As dt coverge is sprse, we give equl weight to ech cell contining dt when clculting rel mens, rther thn weighting y the res of the grid cells. Both polr regions hve exhiited wrming over the periods considered, with lrge vriility superposed. We compre oserved tempertures with simultions from four CMIP3-coupled climte models (UKMO-HdCM3, PCM, CCSM3 nd MIROC3.(medres)). We choose ll CMIP3 models for which n ensemle with nturl forcings lone is ville, nd which include strtospheric ozone depletion in their comined nthropogenic nd nturl forcings simultions. Previous work hs shown tht, s group, these four models perform etter thn verge t simulting Antrctic climte 14, climte nd vriility in the northern extrtropics 15 nd vriility in Arctic temperture 4. Model monthly men lnd temperture nomlies were regridded onto the oservtionl 5 5 grid nd were msked with the oservtionl coverge efore clculting 5-yr men sptil mens. The ensemle men temperture nomly in 7 simultions including nthropogenic nd nturl forcings is shown in red in Fig. 1, nd the ensemle men temperture nomly in 3 simultions including nturl forcing lone is shown in lue. In the Arctic, the simultions including oth nthropogenic nd nturl forcings reproduce the overll oserved wrming trend well. However, lthough these simultions show some predominntly nthropogenic erly century wrming 8, the wrm nomly in the 193s nd 194s is not reproduced, suggesting tht this ws lrgely unforced vriility, consistent with other nlysis 4. The numer of oserved 5-yr men nomlies tht lie outside the.5 97.5th percentile rnge ssocited with the ALL ensemle is not significnt t the 5% level, indicting consistency etween simulted nd oserved vriility. In the Antrctic, the simultions with nturl nd nthropogenic forcing show wrming trend. A wrming trend is lso oserved 16, significnt t the 5% level llowing for 75 nture geoscience VOL 1 NOVEMBER 8 www.nture.com/nturegeoscience 8 Mcmilln Pulishers Limited. All rights reserved.
Temperture nomly Temperture nomly 1 1 19 19 194 1 1 19 19 194 ALL Os 196 198 Yer ALL Os 196 198 Yer Figure 1 Simulted nd oserved Arctic nd Antrctic men temperture nomlies.,, Five-yr men Arctic () nd Antrctic () men lnd temperture nomlies. Solid lines show oserved temperture nomlies (lck), the men simulted response to nturl forcings (solr irrdince chnges nd volcnic erosol, denoted ; lue) nd the men simulted response to nturl nd nthropogenic forcings (greenhouse gs chnges, strtospheric ozone depletion, sulphte erosol, solr irrdince chnges nd volcnic erosol, denoted ALL; red). Dshed lines show the wrmest nd coldest (lue) nd ALL (red) simultion in ech 5-yr period, pproximtely representing.5th nd 97.5th percentiles. Anomlies for 5 9 re sed on oservtions up to July 8. utocorreltion. This wrming trend oserved t sttion loctions, mny situted on the cost nd on the Antrctic Peninsul, my e lrger thn the wrming in n re men over the whole of Antrctic 16, lthough this is difficult to ssess. Although oserved Antrctic men temperture is outside the.5 97.5th percentile rnge ssocited with the ALL ensemle more thn would e expected y chnce (the difference is significnt t the 5% level), we find tht simulted nd oserved Antrctic men temperture trends re consistent, in contrst to study using different suset of CMIP3 models nd n infilled oservtionl dt set 17. The lrge cold nomly in 195 1954 is sed on dt from Frdy (now nmed Verndsky) sttion in the grid cell t 65 S 7 S, 6 E 65 E: we demonstrte tht our detection nd ttriution results re roust to the exclusion of dt from this period. In neither polr region is the wrming trend reproduced in response to nturl forcings lone. Simulted tempertures re plotted only up to 1999, s mny simultions end then; oserved tempertures re plotted up to the present, nd show continued wrming over oth polr regions since. In the Arctic, generlly positive trends re pprent t individul grid cells in the oservtions nd in the simultions with nthropogenic nd nturl forcings (Fig. ), ut, fter sutrcting the men trend, the pttern of trends is not similr in the models (Fig. c) nd oservtions (r =.5). This difference in simulted nd oserved trend ptterns in the Arctic is lrger thn would e expected sed on the difference etween individul model simultions in the multi-model ensemle, t the 5% level. However, we should not expect climte models to relisticlly simulte the grid-ox scle response to climte forcings 18, nd prtly for this reson most detection nd ttriution studies use mens over lrger res 8, such s the 9 sectors we use in this study. In the Antrctic, wheres the ensemle men of the model simultions shows wrming everywhere (Fig. d), the oservtions show strong wrming on the Antrctic Peninsul 14,19, cooling t the South Pole nd mixture of positive nd negtive trends elsewhere, lthough positive trends predominte (Fig. ). A similr, leit smoother, pttern of trends is seen in sptilly complete reconstruction of Antrctic temperture sed on sttion dt nd se surfce temperture oservtions (see Supplementry Informtion, Fig. S). Another sptilly complete reconstruction 16 shows more cooling over est Antrctic over the period 196 1999, proly in prt owing to the lter strt dte (see Supplementry Informtion, Fig. S). The simulted nd oserved trend ptterns re correlted (r =.4), nd consistent to within intr-ensemle vriility. In neither hemisphere re comprle wrming trends simulted in response to nturl forcings lone (Fig. e,f). In oth hemispheres, wrming trends re lrger nd more widespred when clculted up to 8 (Fig. g,h). Prts of the oserved nd simulted trend ptterns re ssocited with n upwrd trend in the Southern Annulr Mode (SAM) index (Fig. 3), which hs proly een forced y comintion of greenhouse gs increses nd strtospheric ozone depletion 8. The pttern of temperture trends ssocited with the trend in the SAM consists of wrming of the Antrctic Peninsul nd cooling elsewhere 19,1, ut the effect on the oserved trends seems weker here thn in previous studies 1, prtly ecuse we re considering nnul men chnges rther thn just summer nd utumn, nd prtly ecuse there ws mximum in the SAM in the erly 196s (ref. ), mking the SAM trend over the period considered here smller thn tht over the 1969 period 1. After the SAM-congruent component of the oserved temperture trends hs een removed, residul trends show wrming everywhere except for t the South Pole (Fig. 3c), mking them more similr to simulted residul trends (Fig. 3d) thn the corresponding rw trends. The CMIP3 models simulte SAM-congruent temperture trend pttern tht is weker thn tht oserved, owing to smller men SAM trend in the models (.5 hp yr 1 ) thn tht oserved (.6 hp yr 1 ) nd weker temperture response to the SAM: the r.m.s. mplitude of the men simulted temperture regression on the SAM is 79% of tht oserved. To ojectively test for the presence of n nthropogenic or nturl response in oservtions of polr temperture, we use detection nd ttriution nlysis to compre simulted nd oserved chnges 3 5. Such methods, first developed to detect nthropogenic influence on glol temperture, hve more recently een used to detect nthropogenic influence on temperture on continentl scles 8. To reduce the numer of sptil degrees of freedom, nd focus on the more relisticlly simulted lrge-scle sptil ptterns 18, we tke four 9 sector mens of simulted nd oserved 5-yr men tempertures over ech polr region ( E 9 E, 9 E 18 E, 18 W 9 W nd 9 W E). We nture geoscience VOL 1 NOVEMBER 8 www.nture.com/nturegeoscience 751 8 Mcmilln Pulishers Limited. All rights reserved.
.9.6.3.3.6.9.9.6.3.3.6.9 c.9.6.3.3.6.9 d.9.6.3.3.6.9 e.9.6.3.3.6.9 f.9.6.3.3.6.9 g.9.6.3.3.6.9 h.9.6.3.3.6.9 Figure Simulted nd oserved grid cell temperture trends. Trends were clculted for grid cells with t lest 7% of 5-yr mens present in the Arctic (19 1999) nd Antrctic (195 1999). f, Oserved trends (,), simulted ALL trends (c,d) nd simulted trends (e,f). In, loclly significnt wrming t the 5% level, llowing for utocorreltion, ws found in the cells contining the sttions Verndsky, Rother, Novolrevskj nd Byrd, with significnt cooling t the South Pole. g,h, Oserved trends clculted in the sme wy up to July 8. regress oserved 5-yr men sector men tempertures onto the simulted response to comined nthropogenic nd nturl forcings, nd the simulted response to nturl forcings lone in totl lest-squres optiml regression 4. We pply trnsform to the regression coefficients 6 to otin seprte regression coefficients for the nthropogenic nd nturl responses. Sectors nd periods contining no oservtions were excluded from the nlysis. We use 19 1999 dt for the Arctic nd 195 1999 dt for the Antrctic, nd n empiricl orthogonl function trunction of 8 (the mximum possile for the Antrctic), lthough comprle results were otined t other trunctions (see Supplementry Informtion, Fig. S3). Figure 4 shows regression coefficients for the responses to nthropogenic nd nturl forcings, together with their uncertinties evluted from control vriility. In oth the Arctic nd the Antrctic, the nthropogenic regression coefficient is significntly greter thn zero, indicting detectle response 75 nture geoscience VOL 1 NOVEMBER 8 www.nture.com/nturegeoscience 8 Mcmilln Pulishers Limited. All rights reserved.
.9.6.3.3.6.9.9.6.3.3.6.9 c.9.6.3.3.6.9 d.9.6.3.3.6.9 Figure 3 Oserved nd simulted SAM-congruent nd SAM-residul temperture trends over the period 195 1999., Oserved SAM-congruent temperture trends, clculted y multiplying the oserved regression pttern of monthly temperture nomlies on the SAM y the trend in the SAM 1. The SAM trend ws clculted from 5-yr mens of non-normlized Mrshll SAM index over the period 1955 1999 nd extrpolted to 195., Simulted SAM-congruent trends, clculted from the ALL ensemle in the sme wy. c,d, Oserved (c) nd simulted (d) residul trends, clculted y sutrcting SAM-congruent trends from ctul trends (Fig.,). 1 8 6 4 Antrctic Arctic 4 6 8 1 ANT Figure 4 Regression coefficients of oserved polr temperture nomlies ginst the simulted response to nthropogenic (x xis) nd nturl (y xis) forcing. Results re sed on 5-yr men Arctic 9 sector men temperture nomlies etween 19 nd 1999 (solid lines) nd 5-yr men Antrctic 9 sector men temperture nomlies etween 195 nd 1999 (dshed lines). One-dimensionl 5 95% uncertinty rnges nd curves enclosing 9% of the estimted joint distriutions were estimted from simulted internl vriility. to nthropogenic forcing. In oth cses, the nthropogenic regression coefficient is lso consistent with one, indicting consistency etween the mgnitude of the simulted nd oserved nthropogenic response nd tht prt of the oserved temperture chnges in oth polr regions is ttriutle to nthropogenic influence. A residul test 6 indictes no inconsistency etween simulted nd oserved vriility in either region. The response to nturl forcings ws lso detected over oth poles t some trunctions, lthough its regression coefficient ws often greter thn one, nd it ws found to e sensitive to vritions in empiricl orthogonl function trunction (see Supplementry Informtion, Fig. S3). As oservtionl coverge ws prticulrly limited t the strt of the record in the Antrctic, we repeted the detection nd ttriution nlysis for the Antrctic over the periods 1955 1999 nd 196 1999 nd found detectle nthropogenic response in oth cses. We lso found tht nthropogenic influence ws roustly detectle in Antrctic surfce temperture fter sutrcting the component of surfce temperture chnge linerly congruent with the SAM (see Supplementry Informtion, Fig. S3). It ws not possile to roustly seprte the greenhouse gs nd ozone influence on surfce temperture in either polr region, proly ecuse of the sprse oservtionl coverge nd low signl-to-noise rtio. The IPCC Fourth Assessment Report 8 concluded tht Anthropogenic influence hs een detected in every continent except Antrctic (which hs insufficient oservtionl coverge to mke n ssessment). Our findings demonstrte tht nthropogenic influence is detectle in Antrctic lnd surfce temperture, nd distinguishle from nturlly forced response, even given the limited sttion network nd short period for which dt re ville, nd tht circultion chnges, which re nture geoscience VOL 1 NOVEMBER 8 www.nture.com/nturegeoscience 753 8 Mcmilln Pulishers Limited. All rights reserved.
lrgely nthropogenic 8, hve reduced wrming rtes over most of Antrctic in models nd oservtions in recent decdes. In the Arctic, some uthors hve suggested tht oserved Arctic temperture chnges re inconsistent with climte model predictions 7, nd dominted y internl vriility 8,9, nd indeed so fr no forml ttriution studies of Arctic temperture chnge exist. We find tht nthropogenic influence on Arctic temperture is detectle nd distinguishle from the influence of nturl forcings. Although climte models my not relisticlly simulte the recent decline in Arctic se ice 3, we find no evidence in twentieth-century lnd temperture chnges to suggest tht climte models systemticlly underestimte high-ltitude climte feedcks. Overll, despite the pucity of oservtions, we find tht humn-induced wrming is detectle in oth these regions of high vulnerility to climte chnge. Received 4 July 8; ccepted 3 Septemer 8; pulished 3 Octoer 8. References 1. Meehl, G. A. et l. in The Fourth Assessment Report of the Intergovernmentl Pnel on Climte Chnge (eds Solomon, S. et l.) (Cmridge Univ. Press, 7).. Symon, C., Arris, L. & Hel, B. (eds) Arctic Climte Impct Assessment (Cmridge Univ. Press, 4). 3. Johnnessen, O. M. et l. Arctic climte chnge: Oserved nd modelled temperture nd se-ice vriility. Tellus A 56, 38 341 (4). 4. Wng, M. Y. et l. Intrinsic versus forced vrition in coupled climte model simultions over the Arctic during the twentieth century. J. Clim., 193 117 (7). 5. Polykov, I. V. et l. Vriility nd trends of ir temperture nd pressure in the mritime Arctic, 1875. J. Clim. 16, 67 77 (3). 6. Overlnd, J. et l. The Arctic nd Antrctic: Two fces of climte chnge. Eos Trns. AGU 89, 177 178 (8). 7. Turner, J. et l. Antrctic climte chnge during the lst 5 yers. Int. J. Climtol. 5, 79 94 (5). 8. Hegerl, G. C. et l. in The Fourth Assessment Report of the Intergovernmentl Pnel on Climte Chnge (eds Solomon, S. et l.) (Cmridge Univ. Press, 7). 9. Brohn, P. et l. Uncertinty estimtes in regionl nd glol oserved temperture chnges: A new dt set from 185. J. Geophys. Res. 111, D116 (6). 1. Jones, P. D. et l. Surfce ir temperture nd its chnges over the pst 15 yers. Rev. Geophys. 37, 173 199 (1999). 11. Lemke, P. et l. in The Fourth Assessment Report of the Intergovernmentl Pnel on Climte Chnge (eds Solomon, S. et l.) (Cmridge Univ. Press, 7). 1. Jones, P. D. Antrctic tempertures over the present century study of the erly expedition record. J. Clim. 3, 1193 13 (199). 13. Trenerth, K. E. et l. in The Fourth Assessment Report of the Intergovernmentl Pnel on Climte Chnge (eds Solomon, S. et l.) (Cmridge Univ. Press, 7). 14. Connolley, W. M. & Brcegirdle, T. J. An Antrctic ssessment of IPCC AR4 coupled models. Geophys. Res. Lett. 34, L55 (7). 15. Gleckler, P. J., Tylor, K. E. & Doutriux, C. Performnce metrics for climte models. J. Geophys. Res. 113, D614 (8). 16. Monghn, A. J., Bromwich, D. H., Chpmn, W. & Comiso, J. C. Recent vriility nd trends of Antrctic ner-surfce temperture. J. Geophys. Res. 113, D415 (8). 17. Monghn, A. J., Bromwich, D. H. & Schneider, D. P. Twentieth century Antrctic ir temperture nd snowfll simultions y IPCC climte models. Geophys. Res. Lett. 35, L75 (8). 18. Stott, P. A. & Tett, S. F. B. Scle-dependent detection of climte chnge. J. Clim. 11, 38 394 (1998). 19. Mrshll, G. J., Orr, A., vn Lipzig, N. P. M. & King, J. C. The impct of chnging Southern Hemisphere Annulr Mode on Antrctic Peninsul summer tempertures. J. Clim. 19, 5388 544 (6).. Chpmn, W. L. & Wlsh, J. E. A synthesis of Antrctic tempertures. J. Clim., 496 4117 (7). 1. Thompson, D. W. J. & Solomon, S. Interprettion of recent Southern Hemisphere climte chnge. Science 96, 895 899 ().. Mrshll, G. J. Trends in the Southern Annulr Mode from oservtions nd renlyses. J. Clim. 16, 4134 4143 (3). 3. Stott, P. A. et l. Externl control of th century temperture y nturl nd nthropogenic forcings. Science 9, 133 137 (). 4. Allen, M. R. & Stott, P. A. Estimting signl mplitudes in optiml fingerprinting, prt I: theory. Clim. Dyn. 1, 477 491 (3). 5. Gillett, N. P. et l. Detecting nthropogenic influence with multi-model ensemle. Geophys. Res. Lett. 9, 197 (). 6. Allen, M. R. & Tett, S. F. B. Checking for model consistency in optiml fingerprinting. Clim. Dyn. 15, 419 434 (1999). 7. Przyylk, R. Temporl nd sptil vrition of surfce ir temperture over the period of instrumentl oservtions in the Arctic. Int. J. Climtol., 587 614 (). 8. Przyylk, R. Chnges in sesonl nd nnul high-frequency ir temperture vriility in the Arctic from 1951 to 199. Int. J. Climtol., 117 13 (). 9. Polykov, I. V. et l. Oservtionlly sed ssessment of polr mplifiction of glol wrming. Geophys. Res. Lett. 9, 1878 (). 3. Stroeve, J. et l. Arctic se ice decline: Fster thn forecst. Geophys. Res. Lett. 34, L951 (7). Supplementry Informtion ccompnies the pper t www.nture.com/nturegeoscience. Acknowledgements We thnk S. Solomon, G. Mrshll nd H. Melling for useful dvice nd discussion; M. Allen for his optiml detection nd ttriution code; nd G. Jones for ssistnce with the provision of model output. This work ws supported in prt y the Climte Chnge Detection nd Attriution Project, jointly funded y NOAA s Office of Glol Progrms nd the US Deprtment of Energy. N.P.G. nd A.Y.K. were lso supported y NERC grnt NE/E6787/1, nd N.P.G. cknowledges support from the Leverhulme Trust. P.A.S. ws supported y the Joint Defr nd MoD Progrmme, (Defr) GA111 (MoD) CBC/B/417 Annex C5. Author contriutions N.P.G. crried out most of the nlysis nd wrote the pper. D.A.S., P.A.S., T.N. nd M.F.W. ssisted with the provision of model dt. A.Y.K. clculted SAM trends nd regression ptterns. G.C.H. proposed the study. P.D.J. provided dvice on oservtions. Author informtion Reprints nd permissions informtion is ville online t http://npg.nture.com/reprintsndpermissions. Correspondence nd requests for mterils should e ddressed to N.P.G. 754 nture geoscience VOL 1 NOVEMBER 8 www.nture.com/nturegeoscience 8 Mcmilln Pulishers Limited. All rights reserved.