R EPORTS energy of acoustic free oscillations will be in 13. P. Lognonne«, E. Cle«ve«de«, H. Kanamori, Geophys. J. Int. 18. R. Kandel et al., Bull. Am. Meteorol. Soc. 79, 765 (1998). part transfered to the resonant modes of seis- 135, 388 (1998); 19. E. L. Fleming, S. Chandra, J. J. Barnett, M. Corney, Adv. mic free oscillations to amplify the latter 14. P. Lognonne«, B. Mosser, F. A. Dahlen, Icarus 110, 186 Space Res. 10, 11 (1990). (1994). 20. We are grateful to a number of people who were amplitudes. This amplification should depend 15. H. Kanamori and J. Mori, Geophys. Res. Lett. 19, 721 associated with IRIS and GEOSCOPE. We thank N. critically on how close the resonant frequen- (1992). Suda, K. Nawa, K. Nakajima, and S. Watada for com- cies are between the solid Earth and the 16. R. Widmer and W. Zu¬rn, Geophys. Res. Lett. 19, 765 ments on this paper. (1992). atmosphere. The eigenfrequencies of acoustic 17. H. Jacobowitz et al., J. Geophys. Res. 89, 4997 (1984). 11 November 1999; accepted 8 February 2000 modes are sensitive to the acoustic structure of the atmosphere (12), which varies annually (19). The above difference between 0S29 and other modes suggests that this annual varia- Simulation of Early 20th tion of the acoustic structure more precisely tunes the resonant frequencies of acoustic Century Global Warming modes to those of seismic modes in the sum- mer of the Northern Hemisphere. We may Thomas L. Delworth* and Thomas R. Knutson alternatively attribute the annual variation of the oscillations to the variation of the source The observed global warming of the past century occurred primarily in two height, to which the excitation of coupled distinct 20-year periods, from 1925 to 1944 and from 1978 to the present. modes is sensitive. Thus, the observed phe- Although the latter warming is often attributed to a human-induced increase nomena are best explained by the atmospher- of greenhouse gases, causes of the earlier warming are less clear because this ic excitation hypothesis, although other pos- period precedes the time of strongest increases in human-induced greenhouse sibilities, such as disturbances of oceanic or- gas (radiative) forcing. Results from a set of six integrations of a coupled igin, cannot be ruled out. ocean-atmosphere climate model suggest that the warming of the early 20th The phenomenon of the background free century could have resulted from a combination of human-induced radiative oscillations represents the hum of the solid forcing and an unusually large realization of internal multidecadal variability of Earth, which we found to be resonant with the the coupled ocean-atmosphere system. This conclusion is dependent on the hum of the atmosphere through the two fre- model’s climate sensitivity, internal variability, and the specification of the quency windows. The excitation source of time-varying human-induced radiative forcing. the hum may be at lowest part of the convec- tive zone of the troposphere, so that it is also Confidence in the ability of climate models to ries of global mean surface air temperature responsible for the hum of the atmosphere. make credible projections of future climate provides a remarkable match to the observed The hum in the resonant windows is louder change is influenced by their ability to repro- record, including the global warmings of both and shows a greater annual variation. The duce the observed climate variations of the the early (1925–1944) and latter (1978 to the phenomenon can be understood only if the 20th century, including the global warmings present) parts of the century. Further, the two systems of the solid Earth and atmo- in both the early and latter parts of the cen- simulated spatial pattern of warming in the sphere are viewed as a coupled system. tury (1). Several climate models accurately early 20th century is broadly similar to the simulate the global warming of the late 20th observed pattern of warming. Thus, we dem- century when the radiative effects of increas- onstrate that an early 20th century warming, References and Notes ing levels of human-induced greenhouse gas- with a spatial and temporal structure similar 1. K. Nawa et al., Earth Planet. Space 50, 3 (1998). 2. N. Suda, K. Nawa, Y. Fukao, Science 279, 2089 es (GHGs) and sulfate aerosols are taken into to the observational record, can arise from a (1998). account (2–4). However, the warming in the combination of internal variability of the cou- 3. T. Tanimoto, J. Um, K. Nishida, N. Kobayashi, Geo- early part of the century has not been well pled ocean-atmosphere system and human- phys. Res. Lett. 25, 1553 (1998). simulated using these two climate forcings induced radiative forcing from GHG and sul- 4. N. Kobayashi and K. Nishida, Nature 395, 357 (1998). 5. K. Nishida and N. Kobayashi, J. Geophys. Res. 104, alone. Factors which could contribute to the fate aerosols. These results suggest a possible 28741 (1999). early 20th century warming include increas- mechanism for the observed early 20th cen- 6. N. Kobayashi and K. Nishida, J. Phys. Condens. Matter ing GHG concentrations, changing solar and tury warming. 10, 11557 (1998). 7. S. W. Smith, Eos 67, 213 (1986). volcanic activity (4–6), and internal variabil- The coupled ocean-atmosphere model that 8. G. Roult and J. P. Montagner, Ann. Geofis. 37, 1054 ity of the coupled ocean-atmosphere system. was used, developed at the GFDL, is higher (1994). The relative importance of each of these fac- in spatial resolution than an earlier version 9. The 17 IRIS stations we analyzed were AAK (Ala tors is not well known. used in many previous studies of climate Archa, Kyrgyzstan), BFO (Black Forest, Germany), COL (College Outpost, AK, USA), COR (Corvallis, OR, Here, we examine results from a set of variability and change (7, 8), but it employs USA), CTAO (Charters Towers, Australia), ENH (Enshi, five integrations of a coupled ocean-atmo- similar physics. The coupled model is global China), ESK (Eskdalemuir, Scotland), HIA (Hailar, Chi- sphere model forced with estimates of the in domain and consists of general circulation na), KMI (Kunming, China), LSA (Lhasa, China), MAJO (Matsushiro, Japan), MDJ (Mudanjiang, China), PFO time-varying concentrations of GHGs and models of the atmosphere (R30 resolution, (Pinon Flat, CA, USA), SSE (Shanghai, China), SUR sulfate aerosols over the period 1865 to the corresponding to 3.75° longitude by 2.25° (Sutherland, Republic of South Africa), TLY (Talaya, present, along with a sixth (control) integra- latitude, with 14 vertical levels) and ocean Russia), and WMQ (Urumqi, China). The eight GEO- SCOPE stations were CAN (Mount Stromlo, Canberra, tion with constant levels of greenhouse gases (1.875° longitude by 2.25° latitude, with 18 Australia), HYB (Hyderabad, India), INU (Inuyama, and no sulfate aerosols. In one of the five vertical levels). The model atmosphere and Japan), KIP (Kipapa, HI, USA), NOUC (Port Laguerra, GHG-plus-sulfate integrations, the time se- ocean communicate through fluxes of heat, Nouvelle Caledonia), SCZ (Santa Cruz, CA, USA), TAM water, and momentum at the air-sea interface. (Tamanrasset, Algeria), and WUS (Wushi, China). Flux adjustments are used to facilitate the 10. A. M. Dziewonski and D. L. Anderson, Phys. Earth Geophysical Fluid Dynamics Laboratory (GFDL)/Na- Planet. Inter. 25, 297 (1981). tional Oceanic and Atmospheric Administration, simulation of a realistic mean state. A ther- 11. D. C. Agnew and J. Berger, J. Geophys. Res. 83, 5420 Princeton, NJ 08542, USA. modynamic sea-ice model is used over oce- (1978). 12. S. Watada, thesis, California Institute of Technology, *To whom correspondence should be addressed. E- anic regions, with ice movement determined Pasadena, CA (1995). mail: [email protected] by ocean currents. 2246 24 MARCH 2000 VOL 287 SCIENCE www.sciencemag.org R EPORTS The first integration is a 1000-year control difference between a single realization and an warming. The four other GHG-plus-sulfate case, with no year-to-year variations in exter- independent five-member ensemble mean oc- experiments show a range of variability in the nal radiative forcing. After a small initial curs approximately 4.8% of the time, demon- early part of the record, illustrating the inter- climate drift over the first 100 years, the strating that although the 1910–1944 trend is nal variability of the model. Interestingly, a coupled model is very stable for the remain- a relatively rare occurrence for this model, it warming at high latitudes of the Northern ing 900 years of the integration. In the other is still within the range of possibilities. Hemisphere is seen in the late 1800s of ex- five integrations, an estimate of the observed We now assess whether internal variabil- periment 1, illustrating that aspects of the time-varying concentrations of GHGs plus ity alone can account for the observed early warming seen in the 1920s and 1930s of sulfate aerosols (9–11) is used to force the 20th century warming, or if the radiative experiment 3 occur at other times. Also no- model over the period 1865–2000. The radi- forcing from increasing concentrations of table is the pronounced high northern latitude ative perturbations associated with sulfate GHGs is also necessary. Over the period cooling in the 1920s and 1930s that occurs in aerosols are modeled as prescribed changes 1910–1944 (which encompasses the warm- experiment 5. A more general warming oc- in the surface albedo. The latter five integra- ing of the 1920s and 1930s), there is a linear curs over the last several decades in all the tions are identical in experimental design, trend of 0.53 K per 35 years in observed experiments and in the observations, suggest- with the exception of the initial conditions, global mean temperature.