GEOPHYSICAL RESEARCH LETTERS, VOL.22,NO. 22,PAGES 3047-3050, NOVEMBER 15, 1995 A non-fluxcorrected transient experiment using the BMRC coupledatmosphere/ocean GCM R.A. Colman, S.B. Power, B.J. McAvaney and R.R. Dahni Bureauof Meteorology Research Centre, Melbourne,Australia AbstractA transient CO2 experiment has been performed with the BMRC coupled atmosphere ocean General Circulation Model (GCM). No flux corrections wereused in the experiment. A roughly linear temporal increase in global surface temperature is found in response to the CO2 increase. The rateof increase isconsistent with the (relatively low) equilibrium response found previously using theAtmospheric GCM anda simple ocean. Ocean surface temperatures increase more at mid latitudes than athigh northern orsouthern latitudes where heat is sequestered into the deep ocean.Despite the secular climate drift which occurs in the model,the majorpatterns of atmospheric and oceanictemperature change are similar to changes noted elsewhere (from flux corrected models). This adds further support to the mainconclusions drawn from transient CO2 experiments performed elsewhere withcoupled GCMs. Introduction In recent years considerable effort has been expended worldwide on the investigation of the possible impacts on climateof increased atmospheric CO•levels resulting fromanthropogenic emissions (IPCC, 1992). Coupled atmosphere ocean General Circulation Models(GCMs) have proven to bepowerful tools in the study of the time evolution of these impacts (Manabe et al., !991, Cubasch et al., 1991,Murphy andMitchell, 1995, Meehl et al., 1993). Earlierexperiments with the Bureau of Meteorology Research Centre(BMRC) atmospheric GCM (AGCM) coupled to a slab ocean found relatively low climate sensitivity toan equilibrium CO2 change compared withsimilar experiments performed elsewhere. It is then of considerable interest to investigate theresponse of the recently developed BMRC Coupled atmosphere ocean GCM (CGCM) toa transient increase in CO2. Many modellers have chosen to apply so-called "flux corrections" to their CGCMs. These are additional surfacefluxes which are added tothose diagnosed bytheAGCM, before being passed to the Ocean GCM (OGCM). They aredesigned to minimise "drift" in the CGCM climate which results from inconsistencies in surface fluxes which thecomponent models require to maintain their present climate (e.g.Manabe et al., 1991). No flux corrections areapplied in theBMRC CGCM. One other non-flux corrected transient experiment has been reported in the literature, that of the NationalCenterof Atmospheric Research (NCAR), (Meehl et al. (1993)). These published results suggest that theimpacts found in non-flux corrected wansient experiments may compare closely withthose found from fluxcorrected experiments, despite theextra climate drift which occurs in the absence of flux correction (see also IPCC, 1992). It is of additional interest, therefore, to consider how the results from the present experiment compare with results from experiments performed elsewhere. Thisis particularly so because the present experiment has different physical parametriz- ations anddifferent (somewhat higher) model resolution thanthe modelused by Meehl et al. (1993). Model and Experiments TheBMRC AGCM wasoriginally described by McAvaney et al. (1978). Themodel uses spectral techniques in the horizontal with a rhomboidal truncation limit of 21 waves, with 9 levels in the vertical. The boundary layer and vertical diffusion parametrizations arebased on theformulations of Louis et al. (1982). Evaporation over theocean is enhanced at low wind speed following Milleretal.(1992). Other parametrizations are as described in ColmanandMcAvaney (1995). The OC•M is a version of the GFDL code which is based on the work of Bryan(1969), Cox (1984) andPacanowski et al. (1991). Thehorizontal tracer gridcorresponds with the AGCM Gaussian grid. There are 12 vertical levels, with greater resolution concentrated in theupper layers.The vertical eddy viscosity and diffusivity areenhanced in thetoptwo levels to simulate a surface mixed layer. Convection is parametrized as enhanced vertical diffusion andviscosity whenever the column becomes statically unstable. In this version of themodel theheat and freshwater fluxesare balanced only by verticaldiffusion (and/or convection) in the uppermost layer. This nevertheless approximates the balance in another version of this model which retains additional transport processes, because of theverylarge vertical diffusion employed between theupper two levels. The effect on themean transient response of themodel is minimal, as it occurs in bothcontrol and transient experiments and its impact issmall compared withtheclimate driftarising fromthe mismatch in surface fluxes. The seaice modelis a "zero layer"thermodynamic model followingSemtner (1976). Prior tocoupling both the OCK2M and AGCM were "spun up" toquasi-equilibrium. After coupling, theAGCM supplies to the OGCM and ice models surface heat and moisture fluxes, continental runoff and surface stresses. These fluxes are accumulated bythe AGCM over the 3 timesteps it takes between OGCM timesteps. Further details of theCGCM are given by Power et al. (1993). In the present experiments, the CGCM was integrated fora period of 105years withatmospheric CO2 held fixed at 330 ppm (hereafter calledthe "control" experiment). After 30 years of coupling a parallel integration was commenced in whichCO2was increased by 1% (compounded) per annum to doubling. Copyright 1995bythe American Geophysical Union. Paper number 95GL01727 0094-8534/95/95GL-01727503.00 Results and discussion Figure 1 shows a time series plotof theglobal mean surface air temperature for thefull experiment. In the control integration a secular climate drift can be seen (of approximately 0.13 3047