A model study of the impact of source gas changes on the stratosphere for 1850-2100

. The long term stratospheric impacts due to emis sions of CO 2, CH 4, N 2 0, and ozone depleting substances (ODSs) are investigated using an updated version of the God dard two-dimensional (2D) model. Perturbation simulations 5 with the ODSs, CO 2, CH 4 , and N 2 0 varied individually are performed to isolate the relative roles of these gases in driv ing stratospheric changes over the 1850-2100 time period. We also show comparisons with observations and the God-40 dard Earth Observing System chemistry-climate model sim-10 ulations for the time period 1970-2100 to illustrate that the 2D model captures the basic processes responsible for long term stratospheric change. The 2D simulations indicate that prior to 1940, the 45 ozone increases due to CO 2 and CH 4 loading outpace the 15 ozone losses due to increasing N 2 0 and carbon tetrachloride (CCI 4) emissions, so that ozone reaches a broad maximum during the 1920s-1930s. This preceeds the significant ozone depletion during ~ 1960-2050 driven by the ODS loading. 50 During the latter half of the 2]8t century as ODS emissions 20 diminish, CO 2 , N 2 0, and CH 4 loading will all have signif icant impacts on global total ozone based on the IPCC A I B (medium) scenario, with CO 2 having the largest individual effect. Sensitivity tests illustrate that due to the strong chem-55 ical interaction between methane and chlorine, the CH 4 im-25 pact on total ozone becomes significantly more positive with larger ODS loading. The model simulations also show that changes in stratospheric temperature, Brewer-Dobson circu lation (BDC), and age of air during 1850-2100 are controlled 60 mainly by the CO 2 and ODS loading. The simulated ac-30 celeration of the BDC causes the age of air to decrease by ~ I year from 1860-2100. The corresponding photochemi cal lifetimes of N 2 0, CFCl: l , CF 2 Ch, and CC1 4 decrease by 11-13% during 1960-2100 due to the acceleration of the BDC, with much smaller lifetime changes «4%) caused by changes in the photochemical loss rates.


Introduction
Changes in the atmospheric abundance of halogenated ozone depleting substances (ODSs) and the greenhouse gases (GHGs) CO 2 , CH 4 , and N 2 0 have been shown to significantly impact the chemical and dynamical structure of the stratosphere [e.g., World Meteorological Organization (WMO), 2007(WMO), , 2011]].For example, much of the decline of stratospheric ozone during the 1980s and 1990s has been attributed to increased atmospheric halogen loading due to anthropogenic forcings.Increases in N 2 0 and the odd nitrogen species decrease ozone in the middle stratosphere [e.g., Crutzen, 1976], while increases in CO 2 and the subsequent cooling reduce the temperature dependent ozone loss rates and cause ozone increases in the upper stratosphere [e.g., Haigh and Pyle, 1979;Rosenfield et aI., 2002].
Recent observational studies have detected the beginning of the ozone recovery process in the upper stratosphere, where ozone is most sensitive to changes in halogen loading [e.g., Reinsel, 2002;Newchurch et aI., 2003].However, detection of the change in the halogen influence on ozone can be complicated by the impacts due to long term changes in GHGs.It is therefore of interest to separate the relative impacts of the different chemical processes that control long term ozone changes.
Another important aspect of the changing atmospheric composition impact on the stratosphere is the quantification of photochemical lifetimes of the ODSs and GHGs.These lifetimes have come under recent scrutiny [Douglass et aI., 2008], as they are important for deriving surface mixing ratio boundary conditions from emissions estimates for use in atmospheric models [Kaye et aI., 1994;WMO, 2011].The potential influence on lifetimes of the Brewer-Dobson circulation (BDC) acceleration due to climate change has also been investigated [Butchart and Scaife, 200 I;Douglass et aI., 2008].
Most studies of past and future stratospheric change now utilize three-dimensional (3D) coupled chemistry-climate models (CCMs) [e.g., Eyring et aI., 2006;Eyring et aI., 2007;WMO, 20111.Some 3D CCM investigations have shown the impact of different processes on long term stratospheric change, such as that due to the multi-decadal changes in ODS and GHG concentrations and sea surface temperatures (SSTs) [e.g., Butchart and Scaife, 2001;Austin et £II., 2007;Olsen et aI., 2007;Li et al., 2008;Eyring et aI., 20 lOa;Eyring et aI., 20 lOb;Austin et aI., 20101.However, performing numerous sensitivity simulations to separate the different chemical processes that control stratospheric changes can be more easily done using two-dimensional (2D) models, given their much smaller computational requirements.2D models have been widely used in international assessments of the stratosphere [e.g., WMO, 2003WMO, , 2007WMO, , 2011]], and past studies have shown that 2D models can resolve much of the large scale stratospheric variability on monthly and longer time scales, as seen in comparisons with observations and 3D models [e.g., Plumb and Mahlman, 1987;Yudin et aI., 2000;Fleming et £II., 2007;Newman et aI., 2009].Previous 2D model studies have investigated the relative roles of the long term changes in CO 2 , CH 4 , and N 2 0, focussing on the stratospheric ozone changes over the next century [e.g., Randeniya et £II., 2002;Chipperfield and Feng, 2003;Portmann and Solomon, 2007].
In this paper we expand on these previous studies and examine in more detail the relative contributions of the long toO term changes in atmospheric GHG and ODS loading using our recently upgraded Goddard Space Flight Center (GSFC) 2D coupled chemistry-radiation-dynamics model.We exploit the computational speed of the 2D model to perform numerous perturbation simulations to investigate the stratospheric impacts due to GHG and ODS loading for the 250year time period, 1850-2100.We examine the ozone, temperature, and age of air impacts, and focus on the time periods 125 prior to 1950 and the latter half of the 21 8t century.We also use perturbation tests to examine the ozone impacts due to the chemical coupling between CH4 and chlorine.We then investigate the long term time dependence of the photochemicallifetimes ofN 2 0, CFC-11, CFC-12, and CCI 4 .Here, we 130 examine the relative importance of changes in the BDC and the photochemical loss rates in controlling these lifetimes.
The recent SPARC Chemistry-Climate Model Validation Activity [SPARC CCMVal, 20101 provided a comprehensive process oriented evaluation of many CCMs.Because 2D models were not included in this activity, and given the recent improvements to our 2D model, we provide in Appendicies A and B, a detailed description and evalua-135 tion of our upgraded model, comparing climatological simulations with observations of various stratospheric tracers.Throughout the paper, we also compare long term simulations from the 2D model with the Goddard Earth Observing System chemistry-climate model (GEOSCCM) and multidecadal observational data sets to illustrate that the 2D model captures the basic processes that drive long term changes in stratospheric ozone, temperature, and age of air.The good 2D model agreement with the measurements and the GEOSCCM then justifies the use of the 2D model for the perturbations addressed in this study.

Model Simulations
For this study, we utilize a series of 2D model experiments in which the surface concentrations of only the ODSs or the individual GHGs are varied time dependently for 1850-2100, while all other source gases are fixed at low (1850) levels.In this way, we separate the individual effects of the ODS and GHG loading.We compare these with the 2D baseline simulation in which all source gases are varied time de-E.L. Fleming et al.: Impact of source gas changes on the stratosphere 3 14D pendently, and with the GEOSCCM baseline simulation for 1950-2100.The GEOSCCM couples the GEOS-4 general circulation model with stratospheric chemistry and has been applied to various stratospheric problems [e.g., Stolarski et aI., 2006;Pawson et ai., 2008;Waugh et aI., 2009;Oman et aI., 2009;Newman et aI., 2009;Li et aI., 2009].The GEOSCCM uses specified time dependent SSTs and sea-ice amounts, and the results presented in this study are comprised of three simulations which utilize somewhat different SSTs for the past and future time periods : 1950-2004,1971-15D 2052, and 1996-2100.For 1950-2100, the 2D and GEOSCCM simulations use surface ODS boundary conditions from scenario A I of WMO [2007], and GHG boundary conditions from scenario A I B (medium) from the Intergovernmental Panel on Cli-155 mate Change (lPCC) Special Report on Emissions Scenarios [IPCC, 2000].The 2D simulations for 1850-1950 use GHG surface boundary conditions from Hansen and Sato [2004].For the ODSs, most are zero prior to 1950, except for the following: CFCb and CF 2Cl2 are set to zero prior to 1935 and HiD 1946, respectively, and are then ramped up slowly to the 1950 WMO [2007] values.CCl 4 is ramped up exponentially from zero in 1900 to the 1950 value ofWMO [2007], approximating the time series ofButleret al. [1999].CH 3 CI and CH3Br are set to 440 pptv and 5 pptv respectively in 1850, and follow the time variation up to 1950 as discussed in Butler et al. [1999] and WMO [20031 Here, EESC is taken as the global average of Cly + 60Bry at 50 km.Note that all 2D and GEOSCCM calculations use fixed solar flux (no solar cycle variations) and clean stratospheric aerosol conditions specified from WMO [2007].

Base model-data comparisons
As a general model evaluation, we first compare the simulated vertical profile ozone trends for 1979-1996 with the SBUY data for the near-global (60 0 S-600N) average (Figure 180 2, top).These are derived from regression fits to the EESC time series (Figure I) for 1979(Figure I) for -2004.At 50 hPa (the lowest level of the SBUY data), the observationally-derived trend is underestimated in the 2D base simulation (all source gases varied time dependently), less so in the GEOSCCM.However above 20 km, both base simulations are mostly in rea-195 sonable agreement with the observations.Note that we do not show GEOSCCM results in the troposphere as ozone is relaxed to an observational climatology in this region.
The recent past and future ozone changes in the 2D model average (Figure 2) and in the latitude-height variations (Figures 3 and 4).This general model agreement is also seen in time series of near-global profile ozone (Figure 5), global total ozone (Figure 6), and tropical total ozone (Figure 7).For reference, we include time series of the BUY/SBUY satellite observations for profile ozone and ground-base data for global total ozone (updatcd from Fioletov et al. [2002]).
The largest 2D-GEOSCCM differences occur in the Antarctic ozone hole region, where the 2D model simulates smaller past (negative) and future (positive) ozone changes compared to the GEOSCCM (Figures 3 and 4).Some of this is likely due to the 2D model not fully resolving the processes that control polar ozone loss, as these can have large zonal asymmetries.Some of this model ditference also reflects the known high ozone bias in the high latitude lower stratosphere in the GEOSCCM.This bias is most pronounced during periods of low chlorine loading so that the chlorineinduced changes in the Antarctic spring are too large by 60-80% [Pawson et aI., 20081.These model differences are reflected in the near-global averaged vertical profiles below 18 km (Figure 2) and in the total column time series (Figure 6) in which the GEOSCCM simulates significantly more past ozone reduction and future ozone increase.However, the generally good agreement between the 2D model base simulations and the observations and GEOSCCM in Figures 2-7 show that the 2D model captures the basic processes responsible for long term stratospheric ozone changes.

2D perturbation simulations
The relative roles of ODS, CO 2 , CH 4 , and N 2 0 loading in controlling the recent past and future ozone changes are illustrated by the 2D model perturbation simulations in Figures 2-7.This includes the well known dominance of ODS loading in controlling the sharp ozone decline in the lower and upper stratosphere globally during ~1970-2000.Because of the strong impact in the upper stratosphere (Figures 3 and  4), ODS loading also largely controls the tropical total ozone time series from ~ 1970 through the early 21 8t century (Figure 7).
CO 2 cooling and subsequent reduction in the ozone loss rates produce a broad ozone increase of 1-2%/decade in the upper stratosphere (Figures 2-4).For the 2005-2095 time period, the ozone increases due to increasing CO 2 and declining ODS emissions are similar in the upper stratosphere (I.5-2%/decade) in Figure 2 (bottom) and Figure 4.By 2100, CO 2 loading is the dominant impact at 40 km, causing a 20% increase in ozone from 1850-2100 (Figure 5, middle, red curve).

241)
Previous CCM studies have shown that increased GHG loading results in an acceleration of the BDC with related impacts on ozone and trace gases [e.g., Butchart and Scaife, 2001;Austin and Li, 2006;Garcia and Randel, 2008;Li e(255 a!., 2009].This feature is seen in the 2D CO 2 -only simulation in Figures 3 and 4, in which lower tropical stratospheric ozone is reduced by 1-2%/decade as ozone-poor air is advected upwards from the tropical troposphere.There is a compensating downward advection of ozone-rich air'60 in the extratropics at 10-15 km which is strongest in the NH.This hemispheric asymmetry is consistent with previous GEOSCCM simulations of the climate change impacts on ozone [Olsen et a!., 2007;Li et aI., 2009;Waugh et aI., 20091.These BDC-driven ozone changes are reflected in the 265 global average in Figure 2, with an ozone decrease (increase) of ~ 0.5o/c/decade centered near 20 km (15 km) in the 2D  C02-only simulation.The CO 2 -induced ozone decrease is dominant by 2100 in the near global average time series at 22 km (Figure 5, bottom, red curve), with a decrease of 5.5% from 1850-2100.At this level, N 2 0 and CH 4 loading have negligible impacts, so that the net result of ODS and CO 2 changes is an ozone increase from 2000-2030 and a decrease from 2030-2100 in the base simulations.
The net impact of CO 2 loading on total ozone is an increase of 12.5 DU (4.2%) from 1850-2100 in the global average (Figure 6, red curve).CO 2 loading also increases total ozone at midlatitudes of both hemispheres (not shown), and as with profile ozone, thc total column increase is more pronounced in the NH compared to the SH midlatitudes owing to the larger enhancement of the BDC north of the equator.In the tropics, the enhancement of the BDC advecting ozonepoor air from the troposphere counteracts the ozone increase in the upper stratosphere caused by the CO 2 cooling.The net effect is a decrease in tropical total column ozone throughout the 21 st century in Figure 7 (red curve), consistent with previous CCM results [Li et aI., 2009;Waugh et aI., 2009;Eyring et ai., 201Ob1.In Figures 2-5, N 2 0 loading and the subsequent increase in stratospheric NO y lead to a decrease in stratospheric ozone, with a maximum decline of -0.5 to -0.6%/decade near 35 km in the global average in both the past and future.29DThese magnitUdes are slightly less than obtained by Portmann and Solomon [20071 who used the IPCC A2 GHG scenario which has larger N 2 0 increases compared to the AlB scenario used here.From 1850-2100, N 2 0 loading results in a total ozone decrease of 8 DU (-2.7%) in the global average 295 (Figure 6) and 4 DU (-I in the tropics (Figure 7).This effect of N 2 0 loading taken in isolation is larger than would be when taking into account the effects of increased CO 2  1970-2009 (excluding 1973-1978) at 22 and 40 km ("+" symbols).To emphasize the model-data comparison after 1970, the data have been adjusted so that the 1970-1972 average matches that of the base simulations.
cooling on future NO y and ozone changes, as has been discussed previously [Rosenfield and Douglass, 1998;Daniel et aI., 20101.We will discuss this in more detail in section 3.5.
Atmospheric CH 4 impacts ozone via three mechanisms: I) increases in CH 4 increase the amount of H 2 0 in the stratosphere and mesosphere which in turn reduces ozone by enhancing the HOx-ozone loss cycles; 2) increases in CH 4 lead to increased ozone throughout the stratosphere by converting active chlorine to the reservoir HCI via the reaction CH 4 +Cl-,HCI+CH: l ; and 3) increases in CH 4 lead to increascd ozone in the troposphere and very lower stratosphere due NOx-induced ozone production [e.g,Brasseur Global total ozone relative to 1860 Fig. 6.Global and annual averaged total ozone time series relative to 1860 values.Shown are the base simulation (all source gases varied time dependently) from the 2D model (black solid line) and GEOSCCM (black dotted line).The GEOSCCM time series has been adjusted to match the 2D base simulation for 1960.Also shown are 2D simulations in which only certain source gases are varied time dependently as indicated, with the other source gases fixed at 1850 levels.The "+" symbols represent ground-based data updated from Fioletov et al. 120021.To emphasize the model-data comparison after 1970, the data have been adjusted so that the 1964-1970 average matches that of the base simulations.Also shown are the combined effects of CH4 and ODSs loading: the 2D-CH4 only(a) curve (orange dashed-dotted) is the difference between simulation with CH4 and the ODSs varied time dependently and that with only the ODSs varied time dependently, showing the effect of CH4 in the presence of time dependent ODS loading; the 2D-ODSs only(b) curve (blue dashed-dotted) is the difference between a simulation with CH4 and the ODSs varied time dependently and310 that with only CH4 varied time dependently, showing the effect of ODSs in the presence of time dependent CH4 (C02 and N2 0 are fixed at 1850 levels in these simulations).Values are in Dobson Unit (DU) change (left axes) and % change (right axes).and Solomon, 19861.This latter mechanism is strongly dependent on the amount of tropospheric N Ox.To more properly account for this in the 2D model, we constrain the model troposphere using output from the Global Modeling Initiative's (GMI) combined stratosphere-troposphere chemistry and transport model (GMI Combo CTM) [Strahan et al., 2007;Duncan et aI., 20071.This includes specifying the lower tropospheric (below 5 km) NO y boundary condition to account for surface emission and loss processes, NO x production from lightning, and washout of HN0 3 , all for present day conditions.The resulting 2D tropospheric NO x and NO y distributions compare favorably with the GMI model.The 2D model utilized in this way simulates a mid-upper tropospheric ozone response to a steady state CH 4 perturbation (0.5 ppmv) similar to that obtained in the GMI model for present day conditions.Figures 2-5 show that the combined effect of the three CH 4 mechanisms outlined above yield a significant ozone increase (1-1.5 %/decade) in the troposphere and very lower stratosphere, a mostly weak positive response at 20-40 km, and a negative ozone response above 40 km with decreases of 1 %/decade and larger above 60 km. Figure 5 (top panel) shows that CH 4 loading is dominant in controlling the ozone time dependence at 60 km, due to the subsequent increase in H 2 0 and HOx-ozone loss.This mechanism also results in significant CH 4 -induced ozone loss at 60 km prior to 1960 so that one-third of the total decrease in ozone during 1850-2050 has occurred by 1960.For total ozone, CH 4 loading results in increases from 1850-2050 of7.5DU (2.5%) in the global average (Figure 6) and 4.5 DU (1.7%) in the tropics (Figure 7).
The future ozone rcsponse to CH4 follows the surface boundary condition, which is significantly larger in 2055 than in 2100 in the AlB scenario (Figure I).This results in a relatively small change from 2005-2095 seen in the latitude-height section in Figure 4 « ±0.5%/decade everywhere).We show vertical profile results for both 2005-2055 and 2005-2095 in Figure 2, illustrating the much different magnitudes of the response for the different time periods.These responses are also qualitatively similar but have smaller magnitudes compared to Portmann and Solomon [20071 who used IPCC scenario A2 which has larger GHG loading.We note that because the 2D model tropospheric NO x and NO y are fixed to the present day GMI simulation, the tropospheric ozone response to the time dependent CH 4 perturbation shown in Figures 2-7 does not account for long term changes in tropospheric NO x emissions.These have undcrgone significant past increases which would lead to more CH4 -induced tropospheric ozone production than we show in Figures 2-7.
By 2100, ODS, C02, CH 4 , and N 2 0 loading all play important roles in controlling global total ozone.CO 2 has the largest impact, leading to a 4% increase from 1860-2100, with CH 4 10ading causing a 2% increase, and N 2 0 and ODS loading each contributing a 2-2.5% decrease from 1860-2100 (Figure 6).The net result is a 1.7% (5DU) increase from 1860-2100 in the 2D base simulation.In the tropics, baseline total ozone increases during 2000-2050.This is followed by a ~ 1 % decrease over the second half of the century, as the combination of increasing C02 and N 2 0 and decreasing CH 4 more than offset the effect of reduced ODS load- ing.This general time variation of 21 st century tropical total ozone is consistent with the GEOSCCM in Figure 7 and other larger by 1960 (+0.75%).The ODS impact, which is due CCM results reported recently [Eyring et aI.,20 lOa,20 lOb; mainly to CCl4 emissions, causes a 1 % depletion in the total Austin et a!., 20101.375 column by 1960, with the vast majority of this decrease oc-

1860-1960 Ozone
As seen in Figures 5-7, the ozone changes prior to 1960 are relatively small, but not insignificant, compared to post-1960.To examine this more closely, Figure 8 shows a close-3aG up of the j 860-1960 ozone time series for the near-global average at 40 km and the global average total column.
From 1860-1960, the total column impacts due to C02 and N 2 0 are approximately equal and opposite, being +0.5% and -0.5%, respectively.The CH 4 impact is slightly 385 curing after 1920.At 40 km, the impact of CO 2 cooling leads to a 2% ozone increase from 1860-1960, which is approximately equal and opposite to the ozone depletion caused by ODSs.
Ozone in the resulting base simulation (black curves) reaches broad pre-modem maxima during 1920-1940, with increases from 1860 of 0.8 DU (0.3%) in the total column and 0.06 ppmv (0.8%) at 40 km.This preceeds the decline in ozone driven mainly by ODS loading, which becomes much more rapid after ~ 1970 (Figures 5 and 6).E. L. Fleming et al.: Impact of source gas changes on the stratosphere

CH 4 sensitivity experiments
As the changes in CH 4 impact ozone via the different mechanisms outlined above.we now examine these effects in more detail.Figure 9 (top panel) shows the global/annual averaged 2D model steady state ozone change due to a 0.5 ppmv CH 4 perturbation for year 2000 conditions.This is expressed in DU/km to emphasize the contribution to the total column change.CH 4 is lost via reaction with OeD), OH. and Cl, and the black curve shows the standard case in which all three reactions use the perturbed CH 4 • This shows that the CH 4 impact on ozone is positive everywhere below 46 km, with only a very small negative change above 46 km, and a total column response of +2.7DU.
To qualitatively separate the effects of the different mechanisms, the green curve is a simulation in which perturbed CH 4 is used for the reactions with 0(1 D) and OH, with unperturbed CH 4 used for the reaction CH 4 +Cl.In this way, the conversion of active chlorine (Cl) to the reservoir (HCl) is not impacted by the additional CH 4 .Therefore, ozone only responds to the enhanced NOx-induced ozone production in the troposphere yielding ozone increases in this region, and the enhanced H 2 0 and HOx-ozone loss cycles above ~35 km and in the very lower stratosphere, which yield ozone decreases.The enhanced H 2 0 in this case also enhances polar stratospheric clouds and the heterogeneous chemical ozone loss in the SH polar region, which results in a negative response at 17-28 km in the global average in Figure 9.
The red curve in Figure 9 shows the opposite case of the green curve, i.e., here, perturbed CH 4 is used for the reaction CH 4 + Cl, with unperturbed CH 4 used for the reactions with 0(1 D) and OH.This significantly reduces the amount of H 2 0 and HOx-ozone loss, as well as the NOx-ozone production in the troposphere generated by the CH4 perturbation.This better isolates the impact of CH4 in eontrolling the chlorine partitioning and subsequent chlorine-catalyzed ozone loss.The resulting ozone response is positive throughout the stratosphere, with the largest impacts occurring in the regions where chlorine-catalyzed ozone destruction is largest, i.e., the upper stratosphere globally, and in the lower stratosphere corresponding to the ozone hole region.Comparing the green and red curves shows that the CH 4 + Cl reaction has the largest impact on ozone at ~ 15-45 km, with 440 the NOx-ozone production mechanism being dominant below 15 km.
The bottom panel in Figure 9 shows the same cases run for 2100 eonditions, which has greatly reduced chlorine loading compared with present day conditions.The effect of the fu-445 ture reduced chlorine loading is evident, as the case in which the perturbed CH 4 is used only for the CH4 + Cl reaction (red curve) exhibits a much weaker ozone response at all levels compared with the simulations for present day conditions.For the total column, this simulation accounts for only 28% of the full CH 4 response (+ 1.8DU) in 2100, compared to 70% of the full response in 2000.In 2100, the full ozone  For time dependent total ozone, the combined effect of CH 4 and Cly loading is illustrated in Figure 6.The solid orange curve (2D-CH4 only) uses fixed 1850 (very low) chlorine loading, and is therefore similar to the green curves in Figure 9 (unperturbed CH 4 +Cl) since the impact of the CH 4 + Cl reaction is minimal in this case.To illustrate the full effect of CH 4 in the presence of time dependent ODS loading (analogous to the black curves in Figure 9), the 2D-CH 4 only (a) curve (orange dashed-dotted in Figure 6) is the difference between a simulation with CH 4 and the ODSs var-ied time dependently and that with only the ODSs varied time dependently (C0 2 and N 2 0 are fixed at 1850 levels in both simulations).Compared to the simulation using fixed 1850 Cly (solid orange line), the full effect of CH 4 loading results in significantly more global total ozone when the ODS loading is highest, i.e., ~ 1985 through the first half of the 21 81 century.

4llO
This effect can also be seen in the ODS-only simulation, i.e., the effect of ODS loading in the presence of time dependent CH4 (the 2D-ODS only (b) curve (Olue dashed-dotted) in Figure 6).This is the difference between the simulation with CH 4 and the ODSs varied time dependently and that 465 with only CH 4 varied time dependently.Comparing the two blue curves illustrates how the impact of the ODS loading is mitigated by the time dependent CH 4 changes.This effect is largest when the ODS loading is largest, i.e., ~2000, when the global total ozone depletion due to ODS loading is 3 DU 470 ( I %) less in the presence of the time dependent CH 4 changes compared to that with fixed 1850 CH 4 levels.

Quantification of the Ozone Impacts due to GHGs
Given the importance of increasing GHG emissions on future ozone levels, we examine these relative impacts in more quantitative detail.To do this, we apply the concept of the Ozone Depletion Potential (ODP) to GHG emissions, as recently done for N 2 0 and the induced NOx-ozone destruction [Ravishankara et aI., 2009;Daniel et aI., 2010].Traditionally, the ODP metric has been used to quantify the 48D change in global ozone per unit mass emission of a specific chlorine-containing compound relative to the change in global ozone per unit mass emission of CFC-Il (CFCl:J) [Wuebbles, 1983;Fisher et aI., 1990;Solomon et aI., 1992].Because the ODP is defined as a ratio to the ozone loss due to 510 485 CFC-ll, many uncertainties in the ozone-loss computation cancel.For this reason ODPs of chlorocarbon compounds are generally less sensitive to photochemical modelling uncertainties than are absolute ozone loss calculations.
Application of the ODP concept to a non-chlorine con-515 taining compound does not benefit in the same way from the cancellation of uncertainties in the ratio of ozone loss to that of CFC-II, e.g.N 2 0 does not cause an ozone hole.More generally, the chemical mechanisms that impact ozone are different for GHGs, and tend to occur in different regions 52o of the atmosphere, compared to those of chlorine containing compounds.
While being cautious about the interpretation of the ODP concept as applied to non-chlorine containing compounds, we use the ODP metric to examine the relative impacts on 525 500 ozone of N 2 0, CH 4 , CO 2 , and CFC-II for 1850, 1950, 2000, and 2100 steady-state atmospheric conditions.These calculations are meant as a guide for evaluating the relative importance of GHG loading on past and future ozone levels.
We use the standard method to calculate ODPs, i.e.,53o change each by an amount that leads to a 1 % change in annually-averaged global total ozone.Table I shows the resulting annual and global average flux of each compound for the years indicated.The ozone change is negative for CFC-II and N 2 0, and positive for CH 4 and CO 2 .Note that the flux changes for N 2 0, CH4 and CO 2 are much larger, by one to three orders of magnitude, than those for CFC-II.Table 2 shows these numbers converted to the standard ODP, i.e., the ratio of the flux change needed to cause a 1% ozone change for each compound to the flux change for CFC-11.Our N 2 0 ODP values for 1950 and 2000 (+0.024 and +0.019, respectively), are similar to those reported in Ravishankara et al. [2009].These results also illustrate that while NzO has a positive ODP, CH4 and CO 2 have negative ODPs relative to CFC-l1.However because the fluxes and ODPs vary by two orders of magnitUde, and the background concentrations also undergo large changes, it is difficult to evaluate these values on an absolute basis or relative to one another.
Tables 3 and 4 illustrate another way to look at the problem of comparing the impact of GHGs on ozone.Table 3 shows the sensitivity of total column ozone to a specified change in the concentration of each GHG.These are expressed in percent change in ozone for a I ppbv or 1 ppmv change in the surface concentration of the GHGs.The impact of each can then be deduced by mUltiplying by the actual change in the concentration of the species over a given time period.This is N 2 0 -0.031 (12) -0.028 (13) -0.020 (16) -0.021 (13) 5£5 +1.8 (70) + 1.7 (49) +2.6 (22) +1.4 (36)   change.However, this time dependence also reveals interesting aspects of the interactions of the perturbations.For CFC-II, increasing levels of N 2 0 (NOy) and CH 4 convert active chlorine to the reservoirs CION0 2 and HCI, respectively, thereby reducing the efficiency of CIOx-ozone destruction.
For N 2 0, the time dependence is partly due to the changing background Cly, i.e., with higher Cly more NO x is tied up in CION0 2 , thereby reducing the efficiency of the NO x - ozone loss.Other factors include: I) the CO 2 cooling of the stratosphere which results in greater chemical destruction of NOy [Rosenfield and Douglass, 1998], and therefore reduced NO x -0;3 loss; and 2) a general decrease in Oe D), by as much as 10%, from ~1970-2100 throughout most of CO2 +0.016 (23) +0.014 (23) +0.0077 (34) +0.0041 (33) the stratosphere which results in less NO x production via the 570 reaction N 2 0+0 e D ).As a result of these factors, our N 20 CO 2 impacts being roughly equal and opposite.The CH4595 impact was also substantially larger than either N 2 0 or C02 during 1940-1960, reflecting the large relative changes in the CH 4 boundary condition during this time.However at the end of the 21 st century, the CO 2 impact is projected to be the largest, and will be more than twice the magnitude of the 600 N 2 0 impact.Table 3 also shows the sensitivities as percent changes to the background levels of each GHG, required to produce a 1 % ozone change (listed in parentheses).For N 2 0, a ~ 15% change in the background is required to cause a 1 % change in ozone, compared to a ~25-35% change for CO 2 , and ~20-count for changes in tropospheric NO x emissions (the model uses fixed present day NOy specified from the GMI simulations).These likely undergo significant past and future changes [IPCC, 2007], with less (more) NO x leading to less (more) tropospheric ozone production and a smaller (larger) negative CH 4 ODP than we report in Table 2.
For CO 2 , the strong time dependence in Tables 1-3 is mostly due to the changing background concentration, which increases by a factor of2.5 from 1850-2100 (Figure I).However, the percentage change relative to the background in Table 3 is somewhat larger in 2000 and 2100 than in 1850 and 1950.This is partly due to the fact that CO 2 cooling and the corresponding increase in upper stratospheric ozone become less efficient at higher CO 2 levels, i.e., a saturation effect [e.g., Ramaswamy et a!.,200Ia].There are other higher order effects that influence this CO 2 time dependence, including the future changing PSC concentrations -due to stratospheric cooling and increasing water vapor -coupled with the decreasing halogen loading.As discussed in previous studies, ODS loading and the corresponding reduction in ozone heating have a large impact on 695 the temperature changes in the recent past, maximizing in the lower and upper stratosphere as seen in the vertical profiles in Figure II (top).By 2100, CO 2 cooling dominates the total temperature change throughout the stratosphere, and is significantly larger than the warming caused by the reduced 700 ODS loading (Figure 11, bottom).
The temperature impact of N 2 0 caused by the associated NOx-ozone depletion maximizes at 30-40 km but is relatively small (-0.05K/decade in Figure 11), corresponding to a cooling of -0.7K from 1850-2100 at 40 km (green curve in 705 Figure 10, middle).
Increases in CH 4 induce temperature changes via several mechanisms.Below ~ 18 km, there is a slight warming (Figure II) caused by the combination of the direct infrared radiative effect of CH 4 and the indirect radiative effect of the 710 resulting ozone increases in the troposphere and lower stratosphere [e.g, Portmann and Solomon, 2007].In the stratosphere, increases in CH 4 lead to warming via the increases in ozone caused by the reduction of active chlorine by the the reaction CH 4 + CI-+ HCI + CRl.Increases in CH 4 also 715 lead to cooling in the stratosphere and mesosphere via the increases in water vapor.In the global average, the combination of these processes results in a net cooling above ~20 km in Figures 10-11.
In the lower mesosphere, the cooling effect of the CH 4 -720 induced H 2 0 increases becomes large and is comparable to the CO 2 cooling during 1850-2000 at 60 km (Figure 10,top).By 2055, the CH 4 -induced cooling is roughly half that due to the CO2 changes at 60 km in Figure 10.We note also that the future CH 4 -induced temperature changes in Figure 725  depletion impact the stratospheric circulation and age of air [e.g., Butchart and Scaife, 2001;Austin and Li, 2006;Butchart et aI., 2006;Kodama et aI., 2007;Garcia and Randel, 2008;Li et aI., 2008;Oman et a!., 2009].Although observations show little or no age trend over the past 30 years 740 in the NH above 24 km [Engel et aI., 2009], simulation of a general decrease in age through the 21"t century appears to be robust among coupled chemistry-climate models [SPARC CCMVal,201O].
Here we briefly examine how the individual source gas perturbations used in this study impact the model age of air simulations.This is summarized in Figure 12 (top panel), which shows time series of global and annually averaged age of air at 25 km from the various model simulations as indicated.These results are representative of altitudes above ~20 km.There is significant interannual variability in the GEOSCCM time series which is comprised of three simulations covering the periods 1950-2004,1971-2052, and 1996-2100.However, the base 2D and GEOSCCM simulations have similar rates of decrease in mean age over the 1960-2100 time period.
As with ozone and temperature, the 2D-simulated changes in mean age are relatively small prior to 1950, with CO 2 loading accounting for about half of the total decrease of 0.1 years from 1860-1950.The base simulated age decreases much more rapidly after ~1970 when the impacts of CO 2 and ODS loading become large and act in the same direction.The total age decrease is 0.3 years from 1980-2005, with 65% of this change due to ODS loading, 25% due to CO 2 , and 10% caused by N 2 0 and CH 4 changes.This dominance of the ODS loading is consistent with the GEOSCCM analysis of Oman et aI., [2009].Figure 12 (top) also shows a 2D-ODS only simulation with all heterogeneous chemical processes tumed off, i.e., no ozone hole is simulated (blue dashed-dotted line).This simulation shows only small mean age change due to gas-phase chlorine-ozone destruction, and suggests that most of the ODS effect on the age of air change is due to the ozone hole.Separation of these ODS impacts on mean age is currently being investigated in the GEOSCCM (L.Oman, personal communication).
In the future, the base simulation age decreases less rapidly as the effects of the reduced ODS loading partially offset the CO 2 impacts, with the latter becoming the dominant mechanism after about 2025.By 2100, C02 loading accounts for ~ 75% of the total age change (~1 year) from 1860.N 2 0, CH4, and the remaining effects of ODS loading have secondary impacts by 2100, with mean age decreases of 0.12, 0.05, and 0.08 years (12%,5%, and 8% of the total) respectively, from 1860-2100.
Previous studies have attributed the BDC acceleration to increased stratospheric wave driving resulting from changes in the zonal mean winds.These are ultimately due to changes in the temperature distribution induced by SST changes, GHG loading, and polar ozone depletion [e.g., Olsen et a!., 2007;Kodama et al., 2007;Garcia and Randel, 2008;Oman et ai., 20091.The 2D model formulation uses parameterizations to account for planetary and gravity wave effects (section AI), and the latitude-height patterns of the long term changes in zonal mean temperature, zonal wind, waveinduced acceleration, lower stratospheric tropical upwelling, and age of air are generally similar to those in WACCM3 and GEOSCCM rGarcia and Randel, 2008;Oman et aI., 20091  Time series for 1860-2100 from the base simulations (all source gases varied time dependently) of the GEOSCCM (black dotted 780 lines) and 2D model (black solid lines).Also shown are 2D simulations in which only certain source gases are varied time dependently as indicated.The blue dashed-dotted line is a 2D-ODS only simulation with all heterogeneous chemical reactions turned off so that no ozone hole is simulated.The right hand axis shows the change 785 relative to 1860.The GEOSCCM time series is comprised of three simulations, which use somewhat different SSTs, for the lime periods 1950-2004, 1971-2052, and 1996-2100 (the first 8-10 years of each simulation have been removed to allow for spin-up).For visual clarity and to account for a systematic offset between the models, we have added .Therefore, it appears that the same basic characteristics of the 795 CCM-simulated BDC and age of air changes are also present in the 2D model.
The C02 perturbation shown in Figure 12 (top) modifies the model temperature field via the direct IR effect, and through changes in surface temperature, latent heating,BOo The impact in the ODS-only simulation (blue solid curve) is driven mainly by the strong cooling trend during 1970-2000 associated with the ozone hole (Figure B2), which induces a positive trend in zonal wind (increasing westerlies) at high SH latitudes.The accompanying trends (enhancements) in the planetary wave driving and BDC acceleration decrease the age of air.For the ODS-only simulation without the ozone hole (blue dashed-dotted curve), the ozone loss and cooling is confined to the upper stratosphere globally, with only small positive trends in zonal wind and wave driving at midlatitudes of both hemispheres, reSUlting in small BDC acceleration and age of air changes in Figure 12.

Photochemical Lifetimes
Given the significance of the long term stratospheric changes caused by GHG and ODS loading, it is useful to examine how these changes impact the modeled photochemical lifetimes of certain compounds.The lifetime is important in determining the length of time over which a molecule of a substance will have a significant impact on ozone depletion or global warming, and in deriving surface mixing ratio boundary conditions from emissions estimates for use in atmospheric models [Kaye et aI., 19941.The lifetime is computed  as the atmospheric burdcn (total number of molecules) divided by the loss rate, both of which are vertically integrated 820 and globally/annually averaged.We have recently shown the impact of new photolysis cross sections on the model-805 computed lifetime of CCl4 [Rontu Carlon et aI., 20lOl Here we examine the time dependence of the lifetimes of various compounds in more detail.Figure 13 (black lines) shows the modeled lifetimes of N 2 0, CFC-li (CFCla), CFC-12 (CF 2 Ch), and CCl 4 from 810 the base 2D simulation.We restrict our analysis here to the 1960-2100 time period since the CFC lifetimes are not well defined prior to 1960 given that emissions began in the late 830 1930s-1940s.While this is not a problem for N 2 0 given it's significant natural source, the computed N 2 0 lifetime decrease from 1860-1960 is small (143-141 consistent with the small age of air decrease shown in Figure 12. Lifetime change due to loss rates 0 4 4  (Top) The difference between the base and fixed loss rate simulations (black lines minus red lines in Figure 13), illustrating the lifetime changes due to the changing loss rates.(Bottom) The difference between the fixed loss rate and fixed chemistry and transport simulations (red lines minus green lines in Figure 13), showing the lifetime changes due to changes in transport.
The present day lifetimes shown in Figure 13 (132, 61, 108, and 51 years, respectively, for N 2 0, CFC-ll, CFC-12, and CCI 4 ), are older than those cited in WMO [2007,20111 and IPCC [20071: 114,45, 100, and 26 years.For CCI 4 , the older lifetimes shown here do not include soil or ocean loss processes.Updated lifetimes for CFC-II and CFC-12 computed from various models (including the 2D model) were presented in Douglass et al. [2008], That study illustrated a strong dependence oflifetime on the modeled circulation and age of air, and showed that models with realistic age of air simulated a relationship between mean age and the fractional release of CFC-II and CFC-12 that compared well with observations.We note that the present 2D model compares well with the age of air derived from observations in Figure B4.
Figure 13 (green lines) also shows a simulation in which the time dependent loading of N 2 0, CFC-ll, CFC-12, and CCl 4 impacts only the atmospheric burden used to compute the lifetimes, while the model transport and chemistry remain fixed at seasonally repeating 1960 values.This illustrates thesso effect of the changing burden on the computed lifetimes.As the surface mixing ratio of a substance increases with time, an increasing fraction of the total atmospheric burden resides in the stratosphere where it is lost, as opposed to the troposphere where the loss is zero, and this results in a decrease in the computed lifetime, Le., the lifetime has not reached equilibrium with the increasing surface boundary conditions [e.g., see Kaye et aI., 1994J.For the CFCs, emissions began in the late 1930s-1940s and ramped up quickly during 895 the following decades [Butler et at., 1999], so that there is a large impact of the increasing burden on the lifetimes prior to ~ 1990.Atmospherie loading of CCl 4 began ~ 1900 so that the influence of the increasing burden on the lifetime is small by 1960.This effect on the lifetimes of the CFCs and CCI 4 900 then levels off as the burdens slowly decrease after 2000, i.e., the lifetime has reaehed equilibrium with the boundary conditions.N 2 0 has a significant natural souree with a slow increase throughout 1960-2100 due to anthropogenic activity (Figure 1), so that the changing burden has little or no 905 impact on the computed lifetime throughout the time period.
The lifetimes are also controlled by the loss rates and the rate of transport of a species through the stratospheric loss region.The loss rates (photolysis and reaction with Oe D» are impacted by the overhead burden of ozone, and the trans-910 port rates are modified via changes in the BDC as discussed previously.Both of these processes incur long term changes via the ODS and GHG loading. The red curves in Figure 13 are the same as the base simulation except with the loss rates of N 2 0, CFC-ll, 915 and CCl 4 fixed at seasonally repeating 1960 values.The effect of the changing loss rates is then isolated by taking the difference between the base and fixed loss rate simulations (black lines minus red lines), i.e., the residual is that due to only the changing loss rates.The results are shown relative to 920 1960 in Figure 14 (top) and reveal small impacts for all four species.These differences approximately follow the changing overhead burden of ozone, with lifetime decreases of ",,2 years (Le., loss rate increase) for all species from 1960-2000 as ozone decreases due to ODS loading.As future ozone in-925 creases with redueed ODS and increased C02, the impact of the loss rates during 2000-2100 increases the lifetimes by 6 years for N 2 0, 4 years for CFC-12, and 2 years each for CFC-11 and CCI4.These changes are all <4% over the 1960-2100 time period.This small impact of the changing 930 loss rates was also reported in Douglass et al. [20081.The effect of the changing BDC is isolated by taking the difference between the fixed loss rate and fixed chemistry and transport simulations (red lines minus green lines), i.e., the residual is that due to only the changing transport.These 935 results are shown in the bottom panel of Figure 14, again relative to 1960.The lifetime decreases are largest during 1960-2000 when the BDC acceleration is fastest (Figure 12), with less pronounced decreases aftcr 2000 as the BDC accelerates more slowly.The net lifetime reductions for 1960-2100 due to the changing BDC are all 11-13%: 16 years for N 2 0, II years for CFC-11, 21 years for CFC-12, and 7 years for CCI 4 •

Summary and Conclusions
In this paper we use an updated version of our GSFC 2D coupled model to study long term stratospheric ehanges caused by souree gas loading for the 250-year time period, 1850-2100.We use numerous sensitivity simulations in which the ODSs, CO 2 , CH 4 , and N 2 0 are varied individually to separate the relative roles of these gases in driving long term changes in ozone, temperature, and age of air.We also compare the 2D model with global ozone and temperature changes observed during the recent past, and with simulations from the GEOSCCM to illustrate that the updated 2D model captures the basie processes that drive long term stratospheric changes.A detailed description and evaluation of the model are provided in the Appendicies.
For long term changes in ozone, the impacts due to ODS, CO 2 , CH 4 , and N 2 0 loading all play important roles in different regions of the atmosphere.GHG loading becomes more important in determining future ozone concentrations as the ODS loading diminishes through the 2Pt century.CO 2 cooling increases upper stratospheric ozone via reduction in the temperature dependent loss rates.CO 2 loading also leads to acceleration of the BDC which redistributes ozone in the lower stratosphere, with less (more) ozone in the tropics (extratropics).The net CO 2 impact on total ozone is a decrease in the tropics and an increase in the global average and at midlatitudes, with a larger enhancement in the NH.For 1850-2100, CO 2 loading has the largest individual impact on global ozone in the upper stratosphere and on the total column, causing increases of 20% and 4.2%, respectively.N 2 0 loading and the subsequent increase in the NO x ozone loss maximizes near 35 km (""O.5%/deeade), with a global total column decrease of -2.7% from 1850-2100.
Methane loading impacts ozone via several processes, and we illustrate the sensitivity of these processes in determining the net result on ozone.The simulations, which use fixed tropospheric NO y , reveal that CH 4 loading leads to a net increase in total ozone at all latitudes throughout 1850-2100.Global total ozone increases by 2-2.5% from 1850 to the latter half of the 2Ft eentury due to CH 4 loading.However when coupled with time dependent ODS ehanges, the methane impact on ozone for present day eonditions is nearly a factor of two larger (+6 DU) compared to the +3 DU response calculated with very low (1850) levels of Cl y .
In the lower mesosphere, the CH 4 -induced HOx-ozone loss dominates the ozone time dependence.Our simulations reveal that this process resulted in significant ozone reductions prior to 1960, so that one-third of the total decrease in ozone during 1850-2050 occurred by 1960.E. L. Fleming et al.: Impact of source gas changes on the stratosphere In the stratosphere prior to 1960, the simulated ozone changes are relatively small, but are not insignificant.ODS emissions, primarily due to carbon tetrachloride, cause 1% depletion in global total ozone from 1860-1960, with the vast majority of this occuring after ~ 1920.From 1860-1960, in..,ooo creasing CO 2 loading results in ozone increases of 2% in the upper stratosphere and 0.5% in the total eolumn, with CH 4 and N 2 0 causing total ozone changes of +0.75% and -0.5% respectively, during this period.The net result is a broad ozone maximum during the 1920s-1930s, as the CO 2 ..,005 and CH 4 -induced ozone increases outpace the ozone losses caused by N 2 0 and CCl 4 emissions.This preceeds the decline in ozone driven mainly by ODS loading, which becomes much more rapid after ~ 1970.
We also present a quantitative analysis of the steady statc,OlO impaets of GHGs on ozone for 1850, 1950,2000, and 2100 atmospheric conditions.These calculations have a significant dependence on the background conditions, reflecting saturation effects as well as the chemical interactions of the perturbations.Using the ozone depletion potential (ODP) concept for GHGs, we compute present day ODP values of +0.019 (N 2 0), -0.00065 (CH 4 ), and -0.0039 (C0 2 ), relative to CFC-II.To more easily interpret the relative impacts of thC'o15 GHGs on ozone, we also derive steady state sensitivity factors in terms of the mixing ratio change of each GHG.These calculations reveal that based on the AlB scenario, CH 4 had the largest impact on ozone for year 2000 conditions, owing to it's strong coupling with Cly.However, CO 2 is projectedo20 to have the largest impact by 2100, and will have twice the magnitude of the N 2 0 impact.
The simulated changes in stratospheric temperature for 1850-2100 are mostly controlled by CO 2 cooling, with the reduced ozone heating caused by ODS loading also playinglO25 an important role from ~ 1980 through the first half of the 975 21 st century.The impact of CH 4 and N 2 0 loading are relatively small below ~45 km.However, the cooling due to CH 4 -induced H 2 0 inereases becomes significant above ~50 km, and the resultant temperature ehanges are comparable to030 those induced by the ODS and CO 2 loading.

980
In the 2D model, the long term changes in surface temperature, latent heating, and tropospheric H 2 0 are parameterized based on the GEOSCCM simulations and the CO 2 loading.These proeesses are likely due, at least in part, to thC'o35 response of the GEOSCCM hydrological cycle to long term 985 SST changes.Parameterization of these processes enables the 2D model to simulate long term changes in the BDC and age of air which are consistent with the GEOSCCM.Changes in GHG and ODS loading ultimately impact the BDC and agelO4o of air through changes in the radiative and temperature dis-990 tributions.We estimate that changes in ODS concentrations account for 65% of the mean age decrease from 1980-2005, with 25% due to CO 2 loading, and 10% due to CH 4 and N20.Our simulations also indicate that the BDC and ageD45 changes caused by ODS loading are mainly due to the forma-995 tion of the ozone hole.CO 2 loading becomes the dominant source of mean age change after ~2025, and explains ~ 75% of the total age reduction (I year)from 1860-2100, with 12% and 5% explained by N 2 0 and CH 4 loading, respectively.
We also examined the time dependent photochemical lifetimes of N 2 0, CFCI 3 , CF 2 CI 2 , and CCI 4 , and found that the impact of the BDC acceleration is significant, and causes the lifetimes to decrease by 11-13% from 1960-2100.The impact of the changing loss rates is generally small and follows the time dependent changes in stratospheric ozone via the changes in photolysis and Oe D).This effect decreases the lifetimes by ~2 years from 1960-2000, and increases the lifetimes by 2-6 years (3-4%) from 2000-2100.The model calculations also allow us to separate these geophysical impacts on the lifetimes from the artifacts caused by the disequilibrium between the rapidly increasing atmospheric burden and the stratospheric loss.

Appendix A GSFC Coupled 2D Model
The GSFC 2D coupled chemistry-radiation-dynamics model was originally discussed in Bacmeister et aI., [1995] and has been frequently used in stratospheric assessments [WMO, 2007], and studies pertaining to the chemistry-climate coupling of the middle atmosphere [e.g., Rosenfield et aI., 1997;Rosenfield et ai., 2002].While this model was not included in the recent CCMVal activity [SPARC CCMVal, 2010], several of the model components are very similar to those used in the GEOSCCM which was evaluated in CCMVal.These components include: the infrared radiative transfer scheme [Chou et aI., 2001]; the photolytic calculations [Anderson and Lloyd, 1990;Jackman et ai., 1996]; and the microphysical model for PSC formation [Considine et ai., 1994].As discussed in Fleming et al. [2007], the model now uses an upgraded chemistry solver that computes a full diurnal cycle for 35 fast chemical constituents.This scheme was shown to be in good agreement with photochemical steady state box model calculations [Park et ai., 1999].The latest Jet Propulsion Laboratory (JPL) recommendations are used for the photolytic cross sections and reaction rate constants rSander et ai., 2006].
The model domain extends from the ground to approximately 92 km.The chemistry calculations are done on a grid resolution of 4° latitude by I km altitude.We have found that for most applications, the model radiation and dynamics calculations can be adequately done on a somewhat coarser grid of ~ 4.9 0 latitude by 2 km altitude.Using a finer resolution only adds to the computational burden but does not improve the model dynamical simulations.
We have recently made extensive upgrades to the model solar radiation and dynamical modules, which are described in the following sections.An evaluation of the model temperature and transport are then provided in Appendix B.

Al Wave Parameterizations 110e
The horizontal mixing (Kyy) and momentum deposition due to dissipating planetary waves is computed using a linearized parameterization similar to that described by Garcia [19911.Previously, the parameterization used a surface boundarY1105 condition based on a representation of the topographic forcing of planetary waves.It was necessary to include adjustable amplitude efficiency factors for each wave number to obtain reasonable seasonal variations of the zonal winds and chemical fields [Rosenfield et aI., 1997].In the present model, this lower boundary condition is based on a geopotential height climatology for 750 mbar as a function of latitude and sea-I1 ,o son, derived from the National Centers for Environmental Prediction (NCEP) reanalysis-2 data [Kistler et aI., 200 I] averaged over 1979[Kistler et aI., 200 I] averaged over -2007. .This provides a more complete representation of the lower boundary forcing of planetary waves, e.g., land-ocean contrasts, in addition to the topo-,"S graphic forcing.We solve for planetary zonal wave numbers 1-4.Compared to that obtained previously, the new methodology provides more realistic model simulations of planetary wave drag and mixing with no artificial wave amplitude adjustment factors necessary.1120 The model also includes the off-diagonal eddy diffusion of constituents (K yz ), following the methodology used in our GSFC 2D fixed transport model.This follows the assumption that horizontal eddy mixing is directed along the zonal mean isentropes, and projects the Kyy mixing rates onto isentropic 1125 surfaces [Plumb and Mahlman, 1987;Newman et aI., 1988].
The momentum deposition due to gravity wave breaking in the mesosphere is parameterized using a ray tracing calculation for waves with non-zero phase speeds, and a cubic drag law for zero-phase speed mountain waves [Bacmeister et al."30 1995].This enables the gravity wave momentum flux to be interactive with the evolution of the zonal mean flow.However to obtain proper tracer simulations, we found it necessary to specify the model vertical eddy diffusion rates (K zz ), which are taken from the GSFC 2D fixed transport mode\'35 as a function of latitude, height, and season [Fleming et aI., 2007].For the upper stratosphere and mesosphere, these are based on the gravity wave parameterization originally developed by Lindzen [1981] and modified by Holton and Zhu [1984].For the troposphere and lower stratosphere, K zz is based on the zonal mean temperature lapse rate as computed from a multi-year average of NCEP reanalysis-2 data.1140

A2 Radiative Transfer
For the absorption of solar radiation in the ultraviolet and visible, we now compute the heating rates consistently with the.'45 model incident solar flux and photolysis calculations [Jackman et a1., 1996].These are computed over a full diurnal cycle with a much finer spectral resolution compared to the broad band calculations used previously [Strobel, 1978;Chou and Suarez, 1999].These heating rate calculationS115o have the further advantage of utilizing the most current recommendations for the photolytic cross sections [Sander et aI., 2006].This new methodology also gives model temperatures that are in somewhat better agreement with observations.For ozone, the heating rate in this way is computed by [e.g, Brasseur and Solomon, 1986  where p is the total atmospheric density, Cp is the specific heat of dry air at constant pressure, [03] is the ozone number density, ).. is the wavelength, CY( 03) is the ozone absorption cross section, and F.,,;' is the incident solar flux (enhanced or reduced) as a function of wavelength at each model grid point.In addition to ozone heating, we also include the heating due to absorption by O 2 , which is important in the mesosphere, and N0 2 which is of secondary importance in the middle stratosphere.For these calculations, we assume that all of the solar radiation absorbed is immediately realized as thermal energy.This is a good approximation below ~80 km, i.e., the region of interest for the current study, where the chemical recombination of Oep) is very fast [e,g, Brasseur and Solomon, 1986].
We also include the minor absorption of solar radiation by ozone in the infrared, and by water vapor in the infrared and visible based on the parameterization of Chou and Suarez fI999].For these calculations, we use surface reflectivity values as a function of latitude and season based on the TOMS climatology compiled by Herman and Celarier, [1997].
For the thermal infrared radiative transfer, we use the parameterization of Chou et al. f200 1], which is the same as that used in the GEOSCCM.This includes the contributions due to 0 3 , CO 2 , H 2 0, CH 4 , N 2 0, CFC-l1, CFC-12, and HCFC-22.For both the solar and thermal IR calculations, the 2D model includes zonally averaged cloud parameters based on a multi-year average of output from the Whole Atmosphere Community Climate Model version 3 (WACCM3) [e.g., Garcia et aI., 2007].

A3 Model Treatment of Longitudinal Variations
Previously, the model-generated zonal mean temperatures were used to compute the gas phase and heterogeneous reaction rates.In the new model version, reaction rates are now computed using a longitudinal temperature probability distribution which is generated from the model-computed planetary wave fields for zonal waves 1-4 (section AI).The rates for the reactions at each model grid point are computed once per day by summing the rates computed for each temperature in the distribution weighted by the probability of occurrence of that temperature.Using the temperature probability distribution instead of the zonal mean temperature is especially important for the heterogeneous chemical reactions as these can have significant non-linearities in temperature.
For the calculation of polar stratospheric cloud (PSC) formation, we utilize the parameterization described in Considine et al. [19941, using longitudinal temperature probability distributions derived from the NCEP reanalysis-2 data averaged over 1979-2006.This climatological average distribution is used for all years in the simulations.This methodology does not allow for the interaction between PSC formation and the chemical/dynamical time evolution of the model stratosphere.However given the strong temperature non-linearity ofPSC formation, we found it necessary to use the observed temperature distributions rather than the model temperatures to properly simulate PSC formation.

A4 Tropospheric Parameterizations
Accounting for the hydrological cycle and surface boundary layer processes is important to properly simulate the dynamical and chemical distributions of the troposphere and lower stratosphere.Since the 20 model framework is inadequate to simulate most tropospheric processes interactively, we specify the surface temperature, tropospheric water vapor and lama tent heating.As described in the following, we first generate monthly and zonal mean c1imatologies of these parameters, and then add on long term changes parameterized in terms of the atmospheric CO 2 loading.
The surface temperature seasonal cycle as a function of latitude is based on the NCEP reanalysis-2 data averaged over 1979-2006.Tropospheric latent heating as a function oflatitude, height, and season is based on a multi-year average of output from WACCM3 simulations.The model water vapor seasonal cycle in the upper troposphere (12-16 km) is based on the UARS reference atmosphere (UARSRA) compiled by Randel et aI., [2001].Below 12 km, water vapor is derived from a 21 year average  of relative humidity data from the European Center for Medium Range Weather Forecasts (ECMWF) updated reanalyses (ERA-40).Values from the UARSRA and ERA-40 data sets are functions of latitude, height, and season and are blended over several pressure levels to obtain a smooth transition in the vertical.Water vapor everywhere above the tropopause is computed in the 20 model (see Figure B6).
In addition to the seasonal variations, the surface temperature, tropospheric water vapor and latent heating undergo substantial long term changes as simulated by the GEOSCCM.This is illustrated in Figure Al which showS!205 zonally averaged time series of GEOSCCM simulations for 1950-2100 (black curves) at the locations indicated.These time series have been deseasonalized and smoothed to reduce the interannual variability of the GEOSCCM.These long term changes are highly correlated with the time de"'210 pendent surface boundary condition of CO 2 (bottom panel) 12DO and are likely a response to the warming of the troposphere and sea surface temperatures caused by the increased atmospheric C02 loading.The past and future temperature changes from the 2D model and GEOSCCM are shown in Figure B2.The 2D model captures most of the latitude-height variations simulated in the GEOSCCM.The main discrepancies occur at high latitudes where the 2D model somewhat underestimates the large temperature changes simulated by the GEOSCCM associated with the ozone hole, i.e., past cooling and future warming.Also, the GEOSCCM simulates a mid-upper stratospheric warming at high SH latitudes for 1960-2000, which was shown to be a dynamical response to the ozone hole [Stolarski et aI., 20061.The 2D model simulates this feature only very weakly (top right).These discrepancies are likely due to the 2D model not fully resolving the large zonal asymmetries characteristic of the polar region, as well as the known high ozone bias at high latitudes in the GEOSCCM [ Pawson et aI., 20081.The corresponding global average vertical profiles (Figure B3) also show good agreement between the 2D model and GEOSCCM as well as radiosonde data for 1960-2000 from the Radiosonde Atmospheric Temperature Products for Assessing Climate (RATPAC-A) [Free et aI., 2005J.We note that the global average GEOSCCM stratospheric temperature trends were also found to be in reasonably good agree-127o ment with those derived from SSU and MSU satellite data for 1979-1999 [Stolarski et aI., 20101.In the troposphere, the 2D model simulates warming throughout 1960-2100, which is due mainly to the parameterized long term changes in surface temperature and latenb275  Fig. B4.Age of air at 20 km derived from ER-2 aircraft measurements of SF6 (asterisks) and CO2 (triangles), and vertical profiles of the age of air derived from balloon measurements of SF 6 (asterisks, plus signs) and CO2 (triangles) at the latitudes indicated.Ages derived from these measurements have been adapted from Hall et al. [1999].Also shown are simulations from the 2D model (red line).

Global Temperature Change
The age is taken relative to the tropical tropopause.

B2 Age of Air
Stratospheric age of air is a widely used diagnostic that tests the overall fidelity of model transport [e.g., Hall et aI., 1999].
Figure B4 shows the mean age of air at 20 km derived from aircraft measurements of SFe (asterisks) and CO 2 (triangles), and a series of vertical profile measurements of SF 6 and CO 2 made from balloon flights in three latitudes zones [e.g., see Hall et aI., 1999].We note that differences in the observations at the middle and higher latitudes may reflect photo-chemical influences on SF 6 which would cause an overestimation in the inferred ages [Hall and Waugh, 1998].Some of the older age measurements at 65°N may also reflect remnants of the winter polar vortex [Ray et aI., 19991-Figure B4 also shows the age of air derived from the 2D model simulation averaged over the 1990s (red line).The age of air in the model is computed from a "clock" tracer ,....... .:5-

Q)
.. that has a surface boundary condition linearly increasing with time, with no other chemical production or loss.This is essentially identical to the age obtained from simulations of SF G or CO 2 as done, for example, in Hall et al. [19991.1300The model somewhat underestimates the observed age in the mid-latitude Northern Hemisphere (NH) (see the 20 km and 400N panels), and at the high latitudes of the NH above 30 km.However for the most part, the model simulates the absolute values and the latitudinal and vertical gradients of thCl305 observations fairly well.This illustrates that the model transport rates in the stratosphere, i.e., the relative magnitudes of vertical motion and horizontal mixing, are generally realistic.
'-" Q) ...  tom).The top panel includes data from AURA/MLS in the polar regions (averaged over 2004-2009) where HALOE lacks data coverage.Contour interval is 0.2 ppmv.The red dashed line separates the regions where the model H2 0 is computed (above) and prescribed to the HALOE climatology (below).

B6
). Figure B6 includes data from AURAIMLS in the polar regions (averaged over 2004-2009) where HALOE lacks data coverage.The model shows good overall agreement with the data in reproducing transport sensitive features in the meridional plane, including the horizontal and vertical gradients.
For example. the model qualitatively simulates the region of strong horizontal mixing during late winter/early spring at midlatitudes of both hemispheres.This is especially pronounced in the SH during September at 20-40 km in the N 20 field.The model tends to underestimate this mixing in the 1310 SH mid-upper stratosphere, as is also seen in the midlatitude vertical profile in Figure B7 (top).Here the model compares well with the MLS N 2 0 below ~27 km, but underestimates the data above this level, which is due to weaker than observed mixing of high-N 2 0 air from lower latitudes.seent within the vortex occurring throughout the winter, witht335 little influence of air in-mixing from midlatitudes [SPARC CCMVal, 20101.The good model-data agreement here illustrates that the magnitudes of vortex deseent and isolation from midlatitudes in the simulation are generally realistic.
The HALOE H 2 0 data in Figure B6 indicate strong pole-'34o ward and downward transport of very dry air from the tropics to midlatitudes just above the tropopause [e.g., Randel et aI.,200I].The model H 2 0 is set to the HALOE climatology in the upper troposphere, below the red dashed line in Fig- ure B6.Above this level, the H 2 0 field is computed in thc,345 model, and reveals that the model transport is resolving fairly well this strong poleward and downward transport from the tropical tropopause region.
In Figures B5 and B6, the model also resolves the isolation of the tropics in the lower stratosphere, as indicated strong horizontal gradients in the subtropics, and a region of low water vapor concentrations at 20-30 km over the equator  associated with the "tape recorder" signal [e.g., Mote et aI., 19961.This feature reflects the slow upward propagation of the water vapor seasonal cycle from the tropical tropopause, and simulation of this feature provides a good diagnostie of model transport.
Figure B8 shows the amplitude variation and phase lag versus altitude of the seasonal cycle in H 2 0 + 2CH 4 at the equator relative to the tropopause from HALOE data (black asterisks).This quantity is quasi-conserved and accounts for both the H20 seasonal cyeIe propagation and the slow photoehemical conversion of CH4 into H 2 0 in the stratosphere.The amplitude attenuation and phase lag with increasing height reflect the strength in the upwelling of the Brewer-Dobson circulation (BDC) combined with the rate of vertical diffusion and entrainment of air from mid-latitudes [Hall et al., 19991. The model (red line) shows an increasingly longer phase lag compared with the data above ~27 km, possibly reflecting weaker BDC upwelling in the trop-" a..
.:5- 1370 ical middle stratosphere than indicated in the observations.However overall, the model shows mostly good agreement 1355 with the HALOE data in simulating this seasonal cycle propagation.This, combined with the good agreement in the tropical age profile (Figure B4), suggests that the model transport ,375 rates in the tropical lower-middle stratosphere appear to be fairly realistic.The simulation compares relatively well with the data in most regards.The model qualitatively reproduces the observed latitudinal and vertical gradients in most places, as well as the magnitude of the ozone amounts.There are some regions of small discrepancy; for example at 15-20 km in the SH midlatitudes where the model slightly overestimates ozone, and in the tropics above 30 km associated with the ozone deficit region where the model underestimates ozone (e.g., Jackman et aI., 19961.The model also underestimates ozone at mid-high NH latitudes near 10-15 km which is likely due to excessive horizontal mixing in this region, i.e., in-mixing of low-ozone air from the tropical troposphere.However, these model-measurement differences are now nificantly smaller compared with previous model versions. Fig. I. Time dependent surface boundary conditions for CO 2 , CH4, and NzO from Hansen and Sato 12004] for 1850-1950 and the IPCC GHG scenario AlB for 1950-2100.The bottom panel shows the upper stratospheric Equivalent Effective Stratospheric Chlorine (EESC) taken as the global average of Cly +60Bry at 50 km.See text for details.
. The corresponding GHG surface boundary conditions for 1850-2100 are shown in Figure I, along with the Equivalent Effective Stratospheric Chlorine (EESC, bottom panel) representing the time dependent concentration of halogen loading from the ODS source gases.
Fig. 2. (Top) Vertical profiles of the annual and near-global average (60 0 S-600N) ozone trend (%/decade) for 1996-1979 derived from the SBUV data and model simulations.Shown are the base simulations (all source gases varied time dependently) of the GEOSCCM (black dashed lines) and 2D model (black solid lines), along with 2D simulations in which only certain source gases are varied as follows: ODSs only (blue lines); CO 2 only (red lines); CH 4 only (orange lines); N2 0 only (green lines).The trends are derived from regression fits to the EESC time series (Figure I) for 1979-2004.(Bottom) As in the top but for the 2095-2005 ozone difference (%/decade) using lO-year averages centered on 2095 and 2005 to reduce the effects of interannual dynamical variability in the GEOSCCM.The change due to CH4 is shown for 2095-2005 (orange solid) and 2055-2005 (orange dashed-dotted).

Fig. 3 .
Fig. 3. Annually averaged ozone trend for 1996-1979 derived from the GEOSCCM and 20 model simulations.The top panels are from the base simulations in which all source gases are varied time dependently.The middle and bottom panels show 20 simulations in which only the OOSs or GHGs are varied time dependently as indicated.The trends are derived from regression fits to the EESC time series (Figure l) for 1979-2004.The GEOSCCM trends are not computed in the troposphere as ozone is relaxed to a climatology in this region.The contour intervals are ±2 %/decade and include the ±.5 and ±1 %/decade contours.

Fig. 4 .
Fig. 4. As in Figure 3, except for the 2095-2005 ozone difference using lO-year averages centered on 2095 and 2005 to reduce the effects of interannual dynamical variability in the GEOSCCM.The contour intervals are ±2 %/decade and include the ±.5 and ±1 %/decade contours.

Fig. 5 .
Fig. 5. Near global (60 0 S-600N) annually averaged ozone time series for 1860-2100, relative to 1860, at 22, 40, and 60 km.Values are in ppmv change (left axes) and % change (right axes).Shown are the base simulations (all source gases varied time dependently) of the GEOSCCM (black dotted lines) and 20 model (black solid lines), along with 20 simulations in which only certain source gases are varied as indicated.The GEOSCCM time series has been adjusted to match the 20 base simulation for 1960.Also shown are the BUV/SBUV satellite observations for1970-2009 (excluding 1973- 1978)  at 22 and 40 km ("+" symbols).To emphasize the model-data comparison after 1970, the data have been adjusted so that the 1970-1972 average matches that of the base simulations.

Fig. 7 .
Fig. 7.As in Figure 6 except for tropical total ozone (10° S-lOoN average), and without the 2D-CH4 only(a) and 2D-ODSs only(b) curves.For visual clarity, the observations are not shown as these are dominated by the large quasi-biennial oscillation, a feature not included in the simulations.Values are in Dobson Unit (DU) change (left axes) and % change (right axes).

Fig. 8 .
Fig. 8. Ozone time series for 1860-1960, relative to 1860 values, from 20 model simulations for the 60 o S-60oN average at 40 km (top), and the 90 o S-90oN average total column (bottom).Shown are the base simulation (all source gases varied time dependently) and simulations in which only certain source gases are varied time dependently as indicated, with the other source gases fixed at 1860 levels.Values are in Dobson Unit (OU) change (left axes) and % change (right axes).

Fig. 9 .
Fig. 9. Vertical profiles of the annual and global averaged steady state ozone change (DUlkm) due to a 0.5 ppmv CH4 perturbation for year 2000 (top) and year 2100 (bottom).The black curves show simulations with all CH4 reactions (reactions with OeD), OR, and Cl) using perturbed CH4.The green curves show simulations using perturbed CH4 for reactions with OeD) and OH, with unperturbed CH4 used for the reaction CH4 + Cl.The red curves show simulations using perturbed CH4 for the reaction ClI4 + Cl with unperturbed CH4 used for reactions with O(lD) and OlI.The total column responses for each case are listed in the upper right corner of each panel.

Fig. 10 .
Fig.10.Global and annually averaged temperature time series for 1850-2100 at 20, 40, and 60 km from the base simulations (all source gases varied time dependently) of the GEOSCCM (black dotted lines) and 20 model (black solid lines).Also shown are 20 simulations in which only certain source gases are varied time dependently as indicated.The right hand axes show the changes relative to 1850.The observations are from the NCEP reanalysis and reanalysis-2 data at 20 km for 1958-2009 (blue "+"), and the new NASA Modem Era Retrospective-analysis for Research and Applications (MERRA) meteorological analyses for 1979-2009 at 40 and 620 60 km (purple "+").To account for the systematic differences between the models and data, we have added the following offsets to the models at 20, 40, and 60 km, respectively: 20 model: OK, -3.4K, -15K; GEOSCCM: -2.7K, +1.3K, -2K.

Figure 10
Figure 10 shows the globally and annually averaged temperature time series from the base GEOSCCM and 2D model simulations as indicated.As an observational reference, we also show the NCEP reanalysis and reanalysis-53s 2 data at 20 km for 1958-2009, and the new NASA Mod-same IR radiative transfer schemes.Further model temperature trend comparisons are discussed in Appendix B (Figures B2 and B3).Following the time change in the surface boundary conditions (Figure I), the 2D simulated temperature changes in Figure 10 are significantly larger after ~ 1970 compared with the 1850-1970 time period.Prior to 1950, the temperature changes in the stratosphere are due almost entirely to the II (bottom) are significantly smaller by 2095 compared to 2055, following the decrease in the CH 4 surface boundary condition during the latter half of the 21 8t century in the AlB scenario (Figure I). 5 Age of Air 730 660 Previous studies have shown that GHG loading and ozone 735

Fig. 12 .
Fig. 12. Global and annually averaged age of air at 25 km.(Top) Time series for 1860-2100 from the base simulations (all source gases varied time dependently) of the GEOSCCM (black dotted 780 lines) and 2D model (black solid lines).Also shown are 2D simulations in which only certain source gases are varied time dependently as indicated.The blue dashed-dotted line is a 2D-ODS only simulation with all heterogeneous chemical reactions turned off so that no ozone hole is simulated.The right hand axis shows the change 785 relative to 1860.The GEOSCCM time series is comprised of three simulations, which use somewhat different SSTs, for the lime periods1950-2004, 1971-2052, and 1996-2100  (the first 8-10 years of each simulation have been removed to allow for spin-up).For visual clarity and to account for a systematic offset between the models, we have added .3 years to the GEOSCCM curves.(Bottom) Time 790 series for 1960-2100, relative to 1960, from several 2D-C0 2 only simulations which use time dependent perturbations for the different processes indicated as discussed in Figure AI.The sum of the curves in the bottom panel equals the 2D-C02 only simulation (red curve) in the top panel.See text for details.
Fig. 12. Global and annually averaged age of air at 25 km.(Top) Time series for 1860-2100 from the base simulations (all source gases varied time dependently) of the GEOSCCM (black dotted 780 lines) and 2D model (black solid lines).Also shown are 2D simulations in which only certain source gases are varied time dependently as indicated.The blue dashed-dotted line is a 2D-ODS only simulation with all heterogeneous chemical reactions turned off so that no ozone hole is simulated.The right hand axis shows the change 785 relative to 1860.The GEOSCCM time series is comprised of three simulations, which use somewhat different SSTs, for the lime periods1950-2004, 1971-2052, and 1996-2100  (the first 8-10 years of each simulation have been removed to allow for spin-up).For visual clarity and to account for a systematic offset between the models, we have added .3 years to the GEOSCCM curves.(Bottom) Time 790 series for 1960-2100, relative to 1960, from several 2D-C0 2 only simulations which use time dependent perturbations for the different processes indicated as discussed in Figure AI.The sum of the curves in the bottom panel equals the 2D-C02 only simulation (red curve) in the top panel.See text for details.

Fig. 13 .
Photochemical Lifetimes145 140 Fig. 14. (Top)  The difference between the base and fixed loss rate simulations (black lines minus red lines in Figure13), illustrating the lifetime changes due to the changing loss rates.(Bottom) The difference between the fixed loss rate and fixed chemistry and transport simulations (red lines minus green lines in Figure13), showing the lifetime changes due to changes in transport.

Fig. AI .
Fig. AI.Time series for 1950-2100 of zonally averaged and deseasonalized surface temperature, water vapor mixing ratio, and latent heating from the GEOSCCM (black curves), and the corresponding fits to the surface CO2 boundary condition (red curves) for the locations indicated.Inclusion of the seasonal cycle to the CO2 fits is depicted by the orange shading.The surface CO2 boundary condition is shown in the bottom panel.See text for details.
heating shown in Figure AI.The magnitude of the warming in the tropics is somewhat underestimated compared with the GEOSCCM in both the past and future (FigureB2), although the 2D-simulated warming over 1960-2000 compares favorably with the radiosonde data in the global average in FigUfe1280 Fig. B5.Latitude-height cross-sections of September N2 0 averaged over 2004-2009 from AURA/MLS (top) and the 2D model (bottom).Contour interval is 20 ppbv and includes the 10 ppbv contour.
Fig. B6.Latitude•height cross-sections of March H2 0 averaged over 1994-2004 from UARS/HALOE (top) and the 2D model (bot-tom).The top panel includes data from AURA/MLS in the polar regions (averaged over2004-2009) where HALOE lacks data coverage.Contour interval is 0.2 ppmv.The red dashed line separates the regions where the model H2 0 is computed (above) and prescribed to the HALOE climatology (below).
Fig. B7.Vertical profiles of September N2 0 at 30 0 S-500S (top) and 80 0 S-88°S (bottom) from AURA/MLS observations (black triangles) and the 2D model (red curve) averaged over 2004-2009.The error bars (I (T) denote the combined effects of measurement uncertainty and interannual variability.The Antarctic MLS profile is adapted from SPARC CCMVaI1201O].

Fig. B8 .
Fig.B8.Equatorial profiles of the amplitude and phase lag of the seasonal cyle in the quantity H2 0 + 2CH4 from UARS/HALOE (black asterisks) and the 2D model (red).Amplitudes are relative to the values at 16.5 km, and the phase lag is defined to be zero at 16.5 km.Values are averaged over 1994-2004 for both the data and model.
Fig. B9.Latitude-height cross-sections of annually averaged ozone expressed in Dobson Units per kilometer.The observations (top panel) are from the climatology compiled by McPeters et al. [2007] covering the time period 1988-2002.Also shown are the 2D model simulation averaged over 1988-2002 (middle panel), and the difference, model minus observations (bottom panel).For the top and middle panels, the contour interval is 2 DU/km and includes the 3 contour level.For the bottom panel, the contour interval is ± 1 DUlkm.
Figure B9 shows latitude-height cross sections of annually averaged ozone from an observational climatology (top panel), the model (middle panel), and the difference, model minus data (bottom panel) expressed in DU per kilometer.This unit is proportional to the number density per cm 2 di4355 E. L. Fleming et al.: Impact of source gas changes on the stratosphcre The corresponding season-latitude sections of total col..,«o umn ozone also averaged over 1988-2002 are shown in Figure B IO.The observations are from ground-based measurements updated from Fioletov et al. [20021.Again the model shows good overall agreement with the data in reproducing the absolute total ozone magnitudes, the latitudinal gradients;445 and the seasonal variations.The main discrepancy occurs at NH high latitudes where the model tends to underestimate total ozone throughout the year by ~ 20-50 DU.This reflects the model underestimation of ozone at high NH latitudes in,450 the \ 0-\5 km region seen in Figure B9 (bottom panel).

Table 1 .
Steady state 2D model calculated annual/global average flux (kg/sec) required to produce a I % change in annually averaged global total ozone for the compounds and years listed.The ozone change is negative for CFC-II and N2 0. and positive for CH4 and CO2 .

Table 3 .
Steady state 2D model calculated percentage change in annual/global average total ozone per unit mixing ratio change for the compounds and years listed.The mixing ratios are in parts per billion by volume (ppbv) for N2 0, and parts per million by volume (ppmv) for CH4 and C02.Shown in parentheses are the percent-560 age changes of each compound, relative to the background levels, required to produce a I % ozone change.

Table 4 .
Percentage change in annual/global average total ozone for the compounds and time intervals listed, based on the sensitivity factors in Table3and the change in the surface boundary condition in the AlB scenario (Figure1).The CH4 -induced ozone change is 575 mospheric chlorine loading, via the reaction CH 4 + CI which affects the partitioning of Cly (e.g., Figures6 and 9).As a result, significantly less CH 4 flux is required (TableI) to get a I % ozone increase in 2000 (large Cly loading) com-580 pared to 1850, 1950, and 2100 (small Cly loading).Similarly, the CH4 perturbation as a percentage of the background (Table 3) is significantly smaller in 2000 compared to the other years.The resulting CH 4 ODP is nearly twice the magnitude in 2000 compared to 1950, 1950, or 2100.As 585 discussed earlier, the CH 4 results presented here do not acsimilar to the method used in Stolarski et al. [2010] for temperature changes.The resulting percentage ozone changes for several20-year periods are listed in Table 4.Note that the 535 CH 4 -induced ozone change is negative for 2080-2100 since 590 the methane boundary condition decreases during this time period (Figure I).These calculations illustrate that CH 4 had the largest GHG impact on ozone during 1980-2000, owing to the large effect of high Cly loading, with the N 2 0 and 540 .
To represent these long term changes in the 20 model, we compute a sensitivity factor to the C0 2 1215