Geophysical Research Letters

Evaluating how photochemistry and transport determine stratospheric inorganic chlorine in coupled chemistry-climate models



[1] This study examines how the stratospheric conversion of organic molecules containing chlorine (CCly) to inorganic forms (Cly) can be comprehensively evaluated in three-dimensional coupled chemistry-climate models (CCMs) using results from two models (SOCOL and UMETRAC) as examples. Stratospheric inorganic chlorine concentrations depend on both photochemistry and transport. The CCly to Cly conversion process is the first step towards chlorine catalyzed destruction of stratospheric ozone. It is therefore important that models used for the prediction of future stratospheric change accurately simulate this process. Also, because there are multiple processes influencing Cly in CCMs, direct comparison of Cly by itself is of limited use as a validation diagnostic for CCMs. Results show that SOCOL's representation of the photochemical conversion of CCly to Cly is more realistic than UMETRAC's. The CCly to Cly parameterization used in UMETRAC masks transport deficiencies in the model which means that Cly in the polar lower stratosphere from UMETRAC compares better with observations than SOCOL.

1. Introduction

[2] Synthetic chlorofluorocarbons (CFCs) are the primary source of anthropogenic stratospheric chlorine. CFCs are relatively inert in the lower atmosphere but they are photolyzed (and to a lesser extent react with O(1D) and OH radicals) in the stratosphere producing Cl atoms which may go on to destroy stratospheric ozone through catalytic loss cycles [Brasseur and Solomon, 2005]. The extent to which stratospheric ozone concentrations are influenced by anthropogenic emissions of CFCs is therefore dependent on the rate of release of Cl atoms from CFCs in the stratosphere, which is primarily determined by the solar actinic flux driving the photolysis. This in turn depends on transport pathways of air through the stratosphere.

[3] Waugh et al. [2007] highlighted the role of stratospheric transport in determining the concentration of stratospheric inorganic chlorine (Cly). They used an off-line chemistry transport model in three configurations with different stratospheric transport characteristics to demonstrate that the Cly distribution within the model is strongly dependent on the modeled stratospheric transport. They also discuss the importance of transport pathways in determining Cly in the lower polar stratosphere [see also Schoeberl et al., 2000; Douglass et al., 2008].

[4] In this work, the conversion of Cl from organic to inorganic forms in two coupled chemistry-climate models, SOCOL (Solar Climate Ozone Links) [Egorova et al., 2005; Schraner et al., 2008] and UMETRAC (Unified Model with Eulerian Transport and Chemistry) [Austin, 2002], is examined. Results from these two models are used to illustrate how diagnostics focusing on the photochemical conversion of CCly to Cly can assist in understanding the processes affecting Cly distributions in CCMs. The strong link between Cly and polar lower stratospheric ozone concentrations has been noted [World Meteorological Organization (WMO), 2007]. A better understanding of the factors influencing Cly in CCMs will in turn improve confidence in projections of future ozone from these models, particularly in the polar regions.

2. Models

[5] The two models used in this work were chosen as examples because they employ rather different strategies for estimating the rate of conversion from CCly to Cly. SOCOL version 2.0 was used in this study. Extensive changes have been made to the SOCOL version 1.0 model used by Eyring et al. [2006, 2007], in particular with respect to mass conservation in the semi-Lagrangian transport scheme [Schraner et al., 2008]. Two families representing short- and long-lived organic chlorine containing species are explicitly advected in SOCOL. 13 photolytic, 14 O(1D) reactions and 8 OH reactions are used to model the photochemical breakup of 16 CFCs. The two families are partitioned after each model time-step into individual CFC species based on the mean age of air (Γ) and the prescribed concentration of each CFC in the planetary boundary layer (for more details see Schraner et al. [2008]).

[6] The chemistry scheme in UMETRAC has been improved since Austin [2002], and is now the same scheme as used in the AMTRAC model [Austin et al., 2007]. UMETRAC results shown here are from the same REF1 simulation described by Eyring et al. [2006]. Photochemical conversion of CCly to Cly is parameterized based on prescribed fractional release rates of Cl. For more detail see auxiliary material. The rate of change of Cly at time t is assumed to be proportional to CCly(t − Γ) where CCly(t − Γ) is the prescribed organic chlorine concentration in the troposphere Γ years prior to time t, where Γ is the mean age of stratospheric air. The proportionality constant varies with latitude and pressure and represents the Cl fractional release rate with different release rates used for different CFCs. If Cly nears complete conversion, Cly/∂t is set to zero.

[7] Both models were run from 1980 to 2000 using the same prescribed boundary conditions (CFCs, well mixed greenhouse gases, stratospheric aerosol surface area densities, sea surface temperatures and sea ice). The boundary conditions are the same as those used in the (2006) REF1 CCMVal integrations described in detail by Eyring et al. [2006].

3. Results and Discussion

[8] The mean age of air in SOCOL was calculated using a passive tracer with a linearly increasing (in time) concentration in the planetary boundary layer as described by Waugh and Hall [2002]. The passive tracer was advected using SOCOL's hybrid advection scheme [Zubov et al., 1999], rather than the advection scheme in the underlying general circulation model (MAECHAM4). UMETRAC estimates the mean age using an 'age-tracer' with an internal source.

[9] Figures 1a1c compare the 1990–1999 average Γ from SOCOL and UMETRAC with estimates from measurements of SF6 and CO2 [Hall et al., 1999; Andrews et al., 2001; Boering et al., 1996; Elkins et al., 1996]. It is clear that outside the tropical lower stratosphere the mean age of air in both models is significantly younger than that derived from measurements. SOCOL Γ is between 0.5 and 1 year younger than UMETRAC. Figure 1 shows that both models have deficiencies in their stratospheric transport characteristics which is expected to impact the spatial distribution of long lived chemical species in the stratosphere, including Cly [Waugh et al., 2007].

Figure 1.

(a) 20 km, (b) 7S, and (c) 65N stratospheric mean age of air. SOCOL and UMETRAC values are 1990–1999 averages. Measurement estimates derived from SF6 and CO2 measurements [Hall et al., 1999; Andrews et al., 2001; Boering et al., 1996; Elkins et al., 1996]. (d and e) Zonal mean 1990–1999 average Cly distributions from SOCOL and UMETRAC.

[10] Cly in SOCOL has weaker latitudinal gradients in the lower stratosphere (200–50 hPa) than UMETRAC (see Figures 1d and 1e). The greatest differences occur in the polar lower stratosphere where the increase in Cly from near zero concentration in the troposphere occurs at a higher pressure (lower altitude) in UMETRAC than in SOCOL.

[11] Figures 2a and 2b compares the two model time series of 50 hPa Cly at 35–60°N and 80°S respectively. In both regions, Cly from UMETRAC is approximately 50% greater than SOCOL. Cly from SOCOL shows reasonably good agreement with measurements at 30–60°N but significantly underestimates measurements at high southern latitudes. UMETRAC on the other hand overestimates Cly at mid latitudes but is in good agreement with observations at 80°S. The October average Cly (dashed lines in Figure 2b) from SOCOL version 2.0 is in better agreement with other CCMs and observations than results from SOCOL version 1.1 [see Eyring et al., 2006, Figure 11; Schraner et al., 2008, Figure 12].

Figure 2.

Time series of zonal mean Cly, (a) 35–60°N, 50 hPa and (b) 80°S, 50 hPa. The dashed line in Figure 2b shows the time series for October only. The diamonds are estimates of Cly from satellite measurements [Eyring et al., 2006].

[12] The Cly fractional conversion [Solomon and Albritton, 1992] for both models was calculated by taking the ratio of Cly(t) to CCly(t − Γ) where Cly(t) is the stratospheric inorganic chlorine concentration at time t and CCly(t − Γ) is the organic chlorine concentration in the troposphere Γ years prior to t.

[13] The Cly fractional conversion is plotted against Γ in Figure 3. UMETRAC results in Figure 3 show a compact relationship between Cly fractional conversion and the Γ. The relationship is roughly linear up to approximately 3 years/0.8 fractional conversion at which point it flattens as the CCly/Cly nears complete conversion. The fractional conversion/mean age relationship from UMETRAC compares reasonably well with the results from SOCOL but SOCOL results are less linear and show a greater spread of conversion fractions.

Figure 3.

Cly conversion fraction (data from 1990–1999) as a function of the stratospheric mean age of air from UMETRAC (black) and SOCOL (blue: tropics (15°S–15°N), red: mid latitudes, lower stratosphere (60°N–30°N, 70 hPa–150 hPa and similarly for the southern hemisphere), grey: all other locations) (see Figure 4). The dashed lines highlight the two distinct compact relationships of points in the SOCOL results.

[14] It appears that there are two distributions in the SOCOL results indicated by the grey dashed lines in Figure 3. The upper distribution corresponds to air that is transported up through the tropical pipe to the upper stratosphere. The second, lower SOCOL distribution represents air that outflows from the tropical pipe through the subtropical barrier [Strahan and Polansky, 2006] and is recirculated between the tropics and the extratropics in the middle and lower stratosphere. This allows the air to age without experiencing the regions of high CFC photolysis in the middle and upper tropical stratosphere [Waugh et al., 2007; Schoeberl et al., 2000] (see Figure 4).

Figure 4.

Cly conversion rates (in units of fractional conversion per year) from (a) SOCOL and (b) UMETRAC. The colored areas in Figure 4(a) correspond to the regions used to discriminate the SOCOL Cly fractional conversion in Figure 3.

[15] The Cly conversion rate, defined as the Cly fractional conversion from Figure 3 divided by Γ, is plotted in Figure 4. There is little spatial variation in the conversion rate in the UMETRAC results (Figure 4b) indicating that the rate of conversion of CCly to Cly is relatively insensitive to the stratospheric transport pathway. This explains why UMETRAC gives such a compact relationship between Cly conversion and Γ seen in Figure 3. SOCOL on the other hand, exhibits relatively more spatial variation in Cly conversion rates. Different transport pathways can lead to different conversion histories and thus to very different conversion fractions (and Cly concentrations) for air parcels with the same mean age (see Figure 3).

[16] The greatest differences in Cly conversion rates occur in the extratropical lower stratosphere where the conversion rates from SOCOL are significantly lower that those from UMETRAC. This explains why the Cly concentration in UMETRAC in the polar lower stratosphere is greater than SOCOL (Figures 1 and 2). In SOCOL, air that is recirculated between the tropics and the extratropics in the middle and lower stratosphere before being transported to the polar region will have a lower conversion fraction and therefore lower Cly. This recirculated air with low Cly is affecting the polar Cly in SOCOL.

[17] This also explains the greater seasonal cycle in SOCOL Cly than UMETRAC at 50 hPa (Figure 2) even though UMETRAC has the greater seasonal cycle in Γ at the same locations (not shown). Because there is a significant difference in Cly conversion for air following different pathways in SOCOL, any periodic annual injection of recirculated air can lead to a decrease in Cly without a corresponding change in Γ [Strahan and Polansky, 2006].

[18] The conversion rates from SOCOL (Figure 4a) capture the spatial variation expected in the CFC photolysis rates, with the broad maximum in conversion rate in the tropical middle stratosphere in qualitative agreement with the results of Ko et al. [1991] and Schauffler et al. [2003] [see also Solomon and Albritton, 1992]. The conversion rates in UMETRAC on the other hand are greatest in lower stratosphere in the mid-latitudes and tropics. The fact that the spatial patterns in UMETRAC's conversion rates do not appear realistic suggests that any changes in simulated Cly resulting from changes in stratospheric circulation in UMETRAC [Butchart and Scaife, 2001] may not be realistic. This is important because future changes in the strength of the Brewer-Dobson circulation are expected with changing climate [Austin et al., 2007].

4. Conclusions

[19] Study of the stratospheric Cly distribution in CCMs is important because of its strong influence on ozone chemistry. To thoroughly diagnose the treatment of Cly in CCMs at the process level, additional diagnostic tests than those currently used for CCM model intercomparisons are required. Because modeled Cly depends on both transport and photochemical conversion, validation methods should be able to discriminate between the individual processes important in determining Cly [e.g., Eyring et al., 2005] Cly versus Γ provides information on how transport affects Cly and the compact relationships between N2O and CFC concentrations [Schauffler et al., 2003] can be used to test the self consistency of model photochemistry.

[20] Over the past 30 years, decadal ozone trends have been largely been driven by changes in halogen loading [WMO, 2007]. Previous CCM intercomparison studies show a wide range of trends and variability in Cly and total column ozone in the polar regions [Eyring et al., 2006, 2007; Bodeker et al., 2007]. For upcoming coordinated CCM intercomparison activities (, it is timely to reconsider how to comprehensively diagnose CCMs treatment of the individual processes determining stratospheric Cly [Douglass et al., 2008]. Better understanding of the factors influencing Cly in CCMs will in turn help the interpretation of simulated ozone trends.


[21] The authors would like to thank John Austin for his helpful comments. H.S. would like to acknowledge the NIWA high performance computing facility for their support in completing the UMETRAC simulations used in this work. The CCM SOCOL development was supported by the ETH Poly-Projects “VSGC I and II” and by the Swiss National Science Foundation (grant SCOPES IB7320-110884). This work was supported by the New Zealand Foundation for Research, Science and Technology under contract C01X0703.