SEARCH

SEARCH BY CITATION

Keywords:

  • Arctic;
  • stratosphere;
  • denitrification;
  • modeling;
  • ozone;
  • SOLVE/THESEO 2000

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Model and Experiments
  5. 3. Results
  6. 4. Discussion and Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

[1] We have used the SLIMCAT three-dimensional chemical transport model together with observations from the Stratospheric Aerosol and Gas Experiment (SAGE III) Ozone Loss and Validation Experiment (SOLVE) and the Third European Stratospheric Experiment on Ozone (THESEO 2000) to quantify the effect of denitrification on Arctic ozone loss. We have used two different denitrification schemes in the model: one based on the sedimentation of ice particles containing cocondensed nitric acid trihydrate (NAT) and the other based on large NAT particles. The model was forced using both UK Meteorological Office (UKMO) and European Centre for Medium-Range Weather Forecasts (ECMWF) analyses. In the Arctic lower stratosphere the UKMO analyzed temperatures are similar to the ECMWF, except at temperatures near the ice point where the UKMO analyses are colder by over 2 K. Consequently, the UKMO analyses predicted large regions of ice clouds, in contrast to the ECMWF. The denitrification scheme based on large NAT particles gives the best agreement with ER-2 NOy observations for both sets of meteorological analyses. Although the ice scheme and UKMO analyses also produce denitrification, the vertical extent of denitrification and renitrification does not agree as well with the observed NOy. Uncertainties in the budget of ClOy observations from the ER-2 prevent an indirect validation of the best model denitrification scheme based on these data. The denitrified model runs give the best agreement with the observed HCl and ClONO2 reservoirs in mid March. However, UKMO-forced runs generally overestimate the observed ClOx during the same period. The denitrified model runs indicate that by late March 56–74% O3 loss had occurred at 460 K and that denitrification contributed 21–30% of this loss. The model runs showing the largest O3 depletion (forced by UKMO analyses) agree well with ER-2 and ozone sonde data, although these runs overestimated ClOx.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Model and Experiments
  5. 3. Results
  6. 4. Discussion and Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

[2] It is well established that polar stratospheric O3 depletion is driven by halogen-catalyzed photochemical loss cycles involving ClO and BrO [e.g., Solomon et al., 1986; Molina and Molina, 1987; Salawitch et al., 1990]. This loss requires sunlight and is efficient where there are large concentrations of active chlorine (ClOx = Cl + ClO + 2Cl2O2), produced by the heterogeneous conversion of reservoir species (ClONO2 and HCl) at low temperatures in/on solid or liquid polar stratospheric cloud (PSC) particles.

[3] Catalytic ozone destruction can continue until either all the O3 is removed, as occurs in the Antarctic, or until the ClOx is “deactivated” by reconversion to reservoir species. In the Arctic, ClOx is deactivated predominantly by reaction of ClO with NO2 to reform ClONO2, while in the Antarctic the conditions of low O3 and low NOx cause the ClOx to be converted directly to HCl [e.g., Douglass et al., 1995]. Heterogeneous conversion of N2O5 to form HNO3 on liquid aerosols and PSCs also contributes to very low NO2 concentrations in winter [e.g., Noxon, 1978; Fahey et al., 1989]. The recovery of NO2 in the springtime, through photolysis of HNO3 and reaction of HNO3 with OH, leads to the deactivation of ClOx.

  • equation image
  • equation image
  • equation image
  • equation image

[4] Denitrification, the irreversible removal of total odd nitrogen (NOy = N + NO + NO2 + NO3 + 2N2O5 + HNO3 + ClONO2 + HO2NO2 + …) from an air mass by the sedimentation of nitric acid-containing particles, will lead to a permanent removal of gas phase HNO3 in the vortex, which in turn will decrease the rate of release of NO2 and ultimately also the rate of ClOx deactivation. Denitrification can therefore lead to enhanced accumulated polar O3 losses (e.g., 1991/92 [Salawitch et al., 1993], 1996/97 [Rex et al., 1997], and 1994/95 [Waibel et al., 1999]). Renitrification may also occur at lower altitudes.

[5] Both remote [e.g., Santee et al., 1995, 1998; Waibel et al., 1999] and in situ observations [e.g., Fahey et al., 1989; Brune et al., 1991] indicate that denitrification is widespread and severe in the Antarctic polar vortex due to the large extent and persistence of extremely low winter temperatures. Recently, Tabazadeh et al. [2000] used satellite data to show that Antarctic denitrification occurs in mid to late June when the lower stratosphere is still relatively warm. This denitrification occurs without dehydration, though dehydration is also observed later in the Antarctic winter/spring.

[6] The Arctic polar vortex is generally warmer and more disturbed than the Antarctic, though there is evidence that denitrification does occur. Observations of denitrification during cold Arctic winters, for example, 1988/1989, 1994/1995, 1995/1996, and 1996/1997, showed it to be less intense and more sporadic than over the Antarctic [e.g., Fahey et al., 1990; Hubler et al., 1990; Rex et al., 1997; Hintsa et al., 1998; Waibel et al., 1999; Kondo et al., 1999, 2000; Dessler et al., 1999; Santee et al., 1999, 2000]. When denitrification is weak it is more difficult to diagnose. Rex et al. [1997] have shown that the signature of weak (2–3 ppbv) denitrification in NOy/N2O correlations [Fahey et al., 1990] may be difficult to distinguish from isentropic mixing of extravortex air, which is further discussed by Plumb et al. [2000].

[7] The Arctic winter of 1999/2000, following two relatively warm winters, was characterized by a cold and persistent vortex [Manney and Sabutis, 2000]. Temperatures remained below the nitric acid trihydrate (NAT) equilibrium temperature TNAT, a nominal PSC formation temperature, from mid-December to early March (see Figure 1). These conditions caused widespread Arctic denitrification, which was observed for the first time. Satellite observations of HNO3 suggest that there was extensive denitrification around the 465 K potential temperature level which persisted long after temperatures rose above the PSC threshold in early March [Santee et al., 2000]. In situ observations of NOy also indicate widespread denitrification [Fahey et al., 2001]. These observations of widespread denitrification without significant dehydration indicate that denitrification may be caused predominantly by nitric acid particle sedimentation, rather than by ice, as assumed in many chemical models.

image

Figure 1. Minimum temperature in the UKMO analyses at 46 hPa poleward of 50°N for the Arctic winters 1994/1995 to 1999/2000 and ECMWF analyses at 44 hPa for 1999/2000. Also shown are the approximate formation temperatures of Type I (195 K) and Type II (188 K) PSCs. Figure updated from Sinnhuber et al. [2000].

Download figure to PowerPoint

[8] A microphysical model study of the contribution of denitrification to Arctic O3 loss during the 1994/1995 winter [Waibel et al., 1999] showed that the observed denitrification caused an additional 55% ozone loss by mid April. The model denitrification scheme was based on the cocondensation of NAT on to ice followed by evaporation of ice at temperatures above the frost point, with most of the denitrification caused by the residual NAT particles. Tabazadeh et al. [2000] used a photochemical box model to show that 50% denitrification is required for any significant effect upon Arctic O3 and the maximum impact of severe denitrification is limited to an additional 30% O3 loss.

[9] Earlier, Brasseur et al. [1997] used a semi-Lagrangian, three-dimensional (3-D) off-line CTM to evaluate the effect of denitrification on O3 depletion in the Antarctic. The model used an equilibrium PSC scheme for the formation of NAT and ice PSCs. Denitrification was found to delay ClOx deactivation by 2 weeks and was responsible for an additional 20% O3 depletion. Chipperfield and Pyle [1998] also found that denitrification tended to enhance Arctic polar O3 loss, but they also found that very severe, early denitrification may decrease ozone loss by promoting deactivation of ClOx to HCl, which is less efficiently reactivated.

[10] Other modeling studies suggest that denitrification may play a more important role in the Arctic than the Antarctic, where continued springtime activation on PSCs is more likely. Portmann et al. [1996] used a 2-D model to show that denitrification is not a prerequisite for large Antarctic ozone losses. In their model, continuous heterogeneous processing at low temperatures on sulphate aerosols was sufficient to depress NOx and maintain elevated ClOx.

[11] Large cumulative ozone losses in the Arctic are associated with a cold and persistent polar vortex. It is important to understand how denitrification occurs, and its chemical effects, in order to improve predictions of future polar ozone amounts. Temperature records [Pawson and Naujokat, 1999] show that there has been a general cooling of the Arctic lower stratosphere during the last decade coupled to an increased persistence of the polar vortex in late winter/early spring. Model simulations show that further radiatively induced stratospheric cooling is likely to increase both the frequency and severity of Arctic denitrification and ozone loss [e.g., Danilin et al., 1998; Waibel et al., 1999; Tabazadeh et al., 2000].

[12] Shindell et al. [1998] used a general circulation model (GCM) with a very simple O3 chemistry scheme to predict future Arctic ozone depletion. Maximum ozone losses were predicted in the 2010–2020 decade based upon expected chlorine loadings and a decrease in stratospheric temperatures of around 10 K causing enhanced chlorine activation. However, their model did not explicitly include PSCs and related processes, though the study illustrates the potential for climate change to affect polar ozone depletion.

[13] The above model studies all support the notion that denitrification in the Arctic increases the amount of chemical ozone loss. However, these studies have employed many different simplifications (e.g., meteorology, chemistry, and microphysics). In this paper, we aim to test the performance of different denitrfication schemes and to quantify the contribution of denitrification to Arctic ozone loss during winter/spring 1999/2000 using a detailed 3-D CTM. Although our adopted denitrification schemes are relatively simple, they contain details not used previously in global CTMs [e.g., Considine et al., 2000], and we constrain them using observations from the Stratospheric Aerosol and Gas Experiment (SAGE III) Ozone Loss and Validation Experiment/Third European Stratospheric Experiment on Ozone (SOLVE/THESEO 2000) campaign. Having demonstrated that our denitrification schemes are realistic, we want to investigate the chemical effects of the denitrification on NOy, Cly, and O3 loss. In the future we plan to improve the treatment of denitrification by including a full microphysical model of particle growth and sedimentation [Carslaw et al., 2002]. However, this will be much more computationally expensive. The scheme presented here serves as a baseline comparison of a computationally efficient scheme suitable for long integrations of a global CTM.

2. Model and Experiments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Model and Experiments
  5. 3. Results
  6. 4. Discussion and Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

[14] We have used the SLIMCAT off-line 3-D CTM [Chipperfield et al., 1996]. Horizontal winds and temperatures are specified using meteorological analyses. Vertical advection is calculated from heating rates using the MIDRAD radiation scheme [Shine, 1987] and chemical tracers are advected by conservation of second-order moments [Prather, 1986]. The model contains a detailed gas phase stratospheric chemistry scheme (for more information, see Chipperfield [1999]). The model uses photochemical data from DeMore et al. [1997] with the updates of some NOx reactions from Brown et al. [1999a, 1999b].

[15] The model also contains a treatment of heterogeneous reactions on liquid aerosols (LA), nitric acid trihydrate (NAT) and ice, which is now described in some detail. PSCs are assumed to be in thermodynamic equilibrium, with no modification of the model temperature to account for supersaturation. The composition of liquid aerosols (containing HNO3, H2SO4, H2O, and HCl) is calculated analytically [Carslaw et al., 1995a, 1995b] and the equilibrium saturation vapor pressure of H2SO4 is taken from Ayers et al. [1980]. NAT is formed at the equilibrium NAT saturation temperature [Hanson and Mauersberger, 1988], and the presence of ice is tested using the following expression:

  • equation image

where ps is the saturation vapor pressure of water over ice in pascal and T is the temperature in kelvin [Murray, 1967].

[16] Ice is assumed to incorporate NAT as a cocondensate (though the heterogeneous chemical rates on the surface assume pure ice) and nitric acid is removed from the vapor phase according to the expression of Hanson and Mauersberger [1988]. Available surface areas for heterogeneous chemistry are determined from the condensed masses assuming either a fixed number density or a fixed radius.

[17] The reaction probabilities (γ) assumed for the different particles are given in Table 1. The rates of reactions (R2) and (R3) are parameterized following Hanson and Ravishankara [1994], using the HCl solubility of Luo et al. [1995]. The γ for reaction (R9) on liquid aerosol is parameterized following the results of Hanson et al. [1996].

Table 1. Heterogeneous Reaction Probabilities Used in the Model
 ReactionReaction Probability, γ
Liquid AerosolNATIce
(R1)N2O5 + H2O0.10.00030.01
(R2)ClONO2 + H2Oac0.1
(R3)ClONO2 + HCla,bc0.2
(R4)N2O5 + HCl 0.0030.03
(R5)HOCl + HCld0.10.3
(R6)HOBr + HBrd0.10.1
(R7)HOBr + HCld0.10.3
(R8)HOCl + HBrd0.10.3
(R9)BrONO2 + H2Oe0.0060.3

[18] Two different denitrification schemes have been used in the model. In both schemes only liquid aerosol particles exist above TNAT and NAT is assumed to form at TNAT. The schemes differ in their treatment of NAT and ice. In the first scheme ice is assumed to form at the ice frost point and to remove gas phase HNO3 as cocondensed NAT. In this scheme, ice particles are assumed to have a radius of 10 μm and fall velocities of approximately 1500 m d−1 and NAT particles released upon ice evaporation are assumed to have a radius of 1 μm [Larsen, 1991]. Denitrification in this scheme is effectively due only to ice particles containing cocondensed HNO3. For nondenitrified model runs, particle sedimentation is switched off.

[19] In the second scheme no ice forms in the model. NAT particles are represented by a bimodal distribution (radii 0.5 μm and 6.5 μm) based on observations of Fahey et al. [2001]. Fall velocities of 1 m d−1 and 1100 m d−1, respectively, are assumed for the two modes. In this scheme, denitrification occurs by the sedimentation of the large NAT particles, while heterogeneous chemical reactions are calculated on the small mode, which has the greatest surface area density. The number density of the small mode is set to be 1 cm−3 and the condensed HNO3 mass is assigned to this mode to produce particles with a mode radius of 0.5 μm. Any additional condensed HNO3 is assigned to the large mode. For nondenitrified model runs using this scheme, the number density of the small mode is not fixed and all the condensed mass is assumed to be in this small mode. Particle formation in saturated air masses is assumed to be instantaneous in all model runs, a process which may not be representative of the real atmosphere, especially for the large particles [Carslaw et al., 2002].

[20] In the experiments shown here the model resolution was 7.5° longitude × 5° latitude × 18 isentropic levels from 330 to 3000 K (approximately 10 to 55 km). We have used both 24-hourly UK Meteorological Office (UKMO) [Swinbank and O'Neill, 1994] and 6-hourly 60-level European Centre for Medium-Range Weather Forecasts (ECMWF) analyses to force the model. A basic model run (using UKMO analyses) was initialized in October 1991 from a 2-D model and integrated until December 1999. Then eight 120-day experiments were initialized from this basic run. A summary of these experiments is given in Table 2. These eight model runs cover the combinations of different analyses and different denitrification schemes with the corresponding “nondenitrified” runs.

Table 2. SLIMCAT Three-Dimensional Model Runs
RunAnalysesHeterogenous ChemistryaSurface AreabDenitrificationc
  • a

    Model runs UKNND and ECNND have all the condensed mass in 0.5 μm NAT particles.

  • a

    Equilibrium LA/NAT/ICE model assumes liquid aerosol only for T > TNAT, NAT only for Tice < T < TNAT, and ice only for T < Tice.

  • b

    Model runs UKNAT and ECNAT have a bimodal distribution of NAT with radii of 0.5 and 6.5 μm with a maximum number density for the small mode of 1.0 cm−3 with the remaining condensed mass redistributed to the large mode [Fahey et al., 2001].

  • c

    For sedimentation NAT particles have a radius of 1 μm and ice particles have a radius of 10 μm with fall velocities of 40 m d−1 and 1500 m d−1 for UKICE and ECICE. Model runs UKNAT and ECNAT have a NAT fall velocity of 1 m d−1 and 1100 m d−1 for the 0.5 and 6.5 μm modes, respectively [Larsen, 1991].

UKICEUKMOLA/NAT/ice10 cm−3 NAT, 10 μm ice10 μm NAT/ice cocondensate
UKNODUKMOLA/NAT/ice10 cm−3 NAT, 10 μm icenone
ECICEECMWFLA/NAT/ice10 cm−3 NAT, 10 μm ice10 μm NAT/ice cocondensate
ECNODECMWFLA/NAT/ice10 cm−3 NAT, 10 μm icenone
UKNATUKMOLA/NAT0.5 μm (max. 1 cm−3) NAT6.5 μm NAT
UKNNDUKMOLA/NAT0.5 μm NATnone
ECNATECMWFLA/NAT0.5 μm (max. 1 cm−3) NAT6.5 μm NAT
ECNNDECMWFLA/NAT0.5 μm NATnone

[21] A modeled passive ozone tracer was initialized from the model ozone field on 1 December 1999 and was used to diagnose chemical ozone loss. Despite the lower resolution, our results are in agreement with those obtained from the higher resolution simulation of this winter [Sinnhuber et al., 2000]. In all comparisons, model output is at 1200 UT and is not temporally interpolated to observations.

3. Results

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Model and Experiments
  5. 3. Results
  6. 4. Discussion and Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

3.1. Temperature

[22] Temperature has a strong influence on the extent of polar ozone depletion, especially through its role in controlling the occurrence of PSCs. Cold winters, for example 1995/1996 and 1996/1997, are associated with large chemical ozone losses in the Arctic stratosphere [e.g., Rex et al., 1997]. Therefore we need to investigate the fidelity of the analyzed temperatures used in the 3-D model.

[23] Figure 1 shows the minimum temperature north of 50°N at 46 hPa from the UKMO analyses for the Arctic winters of 1994/1995 to 1999/2000 and the minimum temperature at 44 hPa for the ECMWF analyses of winter 1999/2000. In the UKMO analyses, winter 1999/2000 was one of the coldest of the past 6 years. As well as showing a long period below the NAT condensation point (approximately 195 K), winter 1999/2000 also showed the longest period below the ice point (Tice). During the coldest period from mid December 1999 to late January 2000, with temperatures well below the NAT point, temperature minima from the UKMO analyses were ∼1–3 K lower than the ECMWF analyses. This difference is also evident in comparisons of the UKMO analyses with the National Center for Environmental Prediction (NCEP) analyses [Manney and Sabutis, 2000].

[24] Figure 2 shows the area below 195 K and 188 K for both the UKMO and the ECMWF analyses on the 460 K potential temperature (θ) surface. The UKMO analyses produce a considerably greater area below 188 K (Tice), though the area below 195 K (TNAT) is similar. Figure 3 compares UKMO and ECMWF analyses inside the polar vortex at 460 K on the SLIMCAT model grid. At temperatures above 200 K the analyses agree well, but the UKMO analyses are consistently colder in the coldest regions of the lower stratospheric vortex, especially in midwinter (Figure 3). Table 3 summarizes the differences at 460 K; between 9 December and 7 January the UKMO temperatures are significantly colder (−2.33 K, standard deviation 0.70 K) when the temperature is below 190 K and when 190 < TUKMO < 195 (−1.37 K ± 0.73 K). Between 8 January and 6 February the UKMO analyses are also colder than the ECMWF analyses below 190 K (−2.46 K ± 1.59), though the scatter is larger.

image

Figure 2. Area of polar vortex with temperatures below 195 and 188 K for UKMO and ECMWF analyses at 460 K during winter 1999/2000.

Download figure to PowerPoint

image

Figure 3. Scatter plot of invortex ECMWF and UKMO analyzed temperature for (a) 9 December 1999–7 January 2000 and (b) 8 January 2000–6 February 2000 interpolated to 460 K. Vortex defined by UKMO potential vorticity >30 PVU (1 PVU = 1 × 10−6 K m2 kg−1 s−1).

Download figure to PowerPoint

Table 3. Mean Vortex UKMO-ECMWF Temperature Difference at 460 K
DateT Range,a KMean ΔT, KStandard Deviation, K
  • a

    Vortex defined by UKMO potential vorticity; >30 PVU (1 PVU = 1 × 10−6 K m2 kg−1 s−1).

  • a

    Temperature range based on UKMO analyses.

9 December to 7 January<190−2.330.70
190 < T < 195−1.370.73
>195−0.350.86
8 January to 6 February<190−2.461.59
190 < T < 195−0.771.18
>1950.461.16
7 February to 7 March<190
190 < T < 1950.240.68
>1951.140.93

[25] The UKMO and ECMWF analyses are clearly different in the 1999/2000 Arctic lower vortex. In order to determine which analyses are more realistic, we have compared them with sonde data. Figure 4 compares the analyses with temperatures from ozone sonde launches at Ny Ålesund. (Note that these temperature observations are not routinely assimilated in the ECMWF analyses, though it is possible for launches before 1100 UT.) During the midwinter period, where large UKMO-ECMWF temperature discrepancies exist in the lower stratosphere, the sondes show better agreement with the ECMWF analyses (Table 4). Although this sonde comparison indicates the ECMWF temperatures are closer to independent observations, our comparisons with the in situ ER-2 Meteorological Measurements System (MMS) [Scott et al., 1990] temperature observations are not as clear. For example, on the flight of 20 January, when temperatures below 191 K were observed by the ER-2, MMS temperature measurements were lower than both ECMWF and UKMO analyses (comparison not shown).

image

Figure 4. Comparison of UKMO and ECMWF temperature analyses with temperatures from ozone sonde flights from Ny Ålesund (79°N, 12°E) at (a) 460 K and (b) 506 K during winter 1999/2000. The sonde temperatures have been averaged over the SLIMCAT gridbox and the error bar shows the standard deviation. The mean differences at different model levels are given in Table 4.

Download figure to PowerPoint

Table 4. Comparison of Ny Ålesund Ozone Sonde Temperatures With UKMO and ECMWF Analyses
Analysisθ, KDatesT RangeaT, KSD, KNobs
  • a

    Mean sonde temperature.

UKMO4609 December–27 January<195−2.311.5816
>195−3.192.005
28 January–31 March<1951.291.5413
>1950.411.8822
5069 December–27 January<195−2.291.0621
>1950
28 January–31 March<1951.941.2110
>1950.282.1822
ECMWF4609 December–27 January<195−0.411.5816
>195−1.651.485
28 January–31 March<1950.711.1913
>195−0.221.8322
5069 December–27 January<1950.121.0921
>1950
28 January–31 March<1950.800.8310
>195−0.711.7522

[26] Previous studies, using analyses for previous years, have shown that both the UKMO [Pullen and Jones, 1997] and the 31-level ECMWF (with a top boundary at 10 hPa) [Knudsen, 1996] analyses have tended to underestimate the possible extent of PSCs. However, it appears that in winter 1999/2000 the discrepancy the meteorological analyses during the coldest period is in the opposite sense and that the new 60-level ECMWF analyses are in good agreement with observations.

[27] In summary, there appear to be significant differences between the UKMO and ECMWF global analyses in the Arctic lower stratosphere during the period of coldest temperatures in January 2000. The UKMO analyses predict a much larger region below Tice, but the warmer ECMWF analyses are in better agreement with sonde observations.

3.2. Modeled Denitrification

[28] The different metorological analyses used in the model, and the different model denitrification schemes, are expected to produce different extents of denitrification. Figure 5 shows the modeled denitrification in the runs UKICE, ECICE, UKNAT, and ECNAT on 5 days from late December to mid January at 460 K.

image

Figure 5. Modeled denitrification (ppbv of HNO3) on 5 days for winter 1999/2000 at 460 K from four model runs, diagnosed by comparison with their nondenitrified model runs UKICE-UKNOD, ECICE-ECNOD, UKNAT-UKNND, and ECNAT-ECNND. Also indicated are the 28 and 30 PVU UKMO potential vorticity contours.

Download figure to PowerPoint

[29] Using the ice denitrification scheme, which was the default scheme in previous SLIMCAT studies, the UKMO analyses produce strong, extensive denitrification in the lower stratosphere, which develops between late December and early January. The calculated vortex mean denitrification for UKICE for March 2000 at 460 K is 65% (diagnosed from UKICE-UKNOD). This is the first Arctic winter since initialization in 1991 that this version of the model has produced denitrification, although observations have indicated that denitrification has occurred in previous winters. The magnitude of denitrification is significantly smaller when ECMWF analyses are used in conjunction with the ice scheme (16% at 460 K as a vortex mean, diagnosed from ECICE - ECNOD) as expected based on the limited period below Tice.

[30] In the model, sedimentation of large NAT particles causes greater denitrification than the equivalent ice-based schemes. The denitrification starts earlier and is more extensive relative to the ice-based model runs. Although the denitrification starts slightly earlier in the UKNAT run, both UKMO and ECMWF analyses give similar extents of denitrification at this level by early January. Using the colder UKMO analyses and the NAT scheme, vortex mean denitrification at 460 K for March 2000 (from UKNAT - UKNND) is 77% whilst the modeled vortex mean denitrification using ECMWF analyses (ECNAT - ECNND) is 60%. Only model run UKICE produces moderately strong dehydration (>1.0 ppmv) at 500 K with weaker (0.3–0.5 ppmv) dehydration at 460 K and rehydration (0.3–0.6 ppmv) at 420 K. All other model runs produce no significance.

3.3. Comparison With NOy Observations

[31] We now compare the different model runs with NOy observations obtained during SOLVE/THESEO 2000 with the aim of evaluating the different schemes. Although the 3-D model denitrification schemes are relatively simple, we are interested in whether they capture the extent and timing of the observed denitrification in winter 1999/2000 so that we can draw conclusions on the subsequent chemical effect on chlorine chemistry and ozone depletion.

[32] Figure 6 compares modeled NOy with ER-2 observations [Fahey et al., 2001] for four in-vortex flights from January to March 2000, chosen to represent the temporal evolution of the denitification profile. Figure 7 compares the profiles of mean denitrification inside the polar vortex from all ER-2 flights from 14 January 2000 to 12 March 2000. For air masses with T < TNAT, small particles are oversampled by the NOy instrument, thus enhancing the NOy signal. Therefore the level magnitude of denitrification inferred from these data is a lower limit.

image

Figure 6. Comparison of SLIMCAT NOy for four model runs (UKICE blue, ECICE green, UKNAT grey and ECNAT orange) and NASA ER-2 NOy (black) for four in-vortex flights (a) 20 January 2000, (b) 3 February 2000, (c) 7 March 2000 and (d) 12 March 2000. Also shown are comparisons of SLIMCAT denitrification (UKICE-UKNOD, ECICE-ECNOD, UKNAT-UKNND and ECNAT-ECNND) and ER-2 denitrification (NOy-NOy*) against θ for the same four flights (Figures 6e–6h). We obtained NOy* using data from the Mark IV balloon flight of 3 December 1999 using the relation NOy* = 17.0959 − 0.02046[N2O] − 0.000105[N2O]2 (following Fahey et al. [1990]) and using Argus tunable diode laser N2O [Jost et al., 1998]. Uncertainties are NOy (20%), N2O (4.2 – 18.0 ppb).

Download figure to PowerPoint

image

Figure 7. Profiles of mean denitrification inside the polar vortex from all ER-2 flights from 14 January 2000–12 March 2000 for ER-2 observations (thick black line, diagnosed from NOy-NOy*), model run UKICE (dashed line, diagnosed from UKICE—UKNOD), model run ECNAT (dotted line, diagnosed from ECNAT–ECNND) and model run UKNAT (thin black line, diagnosed from UKNAT–UKNND). The data have been averaged in 10 K θ bins and the vortex was defined as PV > 19 PVU at 420 K. The error bars show ±1 standard deviation. ER-2 NOy* for 23 January uses ALIAS N2O (estimated accuracy 2–10%).

Download figure to PowerPoint

[33] The observations clearly show regions of strong denitrification (up to 12 ppbv) at 460 K and evidence for renitrification below 400 K. Model run UKICE underestimates the magnitude of the observed denitrification/renitrification (as diagnosed from the NOy* correlation) by approximately 2–3 ppbv, and the denitrification is displaced upwards by around 20 K potential temperature. Note that there is considerable fine horizontal structure in the observed NOy which our coarse grid model and equilibrium denitrification scheme is unable to reproduce.

[34] Forcing the ice model with the warmer, but more realistic, ECMWF analyses (run ECICE) produces considerably less denitrification (around 2 ppbv at 460 K) due to the very short time that air masses are exposed to temperatures below Tice. This model run strongly underestimates the observed extent of denitrification.

[35] Replacing the ice scheme with the denitrification scheme based solely on NAT (runs UKNAT and ECNAT) produces NOy fields that are in better agreement with observations than either ECICE or UKICE. The mean difference between modeled and observed NOy is generally smaller for ECNAT compared to UKNAT, which tends to overestimate NOy at ER-2 altitudes in January and underestimate NOy in March. Runs ECNAT and UKNAT also better reproduce the vertical distribution of denitrification and renitrification. This is especially evident in the “stack flight” (flight consisting of a series of legs at increasing altitudes over Kiruna) of 3 February (Figures 6b and 6f) where the ice denitrification scheme shows NOy increasing along the flight track, while observations show that the sampled air was generally denitrified above 380 K.

[36] When model fields are sampled in a manner consistent with the UARS Microwave Limb Sounder averaging kernel [Santee et al., 2000], vortex mean denitrification is reduced from 50% to around 30% at 68 hPa in model run UKICE for late March. The degraded HNO3 field produced is consistent with MLS observations at this time although the depressed level of HNO3 is due to denitrification not photolysis.

3.4. Effect on Cly Species and Comparison With Observations

[37] Denitrification is expected to delay the deactivation of ClOx, which in turn can lead to increased ozone depletion. Here we investigate the effect of the different denitrification schemes on the behavior of key Cly species. We also evaluate the different model runs against ER-2 observations of ClOx, ClONO2 [Stimpfle et al., 1999], and Aircraft Laser Infrared Absorption Spectrometer (ALIAS) HCl [Webster et al., 1994] to investigate if these Cly species can be used to indirectly evaluate the different denitrification schemes.

3.4.1. Activation

[38] Figure 8a shows that the heterogeneous activation of reservoir species to ClOx on PSCs in early winter is identical in runs UKICE and UKNOD as this process occurs rapidly on NAT at temperatures below ∼195 K. Similarly, for ECICE and ECNOD (Figure 8b) there is no difference between the denitrified and nondenitrified model runs during the activation phase.

image

Figure 8. (a) SLIMCAT runs UKICE (solid lines) and UKNOD (dotted lines) vortex mean ClONO2 (thick black), ClOx (thin black) and HCl (grey) for winter 1999/2000. (b) As Figure 8a but for runs ECICE (solid lines) and ECNOD (dotted line). (c) As Figure 8a but for runs UKNAT (solid lines) and UKNND (dotted line). (d) As Figure 8a but for runs ECNAT (solid lines) and ECNND (dotted line). Vortex defined from potential vorticity at 460 K > 30 PVU.

Download figure to PowerPoint

[39] Interestingly, for model runs based on NAT denitrification, activation of Cly to ClOx is reduced in denitrified model runs (UKNAT and ECNAT) compared to their nondenitrified equivalents (UKNND and ECNND). This is due to the reduced NAT surface area available for heterogeneous processing in model runs UKNAT and ECNAT where the majority of the condensed NAT is in the large mode during this period in the denitrified models (see Figure 9).

image

Figure 9. Time-height cross sections of temperature at Ny Ålesund (79°N, 12°E) from (a) UKMO and (b) ECMWF analyses (note nonlinear scale). Also shown for the same location are SLIMCAT heterogeneous reaction rates on solid PSCs for runs UKICE, ECICE, and UKNAT and ECNAT for (c) ClONO2 + HCl and (d) ClONO2 + H2O. The heterogeneous rates are given as the first order loss of ClONO2 (s−1).

Download figure to PowerPoint

3.4.2. Deactivation

[40] For run UKICE the diagnosed denitrification (65% as a vortex mean) has no effect on the vortex-averaged amount of active chlorine until early February, when a significant portion of the vortex is exposed to solar radiation and photochemical loss of HNO3 reforms NOx. Denitrification slows the reconversion of ClOx to ClONO2 by around 2 weeks for run UKICE compared to run UKNOD by reducing available NOx. The relatively weak denitrification in run ECICE has a barely discernible effect on Cly chemistry. Note the reconversion of ClOx to HCl is enhanced by denitrification.

[41] Delayed deactivation of ClOx is also evident in model runs using the NAT denitrification scheme (UKNAT and ECNAT). Comparison of UKNAT with UKNND shows that recovery of ClOx to ClONO2 is delayed by more than 20 days with a mean vortex denitrification of 8.1 ppbv (77%) at 460 K. Comparison of model runs ECNAT and ECNND, with a mean vortex denitrification of 5.9 ppbv (60%) at the same potential temperature show a reduced recovery time of around 10 days.

3.4.3. Diagnosis of Heterogeneous Rates

[42] In model runs ECICE, ECNOD, UKNAT, ECNND and, especially, run ECNAT, vortex-mean ClONO2 remains significantly greater than zero. This is a result of incomplete activation due to slower heterogeneous processing rates at the higher temperatures in the ECMWF analyses and lower surface area in NAT-based model runs.

[43] Diagnosis of these rates (Figure 9) shows that for the reaction fo ClONO2 with HCl (R3), the heterogeneous rates on solid particles during the activation phase in mid December are around 10−3 s−1 for the UKICE run and 10−4 s−1 for the ECICE run. The denitrified NAT model runs UKNAT and ECNAT show significantly lower rates of around 3.0–7.0 × 10−15 s−1 whilst their nondenitrified equivalents (UKNNND and ECNND) exhibit more rapid processing, with rates around 10−4 s−1.

  • equation image
  • equation image

Reaction (R2) also processes ClONO2, and this reaction is considerably more rapid in the UKICE model run as the heterogeneous rates are considerably enhanced on ice particles (1 × 10−3 s−1) compared to the ECICE run (1 × 10−7 s−1) where there is very limited ice formation and most of the processing occurs on NAT.

3.4.4. ER-2 Comparisons

[44] ER-2 observations of key Cly species are compared to model results from the ice schemes in Figure 10 and the NAT schemes in Figure 11. ER-2 ClO has been adjusted to account for ClO2 formation at low temperatures in the instrument (R. Stimpfle, personal communication). These dates include the ER-2 flights late in the winter where the effects of denitrification are expected to be the largest. Most UKMO model runs overestimate ClOx by 0.2–0.6 ppbv in midwinter when the model has essentially full activation (not shown). It is only during the second ER-2 deployment (late February and March) that the effect of denitrification on ClOx becomes apparent. During this period the denitrified UKMO runs (UKICE and UKNAT) have ClOx values that exceed observations by around 0.2–0.5 and 0.2–0.6 ppbv, respectively. The nondenitrified UKMO run UKNOD appears to agree better with the observations yet underestimate ClOx by ∼0.2 ppbv in mid-March. In mid-March the run UKNND underestimates ClOx by around 0.4 ppbv.

image

Figure 10. Comparison of ER-2 observations (black) with model results from runs UKICE (blue) and UKNOD (green), ECICE (red) and ECNOD (yellow) for (a)–(d) ClOx, (e)–(h) ClONO2, and (j)–(m) HCl for 4 flights. Note, no ER-2 HCl data for 26 February 2000. All flights are within the polar vortex. Uncertainties in the observations are HCl 2–10%, ClOx 17%, Cl2O2 20%, and ClONO2 20%.

Download figure to PowerPoint

image

Figure 11. As Figure 9 but with model results from runs UKNAT (blue), UKNND (green), ECNAT (red), and ECNND (yellow).

Download figure to PowerPoint

[45] The runs forced by ECMWF analyses also overestimate the observed ClOx in midwinter by around 0.4 ppbv on some flights but underestimate ClOx by a similar amount on others (not shown). In mid-March, runs ECICE and ECNOD overestimate ClOx by around 0.4–0.5 ppbv. By mid-March, run ECNAT is in good agreement with observations, while ECNND underestimates ClOx by ∼0.7 ppbv.

[46] All model runs apart from UKNAT tend to overestimate ClONO2 when compared to ER-2 observations. The strongly denitrified runs, especially UKNAT, but also UKICE and ECNAT, produce significantly smaller discrepancies in mid March (up to ∼0.4 ppbv). In fact, the observations of ClONO2 inside the vortex on 12 March (see Figures 10 and 11) give the clearest indication that the denitrified model runs are more realistic than the nondenitrified runs. Under the conditions of 1999/2000, observations of chlorine species even later in the winter would have been desirable.

[47] The agreement between modeled and ALIAS HCl is poor during January and early February, the observations show much larger values than the model (not shown). In March the agreement is much better for the strongly denitrified model runs, and the larger HCl values of run UKNAT, UKICE, and ECNAT are in better agreement than the nondenitrified model runs.

[48] The comparison of the model with the ER-2 chlorine species observations is inconclusive. An important caveat to this comparison is that the sum of ER-2 HCl, ClONO2, and ClOx observations underestimates the expected Cly by an average of 20% (D. M. Wilmouth et al., manuscript in preparation, 2002). This discrepancy precludes any definitive statements about the performance of the different model schemes with the current data. In general, we can say the the denitrified model runs tend to give the best comparison with the HCl and ClONO2 reservoirs in late winter. However, these denitrified runs tend to overestimate ClOx in this period, especially for the runs forced by UKMO analyses.

3.5. Effect on Calculated Ozone Loss

[49] We now quantify the effect that denitrification has on O3 for the different model runs. Figures 12a–12d compare O3 from the denitrified model runs with ER-2 observations, showing that there is generally excellent agreement between modeled and observed O3 in the polar vortex before significant ozone depletion occurs (e.g., 20 January 2000 and 3 February 2000). In early to mid March the comparison with the model passive O3 shows that significant O3 depletion has occurred in UKICE and UKNAT (over 55% by 12 March) and there is good agreement with observed ozone values. Model run ECNAT gives smaller losses (around 40% by 12 March) and overestimates observed ozone by around 0.6 ppmv. Note that the ECMWF-forced runs and UKMO-forced runs show different features near the edge of the polar vortex (3.5 × 104 s – 4.5 × 104 s on 11 March. This difference is due to the higher temporal resolution of the ECMWF analyses which produces tighter horizontal gradients in modeled tracer fields at the edge of the vortex.

image

Figure 12. (a)–(d) comparisons of ER-2 observed O3 (black) with SLIMCAT model runs UKICE (blue), ECICE (yellow), UKNAT (green), ECNAT (red) and the passive O3 tracers from UKICE (dashed blue) and ECICE (dashed yellow) for the dates indicated. (e) Plot shows a comparison of ER-2 O3 (black) with model runs UKICE (blue), UKNOD (dotted blue), ECICE (yellow), ECNOD (dotted yellow) and the passive O3 tracers from UKICE (dashed blue) and ECICE (dashed yellow) for 12 March 2000. (f) As Figure 12e but for model runs UKNAT (green), UKNND (dotted green), ECNAT (red), ECNND (dotted red) and the passive O3 tracers from UKNAT (dashed green) and ECNAT (dashed red). The blue background illustrates where the normalized GSFC potential vorticity at 420 K for the ER-2 flightpath is less than 18 PVU. The 12 UT model output has been interpolated to the ER-2 flightpath. Uncertainties in observations of O3 are 3%.

Download figure to PowerPoint

[50] Figure 13 shows the comparison of the eight model runs with sonde observations at Ny Ålesund at 460 K. Model run UKICE indicates that the accumulated ozone loss by late March is 69%, with denitrification contributing around 30% of the loss (relative to UKNOD). The UKNAT model run shows accumulated losses reaching 74% by the same date with denitrification also contributing 30% of the loss (relative to UKNND).

image

Figure 13. Comparison of O3 sonde observations (+) at Ny Ålesund (79°N, 12°E) with model output from runs (a) UKICE (thick solid black), UKNOD (thick dotted black), UKICE passive O3 (thin dotted black), ECICE (solid grey), ECNOD (dotted grey), ECICE passive O3 (thin solid black) at 460 K. (b) UKNAT (thick solid black), UKNND (thick dotted black), UKNAT passive O3 (thin dotted black), ECNAT (solid grey), ECNND (dotted grey), ECNAT passive O3 (thin solid black) at 460 K.

Download figure to PowerPoint

[51] The accumulated ozone loss in the ECMWF runs at Ny Ålesund by late March is less than the UKMO runs. Accumulated losses for ECNAT is 48% and the effect of denitrification is smaller, contributing around 21% of the O3 loss. Accumulated losses for the ECICE run peaks at 44% with denitrification responsible for only 6% of this loss.

[52] In the ECMWF-forced model runs, Ny Ålesund lies closer to the edge of the vortex and can be seen from Day 90 onwards. Max O3 loss is 56% in the ECNAT model run and 49% in the ECICE run.

[53] The agreement between the different model runs and the O3 is inconsistent with the NOy and Cly comparisons. Even though runs ECNAT and UKNAT give the best simulation of the observed denitrification, and the ECMWF analyses are more representative of stratospheric temperatures, ECNAT does not appear to reproduce the O3 loss as well as run UKICE or UKNAT. However, the ECNAT run does appear to reproduce the ER-2 ClOx observations better than the UKMO-forced runs. Based on these comparisons, it seems that the model with the realistic ClOx evolution underestimates O3 loss. This result is in accord with other studies which point to this model deficiency [e.g., Guirlet et al., 2000; Hansen and Chipperfield, 1999; Becker et al., 1998]. However, there are other uncertainties (e.g., the lack of closure of the observed Cly budget, differences in model vertical transport with different analyses) which need to be investigated in more detail before a firm conclusion can be reached.

[54] In-vortex chemical O3 loss rates for the period 26 February to 12 March 2000 for the UKMO-forced model runs are comparable to the losses calculated by Gao et al. [2001] from ER-2 observations. O3 loss rates of 63 ppbv d−1 for 71% denitrified air parcels and 43ppbv d−1 for 43% denitrified air parcels are reported. Losses averaged 59 ppbv d−1 over this period for the 77% denitrified UKNAT model run and 56 ppbv d−1 for the 65% denitrified UKICE model run. The nondenitrified UKMO-forced model runs UKNOD and UKNND have loss rates of 42 ppbv d−1 and 39 ppbv d−1, respectively. All ECMWF-forced models have O3 loss rates that are considerably lower than their UKMO-forced equivalents.

4. Discussion and Conclusions

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Model and Experiments
  5. 3. Results
  6. 4. Discussion and Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

[55] We have used a detailed 3-D chemical transport model to quantify the effect of widespread denitrification on ozone depletion in the Arctic winter of 1999/2000. In eight model runs, denitrification was parameterized in the model by two methods: (1) sedimentation of ice particles containing NAT and (2) sedimentation of large NAT particles using two sets of meteorological analyses (UKMO and ECMWF).

[56] Observations show that in winter 1999/2000 widespread denitrification occurred in the Arctic vortex. Two of our NAT-based model simulations were able to reproduce this widespread denitrification leading to modeled NOy in good agreement with the ER-2 observations.

[57] The ice denitrification scheme forced with UKMO analyses produces strong denitrification, but the NOy profile is displaced vertically upwards by around 20 K compared to observations. Forcing the ice denitrification scheme with ECMWF analyses produces very little denitrification due to the very limited ice formation that occurs in the ECMWF analyses. This difference illustrates the sensitivity of the ice-based denitrification scheme to relatively small temperature variations below 190 K.

[58] The ECMWF-forced NAT scheme produces earlier and more widespread denitrification than the ice-based schemes and captures the magnitude of denitrification well at all but the highest altitudes, where it underestimates the NOy deficit by ∼2 ppbv. The UKMO-forced NAT-based denitrification results are similar to the ECMWF NAT scheme but with enhanced redistribution of NOy due to the lower temperatures in midwinter resulting in a greater condensed HNO3 mass in the sedimented large NAT mode. Compared to the UKMO-forced ice scheme, both schemes better reproduce the vertical distribution of denitrification/renitrification although they tend to produce renitrification at lower altitudes than observations.

[59] The strongly denitrified UKMO-forced ice-based model run has almost complete activation of ClONO2 around 1 January 2000 on a vortex-wide scale. Vortex-mean ClONO2 levels for the two other strongly denitrified models runs (using the UKMO and ECMWF-forced NAT schemes) remain around 0.2 and 0.3 ppbv, respectively, as the processing on the less abundant and smaller NAT particles was slower compared to ice.

[60] The uncertainties in the ER-2 observations of ClOx, ClONO2, and HCl mean that it is not possible to unambiguously determine which model simulation is the most realistic. All nondenitrified model runs have too much ClONO2 and too little HCl in mid-March when compared to ER-2 observations. The three strongly denitrified model runs show delayed recovery of ClOx into ClONO2 by between 10 and 20 days and enhanced recovery to HCl, which is consistent with in situ ER-2 observations during mid March. The most denitrified model run (the UKMO-forced NAT scheme) gives the best agreement with observed ClONO2 and HCl at this time but the UKMO-forced denitrified schemes appear to have too much ClOx.

[61] In winter 1999/2000 there is a lack of data in the late March period, when the differences in Cly partitioning between the denitrified model runs are greatest. Data during this period would have given a clearer evaluation of the different denitrified models.

[62] The modeled O3 loss inside the vortex is sensitive to the analyses used. Losses reach 69% at 460 K by late March in UKMO-forced ice denitrification model run and 74% with the UKMO-forced NAT denitrification scheme, in excellent agreement with the sondes from Ny Ålesund, although in these runs ClOx is apparently too large.

[63] O3 losses are considerably smaller when the ECMWF analyses are used, reaching around 49% with the ice scheme and 56% with the NAT denitrification scheme at 460 K. This discrepancy is due to reduced levels of ClOx in the critical March period when vortex air masses receive significant sunlight. Incomplete initial heterogeneous activation of Cly and lower denitrification resulting in faster deactivation of ClOx both contribute to this difference with the UKMO runs, though the ClOx was in better agreement with the ER-2 data. Comparison of modeled and experimentally derived O3 loss rates for the period 26 February to 12 March show that, when the UKMO analyses are used, extensive denitrification results in O3 loss rates of around 60 ppbv d−1 and loss rates of around 40 ppbv d−1 for less denitrified air.

[64] Our model shows that widespread and severe denitrification of the Arctic lower stratosphere during the cold winter of 1999/2000 can account for an additional 30% O3 loss at 460 K when the UKMO analyses are used with either the ice or NAT denitrification schemes, and an additional 21% loss when the ECMWF analyses are used with the NAT scheme. Although these contributions are important, they are not a very large enhancement.

[65] The predicted increased atmospheric burden of greenhouse gases may increase the frequency of colder and isolated Arctic polar vortices in the future. This may lead to more extensive denitrification, which will enhance O3 loss. Further studies are required to assess the importance of future enhanced denitrification.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Model and Experiments
  5. 3. Results
  6. 4. Discussion and Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

[66] This work was strongly supported by the UK Natural Environment Research Council. S.D. thanks NERC for a studentship. M.P.C. thanks NERC for a Fellowship. B.M.S. is supported by the NERC UTLS programme. The modeling work at Leeds was also supported by the EU through contract EVK2-1999-00311. We thank Ross Salawitch for help with formatting ER-2 data files. The ozonesonde data at Ny Ålesund are provided through the NILU database. The UKMO and ECMWF analyses were obtained via the British Atmospheric Data Centre.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Model and Experiments
  5. 3. Results
  6. 4. Discussion and Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Model and Experiments
  5. 3. Results
  6. 4. Discussion and Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.