Evolution of inorganic chlorine partitioning in the Arctic polar vortex

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 111,, doi: /2005jd006951, 2006 Evolution of inorganic chlorine partitioning in the Arctic polar vortex D. M. Wilmouth, 1 R. M. Stimpfle, 1 J. G. Anderson, 1 J. W. Elkins, 2 D. F. Hurst, 2 R. J. Salawitch, 3 and L. R. Lait 4 Received 3 December 2005; revised 14 March 2006; accepted 19 April 2006; published 25 August [1] The first simultaneous, in situ atmospheric measurements of ClO, ClOOCl, ClONO 2, and HCl, which together nearly compose total inorganic chlorine, Cl y, were obtained using the NASA ER-2 aircraft, deployed from Kiruna, Sweden, during the SOLVE/ THESEO mission. These chlorine measurements, along with Cl y inferred from in situ measurements of organic chlorine source gases, offer an unprecedented opportunity to observe chlorine activation and recovery in the polar winter stratosphere and evaluate the inorganic chlorine budget, i.e., a test of the quantitative agreement between the sum of ClO x (ClO + 2 ClOOCl), ClONO 2, and HCl with Cl y. Inside the Arctic vortex, inorganic chlorine is rarely fully activated with ClO x /Cl y ranging from approximately 0.25 to An apparent discrepancy in the inorganic chlorine budget for much of the midwinter correlates well with back trajectory solar zenith angle minimum and is consistent with the presence of Cl 2, which was not observable. Evaluation of the midwinter inorganic chlorine budget, excluding data where Cl 2 is present, yields (ClO x + ClONO 2 + HCl)/Cl y = 0.94 ± In contrast, a shortfall in the late winter inorganic chlorine budget is the result of an apparent low bias in the HCl measurements. A diurnal box model constructed to analyze chlorine recovery rates highlights the importance of not only the Cl + CH 4 reaction in HCl formation, but also OH + ClO and potentially HO 2 + ClO. We find that significant HCl production accompanied net ClONO 2 production in the inorganic chlorine recovery phase of this Arctic winter. Citation: Wilmouth, D. M., R. M. Stimpfle, J. G. Anderson, J. W. Elkins, D. F. Hurst, R. J. Salawitch, and L. R. Lait (2006), Evolution of inorganic chlorine partitioning in the Arctic polar vortex, J. Geophys. Res., 111,, doi: /2005jd Introduction [2] Chlorine derived from CFCs, acting alone and in concert with bromine in catalytic cycles, is responsible for the dramatic, widespread destruction of ozone that occurs annually over the poles in late winter and early spring. The observed ozone depletion is the result of heterogeneous reactions on stratospheric aerosols that convert halogen species from their reservoir forms into active forms, which catalytically destroy ozone. Specifically, the reservoirs HCl and ClONO 2 compose more than 95% of inorganic chlorine (Cl y HCl + ClONO 2 + ClO + 2 ClOOCl) in the lower stratosphere outside the polar vortex, while ClO and ClOOCl, which readily participate in reactions that destroy ozone, are the predominant forms inside the activated vortex [World Meteorological Organization, 2003]. [3] The development of the polar vortex each winter effectively isolates the air inside from the rest of the 1 Department of Chemistry and Chemical Biology, Harvard University, Cambridge, Massachusetts, USA. 2 Global Monitoring Division, NOAA Earth System Research Laboratory, Boulder, Colorado, USA. 3 Jet Propulsion Laboratory, California Institute of Technology, Pasadena, California, USA. 4 NASA Goddard Space Flight Center, Greenbelt, Maryland, USA. Copyright 2006 by the American Geophysical Union /06/2005JD stratosphere [Schoeberl and Hartmann, 1991; Schoeberl et al., 1992]. Because of the lack of solar heating during the long periods of darkness in the polar winter, the vortex radiatively cools, and polar stratospheric clouds (PSCs) form at sufficiently low temperatures (195 K). It is on these liquid or solid condensates of HNO 3 /H 2 SO 4 /H 2 O mixtures that the heterogeneous chemical reactions which promote ozone destruction take place. The most important reactions are ClONO 2 þ HCl þ aerosol! HNO 3 þ Cl 2 ClONO 2 þ H 2 O þ aerosol! HNO 3 þ HOCl HOCl þ HCl þ aerosol! H 2 O þ Cl 2 [4] These reactions repartition inorganic chlorine into more photolabile species, which are rapidly converted into ozone-destroying reactive chlorine (ClO x ClO + 2 ClOOCl) in the presence of sunlight. Reactions (1) and (2) additionally sequester nitrogen oxides into the stable product, HNO 3, which contributes to ozone destruction by reducing the amount of NO 2 available to convert active chlorine back into reservoir forms. The reactions are extremely temperature-dependent, with observed ozone losses ð1þ ð2þ ð3þ 1of16

2 being much greater when the winter polar stratosphere is colder [Salawitch, 1998; Newman et al., 2002; World Meteorological Organization, 2003]. [5] Following the winter buildup of reactive chlorine via heterogeneous reactions, ozone is primarily destroyed by one of two catalytic cycles. The first involves chlorine only and has the weakly bound dimer, ClOOCl, as the principal chain carrying species [Molina and Molina, 1987], ClO þ ClO þ M! ClOOCl þ M ClOOCl þ hn! Cl þ ClOO ClOO þ M! Cl þ O 2 þ M ð4þ ð5þ ð6þ 2 ðcl þ O 3! ClO þ O 2 Þ ð7þ Net : 2O 3! 3O 2 The ClO dimer mechanism has long been considered critical to explain the large ozone losses observed in the polar stratosphere, even though historically ClOOCl had not been measured in the atmosphere but was assumed to exist on the basis of laboratory observations. The observations presented here and by Stimpfle et al. [2004] represent the first measurements of ClOOCl in the stratosphere. [6] The second critical ozone loss mechanism involves a coupling of chlorine and bromine cycles [McElroy et al., 1986; Anderson et al., 1991; Wennberg et al., 1994], BrO þ ClO! Br þ Cl þ O 2 Br þ O 3! BrO þ O 2 Cl þ O 3! ClO þ O 2 ð8þ ð9þ ð10þ [8] Once formed, ClONO 2 is lost, in the absence of heterogeneous reactions, via photolysis, ClONO 2 þ hn! Cl þ NO 3! ClO þ NO 2 ClONO 2 þ hn! ClO þ NO 2 ð12aþ ð12bþ In general, as long as ClO x is greater than NO 2, there will be net production of ClONO 2 on a diurnal basis if there is net production of NO 2 from HNO 3 degradation. [9] HCl production in the Arctic vortex recovery phase has historically been believed to be much slower than that of ClONO 2 [e.g., Webster et al., 1993; World Meteorological Organization, 1995, Figure 3.1], proceeding primarily via the reaction of Cl + CH 4, Cl þ CH 4! HCl þ CH 3 ð13þ Since the mixing ratio of Cl is strongly influenced by the reaction of Cl with O 3 to form ClO, the mixing ratio of O 3 plays a pivotal role in controlling the production rate of HCl [Douglass and Kawa, 1999]. [10] The potential importance of other HCl formation reactions has recently been recognized. In particular, studies of the HCl production channel (14b) from the ClO + OH reaction, ClO þ OH! Cl þ HO 2 ClO þ OH! HCl þ O 2 ð14aþ ð14bþ have yielded rate constants significantly higher than previously thought [Lipson et al., 1999; Wang and Keyser, 2001; Tyndall et al., 2002]. The JPL evaluation panel made its first recommendation for the branching ratio of (14b) in the 2000 evaluation [Sander et al., 2000], and the most recent recommendation [Sander et al., 2003] is for a 6% HCl yield at typical stratospheric temperatures. One recent laboratory study reports an HCl yield of 9.0 ± 4.8% independent of temperature between 218 and 298 K [Wang and Keyser, 2001]. [11] A more speculative HCl production channel from the reaction of ClO with HO 2 (15b) has also been suggested, Net : 2O 3! 3O 2 ClO þ HO 2! HOCl þ O 2 ð15aþ [7] These halogen-catalyzed ozone loss cycles continue unabated in the polar vortex until chlorine recovery occurs, i.e., the deactivation of ClO x back to ClONO 2 and HCl via gas-phase reactions. Specifically, ClONO 2 is formed via the three-body reaction of ClO with NO 2, ClO þ NO 2 þ M! ClONO 2 þ M ð11þ Reintroduction of ClONO 2 is limited by the availability of NO 2, which has concentrations at or near zero. The reappearance of gas-phase NO 2 occurs primarily via HNO 3 photolysis and HNO 3 reaction with OH. The average mixing ratio of HNO 3 will vary year-to-year with the severity and spatial extent of denitrification. ClO þ HO 2! HCl þ O 3 ð15bþ The principal product channel produces HOCl (15a), but several studies have indicated an upper limit on the HCl yield (15b) of 1 2% at 298 K [Leu, 1980; Leck et al., 1980; Knight et al., 2000]. Moreover, Leck et al. [1980] determined an upper limit of 3.0% on the HCl yield at 248 K, and Finkbeiner et al. [1995] found the HCl yield to be 5 ± 2% at 210 K. [12] There are numerous other possible HCl production mechanisms, such as the reaction of Cl with CH 2 O, H 2, and other hydrocarbons, but these are all of lesser importance. 2of16

3 Table 1. ER-2 Flight Summary for the SOLVE/THESEO Mission Julian Date Time b Latitude c Longitude SZA The sole known loss mechanism for HCl in the absence of heterogeneous reactions is HCl þ OH! Cl þ H 2 O Flight Character d Flight a W transit intravortex intravortex extravortex intravortex intravortex intravortex intravortex intravortex intravortex W extravortex intravortex W transit a Flights are listed by month and day in 2000 in which they occurred (e.g., 0114 = 14 January 2000). b Time is the duration of the flight in hours. c Latitude, longitude, and solar zenith angle are the minimum and maximum values in degrees characterizing each flight. All latitude values are for the Northern Hemisphere; longitude is east unless indicated otherwise. d Transit flights were between Westover, Massachusetts, and Kiruna, Sweden; all intravortex and extravortex flights originated and terminated in Kiruna; intravortex flights took place entirely within the Arctic vortex; and extravortex flights included segments both inside and outside the vortex. ð16þ [13] Given the important role of chlorine species in controlling ozone loss at the poles, as well as potentially at midlatitudes, accurate measurements of all the forms of inorganic chlorine in the stratosphere are critical. Previous to these observations, it has not been possible to directly monitor the evolution of chlorine activation and recovery in the polar winter because of the absence of ClOOCl measurements and the absence of in situ ClONO 2 measurements in the vortex regions. Attempts to test the inorganic chlorine budget, i.e., the agreement between the sum of the measured inorganic chlorine species and the inferred total inorganic chlorine, Cl y, from tracer measurements have been similarly hampered. Moreover, there are a number of outstanding questions. How often is chlorine fully activated in the Arctic winter stratosphere? How does the partitioning of inorganic chlorine change throughout the course of the polar winter/ spring? Does chlorine recovery in the Arctic take place through ClONO 2 almost exclusively, followed by the subsequent formation of HCl, as suggested by previous work? What are the mechanisms of HCl recovery? Are the observed inorganic chlorine budget and the organic chlorine budget consistent? If not, what is the source of the discrepancy? [14] Here we present the first simultaneous, in situ measurements of ClO, ClOOCl, ClONO 2, and HCl obtained in the atmosphere. These observations along with Cl y, inferred from measurements of organic chlorine source gases, enable us to address the enduring questions listed above. Specifically, for the first time we are able to directly monitor the evolution of chlorine activation and recovery and evaluate the inorganic chlorine budget in the polar winter stratosphere. 2. Measurements [15] The measurements presented here were obtained in situ using the NASA ER-2 aircraft during the joint Sage III Ozone Loss and Validation Experiment (SOLVE) and the Third European Stratospheric Experiment on Ozone (THESEO) 2000 campaign deployed out of Kiruna, Sweden (67.9 N, 21.1 E) [Newman et al., 2002]. There were two ER-2 deployments during the SOLVE/THESEO mission. The first deployment took place in late January 2000 when polar lower stratospheric temperatures were coldest and the vortex was strongest. The second deployment occurred in early March 2000 as the perturbed chemistry of the vortex began to recover [Newman et al., 2002]. The ER-2 was flown 13 times in these midwinter and late winter time periods, as summarized in Table 1. The flights of 0114 (14 January 2000; all flights took place in 2000) and 0316 were transits into and from Kiruna, connecting with Westover, Massachusetts. The other 11 flights originating from Kiruna remained entirely within the polar vortex with the exception of 0127 and 0311, which were flown across the vortex edge. The flight tracks of the ER-2 for the 13 flights listed in Table 1 are shown in Figure 1. [16] The in situ measurements of ClO, ClOOCl, and ClONO 2 were made with the Harvard Halogen instrument using a thermal dissociation/resonance fluorescence technique. Detection of ambient ClO is via a rapid bimolecular reaction with NO followed by ultraviolet resonance fluorescence detection of Cl atoms at nm [Anderson et al., 1980; Brune et al., 1989]: ClO þ NO! Cl þ NO 2 ð17þ Cl þ hnðrf lampþ! Cl* ð18þ Cl*! Cl þ hnðsignalþ ð19þ Figure 1. Map of the Arctic region studied during the SOLVE/THESEO campaign with the flight tracks of all the ER-2 flights indicated. 3of16

4 This core technique is extended to detect other weakly bound Cl y species by incorporating a heater to thermally dissociate ClOOCl or ClONO 2 within the time constraints of the sampling method, 5 to 10 ms: DH¼71kJ=mol ClOOCl! ClO þ ClO ð20þ DH¼113kJ=mol ClONO 2! ClO þ NO 2 ð21þ Figure 2. (a) Relationship between ACATS-IV N 2 O and Cl y for the science flights of the SOLVE/THESEO mission, The least squares polynomial fit given by (22) is shown. (b) Relative difference between ACATS-IV Cl y and the fit data: (Cl y Fit)/Cl y. The standard deviation of the relative residual is The ClO fragments thus produced are titrated with NO and detected in a manner similar to ambient ClO (17 19). The primary method of heater operation involves stepping between three different temperatures: 273 K, where no dissociation takes place; 363 K, where only ClOOCl dissociates; and 493 K, where both ClONO 2 and ClOOCl dissociate. The ClONO 2 concentration is calculated from the difference in the observed Cl signals at 493 K and 363 K. [17] The dissociation heaters, which consist of a grid of 28 resistively heated silicon strips, reside between the front and rear detection axes on each of two 5.08 cm square, mirror image ducts. ClO measurements are made at the front detection axes, where the fluorescence signals are modulated primarily by the addition of NO, while ClOOCl and ClONO 2 measurements are made at the rear detection axes, where the fluorescence signals are modulated by both NO addition and flow temperature. The sampled air is the laminar core of the flow through the primary duct of the aircraft wing pod, extracted and decelerated to m/s. [18] The estimated accuracy of the ClO measurements from the flight instrument is ±17% (1s; all error estimates herein are 1s) with a detection limit of 3 pptv. ClOOCl and ClONO 2 are detected with an accuracy of ±21% and a detection threshold of 10 pptv. ClO measurements are reported every 35 s, the length of the NO addition cycle, while ClOOCl and ClONO 2 are reported less frequently because of the heater cycle. The instrument has recently been described in detail [Wilmouth, 2002; Stimpfle et al., 2004]. [19] The in situ measurements of HCl were obtained by the ALIAS instrument using a tunable diode laser (TDL) infrared absorption spectrometer. The estimated accuracy of the HCl measurements is cited by Webster et al. [1994] to be ±10%. Cl y is determined as the difference between total chlorine, calculated from global tropospheric trends in chlorinated trace gases [e.g., Montzka et al., 1999], and total organic chlorine, calculated from in situ stratospheric measurements of CFC-11, CFC-12, CFC-113, CH 3 CCl 3, CCl 4, and halon These halocarbons are measured by the ACATS-IV instrument using in situ gas chromatography with electron capture detection [Elkins et al., 1996; Romashkin et al., 2001]. Total chlorine computations utilize mean ages of stratospheric air masses calculated from ACATS-IV measurements of SF 6 [Volk et al., 1997] and tropospheric trend data weighted with mean age spectra. Accounting for small discrepancies between ACATS-IV total organic chlorine values and those measured in whole air samples collected onboard the ER-2 [Schauffler et al., 2003], the estimated accuracy of Cl y is ±6 7% at typical cruise altitude Cl y levels of ppt. Measurements of temperature and pressure were made by the Meteorological Measurement System (MMS) with an accuracy of ±0.3 K and ±0.3 mbar, respectively [Scott et al., 1990]. [20] Because Cl y is reported at a relatively large data interval (140 s), a polynomial fit between the Cl y and N 2 O measurements from ACATS-IV is determined for each flight and used to enhance the time resolution on reported Cl y.an all flight fit for the eleven science flights of the SOLVE/ THESEO mission, , demonstrates the tight correlation between the Cl y and N 2 O mixing ratios, as shown in Figure 2a. The least squares polynomial fit is given by Cl y ¼ 3:4290 1: ðn 2 OÞ 2: ðn 2 OÞ 2 2: ðn 2 OÞ 3 ð22þ where Cl y and N 2 O are in units of ppb. Figure 2b shows the relative difference between ACATS-IV Cl y and the fit data. The standard deviation of the relative residual is The compact nature of the Cl y N 2 O relationship strongly suggests that there is no anomalous loss of organic chlorine due to heterogeneous reactions on PSC particle surfaces in the Arctic polar vortex. 3. Inorganic Chlorine Partitioning [21] To provide perspective on the overall data set, chlorine measurements from 0131, a midwinter flight early in the ER-2 portion of the SOLVE/THESEO mission, are depicted in Figure 3. The ER-2 flew in the northwesterly direction to 79 N, 12 E, then southeasterly back to Kiruna. As indicated by the solar zenith angle (SZA) in Figure 3a, 4of16

5 Figure 3. Data from the intravortex flight of (a) Potential temperature (solid) and solar zenith angle (shaded) versus UT. (b) Mixing ratios of observed ClO, ClOOCl, ClO x, and inferred Cl y. (c) Partitioning of active chlorine. the flight began in sunlit conditions, flew into darkness, then returned to sunlight. The changing SZA is clearly manifest in the mixing ratios of ClO and ClOOCl shown in Figure 3b, as ClO is more abundant in sunlight because of ClOOCl photolysis (5 7), and ClOOCl is more abundant in darkness because of the ClO self-reaction (4). ClO x is relatively uniform throughout the flight despite the fluctuations in ClO and ClOOCl mixing ratios. The dip in concentrations at 12.2 hr UT is due to an aircraft dive from 20 km cruise altitude to 15.5 km, as indicated by the coincident drop in potential temperature (Figure 3a). Figure 3c shows the fraction of inorganic chlorine in its active form, ClO x /Cl y, and the ratio of each species to total inorganic chlorine. For this flight, inorganic chlorine reaches a maximum activation of 0.89 with an average of 0.79 ± 0.06, omitting the dive. ClOOCl composes the majority of ClO x for essentially the entire flight, because it was mostly in darkness. [22] The results from 0131, along with data from the other SOLVE/THESEO flights are depicted in Figure 4, where the fraction of inorganic chlorine in the active form is plotted as a function of ambient temperature. In order to provide additional extravortex data, three pre-solve/ THESEO flights are also shown: 0106 and 0109, which originated in Edwards, California, and 0111, which originated in Westover, Massachusetts. Only cruise altitude data are displayed; the ascent, descent, and dive portions of each flight are eliminated in order to prevent introduction of errors from small temporal offsets in the measurements coupled with the rapidly changing concentrations of the observed species. ClO mixing ratio provides an excellent indicator for the vortex edge. Intravortex data are defined as having ClO > 900 ppt and extravortex data as having ClO < 150 ppt for the 0127 and 0311 flights, which left the vortex. [23] A sharp contrast between the data obtained inside the vortex and outside the vortex is evident in Figure 4. Outside the vortex, temperatures are generally warmer and chlorine activation is always less than Inside the vortex, ClO + 2ClOOCl composes a much more significant fraction of Cl y. The large range in ClO x /Cl y is primarily driven by the fact that these measurements were taken in both midwinter and late winter, when chlorine is expected to be at peak activation and when chlorine is recovering, respectively. Of particular interest is that the ClO x /Cl y observations peak at 0.90; whether this represents complete activation cannot be determined within the experimental uncertainty. However, the extent of activation does vary considerably from this peak number, even within the midwinter flights, e.g., Figure 3c, when we expect to observe the highest ClO x /Cl y values. This variability, coupled with the high precision of the measurements, suggests that inorganic chlorine was seldom completely activated in the 1999/2000 Arctic winter vortex. [24] The change in partitioning of inorganic chlorine responsible for the wide range of chlorine activation observed inside the vortex is illustrated in Figure 5. The average chlorine partitioning into ClO x, ClONO 2, and HCl is shown for the intravortex, cruise altitude data from the ER-2 science flights. The lines are linear least squares fits to the data, as described in the figure caption. With two Figure 4. Fraction of total inorganic chlorine in the active form as a function of ambient temperature for flights Solid points correspond to data acquired inside the Arctic polar vortex, and shaded points are from data obtained outside the vortex. 5of16

6 [26] While chlorine recovery was taking place during the second ER-2 deployment of the SOLVE/THESEO mission, rapid catalytic destruction of ozone was occurring as well. Losses of approximately 60% were observed in the lower stratosphere of the Arctic vortex between late January and mid-march [Richard et al., 2001; Newman et al., 2002]. Data from the 0311 flight, for example, depict substantial intravortex ozone loss and further make the case linking stratospheric chlorine to ozone destruction. Figure 6 shows ClO and O 3 on 0311 plotted as a function of latitude, where the lower latitudes sampled by the ER-2 were in extravortex air, and the higher latitudes were in intravortex air. Upon entering the vortex, the ClO concentration rises sharply, the O 3 concentration falls, and there is a clear anticorrelation in their abundances. Figure 5. Partitioning of inorganic chlorine into ClO x, ClONO 2, and HCl for the science flights of the SOLVE/ THESEO mission, The points represent the mean values of the intravortex, cruise altitude data. Data are shown for all flights for which they exist. The lines are linear least squares fits to the mean values. The anomalous 0120 and 0203 flights are omitted from the ClO x /Cl y fit and are discussed in detail in section 4.2. The ClONO 2 /Cl y fit includes the last flight of the first deployment, where ClONO 2 /Cl y is 0, and all of the second deployment flights. The fit lines are shown to display the qualitative trends in the measurements. exceptions, ClO x /Cl y abundances are significantly higher in the first ER-2 deployment, , than in the second, (Table 1 contains a reference to relate Julian Date to Flight Date.) The mean values inside the vortex range from 0.82 ± 0.09 on 0127 to 0.42 ± 0.05 on 0312, representing a significant decrease in the contribution of active chlorine to total inorganic chlorine over this time period. Not only is ClO x /Cl y lower in the second deployment, it is generally decreasing over the two weeks of measurements, indicative of continuing chlorine recovery. [25] In addition to declining ClO x /Cl y, observed increases in ClONO 2 concentrations, evident in Figure 5, are also indicative of chlorine recovery in March. ClONO 2 is below the instrument detection limit for most of the first ER-2 deployment but is a significant fraction of Cl y in the second deployment. In contrast, HCl appears relatively constant throughout the mission, representing 0.1 of Cl y. These data seem to indicate that chlorine recovery in the Arctic polar vortex preferentially takes place through ClONO 2 rather than HCl, which may be formed more slowly. However, the increase in ClONO 2 /Cl y (0.2) is not sufficient to compensate for the observed decrease in ClO x /Cl y (0.4). This discrepancy, as well as the two anomalous flights of 0120 and 0203, will be discussed in more detail in subsequent sections. 4. Inorganic Chlorine Budget 4.1. Overview [27] In addition to monitoring the evolution of chlorine activation and recovery, simultaneous measurements of ClO, ClOOCl, ClONO 2, and HCl during SOLVE/THESEO enable the inorganic chlorine budget to be directly evaluated in the polar vortex for the first time. The analysis of interest is to test whether the sum of ClO x, ClONO 2, and HCl is equivalent to total inorganic chlorine, Cl y. [28] The most recent budget evaluation was conducted by Bonne et al. [2000] using chlorine data from POLARIS, a summer campaign in the Arctic in The measured chlorine species were found to quantitatively account for total inorganic chlorine for this mission, i.e., (ClO + ClONO 2 + HCl)/Cl y = 0.92 ± ClOOCl concentrations were negligible for the conditions of this campaign. No evidence was found to support the presence of missing chlorine species that contribute to Cl y in a significant way. Bonne et al. [2000] further analyzed observations from Figure 6. Observed mixing ratios of ClO (solid) and O 3 (shaded) as a function of latitude for the flight of The vortex edge is located at 66 N. The ClO mixing ratio is much higher and O 3 is much lower in intravortex air. There is a clear anticorrelation in the abundances of ClO and O 3 in the vortex edge and intravortex regions. 6of16

7 fraction of Cl y of each measured species is also shown, such that its contribution to the budget can be readily noted. Omitting the dive from cruise altitude at 15.3 hr UT, the ratio of ClO x + ClONO 2 + HCl to Cl y is between 0.75 and 1.10 for essentially the entire flight. [30] Examination of Figure 7c reveals that there are actually two distinct regions to this flight. The average budget value for the first two thirds of the flight, i.e., prior to 13.9 hr UT, is 0.97 ± 0.06, while after this time, the average value falls to 0.82 ± We demonstrate below that this is a real phenomenon, correlated with a change in sampled air mass. Further examination of the chlorine partitioning plots in Figure 7c reveals that although ClOOCl is the dominant form of inorganic chlorine for most of this flight, much of the variation in the budget is driven by changes in HCl. This frequently holds true, as the observed range in budget values for a given flight is often enlarged by peaks or dips in the HCl measurements. [31] The excellent agreement of the inorganic chlorine budget observed for most of 0202 does not hold true for all the flights of the SOLVE/THESEO mission, as summarized in Figure 8, where (ClO x + ClONO 2 + HCl)/Cl y inside the vortex at cruise altitude is plotted as a function of Julian Date in The vertical points indicate the range of budget values observed for each ER-2 flight day and the dotted horizontal line represents complete budget agreement. The 0123 flight is not shown because of the lack of Figure 7. Data from the intravortex flight of (a) Potential temperature (solid) and solar zenith angle (shaded) versus UT. (b) Mixing ratios of observed ClO, ClOOCl, HCl, and inferred Cl y. ClONO 2 is below the detection limit of the instrument for this flight. (c) Chlorine partitioning. The sum of the measured species, ClO x, ClONO 2, and HCl, divided by total inorganic chlorine, Cl y represents the inorganic chlorine budget for this flight. The dotted line at 1.0 represents complete budget agreement. earlier ER-2 missions dating to 1991, which had substantial shortfalls in the inorganic chlorine budget. The authors concluded that a low bias in measured HCl was the most likely cause of past budget problems. [29] For perspective on the budget data from the SOLVE/ THESEO mission, a budget evaluation for the first deployment, midwinter flight of 0202 is shown in Figure 7. For this flight the ER-2 flew in the easterly direction to 66 N, 63 E, northwesterly to 78 N, 12 E, then southerly back to Kiruna. Potential temperature and solar zenith angle are shown in Figure 7a. The sampled air was in sunlight for approximately the first half of the flight and in darkness for the second half. The observed mixing ratios of ClO, ClOOCl, HCl, and Cl y are shown in Figure 7b. As expected, ClO decreases and ClOOCl increases as the solar zenith angle increases. ClONO 2 was below the detection threshold of the instrument for this flight, so it is not considered. Figure 7c shows the ratio of the sum of the measured species to total inorganic chlorine, (ClO x + ClONO 2 + HCl)/Cl y, where ClONO 2 is taken to be zero. The dotted line at 1.0 represents complete budget agreement. The Figure 8. Inorganic chlorine budget inside the vortex throughout the course of the SOLVE/THESEO mission, The vertical points indicate the range of budget values observed at cruise altitude for each flight. The shaded points represent chlorine budget agreement within the 1s uncertainty of the measurements, and the solid points are outside the 1s error limits. The 0123 and 0226 flights are not shown because of the absence of ClOOCl and HCl data, respectively. The dotted line at 1.0 represents complete budget agreement. 7of16

8 ClOOCl data, and the 0226 flight is not shown because of the absence of HCl data. [32] The shaded points are those that represent agreement of the inorganic chlorine budget within the 1s uncertainty of the measurements. The solid points are when the chlorine budget is not in agreement within the 1s error range. Uncertainties are calculated assuming a worst case scenario in which errors in ClO, ClOOCl, and ClONO 2 are not independent and random. The sum of the errors in these measurements is combined in quadrature with the error in both HCl and Cl y to produce the results shown here. The ratio of ClO x + ClONO 2 + HCl to Cl y necessary for the budget to be in agreement varies slightly from flight to flight because of the different relative abundances of chlorine species present. Typically a ratio of 0.83 is required to be considered agreement. It is immediately clear from Figure 8 that the sum of the measured inorganic chlorine species differs significantly from inferred Cl y for a large portion of the SOLVE/THESEO mission. [33] Analysis of the flight data shown in Figure 8 is aided by considering the flights from the midwinter and late winter independently. In the first deployment, the flights of 0127, 0131, and 0202 agree well with each other and generally represent good overall budget agreement, although there are short periods when the budget disagrees, particularly on 0131 and By contrast, the 0120 and 0203 flights have the inorganic chlorine budget in disagreement most of the time. Moreover, there is a very large spread in the budget values obtained, much larger than those observed on all the other flights of the mission. The four flights of the second deployment are also distinct, as they exhibit a more compact range of budget values, similar to 0127, 0131, and 0202, but generally have poor budget agreement, a characteristic of 0120 and Not only does the inorganic chlorine budget tend to be out of agreement for the second deployment flights, there is generally a downward trend, indicating that the budget became more in error as the second deployment progressed. [34] In order to attempt to reconcile the observations shown in Figure 8, we consider that there are three possible explanations for an inorganic chlorine budget disagreement: (1) a measurement/calculation error in total inorganic chlorine, Cl y ; (2) a measurement error in one or more of the observed chlorine species; or (3) an unmeasured, inorganic chlorine species that is present in significant abundances but not measured by any instrument. Option 1 is unlikely to explain the budget discrepancies here, given the tight correlation between Cl y and N 2 O. The fact that Cl y yields good budget results for some flights in the first deployment also suggests that the measurement is valid. The other two options for the first and second deployment discrepancies will be discussed in turn Midwinter Budget Analysis [35] Focusing on the five first deployment flights in Figure 8, it is unlikely that there could have been a significant measurement error by the Harvard Halogen instrument on some flights and not others, given the consistency maintained in preflight procedure and instrument preparation. Specifically, ClO and ClOOCl compose the vast majority of Cl y for these flights, and the generally good budget agreement for three flights provides strong evidence that these measurements are similarly valid for the other two. Even if the ClONO 2 or HCl measurements were in error, their abundances in the midwinter time period are not sufficient to explain the observed budget discrepancies. This leaves the possibility of an important, unmeasured, inorganic chlorine species present during portions of the first deployment flights, particularly 0120 and Such a chlorine species would need to be present at mixing ratios of approximately ppt for 0120 and ppt for 0203 to bring the budget calculations for these flights into agreement. [36] A notable correlation emerges in analyzing the first deployment data: (ClO x + ClONO 2 + HCl)/Cl y is only outside the 1s uncertainty range when the observation occurred at night. In particular, the 0120 and 0203 flights were almost entirely within darkness, the only exception being the first hour of 0120 when the budget agrees. Despite this correlation, there is no straightforward relationship between solar zenith angle and the inorganic chlorine budget values, as there are flight segments in darkness for other flights when the budget agrees well. No other observed flight variable is significantly different when the inorganic chlorine budget is in agreement versus when it is not; however, a trend does emerge when we use back trajectory data to analyze the history of the sampled air parcels prior to the flights. For the purposes of this analysis, we initially focus on the 0203 flight, as it is the only one that occurred entirely at night. [37] Figure 9 shows flight data and back trajectory data for 0203, the flight of the SOLVE/THESEO mission with the poorest budget agreement. The flight consisted of several segments at increasing altitude levels and took place entirely within the Arctic vortex over Kiruna. Potential temperature and SZA at the time of measurement are shown in Figure 9a. The solar zenith angles that the sampled air parcels experienced in the previous 1-day period, in 2.4-hour increments, are shown in Figure 9b. Back trajectory SZA is computed from the time and position of selected parcels along the flight track using an isentropic trajectory model, run with meteorological analyses of the Goddard Global Modeling Office of Assimilation [Schoeberl and Sparling, 1995]. [38] As shown in Figure 9b, there is a range of solar zenith angles to which the air parcels were exposed prior to the time of measurement for the 0203 flight, with most of the previous 24 hours having been in darkness. The minimum solar zenith angles, indicated by solid circles, show very little variation, as all values are in the range. Figure 9c shows the inorganic chlorine budget, (ClO x + ClONO 2 + HCl)/Cl y, and solar exposure, the fraction of time the air parcels were exposed to sunlight, here calculated for the 1-day period prior to being sampled. A solar zenith angle of 92, indicated by the solid line in Figure 9b, is used to demarcate daytime and nighttime for the solar exposure calculation. [39] There is tremendous variation in the budget values for 0203, exemplified by the starting agreement near 1.0, which drops to below 0.4 just over an hour into the flight. Despite this variability, the calculated solar exposure tracks with the chlorine budget quite well. The solar exposure values have inferior resolution since there are only 10 points in the 1-day back trajectory, but it is clear that the budget 8of16

9 Figure 9. Data from the multiple-level, intravortex flight of (a) Potential temperature (solid) and solar zenith angle (shaded) versus UT. (b) Range of solar zenith angles to which the air masses sampled during the flight were exposed in the 24 hours prior to the flight. Solid circles indicate the minimum solar zenith angles. (c) Solar exposure, the fraction of time the sampled air was exposed to sunlight prior to the flight, with the inorganic chlorine budget, (ClO x + ClONO 2 + HCl)/Cl y, shown for comparison. Solar exposure is calculated for the 1-day back trajectory using a solar zenith angle of 92 to demarcate the light-dark transition. has its greatest shortfalls when solar exposure is 0 and is significantly improved at higher solar exposure values. [40] To more fully explore the inorganic chlorine budget dependence on the history of the sampled air parcels, Figure 10 shows the budget values for the 0203 flight as a function of minimum solar zenith angle in the 1-day back trajectory. Minimum SZA is an excellent indicator of the extent to which the air parcels were exposed to sunlight, and unlike solar exposure, no assumption is necessary regarding which solar zenith angles represent sunlight and which represent darkness in this analysis. There is a salient correlation, as the inorganic chlorine budget values decrease dramatically with increasing SZA. The solid line is a linear least squares fit to the data. Despite the relatively small range in minimum SZA, 4, the corresponding change in the budget is more than a factor of 2. [41] The data in Figure 10 are for the 1-day back trajectory, but the trend is the same for the 2-day and is still evident in the 3-day back trajectory. The trend does not hold for back trajectories greater than 3 days, however. This is primarily because the air masses sampled between hr UT and hr UT, when the budget agreement is the poorest (Figure 9c), had significant exposure to sunlight, i.e., median SZA minimum of 89.5, 4 days prior to the flight. [42] For insight into the results for 0203, it is of interest to test whether a similar correlation between the minimum back trajectory solar zenith angle and the inorganic chlorine budget is present for the other flights of the first deployment. As such, Figure 11 depicts 1-day back trajectory data from the 0120, 0127, 0131, and 0202 flights. The coloring of the points is analogous to Figure 8; that is, shaded points represent where the inorganic chlorine budget is in agreement within the 1s uncertainty of the measurements, and solid points represent where the budget is out of agreement. Overall, there are fewer points in Figure 11 than in Figure 8, because the resolution of the chlorine data has been reduced to match that of the back trajectory data. [43] The range of minimum solar zenith angles spanned by the 4 flights in Figure 11 is significantly larger than for 0203, but the inorganic chlorine budget values are similarly dependent on the degree to which the sampled air parcels were exposed to sunlight prior to the flights. The results suggest three distinct data bins, which are denoted with vertical dotted lines. When the minimum SZA prior to the time of measurement is less than 89.5, the budget is always in agreement, independent of whether the measurement itself took place in sunlight or darkness. These data are primarily from the 0127 and 0202 flights. When the back trajectory SZA minimum is greater than 96, the inorganic chlorine budget is always out of agreement. These data are entirely from the 0120 flight and were also in darkness during the flight. Between 89.5 and 96, (ClO x + Figure 10. Inorganic chlorine budget versus minimum solar zenith angle in the 1-day back trajectory for the cruise altitude data of the 0203 flight. 9of16

10 known chemical mechanism by which an inorganic chlorine species could accumulate in the concentrations necessary to balance the midwinter SOLVE/THESEO chlorine budget. [45] The results in Figures 10 and 11 can be used to distinguish the presence of Cl 2 from that of HOCl. The unmeasured chlorine species transitions from complete photolysis to effectively no photolysis over the solar zenith angle range of Figure 12 shows the photolytic loss rates of Cl 2 (Figure 12a) and HOCl (Figure 12b) as a function of time at solar zenith angles near twilight. The photolysis rates are calculated using a radiative transfer model based upon laboratory-determined absorption cross sections [Salawitch et al., 1993, 1994]. A time range of 6 hours is shown on the abscissa. The length of exposure at the minimum SZA is certainly far less than this, on the order of minutes, but all the solar zenith angles greater than the Figure 11. Inorganic chlorine budget versus minimum solar zenith angle in the 1-day back trajectory for the flights of 0120, 0127, 0131, and The 0123 flight is not shown because of the absence of ClOOCl data. The coloring of the points is the same as in Figure 8; that is, shaded points represent where the chlorine budget is in agreement within the 1s uncertainty of the measurements, and solid points represent where it is out of agreement. Vertical dotted lines at 89.5 and 96 demarcate the regions where the budget is always in agreement (SZA < 89.5 ), both in and out of agreement (89.5 < SZA < 96 ), and always out of agreement (SZA > 96 ). ClONO 2 + HCl)/Cl y generally decreases as SZA increases, with budget values ranging from good agreement to significant disagreement. Data in this range are from portions of all 4 flights, but most of the solid points are from 0120 and the previously mentioned final portion of 0202 (Figure 7c), when there was a change in sampled air mass to one that had not been exposed to sunlight in the 8 days prior to the flight. The number of back trajectory days for which the trend shown in Figure 11 is valid varies for each flight depending on the timing of the last significant exposure to sunlight. [44] The most likely explanation for the results presented in Figures 10 and 11 is that an unmeasured, inorganic chlorine species is present during the midwinter flight segments that took place in darkness and for which the sampled air parcels had a previous history of darkness. The unmeasured chlorine species is photolyzed as a strong function of SZA near twilight and converted into a form of inorganic chlorine detectable by a flight instrument onboard the ER-2. This requirement for darkness at the time of measurement as well as a previous history of significant darkness implicates the presence of Cl 2 or HOCl as an unaccounted for but important species in the inorganic chlorine budget. Both species are products of heterogeneous chlorine activation on PSCs (1 3). These two molecules can occur in large concentrations in air that has been processed but not exposed to sunlight. There is no other Figure 12. Fraction of (a) Cl 2 and (b) HOCl lost via photolysis as a function of time at solar zenith angles near twilight. Photolysis rates are calculated for the SOLVE/ THESEO winter using a radiative transfer model coupled with laboratory-determined absorption cross sections. Dashed lines are used to represent the photolysis of Cl 2 and HOCl at 90 for ease of comparison. HOCl photolysis at solar zenith angles greater than 93 is negligible in 6 hours. Photolysis of HOCl at 80 (shaded) is similar to that of Cl 2 at of 16

11 minimum are sampled at both sunrise and sunset, increasing the effective exposure time. [46] It is immediately clear from Figures 12a and 12b that the rates of photolysis of Cl 2 and HOCl are quite distinct. For example, examination of the 90 data, shown as a dotted line for ease of comparison, reveals that approximately 90% of Cl 2 is photolyzed after 1 hour of exposure, while only 20% of HOCl is lost. Cl 2 is essentially completely photolyzed at a solar zenith angle of 89 within 1 hour, but over 70% of starting HOCl is still present. Even at a solar zenith angle as low as 80, shown as shaded line in Figure 12b, nearly 40% of HOCl remains unphotolyzed after 1 hour of exposure. Moreover, Cl 2 continues to exhibit significant photolysis at solar zenith angles up to 95 on the timescale shown here, while HOCl photolysis is essentially negligible above 93. [47] The response of Cl 2 to sunlight is consistent with the photolysis signature of the missing chlorine species depicted in Figures 10 and 11. Neither HOCl, nor any other inorganic chlorine species examined has a similar photolytic fingerprint. While there may be some HOCl present, it is unlikely to explain the large budget discrepancies observed here. Steady state calculations of HOCl using the ER-2 HO x measurements support this, indicating the presence of 52 ppt HOCl on average inside the vortex during sunlight. Moreover, sunrise measurements of HO x are inconsistent with the presence of a large nighttime reservoir of HOCl [Hanisco et al., 2002b]. [48] We conclude that the apparent error in the inorganic chlorine budget for much of the first deployment of the SOLVE/THESEO mission is due to the presence of significant concentrations of Cl 2. If we reanalyze the midwinter flights, excluding all data where Cl 2 is likely present, i.e., excluding air parcels that were not exposed to sunlight either during the flights or in the 1-day period before the flights, we find (ClO x + ClONO 2 + HCl)/Cl y = 0.94 ± 0.07, representing excellent agreement of the inorganic chlorine budget Late Winter Budget Analysis [49] The mean intravortex values of (ClO x + ClONO 2 + HCl)/Cl y for the second deployment are 0.82 ± 0.06 for the 0305 and 0307 flights and 0.72 ± 0.04 for the 0311 and 0312 flights. The source of the inorganic chlorine budget discrepancy here is necessarily different from that of the first deployment. These four flights occurred much later in the winter, and the sampled air masses had significant exposure to sunlight both before and during the flights. While this does not eliminate the possibility of a missing chlorine species, it does rule out Cl 2 and the mechanism described above. The possibility of a measurement error must be examined more closely. [50] The downward trend in budget values in the second deployment coupled with the overall level of disagreement relative to the first deployment, excluding data where Cl 2 is present, strongly suggests that a measurement error exists in one or both of the reservoir chlorine species, ClONO 2 and HCl. The concentration of reservoir chlorine is expected to increase in proportion to the decrease in concentration of active chlorine when recovery occurs. For SOLVE/ THESEO, however, as ClO x becomes a smaller fraction of total inorganic chlorine during the recovery period, the Figure 13. Inorganic chlorine budget as a function of Cl y for the two extravortex flights: (a) 0127 and (b) Data inside the vortex are shown by solid circles, and data outside the vortex are shown by shaded circles. The dotted lines at 1.0 in each panel represent complete budget agreement. budget discrepancy grows. The most straightforward means to confirm that a measurement error exists in one or both of the reservoir chlorine species is to examine the extravortex data from the mission. Outside the vortex, ClONO 2 + HCl composes 95% of Cl y, so any significant extravortex budget discrepancy implies a ClONO 2 or HCl measurement error. [51] Figure 13 shows the inorganic chlorine budget as a function of Cl y for the flights of 0127 (Figure 13a) and 0311 (Figure 13b). These are the only 2 flights of the mission for which extravortex segments are available; there are no HCl measurements for any of the transit flights or the pre- SOLVE/THESEO January flights in the US. The data shown here were obtained at cruise altitude, with measurements inside the vortex shown as solid circles and the measurements outside the vortex depicted as shaded, open circles. The dotted line at 1.0 in each panel represents complete budget agreement. [52] The inorganic chlorine budget for 0127 is in excellent agreement inside the vortex, having a mean value of 0.95 ± However, the budget outside the vortex is significantly worse, having a mean of 0.76 ± Since these measurements were taken during the same flight, the only explanation for this discrepancy is the change in sampled air mass to one with more ClONO 2 and HCl and less ClO x. The intravortex data were acquired both before and after the extravortex portion, eliminating any possible 11 of 16

12 temporal issue in instrument performance. The 0311 flight yields a similar result to 0127, but with even lower mean values. From these data, it is clear that the inorganic chlorine budget is worse outside than inside the vortex, and more generally, budget agreement is only attained when ClO x composes the vast majority of Cl y. These results, in combination with the increasing budget shortfall as chlorine recovers inside the vortex, support the existence of a measurement error in ClONO 2 or HCl. [53] An independent assessment of the ClONO 2 observations is provided by comparison with a photochemical steady state (PSS) calculation of ClONO 2 using measured ClO, NO 2, temperature, pressure, the production rate constant, and the photolysis rate. The steady state ClONO 2 concentration is defined as ½ClONO 2 Š ss ¼ k ClOþNO2½ClOŠ½NO 2 Š J ClONO2 ð23þ where J ClONO2 is the photolysis rate [Salawitch et al., 1993, 1994], and k ClO+NO2 is the ClONO 2 production rate constant [Sander et al., 2003] from (11). NO 2 concentrations are obtained from an instrument mounted at the rear of the Harvard Halogen instrument, which detects NO 2 with a laser-induced fluorescence technique [Perkins, 2000; Perkins et al., 2001]. [54] Figure 14 shows a comparison of steady state calculated and measured ClONO 2 for the region of the 0311 flight between 9.0 and 12.5 hr UT. This is the extravortex segment of the flight, along with a few points in the transition region. Because measured NO 2 concentrations are very small and relatively uncertain inside the vortex, analysis of extravortex data provides the most reliable values to test the veracity of the ClONO 2 absolute calibration. Measured ClONO 2 is shown as a solid line in Figure 14a, and steady state ClONO 2 is represented as shaded, open circles. Figure 14b depicts the ratio of steady state to measured ClONO 2, with the solid line at 1.0 representing complete agreement. [55] The results in Figure 14b indicate that steady state ClONO 2 values are higher than the measured values by a factor of 1.13 ± 0.28 on average. This represents agreement within the experimental uncertainty of the measurements and serves as a validation of the ClONO 2 measurement accuracy. This result is also consistent with the data from the POLARIS mission in which steady state ClONO 2 values were higher than those measured by a factor of 1.15 ± 0.36 on average for the bulk of the data (80%) and a factor of 1.01 ± 0.30 for the remainder of the data, depending on temperature and latitude [Stimpfle et al., 1999]. Even if measured ClONO 2 for flight 0311 is multiplied by a factor of 1.13, it has only a small impact on the budget, improving agreement from 0.64 to [56] Intravortex steady state calculated ClONO 2 also agrees well with the measured values for all of the regions in the second deployment flights where measured NO 2 is sufficiently large to be reliable. On average, the calculated ClONO 2 values are 13% higher than those measured. This result suggests that ClONO 2 is not the source of the late winter budget discrepancy. Further support for the accuracy of steady state ClONO 2 is derived from the fact that the measured NO 2 values used here agree well with NO 2 Figure 14. (a) Comparison of measured ClONO 2 (solid line) and steady state ClONO 2 (shaded circles) for the extravortex segment of the 0311 flight. (b) Ratio of steady state to measured ClONO 2. The line at 1.0 represents complete agreement. On average, steady state ClONO 2 is 13% higher than measured ClONO 2. calculated from a steady state expression involving measured NO (F. Keutsch, manuscript in preparation, 2006). [57] The implication of this analysis is that measured HCl is the primary source of the observed budget discrepancy during the second deployment of SOLVE/THESEO. This is consistent with the work of Bonne et al. [2000], which found that measurement errors in HCl were likely the cause of shortfalls in the inorganic chlorine budget on several previous field campaigns. The presence of significantly higher intravortex HCl is inconsistent, however, with the generally accepted picture of chlorine recovery in the Arctic vortex in which ClONO 2 forms preferentially before significant HCl formation [e.g., Webster et al., 1993; World Meteorological Organization, 1995, Figure 3.1]. HCl is believed to be small and roughly constant in the vortex throughout the winter, which is what was observed during this mission, as shown in Figure 5. To pursue these differences further, we are interested in reassessing what might be expected, based upon the mechanism defined by (11 16), for the relative recovery of ClO x into ClONO 2 and HCl from highly activated conditions Modeling Inorganic Chlorine Recovery [58] A diurnal box model is constructed and used to examine the ClONO 2 and HCl recovery rates that are feasible for several different photochemical mechanisms. For simplicity, the volume element in the model is fixed in space at a particular latitude and longitude, at a selected 12 of 16