Ozone over McMurdo Station, Antarctica, austral spring 1986 - Altitude profiles for the middle and upper stratosphere

In the austral spring of 1986, a program of measurements of the ozone altitude profile (for the z values between 25 and 55 km), relevant to an understanding of the ozone hole, was conducted at McMurdo Station, Antarctica. The measurements were
of 10
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
  JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 92, NO. Dll, PAGES 13,221-13,230, NOVEMBER 20, 1987 Ozone Over McMurdo Station, Antarctica, Austral Spring 1986' Altitude Profiles for the Middle and Upper Stratosphere BRIAN J. CONNOR, J. W. BARRETT, . PARRISH, P.M. SOLOMON, R. L. DE ZAFRA, AND M. JARAMILLO State University of New York at Stony Brook The ozone ltitude profile or 25 _< _< 55 km was measured rom McMurdo Station, Antarctica, by ground-based illimeter-wave pectrometry, n 34 days between September 2 and October 29, 1986. The mixing ratio peaked at altitudes anging rom 28 to 34 km, with peak values between 5 and 9 parts per million by volume (ppmv). The ozone mixing ratio at 25 km de- creased y 15 q- 5% (la uncertainty) uring he period. No significant ecrease•was bserved at higher altitudes; he upper imit on any secular rend at 33 or 40 km is -•8%. Considerable variability was observed t all altitudes, on time scales of I to a few days. In the most dra- matic case, ozone n a layer 25 _• z _• 40 km nearly doubled between September 0 and 23. This was accompanied y a significant warming and an increase n N20 at the same altitudes. We compare our results o concurrent alloonsonde zone measurements, o a model prediction, and, by combining ur results with the balloonsonde rofiles, o satellite measurements f total ozone. The observation hat ozone depletion was confined o the lower stratosphere s consis- tent with theories nvoking heterogeneous hemistry associated with polar stratospheric louds. It may be consistent ith theories which predict upwelling f ozone-poor ir, if such upwelling is confined o the lower stratosphere. Our data do not support theories predicting depletion n the upper stratosphere, ut the seasonal overage f the measurements as too limited to be definitive. 1. INTRODUCTION The column density of ozone over Antardtica n the spring has declined dramatically during the past decade. In re- cent years mean values during October have been approx- imately two-thirds those which were typical between 1957, when egular measurememts egan, and the mid-1970s Far- man et al., 1985]. The phenomenon, idely known as the Antarctic ozone hole, has been observed to be continent- wide, to begin no later than the end of polar night, and to persist until the breakdown f the polar vortex [$tolarski t al., 1986]. The precipitous decline in ozone has become the focus of a great deal of interest and concern. Important and general issues raised include the following. First, that the decline was totally unexpected by the atmospheric science commu- nity, notwithstanding over a decade of extensive research into stratospheric ozone, raises the possibility that some critical element is missing from our understanding of the stratosphere. We must learn if we can what implications the Antarctic ozone hole has for global, stratospheric ozone. Second, t is of paramount importance to know whether the ozone hole is of natural or anthropogenic srcin, for if hu- mankind caused t, we may still have the option of reversing the process. Any investigation of the Antarctic ozone hole proceeds against the background of its basic phenomenology. Nor- 1 Now at Science & Technology Corp., Hampton, VA. 2 Also at Millitech Corp., S. Deerfield, MA. Copyright 1987 by the American Geophysical Union. Paper number 7D0693. 0148-0227 87 007D-0693 05.00 mally, column ozone amounts are higher in the polar regions than at lower latitudes. Further, the ozone concentration peaks at a lower altitude, typically 15-20 km. The polar ozone mixing ratio profile peaks at 30-35 km, as at lower latitudes, but the peak value is much less, --•6 parts per million by volume ppmv) as opposed o --•10 ppmv in the tropics. Near the poles, column ozone undergoes a distinct annual cycle characterized by an increase to very high val- ues in midspring, when the polar vortex breaks up, followed by a gradual decline through the summer. A minimum is reached in early fall, which is, however, only slightly lower than the values which persist through the winter and into early spring. The development of the ozone hole has distorted the pre- ceding picture as follows. In a single year the ozone hole consists of a decline in total ozone, which begins no later than first sunrise, to values lower than those reached in the fall; indeed, in recent years, to the lowest values ever recorded anywhere. The breakdown of the vortex then fol- lows in normal sequence, apparently ending the anomaly for that year. On a year-to-year basis this process has been observed to repeat itself, becoming on average more pro- nounced with successive years, so that "the hole becomes deeper" with time. The great bulk of the springtime ozone depletion occurs in the region of peak concentration in the lower stratosphere, resulting in an upward shift of the al- titude of the peak of about 5 km [Hofmann et al., 1987]. Whether ozone depletion also occurs in the middle and up- per stratosphere s disputed and has important implications for any theoretical explanation of the ozone hole. In the austral spring of 1986 we performed, at McMurdo Station, Antarctica, a program of measurements relevant to an understanding of the ozone hole. Our ozone measure- ments were made as part of that larger program of mea- 13,221  13,222 CONNOR ET AL.' ANTARCTIC OZONE Ozone, 276.9 GHz 10 ............................. -150 -100 -50 0 5O 100 150 MHz ?rom I•ne cenñer Fig. 1. Ozone spectra observed September 20 (the weaker of the two) and September 23, 1986. In addition to the 144 10- 153 13 line at band center, there is a much weaker ine at approximately +115 MHz. surements with the same instrument. We have elsewhere reported our observations f chlorine monoxide Solomon t at., 1987 and de Zafra et al., 1987] and nitrous oxide (A. Parrish, R. L. de Zafra, M. Jaramillo, B. J. Connor, P.M. Solomon, and J. W. Barrett, Observation of extremely low N20 concentrations in the springtime stratosphere at Mc- Murdo Station, Antarctica, submitted to Nature, 1987, here- after called submitted manuscript 1987a). 2. MEASUREMENTS 2.1 Apparatus and Technique The experimental apparatus is a heterodyne radiometer, with quasi-optical local oscillator injection and a filter bank spectrometer. The radiometer operates at approximately 265-280 GHz; the filter bank has a total bandwidth of 256 MHz and a 1-MHz resolution. Both the instument itself and the observing technique are described n detail by Parrish et al. (A. Parrish, R. L. de Zafra, P.M. Solomon, and J. W. Barrett, A ground based technique for millimeter-wave spectroscopic observa- tions of stratospheric trace constituents, submitted to Radio Science, 987, hereafter alled submitted manuscript 987b). In brief, thermal emission rom the stratosphere s observed at two very different elevation angles. The ratios of the ob- servations at the two angles are the raw data, which are converted to absolute intensity units and corrected for con- tinuum absorption by tropospheric water vapor. The installation of the instrumen[ at McMurdo differed from that described y Parrish et al. (submitted manuscript, 1987b) n that the instrument was ocated ndoors and ob- served the sky through two Teflon windows. This modifi- cation will be described n a forthcoming paper; aside from contributing to the comfort of the operator, the indoor in- stallation greatly improves the thermal stability of the in- strument's environment and is thereby believed to improve observational precision. Millimeter-wave spectral lines have previously been ob- served from several constitutents of the middle atmosphere, including 03 [Penfield et al., 1976; Wilson and Schwartz, 1981; Connor, 985], H20 [Radford t al., 1977; Bevilacqua t al., 1983; Bevilacqua t al., 1985a], N20 [Connor t al., 1987], CO [Kunzi and Carlson, 982; Bevilacqua t al., 1985b], C10 [Parrish et al., 1981; Solomon t al., 1984], HO2 [de Zafra et al., 1984], and HCN (R. de Zafra, unpublished ata, 1986). In addition, de Zafra et al. [1985] derived an upper limit on stratospheric H202 from ground-based millimeter-wave observations. 2.2 The Observed Transition There are a number of 03 lines near 1-mm wavelength which are suitable for ground-based measurements. Even the strongest 03 lines at that wavelength are fairly opti- cally thin, so that the emission depends approximately lin- early on the concentration. Further, the molecular states involved typically have excitation energies comparable to or less than collisional energies at atmospheric temperatures, so that the population of the states depends only weakly on temperature. Because of this, accurate knowledge of the at- mospheric temperature profile is not critical to the retrieval of ozone profiles. We discuss hese points in more detail in the appendix. The observations reported here are of the 144 10- 153 •3 pure rotational transition of 03 in its ground vibrational state. The transition requency s 276.92353 GHz [Poynter and Pickett, 1985]. The 144 10- 153 13 transition s near optimum for ground-based observing. It may be observed at signal-to-noise atios (determined by the amplitude at line center) approaching 00 n 10- to 20-min ntegration, et at the elevation angles used in practice, rv _• 0.15. Also, its temperature dependence is particularly weak. Detailed cal- culation shows hat shifting a standard atmospheric temper- ature profile uniformly by + 1 K changes he line center am- plitude, as observed n a 256-MHz bandwidth, by +0.03%. We conclude that retrieval errors due to uncertainty in the atmospheric temperature profile are negligible. 2.3 Observations Observations were made on 34 days between September 12 and October 29, 1986, or on nearly every day when the weather and the demands of the rest of the observing sched- ule allowed. Time elapsed during a single observation ranged from approximately 15 min to 2 hours. Two typical spectra, recorded September 20 and Septem- ber 23, appear in Figure 1. The difference in amplitude of these spectra will be discussed n section 4. They are presented here to illustrate the signal-to-noise atio typical of the entire data set; measured relative to the line center amplitude, the signal-to-noise ratio is 100-200. The absolute accuracy of the spectral calibration is dis- cussed n detail by Parrish et al. (submitted manuscript, 1987b), where t is given as 12%. The day-to-day alibration precision s estimated at 4%. 3. RETRIEVAL OF OZONE ALTITUDE PROFILES Equation (1) in the appendix defines he observed ig- nal. A substantial literature exists discussing etrieval of re- motely measured quantities by inversion of such equations. References he authors have found particularly useful in- clude Twotory 1977], which reats the general problem at length; Rodgers 1976], which presents n excellent discus- sion of retrieval of atmospheric parameters n particular; and Bevilacqua t al., [1983], which details he application f one  CONNOR ET AL' ANTARCTIC OZONE 13,223 Ozone, 276.9 GHz 15 .... • .... • .... • .... i .... i .... 10 0.25 ...... I ....... I ........ i .... i .... i .... (b) 0.00 -0.25 ............................. -150 -100 -50 0 50 100 150 MHz trorn line center Fig. 2. (a) The ozone spectra observed September 23, 1986, su- perimposed with a synthetic spectrum calculated from a retrieved ozone altitude distribution. (b) The difference of the two spectra shown in Figure 2a. inversion method to ground-based microwave measurements of water vapor. Ground-based millimeter-wave measurements with high signal-to-noise atio, such as those presented here, may be inverted with an altitude resolution of 5-8 km, depending on how esolution s defined Parrish, 986]. The 5-km value de- fines the full width at half maximum of the retrieved profile, if the true profile is assumed o be a single very thin layer. The formal resolution is 8 km; that is, the minimum sepa- ration between two thin layers which can be resolved in the retrieved profile. The upper and lower altitude limits at which such reso- lution may be achieved are determined in the present case by the frequency resolution and total spectral bandwidth of the instrument, respectively. The limits are approximately 52-55 km at the upper end and 22-23 km at the lower end. While these limit the range over which significant altitude discrimination is possible, note that an important contri- bution to the signal may come from outside this range. In particular, ozone as low as z 0• 15 km will contribute a sig- nificant signal within our observed band. The retrievals presented here are derived by the optimal estimation method of Rodgers 1976]. This method allows explicit use of a priori knowledge of the distribution being measured, statistically combining the data and an a priori profile. In practice, the main effect of the a priori profile, within the altitude range where the measurement is most sensitive 25-55 km), is to cause he solution profile to be smooth. Below 25 km, however, the a priori profile has more influence on the solution and, indeed, dominates it below 20 km, as the intrinsic altitude resolution of the measurement decreases. f the a priori profile at 15-20 km does not resem- ble the true profile, small errors are produced in the solution profile at higher altitudes. Therefore the selection of the a priori profile above 25 km is not critical, while the choice of that profile at lower altitudes is more important. With the above considerations n mind, we performed two sets of retrievals. For the first set, we used the same a priori profile for all dates, which profile was based on measure- ments of Holmann et al., [1986, 1987] or z < 30 km and on the reference rofiles f Keating nd Young 1985] or z > 30 km. A second, improved set of retrievals used a different a priori profile for each date. The electro-chemical concen- tration cell (ECC) ozone measurements y a University of Wyoming group Hofrnann t al., 1987; private communica- tion, 1987], which were made at McMurdo Station during the same period as our own measurements, were used to derive the a priori profiles for the second set of retrievals. Each Wyoming profile was assumed correct to its maxi- mum altiude (usually 20-30 km), and was smoothly oined at higher altitudes to the a priori profile used for the first set of retrievals. The resulting composite profiles were then smoothed to remove small-scale vertical structure to which our measurement is not sensitive, and missing dates were filled in by interpolation. The net effect was to produce, for each date, an a priori profile which, in the lower strato- sphere, was based closely on actual measurements made at approximately that date and location, while its upper strato- spheric portion was a climatological average. Thus each a priori profile incorporates the most information where it is most needed, i.e., in the lower stratosphere, where the millimeter-wave measurement is least sensitive. All the retrieved profiles presented here are from the sec- ond set of retrievals, in which the solution was forced to equal the a priori profile for z _< 20 km. They may be thought of as being based entirely on the Wyoming mea- surements for z _< 20 km, entirely on our millimeter-wave measurements for z _> 25 km, and depending on both mea- surements for 20 < z < 25 km. In general, spectra calculated from the retrieved profiles fit the data very well. Figure 2a shows he data of September 23 with a superimposed synthetic spectrum calculated from the retrieval. Figure 2b shows the difference between these two spectra. Any systematic effects are less than the noise. While the preceding statement could also be made for most of our other measurements, in some cases systematic effects, probably instrumental in srcin, are visible in the difference spectrum corresponding to Figure 2b. In all cases, however, such effects are quite small, <_1% of the line amplitude. A summary of the estimated errors in the measured ozone profiles is shown in Table 1. We emphasize that these esti- mates apply specifically to the observations n this paper. Errors resulting from the retrieval process were estimated by performing simulated retrievals. An example of such is shown in Figure 3. A realistic ozone profile was assumed, and a synthetic spectrum calculated from it. Fifty different spectra of normally distributed noise (each with the same variance) were hen added o the synthetic ozone spectrum, producing 50 different simulated observations. The assumed variance was typical of the real data. Altitude profiles were then retrieved from the simulated observations using an- other realistic ozone profile as the a priori profile. Figure 3 displays the mean of the 50 resulting profiles, their 5:1a lim- its, and their extreme values, along with the assumed and a priori profiles. We note first that the retrieved profiles closely follow the assumed profile and, second, that the er- rors in the retrieved profiles are of two types, scatter about the mean and deviations of the mean from the assumed pro- file. We refer to the latter errors as retrieval bias. Several sets of simulations, such as those shown in Figure 3, were performed for different combinations of a priori and assumed profiles. The retrieval bias errors quoted in Table i are ap- proximately the maximum bias errors in the various sets of simulations. The retrieval scatter figures are the average values of the 2a scatter in the various sets. If the true ozone profile remains fixed while repeated mea-  13,224 CONNOR ET AL.: ANTARCTIC OZONE TABLE 1. Ozone Altitude Profile Errors Altitude, km 25 35 45 55 Variable Errors Instrument Calibration 4 4 4 4 Retrieval Scatter(2rr) 2 4 6 8 Instrument Baseline 4 i ...... Retrieval Bias 5 5 10 10 Net Measurement Precision a 6 6 7 9 rss of Variable Errors a 8 8 12 13 Fixed Errors Instrument Calibration 12 12 12 12 Pressure Broadening 2 5 6 6 Dipole Moment 2 2 2 2 rss of Fixed Errors 12 13 14 14 Net Measurement Accuracy 14 15 18 19 (rss of All Errors) All errors are stated in percent; rss stands for root-sum-square. a The net precision is the predicted scatter among repeated measurements of the same ozone profile; it is the rss of the variable errors excluding retrieval bias. The accuracy with which measurements of different profiles may be compared is the rss of all variable errors. surements are made, the retrieval bias error will be identi- cal for each measurement. Therefore retrieval bias does not limit measurement precision in the strict sense. However, if measurements of different ozone profiles are to be compared, retrieval bias must be included in assessing he accuracy of the comparison. The appropriate figures for that purpose are given separately in Table I as the roots of the sums of the squares of the variable errors. The error due to a nonlinear instrumental baseline was precision anges with altitude from 6-9%, while the absolute accuracy anges rom 14-19%. 4. RESULTS The retrieved ozone distributions are shown in Figure 4 in the form of a contour diagram showing mixing ratio in ppmv as a function of date and altitude. The most obvious features are the large variations. For the most part these also erived sing imulated etrievals. he magnitude f appear o be luctuations ccuring n a time cale f a few likely aseline rtifacts as irst stimated y nspection f days. he most ronounced uch luctuation s he zone n- the residuals from the retrievals on the actual observations. Assumed baseline curves, consistent with the actual resid- uals, were then added to a synthetic spectrum calculated from a typical ozone distribution. After adding noise to the result, a retrieval was performed. The resulting profile was compared to a retrieval from the same spectrum without the added baseline curve. The difference in the two profiles is the estimated error due to instrumental baseline effects. Error from uncertainty in the pressure broadened line width is the last component of the error budget derived by simulation. The pressure-broadening coefficients used in our calculations are based on measurements of N2- and 02- broadening f four rotational ines of 03 [Connor and Rad- ford, 1986], and on calculations f N2-broadening f several thousand 3 lines by Garnache nd Rothman, 1985]. (Cal- culations f O2-broadening eem much ess ccurate Connor and Radford, 986].) The average alue of the ratio of mea- sured air-broadening coefficient to calculated N2-broadening coefficient for the four measured lines is 1.03. This factor was multiplied by the calculated coefficients for all lines to derive the coefficients Used n our analysis. The assumed un- certainty is the experimental uncertainty quoted by Connor and Radford, about 5:4%. Retrievals were performed on synthetic spectra with noise added) using both the usual pressure-broadening oefficient and that coefficient plus 4%. The difference in the retrieved profiles is the retrieval error due to pressure broadening shown in Table 1. We end our discussion of retrieval errors by reiterating (section ) that errors due o uncertainty n the atmospheric temperature are negligible for the particular transition we observe. Finally, we note from Table I that the measurement crease between September 20 and 23. The spectra recorded on these days have already been shown Figure 1). In Fig- ure 5 we show the corresponding retrieved profiles. It may be seen that the ozone increased dramatically between 25 and 40 km, roughly doubling at 35 km. We have reported elsewhere Parrish et al., submitted manuscript, 987a) he SIMULATED OZONE RETRIEVALS 60 •... , , • , i , , , , , I•.1 '.• uJ 40 <• 0 -.' .......•y - 20 • '"•' 0 2 4 6 8 I0 12 MIXING RATIO (ppmv) F•g. 3. •esults of a statistical test of the retrieval process used on the ozone spectra. The so]•d line •s •sumed to be the true ozone profile. The dotted line is the a priori profile used for the retrieval. The d•hed line is the mean of retrievals from 50 simulated observations. The error bars show the •1½ scatter of the retrieved profiles, and the •ertJca] bars show their extreme  CONNOR ET AL.' ANTARCTIC OZONE 13,225 6O 55 5O 45 :3 35 "• 30 OZONE, McMURDO STATION, 1986 -"X..., /' .... "'N \ /x _ A ....... .... .... .... ,,": ,"\ , ,, //.r ... :,", ........ .v,,\ ,t ',', ,j .• J '...>, .... ', V .... \...,: .... ..'>j .... .... ','V/ ---'•'4'---x'"'- .... %' .... ' ..- /'X/' V ' ".. '1' 5 15 I0 15 20 25 30 5 I0 15 20 25 30 SEPTEMBER OCTOBE R Fig. 4. Contours of ozone mixing ratio (ppmv), derived from the millimeter-wave data, versus date and altitude. Note that values below 20 km are derived from the University of Wyoming measurements. observation, on September 23 and 24, of a large increase in N20 at the same altitude. Calculations of potential vortic- ity supplied us by R. L. Jones of the British Meteorolog- ical Office private communication, 987) indicate hat on September 23 an airmass srcinating at a latitude _• 50 ø S was at 35-40 km over McMurdo. The observed mixing ratios of both O3 and N20 in this airmass are consistent with a mid-latitude srcin. The high ozone values persisted until September 27, then declined over the next few days to levels comparable to those of September 20. In Figure 5 we also present a profile measured October 24. This profile and that of September 20 illustrate the observed extremes in the altitude of the peak mixing ratio, approximately 28-35 km. Also shown, for comparison, is a profile measured in Hawaii on May 24, 1986, which may be taken as representative of a normal tropical ozone distribution. In Figure 6 we show a contour plot of temperature with axes corresponding to Figure 4. The temperature field was provided by the NOAA Climate Analysis Center (R. Na- gatani, private communication, 987). Comparison f Fig- ures 4 and 6 shows that ozone fluctuations are sometimes correlated with temperature changes and sometimes are not. The substantial warming of September 21-24 accompanied the large increase in middle stratospheric ozone discussed earlier, although the temperature increase extends to lower altitudes than the ozone increase. Conversely, the cool- ing trend of October 14-16 coincides with decreasing ozone mixing ratios. In contrast, the cooling trends of Septem- ber 25-27 and October 5-7 do not seem to be associated with changes n ozone. One significant feature of Figure 6, namely, the warming of September 30 to October 3, reached its peak during a storm which orced a 3-day (October 2-4) hiatus in the ozone record, so we cannot say if there were corresponding ozone changes. However, the typical duration of ozone iuctuations•which o appear n Figure suggests that a significant change n ozone during October 2-4 would have persisted long enough to be observed on October 5. Inspection of Figure 4 also shows hat ozone variations in the middle and upper stratosphere may or may not be corre- lated with changes n the lower stratosphere. In the period October 14-16, ozone decreased at all altitudes from 15 to 50 km. Conversely, during September 20-23, the large increase in ozone in the middle stratosphere has no counterpart at lower altitudes. A most importan• question is whether there was a trend in ozone concentration at a given altitude during the period of observations. While observations from a single location cannot unambiguously follow the chemical evolution within a single air parcel, one hopes that over a period of several weeks meteorological fluctuations will average out, and the secular behavior will reveal the development of the ozone hole, showing both when and at what altitudes depletion occurs. However, in assessing he following observations of secular trends, one must bear this limitation in mind. We observe a secular decrease in ozone at 25 km but not at higher altitudes. Figure 7 shows our measured ozone at 25, 33, and 40 km as a function of date. The decline in ozone at 25 km is statistically significant. A linear regression using all the points shown in Figure 7 yields a change in ozone over 43 days of -15 q- 6% (la uncertainty). We believe t best, however, to exclude from the regression the measurements of September 23, 26, and 27, since these are representative of lower latitude air, and also the four measurements af- ter October 24, since it appears the final warming began shortly after that date. (This is best seen n the University of Wyoming data in Figure 8, which will be discussed n section 5.) If the above seven measurements re excluded, a linear regression on the remaining data yields a nearly identical result, namely -15 q- 5%. In contrast, the corre- sponding result from the millimeter-wave measurements at 33 km is -2 q- 8%; at 40 km, it is 0 q- 8%. Also shown in Figure 7 are the University of Wyoming measurements veraged between 15 and 20 km (D. J. Hof- mann, private communication, 987), which show a steep. progressive decrease in ozone, as reported by Holmann et al., [1987]. One may ask whether the use of the Univer- sity of Wyoming measurements in the a priori profiles for the retrievals influences the observation of the trend. In 6O 5O • 4O •, •o 2O OZONE •\1 ... \ I I. I I I I I • •'•'•.'"'"' x McMurdo /20/8t5 - ',.x• .... , McMurdo /23/86 ... • ....... -, McMurdo 0/24/86--- - •.:. Hawaii 124/86 --- - •.x '.. .•..• x ...... • • ...... .• • '.... •xx , 2 4 6 8 IO MIXING RATIO (ppmv) Fig. 5. Several ozone altitude profiles derived from millimeter- wave observations. The Antarctic profiles were chosen o illus- trate the extremes of variation observed. The Hawaiian profile is shown for comparison. Values below 20 km are based on the University of Wyoming measurements.
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks