POAM III measurements of dehydration in the Antarctic and comparisons with the Arctic



[1] We present measurements of stratospheric water vapor taken at high northern and southern latitudes by the Polar Ozone and Aerosol Measurement (POAM) III instrument. The measurements show the seasonal variations in both hemispheres, including wintertime descent in the polar vortices and dehydration in the Antarctic. While POAM is unable to measure in regions with high aerosol extinction and can therefore not measure the temporary sequestration of water in ice clouds, POAM can measure the effects of such dehydration events if they result in the permanent removal of water vapor from an air parcel. A study of POAM “matches,” where the same parcel is observed more than once by the POAM instrument, shows significant permanent dehydration in the Antarctic for nearly all cases in which the parcel is either exposed to conditions where the mixing ratio either exceeds the saturation mixing ratio by >2 ppmv or experiences temperatures that remain below the saturation temperature for >2 days. These POAM matches suggest that parcels undergo several episodes of dehydration before their water vapor mixing ratio is reduced to the 1–2 ppmv level typically observed within the vortex in the Antarctic spring. Interannual differences in Antarctic vortex temperatures are shown to be qualitatively consistent with interannual differences in dehydration observed by POAM. Although analyses of parcels observed by POAM in the Arctic indicate that a few of these parcels experience conditions which generally cause permanent dehydration in the Antarctic, no significant permanent dehydration is observed in the POAM Arctic measurements.

1. Introduction

[2] During cold polar winters, rapid changes in stratospheric water vapor mixing can sometimes occur, as water freezes and sometimes falls out of the stratosphere altogether [Kelly et al., 1989]. This polar dehydration has a large effect on the water vapor budget of the Antarctic vortex lower stratosphere. At some altitudes, where water vapor mixing ratios ∼6–7 ppmv would be expected in the absence of dehydration, mixing ratios as low as ∼1 ppmv have been measured over monthly timescales [Nedoluha et al., 2000].

[3] The process that causes polar dehydration can have an important effect on ozone mixing ratios. Ice particles that are formed in the stratosphere and then precipitate will remove not just water, but nitrogen as well. Denitrification plays an important role in ozone depletion in polar regions [Crutzen and Arnold, 1986; Solomon, 1999; Tabazadeh et al., 2000b].

[4] Temperatures in the Arctic winter are generally not cold enough to freeze water vapor in the lower stratosphere. However, during the winter of 1999–2000 temperatures were unusually cold, and several measurements during the SAGE III Ozone Loss and Validation Experiment (SOLVE) campaign did indicate dehydration, with Herman et al. [2002] detecting loss of ∼0.2 ppmv and Schiller et al. [2002] finding dehydration of up to ∼0.5 ppmv.

[5] Intensive measurement campaigns such as SOLVE provide a unique opportunity to study in detail the physics and chemistry of the stratosphere in a specific region over a few months. The continuous measurements of stratospheric water vapor at high northern and southern latitudes provided by the Polar Ozone and Aerosol Measurement III (POAM III) instrument provide the opportunity to study seasonal variations in this region over several years. In this paper we shall first show the results of measurements in both hemispheres since 1998, and discuss the interannual variations in Antarctic dehydration. We shall then use trajectory analyses to focus on two specific winters: the Antarctic winter of 1998, and the Arctic winter of 1999–2000. After studying the temperature histories of the air parcels measured by POAM III and identifying instances where POAM measures the same parcel of air more than once, we draw some conclusions about the timescales and levels of supersaturation generally required to cause permanent dehydration.

2. POAM III Measurements

[6] The POAM III instrument, like the POAM II instrument that made measurements from 1993–1996, is a visible/near infrared solar occultation photometer that measures stratospheric constituents in the polar regions of both hemispheres. First measurements were obtained on 17 April 1998. The POAM III measurement complement includes O3, water vapor, NO2, and aerosol extinction. More detailed descriptions of the POAM III instrument and of early validation results are given by Lucke at al. [1999].

[7] The POAM III instrument makes 14 measurements per day in each hemisphere around a circle of latitude, and the latitude of the observations varies with an approximately semi-annual period. In Figure 1 we show the latitudes and equivalent latitudes [Butchart and Remsberg, 1986] of the POAM measurements on the 550 K potential temperature surface based on the UKMO analysis. We also show the vortex center edge threshold value obtained using the Nash et al. [1996] vortex edge analysis. As Figure 1 shows, the POAM measurements in the Southern Hemisphere (SH) are almost completely inside the vortex from July through October. The Northern Hemisphere (NH) POAM measurements are taken at somewhat lower latitudes, and this, coupled with the tendency for the NH vortex to be centered off the pole, allows POAM to sample air masses both inside and outside the vortex on a nearly daily basis. Figure 1 also shows a decrease in the density of measurements beginning on 25 March 1999. Since this date, measurements have generally been taken in only one hemisphere per day in order to reduce wear in the azimuthal pointing mechanism.

Figure 1.

The latitudes (red line) and equivalent latitudes (green crosses) of the POAM measurements at 500 K. Also shown is the latitude of the vortex edge (blue line).

[8] Water vapor is a new POAM retrieval product that was not provided by POAM II. The POAM III version 2 retrieval algorithm water vapor measurements were validated against several instruments in the SPARC Water Vapour Working Group [2000]. The version 3 retrievals used in this paper have been validated against HALOE in the stratosphere (R. M. Bevilacqua et al., Validation of POAM stratospheric water vapor, in preparation, 2001), and against MOZAIC and MLS measurements in the upper troposphere/lower stratosphere (UT/LS) [Nedoluha et al., 2002]. These comparisons showed agreement to within ∼10% from the upper troposphere to the upper stratosphere. The vertical resolution is ∼1 km up to ∼40 km. Above this altitude the resolution broadens and the a priori dependence increases significantly [Lumpe et al., 2002]. The operational retrievals are performed on a 1 km grid.

[9] An important factor that can adversely affect the POAM retrievals of water vapor is the presence of sunspots in the field-of-view. The presence of a sunspot will reduce the intensity of the POAM measurement, thus mimicking the effect of atmospheric absorption [Lumpe et al., 2002]. Because the instrument performs a scan of the entire solar disk above the atmosphere during each measurement the sunspots have, in principal, been accounted for in the retrieval process. Thus, in most cases, the retrievals are reliable even in the presence of sunspots. However, if the sunspot is large, then even small uncertainties in the POAM pointing will cause significant uncertainties in the calculated atmospheric absorption. Measurements made in the presence of large sunspots have therefore been removed from the data set. Because of increasing solar activity during the POAM III observation period, the number of large sunspot events has increased, causing an increase in the number of water vapor measurements which are removed from the data set.

[10] Another factor that increases the uncertainty in the POAM water vapor retrievals is the presence of high aerosol extinctions. We find that the average retrieved water vapor mixing ratio at a given altitude shows no clear systematic bias with variations in aerosol extinction, but that the standard deviation does generally increase with increasing aerosol extinction. Estimates of the error in the water vapor retrieval from aerosol contamination are given by Lumpe et al. [2002]. We have removed from the data set used here those cases for which the aerosol extinctions exceeds a threshold value defined by Nedoluha et al. [2002]. The difficulty of making accurate measurements in regions of high aerosol extinction limits POAM lower stratospheric water vapor observations in the presence of PSCs. Water vapor can sometimes be retrieved reliably despite the presence of Type 1 PSCs (PSCs which contain HNO3 as a major component), provided they are not too optically thick [Nedoluha et al., 2000]. However, Type 2 PSCs (PSCs which contain particles consisting primarily of H2O ice) increase the optical depth to the point where the retrieval suffers an abnormally high altitude termination [Fromm et al., 1999]. The largest amount of data loss from PSC contamination occurs at ∼20 km during July and August in the SH, when 63% of the POAM water vapor measurements are removed from the data set. Averaged over all of the POAM data, 21% of the measurements at 20 km have been removed because of aerosol contamination. Despite the loss of a significant number of observations during periods when PSCs are prevalent, there remain adequate numbers of observations to describe the seasonal behavior of stratospheric water vapor throughout the year.

3. Results

[11] In Figure 2 we show the POAM water vapor measurements for both the NH and SH. The contour plots for both hemispheres show the increase in water vapor with increasing altitude in the stratosphere. This increase is caused primarily by the oxidation of methane. The decrease in water vapor with increasing altitude in the upper stratosphere results from Lyman-α photodissociation. Both the NH and SH POAM measurements clearly show the descent of the peak in the water vapor mixing during the winter. As was shown in Figure 1, the SH observations are confined to the vortex for much of the winter. Figure 2 shows the descent of air within the vortex very clearly. The NH results shown in Figure 2 include measurements from both inside and outside the vortex. Because of the difference in descent rates inside and outside the vortex, the descent of the peak of the mixing ratio in Figure 2 is not as clearly defined in the NH as it is in the SH. Also, a significant portion of the apparent descent in the NH occurs because the POAM measurements move northward during the fall, and the altitude of the peak of the profile drops with increasing latitude. In the spring the peak near the stratopause reappears in the POAM measurements, as air from lower latitudes is transported poleward. In the SH spring the profile shows a clear double peak in the stratosphere. One of these peaks is associated with air that has descended in the vortex, while the other is associated with air that is moving toward the pole from midlatitudes [Nedoluha et al., 2000].

Figure 2.

The water vapor mixing ratio measured by POAM since April 1998. The results shown are obtained from daily averages of the POAM measurements, which optimally include 14 measurements per day. On days when less than 14 measurements are available, the averaging interval is extended to ensure that every point shown includes at least 14 measurements. The white line indicates the daily maximum height of the tropopause (2 PV).

[12] The lower stratospheric variations in the SH POAM water vapor data differ significantly from those in the NH. In the POAM NH data, the mixing ratios in the lower stratosphere are lowest in the spring and summer. This air has recently passed through the tropical tropopause and thus experienced only a limited amount of methane oxidation [Nedoluha et al., 2002]. In the POAM SH data, the driest air in the lower stratosphere is found in the winter and spring. This air is dry not because it has recently passed through the tropical tropopause, but because it has been dehydrated within the vortex.

[13] Dehydration can occur when the temperature drops below the saturation vapor pressure. The dehydration of an air parcel can be either temporary or permanent, depending upon whether the particles that form become large enough to precipitate and fall out of the air parcel before temperatures increase sufficiently to sublimate the water. Because POAM measurements are not possible in regions with the very high aerosol optical depth environments associated with Type 2 PSCs, POAM is unable to measure dehydration as it occurs. If, however, this dehydration causes water to fall out of the air parcel, then the effect of this permanent dehydration will be detectable in those regions where POAM measurements are possible. Since POAM is only able to measure the effects of such permanent dehydration, any analysis of dehydration given here will refer only to the permanent removal of water vapor from an air parcel.

[14] In Figure 3 we show the water vapor mixing ratio as a function of equivalent latitude on the 450 K Θ-surface (∼17 km) during the NH and SH winters. Each POAM measurement (m) has been interpolated to the 450 K Θ-surface, and then plotted onto a grid after being convolved with a Gaussian of the form N(i, j, m) = exp[−((dayi–daym)/5)2] × exp[−((latj–latm)/5)2]. Here dayi–daym represents the difference in days between the date being plotted (dayi) and the date of the measurement (daym), and latj–latm represents the difference in degrees of equivalent latitude, between equivalent latitude being plotted (latj) and the equivalent latitude of the measurement (latm). Thus the mixing ratio plotted on Figure 3 for day = i and lat = j is given by vmrplot(i, j) = Σm{[vmrmeas(m) × N(i, j, m)]/N(i, j, m)}. During periods when a vortex is present in the SH vmrplot(i, j) is calculated separately for regions inside and outside the vortex. Points (i, j) where ΣmN(i, j, m) < 0.1 have been left blank. While Figure 3 provides a good overview of the distribution of water vapor measured by POAM, some of the detailed features may not represent true physical variations. This is especially true for features at very high equivalent latitudes and for those far from the POAM measurement latitude, where the calculated mixing ratio may be dominated by a very limited number of measurements (see Figure 1 for the distribution of measurements at 500 K). We also note that the apparent increase in the variability in the data is probably caused by a decrease in the frequency of available measurements made by POAM after March 1999 (see Figure 1), and by an increase in the number of measurement removed during the PSC season (see section 2). In addition, during the period September through November 2000 there is an increase in the uncertainty in the pointing of the instrument which may affect the water vapor retrievals.

Figure 3.

Water vapor mixing ratio as a function of equivalent latitude at 450 K for the two NH and three SH winters during which POAM measurements are available. Also shown is the latitude of the POAM measurements (black) and the vortex edge (red).

[15] The SH data initially shows the effects of descent, with mixing ratios during the months of May and June generally higher inside the vortex than outside the vortex. In addition to this descent, every SH winter shows extensive dehydration, but there is significant interannual variation in the distribution and timing of this dehydration. In 1998 the dehydration is observed to begin late in July, and affects all equivalent latitudes within the vortex nearly simultaneously. In 1999 the higher equivalent latitudes are affected early in July, while measurements near the edge of the vortex still show mixing ratios near the predehydration levels until early August. The 2000 SH measurements show signs of dehydrated air at very high equivalent latitudes as early as mid-June, but the air near the edge of the vortex is not severely dehydrated until mid-August. In addition, there appears to be less dehydration in 2000 than during 1998 or 1999 (as is also apparent from Figure 2).

[16] The amount of dehydration that occurs is clearly a strong function of the temperature encountered by the parcels measured by POAM. In Figure 4 we show the area covered at 50 mbar (∼450 K) by the coldest temperatures for each of the SH winters shown in Figure 3. The interannual variations in dehydration appear to be qualitatively consistent with the variations in temperature shown in Figure 4. The winter of 1998 shows the latest onset of cold temperatures, but when temperatures fall below 182 K in July they do so over a relatively large region (when compared with 2000) and thus cause vortex wide dehydration to occur quite quickly. In 2000 there were some unusually low temperatures very early in the season, consistent with the appearance of some dehydration in June. However, in July and August when most of the dehydration occurs, the SH winter of 2000 shows a smaller region with very low temperatures. This seems qualitatively consistent with the lack of dehydration observed throughout much of the vortex in July and August 2000 as compared with 1998 and 1999.

Figure 4.

The area, expressed in terms of an equivalent latitude, of various temperature ranges at 50 mbar. Also indicated on the color bar is the saturation mixing ratio at 50 mbar. The temperatures are taken from UKMO data sets.

[17] Since Figure 4 is plotted on a pressure surface it is possible to associate a specific saturation mixing ratio (over ice) with each temperature. It is interesting to compare this saturation mixing ratio with the amount of dehydration experienced during the winter. As is apparent from Figures 2 and 3, the water vapor mixing ratio throughout much of the vortex, which extends to ∼60 S, can drop to ∼1–2 ppmv for several months. Yet the area covered by temperatures cold enough to reduce the mixing ratio to 1.5 ppmv at 50 mbar never extends past ∼80 S, an area with a size equal to only ∼11% of the size of the total vortex (assuming the vortex extends to ∼60 S). Given the area of the 182 K cold pool relative to the size of the vortex, we can speculate as to the rate at which dehydration must take place. Since this relatively small cold pool of air is able to dehydrate air throughout the entire vortex to mixing ratios ∼1–2 ppmv on timescales ∼1 month (with considerable interannual variability), we conclude both that there is sufficient mixing within the vortex to allow a large proportion of the air to pass through this cold pool, and that, since each parcel can only spend a short period in this cold pool, the rate of dehydration is fast enough to dehydrate a parcel to 1–2 ppmv on timescales of at most several days. In section 4 we shall examine the relationship between dehydration and temperature in more detail.

[18] Just as the dehydration shows significant interannual differences in the SH, so does the rehydration in the austral spring. In 1998 and 1999 the vortex persists at 450 K until the end of December, and the water vapor mixing ratio in the vortex remains low throughout this period. In 2000 there is no longer a clear vortex edge at 450 K in December (by the criteria used here), and the typical outside the vortex water vapor mixing ratios reach ever closer to the pole throughout the November–December period.

[19] As in the SH, the NH data in Figure 3 shows the effect of the wintertime descent of the peak in the water vapor mixing ratio at high latitudes, with the highest mixing ratios at 450 K being observed inside the vortex. As is evident from the vortex edge contours, the two NH winters were meteorologically very different. The 1998–1999 Arctic stratospheric winter was abnormally warm and the polar vortex was unusually weak [Manney et al., 1999]. In contrast, the 1999–2000 Arctic winter was unusually cold, especially in the early winter lower stratosphere [Manney and Sabutis, 2000]. These meteorological differences are generally consistent with the POAM measurements differences. The stable vortex early in the 1999–2000 winter results in increased water vapor mixing ratios inside the vortex by the end of January, whereas in 1998–1999 the high mixing ratios inside the vortex are not measured until late February. In general, the difference between the inside and outside the vortex mixing ratios are smaller in 1998–1999 winter than in 1999–2000, however we note that much of this increased difference seems not to be caused by interannual differences inside the vortex, but by the lower mixing ratios outside the vortex in 1999–2000.

[20] Unlike the SH POAM measurements, neither of the two years of NH POAM measurements shows clear dehydration. Nevertheless, we can certainly not discount the possibility that some dehydration, perhaps even permanent dehydration, did occur during the 1999–2000 NH winter. Indeed, balloon-borne instruments did observe water vapor loss of up to ∼0.5 ppmv over a small altitude range during this winter, and simultaneous particle measurements showed that in some cases the missing water was not temporarily sequestered in particles [Schiller et al., 2002]. Their measurement of dehydration of ∼0.5 ppmv near the 500 K Θ-surface in both January and March suggests that this dehydration may have been present over a large area.

4. Trajectory Analysis

[21] In the previous section we have given an overview of the POAM water vapor, and have highlighted some of the differences between measurements in the NH and SH, and between measurements taken in different years. In particular, we have noted that the NH POAM measurements do not show any significant dehydration. A question of interest is whether this absence of measurable dehydration in the NH can be completely understood in terms of temperature differences between the NH and SH.

[22] In order to study dehydration we need to examine not just the temperatures at the POAM measurement locations, but also the temperature history of the parcels measured by POAM. This temperature history is determined using a 3-D trajectory model developed by Bowman and Carrie [2002]. For the results shown here the model has been run using UKMO data with descent calculated from net radiative heating rates [Rosenfield et al., 1994]. Back trajectories have been calculated from the positions of the POAM measurements on several potential temperature surfaces. In order to insure that these primary back trajectories are representative of a reasonably compact parcel of air, we have also calculated complementary back trajectories from four positions displaced by 110 km in each compass direction. The back trajectory calculations are stopped when the position of one of the four complementary back trajectories is more than 500 km from the primary trajectory.

[23] As an example of these calculations, we show in Figure 5 the mixing ratios of inside the vortex POAM NH and SH measurements on the 500 K Θ-surface as a function of the minimum temperature encountered by the parcel during the preceding 5 days. The NH data is taken from the unusually cold Arctic winter of 1999–2000, while the SH data is taken from the Antarctic winter of 1998. We have chosen to show the data on the 500 K Θ-surface because this is the surface on which the NH back trajectories from the POAM measurements show the largest amount of supersaturation.

Figure 5.

The water vapor mixing ratio of POAM measurements at 500 K as a function of the minimum temperature calculated from a 5-day back trajectory for the (top) NH and (bottom) SH. The solid lines are averages, binned in 1 K increments. The colors indicate the length of time each parcel is saturated based on its final measured mixing ratio. Since saturation mixing ratio varies with pressure, we have used two dotted lines in each plot to indicate the range of saturation temperatures as a function of mixing ratio for the range of pressures encountered by those 500 K measurements for which the back trajectories indicate that the parcel became saturated.

[24] Figure 5 emphasizes two important differences between the NH and SH. There are trajectories in the SH with significantly colder temperatures than in the NH, and there are observations in the SH with significantly lower mixing ratios. There is a strong correlation between low mixing ratios and low minimum temperatures in the SH parcels, suggesting that those parcels which experienced low minimum temperatures also experienced dehydration. Observations of very low mixing ratios for trajectories that do not become saturated suggest that these parcels may have become dehydrated at an earlier time, i.e. more than 5 days before the POAM measurement. As was evident in other plots, the POAM measurements do not show any significant dehydration in the NH. Nevertheless, some of the trajectories in the NH in late December 1999 and in January 2000 indicate that the parcels observed passed through regions where the temperature was low enough so that some dehydration could have occurred. On two occasions (1 and 2 January 2000) POAM measured parcels in the NH that experienced temperatures several degrees below saturation and remained supersaturated for >2 days, but still show no clear dehydration above the noise level of the measurements.

[25] There are several possible reasons why a parcel that has been supersaturated may not exhibit any permanent dehydration. It could be that the POAM water vapor measurement is in error, but it is unlikely that the error would by chance raise the mixing ratio of a dehydrated parcel to a level very similar to that of undehydrated parcels. Another possibility is that the temperatures on the back trajectories for these supersaturated cases are inaccurate, either because of errors in the temperatures themselves or because the trajectories do not, in these cases, provide an accurate model of parcel motion.

[26] Alternatively, the POAM NH measurements and the temperature minima could be correct, but any dehydration that occurred may have been temporary. This would imply that while ice particles formed, they did not grow to sufficient size to fall out of the layer. Calculations by Kasten [1968] show that a spherical particle with a mass density of 1 gm-cm−3 and a radius of 1 μm will fall at a velocity ≈2.3 × 10−2 km-day−1 at 17 km. While the fall rate is not very sensitive to changes in altitude, it grows very quickly with increasing particle size, so that particles with a radius of 10 μm will fall at ≈1.4 km-day−1 at 17 km. Particles with sizes less than a few μm would therefore not have time to precipitate out during the period that NH parcels were at temperatures below saturation.

[27] It is also possible that the NH parcels failed to experience dehydration because of a nucleation barrier. Tabazadeh et al. [2000a] calculate that, for water vapor pressures comparable to those observed here, supercooling of ∼3 K is required in order to for ice nucleation to occur on an aqueous H2SO4 solution droplet. Carslaw et al. [1998], using airborne lidar measurements of PSCs, find that the optical properties are best fit by assuming a supercooling of ∼4 K. The coldest temperatures in the POAM back trajectories shown in Figure 5 are ∼3 K below the saturation temperature.

[28] While Figure 5 does show that there was significant dehydration in the SH, it does not provide information as to when the dehydration occurred. In order to obtain this information we search for cases in which multiple POAM measurements of a single air parcel are available. This is similar to the “match” technique [e.g., Rex et al., 1999] used to investigate ozone loss. The measurements are assumed to sample the same air parcel if the back trajectory from the later POAM measurement passes within 500 km of the earlier measurement at a time differing by no more than 2.4 hours. In addition, none of the complementary back trajectories must be more than 500 km from the primary trajectory when the match occurs. Back trajectories of up to 10 days are used to search for these matches.

[29] In Figure 6 we show the POAM matches for the 1999–2000 NH winter and the 1998 SH winter. We have binned the data in Figure 6 in 1 ppmv increments, and the average change in water vapor in each increments is indicated by the solid line. The error bars shown are based solely on the variability within each bin and provide an estimate of statistical significance of this average, but do not take into account any possible systematic errors. Parcels for which vmri–vmrsat < 0, i.e. the initial mixing ratio (vmri) is smaller than the saturation mixing ratio at the minimum temperature encountered by the parcel (vmrsat), do not become saturated during the period between the measurements, and hence would not be expected to show any loss of water vapor. The presence of the solid line near vmri–vmrf = 0 when vmri–vmrsat < 0 indicates that, on average, very little change in the average water vapor mixing ratio has taken place among these parcels. For values of vmri–vmrsat > 0 some dehydration may occur, hence we might expect vmri–vmrfinal > 0.

Figure 6.

The difference in water vapor mixing ratio between the successive POAM measurements of the same air parcel, plotted against the difference between the initial water vapor mixing ratio and the saturation mixing ratio based (calculated from the minimum temperature encountered by the trajectory). Results are shown for the 450 K, 500 K, and 550 K levels. Also indicated are the limiting cases of no dehydration and of the dehydration required to bring the mixing ratio down to the saturation level at the minimum temperature (dotted lines). The solid line is calculated from the average change for 1 ppmv bins, with error bars indicating the random uncertainty in the average of each bin (σ/n1/2). The colors of the points indicate the lengths of time that a particular parcel has spent below the saturation temperature based on its initial mixing ratio.

[30] An important point to consider in analyzing Figure 6 is that the axes used to display the data in Figure 6 are not independent (vmri is included in both ordinate and abscissa), hence a random error in the POAM measurements will result in a positive correlation in these plots. This problem becomes especially pronounced near the most extreme values of vmri–vmrf and vmri–vmrsat. In order to estimate the importance of this effect we produced a model data set with vmr*f = vmri and vmr*i = vmri + σ, where σ is a random number with a standard deviation equal to the estimated random error in vmri. We then compared the results for vmr*i–vmr*f as a function of vmr*i–vmrsat to those in Figure 6. For the SH, this model binned data showed that, for temperatures above saturation, vmr*i–vmr*f was not significantly different from zero. At temperatures below saturation the model vmr*i–vmr*f in the saturated regime was ∼1/2 to 1/3 the size of the difference calculated from the actual data. Thus, in the SH, the lack of independence of the two axes does not significantly alter the value of vmri–vmrsat at which water vapor loss begins to occur, but may cause the average water vapor loss suggested by the matches in Figure 6 to be larger than the average true loss by as much as a factor of 2. A similar calculation based on the NH data shows that the small positive slope near saturation in the NH data shown in Figure 6 may be caused by random errors in vmri.

[31] We have indicated by the two dotted lines in Figure 6 the two limits for the changes in water vapor that a parcel could experience when the temperature becomes colder than the saturation temperature. The one limit is that of no dehydration, which would imply that, despite the exposure of the parcel to temperatures below saturation, no water has been permanently lost during the period between the two POAM measurements of this parcel. As was discussed above, some possible reasons for the absence of dehydration under these conditions are the presence of a nucleation barrier or the failure of ice particles to precipitate out. The other limit is to assume that the parcel loses all of the water required to bring it to saturation when it encounters the minimum temperature along the trajectory. While there is clearly a large amount of scatter, the measurements do show that the average observed dehydration falls in between these two limits.

[32] In the SH the amount of dehydration is positively correlated both with the amount of supersaturation and the time spent at temperatures below saturation, but since the latter two quantities are themselves highly correlated it is difficult to determine their relative importance. If we segregate parcels according to their maximum level of supersaturation, then we see that of those twenty SH parcels that experience temperatures such that vmri–vmrsat > 2 ppmv, all but one exhibit some dehydration, even though many of the parcels were supersaturated for <1 day. If, on the other hand, we segregate parcels according to the length of time for which they are exposed to temperatures below saturation, then we find that eleven of the twelve matches which were supersaturated for periods of between 2 and 6 days lost >70% of their water in excess of the saturation level (as discussed above, this fractional loss may be too large by a factor of 2). This timescale is somewhat deceptively long, since while these parcels are supersaturated for several days, the amount of time for which that they are at a temperature cold enough to dehydrate them to their final water vapor mixing ratio is obviously shorter. In fact, only six of the parcels that lost water vapor experienced temperatures below the saturation temperature of their final mixing ratio for >1 day, and only one experienced such temperatures for >2 days. This suggests that particles can grow quickly to the ∼10 μm size at which they fall at least ∼1 km (the POAM vertical resolution) in ∼1 day.

[33] It is interesting to compare the dehydration observed in these matches with the overall amount of dehydration that occurs in the Antarctic winter. The total loss of water vapor during the SH winter is ∼5–6 ppmv, while according to Figure 6 the average mixing ratio lost along those trajectories showing supersaturation of >2 ppmv is ∼1.5 ppmv (which, taking into account the nonindependence of the axes, suggests an average true loss of ∼0.75–1.0 ppmv). Thus, while a parcel can be significantly dehydrated during one episode of supersaturation, it most likely undergoes several such episodes before reaching its final level of dehydration. Of course, as the parcel becomes partially dehydrated, the temperature required for further dehydration continues to decrease. As was shown in Figure 4, the pool of air required to dehydrate air down to the ∼1.5 ppmv level is small relative to the size of the vortex, suggesting that it is possible for a parcel to be dehydrated down to this level on timescales of at most several days. Since we do not observe any matches with losses ∼5–6 ppmv, these results suggest that most parcels enter this very cold air having already lost much of the water with which they began the winter.

[34] An interesting feature of Figure 6 is that, in the SH, the first signs of dehydration appear to occur near vmri = vmrsat. Thus this data suggests that there is no nucleation barrier. This contrasts with the NH measurements, which are consistent with the presence of a nucleation barrier. There are, however, significant uncertainties in the data that could affect our estimate of the size of the nucleation barrier, and these uncertainties are probably large enough to explain the hemispheric differences. We examine some of these uncertainties below.

[35] Nucleation barriers are generally presented in terms of a temperature threshold that must be overcome, but because the results shown in Figure 6 combine measurements at different pressures and initial mixing ratios there is no simple formula that can be used to convert the horizontal axis from vmri–vmrsat into a nucleation barrier measured in Kelvin. However, we note that if we apply a 3 K shift to the saturation temperature, then statistically significant dehydration does not occur in the 0 ppmv < vmri–vmrsat < 1 ppmv bin, but does occur in the 1 ppmv < vmri–vmrsat < 2 ppmv bin. Thus the conclusion that the nucleation barrier is smaller than 3 K relies upon the significance of the dehydration observed in the 0 ppmv < vmri–vmrsat < 1 ppmv bin. The dehydration in this mixing ratio bin is statistically significant even when the nonindependence of the axes is taken into account. In addition to uncertainties in the measurements of dehydration, there are several other possible sources of error that could all shift the data in Figure 6 in the direction of reducing any apparent nucleation barrier. Most obviously, there could be a high bias in the UKMO temperatures. Manney et al. [1996] have pointed out that neither the UKMO nor the NMC analyses captures the very lowest temperatures observed by radiosondes. Even a random temperature error with no bias will tend to lower the apparent nucleation barrier, provided that some of the observed parcels actually do experience temperatures cold enough to cause dehydration. This would occur because some trajectories that do experience cold temperatures and dehydration will, by chance, have model minimum temperatures that are significantly higher than their true minimum temperatures. Since the relationship between temperature and dehydration is nonlinear, the failure to include the effects on the parcels of those temperatures that are colder than the model temperature will not be cancelled out by random temperature errors of the opposite sign. Furthermore, even in the absence of any random temperature error in the calculated back trajectory, some portions of the air in the measured parcel will certainly have passed through temperatures that are slightly colder than the average temperature for the parcel as a whole, and these temperatures could have fallen below the nucleation barrier even while the average temperature remained above the nucleation barrier.

[36] As in Figure 5, there is no clear indication in Figure 6 of any significant dehydration in the NH. There are, however, only a limited number of matches available with trajectories that encounter regions of supersaturation, and the vast majority of these are supersaturated for <1/2 day. While a comparison of NH and SH results in Figure 6 may not suggest any significant difference in the relationship between supersaturation and dehydration in the two hemispheres, the relative stability of the NH water vapor mixing ratios allows us to make useful estimates of the effects of supersaturation from Figure 5. There are a few NH measurements in Figure 5 which do undergo conditions for which the SH measurements shown in Figure 6 almost always show significant dehydration. As we noted earlier, eleven of the twelve matches in the SH which were supersaturated for more then 2 days lost >70% their available water above the saturation level, but neither of the NH measurements which experienced similar periods of saturation similar signs of dehydration. In addition, there are several NH measurements of parcels whose mixing ratios exceed the saturation mixing ratio by ∼2 ppmv, but again these show no clear dehydration. The differences between the observed absence of dehydration in the NH, and the presence of such in the SH could be caused by many factors, including some of those discussed above in reference to the apparent absence of a nucleation barrier in the SH. For example, a difference in the temperature bias in the two hemispheres would lead to an apparent interhemispheric difference in the nucleation barrier. The study by Manney et al. [1996] showed that in the Arctic winter of 1994–1995 the UKMO temperatures were warmer than the NCEP temperatures, while in the Antarctic winter of 1993 the opposite was the case. If these differences between the UKMO and NCEP data were representative of a typical bias in the UKMO data, then they could result in an apparent nucleation barrier for the POAM SH measurements, and no such barrier in the NH; the exact opposite of what is suggested by the results that we have presented. However, the analysis by Manney and Sabutis [2000] shows that the coldest UKMO temperatures at 46 hPa were ∼1–3 K lower than NCEP temperatures from mid-December 1999 through mid-January 2000, while Bevilacqua et al. [2002] found only a very small cold bias of ∼0.33 ± 0.25 K in a comparison of UKMO data with sondes near the POAM measurements. Thus it is possible that the UKMO data during the NH winter of 1999–2000 are accurate and result in a more accurate estimate of the nucleation barrier in the NH than in the SH.

5. Summary

[37] We have documented several years of POAM water vapor measurements in the NH and SH. While the latitudinal coverage of the POAM measurements in the NH and SH differs, the data does cover a large range of equivalent latitudes in both hemispheres, and there is good coverage in both the Arctic and Antarctic vortices. We have therefore been able to investigate in detail the dehydration in the SH, and the occurrence of supersaturation without measurable dehydration in the NH.

[38] Analysis of the size of the pool of coldest air relative to the size of the vortex suggests that permanent dehydration must occur over timescales no more than several days. This timescale is consistent with calculations from POAM matches which show that significant dehydration occurs in almost all cases when a parcel remains supersaturated for >2 days, and that parcels which lose water generally spend <1 day at temperatures below their final saturation level. In order for ice particles to drop quickly enough to permanently dehydrate a parcel on this timescale they must grow to a size of at least ∼10 μm in <1 day. While many of the POAM matches in the SH show significant dehydration, the average amount of water vapor lost in even the most supersaturated set of POAM matches is still only ∼1 ppmv; thus these parcels must undergo several episodes of dehydration before reaching the 1–2 ppmv mixing ratio typically observed within the vortex in the Antarctic spring.

[39] The POAM NH measurements do not indicate any significant permanent dehydration. POAM does, however, observe a few parcels in the NH whose back trajectories indicate that they experienced supersaturation. A very small number of these parcels have experienced levels and periods of supersaturation for which dehydration is almost always observed in the SH. The apparent NH/SH asymmetry is probably caused by uncertainties in the POAM measurements and in the temperatures to which the measured parcels were exposed.


[40] We thank the UK Met. Office for providing the meteorological data used in this paper, J. D. Lumpe for providing water vapor cross sections, and D. R. Allen for his helpful comments on the manuscript. The POAM III instrument was sponsored by the Office of Naval Research. The French Centre National d'Etudes Spatials operates the SPOT 4 spacecraft and has generously waived the uplink fees. Launch and initial operation of POAM were sponsored by the Air Force Space Test Program; continuing operations are performed by the Air Force Space and Missile Command. Support for scientific analysis of the data comes from the Naval Research Laboratory and from the NASA Atmospheric Chemistry Modeling and Analysis Program.