We present an analysis of polar stratospheric water vapor from measurements of the Improved Limb Atmospheric Spectrometer (ILAS) onboard the Advanced Earth Observing Satellite (ADEOS). The variability of water vapor is examined using contemporaneous potential vorticity fields to separate the vortex from the extravortex air. The overall water vapor distribution from the ILAS measurements agrees qualitatively with the water vapor climatology in the UARS reference atmosphere (covering 1992–1999), and the space-time variability is consistent with the known stratospheric circulation. Quantitatively, comparisons between ILAS and Halogen Occultation Experiment (HALOE) equivalent latitude zonal means show agreement to within 5% in the lower stratosphere (400–800 K potential temperature or 15–30 km). In the upper stratosphere (800–1800 K potential temperature or 30–45 km), however, ILAS data have a 10% to 20% high bias compared to the 8-year (1992–1999) UARS climatology and to HALOE data for the same time period. The ILAS data, covering the time period November 1996 to June 1997, were measured during a winter-spring period when the Arctic polar stratosphere was anomalously cold. The ILAS water vapor data suggest an Arctic vortex that is more isolated and persistent in the spring than that seen in the UARS climatology. The spatial and temporal extent of the Arctic vortex this year was very similar to that of the Antarctic vortex. This feature is supported by HALOE data for the same measurement period. Using a trajectory model, we show that the NH vortex in spring 1997 was significantly more isolated compared to other years in the decade, and this explains the unusual confinement of high water vapor within the Arctic vortex observed by ILAS.
 Water vapor is a long-lived tracer in stratosphere. Its distribution and variability has provided a wealth of information on the atmospheric circulation. The main chemical source of water vapor in the stratosphere is methane oxidation [Brasseur and Solomon, 1986; LeTexier et al., 1988]. The primary sink in the lower stratosphere is due to the formation of ice particles, which occurs mainly in the polar region. An additional sink of water vapor due to photolysis in the mesosphere [Brasseur and Solomon, 1986] also effects the stratospheric water vapor distribution near the stratopause. The quantity (H2O + 2CH4), often referred to as the total hydrogen index or total water, is approximately conserved in the lower stratosphere with a value around 7.5 ppmv [Dessler et al., 1994; Hurst et al., 1999], except when dehydration occurs in the polar region due to the formation of ice particles. The climatological distribution of water vapor in the stratosphere [Remsberg et al., 1984, 1996; Mote et al., 1996; and Randel et al., 1998] has several well known features: (1) low values near the tropical lower stratosphere as a result of air ascending through the cold tropical tropopause region with an annual average around 3.8 ppmv [Dessler and Kim, 1999, and references therein], (2) increased mixing ratio values with altitude due to the production of water vapor by methane oxidation, reaching ∼7 ppmv near the stratopause, and (3) high values in the polar regions due to the descent of air from high altitudes and confinement within the polar vortices, except when dehydration occurs in the polar lower stratosphere. Quantitative information on water vapor distribution is important for understanding both the radiation budget of the Earth and atmospheric chemistry. Variability of water vapor in the polar region is especially relevant for the latter, since the formation of polar stratospheric clouds (PSCs) and the subsequent heterogeneous chemistry is chiefly responsible for ozone destruction [Solomon et al., 1986].
 In this paper, we present the monthly variability of polar stratospheric water vapor derived from the version 5.20 data of the Improved Limb Atmospheric Spectrometer (ILAS) onboard the Advanced Earth Observing Satellite (ADOES). As described by Sasano et al. , ILAS is a solar occultation spectrometer that consists of infrared and near visible channels. ILAS made profile measurements of ozone, water vapor, nitric acid, as well as aerosol extinction in the polar stratosphere from November 1996 to June 1997. Descriptions of the instrument performance and retrieval algorithm can be found in companion papers in this issue [Nakajima et al., 2002; Yokota et al., 2002]. Overall description and validation of the version 5.20 water vapor data are given by Kanzawa et al. , where the differences between ILAS water vapor and correlative measurements, including both in situ measurements from balloon borne, airborne platforms and satellite measurements, are quantified.
 The objectives of this paper are (1) to examine the climatology and variability of the ILAS water vapor data for consistency with the known stratospheric circulation, (2) to make comparisons with the water vapor climatology derived from the UARS Halogen Occultation Experiment (HALOE) and MLS data [Randel et al., 1998], and (3) to demonstrate the dynamic variability of water vapor in relation to meteorology of the Arctic polar vortex between the ILAS and UARS time periods. Examination of the climatology or zonal average is a good approach for evaluating data quality and is complementary to detailed comparisons with correlative measurements. Since the distribution of water vapor is largely controlled by transport in the stratosphere, the observed variability in the ILAS zonal mean is an important validation for the data set as a whole.
 The 1996–1997 NH winter-spring period has been known for its anomalously low temperature in the Arctic polar stratosphere. The spring Arctic polar vortex was anomalously strong, similar to the Antarctic polar vortex [Coy et al., 1997]. This has been shown to have significant impact to the ozone chemistry. Using ILAS observations, Sasano et al.  have estimated that up to 50% of the ozone inside the vortex was lost. Arctic PSCs were observed by ILAS during January–March 1997 [Hayashida et al., 2000]. In connection with observed PSCs, ILAS measurements have also found occurrences of dehydration inside the Arctic polar vortex during this extremely cold NH winter [Pan et al., 2002]. In this paper, we examine how the meteorological conditions were manifested in the water vapor zonal mean, which provides a useful perspective of the vortex evolution this year.
 The paper is organized into four sections. This section serves to define the scope of the paper. The remaining three sections cover the motivation and methodology of equivalent latitude mapping for ILAS data, comparisons of the ILAS zonal mean with the UARS climatology, and examination of water vapor variability in Arctic polar region in association with transport across the vortex. A brief summary is given to conclude the paper.
2. Data and Analyses
2.1. ILAS Data and Equivalent Latitude Mapping
 The ILAS instrument made measurements at high latitudes near both the Arctic and Antarctic polar vortices during 8 months from November 1996 to June 1997. The latitude ranges of sampling, varying month-by-month, were from 57 to 73 degrees North and from 64 to 88 degrees South [Sasano et al., 1999]. In this section, we show that although the measurements were made within relatively narrow latitude bands geographically, they covered a wide range dynamically, both because of steep potential vorticity gradients near the vortex edge and the fact that the vortex shapes were often distorted with respect to the pole.
 As an example, Figure 1 shows the relationship of the ILAS sampling location to the polar vortex. Given in the figure are the modified potential vorticity (MPV) [Lait, 1994] at the 750 K level for NH for 22 January 1997 and the locations of ILAS sunrise measurements for 22 and 23 January. ILAS measurements were near the 65 N latitude circle for the two days. The polar vortex was asymmetric (using the 20 PVU contour to represent the edge) and extended from 45°N at the lowest latitude (near 30° east) to 80°N at the highest latitude (near the Date Line). Due to the asymmetry of the polar vortex, the measurements were made both inside and outside the vortex. Figure 2 gives the vertical cross-section of the measured water vapor profiles during the three days of 22–24 January, arranged in sequence of the adjacent measurement events, which were approximately 25 degree apart in longitude. Overlaying contours in the figure are the PV field. The horizontal gradient of the PV field and the water vapor distribution are well correlated and mark the edge of the vortex. High values of water vapor mixing ratio mark the interior of the vortex, where water vapor values ∼6 ppmv or higher reflect the descent from near the stratopause, and the low values indicate the region outside the vortex.
Figures 1 and 2 illustrate the wide dynamic range that the ILAS measurements cover in a given day and the importance of separating the data by dynamical variables when computing statistics or making physical interpretations. To connect the space defined by the dynamic variable of PV to the space defined by the geographical latitude, we map PV values into their equivalent latitudes and use the equivalent latitude as an alternative way to represent the data. The equivalent latitudes are determined by comparing the areas enclosed by the PV contours with those enclosed by the latitudinal circles [e.g., Butchart and Remsberg,1986; Lait et al., 1990; Schoeberl et al., 1992; Randel et al., 1998; Manney et al., 1999]. Figure 3 illustrates the effectiveness of using equivalent latitude as the coordinates to display the data. Displayed in Figure 3 are ILAS NH middle stratosphere water vapor data (between 600 and 800 K potential temperature range) as a function of latitude, equivalent latitude and modified PV for the month of January 1997. As shown in the figure, the measurements made within a narrow latitude range near 65°N are well correlated with the dynamic variable of PV. When the data are sorted by PV (expressed as the equivalent latitude), the ILAS sampled from approximately 25 degrees to the pole in the equivalent latitude space, a range representative of both hemispheres. The change of slope around 65°N equivalent latitude in the water vapor distribution also marks the position of the vortex edge for this month.
 As illustrated in Figures 1–3, under these conditions, it is advantageous to use the equivalent latitude instead of the geographical latitude when computing zonal mean statistics and for comparison with other satellite climatologies that are compiled in the same dynamical coordinates. We have computed ILAS water vapor monthly zonal mean in equivalent latitude and potential temperature (θ) coordinates. The data are binned in 4-degree equivalent latitude intervals and θ layers varying roughly according to the standard UARS pressure levels (six per decade of pressure). The PV values used for computing equivalent latitudes are derived from UKMO meteorological analyses [Swinbank and O'Neil, 1994], and we use the modified PV of Lait  and have applied a scaling factor of (θ/420)−9/2.
2.2. UARS Reference Atmosphere
 The UARS water vapor climatology used in this paper is a subset of the UARS reference atmosphere (URL for the project is http://code916.gsfc.nasa.gov/Public/Analysis/UARS/urap/home.html), which includes a set of climatologies of atmospheric trace gas profiles compiled from UARS measurements. The water vapor climatology is derived from a combination of HALOE and MLS data [Randel et al., 1998], using HALOE [Russell et al., 1993; Harries et al., 1996] version 19 (96 months from 1992 to 1999) and MLS prototype version 5 (nonlinear retrieval) data (September 1991 to April 1993) [Pumphrey, 1999]. The 8-year climatology shown in this paper is derived using a harmonic fit of the seasonal cycle to the 96-month binned averages, as described by Randel et al. .
2.3. HALOE Data
 To examine and validate the atmospheric variability shown in ILAS data, we have computed a HALOE monthly zonal mean using HALOE version 19 level 2 data for the same measurement period as ILAS (November 1996 to June 1997). The average is computed in identical fashion as for the ILAS equivalent zonal mean calculations. The equivalent latitude mapping of the HALOE data used UKMO PV data.
3. Seasonal Variability of Water Vapor Observed by ILAS
 Displayed in Figure 4 is ILAS water vapor monthly zonal mean for November 1996 to June 1997 period in equivalent latitude-θ cross-section. Consistent with the previous section, the ILAS measurements cover equivalent latitudes from approximately 25° to the pole in each hemisphere with no data in the tropics. As a comparison and to provide a global perspective, water vapor climatology in the UARS reference atmosphere for November, March and June is shown in Figures 5a–5c. We do not expect the ILAS and the UARS climatology to be identical, since the length of the data records and the sampling intervals for the data sets went in the climatologies are all very different. Rather, qualitative agreements of the general features are expected. As an additional comparison in a more quantitative level, HALOE data for November 1996 and March, June 1997 are displayed in Figures 5d to 5f.
 As shown in Figures 4 and 5, the global scale structure and seasonal variation of water vapor evident in ILAS data are consistent with the UARS climatology, and reflect features of stratospheric circulation. The overall patterns show the general circulation of air rising in the tropics and sinking in the pole. The high value of water vapor mixing ratio in the upper stratosphere (represented by yellow and red in the figures) is due to methane oxidation. The effect of mesosphere photolysis is visible near the winter polar stratopause (represented by green right below the 2000 K level in the figures). Although the ILAS measurements do not cover the tropics, the general pattern of upward bulging contours as a result of the circulation and the weak latitudinal gradient in the midlatitude as an effect of mixing in the “surf-zone” are readily observed in the zonal mean.
 The formation and the decay of the polar vortex are clearly revealed in the ILAS water vapor data. In both hemispheres, the vortex edge is marked by the steep gradient of water vapor, with high vortex interior values due to the seasonal downwelling. The decay of the vortex in the middle stratosphere (∼800 K or ∼30 km) in the spring hemisphere is shown in NH March and April, with the vortex air persisting longer in the lower stratosphere (∼600 K or ∼23 km). This feature is consistent with the seasonal variation of vortex breakdown derived from previous observations and model simulations [e.g., Schoeberl et al., 1992; Manney et al., 1994], showing breakdown first in the upper stratosphere and progressing downward in time.
 The ILAS data also revealed very dry air associated with dehydration in the Antarctic polar lower stratosphere (around 400 K) during November–December, together with a persistent dry layer in the lower stratosphere over much of the globe. These features are consistent with the UARS climatology, as shown in Figure 5. The climatological feature of dehydration inside the Antarctic vortex is well documented from previous observations [e.g., Tuck et al., 1993; Pierce et al., 1994; Rosenlof et al., 1997; Randel et al., 1998; Pumphrey, 1999; Nedoluha et al., 2000]. Although ILAS only observes this feature for November–December, the near global dry layer above the tropopause is mapped more completely in the UARS data and is formed by rapid isentropic transport following the tropical dehydration [Randel et al., 2001].
 There are also marked differences between the ILAS data and UARS climatology. Quantitatively, ILAS data show higher water vapor mixing ratios in the upper stratosphere and inside the vortex, especially in the SH. The amount of high bias is consistent with the statistical comparisons with the correlative measurements, which indicate that ILAS water vapor is ∼10% higher in the altitude region of 30–60 km for a majority of the comparisons and within 20% for all comparisons with isolated exceptions [Kanzawa et al., 2002], Qualitatively, the descent of air within the NH vortex is much more pronounced in the 1997 ILAS data compared to the 1992–1999 UARS data. A much more isolated and persistent vortex in the NH polar region is indicated by the March and April ILAS data as compared to the UARS climatology.
 To investigate whether the qualitative difference between the 1996–1997 ILAS data and the 8-year UARS climatology is due to physical variability of the atmosphere, we have examined the HALOE data for the same time period. A comparison of HALOE data for November1996 to June 1997 (three of the eight months shown in Figure 5) with ILAS data shows many similar features that are different from the 8-year UARS climatology. For example, the Antarctic vortex in the lower stratosphere is more pronounced in November 1996 data from both ILAS (Figure 4) and HALOE (Figure 5d) as compared to the UARS (Figure 5a). More significantly, strong water vapor gradients across the Arctic vortex are found in both ILAS and HALOE March 1997 data (Figure 5e), strikingly different from the UARS climatology (Figure 5b).
 A more quantitative perspective of these comparisons is given in Figure 6, which shows the March NH water vapor equivalent latitude zonal mean at the 655 K θ level for all three data sets. The figure shows a reasonable agreement between ILAS and HALOE data for the measurement made during the same time. Both data sets show a near 2 ppmv increase of H2O from 55 to 75 degree equivalent latitude. The steep gradients near 60 N equivalent latitude mark the vortex edge in March1997, which do not appear in the multiyear UARS climatology. In the equivalent latitude range 25–50 N, on the other hand, ILAS water vapor is ∼0.5 ppmv higher than the HALOE data. Latitudinal sampling difference could be a factor that contributed to the apparent high bias in ILAS data. ILAS sampled mostly high equivalent latitude air during this month, and the averages for lower equivalent latitudes are calculated from a small sample and are subject to a higher uncertainty. As a specific example, the 655 K equivalent latitude zonal averages are computed from ∼100 data points per 4-degree equivalent latitude band for the 50–80 N range, but are from ∼10 data points per equivalent latitude band between 25 and 50 N. Indeed, such latitudinal sampling induced biases are found in the ozone data comparisons using equivalent latitude/θ coordinates between POAM II data and HALOE data [Manney et el., 2001].
 Additional quantitative comparisons for different altitudes are given in Figure 7, where ILAS, UARS and HALOE March equivalent zonal mean profiles for 68 S and N are shown. Evident in Figure 7, ILAS and HALOE mean profiles show good agreement in the middle to lower stratosphere (below 800 K or 30 km), where the differences are well within 5%. Above 800 K, ILAS data begin to show high bias compared to HALOE for both hemispheres, up to ∼10% in the upper stratosphere. The amount of high bias varies with latitudes and is different for each month. The biases in the NH data are typically less than 10% but are higher for the SH data at higher equivalent latitudes, up to 25% in March, consistent with that discussed by Kanzawa et al. .
 In the lower stratosphere, both ILAS and HALOE 1997 mean profiles are similar to the UARS climatology in the SH, but significantly different in the NH. The agreement between ILAS and HALOE data provide strong evidence that the difference between the ILAS and UARS data in the NH polar region is a result of the meteorological conditions of the 1996–1997 winter-spring seasons, rather than a peculiar bias in the ILAS data. In the following section, we further examine the atmospheric variability manifested in water vapor observations during this time period.
4. Transport Near the Arctic Polar Vortex for NH Spring 1997
 To compare the time evolution of the polar vortices as inferred from water vapor data, we present, in Figure 8, UARS climatology and the ILAS monthly zonal mean water vapor time-height cross-section at 72°NH and SH equivalent latitudes. To facilitate a more direct comparison between the NH and the SH seasonality, we display SH data from January to December and NH July to June. Although the ILAS measurements only cover two thirds of the seasonal cycle and have higher mixing ratios in the upper stratosphere, there is a qualitative agreement between ILAS and UARS in the vortex evolution for the summer to early winter transition in the SH. In the NH vortex, however, the ILAS data show deeper descent of the winter maximum, indicating a more isolated and persistent Arctic vortex during 1997. ILAS NH data indicate the descent of high H2O vortex air to about 500 K in θ and to April–May in time, while the UARS data show a reduced scale in the descent and a much shorter vortex season at the lower altitudes. On the other hand, similarities of 1997 Arctic vortex shown in ILAS water vapor to the multiyear UARS climatology of the Antarctic vortex is readily observed in Figure 8.
 To further demonstrate that the anomalous water vapor distribution observed by ILAS is consistent with the meteorological conditions of the 1996–1997 vortex, we have performed a tracer transport simulation, using a three-dimensional trajectory model. The calculation uses isentropic coordinates with horizontal velocities from the UKMO wind data, and vertical velocities derived from the radiative heating code of Olaguer et al. . These calculations are very similar to those shown by Manney et al. . A set of tracers is initiated inside the vortex (choosing 20 PVU contour as the boundary) at the 665 K level on 1 March and we examine the change of the distribution after one month. Among the 7 years we investigated (1993–1999), March 1997 stands out as an extreme case of vortex isolation in terms of the confinement of the tracers.
 Displayed in Figure 9 are comparisons of tracer distributions at the beginning and the end of March for 1993 (Figure 9a) and 1997 (Figure 9b). We choose 1993 as a representative for other years, both because the 1993 results are typical, and because the UARS climatology in the polar region is largely compiled from the MLS measurements (made between September 1991 and April 1993). Figures 9a and 9b show the number of tracers as a function of the equivalent latitude on 1 and 31 March for the two selected years. Figures 9c and 9d show the tracers latitudinal-longitudinal position on 1 March for the 2 years at the starting potential temperature level 665 K. Figures 9e and 9f give the tracer positions at the end of March and the lowest potential temperature level the parcels reached though descent. As shown in the figure, during March 1993, about 35% of the tracers were transported out of the vortex, but almost none for 1997. In fact, there is a small poleward change of tracer distribution from the beginning to the end of March 1997, which reflects the fact that although maintaining its strength, the vortex became smaller by the end of the month.
 To reveal the mechanism of the NH vortex isolation during spring 1997, we have examined the March PV and wind fields using multiple year NCEP data. The results are given in Figure 10, where we show PV and wind speed averaged along equivalent latitudes for 1 and 31 March, 1993–1999, at the 665 K level. The change of PV gradient indicates the vortex edges to be around 60° equivalent latitude. The maximum of the zonal wind also corresponds to the position of the vortex edge and gives the strength of the polar night jet. At the beginning of March, both 1997 PV and equivalent latitude zonal wind are similar to other years. At the end of March, however, the 1997 the wind became completely anomalous: while the jet was significantly weakened by the end of the month for all other years, the 1997 jet was as strong as at the beginning of the month. This is also reflected in the higher PV value at the end of March 1997. These results suggest that the more persistent NH polar night jet in early spring 1997 is responsible for the isolation of the Arctic vortex this year, as reflected in the ILAS water vapor data. The comparisons in Figures 9 and 10 also suggest that the strength of the polar jet is a more determinant characteristic for the vortex confinement than the PV structure alone.
 Using equivalent latitude mapping, we have demonstrated that the ILAS measurements at high latitudes covered a wide dynamic range of the atmosphere. The water vapor monthly equivalent latitude zonal mean compiled from the ILAS data contains many key features of stratospheric circulation from around 25 degrees equivalent latitude to the poles. Quantitatively, ILAS equivalent zonal mean water vapor agrees well with the HALOE measurements during the same time period in the lower stratosphere (within 1 standard deviation of both data sets below 800 K) and shows high bias in the upper stratosphere (around 10% for most months for 800–1800 K, higher in the SH February to March time period). Comparisons with the UARS water vapor climatology show qualitative agreements as well as differences. The most remarkable difference is the spatial and temporal extent of the 1996–1997 Arctic polar vortex, as inferred from the distribution of water vapor. Compared to the UARS climatology, the ILAS data show that the extent of the 1996–1997 Arctic vortex is similar to that of the Antarctic vortex. HALOE water vapor observations during the same period support that this difference reflects atmospheric variability rather than ILAS data biases. We use a tracer transport simulation to show that this difference is due to a more persistent NH polar vortex for the record cold 1996–1997 NH winter-spring period. The wind and PV statistics show that the polar jet during spring 1997 was much stronger and provided a stronger barrier to the transport across that vortex edge.
 Complementary to the correlative data comparisons [Kanzawa et al., 2002], the comparisons of climatological features shown in this paper serve to characterize the overall quality and the limitations of the ILAS water vapor data. The good agreement between ILAS and HALOE monthly equivalent zonal means in the Arctic lower stratosphere and the differences between the ILAS climatology and the long term UARS climatology in this region provide an excellent example of trace gas distribution in response to the changes in stratospheric circulation. The consistency of the observed water vapor with the atmospheric variability validates the scientific value of the ILAS V5.20 water vapor product.
 The authors thank three anonymous reviewers for their helpful comments on the manuscript. This work is supported in part by the National Science Foundation through its support to the University Corporation for Atmospheric Research, by the NASA Upper Atmosphere Research Satellite guest investigator program, and by the NASA Atmospheric Chemistry Modeling and Analysis Program.