2.4.7 INTERCOMPARISON OF SAGE II AND HALOE O3 RETRIEVALS USING DYNAMICAL MAPPING TECHNIQUES. 2.4.7.1 INTRODUCTION. The concept of employing dynamical techniques for the intercomparison of sparse data sets was introduced in Section 2.2.4 above. Here the Trajectory Mapping (TM) and Coordinate Mapping (CM) techniques will be considered in more detail, and the results of their application to an intercomparison of HALOE and SAGE II ozone data will be presented and discussed. Sections 2.4.7.2 and 2.4.7.3 describe the CM and TM techniques, respectively, and discuss their advantages and disadvantages as data intercomparison tools. In Section 2.4.7.4 their effectiveness is verified through comparison with the more conventional approaches for a test period around March 1995. Because SAGE and HALOE made observations at similar latitudes for an extended portion of this time period, the conventional techniques are expected to produce their most accurate results. Section 2.4.7.5 deals with the results of a CM-based intercomparison of SAGE and HALOE ozone, over the full 5-year UARS period, and examines the long-term mean differences, the seasonal variations and UARS-period-mean trends in SAGE, HALOE and their differences. In Section 2.4.7.6 the implications for long-term trend analysis are considered. 2.4.7.2 COORDINATE MAPPING. The general concept of employing quasi-conservative coordinate mapping as a means of increasing the utility of available stratospheric trace gas data dates back to McIntyre (1980), who referred to it as a “modified Lagrangian mean” approach to stratospheric analysis. The method, formally developed by Schoeberl and Lait (1993), has been utilized in recent years as a powerful tool with numerous applications. Amongst these are stratospheric trace gas data analysis [Lait et al. 1990; Schoeberl et al., 1989; Atkinson, 1993], transport case studies [Atkinson and Plumb, 1997], two and three dimensional chemical transport model initialization and verification [Lary et al., 1995], trend analysis [Schoeberl et al., 1989; Randel and Wu, 1995], data intercomparison [Redaelli et al., 1994], and operational total ozone forecasting [Atkinson et al., 1997]. It has also been proposed as a useful tool for meteorological data assimilation and numerical weather prediction [Riishojgaard et al.,1996]. The technique involves a transformation of trace gas data from physical space (latitude- longitude-pressure) into a flow-following “dynamical” reference frame such as potential vorticity (PV) – potential temperature (PT) coordinates, which has a number of advantages over classical techniques, particularly in application to data intercomparison. Because the dynamical coordinates move with the meteorology when the flow is adiabatic, it is possible in this reference frame to reduce the substantial contribution to “local” variance in ozone concentrations due to short-period meteorological variability, which hinders intercomparisons using classical techniques. Also, since the instantaneous large-scale stratospheric wind field is typically closely-aligned along isopleths of isentropic PV, ozone mixing tends to be very rapid along a PV contour as compared to mixing across it. The isentropic ozone mixing ratio contours therefore closely mimic those of the PV [Leovy et al., 1985; Butchart and Remsberg, 1986; Manney et al., 1995]. So over suitably short periods (short compared to photochemical and diabatic time-scales), “longitudinal” and temporal variations in this dynamical reference frame (along PV contours) will be small and should contribute little to local variance. Dynamically-mapped data can provide a picture of the “modified Lagrangian mean” meridional cross-section of ozone, which, while generally similar to the classical zonal-mean, period-mean picture (identical in the absence of meteorological variability), should be both relatively unclouded by the presence of meteorological fluctuations, and more geographically extensive. In this way, the CM approach to data intercomparison is equivalent to the historical intercomparison of zonal mean, monthly mean distributions, except that by choosing coordinates which themselves move with the meteorology, the results obtained should come with higher statistical significance and greater global coverage. The only disadvantages of the technique are the need for access to daily global three-dimensional meteorological analyses, and an added contribution to local variance arising from errors in those meteorological analyses. But given the availability of high quality analyses, the technique is a computationally inexpensive and efficient means of intercomparing the sparse data sets, as will be shown below. 2.4.7.3 Trajectory Mapping As noted in Section 2.2.4 above, trajectory mapping (TM) is another dynamical technique useful for the intercomparison of sparse data sets. Whereas the CM approach discussed above has its equivalent in the intercomparison of monthly zonal mean data, the TM technique is the “dynamical” equivalent of the traditional coincidence approach. Like coordinate mapping, it takes advantage of quasi-conserved quantities following the motion, namely the mixing ratio and potential temperature. Morris et al. (1995) noted that the latter of these is reasonably well conserved for about 7 to 10 days. To create a synoptic trajectory map from satellite data, air parcels are initialized in the model at the time and location of each satellite observation. These air parcels are then isentropically advected forward or backward in time by the model using analyzed wind fields. Synoptic, global constituent maps can then be produced at any time during the analysis period. Because trajectory maps include observations from a number of days, they yield substantially enhanced coverage compared to asynoptic maps produced from a single day of measurements. Furthermore, by correctly accounting for dynamical changes in the atmosphere between observations, TM, like CM, provides better representations of the measured constituent fields than asynoptic schemes. Trajectory mapping can be straightforwardly applied to data intercomparison, as shown in Morris (1994), Morris et al. (1995) and Morris et al. (1997). By creating trajectory maps of observations from one instrument at the time of the observations of a second instrument, individual pairs of observations (which may have been made at widely different locations and times) can closely coincide and be directly intercompared. Frequently, these pairs would have failed the coincidence criteria used in the traditional approach and would therefore have been eliminated from consideration. The use of trajectory mapping increases the coincidences available for data validation while reducing the dynamical component of the variance, as shown in Morris et al. (1997) and will be demonstrated below. The main disadvantages of the TM technique are the need for accurate global wind velocity fields (like the CM approach) and the high computational cost (unlike CM) which makes the technique impractical for intercomparisons over decadal time-scales. 2.4.7.4 Technique Verification. The period February 26 to April 19, 1995 (hereafter called “March 1995”) was selected to assess the effectiveness of each dynamical technique. Figure 2.4.7-1 shows the data gathering patterns of HALOE and SAGE during this time period. Note that during this time period, HALOE and SAGE actually observe at the same Northern Hemisphere latitudes (64o-66oN) for an unusually long period of time. The analysis comprised several parts. First, standard coincidence criteria were applied to compare the data sets on nine isentropic surfaces between 400 and 1200 K. Then the trajectory mapping technique was applied to the same data. Finally, the coordinate mapping technique was applied to the two data sets, and the modified lagrangian mean ozone distributions were compared. Global National Centers for Environmental Prediction (NCEP) analyses were used for both dynamical techniques. Unless noted otherwise, only HALOE data with quality flags less than the corresponding observations and SAGE data with quality flags less than 12% of the corresponding observations were included in these studies. 2.4.7.4.1 Coincident Comparisons. When HALOE and SAGE retrieve ozone in the same latitude bands, the traditional coincidence approach should enjoy its greatest success. Figure 2.4.7-2 shows a scatter plot of the coincident ozone measurements retrieved by HALOE and SAGE on the 800 K (~30 km) potential temperature (PT) surface for the period March 12 – April 5, 1995. If we consider HALOE and SAGE to be coincident when the HALOE measurement is made within 400 km and 12 hours of the SAGE observation we find 120 coincidences (a stricter time criterion of 10 hours results in fewer than 10 coincidences). Roughly 20% of the 620 measurements made by each of HALOE and SAGE during this time period, therefore, can be correlated in this way. Furthermore, the coincident approach can only comment on the relationship between the two instruments in the restricted latitude bands of concurrent observation (e.g., 64o – 66oN, 29o – 35oS). Figures 2.4.7-3a and b depict the number of traditional coincidences in the Northern and Southern Hemispheres respectively on the 800 K PT surface as a function of latitude using 5o latitude bins for the March 1995 period. Notice that only 6 observations satisfy the coincidence criteria in the Southern Hemisphere during the entire study period. Such infrequent coincidences hamper validation efforts. The temporal differences between and sparse nature of the two data sets virtually necessitates use of the standard zonal mean to produce meaningful comparisons. 2.4.7.4.2 Zonal Mean Comparisons. The left-hand panels of Figures 2.4.7-4a and b show a comparison of standard zonal mean ozone profiles from HALOE and SAGE for the period midnight March 18 – midnight March 21, 1995 in the Northern and Southern Hemispheres respectively. During this 3-day period, HALOE and SAGE are observing at the same latitudes in both hemispheres. As a result, the traditional zonal mean approach should produce its most reliable results. The right-hand panels of these figures show the percent difference between HALOE and SAGE ozone measurements as given by (1) Comparisons were performed on both pressure and potential temperature surfaces with generally only small differences in the results. The agreement between the two data sets generally appears quite good (to within 5%), particularly above 25 km. In the Southern Hemisphere below 22 km, larger differences exist between SAGE and HALOE, with HALOE generally lower than SAGE. Disagreements in the lower stratosphere may be due to the presence of aerosols. 2.4.7.4.3 Trajectory Mapping Approach. The use of trajectory mapping (TM) in satellite data validation [Morris et al., 1995; Morris et al., 1997] is summarized in Figure 2.4.7-5. This figure shows a synoptic trajectory map of the HALOE ozone measurements on the 800 K potential temperature surface for 12:00 G.M.T. on March 19, 1995. Four weeks of HALOE data (March 5 - April 2) were advected with a combination of forward and backward trajectory calculations to produce this map. Each HALOE measurement is initialized in the model as a tightly packed cluster of 5 parcels (small dots in the figure). The model then advects these parcels to 12:00 G.M.T. on March 19. Overplotted in the map are the SAGE ozone observations (larger triangles) made on March 19. Those HALOE parcels within 400 km and observed within n days of the SAGE observations on March 19 are considered to be coincident. As shown in Morris, et al. (1995), the advection of parcels over short time periods should not lead to substantial errors. The accuracy of the comparisons, therefore, should not be significantly reduced by short advection times. Figure 2.4.7-6 shows scatter plots of coincident, 3-day, trajectory-mapped HALOE averages versus the SAGE observations during the period March 12 – April 5, 1995 at the 800 K level. This figure can be compared to that produced with the traditional approach (Figure 2.4.7-2). First, we notice an increase in the number of coincidences over the traditional coincidence approach (202 with TM vs. 120 for the traditional approach). The longer the trajectory calculations are, the larger the increases are, as demonstrated in Figure 2.4.7-3a and b. These figures show the latitude distribution of coincident measurements as determined with a traditional approach and the TM approach with a variety of trajectory calculation durations. As can be seen from the figures, trajectory mapping substantially expands the latitude range at which comparisons can be made, so that the 14-day trajectory calculations provide comparisons at essentially all latitudes from 70oS to 80oN. Unlike the traditional approaches, trajectory mapping allows us to comment on the agreement between HALOE and SAGE at latitudes outside of those at which they concurrently measure. Further, the impact of the technique on the uncertainty of our comparisons is small if not beneficial. Figures 2.4.7-7a (Northern Hemisphere) and b (Southern Hemisphere) show root-mean-square (RMS) differences between HALOE and SAGE ozone observations as computed using traditional and trajectory mapping coincidences as a function of altitude during the March 12 – April 5, 1995 period. For the traditional approach, the RMS difference is calculated on both pressure and potential temperature surfaces. The differences between the results for the two coordinates are minimal in the Northern Hemisphere and likely insignificantly different in the Southern Hemisphere (though larger in magnitude, the number of coincidences in the Southern Hemisphere found using the traditional approach hinders our ability to make any firm conclusions). Notice that for much of the profile in the Northern Hemisphere, the RMS errors associated with trajectory advection periods of up to 7 days duration are actually smaller than those found using the traditional coincident comparison technique. The reason for this behavior can probably be attributed to the asynoptic nature of the traditional comparisons [Morris et al., 1997]. For most of the traditional coincidences, the HALOE and SAGE observations are separated by 10 – 12 hours. During March 1995 at 800 K, mean wind speeds from NCEP are 18 – 20 m/sec (or 68 – 72 km/hour). Over the course of 12 hours, an air parcel observed by HALOE will move an average of ~800 km before the SAGE observation. Such distances amount to twice our correlation criteria of 400 km, which suggests that the instruments often make their measurements in entirely different air masses. The trajectory mapping approach takes into account the dynamical motion occurring in the atmosphere between the time of the two measurements, thereby reducing the meteorological component of the observed variance. In the Southern Hemisphere, the 1.5-day and 3-day TM approaches show decreases in the RMS statistic as compared to the traditional approach on pressure surfaces below 25 km. It is in the lower stratosphere that diabatic and photochemical time scales are longest. Therefore, the TM approach is expected to provide its greatest benefit in this region. Larger RMS errors are observed above 25 km for the longer (7-day and 14-day) trajectory calculations compared to the traditional coincident RMS estimates. The increase is consistent with known diabatic time scales over which the isentropic calculations used in this analysis become unreliable [Morris et al., 1995]. To further examine the impact of the trajectory calculations on the uncertainty of the comparisons, profiles of zonal means of the raw SAGE and HALOE data (March 18 – 20, 1995) and the 3-day trajectory-mapped HALOE data were constructed in the coincident latitude bands of observation (64o – 66oN, 29o – 35oS). The zonal mean profiles for (a) the Northern Hemisphere and (b) the Southern Hemisphere are plotted in left-hand panels of Figure 2.4.7.4. The right-hand panels show the percent differences between HALOE and SAGE as defined earlier in Equation (1). The two raw zonal mean profiles generally agree to within 5% throughout the profile in the Northern Hemisphere and above 22 km in the Southern Hemisphere. In both hemispheres, the trajectory mapping results are consistent with those from the traditional approach. In the Northern Hemisphere, the TM zonal-mean HALOE data agree to within 5% of the traditional values between 17 and 35 km. In the Southern Hemisphere, the agreement is to within a few percent throughout the profile. These results suggest that the TM approach is consistent with the traditional zonal mean approach during periods of overlap between the data gathering patterns of the HALOE and SAGE instruments when the traditional approach is expected to perform well. The advantage of using the trajectory approach over the traditional approach, however, is readily apparent during time periods when the two instruments are not making observations at the same latitude, and in the substantial increase in the occurrence of coincidences. Finally, it should be noted that between 18 and 25 km, where the lifetime of ozone is long, the trajectory method appears to introduce errors less than 5% as compared to the standard zonal mean differences. The technique is best used in regions where the constituent lifetime is long and the magnitude of the diabatic circulation is relatively small. Both conditions are better satisfied below 27 km in our HALOE – SAGE ozone comparison. 2.4.7.4.4 Coordinate Mapping Approach. In this study, HALOE ozone data for the period February 26 – April 19, 1995 are transformed using a coordinate mapping (CM) technique from latitude – longitude – pressure coordinates to potential vorticity – isentropic (PVI) coordinates using Modified PV (MPV; Lait, 1994) and potential temperature (PT) from NCEP analyses interpolated to the HALOE measurement locations. The HALOE data are then interpolated onto PT surfaces with ~1 km vertical resolution and separated into bins of 2 PVU (1 PVU= 3 X 10-6 K m2 kg-1 s-1) on each level. Measurements in each bin are averaged to define a composite field in PV – PT space, discarding bins with fewer than 2 measurements. Each composite field consists of data collected for periods of +/- 10 days (21 days total). Composite fields are constructed every 5 days, and in-between days are linearly interpolated from the nearest two days on which composite fields have been constructed. Finally, synthetic HALOE ozone is calculated from the reverse transform back to physical space using the appropriate NCEP dynamical data. Note that HALOE and SAGE data are excluded which have corresponding quality flags in excess of 12% of the observation values. Figure 2.4.7-8 shows the 2o latitude by 5o longitude reconstructed HALOE data on March 19, 1995 at 800 K. This map can be compared to the HALOE trajectory map of Figure 2.4.7-5. Both techniques use ~3 – 4 weeks of data and both give nearly global coverage for the day. The absolute values of ozone and the large-scale structures agree well between the two techniques. However, many smaller-scale features appear quite different. The CM reconstruction technique depends on daily dynamical data to produce an ozone map. Noise in these data fields will result in a noisy reconstructed ozone field. This may contribute to the noise observed in the subtropical CM ozone, where the uncertainty in the NCEP analysis can be comparatively large. Near the equator, the CM ozone is smoother than the trajectory-mapped ozone. The lower variability of the PV field in this region results in lower correlations with ozone. As a result, the CM technique will not capture variability in the tropics. Fortunately, the ozone field is also relatively undisturbed in this region. Figure 2.4.7-9 shows a root-mean-square (RMS) statistic of the variability of HALOE and SAGE observations about their respective zonal mean values (in the case of the traditional approach) and modified Lagrangian mean values (in the CM case) for (a) the Northern Hemisphere and (b) the Southern Hemisphere. For the traditional approach, data were binned and averaged in 5o latitude bins while for the CM approach, data were binned and averaged in 5o equivalent latitude (see Section 2.4.7.5.1 for an explanation) bins on each of 24 PT surfaces between 300 and 1200 K. The sum of the squares of the differences between each observation and its associated zonal mean value was computed in each bin on each PTS and the quantity RMS reported as a function of altitude given by (2) where i is the latitude bin, nlat is the total number of latitude bins in each hemisphere, Nk is the total number of points on the kth PT surface in the hemisphere, ?i,j,k is each individual measurement and ?i,k is the mean bin value. The figures demonstrate that in the Northern Hemisphere, the CM approach produces lower RMS differences than the traditional approach for both SAGE and HALOE throughout the profile, with the exception of a small region around 500 K where the traditional approach is marginally better. In the Southern Hemisphere, the CM approach produces results comparable to the traditional approach for altitudes below about 700 K. As with the trajectory mapping approach, the CM approach is expected to perform best in the lower stratosphere when photochemical and dynamical time scales are relatively long. Figure 2.4.7-10 shows the comparisons between coordinate mapped HALOE and raw SAGE measurements for the same time period as shown for the trajectory mapping case above. Because the technique produces near global fields from the HALOE observations, nearly every SAGE measurement can be validated. Note that the CM technique is the only technique to produce coincidences in the tropics (20oS – 20oN) during this time period (see Figures 2.4.7-2, 2.4.7-6, and 2.4.7-10). Zonal means of the coordinate mapped HALOE ozone are compared to zonal means of the coordinate mapped SAGE measurements as a function of height in Figure 2.4.7.4. Here the zonal mean, coordinate mapped HALOE data (denoted CM HALOE in the figure) on March 19, 1995 is plotted for 65oN. The agreement between the CM HALOE and CM SAGE zonal mean profiles is within 5% of the traditional results from 15 – 30 km in both hemispheres. Again, this demonstrates the validity of the technique as compared to the traditional approach in a latitude band and during a time period when the traditional approach is performing its best. However, unlike the traditional approach, CM can comment on the agreement between SAGE and HALOE at latitudes and during time periods outside of those when SAGE and HALOE are coincidentally observing the same latitudes. 2.4.7.5 COORDINATE MAPPING BASED INTERCOMPARISON OF SAGE II WITH HALOE OVER THE UARS PERIOD. The previous section dealt with verification of the performance of the alternative dynamical intercomparison techniques. Here we focus on the results of a CM-based intercomparison of the ozone profile data sets obtained by SAGE II and HALOE during the entire UARS period. 2.4.7.5.1 ANALYSIS TECHNIQUE. Each SAGE and HALOE profile from 20 September 1991 to 31 December 1996 was first interpolated to each of 24 isentropic levels between 300K and 1200K. Daily NCEP stratospheric analyses were then used to transform all the data for each level into PV-like coordinates. The specific coordinate used here was “equivalent latitude”, originally proposed by Butchart and Remsberg (1986) and later used by Lary et al. (1995). It is a PV-like coordinate obtained from the PV distribution, which allows particular ease of interpretation, since it is the “dynamical” equivalent of geographic latitude itself, and reduces to latitude when the flow is zonally symmetric, or where the dynamical time-scale is much longer than chemical or radiative time-scales. To reduce the effects on the analysis of aerosol-affected ozone retrievals, and on the advice of the SAGE instrument team, all SAGE retrievals were eliminated from the analysis if their estimated error was more than 12% of the retrieved value itself. On similar advice from the HALOE instrument team, a 99% error limit was imposed on the HALOE data. For each instrument, each quality-filtered isentropic data set was next interpolated to a regular two dimensional grid in time versus equivalent latitude, to provide a detailed broad-scale picture of the daily evolution of the global ozone distribution depicted by each instrument over the UARS period. A Barnes-type, observational error-weighted, two-dimensional gaussian interpolation scheme was used for the analysis, with temporal and "latitudinal" gaussian half widths of approximately six weeks and four degrees, respectively. The analysis grid dimensions were 1 day (UARS day number 1 to 2000) by two degrees (90 equivalent latitude bins from 90oS to 90oN). Explicitly, the isentropic ozone mixing ratio (?) analyzed at grid point (i,j) was determined from the N quality-filtered values observed by the instrument at that isentropic level (anywhere on the globe and at any time during the UARS period), as (3) where (4) and where ?n and tn are the equivalent latitude and UARS day number of observation n, ?i,j and ti,j are the equivalent latitude and UARS day number of grid point (i,j), X is the equivalent latitudinal decorrelation scale (approximately 2 degrees), ??is the temporal decorrelation scale (approximately 20 days), and ?(?n) is root-sum-square error of the retrieved ozone for observation n (as a fraction of the retrieved value). By retaining both the analyzed values themselves and the sums of the weights at each grid point, a semi-quantitative measure of “analyzed data quality” was preserved for the subsequent intercomparison of the analyses from either instrument. A “minimum weighting” limit (of 10) was imposed to eliminate spurious extrapolated values from consideration. At each isentropic level and equivalent latitude the time series of analyzed values from each instrument were then intercompared to provide meridional cross sections (equivalent latitude versus potential temperature) of UARS-period mean differences between the data sets, seasonal variations in these differences, the simple least squares linear trends in ozone depicted by each instrument, and corresponding trends in the differences between the instruments, the latter of which should be indicative of UARS period mean drift between SAGE and HALOE. 2.4.7.5.2 RESULTS. Figure 2.4.7-11 shows a selection of UARS-period time series of the data from each of SAGE and HALOE, and their percentage differences, for selected values of potential temperature and equivalent latitude. The “locations” shown were chosen for their relevance to the discussion following, but together they demonstrate the degree of effectiveness of the analysis technique in describing the “modified Lagrangian mean” ozone evolution observed by each instrument during the period. The seasonal data gaps evident in Figures 2.4.7-11a and d, for the high southern latitudes, suggest that any results for this region are likely to be seasonally biassed, and the analyzed curves suggest that a harsher minimum weighting criterion might have been used to limit data extrapolation. Figures 2.4.7-11e, g, and h reveal the very limited record available for the calculation of trends in the lower stratosphere. Figure 2.4.7-12 shows (a) the UARS period mean meridional cross section of SAGE ozone mixing ratio, and (b) the period mean percentage difference between SAGE and HALOE given by Equation (1) above. Several interesting features are evident in Figure 2.4.7-12b. In general, SAGE sees considerably more ozone than HALOE throughout the lower stratosphere, more so with decreasing altitude (see Figures 2.4.7-11g and h). Above about 600K (~25 km) SAGE and HALOE typically agree to within a few percent, though larger differences are present in the mid- to high latitude upper stratosphere and the tropical middle stratosphere (around 700K) where HALOE ozone exceeds SAGE by 2 to 5%. In the mid-latitude middle stratosphere, SAGE exceeds HALOE by up to 5%. An examination of the seasonal variations in the differences (not shown) indicates that the upper stratospheric differences are largely a summer feature (see Figure 2.4.7-11a and c). Figure 2.4.7-11e suggests that HALOE may see a stronger QBO cycle in ozone than SAGE in the tropical middle stratosphere below the ozone maximum. Figure 2.4.7-13 shows the pattern of simple least squares linear trends in the differences between the two instruments. The shading is indicative of the relative level of confidence that can be placed in the trend values (light/dark shading – high/low confidence), but because the temporal gaussian smoothing carried out in the analysis has reduced the independence of the analyzed values from one day to the next, the confidence statistics will generally be gross overestimates and can only be used as a crude comparative guide. The analysis suggests a tendency during the UARS period for HALOE observations at most latitudes in the middle stratosphere to have drifted downward relative to SAGE, by up to a percent or so per year. This is where, according to Figure 2.4.7-12b, period mean HALOE ozone is lower than SAGE, suggesting the gap between the two instruments has widened during the UARS period. In the lower stratosphere, the short record length considered makes it meaningless to interpret more than the period mean differences. In the upper stratosphere, particularly in those regions where HALOE ozone is higher than SAGE, the analysis suggests an absence of drift between the instruments. But the above results can not be considered conclusive, because, and as already noted, it is not possible to attach reliable confidence levels to the indicated trends. To address this issue, the analysis was repeated at lower temporal and spatial resolution, to provide temporally independent, data quality-weighted, mid-monthly ozone values at a meridional resolution of five degrees. In this case, although the two-dimensional gaussian weighting scheme was retained, only observations made in the same month were used for the calculation of the monthly value. A selection of the resulting monthly time series is shown as the squares plotted in Figure 2.4.7-11. The temporal decoupling of the data has led to much greater scatter in these values than in the higher resolution, smoother analysis, but on the whole the analyzed values provide a reasonable fit to the “raw” observations. Simple least squares linear trends were again calculated for each instrument and for the differences between the two, and are shown in Figures 2.4.7-14 and 2.4.7-15, respectively. By reducing the degree of smoothing, a much noisier picture results, as was the case for the individual time series, but this analysis provides trends to which we can attach quantitative estimates of uncertainty. In the Antarctic upper stratosphere, Figures 2.4.7-11a, 2.4.7-14 and 2.4.7-15 show a region of statistically significant (at the 95% confidence level) summertime drift between SAGE and HALOE, with SAGE, but not HALOE, showing a significant positive trend in ozone of as much as 1.5% per year during the UARS period. A similar feature is evident in the Arctic middle and upper stratosphere. The two instruments agree on a significant downward trend in tropical upper stratospheric ozone (above 750K) of up to 2% per year during the period (see Figure 2.4.7-11b). Figure 2.4.7-15 shows three regions of significant drift between HALOE and SAGE in the middle stratosphere, where HALOE decreases (relative to SAGE) by up to a few percent per year. The tropical feature, as suggested earlier and in Figure 2.4.7-11e, may be largely attributable to the short SAGE data period and a relatively strong HALOE QBO signal, but the two mid-latitude features are more difficult to discount. At the higher northern mid-latitudes, both SAGE and HALOE show an overall ozone decrease of a few percent per year, but they disagree on the latitudinal extent of this region; HALOE shows a decrease as far equatorward as 40oN, while SAGE only shows it to 50oN. This disagreement is responsible for the region of significant drift shown in Figure 2.4.7-15. Similarly, the Southern Hemisphere feature is the result of a statistically significant difference between the negative trend in ozone seen by HALOE (of a percent or so per year), and the insignificant (positive) trend seen here by SAGE. Again, this is a region where Figure 2.4.7-12b shows HALOE ozone to be lower than SAGE, so the trend analysis suggests this gap has widened during the UARS period. As noted above, and despite the statistical significance of the results, three years’ data coverage in the lower stratosphere is insufficient to draw any meaningful conclusions about trends. Of note in Figure 2.4.7-11g, though, and to a lesser extent in Figure 2.4.7-11h, is an indication that the trends derived for these locations may be unduly influenced by aerosol-affected observations from the Pinatubo period which have escaped rejection by the data quality filter. 2.4.7.6 DISCUSSION AND IMPLICATIONS Both trajectory mapping (TM) and coordinate mapping (CM) techniques are effective methods for comparing “sparse” data sets. They enable (1) a greater number of observations from each instrument to be intercompared (more "coincidences"), (2) increased statistical certainty associated with the intercomparisons, and (3) greater spatial resolution and more global coverage to be achieved. This analysis suggests the TM approach to be more accurate than the CM technique, but it is also much more computationally expensive than CM. As a result, TM is probably better used in closely examining particular pairs of observations, whereas the computational efficiency of the CM approach makes it ideally suited for intercomparing data sets over extended periods. The results obtained for the SAGE/HALOE intercomparison using the CM technique generally support those obtained using the classical “zonal mean” and “coincidence” approaches, but tend to be more significant. The CM intercomparison has also provided insight into instrument performance in regions of the atmosphere where data sparsity inhibits intercomparison by the classical techniques. The CM intercomparison shows SAGE ozone data to be higher than HALOE in the lower stratosphere (at potential temperature levels below 550K, or about 22 km altitude), particularly in the tropics. These differences increase from about 5% at 550K to more than 20% below 450K. Note, however, that the period of intercomparison is only two to three years at the lowest levels, so the applicability of these differences to other periods is not known, and no confidence can be placed in indicated trends for this region, no matter how significant. At higher levels in the atmosphere (above 550K) SAGE and HALOE generally show “UARS period mean” agreement to within a few percent, except in localized regions where differences of up to several percent exist. These include the summertime extra-tropical upper stratosphere, where HALOE ozone is up to 5% higher than SAGE, and the mid-latitude middle stratosphere, where HALOE is up to 5% lower than SAGE. Above 800K the CM analysis suggests that the UARS-period mean drift between SAGE and HALOE is generally small (less than 1/2% per year) and insignificant, except at polar latitudes. In the polar regions, summertime HALOE observations have drifted downward significantly compared to SAGE (about 2% per year), with SAGE during the period showing a significant ozone increase not detected by HALOE. In the middle stratosphere (600 to 800K) the analysis shows a significant downward drift of up to 2% per year in mid-latitude HALOE ozone (relative to SAGE). A similar feature was evident in both of the analyses performed, and is in agreement with the marginally significant results obtained from the classical intercomparisons reported in the previous section. The significant drift between instruments indicated by this analysis in the tropical middle stratosphere, while statistically significant, may be largely a reflection of a QBO cycle in the data. Below 550K the analysis shows large and significant SAGE/HALOE drift (SAGE generally drifting downward towards HALOE), but the period examined is too short for these results to be meaningful and the results may anyway be influenced by aerosol-affected data. Two important qualifications should be made regarding the above trend results. First, they are based on just five years of data, at most: whether or not the analyzed drifts between the two instruments, for example, are representative of instrument behavior over longer time scales, has not been addressed here. Second, were there to be erroneous temperature trends in either the NCEP stratospheric analyses or the HALOE temperature retrievals, this could have a significant impact on the trends analyzed here. The SAGE ozone retrieval is based on first guess temperature and pressure profiles interpolated from the NCEP analyses, whereas HALOE retrieves its own temperature and pressure profiles. The first step in the CM-based analysis was the interpolation of all ozone profiles to isentropic levels. For this, the temperature profiles provided by the respective instrument team with each ozone profile were used (NCEP for SAGE, HALOE for HALOE), so any overall drift between HALOE and NCEP temperatures would induce an isentropic ozone trend wherever there is a vertical gradient in ozone. Perhaps as important, were the NCEP analyses to show erroneous trends, these could impact directly on the SAGE retrievals themselves, through the pressure dependence of Rayleigh scattering in the SAGE ozone retrieval. On the other hand, real temperature trends depicted by both NCEP and HALOE should not influence the results presented here. With the above qualifications in mind, there are some implications of the present analysis for long- term trend analysis using the SAGE data set. The CM intercomparison results for the mid-latitude middle stratosphere (between about 600 and 800K) and their general agreement (at low confidence) with the more classical analyses of SAGE, HALOE, and MLS data raise doubts about the ability of SAGE to detect mid-stratospheric ozone trends at middle latitudes to better than a percent or so per year. The analysis for the lower stratosphere also suggests care should be exercised in interpreting trends based on data that include aerosol-affected measurements. Given the documented sensitivity of SAGE ozone retrievals to stratospheric aerosol, this analysis suggests, in support of the inferences drawn from the classical analyses considered elsewhere in this assessment, that it may be advisable to time-filter, rather than quality-filter, SAGE data to remove those retrievals which might be so affected. REFERENCES: Atkinson, R.J., An observational study of the Austral spring stratosphere: dynamics, ozone transport and the “ozone dilution effect”, Sc.D. Thesis, Massachusetts Institute of Technology, 1993. Atkinson, R.J., Grainger, S., and P..M. Udelhofen, A scheme for the routine analysis and deterministic prediction of the global ozone distribution, Proc. AMS Conference on the Meteorology of the Southern Hemisphere, Pretoria, South Africa, 1997. Atkinson, R. J., and R. A. Plumb, Three-dimensional ozone transport during the ozone hole breakup in Dec. 1987, J. Geophys. Res., 102, 1451-1466, 1997. Butchart, N., and E. Remsberg, The area of the stratospheric polar vortex as a diagnostic for tracer transport on an isentropic surface, J. Atmos. Sci., 43, 1319-1339, 1986. Lait, L. R., An alternative form for potential vorticity, J. Atmos. Sci., 51, 1754-1759, 1994. Lait, L. R., M. R. Schoeberl, P. A. Newman, M. H. Proffitt, M. Loewenstein, J. R. Podolske, S. E. Strahan, K. R. Chan, B. Gary, J. J. Margitan, E. Browell, M. P. McCormick, and A. Torres, Reconstruction of O3 and N2O fields from ER-2, DC-8, and balloon observations, Geophys. Res. Lett., 17,521-17,524,1990. Lary, D.J., M.P. Chipperfield, W.A. Norton, and L.P. Riishojgaard, Three-dimensional tracer initialization and general diagnostics using PV latitude-potential-temperature coordinates, Quart. J. Roy. Meteor. Soc., 121, 187-210, 1995. Leovy, C. B., et al., Transport of ozone in the middle stratosphere: Evidence for wave breaking, J. Atmos. Sci., 42, 230-244, 1985. Manney, G.L., L. Froidevaux, J.W. Waters, and R.W. Zurek, Evolution of microwave limb sounder ozone and the polar vortex during winter, J. Geophys. Res., 100, 2953-2972, 1995. McDermid, I. S., et al., Comparison of ozone profiles from ground-based lidar, electrochemical concentration cell balloon sonde, ROCOZ-A rocket ozonesonde, and stratospheric aerosol and gas experiment satellite measurements, J. Geophys. Res., 95, 10,037-10,042, 1990. McIntyre, M. E., Towards a Lagrangian-mean description of stratospheric circulations and chemical transports, Phil. Trans. R. Soc. Lond., 296, 129-148,1980. Morris, G. A., A Demonstration and Evaluation of Trajectory Mapping, Ph.D. Thesis, Rice University, 1994. Morris, G. A., et al., Trajectory mapping and applications to data from the Upper Atmosphere Research Satellite, J. Geophys. Res., 100, 16,491-16,505, 1995. Morris, G.A., et al., Satellite data validation using trajectory mapping, Part I: Accuracy assessments, J. Geophys. Res., to be submitted, 1997. Randel, W.J. and F. Wu, TOMS total ozone trends in potential vorticity coordinates, Geophys. Res. Lett., 22, 683-686, 1995. Redaelli, G., L. R. Lait, M. Schoeberl, P. A. Newman, G. Visconti, A. D'Altorio, F. Masci, V. Rizi., L. Froidevaux, J. W. Waters, and A. J. Miller, UARS MLS 03 soundings compared with lidar measurements using the conservative coordinates reconstruction technique, Geophys. Res. Lett., 21, 1535-1538, 1994. Riishojgaard, L.P., On four-dimensional variational assimilation of ozone data in weather prediction models, Q.J.R. Meteorol. Soc., 122, 1545-1571, 1996. Schoeberl, M. R., and L. R. Lait, Conservative-coordinate transformations for atmospheric measurements, The use of EOS for studies of atmospheric physics, Proceedings of the International School of Physics Enrico Fermi, Italian Phys. Soc., 419--431, 1993. Schoeberl, M. R., L. R. Lait, P. A. Newman, R. L. Martin, M. H. Profitt, D. L. Hartmann, M. Loewenstein, J. Podolske, S. E. Strahan, J. Anderson, K. R. Chan, and B. Gary, Reconstruction of the constituent distribution and trends in the Antarctic polar vortex from ER-2 flight observations, J. Geophys. Res., 94, 16,815-16,845, 1989. FIGURE CAPTIONS Figure 2.4.7-1. The latitudes at which HALOE and SAGE make their observations during the March 1995 study period. The black lines represent HALOE while the gray lines represent SAGE. Dashed lines indicate sunset measurements while solid lines indicate sunrise measurements. Figure 2.4.7-2. HALOE vs. SAGE ozone using traditional coincidence criteria of 400 km maximum separation and 12 hour maximum time difference for the period March 12 – April 5, 1995 at the 800 K potential temperature surface (~10 mb or 30 km). Figure 2.4.7-3. Probability distribution functions of coincidences in the (a) Northern Hemisphere and (b) Southern Hemisphere using traditional criteria (black, solid lines) and the trajectory mapping approach (gray lines) as a function of latitude. Note that the trajectory approach greatly expands both the number of coincidences and the latitude-range over which coincidences occur as compared to the traditional approach. As the duration of the trajectory calculations increases, the coverage increases. With 14-day trajectory calculations, coincidences can be found at nearly all latitudes between 70oS and 70oN. Figure 2.4.7-4. HALOE vs. SAGE ozone profiles (left panels) and percent differences (right panels) in the (a) Northern Hemisphere and (b) Southern Hemisphere. Zonal mean ozone profiles are calculated and shown between 400 and 1200 K (15 – 40 km) using several different methodologies. Traditional approaches are taken on both pressure and potential temperature (PT) surfaces for HALOE and SAGE (black and light gray). 3-day trajectory mapping (TM; dark gray, solid lines) and coordinate mapping (CM; dark gray, dashed lines) are also applied. Only those measurements found in coincidence pairs are included in the zonal means for the traditional and TM approaches. CM HALOE observations are compared to CM SAGE. In the right panels, the light gray solid line shows the difference between HALOE and SAGE as compared using a traditional approach on pressure surfaces; the light gray, dashed line shows the difference between HALOE and SAGE compared using a traditional approach on PT surfaces. Figure 2.4.7-5. Synoptic trajectory map at 12:00 G.M.T. on March 19, 1995 of HALOE ozone observations (small dots) made between March 5 and April 2 on the 800 K potential temperature surface (~30 km). SAGE observations (large triangles) from March 19 are overplotted. Figure 2.4.7-6. HALOE vs. SAGE ozone on the 800 K (~30 km) potential temperature surface using the trajectory mapping approach. Trajectory calculations are limited to 3 days duration. Advected HALOE measurements within 400 km of the SAGE observations are considered coincident. Figure 2.4.7-7. Root-mean-square (RMS) differences between HALOE and SAGE as a function of altitude (~15 – 40 km) and the duration of the trajectory calculations (1.5 to 14 days) for the (a) Northern Hemisphere and (b) Southern Hemisphere. Also shown are the RMS differences found using a traditional coincidence approach on both pressure surfaces (solid black line) and potential temperatures surfaces (dashed black line). Notice that in the Northern Hemisphere, trajectory calculations of up to 7 days perform as well or better than the traditional approach throughout the profile. In the Southern Hemisphere, the trajectory approach shows improvement in the lower- most stratosphere (below 25 km) as compared to traditional approaches. Figure 2.4.7-8. Synoptic, reconstructed ozone field from coordinate mapped HALOE data for March 19, 1995 at the 800 K (~30 km) potential temperature surface. 21 days of HALOE data were included in the construction of this map. Figure 2.4.7-9. Root-mean-square differences (see text) of individual satellite observations from their zonal mean (traditional) and modified Lagrangian mean (coordinate mapping) value for both HALOE and SAGE for (a) the Northern Hemisphere and (b) the Southern Hemisphere during March 1995. Figure 2.4.7-10. CM HALOE vs. SAGE ozone on the 800 K (~30 km) potential temperature surface using the coordinate mapping approach. Figure 2.4.7-11. UARS period time series of raw and analyzed ozone mixing ratio for HALOE, SAGE, and the percentage difference between the two, at a range of equivalent latitudes and potential temperatures. (a) 76oS at 1100K, (b) 4oS at 1000K, (c) 60oN at 1000K, (d) 64oS at 650K, (e) Equator at 650K, (f) 40oN at 700K, (g) 60oS at 400K, and (h) 4oS at 475K. Crosses denote all quality-filtered observations made within 2 degrees equivalent latitude of the central latitude, the solid curve in each plot is the daily time series of analyzed values, and the envelope described by the dashed curves shows the relative confidence which can be placed in the analyzed value at that point (the smaller the envelope, the higher the confidence). Also shown as filled or open squares are the monthly analyzed values from the alternative analysis (see text). Figure 2.4.7-12. UARS period mean, modified Lagrangian mean cross-section (equivalent latitude versus potential temperature) of (a) SAGE ozone mixing ratio (ppmv), and (b) percentage difference between HALOE and SAGE ozone mixing ratio (i.e. 100*(HALOE-SAGE)/SAGE) Figure 2.4.7-13. UARS period mean, modified Lagrangian mean cross-section (equivalent latitude versus potential temperature) of the drift (%/year) between analyzed SAGE and HALOE ozone mixing ratio. The contours show the drift and the shading provides a crude measure of relative trend certainty (lighter/darker shading indicates higher/lower confidence). Negative drift (HALOE decreasing relative to SAGE) is shown by the dashed contours. Figure 2.4.7-14. UARS period mean, modified Lagrangian mean cross-section (equivalent latitude versus potential temperature) of (a) SAGE and (b) HALOE ozone mixing ratio trends (%/year) obtained from the monthly analysis. Contours show the trends (negative trends dashed) and the shading indicates normalized 95% confidence limits (confidence limit/trend). Values less than unity (light shading) therefore indicate trends that are significant at the 95% confidence level. Figure 2.4.7-15. As for Figure 2.4.7-13, but showing the UARS period mean, modified Lagrangian mean cross-section (equivalent latitude versus potential temperature) of the drift between monthly analyzed SAGE and HALOE ozone (%/year).