Polar stratospheric cloud (PSC) detection, top height, and composition as derived from measurements by the infrared emission limb sounder Michelson Interferometer for Passive Atmospheric Sounding (MIPAS) on Envisat and the Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) lidar have been compared. The comparisons are based on coincident observations from the 2006 and 2007 Antarctic and the 2006/2007 and 2007/2008 Arctic winters. During the middle of the Antarctic season, very good agreement in common PSC detection (around 90%) has been found. At the beginning and end of the Antarctic PSC season and in the Arctic, the frequency of common PSC detection is generally less (60–70%) which can be explained by cloud inhomogeneity and viewing geometry differences. MIPAS PSC top heights are about 0–2 km lower than CALIPSO top heights with larger offsets at higher altitudes. The negative bias of MIPAS PSC top heights can be modeled under the assumptions of limited horizontal cloud extent and a field-of-view–dependent sensitivity. The comparisons further show a high degree of consistency between PSC composition derived from the fundamentally different classification approaches of the two instruments. Remaining differences can be explained considering the physical limitations of each approach and the definition of composition boundaries within the classification scheme of each instrument.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Polar stratospheric clouds (PSCs) are key components within the chemical processes leading to polar ozone depletion [Solomon, 1999]. First, through heterogeneous reactions, chlorine reservoir species like HCl and ClONO2 are converted into active forms involved in the catalytic ozone destruction cycles. Second, the removal of NOy by sedimentation of HNO3-containing PSC particles inhibits a fast chlorine deactivation via the reformation of ClONO2. Since the formation of PSCs depends strongly on the temperatures over the polar regions in winter, a cooler stratosphere due to climate change might increase polar ozone depletion in the future [World Meteorological Organization (WMO), 2007; Hitchcock et al., 2009].
 A further problem is the determination of stratospheric heating rates due to PSC coverage over the polar regions. Hicke and Tuck  state that the major uncertainty in their model calculations is the assumptions about the Antarctic PSC coverage, mainly about ice PSCs, and that only global PSC observations covering the polar region also in the absence of sunlight would solve this problem.
 The only methods to obtain these goals, global PSC measurements independent of sunlight and PSC composition discrimination, are spaceborne lidars and limb sounders measuring atmospheric emission in the midinfrared region.
 The first observations of PSCs by lidar from space were performed by the Geoscience Laser Altimeter System (GLAS) on the Ice, Cloud and land Elevation Satellite (ICESat) [Spinhirne et al., 2005; Palm et al., 2005]. However, since GLAS did not have polarization sensitivity, a PSC composition discrimination was not possible. Such measurements, however, have been conducted since June 2006 by the Cloud-Aerosol LIdar with Orthogonal Polarization (CALIOP) as part of the Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) mission [Pitts et al., 2007; Noel et al., 2008; Wang et al., 2008; Pitts et al., 2009].
 The goal of the present study is the intercomparison of the simultaneous CALIPSO and MIPAS PSC data sets from June 2006 through February 2008 with respect to PSC detection (section 4.1), PSC top altitude (section 4.2), and PSC composition (section 4.3). With regard to the last item it is of interest how well the PSC compositions derived from totally different methods (i.e., lidar depolarization versus IR spectroscopic) compare. An initial such study was performed by Höpfner et al. [2006a], who compared MIPAS data during the Antarctic winter 2003 with ground-based lidar observations from McMurdo. In that paper a relatively good correspondence of the PSC composition was found, albeit with a very limited number of matching observations in time and space. The present comparison between the two spaceborne instruments overcomes this limitation due to the much larger number of coincident PSC observations.
2. Instruments and PSC Processing
2.1.1. Instrument and Data Set
 The primary instrument on CALIPSO is a dual wavelength, polarization sensitive lidar that provides high vertical resolution profiles of backscatter coefficient at 532 and 1064 nm, as well as two orthogonal (parallel and perpendicular) polarization components at 532 nm [Winker et al., 2007]. The lidar pulse rate is 20.25 Hz, corresponding to one profile every 333 m along the orbit track, and the lidar beam has approximately a 100 m diameter spot size on the Earth's surface. Although the fundamental sampling resolution is 30 m in the vertical and 333 m in the horizontal, an altitude-dependent on-board averaging scheme is employed that provides highest resolution in the troposphere and lower resolution in the stratosphere. For PSC analyses, the lidar data are averaged to a uniform grid with 5 km horizontal and 180 m vertical resolution [Pitts et al., 2009]. PSC detection is currently limited to nighttime CALIPSO observations due to the presence of higher levels of background light during the daytime, which reduces the PSC detection sensitivity. As part of the A-train satellite constellation, the CALIPSO spacecraft is in a 98° inclination orbit that provides extensive coverage over both polar regions with measurements extending to 82° latitude. On average, over 300,000 lidar profiles are acquired per day in each hemisphere at latitudes poleward of 55°, providing a unique data set for studying PSCs. The CALIPSO lidar has been operating nearly continuously since routine data acquisition began in June 2006 and has already produced an extensive set of PSC observations for both the Antarctic and Arctic.
2.1.2. PSC Detection
 PSCs are identified in the CALIPSO lidar data as enhancements above the background aerosol ensemble in terms of either lidar scattering ratio (ratio of total and molecular backscatter coefficients) at 532 nm, R532, or perpendicular backscatter coefficient at 532 nm, β⊥532. PSC detection thresholds are calculated daily as a function of altitude based on statistical analyses of the background aerosol ensemble measurements. Individual profiles of R532 and β⊥532 at the standard 5 km horizontal by 180 m vertical resolution are scanned from the top altitude (30 km) downward for data points that exceed the cloud thresholds. PSCs with R532 greater than 2.75–3.0 are easily detected at the 5 km horizontal resolution. A successive horizontal averaging scheme enables detection of more tenuous clouds at increasingly coarser scales (15, 45, 135 km), the results of which are replicated onto the 5 km grid. Details of the CALIPSO PSC detection algorithm are described by Pitts et al. .
2.1.3. PSC Composition
 The composition of PSCs detected by CALIPSO is specified using the algorithm described in detail by Pitts et al. , which was developed based on optical model calculations for mixtures of spherical STS droplets and oblate spheroidal NAT (or ice) particles, assuming nominal vapor mixing ratios of 10 ppbv HNO3 and 5 ppmv H2O at 50 hPa atmospheric pressure. The algorithm is applied in a coordinate system of aerosol (particle only) depolarization ratio (δaerosol) versus inverse scattering ratio (1/R532), which has been used successfully in many previous lidar-based PSC studies [e.g., Adriani et al., 2004; Massoli et al., 2006]. Figure 1 shows a composite 2-D histogram of all PSCs observed by CALIPSO during the four winters included in this paper, with the domains of the four CALIPSO composition classes superimposed. “Mix1” and “Mix2” denote mixed-phase STS-NAT clouds characterized by generally lower and higher NAT particle number density/volume, respectively. Data falling in the “ice” class are clearly indicative of H2O ice; however, these are very likely mixtures of STS and ice, except in the case of wave clouds. Data falling in the “STS” class have an ensemble optical signature dominated by STS; these may be purely STS clouds or STS-solid mixtures in which the presence of solid particles is masked by the more numerous liquid droplets.
 In terms of comparison with more traditional lidar-based PSC classification schemes, the CALIPSO STS and ice classes can be viewed as analogs to Type 1b and Type 2 PSCs, respectively. Together, the CALIPSO Mix 1 and Mix 2 classes can be thought of as analogous to the traditional Type 1a PSC class. However, we strongly believe that almost all PSCs are external mixtures of liquid and solid particles, and in particular, it is very unlikely that pure NAT (i.e., “Type 1a”) PSCs exist in the atmosphere.
2.2.1. Instrument and Data Set
 MIPAS is a Fourier transform spectrometer sounding the midinfrared radiation emitted (in case of trace gases) or emitted and scattered (in case of clouds) [Höpfner et al., 2002] between 685 and 2410 cm−1 (14.6–4.15 μm) in limb geometry from the Sun-synchronous polar orbiting satellite Envisat [Fischer et al., 2008]. The field of view of the instrument is 30 km in the horizontal across track and 3 km in the vertical at the tangent points. Here, MIPAS data with a tangent altitude grid of 1.5–2 km and horizontal along track sampling steps of 410 km and 290 km have been used. Due to the orbital inclination, the subsatellite points of Envisat cover latitudes only up to 81.45°N/S. By varying the azimuthal pointing of the line of sight out of the orbital plane in the polar regions, the true latitudinal coverage of MIPAS tangent points ranges from 87.5°S to 89.3°N. Thus, in combination with its sensitivity to optically very thin clouds due to the limb view and the independence from any external light source, MIPAS is especially suited for temporally and spatially continuous observations of PSCs during polar winter.
 From July 2002 until March 2004 MIPAS took measurements with maximum optical path difference (OPD) of 20 cm corresponding to a unapodized spectral resolution of 0.025 cm−1. Due to problems with the interferometer drive system, MIPAS observations were interrupted from March 2004 until January 2005, when regular observations were resumed with a reduced spectral resolution (OPD = 8.2 cm) allowing a denser tangent altitude grid. From January 2005 onward, the duty cycle was increased steadily and reached 100% again in December 2007.
 For the present investigation we have used MIPAS measurements of two southern (2006, 2007) and two northern (2006/2007, 2007/2008) polar winters in parallel to the first two years of CALIPSO operation. MIPAS data of the NOMINAL and the upper troposphere and lower stratosphere region 1 (UTLS-1) limb-scan pattern (http://www.atm.ox.ac.uk/group/mipas/rrmodes.html) were analyzed. Compared to MIPAS observations from 2002–2004, which have been used in previous studies on PSCs, the NOMINAL and UTLS-1 scan patterns of the data since 2005 have the advantage of a finer vertical sampling in the lower stratosphere with tangent altitude differences of 1.5–2 km up to about 30 km altitude compared to 3 km steps in the period 2002–2004. The horizontal spacing of subsequent limb scans is about 410 km in case of NOMINAL and 290 km in case of UTLS-1 observations.
 As Kiefer et al.  have shown, the engineering tangent pointing information of MIPAS can be in error by as much as 3 km. What is mostly affected here are all tangent altitudes of one limb scan as a whole, while the relative distances between two nearby tangent points is much more accurate (about 100 m). Thus, using the original tangent altitudes would result accordingly in a wrong assignment of the PSC altitudes. Therefore, we have applied a tangent altitude correction which is based on line-of-sight pointing retrieval performed with the IMK scientific MIPAS data processor [von Clarmann et al., 2003]. Within the applied sequential retrieval chain, following the retrieval of spectral shift, MIPAS line of sight is determined jointly with temperature. This is possible due to the independent information on pressure and temperature contained in the rotational-vibrational spectral signatures of CO2. For the MIPAS data version used in this paper (post-2005), the retrieval scheme described by von Clarmann et al.  had to be only slightly modified to account for the new measurement mode [von Clarmann et al., 2009]. The error of the corrected tangent altitudes is estimated to be within a few hundred meters. The validity of the tangent height corrections for the post-2005 data set has been verified by a comparison of temperature and trace gas profiles with the MLS (Microwave Limb Sounder) instrument, wherein no indications for an incorrect altitude assignment were found [Chauhan et al., 2009].
2.2.2. PSC Detection
 For the detection of PSCs within MIPAS spectra the time-efficient cloud index (CI) method has been applied. This method was used previously to detect clouds in CRISTA [Spang et al., 2001a, 2001b, 2002] and in MIPAS measurements [Spang et al., 2004]. The CI is the ratio between spectral radiances around 792 cm−1 and those at 832 cm−1. It is a measure for the presence of any cloud, independent of its composition: the smaller the CI, the optically thicker the cloud. The detection limit for PSCs for the present study has been set to CI ≤ 4.5 [Spang and Remedios, 2003]. Previous simulations have shown that this corresponds to a detection limit of horizontally homogeneous PSCs with volume densities of 0.2–0.4 μm3 cm−3 along the ∼300 km limb path of MIPAS inside the cloud. PSCs with volume densities of less than 0.2 μm3 cm−3 are not detected, while all PSCs with volume densities >0.4 μm3 cm−3 are identified [Höpfner et al., 2006a]. Of course if only a part of the field of view of MIPAS is covered by the PSC, or if the PSC does not extend horizontally throughout the tangent layer, these volume density limits increase accordingly.
2.2.3. PSC Composition
 To differentiate the PSC composition directly from MIPAS spectra we have applied the method which Spang and Remedios  used to analyze PSC observations by CRISTA. This color ratio approach discriminates measurements with the spectral band signature of NAT at 820 cm−1 by plotting the ratio of the radiances at around 820 cm−1 and 792 cm−1 versus the cloud index CI defined in section 2.2.2.
Höpfner et al. [2006a] analyzed this color ratio method more quantitatively by simulated observations [see Höpfner et al., 2006a, Figure 9]. Figure 2 shows a histogram of MIPAS PSC observations within the color ratio space during the four winters investigated here. The domains of the MIPAS composition classes are superimposed. The different regions R1–R4 represent the typical color ratio classes which have been assigned to different PSC compositions [Höpfner et al., 2006a]. R1 contains NAT particles with mean radii smaller than 3 μm. Ice PSCs are located in R3. R2 data are most likely STS, but large NAT or thin ice layers cannot be excluded. In R4 it is difficult to distinguish among ice, STS and large NAT. The calculations by Höpfner et al. [2006a] were performed only for PSCs of pure compositions. Here, we have extended them to account also for mixtures of NAT and STS. For that, MIPAS spectra of mixed-phase clouds of NAT and STS with various total volume densities, particle radii, and volume fractions partitioned between NAT and STS were simulated by radiative transfer calculations. These synthetic spectra were analyzed using the same scheme applied in case of real MIPAS observations. The new simulations showed that for NAT volume density fractions of 30–40% for particles with radii smaller than 3 μm, the cloud would be identified as NAT (R1) by the method applied. For smaller fractions it would fall into regions R2 and R4.
 For the present paper, the PSC compositions derived from MIPAS are classified as follows: R1 as NAT, R3 as ice. Data points in R2 and R4 are merged together into STS/Mix, since the classification into these two sectors can contain STS, NAT or ice PSCs.
3. Comparison Data Set
 PSC observations of both instruments are compared for two northern and two southern polar winters between June 2006, when the CALIPSO observations started, and February 2008, when the last PSCs were detected in Arctic winter 2007/2008.
Figure 3 shows an example for a typical day of Antarctic and Arctic PSC observations of both instruments on 11 July 2007 and 3 January 2008. The color code indicates the PSC top height. Note that the different PSC coverage near the poles is due to the different latitudinal coverage of MIPAS and CALIPSO. In general the detected PSC distribution and top heights of both instruments look quite similar.
 For more quantitative comparisons we selected a subset of these data points matched in space and time. As a compromise between matches as close as possible and good statistics we have chosen a maximum time difference of 6 h and a match radius of 200 km. Further we have restricted these investigations to latitudes equatorward of 82° in an attempt to avoid clouds detected by MIPAS at higher latitudes that cannot be observed by CALIPSO. Results from tests with tighter match criteria (e.g., an additional maximum temperature difference criterion of 1 K) did not differ significantly from those presented herein.
 For the example shown above, Figure 4 now includes only those PSC observations which meet the selected coincidence criteria. In total these criteria result in 12088 MIPAS limb scans matching with at least one CALIPSO profile. Of those there are 4683 cases where at least one instrument detects PSCs as described in more detail in section 4.
4. Results and Discussion
4.1. PSC Detection
 Here we compare the two instruments with regard to the detection of PSCs independent of altitude. The height range is restricted to altitudes above 14 km for MIPAS and above 12.5 km, i.e., 14 km minus one half the MIPAS field of view, for CALIPSO.
 For the coincidence criteria of 6 h and 200 km, Table 1 gives the statistics of PSC detection for both instruments. All numbers have been calculated with respect to MIPAS limb scans; one must keep in mind that for each MIPAS observation there are typically about 50 matching CALIPSO profiles, which would represent a segment of 250 km along a CALIPSO orbit in cases where only one orbit matches the temporal criterion. A positive CALIPSO PSC detection means that a PSC is found in at least one of the coincident CALIPSO profiles.
Table 1. Statistics of PSC Detection for MIPAS and CALIPSO Matches Within 6 h, 200 km and Equatorward of 82° Northern/Southern Latitudea
Region and Date
Total Number PSC
PSC Both (%)
PSC Only MIPAS (%)
PSC Only CALIPSO (%)
Dates are given as year and month. Total Number is total number of MIPAS matching limb scans with at least one coincident CALIPSO observation. Total Number PSC is number of MIPAS matching limb scans where MIPAS or matching CALIPSO or both instruments detected PSCs. PSC Both is percentage of MIPAS matching limb scans where both instruments see PSCs relative to Total Number PSC. PSC Only MIPAS is percentage of MIPAS matching limb scans where only MIPAS detects PSCs. PSC Only CALIPSO is percentage of MIPAS matching limb scans with only CALIPSO detecting PSCs. A positive CALIPSO PSC detection indicates that a PSC is found in at least one coincident CALIPSO profile.
 In the case of Antarctic observations, there is very good agreement in both years regarding the total number of PSCs detected: in nearly 90% of the matches both CALIPSO and MIPAS detect PSCs. However, the fraction of matches in which both instruments detect PSCs varies significantly with time. From June to August, both detect PSCs in more than 85% of the matches, whereas in May the fraction is only around 60%, in September 82–84%, and in October 63% in 2006 and 29% in 2007. In the Arctic, MIPAS and CALIPSO both detect PSCs in 59% of the matches in 2006/07 and 72% of the matches in 2007/2008, which are generally lower than in the Antarctic, but comparable to that found in the Antarctic in October 2006.
 In general, there are more CALIPSO-only than MIPAS-only PSC detections: about 6% more in the Antarctic and 14% more in the Arctic. However, there is a trend during the Antarctic winters, with more MIPAS-only PSCs in May, similar MIPAS-only and CALIPSO-only numbers in June, and increasingly fewer MIPAS-only PSCs than CALIPSO-only PSCs from July through October.
 We can explain the fact that more PSCs are detected by CALIPSO late in the Antarctic season and in the Arctic by the patchier nature of the clouds at these times. Since their geometrical extent is smaller than during the June–August period, these PSCs are better sampled by CALIPSO due to its high spatial resolution. The observation that there were more common PSC detections in October 2006 than in October 2007 is likely related to the longer-lasting period of cold temperatures in 2006. The argument of a patchier nature of PSCs late in the Antarctic season and in the Arctic should also be true for May, when, in contrast, more PSCs are detected by MIPAS alone. However, in May PSCs in the Antarctic are largely confined to a low-temperature region centered near the South Pole.
 Nearly all of the MIPAS-only PSCs detected in May occur very close to the latitude limit of CALIPSO, where the MIPAS field of view is pointed toward the South Pole. For MIPAS, only tangent point locations equatorward of the lidar latitude limit (82°) have been used. However, since MIPAS is a limb sounder, rays along its line of sight also pass through air masses poleward of the tangent points, while the CALIPSO lidar samples only the atmospheric column directly beneath the satellite. Hence, MIPAS is likely able to detect PSCs occurring poleward of the tangent point location and beyond the latitudinal viewing limit of CALIPSO.
 To investigate in more detail the cases where only one of the instruments detected PSCs, we reexamined those data points in the related CALIPSO δaerosol versus 1/R532 and MIPAS color ratio histograms. For the cases where CALIPSO detected a PSC but MIPAS did not, the majority of the lidar data points have very small 1/R532 values and slightly lower depolarization than synoptic ice PSCs. This typically indicates the presence of small-scale wave-induced PSCs. This points toward a sampling issue as an explanation for the nondetection of those PSCs by MIPAS. On the other hand, for PSCs detected by MIPAS but not by CALIPSO, most of those data points have CI values larger than about 3.3. These are indicative of PSCs with small volume densities along the MIPAS line of sight, which might be due to either tenuous large-scale clouds or patchy clouds that only partially fill the MIPAS limb field of view.
4.2. PSC Top Height
 For comparison of PSC top altitudes, the MIPAS PSC top height is set to the highest tangent altitude where a PSC signal is detected in the spectrum. The comparative CALIPSO PSC top height is defined as the maximum cloud top altitude in the ensemble of coincident profiles and is subtracted from the MIPAS top altitude to determine the offset (MIPAS-CALIPSO). These offsets were then grouped into altitude bins according to the lidar maximum top height and averaged.
Figure 5 shows mean offsets calculated from the ensemble within each altitude bin. Thin horizontal bars denote the 1-σ standard deviation of the distribution, while overlaid thick bars denote the estimated 1-σ standard deviation of the mean offset (calculated as the standard deviation of the distribution divided by the square root of the number of points within each bin). Thus, for all cases where the bold error bar does not include the zero line the differences in mean offset are statistically significant. The results for the entire data set are summarized in Table 2.
Table 2. Altitude-Dependent Comparison Between MIPAS and CALIPSO PSC Top Heights for the Antarctic 2006 and 2007 and Arctic 2006/2007 and 2007/2008 Winters
Mean Difference MIPAS-CALIPSO (km)
Standard Deviation (km)
Estimated Error of Mean Difference (km)
 The main results are (1) at all altitudes, with the exception perhaps of the lowest bins around 15 km, the lidar PSC top altitude is higher than the MIPAS top altitude; (2) the mean offset increases with increasing altitude, to a maximum of about 2.5 km at the 27 km altitude. There are various possible reasons for the offset in PSC top altitudes between the two instruments, which we will discuss in the following paragraphs.
 First, the altitude assignment of either the CALIPSO or MIPAS data could be wrong. This can be ruled out for CALIPSO, since altitudes can be determined very precisely by the lidar. The cloud top altitude is accurate to within one range bin (180 m) for the CALIPSO PSC data set [McGill et al., 2007]. For MIPAS, as described earlier, we have performed a tangent altitude correction by shifting the comb of tangent altitudes as a whole. Thus, a constant offset could be induced by this correction. However, since the estimated accuracy of the correction is a few hundred meters, it is unlikely that this is the only explanation for the observed offset. Moreover, it is unlikely that the altitude dependence of the offset is due to an incorrect MIPAS tangent point assignment, since the relative pointing of single altitudes is much more accurate than the common offset.
 To investigate additional explanations for the bias in cloud top height, we performed Monte Carlo simulations to estimate the altitude-dependent offset for the MIPAS tangent point grid assuming a vertical field-of-view (FOV) extent of 3 km. Figure 6 shows the case of PSCs with infinite horizontal extent. It is further assumed that MIPAS detects the cloud as soon as any part of it appears within the FOV. The only parameter which is being varied here is the PSC top height. Due to the FOV oversampling tangent point pattern, the cloud top height would be overestimated by 0.5–1 km in this case, with a slight dependence on tangent altitude. In Figure 6 (middle) the PSCs are assumed to have a horizontal extent of 50 km. Additionally to the top height, the position of the cloud along track relative to the tangent point has been varied. Due to the MIPAS geometrical sampling pattern, this results in a shift by −1.0 to −1.5 km to a slight underestimation of cloud top height, as was shown previously by Kent et al.  in a comparison of SAGE II and lidar cloud observations. This effect depends, however, strongly on the assumed horizontal size of the PSC: for an extent of 100 km the shift in altitude reduces to 0.7 km and for a 200 km cloud to only 0.3 km. The results shown in the right panel of Figure 6 include additionally the cloud top error due to cloud detection sensitivity being proportional to the covered MIPAS FOV. It is assumed that the probability for cloud detection increases linearly from zero when there is no cloud within the FOV to 1 in the case when the FOV is totally covered. Beside FOV effects, this assumption also includes cases where, for example, the cloud has a diffuse boundary and the signal which reaches the detection limit stems from farther inside the PSC. This shifts the offsets additionally by about −1.5 km, reaching values on the order of those observed in our comparison between CALIPSO and MIPAS.
 The calculations have shown that the effects of finite horizontal cloud extent and sensitivity of cloud detection to coverage of the MIPAS FOV lead to offsets comparable to those observed. However, the calculated offsets do not vary as much with altitude as the observed offsets. One reason could be an altitude-dependent sensitivity to PSCs on the part of either instrument due to different PSC microphysical properties as a function of altitude. However, since altitude-dependent offsets are observed regardless of the time within a season (not shown here) and in both the Arctic and Antarctic (see Figure 5), we do not consider this explanation to be robust. A more convincing argument could be that higher-altitude clouds are generally patchier than those at lower altitudes and, thus, an altitude-dependent horizontal cloud extent would result in an altitude-dependent cloud top offset determined by MIPAS.
4.3. PSC Composition
 Examples for typical geographical distributions of PSC composition derived from both instruments for one day over the Antarctic and over the Arctic are shown in Figure 7. Note that poleward of 82° there are no observations from CALIPSO. Apparent similarities between the two data sets are (1) the clear indication of ice PSCs in the region of the Antarctic Peninsula on 11 July 2007 and over Svalbard, the Barents Sea and Novaya Zemlya on 3 January 2008, (2) the correlation between CALIPSO-STS and MIPAS STS/Mix over West Antarctica, and (3) the agreement between MIPAS-NAT and CALIPSO-Mix2 over larger areas. However, some differences are also apparent, such as the green MIPAS-STS/Mix data points corresponding with CALIPSO-Mix1/Mix2 (Figure 7, orange/yellow) observations over some regions of East Antarctica and over the Kara Sea/Laptev Sea and northern central Siberia in the Arctic.
 Additional examples for each PSC composition are shown in Figures 8, 9, and 10 as PSC cross sections along selected CALIPSO orbit segments, with coincident MIPAS observations overlaid. A large part of the PSC data points in Figure 8 are identified by CALIPSO as ice, which is indicated by the blue color of the lidar observations. On top of the ice cloud near 25 km, a thinner STS cloud is indicated by CALIPSO. Both the ice cloud and the STS layer on top are also well captured by MIPAS. The largest part of the PSC shown in Figure 9 is identified by CALIPSO as Mix2, with some STS layers on the left side of the scene at lower altitudes and increasing patches of ice at all altitudes farther to the right. Overlaid MIPAS limb scans identify the higher-altitude Mix2 portions as NAT and the lower-altitude lidar-STS layer as STS/Mix. Figure 10 presents one of the few examples of a PSC identified by CALIPSO as mostly STS, which occurred near the beginning of the 2007 Antarctic winter. Here MIPAS classifies the cloud as STS/Mix and thus agrees well within the altitude range of the PSC as observed by CALIPSO. However, MIPAS also indicates the presence of PSC at much lower altitudes where there are no PSC particles detected by the lidar. This is a typical feature of the fast cloud identification approach applied to MIPAS data; due to the limb-sounding geometry, higher-altitude cloud layers affect the detection of PSCs at lower tangent altitudes. For this reason, the comparisons of PSC compositions presented herein are based on MIPAS data from only the highest tangent altitudes affected by PSCs. However, the previous two examples (Figures 8 and 9) clearly show that inhomogeneous PSCs with large vertical extent are often described correctly by the MIPAS data even at lower altitudes.
 In the following paragraphs, we investigate the agreement of the PSC compositions between CALIPSO and MIPAS statistically using all coincidences where both instruments detected PSCs. For each MIPAS profile, the PSC composition was set to the composition determined at the second highest tangent altitude where PSCs were detected in order to (1) avoid the possibility of only partial FOV coverage and (2) exclude lower tangent altitudes which might be affected by higher-altitude PSCs as discussed earlier. The composition of PSCs as determined by CALIPSO is shown in Table 3 relative to each MIPAS PSC composition. The numbers shown are the percentage of all coincident lidar points lying at altitudes within the MIPAS FOV that fall in each of the CALIPSO composition classes. All coincident points for the Antarctic are plotted in the lidar δaerosol versus 1/R532 optical space in Figure 11, partitioned into frames according to the MIPAS PSC composition. Conversely, Table 4 shows the percentage of MIPAS compositions determined for each CALIPSO PSC composition class, where the CALIPSO composition for each coincident ensemble was specified as the composition of the majority of points in the ensemble. Figure 12 accordingly shows coincident points for the Antarctic plotted in the color ratio optical space of MIPAS, partitioned according to the CALIPSO PSC composition.
Table 3. CALIPSO PSC Composition Relative to Each MIPAS PSC Compositiona
Numbers are the percentage of all coincident lidar points lying at altitudes within the MIPAS FOV that fall in each of the CALIPSO composition classes. For guidance, the expected probability of combinations is indicated by boldface, italics, and normal face in decreasing order.
Antarctic 2006, 2007
Arctic 2006/2007, 2007/2008
Table 4. Percentages of MIPAS Compositions Determined for Each CALIPSO PSC Composition Class, Where the CALIPSO Composition for Each Coincident Ensemble was Specified as the Composition of the Majority of Points in the Ensemblea
MIPAS STS/Mix, %
MIPAS NAT, %
MIPAS Ice, %
For guidance, the expected probability of combinations is indicated by boldface, italics, and normal face in decreasing order.
Antarctic 2006, 2007
Arctic 2006/2007, 2007/2008
4.3.1. MIPAS-NAT and CALIPSO-Mix1/Mix2
 For cases where the spectral signature of NAT has been detected in the MIPAS measurements (rows indicated as NAT in Table 3) about 90% of all coincident CALIPSO data points are Mix1 or Mix2. This result is very promising since it confirms the consistency between the spectroscopic identification of the presence of NAT particles by MIPAS and the detection of particles with significant aerosol depolarization and small/intermediate backscatter ratio by CALIPSO. In other words, when NAT has been clearly identified in infrared spectra, nonspherical particles are clearly present in the majority of the lidar observations.
 In Figure 11, the “MIPAS-CLASS: NAT” plot shows where coincident Antarctic data identified as MIPAS-NAT fall in the CALIPSO δaerosol versus 1/R532 space. As already pointed out, most points fall into the regions of Mix1 or Mix2. The relatively few points (7–8%) that fall into the CALIPSO STS domain are explained by the fact that lidar profiles with only Mix1/Mix2 composition at all altitudes within the MIPAS FOV occur rarely. Thus, there are layers/points within the cloud identified as STS by CALIPSO, but the majority of points at other altitudes are identified as Mix1/Mix2, leading to an identification of the entire cloud as NAT by MIPAS.
 On the other hand, PSCs identified as Mix1 and Mix2 by CALIPSO (see rows in Table 4) are not clearly identified as MIPAS-NAT but also as MIPAS-STS/Mix. This is expected, since as described earlier, mixtures of NAT and STS with either a majority of particle volume in STS or large NAT particles are classified as STS/Mix by MIPAS. This would also be the case for particles which are nonspherical, but do not have the composition of NAT but, for example, of NAD (Nitric Acid Dihydrate). However, this PSC composition has not been observed in atmospheric measurements to date [Höpfner et al., 2006a].
Figure 12 shows coincident Antarctic data identified as CALIPSO-Mix1 (Figure 12, top right) and Mix2 (Figure 12, bottom left) plotted in the color ratio space of MIPAS. Obviously for Mix1 the data points are located at higher CI values indicating that their volume densities are smaller compared to Mix2, which in general have smaller CI values and, thus, larger volume densities. This is consistent with the CALIPSO definition of the Mix1 and Mix2 classes in terms of NAT particle volume.
4.3.2. MIPAS-STS/Mix and CALIPSO-STS
 For the case of STS detection we first inspect the CALIPSO-STS observations of Table 4. Here for all seasons the large majority of coincident MIPAS data are also identified as STS/Mix. As in the previous case where NAT was unambiguously identified by MIPAS due to its spectral signature, here CALIPSO classifies the PSCs as STS due to the low depolarization. On the other hand, clouds identified as STS/Mix by MIPAS (see Table 3) are classified as STS, Mix1, Mix2, and even ice by CALIPSO. In many of these cases, the clouds are likely STS/NAT mixtures, but with either a minority of particle volume in NAT or large NAT particles, which results in STS/Mix rather than NAT classification by MIPAS. These results are additionally illustrated in Figures 11 and 12.
 Interestingly, at the beginning of the Antarctic PSC season in May and early June, there is better agreement between MIPAS-STS/Mix and CALIPSO-STS (61%). During this period STS PSCs are relatively more abundant and spatially homogeneous than later in the winter as shown in the example of Figure 10.
4.3.3. MIPAS Ice and CALIPSO Ice
 For coincident measurements where MIPAS detects ice PSCs, ice is also identified as the composition by CALIPSO in about 60% of the points during the two Antarctic winters (Table 3 and Figure 11). Most of the remaining coincident MIPAS ice points are assigned to the Mix2 class by CALIPSO. However, in about 90% of the cases where ice is not the majority composition in the coincident CALIPSO data, it has been detected in some of the lidar points. Since layers of ice are in general optically thicker than those of STS or NAT, localized patches of ice would contribute more to the observed MIPAS radiances but would be weighted only by their number in terms of the CALIPSO statistics. Coincidences including such layers would, thus, be identified as ice by MIPAS. Another reason might be a misclassification of ice layers as Mix2 by CALIPSO due to a conservative setting of the separation line between Mix2 and ice in the δaerosol versus 1/R532 space. In the Arctic, there are too few coincidences classified as ice to derive meaningful statistics.
 From the point of view of CALIPSO ice observations (Table 4), considerably more of these are identified by MIPAS as STS/Mix than as ice, but none as NAT. In Figure 12 this case is illustrated in the panel denoted as “CALIPSO ice,” where it is clear that all data points are located close to the separation line between MIPAS ice and STS/Mix. Thus it is probable that these cases are really ice PSCs, but due to the rather conservative choice of the MIPAS separation between the two classes, they are classified as STS/Mix.
5. Summary and Conclusions
 We have compared PSC detection, top height and composition observations from two spaceborne remote sensors, CALIPSO and MIPAS, with entirely different measurement principles: active lidar in the visible versus passive spectroscopy in the midinfrared spectral region.
 Regarding the detection of PSCs we have found that there is a very good consistency between the two instruments during the midwinter Antarctic season. In the Arctic and late in Antarctic winters, CALIPSO generally detects PSCs in a greater fraction of coincidences than MIPAS. This can be explained by the patchier nature of PSCs during these periods, which makes the clouds more likely to be detected by the very high spatial resolution lidar. During the early Antarctic winter season in May, MIPAS detects PSCs more frequently than CALIPSO, probably due to the measurement geometry of MIPAS favoring detection of PSCs which are confined early in the season to latitudes nearer the South Pole.
 Cloud top altitudes observed by MIPAS are generally lower than those observed by CALIPSO, with the absolute bias decreasing at lower altitudes. We have shown that negative offsets similar to the observed ones in cloud top height determination might be caused by a limited horizontal extent of the clouds and an increasing sensitivity of the MIPAS PSC detection when sounding deeper inside the cloud. The altitude dependence of the cloud top offsets, however, can only partially be explained and thus remains an open issue.
 The characterization of PSC composition by MIPAS is based on a spectroscopic signature in the case of NAT and large optical depth in the case of ice PSCs; while the lidar distinguishes liquid (STS) and solid or mixed-phase (Mix1/Mix2, ice) clouds by their aerosol depolarization and backscatter intensity. Intercomparisons are complicated by the fact that, in general, CALIPSO observes multiple PSC compositions within the same spatial domain from which MIPAS retrieves a single composition. In spite of this, we found composition categories where there was a high degree of consistency between the two instruments.
 1. The spectral identification of NAT by MIPAS is robust as most coincident lidar measurements were identified as Mix1/Mix2, i.e., NAT/STS mixtures. Furthermore, the distinction between CALIPSO Mix1 and Mix2 is well correlated with MIPAS cloud indices in terms of particle volume densities.
 2. For PSCs classified as STS by the lidar due to low aerosol depolarization, a large majority of the MIPAS coincidences are identified as STS/Mix.
 There are, however, some inconsistencies between the two instruments with respect to STS/NAT classification. For example, STS/NAT mixtures with low NAT volume densities or NAT particles greater than 2–3 μm in radius are classified as STS/Mix by MIPAS, but would be classified as Mix1/Mix2 by CALIPSO.
 3. The MIPAS detection of ice generally corresponds well to ice detection by the lidar. However, both the CALIPSO ice–Mix2 and MIPAS ice–STS/Mix boundaries are likely conservative and may result in misclassification of ice PSCs. In addition, even small patches of ice within the MIPAS FOV will likely result in ice classification by MIPAS, while ice would be the minority composition class in the coincident CALIPSO data ensemble.
 Future issues regarding the analysis of the CALIPSO PSC observations are, thus, refinements in the definition of the boundary line between ice and Mix2 cases, possibly by incorporating measurements of gas-phase H2O and HNO3 by other satellites. In the case of MIPAS, a further separation of ice from the other compositions might be achieved by utilizing the spectroscopic behavior of ice in the midinfrared.
 Further work which is foreseen is the identification of PSCs from lidar daytime observations which could be compared, based on the present work, with MIPAS data, which are in general not strongly affected by the presence of sunlight. For MIPAS, the retrieval of altitude profiles of microphysical PSC parameters like particle size and volume density is a future option, the results of which could also be compared to CALIPSO measurements following the approach of Höpfner et al. [2006a].
 Overall, this work has demonstrated a very promising consistency between PSC parameters derived from instruments probing different physical particle properties. This further leads to the conclusion that the microphysical properties of PSCs are reasonably well understood.
 We thank the European Space Agency for providing MIPAS spectra. CALIPSO PSC research is supported by Hal Maring, NASA Radiation Sciences program manager, and the NASA Headquarters Earth Science Division. Support for L. Poole is provided under NASA contract NNL07AA00C.