By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
 Polar Mesospheric Clouds (PMCs) were observed by a limb-scanning ultraviolet spectrometer on the Student Nitric Oxide Explorer (SNOE). Radiance profiles at 215 and 237 nm are analyzed to determine the presence of clouds. Once detected, the altitude and brightness of a cloud relative to the background atmosphere is determined. SNOE observations provide the frequency of occurrence of PMC as a function of location and time for the years 1998 through 2003. The observations show at high latitudes a general rise in frequency of occurrence beginning approximately 3 weeks before summer solstice in both hemispheres and lasting for approximately 1 week. These rises are followed by approximately 60 days of relatively high but variable occurrence frequencies. The declines in frequency of occurrence at the ends of the seasons are generally slower and more structured then the beginning of the seasons. One of the major results from the SNOE observations is that significantly more PMCs are observed in the Northern Hemisphere than in the south, leading us to conclude that the southern polar mesosphere must be on average less saturated than the northern polar mesosphere. The SNOE observations also suggest that the frequency of occurrence of PMCs is strongly modulated by local dynamical influences. The SNOE results are in general agreement with results from the Solar Mesosphere Explorer which observed PMC with similar instrumentation in the years 1981 through 1986.
 Polar Mesospheric Clouds (PMCs) are clouds known to occur at altitudes of 81–85 km, just below the mesopause, at times near the summer solstice. A subset of PMCs are sometimes visible to appropriately located ground observers after sunset and before sunrise and are referred to as Noctilucent Clouds (NLCs). The first published sighting of an NLC occurred in the late-nineteenth century [Leslie, 1885]. The data from organized observing networks established over the last 50 years have shown that NLCs are generally confined to the summer months, and occur at latitudes poleward of 55°. They can occur over a roughly 3-month “season,” although 80% occur over the shorter period of 54–62 days centered about the 2 weeks following summer solstice [Gadsden, 1990]. Cloud particles have radii (spheres of equivalent volume) on the order of 40 to 70 nm [Thomas and McKay, 1985; Rusch et al., 1991; von Cossart et al., 1999], and are primarily made of water ice [Hervig et al., 2001]. The seasonal occurrence of supersaturated conditions near the mesopause tends to confirm the basic theory of PMC formation [Lübken, 1999], described in detail by Turco et al.  and more recently by Rapp et al. .
 The basic seasonal behavior of PMCs was deduced from analysis of ultraviolet observations made by the Solar Mesosphere Explorer (SME) [Barth et al., 1983] ultraviolet spectrometer (UVS) from 1981–1986 [Thomas and Olivero, 1989]. The SME statistics showed that a typical PMC season lasts approximately 90 days, usually beginning about 20 days before summer solstice. PMCs first appear at the highest latitudes and gradually appear lower in latitude to about 55°. SME also showed that there are differences between the north and south equatorward boundaries; PMCs occur about 5° lower in latitude in the north. Since SME, there have been numerous space observations reported, all based on serendipitous measurements made by experiments intended to study other atmospheric phenomena [e.g., Thomas et al., 1991; Evans et al., 1995; Debrestian et al., 1997; Carbary et al., 1999, 2003; Stevens et al., 2001; Shettle et al., 2002; DeLand et al., 2003].
 The Student Nitric Oxide Explorer (SNOE) [Barth et al., 2003], like SME, observed the Earth's UV limb radiance and thus also observes PMCs. The SNOE and SME instrumentation and observation geometries are nearly identical, making the observations comparable. SNOE was launched in February 1998 and operated until December 2003; thus the combination of SNOE and SME measurements cover a span of nearly 2 decades. Like SME, SNOE was in a Sun-synchronous polar orbit, but unlike SME, it was operated at 100% duty cycle, providing the opportunity to view PMCs over the entire Earth at 15 longitudes each day. In addition, the local times of the ascending nodes of SME and SNOE were different.
Hines et al.  first proposed the importance of dynamical forcing of the atmospheric variables influencing PMCs by suggesting that atmospheric gravity waves are responsible for the wave patterns in NLCs. Haurwitz and Fogle  added further support to this idea by using information from ground-based photographs to quantify the wave structure and period. Jensen and Thomas  and later Rapp et al.  further established that gravity waves are indeed a fundamental forcing not only of the heights of NLCs, which influence visible wave structure, but also of the cloud microphysics itself. On larger spatial and temporal scales, Merkel et al.  showed from SNOE observations that PMC formation is highly variable in time and space, and that PMC brightness shows evidence of a response to 5-day planetary wave activity. This work confirmed earlier ground-based studies by Sugiyama et al.  as well as Kirkwood et al.  and Kirkwood and Stebel , who found planetary wave effects in NLCs and Polar Mesospheric Summertime Echoes (PMSE). PMSE are a radar phenomenon related to PMC in that they both require the presence of ice particles. On still longer timescales, Thomas et al. , DeLand et al. , and Thomas  have shown that PMC interdecadal activity and brightness is anti-correlated with solar Lyman-α irradiance, which is the primary source of water photolysis at cloud heights.
 The forcing of PMCs is probably a combination of processes that produce variations of water vapor and temperature on various spatial scales and timescales. These processes may include dynamical process such as gravity waves, tides, and planetary waves that affect water vapor and temperature as well as longer timescale phenomena such as seasonal upwelling of water vapor, variable UV photolysis of water vapor, long-term changes in methane-derived water, volcanic influences, etc. There is also evidence of water vapor enhancements from shuttle exhaust leading to the production of PMCs [Stevens et al., 2003]. We will focus on day-to-day variations of PMCs, primarily using zonally averaged daily cloud parameters. Observed variability in PMCs most likely reflects the intermediate-scale processes of planetary waves, and of seasonally varying upwelling. We will show that such variations are substantial, and vary in important ways with latitude, between hemispheres, and interannually. It should be noted that the term “PMC climatology” has been used in past references. We suggest that this term be reserved only for annually averaged quantities. We recommend the term “PMC morphology” (spatial or temporal) to refer to behavior within a season, or with latitude, and which may or may not refer to zonal averages.
 In this paper we describe the SNOE observing technique and the development of the SNOE PMC database. We describe the algorithms for determining PMC scattering properties from the background limb radiance. The SNOE results are compared to those from SME as a partial validation. We will present the zonally averaged morphology of PMC activity, brightness, and altitude and discuss them in the context of the standard theory for PMC formation and evolution. Finally, we will provide speculation to explain the observed behavior.
2. Instrumentation and Observations
 SNOE was launched on 26 February 1998 into a 1030 LT/2230 LT sun-synchronous near-circular orbit with an average altitude of 556 km. SNOE continued to function until it reentered the atmosphere on 13 December 2003. The spacecraft spun at 5 rpm with the spin axis normal to the orbit plane. The goals of SNOE involved the study of nitric oxide in the lower thermosphere and the energy sources that drive NO variability, in particular solar soft x-ray irradiance and auroral energetic particles. Details of the observations and reports of some results have been described by Merkel et al.  and Barth et al. .
 SNOE employed a UVS designed to measure the NO (1,0) and (0,1) γ band emissions at 215 nm (Channel 1) and 237 nm (Channel 2). The UVS was a fixed grating spectrometer with dual exit slits. The spectrometer was fed by an off-axis spherical mirror with a 0.25-m focal length. The telescope projected the image of the spectrometer slit onto the atmosphere. The slit image was 3.3 km in altitude and 33 km in the horizontal. The exit slits were positioned so that 215- and 237-nm radiances emerged onto two phototube detectors. The slits were sized to provide a spectral bandpass, Δλ, of 3.2 nm onto each detector. The detectors had cesium-telluride photocathodes. The sensitivity of the instrument was approximately 1 count per kR/Δλ in both channels. Details of the instrument and its calibration were described by Merkel et al.  and Barth et al. . In-flight monitoring of the instrument calibration has been described by Barth et al. . For altitudes below tangent heights of 175 km, instrument dark counts are negligible and the dominant source of noise in any data point is due to photon counting statistics.
 The UVS pointed radially outward from the spacecraft and scanned the limb each spin. A fixed integration period of 2.4 ms was employed. This value corresponds to an altitude coverage of approximately 3.3 km during one integration period for the nominal spin period of 12 s. A total of 100 samples were collected per spin. All observations described in this paper were made with the UVS pointed within the plane of the orbit. Data were stored from a fraction of each spin such that radiance from approximately 200 km down to the Earth's surface was captured. Figure 1 shows sample altitude profiles from both channels. In Figure 1 (left), a UVS profile that includes a PMC at the threshold for detection is shown. In Figure 1 (right), a profile that includes a bright PMC is shown. For times when only dim or no PMCs are present, the UVS signal below 90 km is dominated by Rayleigh scattered sunlight as is the case in Figure 1 (left). Local maxima occur in the Rayleigh scattered sunlight profiles at approximately 60 km. These peaks are created by absorption from ozone and molecular oxygen at lower altitudes and are critically important for the determination of SNOE pointing. During the summer in each hemisphere, scattered sunlight from PMCs often appears near 83 km. The PMCs may appear as small increases in the Rayleigh scattered sunlight profile or they may dominate as is shown in Figure 1. Above 90 km, NO radiances emerge from the Rayleigh scattered sunlight background or from the PMC signal. The PMCs are brighter at 237 nm than at 215 nm primarily because of the higher solar irradiance. The Rayleigh scattered radiance at altitudes below the PMCs is less bright at 237 nm because of the increased absorption by ozone.
 Because SNOE was a spinning satellite, proper registration of the UVS altitude profiles is crucial. Horizon sensors onboard the spacecraft provide only coarse altitude registration. More accurate registration is obtained through analysis of the Rayleigh scattered sunlight profile near 60 km (see Figure 1). The altitude registration procedure is described in detail by Merkel et al. . Each UVS profile is compared to a model calculation of the Rayleigh scattered sunlight profile. The peaks near 60 km are well understood and serve as altitude fiducials. The raw profiles are shifted in altitude until the peak is at 60 km. The profiles are then linearly interpolated onto a high-resolution altitude grid and shifted in small increments until a best fit between the measured profile and the model profile is obtained. The 1-standard-deviation uncertainty in the resulting registration of an individual UVS profile is 1.5 km. This uncertainty is primarily driven by uncertainty in the assumed ozone climatology [Keating et al., 1996].
 In addition to determining the altitude registration, a correction to the radiances must be made before PMCs can be clearly identified. While nearly all of the NO emission observed by SNOE is produced by molecules above 95 km, observations at lower tangent altitudes include contributions along the line of sight from higher altitudes. These contributions must be removed from the data so that the resulting profiles contain only scattered sunlight from the atmosphere (Rayleigh scattering) and from PMCs (Mie scattering). The NO densities above 95 km are determined in an independent analysis which is described by Barth et al. . Knowing the NO densities above 95 km, it is straightforward to calculate the airglow signal which would be observed at tangent altitudes less than 95 km [see Merkel et al., 2001; Barth et al., 2003]. It was found that below 95 km, the shape of the profile is nearly independent of the column abundance of NO and that an average profile normalized at 103 km adequately reproduces the modeled NO contribution. The observed radiance at 103 km is entirely due to scattered sunlight by NO. Figure 2 illustrates the altitude variation of the average NO contribution to the 215-nm channel relative to the observed radiance at 103 km. The figure plots the average ratio of the radiance from NO observed at a given altitude to that radiance observed at 103 km. The error bars are 1 standard deviation and are determined by comparing the calculated NO contributions for 1 year of profiles to those obtained by scaling an average profile. The uncertainty is primarily the result of the natural variability of the altitude profile of lower thermospheric NO. The magnitude of the emission rate of NO can vary by as much as a factor of 5 [see Barth et al., 2003], but the change in the shape of the profile, as shown by Figure 2, is much smaller. Thus for each observed profile, the ratio shown in Figure 2 is scaled by the observed radiance at 103 km and then subtracted from the profile. The use of a single correction to the data below 103 km results in an 11% uncertainty in the radiance profile at 83 km due to the removal of contributions by NO. The result of this process is an observed altitude profile of Rayleigh plus PMC scattered sunlight. This profile can then be analyzed to determine PMC presence and PMC properties.
3. Cloud Detection
 The UVS profiles, once registered in altitude and corrected for NO, provide a database which can be searched for the presence of PMCs. Each observed radiance profile is processed individually. From a given profile the computer algorithm is able to detect PMC presence, altitude, and apparent brightness. The absolute brightness of any given PMC cannot be obtained because there is no information regarding the extent to which the cloud fills the 3.3 by 33 km image of the slit on the horizon. However, the statistical distributions of apparent cloud brightness provide interesting information. The most useful unit is the ratio of the observed cloud brightness to the background Rayleigh scattering profile, or Limb Scattering Ratio (LSR). In this unit, the determined brightness is not dependent upon calibration and can be compared to other observations, both past and present. The LSR can also potentially be compared to LIDAR observations. Note that a retrieved PMC profile that contains both scattering by PMCs and Rayleigh scattering, when divided by the background profile yields LSR + 1. Because the upper mesosphere and even the brightest PMC are optically thin [Thomas and McKay, 1985], the total signal is just the sum the two contributions.
 In order to separate PMCs from the background, we adopt an algorithm based on that used for SME data and described by Olivero and Thomas  and Thomas and Olivero . The assumption is made that the background Rayleigh scattering is much less variable than the PMC radiances; therefore this background can be adequately represented by an average background radiance profile over the entire season. This profile is obtained by averaging all of the NO-corrected profiles between 50.0° and 55.0° latitude (north or south) for the time period encompassing each PMC season. The period over which the profiles are averaged is 40 days prior to solstice through 80 days after solstice. For convenience, the reference profile is applied at all latitudes and times within the season. Any deviation between the measured profile and the reference profile by more than a specified amount near a typical cloud altitude is considered to be due to the presence of a PMC. In this work, any deviation greater than 4 standard deviations (4σ) above the average is flagged as a potential PMC. This criterium results in a minimum detection threshold of LSR equal to about 0.7. Further criteria that must be met before a deviation is identified as a PMC are: (1) that the deviation must exceed 4σ in both UVS channels and (2) that the altitudes in the two channels must be within 1.5 km of each other (half the UVS altitude resolution). All UVS profiles for which the solar zenith angle is less than 90° are considered.
 The use of a single reference background profile applied to all locations and times represents a source of uncertainty for cloud detection. Variations in atmospheric density at 83 km will modify the brightness of the Rayleigh scattered radiance profile. In the cases where the background is reduced, dim PMCs may go undetected by the SNOE algorithm. The MSIS empirical model [Hedin, 1991] indicates that the atmospheric density at 83 km may vary by as much as ±10% over the polar regions during the PMC season; however, this value is based on limited data. The systematic error in cloud detection incurred by the constant background assumption may be at least partially removed in the future when more information is available regarding the atmospheric density variability during polar summer.
Figure 3 illustrates the stages in analyzing 215 nm UVS profiles for PMC presence. Figure 3 is separated into two columns, demonstrating the application of the algorithm for clouds that are both near and well above the detection threshold. The first step is removal of the contribution from NO. This is shown in Figure 3 (top) for the dim and bright clouds. The dotted line in both of these plots is the contribution from NO emission to the two observed profiles arrived at from the algorithm described earlier. Note that even in the case of the weak detection (Figure 3, top left), the contribution from NO is small at 85 km and below.
 In Figure 3 (middle) the corrected radiance profiles are shown with symbols and the reference 215 nm radiance profile is shown in thick lines. The dashed line in both plots is the reference profile plus 4σ. If a point in the observed profile between tangent altitudes of 75 and 95 km lies above the detection threshold (dashed line), that profile is flagged as potentially containing a PMC. Once flagged, the ratio of the observed radiance profile to the background profile is fit to a parabola using the maximum and adjacent data points. This process is shown in Figure 3 (bottom). The LSR is the maximum of the parabolic fit minus 1 and the altitude of the PMC is the altitude of the maximum (identified by the square in Figure 3 (bottom)); both are obtained from the fit parameters. The parabolic fit is employed because the 3.33-km altitude resolution of the observations does not resolve the cloud.
 The above process was performed for both UVS channels. The profile is registered as a PMC if the procedure finds a peak in LSR above their respective thresholds in both channels and the altitudes are within 1.5 km. The threshold values of LSR are about 0.7 for both channels. Typically the two channels yield PMC altitudes that are within 0.33 km. All of the UVS profiles undergo this procedure. The result is a database of PMC detections, their location, time, LSR, and altitude.
 The uncertainties in inferred cloud altitude and brightness have been determined through simulation of an ideal cloud profile as observed by the SNOE UVS. Simulations were performed for a range of cloud thicknesses and are described in detail by Merkel . Table 1 shows results for a cloud of thickness 1 km and brightness at the threshold for detection. The errors in cloud altitude are due to the altitude registration of the raw profile and the random sampling of the cloud in 3.33-km steps with the UVS field of view. The sampling provides a systematic lowering of the cloud altitude (apparent height is lower than true height) by 1.1 km. A random uncertainty of 0.4 km is also due to the sampling. The largest uncertainty in cloud brightness arises from the smearing of the true cloud radiance profile by the UVS field of view. For cloud thicknesses of 0.5 km to 2.0 km, the apparent LSR may be reduced from the true LSR by factors of 3.1 to 1.7. A further systematic error of 5.5% arises from the sampling and the parabolic fitting. Random errors in the inferred brightness arise from the NO radiance subtraction, the sampling and parabolic fitting, and counting statistics.
Table 1. Uncertainties in Inferred PMC Parameters
Reported values are appropriate for a cloud with brightness near the threshold for detection and become negligible for brighter clouds.
This is a factor relating SNOE observed LSR to an instrument with infinitesimal field of view. This factor varies with cloud width.
 In the following sections, the systematic errors in the inferred cloud altitude are corrected for in the observations; however, no correction for systematic errors is applied to inferred cloud brightness. These systematic errors are important only when comparing different observations of cloud brightness and do not affect the results described in this paper. The LSR and PMC altitudes quoted in the following sections are from the 215-nm channel.
 The most useful application of the database derived above is the determination of the seasonal, spatial, and temporal morphology of PMCs. Four statistical quantities have been used for this purpose. Olivero and Thomas  introduced the frequency of occurrence, average limb radiance, and maximum brightness for a given period of time. Thomas  defined the statistical distribution of brightness, or g-distribution, which was applied by DeLand et al. . For the present study we focus on the frequency of occurrence which has been widely used as a quantitative indicator of PMC activity. We define the daily occurrence rate (or simply frequency of occurrence) as the ratio of the number of PMC detections to the number of possible detections in a given latitude interval on a given day. At a given latitude, the average PMC brightness varies little with season because most of the observed PMCs are near the detection threshold. We will show, however, that the daily maximum PMC brightness is illustrative of the changing environment in which PMCs form. Daily zonal values of frequency of occurrence, altitude, and maximum brightness have been determined for 5°-wide latitude bins (centered on 60°, 65°, etc.). In a given latitude bin, at least 100 observations were made each day.
 We will now describe the behavior of PMCs over the course of a single season. We will first examine the behavior of several seasons averaged together, so that we may describe the general properties of the PMC season and compare to earlier results.
4.1. Seasonal Results
 Because the SNOE observation technique is so similar to that employed on SME, comparing the general features of PMC spatial and temporal morphology obtained from SNOE observations to those obtained from SME provides a crude validation of the SNOE results. The correspondence of the two data sets cannot be exact because of many factors, described below. Figure 4 shows the average seasonal PMC frequency of occurrence from SNOE as well as from SME. These are shown at 80°, 70°, and 60° latitude for the period from 50 days prior to solstice to 100 days afterward. Figure 4 shows the PMC frequency of occurrence for the (left) Northern and (right) Southern Hemispheres. The SNOE results (solid lines) are non-weighted zonal averages of daily values calculated during the six northern seasons (1998–2003) and five southern seasons (1998/1999, 1999/2000, 2000/2001, 2001/2002, 2002/2003) observed by SNOE. The SME values (dashed lines) are polynomial fits of averaged seasonal results developed by Shettle et al.  for use in analyzing irregularly sampled observations. Error bars on the SNOE results are the standard deviations of the binned data, including both day-to-day and interannual variability. To facilitate comparison of the seasonal behavior of PMCs in both hemispheres, time is shown on the abscissa of both plots in terms of days relative to summer solstice. The PMCs appear first at high latitudes about 3 weeks before summer solstice. Over a period of about 10 days, activity rises to the point where PMCs are detected in nearly 90% of the observations made at 80° latitude. This level of activity is maintained for about 2 months before the occurrence frequency decreases to about 50%. By about 70 days after summer solstice, all PMC activity is concluded.
 The agreement between SNOE and SME is reasonable. The SME occurrence frequencies are higher than the SNOE frequencies at the beginning and end of the seasons; however, this is mostly likely due to the inability of the polynomial fits to represent the periods of rapid change. In the north, the SME values are slightly higher than SNOE at the peak of the season. In the south, the differences between the two data sets are larger. The SNOE and SME instruments and observation geometry are nearly identical; however, there are several sources of possible differences. The observational factors include differing (1) eras (1981–1986 versus 1998–2003), (2) longitudinal coverage, (3) wavelengths of observation, and (4) local times of observation. Considering each of these sources in more detail, we find the following. (1) A long-term trend in PMC brightness [Thomas, 2003] of 3–7%/decade over this ∼15-year time period could result in an overall 5–10% increase in observed SNOE brightness. Solar cycle changes are more important, as the two data sets do not coincide with the same phase of the solar cycle. However, such systematic differences in overall brightness or occurrence frequency are not apparent in this initial comparison. (2) SNOE operated in a mode that permitted PMC detection throughout all of its 15 orbits per day while the SME UVS operated in its primary mode for 3 to 4 orbits per day. (3) SME observed PMCs at 265 nm and 296 nm while SNOE observed PMCs at 215 nm and 237 nm. This results in a difference in observed brightness because of the dependence of particle scattering on wavelength and the spectral variation of solar irradiance. The differing solar irradiance is cancelled out in the LSR; however, at the shorter wavelengths, using Mie-scattering theory and standard particle-size distributions [e.g., von Cossart et al., 1999] we estimate that the LSR should be ∼15% greater than at the SME wavelength (265 nm) because of the higher particle cross section at 215 nm. (4) SNOE observed PMCs at local times closer to local noon and midnight than did SME. A detailed comparison between SNOE and SME is best done by comparing the distribution of relative PMC brightness, and allowing for the various differences of observation described above. This effort is presently underway and will be described in a future report. For the present work, the reasonable agreement between SNOE and SME serves as a partial validation of the SNOE results and permits the use of SNOE observations to study PMC properties.
 In addition to the frequency of occurrence of the PMCs, we may examine the brightnesses of the PMCs observed and their dependence on latitude. Figure 5 displays the LSRs for the 215-nm channel of individual PMC observations for all seasons. LSR is shown as a function of both latitude and scattering angle. Only each fifth observation is plotted for clarity of presentation. The number and brightness of the PMCs increase significantly with latitude toward the poles. Weak PMCs are detected at latitudes as low as 45° north and south. PMCs are detected as high as 82.5° latitude in both hemispheres. The high-latitude cutoff is not geophysical but rather a result of SNOE's orbit which does not permit observations above this latitude.
 A striking result of Figure 5 is that the brightest PMCs observed in the Southern Hemisphere are nearly an order of magnitude brighter than those observed in the Northern Hemisphere. The maximum LSR in the north is on the order of 10 while in the south, LSR values of 200 are observed. While this may be partially due to real differences in the altitudes or optical depth of the PMCs in the two hemispheres, the bulk of the difference is almost certainly due to observational geometry. As shown in Figure 5 (bottom), Northern Hemisphere PMCs are observed at large scattering angles while southern PMCs are observed at small scattering angles. Mie theory of scattering from sub-micron ice particles predicts that scattering is much more efficient at smaller scattering angles [Thomas and McKay, 1985]. This effect was also seen by SME and first interpreted by Thomas . Some structure is apparent in Figure 5 in the distribution of the number of clouds with high scattering angle. This structure is not necessarily geophysical, but is due to the changing local time of the spacecraft's orbit with successive seasons which modifies the range of scattering angles.
4.2. Interannual Variability of PMC
 Characteristics for the individual PMC seasons observed by SNOE are listed in Table 2. The beginning day number relative to solstice, end date, and length of season at 80°, 70°, and 60° latitude are listed for each observed PMC season. The maximum frequency of occurrence, maximum observed brightness, and the Mean Local Time (MLT) of the observations for each season are also listed. Note that only the ascending node observations (usually the AM portion of the orbit, see Table 2) are discussed in this paper. The beginning of the season is defined as the day (relative to solstice) on which 10% of the total number of PMCs observed for that season and latitude has been detected. The end of the season is defined as the day on which 90% of the total number of clouds has been detected.
Table 2. Characteristics of PMC Seasons Observed by SNOE
 The start and end of a PMC season can be described by observations at 80° latitude where the observed seasons begin the earliest and end the latest. Table 2 shows that the start of the season is fairly constant in the north ranging from 2 to 11 days prior to solstice. In the south, the date of onset is more variable with the start of the season ranging from 16 days prior to solstice to 2 days after solstice. In both hemispheres, the end of the season occurs 40 to 50 days after solstice. The time between the beginning and end of each season ranges from 40 to 50 days. SME showed that the southern PMC season began about 9 days earlier than the northern, relative to solstice. SBUV data for PMCs (M. DeLand, personal communication, 2004) also show an earlier start (on the average) of the southern PMC season by about 1 week. Care must be taken, however, when comparing the observed beginnings and ends of seasons between different instruments. These values are dependent upon the sensitivity of the instrument and the brightness threshold for cloud detection. Figure 4 shows that the seasonal behavior observed by SNOE is very similar to that observed by SME, especially in the south.
Figures 6 and 7 illustrate the zonally averaged properties of the observed northern and southern PMC seasons, respectively. For each season the occurrence frequency at 80°, 70°, and 60° latitude is shown as a function of time in days relative to solstice. Both the daily (dotted line) and the 7-day averages (solid line) of occurrence frequency are shown. The LSR of the brightest PMC observed on each day is also plotted, as is the zonally averaged altitude of all observed PMCs. The maximum brightness and average altitude are shown for 80° latitude only because the results at 70° and 60° are very similar. Occasional gaps can be seen in the occurrence frequencies at times when observations are not available.
 The occurrence frequencies for each season follow the same general trend as shown in Figure 4. PMCs first appear about 20 days before summer solstice. They always appear first at high latitude and within 10 to 20 days appear at 60° latitude. At the highest latitudes the occurrence frequency reaches about 90% in the north. The decline in occurrence frequency is typically slower during the second half of the season than during the first half. The core part of each season displays significant structure compared to the average shown in Figure 4. In the 1999, 2001, and 2002 northern seasons, and the 1999/2000 and 2000/2001 southern seasons, prominent episodic patterns are seen in the occurrence frequency. These patterns have typical timescales on the order of 10 to 20 days and amplitudes (peak-to-valley) on the order of 10% to 25% in occurrence frequency, depending on latitude. The relative changes in frequency increase with decreasing latitude.
 The daily maximum LSR is generally correlated with occurrence frequency. At the beginning of the season, the brightest PMCs are near the detection threshold of LSR, which is 0.7. The maximum brightness generally increases to a plateau at about solstice which lasts until about 30 days after solstice. It then declines until the end of the season when the brightest clouds detected are again near the detection threshold. The values of the brightest LSR observed in a northern season are of order 10, while the largest LSR observed in a given southern season (with observations made at a different scattering angle as described earlier) is at least 50, and can be as high as 200. Some structure appears, especially during the plateau period, that mimics structure in the occurrence frequency.
 In the 1998 northern season as well as the 1998/1999 and 2002/2003 southern seasons, dramatic depletions in frequency of occurrence followed by resurgences in activity are seen. In each case, these features occur 30 to 40 days after solstice. The magnitude of the depletion ranges from 30 to 100% with stronger depletions occurring at lower latitudes. In each case the minima in observation frequency coincide with reductions in the maximum observed brightness and increases in the average PMC altitude. In the 2000 and 2001 northern seasons and the 2001/2002 southern season, features are seen in the maximum brightness and average altitude time sequences, but related depletions in frequency of occurrence are less obvious.
 The average altitudes of PMCs vary throughout the season. At the beginning of the season, PMCs are detected at 84 to 86 km in the Northern Hemisphere and as high as 86 km in the Southern Hemisphere. The average altitude lowers as the season progresses. In the north, a minimum of about 83 km occurs about 40 days after solstice. The average altitudes of southern PMCs, like the maximum brightness, are more structured in time than are those in the north. While the northern progressions of average altitudes are very similar from year to year, there are greater interseasonal differences in the south. However, care must be taken in comparing the northern and southern PMC altitudes because the smaller scattering angles in the south allow for increased numbers of detections and may therefore introduce weaker clouds at a greater range of altitude. There is also greater sensitivity to horizontally patchy clouds which are not viewed at the tangent altitude.
5.1. Seasonal Characteristics
 The standard model for PMC formation [Turco et al., 1982; Jensen and Thomas, 1988; Reid, 1989] predicts that ice particles form near the mesopause where the partial pressure of water vapor far exceeds the saturation vapor pressure and facilitates the nucleation process. As water is deposited onto the particle and the size increases, gravity forces the particle to drop in altitude. The particles continue to grow as the abundance of water increases with decreasing altitude. Offsetting this tendency for greater growth rates, the temperature also rises with decreasing altitude so that the particles eventually sublime as they move downward. Vertical winds loft the water (and nucleation particles) back to higher altitudes so that the process can be repeated. This model suggests that PMC particles have their largest size at the lower altitudes. PMCs should therefore be brightest at the lower altitudes because while the scattering efficiency of the PMC is linearly dependent upon the number density of ice particles, it is proportional to the particle size to the sixth power [Jensen and Thomas, 1988, Figure 7]. The height of maximum particle size is thus conventionally identified as the height of maximum brightness as, for example, observed by a LIDAR. As shown in Table 1, the tangent height of maximum PMC brightness as observed by SNOE is systematically lower than the true height, by about 1 km.
 From the above model we expect that early in the PMC season, as the temperature drops below the frost point, relatively dim PMCs will appear at higher altitudes. Presumably, this occurs because the height of the surface where saturation is sufficient for nucleation is higher during the onset of summertime conditions. As the season progresses, and as this saturation level sinks, brighter PMC will begin to appear at lower altitudes. As the PMC season ends, this progression will be reversed. The SNOE observations shown in Figures 6 and 7 generally exhibit the expected seasonal patterns described above. Ten to 30 days before solstice, weak PMCs begin to appear at 85–86 km heights. As the season progresses, the occurrence frequency increases, the maximum brightness increases, and the altitudes decrease to about 83 km depending upon the specific year, latitude, and hemisphere. By the time of solstice and for a period thereafter, the occurrence frequency, brightness, and altitudes vary about their respective plateaus. Thirty to 50 days after solstice the frequency of occurrence decreases, the brightness decreases, and the average altitude generally rises. The rates of change are not usually the same at the end of the season as at the beginning, and the last clouds of the season have lower altitudes than the early season clouds.
 The statistical anti-correlation between cloud brightness and altitude observed by SNOE was also observed by SME, and by ground-based LIDAR experiments. In the Northern Hemisphere, von Zahn et al.  found from measurements at the ALOMAR Observatory (69.3°N) that over a 24-hour period, NLC brightness was anti-correlated with altitude. Thayer et al.  reported the same behavior for their measurements in Sondestrom, Greenland (67°N). At the south pole, LIDAR measurements by Chu et al.  showed the same correlation as SNOE when averaged over the season. A difference seen at the South Pole was a generally higher (by about 1.5 km) PMC altitude than in the north at lower latitudes. Because SME and SNOE measurements are global averages, taken at a fixed local time, it is not clear whether the same phenomena are being observed. However, all the data are compatible with the standard theory.
 During the periods of highest occurrence frequency, the period from about 10 days before solstice until about 40 days after, significant variability is seen in PMC activity. The occurrence frequency shows periodic reductions and increases taking place on timescales of 5 to 20 days as well as dramatic reductions and resurgences in occurrence frequency. An example of a major reduction in PMC activity can be found in the 1998 northern season 40 days after solstice. A very dramatic example occurs in the 1998/1999 southern season, also 40 days after solstice, where the occurrence frequency drops to nearly zero. On this date (27 January 1999), PMCs were virtually absent at all longitudes. The 1999/2000 southern season shows a quasiperiodic variability. Merkel et al.  have shown that the brightness of PMCs is modulated by large-scale waves with periods of about 5 days. Periodicities of 5 days are not obvious in the zonally averaged observations described in Figures 6 and 7, but could be largely eliminated in the zonal averages. Although we have not analyzed energetic particle and solar UV inputs for correlations with PMC activity, it seems more likely that dynamical influences, either through perturbation of temperature and/or water abundance, are responsible for much of the observed variability seen in PMCs. Work is in progress to identify these special periods, to determine whether concurrent measurements of temperature, water vapor, or other external forcings were anomalous.
 Much of the variability seen in occurrence frequency is also observed in maximum brightness and average altitude. The features mentioned above in the 1998 northern and 1998/1999 southern seasons are also seen in maximum brightness and average altitude. In the latter example, when the occurrence frequency dropped dramatically, the PMCs that were observed were relatively dim and occurred at a high altitude. The behavior is also consistent with the standard theory described above. At times when either the temperature or water abundance are less favorable for PMCs, the maximum observed brightnesses are low and the average altitudes are high.
5.2. Differences Between Northern and Southern Hemisphere PMCs
 A comparison of Figures 6 and 7 shows that PMC maximum brightness and average altitude exhibit larger variability in the south than in the north. While this may be partially a result of geophysical differences, it is expected because of the different scattering geometry for observations in the Northern and Southern Hemispheres. Because the southern observations are made in a forward-scattering geometry for which there is a much larger scattering coefficient, SNOE has a much greater sensitivity for detecting scattered radiance from Southern Hemisphere ice particles than Northern Hemisphere particles. This greater sensitivity allows the detection of clouds which would fall below the detection threshold in the Northern Hemisphere. This also allows SNOE to observe southern clouds earlier in their development than is possible in the north and probably explains why clouds can be seen to a higher altitude in the south.
 Interestingly, the occurrence frequencies in the south are smaller than those in the north. This is true at all latitudes. While the geographical coverage is identical in the two hemispheres, SNOE detects on average 25% more clouds in the north than in the south. There are no instrumental factors that could cause this difference. Altitude registration is more difficult in the presence of bright PMCs as observed in the south [Merkel et al., 2001]; however, while a small number of profiles contain PMCs so bright that the Rayleigh scattering profile is inverted at all altitudes, these cases are too few in number to explain the dramatic differences between northern and southern detections. In these cases, we treat the profile as containing a PMC and we place the altitude at the arbitrary value of 83.0 km. In this way, no clouds are lost, particularly the very brightest which could be of great scientific interest. DeLand et al.  analyzed SBUV observations and detected fewer PMCs in the Southern Hemisphere relative to the north, providing further confirmation of the SNOE data. SAGE-II results also indicate about 50% fewer clouds in the south than in the north [Shettle et al., 2002]. SME observed little if any variation in occurrence frequency between the two hemispheres. This suggests that north-south differences may themselves have a long-term trend. Such a trend is evident in the longer-term data sets from SBUV and SAGE-II.
 Fewer PMCs in the Southern Hemisphere suggests that the southern summer polar region is an environment less conducive to PMC formation compared to the northern polar region. We might expect from this that the southern polar mesosphere must be less saturated than the northern summer polar region. Chu et al.  compared lidar observations from the South Pole with observations in the northern polar region. They found a significant hemispheric difference in PMC altitudes. They concluded that the closer proximity of the Earth to the Sun in the southern summer, due to the eccentricity of the Earth's orbit, produced a higher colder mesopause but a smaller region of supersaturation that then led to different PMC altitudes. Siskind et al.  examined the possibility of summertime hemispheric differences in temperature and water vapor concentration as a result of the eccentricity of the Earth's orbit and also by hemispheric differences in gravity-wave filtering due to hemispheric asymmetries in the mean winds of the troposphere and lower stratosphere. They found that while radiative effects due to the eccentricity of the Earth's orbit were non-negligible between 20 and 65 km, the asymmetries in dynamical behavior produced larger hemispheric differences in temperatures above 65 km. They concluded that the southern summertime stratosphere and mesosphere should be 3 to 8 K warmer than the Northern Hemisphere. They pointed out that these hemispheric differences in temperature and humidity suggest that fewer PMCs will form in the Southern relative to the Northern Hemisphere. The SNOE, SBUV, and SAGE-II observations support this conclusion.
 In more recent studies, Siskind et al.  suggest that the clouds are also brighter in the north. Given the different observation geometries between the north and the south used for SNOE, it is not possible to determine if the clouds are brighter in the north. For the 2000/2001 southern season, however, special SNOE observations were made at scattering angles similar to those in the north. These observations are currently being analyzed and will be described in a future publication.
5.3. Dynamical Influences on PMC Morphology
 The simple description of the PMC formation cycle described in section 5.1 suggests an orderly seasonal evolution of PMC occurrence, consisting of a rise to maximum, a relatively stable period, and a period of decrease. The duration of this evolution is on the order of 90 days. This view is supported by the seasonally averaged results of Figure 4. Examining seasons individually as in Figures 6 and 7 reveals more structure. Rapid growth and decay in the PMC frequency occurs over timescales of 10 to 20 days. The PMC brightness also shows these oscillations. Kirkwood and Stebel  suggested that planetary waves affect mesospheric ice formation. DeLand et al.  have also inferred the presence of planetary wave effects in SBUV observations. Merkel et al.  examined spatial maps of SNOE PMC scattering ratios and showed with Fourier techniques that PMC brightness was modulated by 5-day planetary waves. Could the episodic behavior of the SNOE frequencies be caused by planetary-scale waves? Espy et al.  identified a 16-day normal mode in their OH summertime temperature data at Stockholm, Sweden (55°N). However, since this particular mode is a westward traveling wave 1, a zonal-average of a sinusoidal perturbation would dampen or completely remove the effects of a longitudinally periodic perturbation. While we suggest that planetary waves may influence the morphology of PMC formation, we recognize that PMC formation is also influenced by gravity waves and tides [Thayer et al., 2003]. An important clue is that the influences on PMC occurrence frequency appear to increase toward the equator. Work is underway to identify possible underlying causes.
 We have described the detection of PMCs by the SNOE satellite and the database of PMC detections. SNOE provides a rich database for the spatial and temporal morphology of PMCs. We have shown that the observed SNOE PMC morphology is in general accord with SME observations. The observed frequencies of occurrence, maximum brightness, and average altitude of PMCs for several seasons are consistent with the standard model of cloud formation in that PMC particles form at high altitudes in the coldest air, gain in size as they descend in altitude, and finally sublime below the lower brightest portion of the cloud. A major result of this study is that significantly fewer PMCs are observed in the Southern Hemisphere relative to the north. This is true given identical geographic coverage and despite the increased sensitivity to scattered radiance by PMCs in the south. We conclude that the southern polar mesosphere must be on average less saturated compared to the northern polar mesosphere. The SNOE observations also suggest that while the general characteristics of the PMC season are controlled by large-scale changes in temperature, episodic wave influences also play a strong role, particularly near the equatorward cloud boundary.
 This work was supported by NASA grant NAG5-12648. A. W. M. was supported by the National Science Foundation as a CEDAR post doctoral researcher. G. E. T. was supported by the NASA Living With a Star program. We have benefited greatly from conversations with Charles Barth as well as support from Kenneth Mankoff. We thank the referee for helpful comments that improved the manuscript. We thank the entire SNOE team for their support. SNOE was managed for NASA by the Universities Space Research Association.