Location controls the findings of ground-based PSC observations

Abstract. Spaceborne observations of Polar Stratospheric Clouds (PSCs) with the Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) aboard the Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) satellite provide a comprehensive picture of the occurrence of Arctic and Antarctic PSCs as well as their microphysical properties. However, advances in understanding PSC microphysics also require measurements with ground-based instruments, which are often superior to CALIOP in terms of, e.g. time resolution, measured parameters, and signal-to-noise ratio. This advantage is balanced by the location of ground-based PSC observations and their dependence on tropospheric cloudiness. CALIPSO observations during the boreal winters from December 2006 to February 2018 and the austral winters 2012 and 2015 are used to assess the representativeness of ground-based PSC observations with lidar in the Arctic and Antarctic, respectively. Information on tropospheric and stratospheric clouds from the CALIPSO Cloud Profile product (05kmCPro version 4.10) and the Polar Stratospheric Cloud (PSC) mask version 2, respectively, is combined on a profile-by-profile basis to identify conditions under which a ground-based lidar is likely to perform useful measurements for the analysis of PSC occurrence. It is found that the location of a ground-based measurement together with the related tropospheric cloudiness can have a profound impact on the derived PSC statistics and that these findings are rarely in agreement with polar-wide results from CALIOP observations. Considering the current polar research infrastructure, it is concluded that the most suitable sites for the expansion of capabilities for ground-based lidar observations of PSCs are Summit and Villum in the Arctic and Concordia, Troll, and Vostok in the Antarctic.


enable meaningful PSC observations with a ground-based lidar. In addition, all-sky refers to the use of all profiles independent of tropospheric cloudiness.
The PSC mask v2 is processed analogous to the Vertical Feature Mask for tropospheric clouds by accumulating the number 90 of height bins with different PSC types for each CALIPSO profile. PSCs that extend over just one height bin are excluded from the analysis. Profiles are referred to as containing a certain a PSC type, for instance STS, if this types was identified in at least one of the PSC height bins. Maps of the occurrence of the accumulated number of height bins related to different PSC types are normalised by the total number of PSC height bins per considered profile or grid box.
To enable a combined analysis of cloudiness in the polar troposphere and stratosphere, the data extracted from the 05km-95 CPro.v4.10 and PSC Mask v2 products are temporally matched and reduced to only those profiles with detected PSCs. The data set is then filtered according to the occurrence of tropospheric clouds and different PSC types. The filtered data is gridded into cells of 1.25 • latitude by 2.50 • longitude for visualisation of PSC occurrence.
The matched observations of tropospheric and stratospheric clouds allow for a direct comparison of PSC statistics as seen from ground and space independent of the considered instruments. To assess the representativeness of ground-based PSC 100 measurements, PSC statistics are obtained for boxes of 2 • latitude by 2 • longitude around the sites in Figure 1 and Table 1.
True PSC statistics unaffected by tropospheric cloudiness, i.e. during all-sky conditions, can only be obtained with a spaceborne lidar. In contrast, filtering with respect to tropospheric cloudiness is applied to emulate the likely conditions for meaningful ground-based PSC measurements in the CALIPSO data set. Specifically, we assume that a ground-based lidar would only provide meaningful results during conditions with no clouds or only transparent clouds that would not already attenuate the 105 laser beam before it can reach PSC altitudes. This is referred to as the ground-based view. We subsequently separate between observations of (i) a continuously operating ground-based lidar for which all cases of the ground-based view are considered and (ii) a manually operated system for which one third of the cases of the ground-based view was randomly selected. The two ground-based configurations are used to account for sampling effects related to the fact that most ground-based instruments are operated manually and on campaign basis and that the decision to start a measurement, i.e. the assessment of tropospheric 110 cloudiness, is made subjectively by the operator. The purpose is hence to provide an estimate of the potential effects of, e.g. system downtime, logistical problems, and lack of personnel (to list just a few infrastructural challenges in operating a groundbased lidar at a remote location and under harsh conditions) on the inferred PSC statistics.  occurrence. This is levelled by the normalised occurrence rate of suitable conditions for ground-based observations presented in Figure 2b. The region of highest PSC occurrence rate over the north Atlantic coincides with the highest occurrence of opaque tropospheric clouds. While Ny Ålesund could potentially observe the most PSCs in the Arctic, the occurrence rate of good conditions for ground-based lidar measurements is much lower than at the other Arctic sites. In contrast, sites on Greenland and in the Canadian Arctic show almost no opaque clouds but -with the exception of Villum -also feature a low occurrence rate 125 of PSCs. A similar situation though with a generally lower rate of suitable conditions for ground-based observations is found for Alomar, Esrange, and Sodankylä. However, these sites provide much easier access than the other more remote locations.
Tiksi is a site that could potentially provide information on PSCs over the Siberian Arctic.
The occurrence rate of different PSC types in the Arctic for all-sky conditions is shown in Figure 3. The figure reveals that STS and NAT mixture are most abundant with a region of maximum STS occurrence over the north Atlantic and southern 130 Greenland. The occurrence rates of NAT enhanced and ICE are well below 10% and neither type shows an area of pronounced occurrence. The distribution of wave ICE in Figure 3e shows that this type is restricted regionally to southeastern Greenland, around Iceland, southern Svalbard, the Scandinavian mountain range, and Novaya Zemlya. Figure 4 provides a local quantification of the Arctic-wide display in Figure 3 for the selected Arctic sites in Table 1 in the form of the occurrence rate of different PSC types as seen by a spaceborne instrument (all-sky conditions, same as in 135 Figure 3), a continuously operating ground-based instrument (no or only transparent clouds), and a manually operated groundbased instrument (one third of randomly selected CALIPSO profiles in the presence of no or only transparent clouds). For the entire Arctic, the spaceborne view gives a smaller fraction of NAT mixture compared to the ground-based view because the regional minimum in the occurrence rate of NAT mixture ( Figure 3b) covers the location of most of the considered ground sites. This is balanced by a larger fraction of STS for the entire Arctic compared to most ground sites. The occurrence rates of 140 NAT enhanced, ICE, and wave ICE are marginal with a total contribution of less then 10% of all observed PSC height bins.
Tropospheric cloudiness would allow for ground-based observations in only about 42% of all Arctic CALIPSO PSC profiles.
This causes the slight difference between the three bars related to Arctic-wide observations in sensitive to cloudiness and further sub-sampling. A considerable difference between the spaceborne and ground-based view is found in the European Arctic, particularly at Myvatn and Sodankylä. The occurrence rate of STS (ICE) is underestimated (overestimated) at Esrange, Myvatn, and Sodankylä while the opposite is found at Alomar and Ny Ålesund. The ratio of the number of PSC height bins representing the ground-based versus the spaceborne view is given in the third column of Table 1 a difference of also about 10 percentage points compared to the Antarctic mean. Casey is also the station with the highest occurrence rate of NAT mixture followed by Mirny. In addition, these two stations show almost no ICE PSCs. The lowest rate 190 of NAT mixture and the highest rate of ICE (45%-50%) is found at Belgrano II, as this is the only site located in the regional minimum (maximum) of the occurrence rate of NAT mixture (ICE) revealed in Figure 6. All other sites show ICE occurrence rates below the Antarctic average. Wave ICE is found only at Jang Bogo (1%) and McMurdo (0.5%). the high occurrence rate of tropospheric clouds is levelled by the also high PSC occurrence rate (see Figure 2). Note that the assessment in Figure 8 is based entirely on atmospheric conditions and does not consider infrastructural challenges such as the accessibility, power supply, or availability of facilities at the respective sites; or the training and work load of the stationed 205 personnel. It is because of this that most of the established PSC observatories fall into a region that could be considered as less suitable for establishing a ground-site for PSC observations. Nevertheless, the trade-off between PSC occurrence and tropospheric cloudiness at those sites still creates conditions that allow for meaningful amounts of PSC observations -as witnessed by the available literature. If new PSC observatories were to be established, the most suitable choices -based solely on atmospheric conditions -would be Villum, Summit, Zackenberg, Thule, and Alert in the Arctic; and Vostok, Concordia, 210 Troll, Jang Bogo, Belgrano II, and Neumayer III in the Antarctic.

Summary and conclusions
There is a rich literature on airborne and ground-based PSC measurements going back to the 1980s. The thus collected time In the Arctic, there is the additional combined effect of the inhomogeneous distribution in the occurrence of both PSCs and tropospheric clouds on the representativeness of ground-based PSC observations with respect to all-sky conditions.
The combination of the occurrence rate of PSCs and of suitable conditions for ground-based PSC observations allows to assess the suitability of a ground site for long-term lidar measurements of PSCs. This suitability is related solely to atmospheric conditions and does not consider challenges with respect to logistics, personnel, or training. According to this definition, measurements at more suitable sites will require less measurement effort to obtain a data set that can be used to infer statistically 225 significant PSC data. This knowledge is important as ground-based lidars are generally more advanced than spaceborne instruments and allow to independently retrieve backscatter and extinction coefficients as well as the particle linear depolarisation ratio at multiple wavelengths and at a better signal-to-noise ratio. Their measurements are therefore invaluable for a better understanding processes related to PSC formation and persistence.
Of the established PSC observatories only Eureka, McMurdo, and Ny Ålesund are found to fall into a category that provides 230 a good balance between PSC occurrence and tropospheric cloudiness. Dumont d'Urville is at the lower end of available PSC observations while Esrange, Sodankylä, and Syowa all show only about 1000 CALIPSO PSC profiles during conditions for ground-based measurements. The occurrence rate of PSCs in the Arctic is much lower than in the Antarctic. Hence, the assessment prevented here is particularly important for Arctic sites. Considering only atmospheric conditions, it is found that Villum, Summit, Zackenberg, Thule, and Alert would be the best choices for establishing new PSC observatories with state-235 of-the-art lidar instruments the Arctic. In the Antarctic, this is that case for Vostok, Concordia, Troll, Jang Bogo, Belgrano II, and Neumayer III.
The strong dependence of PSC formation on temperature suggests a crucial role of processes that enhance local cooling (Carslaw et al., 1998;Teitelbaum et al., 2001). These include synoptic or mesoscale events that are generally linked to specific types of tropospheric cloudiness. It is therefore reasonable to expect a connection between tropospheric cloudiness and the 240 occurrence of PSCs and maybe even different PSC types. Initial studies focussed on individual winters in the Arctic (Achtert et al., 2012) and Antarctic (Wang et al., 2008;Adhikari et al., 2010) show that particularly high and deep-convective cloud systems have a strong effect on PSC formation. This indicates that tropospheric meteorology might be an important driver for the interannual variability in PSC formation and ozone hole recovery. While CALIPSO is operational since 2006, there has not yet been a thorough assessment of the dependence of the occurrence of different PSC types on tropospheric cloudiness. In the 245 future, the combined CALIPSO data set of clouds in the troposphere and stratosphere presented here will be used to investigate this connection.   Table 1 versus the ratio of PSC height bins as observed by a ground-based and a spaceborne lidar (columns 3 and 6 in Table 1). Filled symbols mark sites with published PSC climatologies. Horizontal lines mark the values for the entire Arctic and Antarctic, respectively. The vertical dashed line separates stations with more than 2000 CALIPSO PSC profiles from those with fewer observations. Grey lines mark the ratios 0.6:1.0, 0.8:1.0, 1.0:1.0, 1.0:1.2, 1.0:1.4, and 1.0:1.6. Stations abbreviations are given in Table 1. Table 1. Overview of the location of Arctic and Antarctic research stations. Station abbreviations in columns 1 and 5 are used to mark the corresponding sites in Figures 1 and 8. Stations with a deployment of atmospheric lidar instruments are marked with ♣ while those with existing PSC data sets are marked with ♠. R gives the ratio of PSC height bins for tropospheric cloudiness that relates to the data coverage of a ground-based (cloud-free and transparent clouds) and a spaceborne lidar (all-sky).