3.1 Instrumentation

During the campaign, HALO was equipped with a wide variety of in situ and remote sensing instruments to measure cloud particles, aerosols, trace gases (H₂O, NO, NO_y, CO, CO₂, CH₄, and O₃), radiation, and basic meteorological parameters (pressure, temperature, position, wind direction, and velocity). The data used in this study were collected by the Cloud Combination Probe (CCP) and the Precipitation Imaging Probe (PIP) mounted under the wings of the HALO aircraft, and they are presented below. Detailed information about the other cloud probes and instrumentation installed for this campaign is given in Voigt et al. (2017).

3.2 Methodology

4.1 Overview and differences in the microphysical properties of cirrus (N, ED, and IWC) at mid- and high latitudes

The cirrus measurements are classified by latitude in order to investigate and compare microphysical properties of cirrus at mid- and high latitudes. For this first classification, we consider the latitude at the measurement point. Data obtained at latitudes <60^∘ N are considered to be mid-latitude (ML) cirrus, and data collected at ≥60^∘ N are considered to be high-latitude (HL) cirrus. This differentiation is somewhat arbitrary, as there is no universally accepted definition of the three general latitude zones. Perry (1987), for example, defined the mid-latitude in the Northern Hemisphere as the zone between 35 and 56^∘ N, but the limits vary according to the month and year. In Sect. 5.1, we discuss the variation in the latitude threshold, and we find no major influence on the conclusions of this work (see also Fig. S4 in the Supplement).

Figure 2 shows an overview of the frequency distribution of the measured cirrus microphysical properties with respect to latitude. The normalized frequencies of occurrence allow the comparison between latitudes, regardless of the number of observations in each bin. All together, we provide a cirrus data set from 41 564 cloud samples in 2 s time resolution, of which 15 211 samples are HL cirrus and 26 353 samples are ML cirrus. Overall medians of N, ED, and IWC (denoted with a tilde) for ML and HL cirrus are indicated in Fig. 2 and Table 2. We consider the median instead of the mean, since the mean is more affected by outliers. The overall median N for ML cirrus (0.0086 cm⁻³) is higher than for HL cirrus (0.001 cm⁻³) by an order of magnitude. The median ED in ML cirrus is 165 µm, which is smaller than the HL cirrus at 210 µm, and the median IWC does not differ much between ML and HL cirrus, as it represents a combined effect of N and ED. A similar behaviour is observed in the normalized frequency distribution of β_ext, which is shown in Fig. S5 in the Supplement.

Figure 2Normalized frequency distribution of (a) N, (b) ED, and (c) IWC observations as a function of the latitude of the measurement during CIRRUS-HL. Latitude bins are 1^∘ wide, and vertical bins are logarithmic. The colour code indicates the frequency of occurrence in percent (%) per 1^∘ latitude bin, which is normalized by the total counts per latitude bin. The vertical dashed line marks the threshold of 60^∘ N for the differentiation of ML and HL cirrus. Triangular markers are medians per latitude bin ($\stackrel{\mathrm{̃}}{N}$, $\stackrel{\mathrm{̃}}{\text{ED}}$, and $\stackrel{\mathrm{̃}}{\text{IWC}}$). The top-most panel and right column contain histograms of the corresponding variables. The solid black line in panel (b) is a linear fit of the ED medians, and the dashed lines are the corresponding linear fits of the 5th and 95th percentiles.

Download

Table 2Median, 25th percentile, and 75th percentile of the microphysical properties N, ED, and IWC in mid- and high latitudes during CIRRUS-HL.

Download Print Version | Download XLSX

Cirrus measured at mid-latitudes are characterized by higher $\stackrel{\mathrm{̃}}{N}$ than those measured at high latitudes. $\stackrel{\mathrm{̃}}{N}$ in HL cirrus mostly takes values between 0.0001 and 0.01 cm⁻³. The central range of the ML cirrus measurements is shifted to values that are 1 order of magnitude larger (0.001–0.1 cm⁻³). In addition, Fig. 2a also indicates that high particle number concentrations ($N>\mathrm{1}{\mathrm{cm}}^{-\mathrm{3}}$) were more frequently observed at mid-latitudes. A linear correlation of the $\stackrel{\mathrm{̃}}{\text{ED}}$s is drawn in Fig. 2b and exhibits a slight and clear positive slope of ≈4 µm per degree. The correlations of the 5th and 95th percentiles show that the distribution of ED in ML cirrus is broader, and smaller EDs are more frequently observed with decreasing latitude. HL cirrus contain a lower IWC; however, no significant correlation with latitude is noticed in Fig. 2c. The range of values agree well with the findings by Schiller et al. (2008) for the Arctic and mid-latitudes. Voigt et al. (2017) also reported IWC between 10⁻⁶ and 0.2 g m⁻³, with a high variability during the ML-CIRRUS campaign at mid-latitudes in spring.

The cirrus properties from our measurements agree well with observations from previous campaigns. Brown and Francis (1995) and Heymsfield et al. (2010, 2013) found, in general, similar median values of N and IWC (mainly at mid-latitudes and the tropics). However, Brown and Francis (1995), for example, only included data from forward-scattering probes. Here we use also the CIPgs and PIP, which have higher sample volumes. For this reason, we also measure lower IWC than 0.001 g m⁻³. In general, our results also agree with those reported by Krämer et al. (2016, 2020) and Luebke et al. (2016) at mid-latitudes, but the larger EDs that result from our analysis are due to the addition of the PIP data and due to different sizing methods applied in the calculation of the diameter from the CIPgs images. Here, we consider the maximum dimension of the ice crystals and not the area-equivalent diameter. The lower N observed in HL cirrus compared to ML cirrus is consistent with the observations from Wolf et al. (2018). Marsing et al. (2023) showed averaged IWCs (0.003 and 0.005 g m⁻³) of two case studies during the POLSTRACC (Polar Stratosphere in a Changing Climate) campaign, which is in line with our measurements at high latitudes.

According to our measurements, the combination of higher N and smaller ED ice crystals is characteristic for ML cirrus. The cause of that can be manifold; more anthropogenic activities, including aviation, at mid-latitudes result in higher aerosol and ice nuclei loads in the upper troposphere. Gayet et al. (2004) also found higher N and lower ED in the more anthropogenically influenced Northern Hemisphere with respect to the more pristine Southern Hemisphere. Aviation-induced cirrus tend to have high number concentrations and lower ED for several hours after their formation (Voigt et al., 2017; Schumann and Heymsfield, 2017; Schumann et al., 2017). In general, it is difficult to distinguish aged contrails from thin natural cirrus unaffected by aviation (Li et al., 2022) or embedded contrails in natural cirrus (Unterstrasser et al., 2017a, b), in particular, as the influence of air traffic over Europe is omnipresent (Voigt et al., 2017; Schumann and Heymsfield, 2017).

In contrast to the ML cirrus, we often find that HL cirrus have lower N but larger ED, which could be attributed to the uptake of the ambient water vapour by few INPs and the further growth of a smaller number ice crystals that thus allow larger sizes. The resulting number concentration of the ice crystals depends on the temperature, the updraft, and the number of INPs, as well as their capability to nucleate ice (Kärcher et al., 2006). Wolf et al. (2018) also found lower N at high latitudes and identified higher concentrations of INPs at the mid-latitudes compared to the Arctic as a possible explanation. Beer et al. (2022) analysed global model simulations under cirrus formation conditions and showed higher number concentrations of INPs (about 0.1 cm⁻³) and higher number concentrations of newly formed ice crystals (about 0.001 cm⁻³) between 30–60^∘ N compared to 60–90^∘ N. However, a confirmation from airborne measurements of INPs is very challenging, as measurements in the cirrus regime with temperatures lower than −38 ^∘C are very scarce. DeMott et al. (2003, 2010) reported different measurements of ice particle residuals at temperatures $>-\mathrm{40}{}^{\circ }\mathrm{C}$ and showed a large variability in INP number concentrations, depending on regional and seasonal changes in the aerosol sources. Typically, mineral dust particles are found to have a high nucleation efficiency (DeMott et al., 2003), and black carbon (BC) particles from anthropogenic sources also act as INPs. In particular, the importance of aviation soot particles as ice nuclei is not yet well determined. A number of studies investigated the aviation soot impact on cirrus and showed a large range of possible model results, evidencing the uncertainty that exists in this regard (Hendricks et al., 2011; Gettelman et al., 2012; Zhu and Penner, 2020; Righi et al., 2021).

A comparison between the temperature profiles of ED and N in ML (red) and HL (blue) cirrus is presented in Fig. 3 and sorted in 2.5 ^∘C temperature bins. We show the temperature along the vertical axis as a representation of altitude. In Fig. 3a, we observe a clear decrease in the ED with decreasing temperature for both ML and HL cirrus. Both profiles cover a similar temperature range and correspond to altitudes between 8.5 and 13.5 km, with fewer measurements at high latitudes. The same decrease in the ED is also observed as a function of altitude. This effect has already been observed in previous studies and can be explained by reduced atmospheric water content at lower temperatures (or higher altitude), together with the different altitudes at which in situ and liquid origin cirrus exist (Luebke et al., 2016; Wernli et al., 2016). With increasing altitude (decreasing temperature), cirrus origins change from cirrus dominated by the liquid origin regime with larger ice particles to in situ origin cirrus with smaller particles (see also Voigt et al., 2017). This is expected, due to the low-altitude origin of the liquid origin cirrus (Luebke et al., 2016). Larger particles can also be found at higher temperatures (lower altitude) resulting from sedimentation processes.

Figure 3Distribution of (a) ED and (b) N observations sorted in temperature bins of 2.5 ^∘C for ML (red) and HL cirrus (blue) and (c) a histogram of the total frequency of observations per temperature bin for each group. The thicker solid lines correspond to the median values per temperature bin. The dark shaded areas between the dashed lines comprise the 25th and 75th percentiles of the population. The lighter shaded areas mark the extension between the 1st and 99th percentiles. The black asterisk indicates a low statistically significant difference between medians. Blue and red asterisks indicate low statistics in HL and ML data, respectively (see the text for further details).

Download

Between −50 and −38 ^∘C, we find a remarkable difference between the 25th and 75th percentiles in the ED between ML and HL cirrus. This fact probably indicates a more dominant influence of liquid origin cirrus at the high latitudes with enhanced sedimentation as a result of the larger particles at these latitudes. From the temperature profiles, we can confirm the larger EDs observed at high latitudes in the general picture of Fig. 2b. Higher EDs in HL cirrus compared to ML cirrus are observed in almost all temperatures. To confirm this finding, we use the U test according to Wilcoxon, Mann, and Whitney (mannwhitneyu function from the Python sub-package scipy.stats) to assess whether there is a statistically significant difference between the ML and HL medians per temperature bin for ED (and also for N). The null hypothesis that the distributions are equal was rejected for all cases with p value $<{\mathrm{10}}^{-\mathrm{6}}$, except for the ED medians between −58 and −55.5 ^∘C with a p value =0.028. Therefore, we conclude that the observed differences are statistically significant. Between −45 and −38 ^∘C at mid-latitudes, we observe high ED values in the 1st to 99th percentiles. These events are connected to isolated convective systems over Germany, with an enhanced growth of the ice particles.

Contrary to the ED, N does not show a clear tendency with decreasing temperature in either case. However, the N in ML and HL cirrus differ at all temperatures by about an order of magnitude, which is more pronounced at lower temperatures. However, the ML cirrus coldest temperature bin, which mostly contains high N, has reduced statistics with a total of 119 2 s samples measured from four different cloud sequences. Between −62 and −55 ^∘C, there is an enhancement of N in ML cirrus associated with a steeper decrease in ED, which is probably connected to contrail formation. According to Bräuer et al. (2021a), contrail conditions are favourable between −50 and −60 ^∘C. Here contrails have the highest extinction coefficients. This corresponds to altitudes between 9.5 and 11 km, where short- and medium-range commercial flights are typically located. Contrails are frequently present in ML cirrus above central Europe. Here, we mostly find high N and low ED around −55 and −60 ^∘C in our data set. High N values ($>\mathrm{1}{\mathrm{cm}}^{-\mathrm{3}}$) in the other temperature ranges are rare and represented by the outliers, which do not impact the medians. Measurements of HL cirrus below −55.5 ^∘C are less representative, with 210 counts in total. In particular, between −63 and −60.5 ^∘C, we find 22 2 s consecutive samples from the same cloud sequence. Between −60.5 and −58 ^∘C three different cloud sequences were probed, with a total of 40 counts. Fluctuations in ED in HL in these temperature ranges in Fig. 3a can thus be explained by the reduced statistics.

In addition, we examined whether part of the cirrus were measured in the stratosphere, as stratosphere-penetrating extratropical convection is most likely to occur during summer (Dessler, 2009). In addition, the tropopause (TP) height vary strongly between 35 and 50^∘ N, and the cirrus measurements of high and mid-latitudes could be differently distributed with respect to the TP height (Dessler, 2009). In order to derive the altitude of the cirrus measurements with respect to the TP, the TP height was extracted from the data set Hoffmann and Spang (2021), and the definition used was from the WMO (World Meteorological Organization), which includes the first TP height from ECMWF's reanalysis ERA5 data (Hoffmann and Spang, 2022). The absolute frequency distribution of the HL and ML cirrus altitudes with respect to the TP is shown in Fig. S2 and reveals similar profiles between mid- and high latitudes. From all data points, 2.6 % were measurements above the TP, with only 0.2 % belonging to HL cirrus, and most of the measurements were performed in the upper troposphere at around 1 km below the TP. We do not observe a substantial influence of stratospheric clouds, and there is an even smaller contribution at high latitudes.

4.2 Relating ice cloud microphysical properties to the location of cloud formation

The main approach of this study is the analysis of cirrus microphysical properties, depending on their latitude. So far, we have classified the data in ML and HL cirrus using a latitude threshold at 60^∘ N regarding the measurement point. Now, we investigate the meteorological conditions at the time of cirrus formation and how they affect the observed microphysical properties. Following this strategy, we obtain the following four categories: cirrus formed and measured at mid-latitudes (M–M), cirrus formed at mid-latitudes and measured at high latitudes (M–H), cirrus formed at high latitudes and measured at mid-latitudes (H–M), and cirrus formed and measured at high latitudes (H–H). Examples of backward trajectories for each group are shown in Fig. 4. The H–M cirrus is excluded from the analysis, as it only contains 36 data points, and all of them are just at the limits of the boundaries (see Fig. 5). In Fig. 4, the points with IWC>0 in the trajectory before the purple star indicate a previous formation of ice within the air parcel which evaporated later.

Figure 4IWC (colour coded) along three example trajectories from the cirrus classifications of formation and measurement at high latitudes (H–H, blue); formation at mid-latitudes and measurement at high latitudes (M–H, green); and formation and measurement at mid-latitudes (M–M, orange). The formation point is indicated with a purple star.

Figure 5 shows the distribution of 2 s measurements for the four categories delimited by the 60^∘ N threshold and colour coded according to ED. The three important categories are M–M in the bottom left, M–H in the bottom right, and H–H in the upper right. Using this categorization, we separate the measurements previously classified as HL cirrus into cirrus formed at high latitudes (H–H) and cirrus formed at mid-latitudes and transported to high latitudes (M–H). The number of data points found in each category (4450 for H–H and 10 753 for M–H) indicates that the category M–H contributes more frequently to the Arctic cirrus observed in our measurement campaign.

It becomes evident that the distribution of the measured ED is connected with the latitude at the measurement point. Yellow colours represent high EDs of up to 1000 µm, and the dark purple dots represent EDs of about 10 µm. Higher occurrences of smaller particles are measured in ML cirrus and higher occurrences of larger particle sizes are found in HL cirrus. The differences are smaller when looking at the latitude of the formation of cirrus measured in high latitudes. We can conclude that the measurement location has a larger influence on the particle size than the formation region.

Figure 5Correlation of latitude at the formation and latitude at the measurement for 2 s measurement points. The colour indicates the associated ED. Four regions are delimited with dashed black lines and define the groups of H–H, M–H, and M–M (see the text for details).

Download

In order to investigate in more detail how the cirrus properties depend on their source region, Fig. 6 illustrates the normalized frequency distributions of N, ED, and RH_i of the three categories. The maximum of the N distribution of the M–M cirrus lies between 0.01 and 0.1 cm⁻³. In contrast, the highest probability in the H–H cirrus is shifted to significantly lower values between 0.0001 and 0.001 cm⁻³. The M–H cirrus distribution exhibits an intermediate behaviour. The medians reflect the same observation, with 0.0086, 0.0018, and 0.0004 cm⁻³ for the M–M, M–H, and H–H cirrus, respectively. The ED profiles in Fig. 6b show the opposite trend, as do the medians (164, 206, and 225 µm for M–M, M–H, and H–H cirrus, respectively).

Figure 6Relative frequency distributions of (a) N, (b) ED, and (c) RH_i of the cirrus formed and measured at high latitudes (H–H in blue), cirrus formed at mid-latitudes and measured at high latitudes (M–H in green), and cirrus formed and measured at mid-latitudes (M–M in orange). The distribution of N is given in logarithmic bins. Linear bins of 15 µm and 2.5 % width have been chosen for ED and RH_i, respectively. The corresponding median values of each variable and category are depicted with dashed lines.

Download

In addition, Fig. 6c shows the measured RH_i for the three categories. In general, the tropopause region in summer 2021 had a high occurrence of supersaturation both at mid- and high latitudes. We observe a clear difference in the RH_i distribution with lower RH_i in M–M cirrus, with a median of 107 %, and higher values in the H–H cirrus, with a median of 125 %. This can be explained by the lower ice crystal concentrations at high latitudes due to the reduced availability of INPs. The relative humidity is not sufficient to homogeneously nucleate new particles, and instead, the small number of ice crystals available takes up the abundant water vapour that increases their size. On the contrary, water vapour at mid-latitudes is rapidly consumed by the availability of many ice particles present in the upper troposphere, which only leads to a moderate growth.

The M–M cirrus is 57 % of liquid origin and 43 % of in situ origin in our data set. Regarding HL cirrus, on the one hand we find that 90 % of the measurements corresponding to the M–H cirrus have a liquid origin with a larger cloud age. On the other hand, 86 % of the H–H cirrus data points are classified as in situ origin cirrus. Therefore, M–H cirrus were formed from the liquid phase in a ML environment rich in nucleating particles. Here many small particles were nucleated in a first stage. Then, the air parcels are advected northwards to high latitudes, bringing the clouds into a highly supersaturated atmosphere and allowing the formed ice particles to grow. Cirrus measured at high latitudes can be clearly distinguished from ML cirrus by the bimodality of the ED distribution. The bimodality of the ED distribution of M–H and H–H cirrus can be understood by two different processes; these are higher ice crystal growth at high latitudes due to the increased supersaturation, leading to the sedimentation of large ice crystals from the upper layers, or different nucleation times during their lifetime with an enhanced growth of the larger mode, due to fewer INPs and high supersaturation.

These findings together clearly show an influence of the ML origin in the development of cirrus at high latitudes, which conserves the microphysical properties of the ML cirrus while being transported to higher latitudes. In this way, the formation region influences and defines the initial ice crystal properties, which then mix or are modified during their lifetime due to the different atmospheric conditions. Thus, the largest fraction of clouds observed during CIRRUS-HL originate from the mid-latitudes and, through long-range transport, modify the HL cirrus. Opposite influences from high to mid-latitudes have not been observed.

We can conclude that latitudinal differences in microphysical parameters are better understood when looking at both the location of the ice crystal formation and the location of the measurement. The region in which the ice crystals are formed influences the initial cirrus properties, in particular the initial N. The latitude at which the cloud particles were measured determines the resulting state of the measured ice crystal properties, which is strongly influenced by the RH_i throughout the cirrus life cycle and mainly affects the measured ED. Both processes are largely influenced by the updraft on small and large scales, which are discussed below.

4.3 Liquid and in situ origin cirrus at mid- and high latitudes

Some of the differences in the microphysics of the cirrus measured during CIRRUS-HL discussed in Sect. 4.1 could be attributed to a greater or lesser presence of in situ or liquid origin cirrus at mid- or high latitudes. Here, we classify the measurements making use of the information on LWC and IWC along the backward trajectories after the formation of ice in the cloud, as described in Sect. 3.2.2. We obtain the following four groups with this division: in situ origin HL cirrus, liquid origin HL cirrus, in situ origin ML cirrus, and liquid origin ML cirrus.

The frequency of measurements of each group and their microphysical properties in terms of ED and N are presented in Fig. 7. Liquid origin cirrus were frequently measured at mid-latitudes and were also the dominant type at high latitudes. Liquid origin HL cirrus were observed almost as frequently as the in situ origin ML cirrus. In situ origin HL cirrus, in turn, were the least measured. However, isolated rare events are not well represented in the diagrams, due to the calculation of contours by interpolation.

Figure 7Frequency of ED observations in counts as a function of N. The observations are separated into (a) in situ origin HL cirrus, (b) liquid origin HL cirrus, (c) in situ origin ML cirrus, and (d) liquid origin ML cirrus. The method to draw the contours uses a marching squares algorithm. Vertical and horizontal solid red lines indicate the overall median values of ED and N, respectively, and are shown to better illustrate differences. The total number of 2 s data points are given in the upper right corner of each panel.

Download

Larger particles than the total ED median value (180 µm) were found in both cirrus types at high latitudes, with larger values for liquid origin cirrus. In situ origin HL cirrus were rather thin, with N often under 0.004 cm⁻³, which is the median value of all data. The ED distribution of both cirrus types at high latitudes exhibits two distinct maxima, as seen in the M–H and H–H cirrus in Fig. 6b. Large EDs representative of liquid origin cirrus are present in mid-latitudes, with some outliers of high EDs and high N in convective cells also involving precipitation particles. However, smaller EDs than the median, even for the liquid origin cirrus, are more frequent at mid-latitudes.

The observations at high latitudes generally agree well with the measurements in the Arctic from Wolf et al. (2018). However, high number concentrations are not found in our in situ origin HL cirrus, in contrast to Wolf et al. (2018). The reason is that their measurements of Arctic in situ origin cirrus were dominated by homogeneous nucleation events driven by high updraft motions. Our analysis of the updraft speed along the backward trajectories and the measured vertical velocities indicate rather low vertical velocities. Specific information on the updraft speeds along the backward trajectories of each of the four categories can be found in Fig. S3. In addition, orographic cirrus were not a target of the campaign, and therefore, we do not expect mountain wave cirrus with high number concentrations. We assume that mainly heterogeneous nucleation defined the cloud formation in our data set. However, it should be noted that the temporal resolution of the parameters calculated along the trajectories is 1 h, and therefore, the effects on the updrafts from small-scale fluctuations and turbulence are not taken into account here. The prevalence of heterogeneous freezing was also shown by Froyd et al. (2022) at mid- and high latitudes in the Northern Hemisphere during spring and summer, due to the suppression of homogeneous nucleation by intense dust emissions. Saharan dust emissions were forecasted and reported during the campaign, but their influence on cirrus formation have not been investigated in this work.

Greater differences are found among the in situ origin cirrus. While in situ origin ML cirrus have higher N with smaller ED, mainly between 70 and 250 µm, in situ origin HL cirrus are mostly composed of low number concentrations that have large particles with ED between 150 and 450 µm. As for the ED medians, the values of liquid origin cirrus change from 188 µm at mid-latitudes to 220 at high latitudes, while the in situ origin cirrus change from 128 at mid-latitudes to 189 µm at high latitudes. We also draw the reader's attention to a feature in the bottom-right square of Fig. 7c of in situ origin ML cirrus, which differs from the in situ origin HL cirrus observations. Although we did not quantify the contribution, we identified high concentrations of small particles connected to young contrails, aged contrails, contrail cirrus, or embedded contrails in natural cirrus. This feature is also weakly present in Fig. 7d of liquid origin ML cirrus but substantially less pronounced. We discuss the aviation influence in more detail in Sect. 5.2.

Our previous analysis in Sect. 4.1 already suggested that ML cirrus are more affected by contrail cirrus. As in Li et al. (2022), we observe that in situ origin cirrus are more strongly influenced by contrails than liquid origin clouds. Liquid origin clouds are formed at lower altitudes, where liquid droplets and ice crystals can coexist. Ascending motion, isobaric mixing, or radiative cooling can reduce the temperature in the cloud with a subsequent increase in relative humidity. The Wegener–Bergeron–Findeisen (WBF) process leads to the growth of the ice crystals by water vapour uptake from the evaporated liquid droplets and finally to the complete glaciation of the cloud (Korolev, 2007; Costa et al., 2017). The evolution of the liquid origin cirrus has sufficient time to consume the available water vapour when the cloud reaches the cirrus regime. Li et al. (2022) found that the Schmidt–Appleman criterion for contrail formation (SAC) is mostly not fulfilled in liquid origin clouds due to the warmer temperatures at which they are located. Therefore, it is more difficult to find the appropriate conditions for a contrail to develop within a liquid origin cloud.

On the contrary, in situ origin cirrus form directly at high altitudes, where air traffic is also present. If we imagine a scenario in which an in situ origin cloud starts forming heterogeneously by the deposition of water vapour on INPs and an aircraft exhaust jet is introduced, then the natural ice particles to be formed would be substituted by the freshly formed contrail that will severely affect the resulting microphysical properties. Contrail formation within cirrus can increase the cirrus ice crystal number by a few orders of magnitude (Schröder et al., 2000; Voigt et al., 2017; Schumann et al., 2017), in particular if the pre-existing cirrus has high ice water content and the supersaturation is low (Verma and Burkhardt, 2022). In this sense, we can expect an influence of contrail formation and growth on the formation process of in situ origin cirrus at cruise altitudes by an increase in the N and reduction in ED.

Additional information on the properties of the cirrus groups is given by their particle size distributions (PSDs). Knowing the ice crystal size distribution is vital, in order to derive the radiative impact of cirrus. The representation of PSDs of the four groups is not a trivial task in this case. We usually find, in other publications, the mean PSD being representative of a considered data set. However, our campaign deals with a wide variety of situations, including outliers, which were in fact infrequently observed but that strongly dominate the mean. These events are, in our case, contrails and convection encounters. This implies that the mean PSD fails to identify the averaged behaviour in the group, and the use of medians appears to be a better choice for this case. However, due to the low concentrations in cirrus and the lower sampling efficiency of the CDP compared to the CIPgs and PIP, the lower bins are frequently empty in the 2 s sampling rate and do not allow a median value calculation. Therefore, we increase the averaging intervals to 180 s to calculate the median concentration in each size bin. Sensitivity analyses were performed to select the 180 s averaging interval.

The calculated PSD median values are shown for each cirrus group in Fig. 8a and the means in Fig. 8b are also included for comparison. The median concentrations in each bin of the liquid origin cirrus are higher than for the in situ origin cirrus for particles >20 µm at both mid- and high latitudes, and the former covers a larger size range than the latter, which is consistent with previous observations (Krämer et al., 2016; Luebke et al., 2016). The median PSDs of liquid and in situ origin ML cirrus have higher concentrations than HL cirrus. Both ML cirrus profiles have the same behaviour in the smaller bin range (2–100 µm), and they differ in the larger sizes. The liquid origin cirrus reveal a second mode, which is slightly less pronounced than at high latitudes, and the in situ origin ML profile decays fast without a second mode.

Figure 8(a) Median, (b) mean particle size distributions (PSD), and (c) extinction coefficient (β_ext) of the following groups: in situ origin HL cirrus (light blue), liquid origin HL cirrus (navy), in situ origin ML cirrus (light red), and liquid origin ML cirrus (dark red). Median PSDs are indicated with thick lines and means with thinner lines. The frequency of occurrence of β_ext was calculated in logarithmic bins and fitted with B splines.

Download

As for the mean values, we detect two clear outlying features. First, the size range in liquid origin ML cirrus is larger, which is a consequence of the isolated convective events that were encountered. Second, the large number concentrations of small crystals for the in situ origin ML cirrus clearly show that the microphysics of in situ origin ML cirrus were affected by contrails. The difference in the concentration of the lower bins (from 2 to 10 µm) between the median and the mean is about 2 orders of magnitude. Then, the mean concentration decays and almost converges with the median at about 40 µm. In general, the median PSDs exhibit lower concentrations than the means in all bins for the four groups, which is expected due to the right skewness of the lognormal distribution. The mean and median show that the larger the differences, the more outliers there are in the distribution. The PSDs of the in situ and liquid origin HL cirrus show a weaker influence of outliers, as the medians profiles do not differ much from the means.

In Fig. 8c, we show another way to represent the radiative impact of cirrus through β_ext, which is often used to assess the climate impact of contrails (Bräuer et al., 2021a; Voigt et al., 2021). For a certain IWC, the extinction is larger when the ice mass is distributed in smaller crystals (Unterstrasser and Gierens, 2010). We observe higher extinction coefficients for the liquid origin cirrus at HL and ML, with a slightly higher median for ML, when compared to the in situ origin cirrus. This is expected, due to the higher IWC of the liquid origin cirrus. The lowest extinction corresponds to the in situ origin HL cirrus, since the N is lower, and the ice particles are larger, which reduces the extinction and the radiative impact. In line with the PSDs, here we also see a stronger difference between the in situ cirrus at HL and ML with a shift to larger extinctions at ML. In addition, both frequency distributions of ML cirrus show a small mode at 1 km⁻¹, which is frequently observed in young contrails (Febvre et al., 2009; Voigt et al., 2011; Gayet et al., 2012; Bräuer et al., 2021a).

The increased humidity in the upper troposphere during the campaign period can contribute to more persistent contrails, which can be particularly warming (Wilhelm et al., 2022; Teoh et al., 2022; Wang et al., 2023). However, the potential contrail cirrus cover is the lowest during summer over Europe (Dischl et al., 2022). The increase in air temperature in summer hinders the formation and persistence of contrails, and their radiative effect is mostly cooling because of the increased Sun hours in summer in the North Atlantic corridor (Teoh et al., 2022). Our study confirms a low frequency of direct observations of contrail encounters. Even though they do not dominate the campaign observations, and therefore are not representative of the general picture, the mean PSD of in situ origin ML cirrus shows how strongly contrail encounters can affect the cirrus microphysics with an enhancement of smaller particles. More importantly, our analyses indicate that the higher number concentration and smaller particles indeed describe the microphysical properties of ML cirrus compared to lower concentrations and larger ice crystals at the more pristine high latitudes.

5.1 Influence of the method

The selection of the latitude threshold of 60^∘ N is a critical aspect in our study, as it determines the differentiation between mid- and high latitudes and, at the same time, our cirrus classification. We performed a sensitivity study and varied the threshold between 55 and 65^∘ N (see Fig. S4). Here, we show that the variation in the threshold does not alter the existence of the described differences between the three cirrus groups, even though it influences the absolute values of the medians. In addition, Fig. 2b shows a continuous tendency of increasing effective diameter from mid- to high latitudes, which shows that the differences between ML and HL cirrus are continuous and do not depend on the threshold.

The classification of cirrus origins of our in situ measurements is susceptible to uncertainties, as trajectories were initiated along the flight track every 10 s, which corresponds to ≈2 km in horizontal scale. We assume that cirrus measured in this interval have the same origin. In addition, the model parameters calculated along the trajectories are averages of grid points in a $\approx \mathrm{50}\mathrm{km}\times \mathrm{50}\mathrm{km}$ domain that does not capture the smallest local variations. We use the vertical motion along trajectories to separate processes connected to our measurements. However, our trajectory approach excludes the analysis of the impact of small-scale temperature and wind fluctuations on cirrus formation. The accuracy of the approach is difficult to assess, as the associated uncertainty is very inhomogeneous; it depends on the location and the meteorological situation. Nevertheless, we yield convincing results from our classification method, which are in line with previous publications (Krämer et al., 2016, 2020; Luebke et al., 2016; Wernli et al., 2016).

5.2 Aviation influence

Even though the conditions for contrail formation are not as favourable in summer compared to winter (Dischl et al., 2022), and even though the air traffic was still reduced in 2021 due to the COVID-19 pandemic (ICAO, 2022; Schumann et al., 2021; Voigt et al., 2022), we could identify an influence of contrail formation and interaction with natural cirrus in our data set. We did not focus on a methodical approach to differentiate contrails or contrail cirrus from natural cirrus (Li et al., 2022), but we considered nonnatural cirrus to have measurements with $N>\mathrm{0.1}{\mathrm{cm}}^{-\mathrm{3}}$ and ED>40 µm (lower right section of Fig. 7c). In general, we did not observe high updraft speeds during the campaign – neither from the trajectories nor from the in situ measurements (see Fig. S3). In particular, the events with high N and low ED were also not associated with high updraft speeds that could lead naturally to strong cooling rates and an outburst of small ice crystals. Additionally, we supported the analysis with flight reports and contrail cirrus predictions. With this information, we are able to indicate the existence of an aviation influence and identify this influence on the properties of the described cirrus groups.

5.3 Representativeness

This study analyses the differences in the properties of the cirrus that we measured during the CIRRUS-HL campaign according to the latitude. The results are representative of summer 2021; of the regions measured over the Arctic, North Atlantic, and central Europe; and of the meteorological situations observed during the campaign. In this context, our cirrus observations cover a wide range of latitudes between 38 and 76^∘ N and temperatures between −38 and −63 ^∘C. The statistical tests we performed indicate a sufficient statistical significance of the differences that we found in these ranges, but conclusions drawn on the coldest temperatures (between −58 and −63 ^∘C for HL cirrus and −60.5 and −63 ^∘C for ML cirrus) have to be viewed with caution, as the number of observations is limited, and the samples are mainly from the same cloud sequence. This implies that the observations are correlated to some extent, as they are measured at approximately the same temperature and height. However, each 2 s sample is horizontally separated by ≈800 m, and thus, the cloud properties are not necessarily connected. In any case, our measurements can be combined with future observations to build a larger data set and gain a broader perspective.