An extensive data set for in situ microphysical characterization of low-level clouds in a Finnish sub-Arctic site

Continuous, semi-long-term, ground-based in situ cloud measurements were conducted during eight Pallas Cloud Experiments (PaCEs) held in autumn between 2004 and 2019. Those campaigns were carried out in the Finnish sub-Arctic region at the Sammaltunturi station (675824 N, 240658 E; 560 m a.m.s.l.), the part of the Pallas Atmosphere–Ecosystem Supersite and Global Atmosphere Watch (GAW) program. Two cloud spectrometer ground setups and a weather station were installed on the roof of the station to measure in situ cloud properties and several meteorological variables. Thus, the obtained data sets include the size distribution of cloud droplets as a measured cloud parameter along with the air temperature, dew point temperature, humidity, pressure, horizontal wind speed and direction, (global solar) sun radiation, and visibility at the station. Additionally, the number concentration, effective diameter, median volume diameter, and liquid water content from each instrument were derived. The presented data sets provide a insight into microphysics of low-level clouds in subArctic conditions over a wide range of temperatures (−25.8 to 8.8 C). The data are available in the Finnish Meteorological Institute (FMI) open data repository for each campaign and each cloud spectrometer ground setup individually: https://doi.org/10.23728/FMI-B2SHARE.988739D21B824C709084E88ED6C6D54B (Doulgeris et al., 2021).


Introduction
Clouds are considered a major component of both the climate system and the hydrological cycle. Nevertheless, our level of understanding of the fundamental details of the cloud microphysical processes is still very limited (Boucher et al., 2013). To gain a deeper knowledge of the formation and development of the clouds, more in situ measurements are needed . In addition, a correct representation of cloud microphysics in general circulation models for numerical weather and climate prediction is of great importance (Guichard and Couvrex, 2017;Morrison et al., 2020). Despite the fact that cloud processes can now be studied with much more confidence (Bony et al., 2015), representing the formation and evolution of cloud droplets and the effects of aerosols on clouds at various meteorological conditions remains a challenge (Grabowski et al., 2019). The number concentration and size distribution of cloud droplets are considered key parameters for a quantitative microphysical description of clouds (e.g., Rosenfeld and Ulbrich, 2003;Komppula et al., 2005;Lihavainen et al., 2008;Pruppacher and Klett, 2010;Chang et al., 2019) and are connected with the cloud lifetime and radiative effects as well as precipitation (e.g., Albrecht, 1989;Devenish et al., 2012;McFarquhar et al., 2020).
Data sets that have been obtained from measurements in sub-Arctic clouds are significant as cloud processes are of high value since cloud processes are considered an important component of climate change in the Arctic region (Wendisch et al., 2019). Pallas Cloud Experiments (PaCEs) took place in the Finnish sub-Arctic. The main objective during PaCE was to study low-level clouds and their microphysical properties in a background sub-Arctic environment. In this work, we present a unique data set of ground in situ cloud measurements along with several meteorological variables collected at the Sammaltunturi station in eight autumn campaigns conducted between 2004 and 2019. This data set can be used in studies of cloud microphysics, climate change in the sub-Arctic, and performance evaluation and improvement of existing models, in particular at higher altitudes. In the next section, we provide a description of the sampling location, instrumentation, and the measurement methodology we used for sampling, data processing, and quality control.

Measurement site and PaCE campaigns overview
The Sammaltunturi station (67 • 58 24 N, 24 • 06 58 E) is hosted by the Finnish Meteorological Institute (FMI) and is located on a top of an arctic fell (560 m a.m.s.l.) in the Finnish sub-Arctic region inside the Pallas-Yllästunturi National Park (Fig. 1). The Pallas area is located around 180 km above the Arctic circle, and it has no significant local or regional air pollution sources. Thus, the Sammaltunturi station provides an excellent location for the monitoring of background air composition in northern Europe. The station is about 100 m above the tree line, and the vegetation around it consists mainly of low vascular plants, mosses, and lichen. There is a long history of atmospheric data collection in the area (see Lohila et al., 2015). Monitoring activities of atmospheric composition at Sammaltunturi started in 1991 in a building that originally served the Finnish Broadcasting Company. The new station (102 m 2 ) opened in July 2001. Since 1994, Sammaltunturi has been established as a node of the Pallas-Sodankylä supersite that contributes to the GAW program of the World Meteorological Organization. The site was described in detail in Hatakka et al. (2003). The main research measurements focus on greenhouse gas concentration, climate effects of atmospheric aerosols, aerosol cloud interaction, and air quality (e.g., Komppula et al., 2005;Lihavainen et al., 2008;Asmi et al., 2011;Backman et al., 2017;Doulgeris et al., 2020). The predominant origin of air masses arriving at Sammaltunturi is from the Arctic (Asmi et al., 2011).
The main motivation to perform in situ cloud measurements at the Sammaltunturi was that the station was occasionally immersed in a cloud. Based on analytical data the most suitable time of the year for in situ cloud measurements was autumn when the horizontal visibility drops below 1 km around 40 % of the time (Hatakka et al., 2003). Once the preferable time of the year was identified, we started to conduct ground-based in situ measurements and study cloud formation. The Pallas Cloud Experiments were, usually, 6-8 weeks long and lasted approximately from the beginning of September until the end of November, occasionally extended to the beginning of December. The first attempt at measuring in situ cloud properties was made in 2004 using the forwardscattering spectrometer probe (FSSP-100) ground setup that was the only available cloud spectrometer at that time. The next campaigns, in 2005 and 2009, were done using the same instrument setup (Lihavainen et al., 2008). Later, in 2011 the cloud, aerosol, and precipitation spectrometer (CAPS) ground setup was added. In January and February 2012, it was tested for the first time for two short periods during winter at the Sammaltunturi site. In 2012, 2013, and 2015 both instruments were installed and used during PaCE . In 2017 and 2019, only CAPS was used (Girdwood et al., 2020). An overview of each year's campaign duration and the cloud spectrometer ground setups' availability is presented in Fig. 2. Instruments that were used for measuring the meteorological variables and the solar radiation were operating continuously during all PaCE years. The instrumentation used during PaCE campaigns is described in detail in the following section.

Instrumentation
In order to monitor meteorological variables, the station was equipped with an automatic weather station (Milos 500, Vaisala Inc.). A weather sensor (model FD12P, Vaisala Inc.) was used for measuring the horizontal visibility, the Vaisala HUMICAP was used for measuring the relative humidity, BAROCAP sensors were used for measuring the barometric pressure, and PT100 sensors were used to measure temperature at 570 m. Global radiation and photosynthetically active radiation were measured with a pyranometer and a photo- voltaic detector, respectively. Additionally, the wind speed was measured with a heated cup and the wind direction with a heated wind vane. All the above meteorological variables were saved as 1 min averages. A detailed description of the weather sensors can be found in Hatakka et al. (2003).
In order to conduct in situ cloud ground-based measurements, we deployed two instruments. The cloud, aerosol, and precipitation spectrometer (CAPS) and the forwardscattering spectrometer probe (FSSP-100; Droplet Measurement Technologies (DMT), Boulder, CO, USA) (Fig. 3). The FSSP (model SPP-100, DMT) was originally manufactured by Particle Measuring Systems (PMS Inc., Boulder CO, USA). Both instruments were originally developed for airborne measurements but modified as ground setups by the manufacturer (DMT, USA). They were installed on the rooftop of the Sammaltunturi station. The CAPS was fixed with a heading always to the main wind direction of the station southwest, ∼ 225 • , while the FSSP-100 was installed on a rotating platform to continuously face the wind. The CAPS had a total height of 0.6 m above the roof where it was in-stalled and a height of 4.5 m from the ground. FSSP had a total height of 0.6 m above the roof where it was installed and a height of 5.5 m from the ground. The two setups had a horizontal distance of ∼ 10 m and vertical distance of ∼ 1 m between them. From 2004 until 2012 a flow laminator was used inside the FSSP inlet (Lihavainen et al., 2008). However, the flow laminator was often blocked by freezing or supercooled cloud droplets at sub-zero temperatures, and for this reason it was cleaned every hour if occurrence of subcooled water was detected. The laminator blockage was evident both during everyday instrument inspection and from the raw data. Only data cleaned of this artifact were used in the FSSP data set. However, even without placing the laminator, the Reynolds number indicated that the flow inside the inlet was still laminar. As a result, in 2012 we decided that the laminator would not be used in the FSSP setup anymore. Thus, the number of data after 2012 were more extensive, and the number of cases when the FSSP would have been blocked was significantly reduced. A detailed description of both ground setups and the methodology we used for obtain- ing the ground-based cloud microphysical properties with in situ method was documented in Doulgeris et al. (2020). Only a short overview is given here.
The CAPS has been widely used in airborne measurements of the microphysical properties in clouds (e.g., Baumgardner et al., 2001Baumgardner et al., , 2011Droplet Measurement Technologies Manual, 2011;Lachlan-Cope et al., 2016). The CAPS probe includes three instruments: the cloud and aerosol spectrometer (CAS) which measures smaller particles, the cloud imaging probe (CIP), and the hot-wire liquid water content (LWC hw ) sensor. For the ground setup we deployed, the hot-wire LWC faced difficulties operating in such extreme conditions; after operating in supercooled liquid clouds (even for a short time) the sensor was accreting ice. In addition, the lifetime of the sensor is limited and significantly shorter than the duration of the campaign. The FSSP-100 was widely used for measuring droplet size distribution (e.g., Brenguier, 1989;Lihavainen et al., 2008;Lloyd et al., 2015;Doulgeris et al., 2020). CAS and FSSP-100 derive the size of the particle from the intensity of the scattered light, using the Mie theory (Mie, 1908). Furthermore, backscatter optics measure light intensity in the 168 to 176 • range. This allows the determination of the real component of a particle's refractive index for spherical particles. The CIP is a single particle optical array probe. Its design is based on optical measurement techniques whereby single particles pass through a collimated laser beam and their shadow is projected onto a linear array of 64 photodetectors. The count of the particle is dependent on the change in the light intensity of each diode.
All the instruments were calibrated before and after each campaign. Until 2011, we relied on the manufacturer calibration that was done at DMT. After 2011, we also started to perform calibration at the FMI, on top of manufacturer calibration, to ensure the quality of the collected data. For the calibration of the CAS and FSSP-100, glass beads in the diameter size range 2-40 µm and polystyrene latex sphere (PSL) standards in the diameter size range 0.74-2 µm were used. Cloud spectrometers (in our case CAS and FSSP-100) are calibrated for size measurements but not for number concentration measurements. The instruments faced extreme conditions during the whole campaign, in terms of frequent changes in wind direction, wind speed and sub-zero temperatures. Despite the calibration procedures we should always keep in mind that extreme meteorological conditions could possibly lead to unexpected performance. To calibrate the CIP, a spinning glass disk with opaque dots of known size was used.
The CAPS ground setup included a high-flow pump (Baldor, Reliance, USA) which worked as an aspiration system. The aspiration system was made and provided by the manufacturer (DMT). A custom aspiration system with a high-flow ventilator was also made by the manufacturer (PMS) and employed through FSSP-100 inlet to ensure constant flow through it. A digital thermo-anemometer (model 471, Dwyer Inc.) was used in each campaign for checks of daily cloud spectrometers' air speed. The FSSP air speed inside the inlet was calculated from the measured air speed in front of the inlet, except in 2004 and 2005 when the air speed was calculated with measured volume flow rate through the in- let. A necking inside the inlet led the flow from inner diameter 3.8 to 2.0 cm. Both spectrometers were equipped with anti-ice systems as they were modified by the manufacturers (DMT for CAPS and PMS for FSSP-100) for ground-based use. Despite the existing anti-ice features, due to the subzero temperatures that they were facing, snow or ice could accrete and affect the air speed inside the probe inlets. For this reason, to ensure the proper operation of the instruments, they were inspected and cleaned twice per day, every morning and evening (approximately every 12 h).
The ground-based in situ cloud measurements provided the cloud and precipitation size distribution. The PADS 2.5.6 software that was used for the data acquisition of CAPS measurements (Droplet Measurement Technologies Manual, 2009) provided the number concentration (N c , cm −3 ), liquid water content (LWC, g cm −3 ), median volume diameter (MVD, µm), and effective diameter (ED, µm). For the FSSP-100, N c , LWC, MVD, and ED were also derived using the same equations , since we used an older software for data acquisition (PACS 2.2, DMT).
The major sources of uncertainties of the cloud spectrometers can be coincidence, dead-time losses, and changing velocity ratio (Guyot et al., 2015). The uncertainty of estimation of sizing at the cloud spectrometers was 20 % and that of the number concentration was 16 % (Baumgardner, 1983;Dye and Baumgardner, 1984;Baumgardner et al., 2017). According to Lance (2012), it was observed that for CAS at ambient droplet concentrations of 500 cm −3 there was 27 % undercounting and a 20 %-30 % oversizing bias. In our case, during PaCE campaigns the droplet number concentration values we monitored were in the majority of cases less than 300 cm −3 . These number concentration values lead us not to take coincidence, dead-time losses, and velocity acceptance ratio (VAR) uncertainties into consideration in this analysis. LWC has a significant uncertainty of 40 % (Droplet Measurement Technologies Manual, 2009). The FSSP-derived ED and LWC had an uncertainty of 3 µm and 30 % in mixedphase clouds (Febvre et al., 2012). An overview of the instrumentation and their operational characteristics we used for cloud measurements are summarized in Table 1.

Overview of data set and quality control description
The current data set contains only in-cloud measurements when the station was immersed in a cloud. Data from each cloud probe and the weather station were quality controlled and unified in a common format for release and further analysis. The presence of a cloud at the station was identified with three different factors. First, we checked the droplet size distribution measured in both of the cloud spectrometers. This was the main parameter to consider that the station was inside a cloud. Then, to confirm this assumption, we cross-checked the droplets counts with two meteorological variables -the relative humidity at the measurement site which was expected to be ∼ 100 % and the horizontal visibility which should be less than 1 km, when the Sammaltunturi station is in the cloud. In the event that one of the factors was not fulfilled, a final inspection was done visually using pictures recorded by an automatic weather camera installed on the roof of the station. During PaCE 2004 and 2005 the sampling time of the FSSP-100 was 15 s. During PaCE 2009 the instrument was set to sample at 10 s. From 2009 until 2019 the sampling time was set to sample each 1 s (1 Hz) for both instruments. The PT100 sensor, Vaisala HUMICAP and BAROCAP sensors, the pyranometer, and the heated cup and wind vane were also set to sample to 1 s. The FD12P Vaisala weather sensor sampling time was 15 s. For every year, 1 min averages were calculated for each cloud spectrometer and each meteorological variable. As a result, we obtained the cloud droplet size distribution and several meteorological variables for each minute and as derived parameters the N c (cm −3 ), LWC (g cm −3 ), MVD (µm), and ED (µm). All data sets were converted to NetCDF format. All times in this work are given in UTC time. Our data set includes a separate NetCDF and CVS file for each cloud spectrometer and for each year under the file name PACE.yyyy.cloud_spectrometer.nc and PACE.yyyy.cloud_spectrometer.cvs (example names). For every file, the sampling area (mm 2 ) and the probe air speed (ms −1 ) that were used to derive each parameter are provided. In addition, it includes the cleaned timeline data set of the fol-  ±10 %, 10-10 000 m ±20 %, 10 000-50 000 m lowing cloud properties and meteorological variables: year (YYYY), day (DD), month (MM), hour (HH), min (MN), size bin lower limit, size bin higher limit, number concentration (cm −3 ), liquid water content (g cm −3 ), effective diameter (µm), median volume diameter (µm), the calculated MSD (cm −3 ) values in each bin, temperature at 570 m ( • C), dew point ( • C), humidity at 570 m (%), pressure (hPa), wind speed (m s −1 ), horizontal wind direction (degrees), global solar radiation (W m −2 ), photosynthetically active radiation (µmol m −2 s −1 ), and the horizontal visibility (m). The derived cloud parameters -number concentration (cm −3 ), liquid water content (g cm −3 ), effective diameter (µm), and median volume diameter (µm) -were not included in the CIP files. The number of cloud droplets per minute in CIP size range leads to statistically biased values, and for this reason we decided to exclude them. The variables, naming abbreviations, and units are summarized in Table 2.
As it is shown in Fig. 4, the observation hours after PaCE 2013 when the campaigns had longer duration are significantly higher. The number of data in these years is excessive, serving as an important source of information for Arctic studies. An overview of meteorological variables is presented for each campaign when the FSSP-100 and CAPS ground setups were operational. In Fig. 5, a statistical description of the temperature at 570 m a.m.s.l. for each campaign is illustrated.
Each PaCE year the temperature trends and ranges were similar (around −10.0 to 8 • C). In Fig. 6, we show the percentage of the data set for each year in which the global solar radiation was higher than 0. It was used to estimate the number of data collected in each campaign in daylight. In addition, an overview of the microphysical derived cloud properties data from each campaign is presented. Thus, in Figs. 7, 8, and 9, the number concentration, the effective diameter, the medium volume diameter, and the liquid water content are presented for each campaign and for the FSSP-100 and CAS ground setups, respectively. Number concentration averaged values were similar for every year of the measurements and reach scales around 100 cm −3 . However, there were some cloud cases during each campaign that number concentration had values around 300 cm −3 . The averaged ED and MVD values were ranging approximately from 10 to 20 µm. The liquid water content was less than 0.2 g cm −3 in most cases.

Code availability
Software developed to process and display the data from the cloud ground-based spectrometers is not publicly available and leverages licensed data analysis software (MATLAB). This software contains intellectual property that is not meant for public dissemination.

Data availability
Each described data set was collected by Finnish Meteorological Institute during PaCE campaigns and was published in the described form at the FMI open data repository. All data sets have undergone thorough quality control, and false readings were eliminated. Data sets can be all found at https://doi.org/10.23728/FMI-B2SHARE. 988739D21B824C709084E88ED6C6D54B (Doulgeris et   al., 2021). When the CIP was operational, we also collected the CIP images. However, we did not include the raw images in the data set for two reasons. First, they were in binary format. To read them, we used a proprietary image analysis software that was provided by DMT. Secondly, the upper limit of the open data repository is 10 GB, which was not enough to include the CIP raw images which were approximately 0.5 GB per case per day. However, RAW CIP images could be provided by the authors upon request.

Summary
In this study we produced and summarized data sets obtained from two cloud ground-based spectrometers (CAPS and FSSP-100 ground setups) owned by the FMI during 8 years of PaCE campaigns conducted in autumn from 2004 until 2019 along with several meteorological variables. PaCE campaigns took place in the Finnish sub-Arctic region in a clear environment in temperatures that were usually below zero. In Sect. 2, we describe the measuring site where PaCE campaigns took place and the cloud ground spectrometers  setups that were used to obtain the cloud data along with the instrumentation that was used to monitor the weather conditions. In Sect. 3 an overview of the data set is presented. These observations gathered in sub-Arctic conditions are a unique source of in situ cloud measurements, which can contribute to the understanding of the cloud dynamics and formation in a sub-Arctic environment in different meteorological conditions. Such semi-long observations are difficult to obtain in similar environments due to current lack of instrumentation which would allow continuous unattended operation at temperatures below 0 • C. Cloud droplet spectrometers with surface installation has been identified as a  potential method for continuous cloud in situ measurements (Wandinger et al., 2018). Thus, due to the increased demand for long-term continuous ground-based in situ cloud measurements, we provide a data set of in situ cloud measurements in a harsh sub-Arctic environment. Each data set includes a combination of cloud microphysical properties along with several meteorological variables. Even though the data set includes measurements from eight campaigns, we would propose a case-by-case cloud investigation. Due to the inhomogeneity of the presented cloud cases, it is challenging to retrieve any trend that can be unambiguously connected to changes in the atmosphere. Also, the quality of data set may differ for each campaign due to the different number of observations per year and operators' experience running the ground-based spectrometers through the years. In addition, each cloud case could be of different mass origin. We therefore discourage from any trend analysis based only on the presented data set. At least thorough back-trajectory analysis and subsequent segregation of data set according to air mass origin is recommended. However, this was not an objective of this paper. The data set in current form provides a helpful contribution to cloud microphysics processes on shorter timescales. Microphysical processes can strongly influence cloud-climate feedbacks in global climate models (Bodas-Salcedo et al., 2019). Furthermore, it can be used as complementary in model development. Representation of cloud microphysics is considered significant for large eddy simulation (LES) models (Morrison et al., 2020). There is a need for in situ cloud data sets due to two significant problems that the modeling community is facing: the representation of the population of the cloud and precipitation particles and the uncertainties due to fundamental gaps in knowledge of cloud physics (Morrison et al., 2020). In this data set, the cloud size distribution was monitored in different stages of its evolution. Competing interests. The contact author has declared that neither they nor their co-authors have any competing interests.

Appendix A: Abbreviations
Disclaimer. Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.