Measurement report: Introduction to the HyICE-2018 campaign for measurements of ice nucleating particles in the Hyytiälä boreal forest

The formation of ice particles in Earth’s atmosphere strongly influences the dynamics and optical properties of clouds and their impacts on the climate system. Ice formation in clouds is often triggered heterogeneously by ice nucleating particles (INPs) that represent a very low number of particles in the atmosphere. To date, many sources of INPs, such as mineral and soil dust, have been investigated and identified in the lower latitudes. Although less is known about the sources of ice nucleation at higher latitudes, efforts have been made to identify the sources of INPs in the Arctic and boreal environments. In this study, we 5 investigate the INP emission potential from high latitude boreal forests. We introduce the HyICE-2018 measurement campaign conducted in the boreal forest of Hyytiälä, Finland between February and June 2018. The campaign utilized the infrastructure 1 https://doi.org/10.5194/acp-2021-744 Preprint. Discussion started: 8 October 2021 c © Author(s) 2021. CC BY 4.0 License.

for instrument inter-comparison analyses to test the reproducibility of results across instruments and scientific teams in the field setting.
The primary objectives of the campaign were to • quantify and characterize INPs in a boreal environment within different thermodynamic forcing regimes (i.e., different 25 temperatures, T , and supersaturations with respect to ice, S i ) • determine and quantify the existence of any seasonal variation of INPs and • assess the vertical distribution of INPs above the boreal forest. This paper gives an overview of the campaign setting and design as an introduction to the Copernicus Special Issue, "Ice nucleation in the boreal atmosphere", which is anticipated to include several contributions from HyICE-2018. Data from 30 several days dedicated to evaluate instrument inter-comparison is presented to illustrate instrument-to-instrument agreement and to facilitate future presentation and interpretation of data from subsets of instruments.

Measurement Site -SMEAR II
The HyICE-2018 campaign took place at the SMEAR II station in Hyytiälä, Finland (Hari and Kulmala, 2005). The station is located in Southern Finland (61 • 51'N, 24 • 17'E; 181 m a.s.l.) and is surrounded by boreal coniferous forest (Fig. 1). The conditions at the site are typical for a background location, with the main pollution sources being the city of Tampere (60 km 5 to the southwest with ≈ 238, 000 inhabitants as of 2019) and the activity and buildings at the station Boy et al., 2004). The station has several operational units that span a wide forest area and reach into and above the tree canopy. Multiple towers include a 128 m mast used for atmospheric flux measurements, a 18 m tower for irradiation and flux measurements, a separate 18 m tower for tree physiology measurements and a 35 m walk-up tower for aerosol measurements.
There are also several measurement cottages and containers spread throughout the forest, with a cottage located near the walk-   The SMEAR II station is equipped to monitor the physical and chemical properties of aerosols and gas phase precursors to aerosol formation with a suite of state-of-the-art monitoring instrumentation. Measurements also cover meteorology, radiation, soil, snow cover and gases. An overview of the instruments in operation at the site during the HyICE-2018 campaign is available in the supplemental materials (Tables A1,A2 and A3). In this study, data from the SMEAR II Differential Mobility Particle Sizer (DMPS) and Aerosol Particle Sizer (APS; TSI model 3321) were used. The DMPS measures aerosol particle 5 size distributions from 3 to 1000 nm in mobility diameter, with a 10 min time resolution Jokinen and Mäkelä, 1997). During HyICE-2018, the instrument was sampling through a total suspended particle inlet 8 m inside the forest canopy and was operated following the guidelines from Aerosols, Clouds and Trace gases Research InfraStructure (ACTRIS; Wiedensohler et al. 2012). The APS is used to measure the super-micron aerosol particle size distribution from 0.5 to 20 µm in aerodynamic diameter. The instrument was sampling through a total suspended particle inlet (DIGITEL Elektronik GmbH) 6 10 m above ground level, and a vertical sampling line was used to avoid particle losses. In addition, the APS inlet was heated to 40 • C to ensure that the relative humidity in the sampling inlet remained below 40%, which prevents condensation and dries the aerosols before measurements.
The instruments deployed specifically for HyICE-2018 are summarized in Table 1, and the dates they were operational are depicted in Fig. 2 against a background of the changing seasons illustrated by the evolving average daytime temperatures 15 T day (mean surface temperature 08.00-20.00, UTC+2). The campaign-specific instrumentation combined with the measurements from SMEAR II provides a rich dataset for the analysis of INPs. Additionally, the large parameter space allows for advanced machine learning techniques to be applied, where the parameter space dimensionality is reduced to illuminate pro-

Augmented sampling for Ice Nucleating Particles
One of the motivations behind HyICE-2018 was to compare different INP measurement techniques in a field setting. Previous studies have performed inter-comparisons of INP instrumentation in a number of intensive measurement campaigns (DeMott et al., 2011;Wex et al., 2015;Hiranuma et al., 2015;Burkert-Kohn et al., 2017;DeMott et al., 2018;Hiranuma et al., 2019).
However, these efforts focused on well-controlled laboratory measurements to assess sampling procedures and to calibrate 5 instruments relative to one another (Hiranuma et al., 2015;Wex et al., 2015). Only a few INP instruments have been co-located for long field measurements of real atmospheric aerosol (DeMott et al., 2017), and continued efforts in multi-instrument measurements can be beneficial for the entire community (Lacher et al., 2020

Continuous Flow Diffusion Chambers
Three CFDCs with parallel-plate designs (PINC, PINCii, SPIN) were deployed for online INP measurements (Fig. 3a). The 15 three models are iterations of a design that consists of two parallel ice-coated walls which are cooled to below freezing temperatures (Rogers, 1988;Stetzer et al., 2008;Chou et al., 2011). In the upper part of the chamber, referred to as the main chamber,  The PINE expansion chamber samples a volume of ambient air before a short expansion cycle activates INP within the trapped volume. In both (a) and (b), optical particle counting is used at the chamber exits to measure particle number. (c) The droplet freezing assays rely on filter sampling of ambient aerosols. Collected particles are washed from the filters and analyzed for INP content, generating curves of INP temperature spectra. (d) The thermodynamic space typically accessible to different measurement techniques. the walls are held at different temperatures ( Fig. 3a), to produce vapor supersaturation in the region between the walls (Rogers, 1988). The lower chamber, or evaporation section, is held at an isothermal condition of ice saturation. During measurements, a continuous flow of sample air, referred to as the sample lamina, is sandwiched between two particle-free sheath flows and is drawn through the center of the chamber, thus exposing any airborne particles to the supersaturated conditions. At a given lamina temperature, T l , and saturation condition (e.g., supersaturation with respect to ice, S i ) some particles will induce water 5 condensation and/or ice formation by accommodating excess vapor. When the flow leaves the main supersaturated chamber, it immediately enters an evaporation section, which is held at ice saturation, and is thus sub-saturated with respect to liquid water. Within the evaporation section, liquid droplets evaporate, creating a size difference between frozen and unfrozen particles. When paired with an inlet size cutoff, particles within the exit flow that exceed the cutoff are determined to be ice. Above a certain S i , CFDCs have an instrument-specific point of "droplet breakthrough" where the residence time of the evapora-10 tion section does not enable adequate evaporation for phase differentiation based on cut-off size with standard optical particle 7 https://doi.org /10.5194/acp-2021-744 Preprint. Discussion started: 8 October 2021 c Author(s) 2021. CC BY 4.0 License.
counting. During HyICE-2018, the CFDCs were operated at relative humidities below the point of droplet breakthrough in order to prevent such a situation.
CFDCs are online instruments that measure INP concentration in real time with a minimum time resolution determined by the instrument specific particle counting method. However, since ambient INP concentrations are generally low, measurements typically consist of multi-minute counting averages. For the CFDCs used in this study, sampling intervals of 5 to 20 minutes 5 were separated by background measurements of clean, filtered air. Overall, CFDC measurements are time-limited by the quality of the thin ice layer coating the chamber walls, which deteriorates over time and contributes to increasing particle counts as the instruments operate. Single experiments typically last three to five hours, after which the ice coating needs to be regenerated to restore low background conditions. This is done by warming and purging the chamber, before re-cooling and re-coating the walls with ice. This process can last one to three hours, allowing for two to four daily measurement cycles if continuous 10 operation is desired (Paramonov et al., 2020). Concentrations obtained during background measurements are subtracted from the concentrations measured during each sampling window to compute the measured INP concentrations, where the lower limit of detection is defined as one standard error above the background mean. In the following sections we describe the instrument specifics for each CFDC run during HyICE-2018.

15
The Portable Ice Nucleation Chamber (PINC) was the first generation field-deployable, parallel-plate CFDC based on the designs of Stetzer et al. (2008), and it has been operated in many locations around the world for more than a decade (Chou et al., 2011(Chou et al., , 2013Boose et al., 2016;Kanji et al., 2019). During HyICE-2018, PINC was operated from February 19 to April 2 ( Fig. 2) in the main aerosol cottage (Figs. 1 & 4), with one to two experimental cycles conducted per day, always during daytime. For the duration of the campaign PINC was operated at a fixed lamina temperature of T l = -31 • C and a relative 20 humidity with respect to water RH w = 105%. These conditions were selected to simulate mixed-phase cloud conditions and correspond to the condensation/immersion freezing mode(s) of ice nucleation (Vali et al., 2015). Sampling was performed from a total aerosol inlet mounted outside of the building, 6 m above ground level (Fig. 4). The inlet was heated to 25-30 • C to evaporate droplets and ice crystals and sampled ambient air with a flow rate of 250 L min −1 . From the large inlet flow, a smaller 4 L min −1 sample flow was extracted, and a cyclone was used to eliminate particles larger than 2.5 µm. A molecular 25 sieve dryer was installed to further reduce the sample relative humidity. After the dryer, the sample flow was split into four: 1 L min −1 to a Condensation Particle Counter (CPC; TSI model 3010), 1 L min −1 to an Aerodynamic Particle Sizer (APS; TSI model 3321), 1 L min −1 to a Scanning Mobility Particle Sizer (SMPS; with one Hauke-type differential mobility analyzer (DMA) and one TSI model 3772 CPC) and 1 L min −1 to PINC (Paramonov et al., 2020).
On several occasions, a Portable Fine Particle Concentrator (PFPC) was used to concentrate aerosol particles in the sample 30 flow upstream of PINC (Fig. 4). The PFPC, described by Gute et al. (2019) and based on the design by Sioutas et al. (1995), is a multi-stage concentrator based on virtual impaction. It concentrates aerosol particles with a certain size-dependent enrichment factor where larger particles are concentrated more efficiently than smaller ones (Gute et al., 2019). For HyICE-2018, the size-dependent enrichment factor was determined by measuring the particle size distributions before and after the PFPC, as When the PFPC was used, the 2.5 µm cyclone was removed and replaced by the 2.5 µm impactor located inside the PFPC (see Fig. 4). On March 22, for inter-comparison purposes, the cyclone was removed so that PINC directly sampled the inlet air, and the operating conditions were changed to T l = -29 • C and RH w = 105%.

5
In this article we summarize the performance of PINC during the targeted days of instrument inter-comparison, while the longer-term results from the PINC measurements during HyICE-2018 are presented and discussed by Paramonov et al. (2020).

II. Portable Ice Nucleation Chamber II -PINCii
The Portable Ice Nucleation Chamber II (PINCii) is a parallel-plate CFDC developed as an upgrade to the PINC instrument.
Although many specific engineering details differ, the primary differences between PINC and PINCii are the chamber dimen-10 sions and cooling power. While PINC's main chamber and evaporation section are 568 and 230 mm in height respectively (Chou et al., 2011), PINCii is approximately twice as large with a main chamber of 1000 mm and an evaporation section of 440 mm. A manuscript outlining the engineering and experimental/operational details of PINCii is in preparation .
During the campaign, PINCii was operational from April 22 to June 10, 2018 and measured INP concentrations at a fixed 15 T l =-32 • C and RH w = 105% in order to generate results comparable with the earlier PINC measurements. PINCii was located in the main cottage and sampled from the heated total aerosol inlet, essentially acting as a substitute for the early PINC measurements as depicted in Fig. 4. However, during this later sampling interval, the PFPC was not used and the additional SMPS was used for the first two weeks of June only. For the typical setup during the campaign, PINCii sampled downstream of the 2.5 µm cyclone.

III. Spectrometer for Ice Nuclei -SPIN
The Spectrometer for Ice Nuclei (SPIN, DMT) is a commercially developed CFDC based on the parallel-plate PINC design.

5
The design and use of SPIN has been previously documented (Garimella et al., 2016(Garimella et al., , 2017(Garimella et al., , 2018 and some SPIN instruments   The use of the VI-concentrator may have generated biases in the SPIN measurements. Indeed, the magnification factor of the VI-concentrator was most prominent for particles with diameters from 1.3 to 2.5 µm, with an average of 8.45±0.43. For smaller particles, the magnification factor decreased steeply towards 1, being 4.2 for 850 nm particles. This resulted in larger super-micron particles being over-represented in the sampled particles in comparison to their ambient number concentrations.
On average, the APS data showed very low number concentrations for particle diameters larger than 1.3 µm, and it can be expected that the role of such particles as INPs is more pronounced in the SPIN observations. Note that, due to the absence of correction method for the SPIN measurements during HyICE-2018, the SPIN data presented here is the concentrated INP concentration per volume of sampled air and not the back-calculated ambient INP concentration.

5
During its first three measurement days, SPIN was programmed to measure using automated RH-scans where the relative humidity with respect to water, RH w , was increased from 70 to 110% while keeping a constant lamina temperature. Such scans were realized for lamina temperatures of T l = -40 • C, -36 • C and -32 • C. On these three days, a total of 33 RH-scans were performed with 5-minute background checks performed between successive scans. Thereafter, the sampling program was modified to be more suitable for ambient particle measurements, and a modified scanning protocol was used for the remainder 10 of the campaign. The modified sampling protocol consisted of longer scans at T l = -32 • C and -28 • C, where at each T l , measurements were made with two saturation conditions (RH w = 95% and 110%). At each of the 4 conditions, 20-minute sampling intervals followed by 5 minutes of background sampling were used. Measurements continued until the background signal exceeded 10-15 particles L −1 , after which the chamber was re-iced to return to a lower background signal (Garimella et al., 2016(Garimella et al., , 2017(Garimella et al., , 2018. 15 For inter-comparison purposes, SPIN sampled at T l = -29 • C and RH w = 105% on March 22, and T l = -31 • C and RH w = 105% on March 28.

Expansion Chamber -PINE
The Portable Ice Nucleation Experiment (PINE) chamber ( Fig. 3b), as described in , has been developed based on the working principle of the Aerosol Interactions and Dynamics in the Atmosphere (AIDA) cloud chamber (Bunz 20 et al., 1996;Möhler et al., 2003Möhler et al., , 2006, which simulates cloud formation in rising atmospheric air parcels. The PINE chamber is operated in a cycled mode during which it is first flushed with ambient aerosol particles to renew the sampled volume of air under investigation. During HyICE-2018, this first mode was run for 4 minutes with a flow rate of 3 L min −1 . Then, the expansion mode is initiated by sealing the main inlet valve and pumping air out of the chamber, leading to a decreasing pressure within the chamber. As the pressure decreases to 700 mb in approximately 40 seconds, the volume of gas expands and the gas 25 temperature decreases, leading to an increase in the relative humidity within the chamber. Upon reaching water saturation, aerosol particles within the chamber activate to form supercooled liquid droplets. If aerosol particles immersed within the droplets are also active as INPs at the respective droplet temperature, then those droplets will freeze via the immersion freezing pathway. Frozen droplets are optically detected at the exit of the chamber using an Optical Particle Counter (OPC, Welas model 2500p) and are differentiated from liquid droplets based on their size. After the expansion, a final refill mode is conducted, 30 wherein the chamber continues to be depressurized using dry, filtered air to avoid any icing. The total experimental time of the 3-mode cycle is approximately 6 minutes and each cycle generates one measurement point. Longer averaging times (e.g., 1/6/24 hr) are commonly implemented in post-processing for purposes of statistical analysis. PINE is operated in a way that ensures frost-free walls such that a subtraction of any background ice counts is not needed.
During HyICE-2018, PINE was located in the main cottage (Figs. 1 and 4) and was operated continuously from March 13 until May 11, 2018 at measurement temperatures between -24 • C and -32 • C. During this time, PINE was operating with either a constant temperature or with a stepwise temperature ramping, during which the temperature was lowered three times by 2 -3 • C each hour. In the main cottage, PINE sampled from the heated total aerosol inlet without using a size cut-off in order to sample the total aerosol. Particle transmission efficiency as a function of particle size ( Fig. 6) was investigated by measuring 5 particle concentrations at the inlet of the PINE chamber and in the ambient air using an OPC (MetOne model GT 526S). The results illustrate that PINE was effectively sampling PM 5 aerosol; although the cut-off was not as sharp as would be expected from a traditional impactor.
During the inter-comparison days, PINE was operating at a constant temperature close to the lamina temperature, T l , selected for the CFDCs. On March 22, PINE sampled at a temperature of T =-29 • C while on March 28, the temperature was lowered 10 to T =-30 • C. For the inter-comparison conducted in April, PINE sampled at T =-29 • C.
While this study focuses on PINE's results during the inter-comparison days, the long-term measurements will be presented and discussed in the Adams et al. (2022) manuscript in preparation. Figure 6. Transmission efficiency as a function of particle size for the PINE sampling system. The measurements were made with an OPC (MetOne, GT 526S) with an accuracy of ±10%.

Filter sampling for droplet freezing assays
Numerous droplet freezing assay techniques exist for offline measurements of INPs in sampled aerosol. Techniques vary in 15 terms of aerosol collection methodology, which can be filter collection (Conen et al., 2012), particle impaction into liquid (Šantl-Temkiv et al., 2017), electrostatic deposition (Schrod et al., 2016) and/or collection of bulk materials (Hill et al., 2014(Hill et al., , 2016. All sampling techniques presented here rely on the re-suspension of collected material into known volumes of liquid ( Fig. 3c). The liquid is generally separated into sample aliquots which are exposed to decreasing temperatures and monitored for freezing using optical techniques. Aliquots are assumed to freeze when the most "active" INP in any given sample volume initiates ice formation. A series of assumptions allows the concentration of INP at a given freezing temperature to be calculated from the number of frozen aliquots (Vali, 1971b). It is important to note that these techniques typically assess the nucleation 5 tendencies of sample material that is immersed within the liquid (Fig. 3d). However, the strength of droplet freezing assay techniques is that the sampling and analysis can be disconnected and thus relatively simple sampling units can be deployed with the labor intensive analysis done at dedicated laboratory facilities. Another advantage of droplet freezing assay techniques is that they provide a complete INP temperature spectra, as opposed to the INP concentration at one specific temperature.
During HyICE-2018, two droplet freezing assay techniques using similar sampling protocols but different freezing systems 10 were deployed, and the analyses were performed in the SMEAR II laboratories.

I. Ice Nucleation SpEctrometer of the Karlsruhe Institute of Technology -INSEKT
The INSEKT is a droplet freezing assay based on the Colorado State University Ice Spectrometer (CSU-IS) design, which was originally developed by Hill et al. (2014Hill et al. ( , 2016. instruments. For both filters, the sampling time was approximately 3 hours, and sampling was carried out using a total aerosol inlet. After sampling, the collected aerosol particles were washed off the filters and suspended into 8 ml of nanopure water that had Brightness changes of the small sample volumes, which correspond to freezing events, were detected using a camera and an image acquisition and analysis software. The number of frozen volumes, which increases as the temperature decreases, is used to determine a temperature spectrum of INP concentration (Vali, 1971a). The INSEKT is able to measure INP concentrations at temperatures between -5 and -25 • C, which is relevant for heterogeneous freezing conditions within supercooled mixed-phase clouds.
After quantifying the INP content of the aerosol samples, some heat treatment tests were performed to investigate the heatsensitivity of the sampled aerosol with respect to their freezing ability, which can be used to investigate if biological particles 5 contributed to the INP population (Hill et al., 2016). For the heat treatment, the liquid aerosol suspensions were placed in boiling water (100 • C) for about 20 minutes. The heat-treated samples were then re-analyzed with the INSEKT to quantify changes in the INP temperature spectra.

II. microlitre Nucleation by Immersed Particle Instrument -µL-NIPI
The µL-NIPI, similar to INSEKT, is also a droplet freezing assay instrument for offline measurements of INP concentrations Research Kit (SHARK) (Porter et al., 2020) to quantify more detailed, size resolved measurements of the INP population, 15 although this was done only sporadically for a few different sampling periods.
After sampling, collected aerosol particles were washed from filters using 5 mL of nanopure water that had been filtered through a 0.2 µm filter (Sartorius, model Minisart). Droplets of the sample solution containing particles were pipetted onto a hydrophobic glass slide that holds approximately 50 droplets of 1 µL (Fig. 3c). The glass slide was placed on the temperaturecontrolled plate of the µL-NIPI, which was cooled to -40 • C at 1 • C per minute. The freezing temperature of each droplet was 20 recorded via a digital camera using changes in contrast to determine when a droplet had frozen .
Heat tests were performed on the samples by increasing the aerosol suspension temperature to 100 • C for 30 minutes (Hill et al., 2016;O'Sullivan et al., 2018). The heated samples were analyzed using the µL-NIPI and the results were compared to the original unheated sample results in order to quantify any changes in INP activity, which is used to infer information about INPs of biogenic origin.

Additional Aerosol Particle Characterization
Beyond INP measurements, further enhanced efforts were made to quantify, classify and assess aerosol properties at SMEAR II during the HyICE-2018 campaign. Multiple additional particle counters were installed to provide physical characterization of particles (Table 1). In addition to these ground and tower based instruments, airplane, drone flights and remote sensing retrievals were used to provide insight into the vertical distribution of aerosols and INPs. Moreover, the SMEAR II station 30 has a unique boreal location and is known for the documented occurrences of NPF events driven by biogenic volatile organic compounds emitted by the vegetation of the boreal forest (Lehtipalo et al., 2018). Hence, extra effort was made to assess whether any links between biology and INP emerge, with a special focus on the seasonal transition of the forest biome.

Search for bio-ice nucleators
The boreal forest is a diverse ecosystem in which biological drivers of aerosol properties have been previously identified.
For example, past research at the SMEAR II research station has shown that gas phase BVOCs can act as precursors to NPF (Kulmala et al., 2013). In addition, previous studies have identified several biological influences on INPs, including biological detritus (Hiranuma et al., 2019;O'Sullivan et al., 2015), pollen (Dreischmeier et al., 2017), bacteria such as Pseudomonas 5 Syringae (Morris et al., 2004), fungi (Morris et al., 2013), and other microorganisms such as viruses .
Therefore, during HyICE-2018, we aimed to assess potential links between biological activity and INPs.
In order to evaluate INPs of biological origin, a Wideband Integrated Bioaerosol Sensor (WIBS; Droplet Measurement Technology model WIBS-NEO) was used for online analysis of particles during the campaign. The WIBS measures particles between 0.5 and 30 µm using a light scattering technique. In addition to particle counting, particles trigger two optically 10 filtered (280 nm and 370 nm) Xenon lamps that excite the fluorescence of biological particles. The emission is monitored in two detection bands (310-400 nm and 420-650 nm), whereafter an additional fluorescence threshold can be used to distinguish biological particles (strong fluorescence response) from other materials (e.g., some types of dust, soot, black carbon) that are typically more weakly fluorescent. In this study, a threshold of FT + 9σ (where FT is the mean value of the forced trigger intensities and σ is their standard deviation) was used to determine if a particle is fluorescent (details in Savage et al. 2017).

15
The use of two excitation wavelengths and two detection channels allows for additional resolution in the fluorescence analysis, because different particle types often have different fluorescence intensities and emission bands (Savage et al., 2017). An additional benefit of the WIBS instrument is that it allows particle asphericity to be calculated (Savage et al., 2017). During HyICE-2018, the WIBS was operated from March 11 to June 25 at a flow rate of 0.3 L min −1 with an acquisition rate of 13 Hz, and the data was later averaged into user-defined time intervals of 10 minutes. The WIBS was first installed in the main cottage 20 (March 11 -April 3; Fig. 4) where it was sampling from the total aerosol inlet, as previously described. Later, the instrument was moved into the aerosol cottage (April 3 -June 25) and attached to a PM 10 inlet described in Schmale et al. (2017).
In addition to online measurements, several offline sampling techniques were used to investigate potential biological contri- and liquid water content (Illingworth et al., 2007). Moreover, radar observations can be used to identify ice particle growth processes like riming (Kneifel and Moisseev, 2020) and the onset of ice particle formation (Oue et al., 2015;Li and Moisseev, 30 2020). In combination with INP measurements, these observations are useful for identifying cases that may be attributed to secondary ice production (Field et al., 2017;Sinclair et al., 2016).
The wide range of techniques and instrumentation employed during HyICE-2018 have led to many results. In this paper, we summarize the general results for the duration of the intensive campaign and focus on a few instances where defined efforts were made to inter-compare different INP measurement techniques and instrumentation. Longer term and more detailed studies that have emerged from specific instruments and/or activities have been or will be published independently, with many contributions 5 aimed at the Copernicus special issue, "Ice nucleation in the boreal atmosphere" (Paramonov et al., 2020;Schneider et al., 2021).

Meteorological Conditions and Seasonal Change
The HyICE-2018 campaign aimed to capture the seasonal transition from winter to summer conditions. As expected for the boreal region, during winter the campaign was characterized by deep snow cover and cold temperatures (Fig. 7). In 2018, the 10 transition from winter to summer was rather abrupt at the measurement site, and as depicted in Fig. 7, snow cover went from a near maximum to completely melted away within a couple of weeks in April 2018. This transition coincides with an increase in the fraction of fluorescent particles (Fig. 8), which are used as a proxy for biological particles (Savage et al., 2017). By mid-May, the forest ecosystem was fully transitioned into summer and ambient temperatures reached nearly 30 • C. In fact, May 2018 was declared anomalously warm over all of Finland (Sinclair et al., 2019). The seasonal change is also noticeable 15 from particle number size distribution measurements (Fig. 9), with an increase in particle concentrations beginning in April.
The seasonal change is less evident in the NPF event frequency, which is also plotted in Fig. 9

INP measurement inter-comparisons
To maximize the number of instruments that were available for the inter-comparison study, March 22, March 28, April 26 and April 28, 2018 were chosen for instrument inter-comparison. During these days, the online chambers PINC, PINCii, SPIN and PINE were operated with thermodynamic conditions close to one another and an effort was made to maximize the temporal overlap of measurements. Filter sampling for droplet freezing assay inter-comparison measurements were carried out on March 5 28, 2018.

Online instrument inter-comparison
A summary of the thermodynamic conditions used with the online chambers during the inter-comparison days is shown in   Figure 11 presents the time series of each chambers' measurements during these inter-comparison days. The PINE data is presented as a 5 point moving average to smoothen the high frequency variability, with error bars that represent 20% uncertainty in absolute INP concentrations (cf. Möhler et al., 2021). The SPIN data is generated from the difference between 15 min 15 sampling averages and 5 min interpolated background concentrations, with error bars that represent ±1 standard deviation of the processed signal. The PINC and PINCii data is processed in an analogous manner to the SPIN data, but with sampling windows of 20 min and 15 min, respectively, and background windows of 5 and 15 minutes, respectively.
As seen in Fig. 11 Table 2 for more information concerning the thermodynamic conditions at which the chambers were running. Traces of aerosol number concentration, N p(>0.5 µm), and ambient air temperature, Tair, are plotted as solid and dotted lines, respectively. Note that the single SPIN data point that includes a black cross on March 22, 2018 is deemed to be below the level of detection, defined as the average background signal plus one standard error of the mean.
partially due to instrumental differences. In the cases presented here, the PINC data is corrected by a uniform scaling factor of 1.14 determined from well-characterized particle losses and lamina spreading measured in the instrument (Paramonov et al., 2020). Currently no simple scaling factor is available for the SPIN during HyICE-2018, and thus, no correction factor is used for the SPIN results. However, the biases in the ice-activated fraction from the SPIN chamber were discussed in Korhonen et al. (2020), when the chamber was used in a separate laboratory experiment. The study shows that at approximately -31 • C, 5 the activated fraction is biased low by a factor of ≈ 3 due to lamina spreading and particle losses, which may partly explain why SPIN measures systematically lower INP concentrations than PINE and PINC during HyICE-2018. obtained from the SMEAR II APS is plotted for comparison (Fig. 11, solid black lines) and suggests a reasonably constant activated fraction throughout the inter-comparison days. On March 22, the aerosol number concentration gradually decreases between 10:00 and 14:00 (UTC+2), and it is interesting to observe that such variability is reflected in both the PINE and PINC data. The ambient air temperature is also represented (Fig. 11, dotted lines) and shows a normal diurnal variability.

Offline instrument inter-comparison
Filter sampling for the droplet freezing assay inter-comparison was conducted on March 28 and extended through the morning 5 of March 29. Figure 13 depicts the time periods during which sampling took place. Although the filters were not collected at the exact same time, efforts were made to maximize the temporal overlap of the samples. The INP temperature spectra of the collected filters are presented in Fig. 14. Although INSEKT detects INPs at temperatures up to 5 • C warmer than µL-NIPI, the INP temperature spectra show good overlap between the two techniques and strong temporal agreement. An increase in the INP concentration is observed for the filters that were collected exclusively during daytime hours, with no clear bias between the morning and the afternoon. As reflected in Fig. 11, the aerosol number concentration remained nearly constant on that day, while other meteorological parameters, like the ambient air temperature, were fairly stable with expected diurnal variability. 08:00 12:00 16:00 20:00 00:00 04:00 08:00 Mar 28  14 and 16.
It should be noted that dilutions were used to extend the INSEKT detection range to lower temperatures. For the measurements presented in Fig. 14, the data points between -12 and -18 • C are coming from the non-diluted aerosol solution while the data points at lower temperatures are coming from a solution that was diluted by a factor of 10. Although the decreasing "step" 5 we observe between the series of data points is nonphysical, we nevertheless decided to present all data points here as measured. Moreover, when considering the spectra in the range of the error bars, the data points overlap and the INP concentration still increases with the decreasing temperature.
In Fig. 15, the INSEKT and µL-NIPI methods are directly compared for samples selected with as much temporal overlap as possible. The results are very similar, and a good agreement is observed over the entire range of temperatures. The primary 10 source of deviation in the agreement between the two methods is shown in Fig. 15c, which is expected due to a shorter temporal overlap in the sample collection for these two filters. More specifically, the data shown in Fig. 15c was obtained from one filter collected between 12:30-17:00 (UTC+2) for by µL-NIPI and one filter collected from 16:30-20:00 (UTC+2) for INSEKT, representing only 30 minutes of overlap between the measurements.   Fig. 13) versus µL-NIPI afternoon measurements (dark green in Fig. 13) (c) INSEKT afternoon (light green in Fig. 13) versus µL-NIPI afternoon measurements (dark green in Fig. 13) and (d) INSEKT night (dark blue in Fig. 13) versus µL-NIPI night measurements (light blue in Fig.   13). In each case the dashed line represents the 1:1 line.

Inter-comparison summary
In Fig. 16     Soil and water balance (includes snow and water collection; soil temperature, water content, matric potential and heat flux measurements; measurements of CO 2 , CH 4 and VOC concentrations and fluxes in the forest floor)   (Paramonov et al., 2020) and are available at https://doi.org/10.3929/ethz-b-000397022.