2.2.1 The Colorado State University Continuous Flow Diffusion Chamber (CSU-CFDC)

The CSU-CFDC is the longest-existing instrument for online detection of ambient INPs, with a legacy of versions for ground- and aircraft-based measurements starting from the late 1980s (Rogers, 1988; Rogers et al., 2001; DeMott et al., 2018). Its working principle is based on the establishment of supersaturated water and ice conditions in flowing air between two ice-coated walls of cylindrical shape in a vertical orientation. Those walls are held at different temperatures, and while the air temperature in the central lamina region is a linear function between these temperatures, the water vapour pressure is a non-linear function of temperature, resulting in a supersaturated region with respect to ice and water between the walls. The aerosol lamina is surrounded by particle-free sheath air through this region, where particles can activate into water droplets and ice crystals. While cloud droplets are evaporated downstream using an evaporation section, the remaining ice crystals are detected by their larger size using an OPC (Climet CI-3100). The size threshold to determine ice crystals was thereby 4 µm. The CSU-CFDC uses a pair of single-jet impactors upstream of the chamber for this study, with inserts defining 50 % aerodynamic size cuts at 2.5 µm, such that effectively only aerosol particles smaller than this size enter the system. This allows ice crystals to be differentiated from larger ambient aerosol. The measurement uncertainties at −30 °C with regard to temperature and relative humidity with respect to water are stated as ±0.5 °C and 2.4 %, respectively (DeMott et al., 2015). The residence times of aerosols in the supersaturated region are 5 s for the flow rate used (1.5 L min⁻¹). For this study, water supersaturation was controlled to be sufficiently high to promote comparison to the results of immersion freezing methods (DeMott et al., 2017). The mean and median supersaturations employed for this study were both equal to 6.5 % (i.e. 106.5 % relative humidity with respect to water, RH_water), with a standard deviation of 1.4 %. At this value, it is likely that maximum INP concentrations are not captured, although underestimations would be expected to be less than the factor of 3 noted for mineral dusts in comparing data collected at 105 % versus 109 % in DeMott et al. (2015). The 1 Hz data were accumulated and averaged over a time period of 1 min for this study. CSU-CFDC is typically operated for ∼4 h before refreshing the ice surfaces on the walls. Operation times in excess of 4 h can result in an increase in background ice counts (due to frost) in the chamber and thereby degrade the signal-to-noise ratio. CFDC background corrections are needed to account for INP signal contamination that may come in the form of frost crystals flaking from the ice walls (Rogers et al., 2001). Infrequent, high-concentration bursts may occur, typically in the time just following wall icing or after a number of hours of operation. These are accounted for with a data pre-screening method to search for outliers in ice crystal arrival rates at the optical particle counter (Moore, 2020). The more common intermittent, low-concentration frost events are corrected for by comparing ambient measurements with measurements of high-efficiency particulate absorbing (HEPA)-filtered air. For PICNIC, these filter periods were 5 min long, thus bookending each 10 min ambient air sample period. The correction for intermittent frost events has recently been modified to improve the estimates of statistical significance and confidence intervals over previous techniques, following Krishnamoorthy and Lee (2013). The background INP counts from filter periods that bracket each ambient measurement are combined into a single Poisson distribution with a characteristic rate parameter. The difference between the ice crystal arrival rates during the ambient measurement and the combined filter period is used to calculate the background-corrected INP concentrations (Moore, 2020). Statistical significance and confidence intervals for each ambient measurement are determined using the moment-based Z statistic defined in Krishnamoorthy and Lee (2013).

2.2.2 The Spectrometer for Ice Nuclei (SPIN)

SPIN is a commercially available CFDC-style instrument developed by Droplet Measurement Technologies (Garimella et al., 2016). It is based on the design of the laboratory instrument ZINC (Zurich Ice Nucleation Chamber; Stetzer et al., 2008) and its mobile version PINC (Portable Ice Nucleation Chamber; Chou et al., 2011). Briefly, two parallel flat plates are separated by 1 cm and each coated with 1 mm of ice prior to experiments. A temperature gradient between the two plates establishes a supersaturation with respect to ice and potentially liquid water. The supersaturation employed for this study was 2.8±1.9 % (102.8 % RH_water ± 2.5 %), with an uncertainty in temperature of ±0.5 °C. Aerosols are fed into the chamber at a sampling rate of 1 L min⁻¹ and constrained to a lamina centreline with 9 L min⁻¹ of sheath air. The residence time of the particles in the chamber is 10 s. An impactor with a 50 % size cut at 2.5 µm (BGI Inc., SCC1.062 Triplex) was installed before the SPIN inlet. Activated INPs are detected using a light-depolarization OPC (Garimella et al., 2016; Droplet Measurement Technologies). Due to the sigmoidal shape of the impactor's size cut, OPC counts larger than 5 µm in diameter were considered to be activated INPs. Although SPIN is operated at a lower supersaturation as compared to the CSU-CFDC, the ice crystals have a longer residence time (10 s), such that they grow to sizes larger than 5 µm.

Aerosol spreading due to turbulence at the inlet can cause some sampled aerosol to spread outside of the aerosol lamina, where they are exposed to a lower relative humidity. This phenomenon is ordinarily accounted for by applying measurable correction factors to the CSU-CFDC and SPIN data (DeMott et al., 2015; Garimella et al., 2017; Wolf et al., 2019). The degree of aerosol lamina spreading, and therefore the correction factor applied to observed INP concentrations, depends on several variables such as inlet pressure, chamber temperature, and degree of supersaturation. The correction factor for SPIN has been quantified to vary from approximately 1.5 to 9.5 for immersion freezing conditions (Garimella et al., 2017). As the degree of aerosol lamina spreading was not quantified in this study, no correction factor was applied.

The uncertainty in INP concentration for SPIN represents the standard deviation during a 10 min sampling period. SPIN's limit of detection is dependent on background ice concentrations resulting from ice shed from the walls. Backgrounds were measured for 5 min on both sides of a 10 min sampling period. Average backgrounds before and after a sampling period were subtracted from the average measured INP concentration. Only data from when backgrounds were less than half of measured INP concentrations are reported. The campaign-averaged background concentration was ∼3 L⁻¹. The limit of detection of SPIN sampling at the concentration is lower (∼0.6 INP L⁻¹) when compared to not sampling at the concentrator (∼6 INP L⁻¹), as more sampled air is analysed, while the ice background counts remain the same. SPIN can typically be operated for 4 to 6 h before backgrounds are too high to prevent measurement of ambient INP concentrations. Beside the results from SPIN presented in this article, focusing on mixed-phase cloud conditions, SPIN also measured cirrus-relevant INP concentrations, which are discussed elsewhere (Wolf et al., 2020).

2.2.3 The Portable Ice Nucleation Experiment (PINE)

The PINE is a new type of mobile instrument to measure INPs (Möhler et al., 2021). It is based on the AIDA (Aerosol Interaction and Dynamics in the Atmosphere) chamber and mimics cloud formation upon air mass lifting by expansion. The instrument is fully automated and can be operated continuously. During the PICNIC campaign, the PINE-1A version was deployed. This version consists of a 7 L cylindrical chamber, which is cooled by an external ethanol cooling chiller (Lauda RP 855; Lauda-Königshofen, Germany). PINE operates in a cycled mode of flush, expansion, and refill. During the so-called flush mode, aerosol particles are guided through the chamber at a flow rate of 2 L min⁻¹ for 5 min. Before entering the chamber, the sampled air is dried to a frost point temperature of below $\sim -\mathrm{13}$ °C, which avoids accumulation of ice on the chamber wall. An OPC (welas^®  2500, Palas GmbH, Karlsruhe, Germany) attached to the outlet of PINE counts larger unactivated aerosol particles. The flush mode is followed by the expansion mode when a valve upstream of the chamber is closed while the volumetric flow out of the chamber is set to a constant value of 3 L min⁻¹. Please note that the inlet flow rate during the expansion is maintained by a bypass flow, which is the same as the flush flow rate, such that no change in the sampling flow at the WAI occurs. A total pressure reduction of ∼300 mbar is thereby induced over a time of ∼50 s. During this expansion, the air temperature in the chamber is decreased by expansion cooling. As the wall and air temperatures are below the frost point temperature, the chamber is ice-saturated at the start of the expansion and achieves supersaturation with respect to ice and water during the course of the expansion, such that cloud droplets (on cloud condensation nuclei) and ice crystals (on INPs) can form. The temperature during one expansion typically decreases by 6 °C. The coldest temperature is thereby used as the nucleation temperature for each experiment, as it is assumed that the coldest temperature dominates the INP number concentration. After completing an expansion, the chamber is set to the refill mode, where the chamber is refilled with filtered sample air to reach ambient pressure conditions. Then another cycle of flush, expansion, and refill is started.

During the expansion, the ice crystals are detected by their comparably large optical size in the OPC, which makes a distinction with cloud droplets possible. As the OPC has a sideward-scattering geometry, aspherical ice crystals are detected with a higher scattering intensity than spherical cloud droplets of the same volume and refractive index. No ice background correction is needed for the INP measurements, since no ice crystals form from frost forming on the walls, which is confirmed by regular background experiments when the sampled air is passed over a filter to remove all particles before entering the chamber for several consecutive expansions.

In the PINE instrument, the residence time of aerosol particles at supersaturated conditions or in supercooled droplets is more variable compared to CFDC instruments. The time during which cloud droplets are present during an expansion is 33 s. However, it should be noted that this is an upper limit for the residence time, as ice crystals formed by INPs are detected during the whole expansion period, and each INP has its own trajectory within the cloud chamber. In the presented study, the INP concentrations are averaged over two consecutive experiments (two cycles of flush, expansion, and refill) to increase the detection limit for INPs. During the course of one expansion, about 2 L of air are continuously taken out of the chamber and analysed for forming ice crystals. The welas^® 2500 OPC has an optical detection volume of 10 % and thus has a limit of detection of 2.5 INP per litre for two consecutive experiments. The uncertainty for the INP concentration is 20 %, which is an upper estimate from the uncertainties in the determination of the optical detection volume. The uncertainty in temperature is ±1 °C (see Möhler et al., 2021, for further details about the specifications of PINE).

The majority of aerosol particles with an aerodynamic diameter of <2 µm are sampled with PINE (80 %), which decreases to <50 % for particles with an aerodynamic diameter of >4 µm. No impactors were used with the PINE instruments. However, when sampling at the PFPC, which is operated with an impactor with a 50 % size cut at 2.5 µm, the sampled particle size was limited to this size.

In order to compare the PINE measurements to the offline methods for a perfect time overlap, PINE joined the offline intercomparison times for some nighttime measurements and measured at a constant temperature during the 8 h.

2.3.1 FRankfurt Ice nucleation Deposition freezinG Experiment (FRIDGE)

For the FRIDGE measurements, aerosol particles were collected in the laboratory from the WAI inlet. Aerosol was collected by the use of a custom-built semi-automated multi-filter sampling device. The unit consists of eight individual filter holders, the 45.7 cm housing, valves, a pump, and electronics. The sampling time of each filter can be programmed separately. The flow rate through the filters was determined to be 4.8±0.4 SD L min⁻¹ on average. This is more than 50 % lower than the flow rate that was originally targeted due to a miscalibration and a leakage in the system. Accordingly, the flow rate needed to be corrected to the above-mentioned value and carries a rather high uncertainty. Aerosol particles were collected onto 47 mm hydrophobic PTFE Fluoropore membrane filter of 220 nm pore size (Merck Millipore). Filters were not pre-cleaned in any way. It was decided to limit the sampling time for FRIDGE to 4 h during daytime (10:00–14:00 LT; i.e. termination in the middle of the total sampling time of other instruments), as we initially expected higher INP concentrations compared to nighttime sampling, and to better capture potential variability in INP concentrations. The nighttime sample was the same as for the other groups (8 h). Moreover, on 18 October, the sampling time was not increased to 24 h, as for other methods. Filters were stored frozen at −18 °C after collection at the site. The samples were not actively cooled during transport; however, given the relatively short travel time of ∼8 h to the laboratory in Frankfurt, we do not consider that this impacts the results, but it cannot be excluded for certain (Beall et al., 2020). After transport, they were stored in a refrigerator at 4–7 °C until freezing measurements were performed. The analysis was performed using the droplet freezing mode of FRIDGE (Hiranuma et al., 2015). Before starting a measurement, a filter containing the sampled aerosol was placed in a sterile Eppendorf tube, which was filled with 5 mL of ultrapure water (ROTIPURAN^® Ultra, Carl Roth). Particles were then extracted into the ultrapure water by repeated steady shaking for several minutes, without dilutions. Using an Eppendorf Reference 2 pipette, a total of about 200 (184–231) 2.5 µL droplets were manually pipetted onto a 45 mm silanized (dichlordimethylsilan) silicon wafer substrate placed on a cold stage inside a 500 cm³ measurement cell. About 65 droplets of 2.5 µL fit onto the substrate at a time; therefore, three individual runs per sample were performed to improve the freezing statistics. Before and after each measurement run, the substrate was thoroughly cleaned with pure non-denatured ethanol (ROTIPURAN^®, >99.8 %, Carl Roth). During the experiment, the measurement cell was constantly flushed with dry synthetic air at 1 L min⁻¹ to prevent condensation and riming. The temperature was decreased at a constant rate of 1 °C min⁻¹ until every droplet was frozen, using a proportional–integral–derivative (PID)-controlled Peltier element. An ethanol cryostat cooling system supported the Peltier by dissipating the heat. The surface temperature was measured with a Pt100 sensor, which has an accuracy of ±0.2 °C. A camera saved measurement images every 10 s, and a change in brightness was detected when droplets were freezing. A detailed description of the FRIDGE immersion freezing method can be found in Schrod et al. (2020).

Table 2Specifications of the offline freezing methods.

^* INDA and LINA use the same collected filter.

Download Print Version | Download XLSX

2.3.2 Ice Nucleation Droplet Array (INDA)

For analysis with INDA and also the below-discussed LINA (Leipzig Ice Nucleation Array; see Sect. 2.3.5), three different types of filters and two different samplers were deployed, with both samplers operating in parallel. All filters were taken at the WAI. Quartz fibre filters (Munktell, MK 360; 47 mm diameter) were used for sampling, as well as polycarbonate filters (Nuclepore, Whatman, 47 mm diameter with pore sizes of 200 or 800 nm). One sampler was a simple standard filter holder. The sampling flow was deliberately set to different values for different sampling periods, varying between 12 and 37 L min⁻¹ and resulting in total collected air volumes between 6 and 18 m³. The other sampler was HERA (High Volume Aerosol Sampler; Hartmann et al., 2020; Grawe et al., 2023), which was developed for airborne sampling and enables the subsequent sampling of six filters. For HERA, the sampling flow varied between 15 and 41 L min⁻¹, resulting in collected air volumes between 7 and 20 m³. All samples and blank filters were stored in separate Petri dishes right after sampling and stored and shipped frozen until the analysis was done at the laboratory in Leipzig.

INDA is based on a measurement technique that was introduced by Conen et al. (2012) and modified as suggested by Hill et al. (2014). A suspension is obtained by washing particles off a polycarbonate filter. For this, the filters are put in 3 mL of ultrapure water, followed by shaking for 15 min in a flask shaker. Subsequently, typically 0.1 mL of the suspension is used for a LINA experiment (Sect. 2.3.5). Then 3.1 mL of ultrapure water is added, and 50 µL droplets of this suspension are placed into 96 wells of a polymerase chain reaction (PCR) tray. For the quartz filter samples, each well is filled with 50 µL of ultrapure water, together with a 1 mm diameter filter punch from the quartz fibre filter. The PCR tray is then immersed in a temperature-controlled cooling bath of a thermostat and is illuminated from below. During cooling, typically done at 1 °C min⁻¹, a picture is taken every 6 s from above. Changes in the colour of wells occur during freezing and are automatically detected. More information can be found in Gong et al. (2020) for the INP analysis of quartz fibre filters and in Hartmann et al. (2020) for polycarbonate filters.

2.3.3 The Colorado State University Ice Spectrometer (IS)

The Colorado State University (CSU) Ice Spectrometer (IS) analyses arrays of liquid suspensions from filter samples to quantify immersion freezing INP concentrations (e.g. DeMott et al., 2018). Aerosol filter samples were collected on the roof of the laboratory using precleaned (5 % H₂O₂, followed by two 100 nm filtered deionized (DI) water rinses), 200 nm pore diameter, 47 mm diameter Nuclepore polycarbonate filter membranes (Whatman, GE Healthcare) held in open-faced sterile Nalgene sampling heads. Mass flow rates (at 101.3 kPa and 0 °C) were recorded at the start and end of the sample period to calculate the total volume filtered. The average sample volume collected was 6 m³. Filter samples were immediately placed into sterile Petri dishes (Pall) and stored and transported frozen until analysis of INPs in Fort Collins, Colorado.

For analysis, 10 mL of 0.1 µm-filtered (Whatman Puradisc, PTFE membrane) DI water was added to a pre-rinsed polypropylene 50 mL tube (Corning) and shaken in a Roto-Torque rotator (Cole-Parmer) for 20 min to create a suspension. For each sample, serial 20-fold dilutions were made to 8000-fold. Next, 32 50 µL aliquots of each sample, with corresponding dilutions, and a 0.1 µm-filtered DI water blank were dispensed into 96-well PCR trays (OPTIMUM^® µLTRA Brand, Life Science Products) in a laminar flow hood. The trays were then placed into aluminium blocks in the IS and cooled at a rate of ∼0.33 °C min⁻¹. Freezing was detected by a CCD camera and the corresponding temperature was recorded with a LabVIEW interface. Frozen fraction results were corrected for the number of INPs in the DI water blank, resulting in the lowest freezing temperature achievable (generally between −27 and −30 °C). Temperature uncertainty is estimated at <±0.5 °C. The proportion of frozen wells was converted to a number of INPs per millilitre of suspension, using Eq. (13) in Vali (1971), and subsequently scaled to the number of INPs per filter. The average number of INPs on three field blanks (cleaned, handled, transported, and processed in the same way, with the exception of airflow) was subtracted from all samples before conversion to INPs per litre of air, considering the volume collected. Two-tailed 95 % confidence intervals for binomial sampling were calculated, based on Agresti and Coull (1998). Some samples of IS were investigated for the size of INPs by filtering the suspensions at 3 or 0.8 µm.

2.3.4 The Ice Nucleation Spectrometer of the Karlsruhe Institute of Technology (INSEKT)

INSEKT is a rebuild of the IS freezing method (e.g. Schneider et al., 2021). During PICNIC, aerosol particles were collected in the laboratory via the WAI with a standard filter holder. The aerosol particles were sampled with a flow rate of 11.3 (±0.2) SD L min⁻¹ on 47 mm diameter Nuclepore filters (Whatman), with a pore size of 200 nm. The filters were pre-cleaned (10 % H₂O₂ solution) and kept frozen after aerosol particle collection and during transport until analysed in the laboratory in Karlsruhe. For INSEKT analysis, aerosol particles are washed off the filter using 8 mL filtered nanopure water (100 nm pore diameter filter and 18 MΩ deionized water) and shaken in a rotator for 20 min to ensure the release of all particles from the filter. The resulting suspension is then diluted by factors of 1, 15, and 225, and volumes of 50 µL are placed in the wells of a sterile PCR tray, alongside filtered nanopure water samples, to determine its freezing behaviour for a background correction. The PCR tray is then placed in an aluminium block cooled with an ethanol cooling bath (Lauda RP 890; Lauda-Königshofen, Germany). From a starting temperature of 0 °C, the wells are cooled down at a rate of 0.33 °C min⁻¹. Four Pt100 temperature sensors are placed inside the aluminium blocks for each PCR tray, measuring with an accuracy of ±0.1 °C and a deviation to the edges of the wells of ±0.1 °C, resulting in an uncertainty in temperature of ±0.2 °C. A camera detects brightness changes in the wells that correspond to their freezing.

Washing water from handling filter blanks that were taken prior to the 8 October 2018 started to freeze at −7 °C, which was traced back to using non-powder-free gloves during the filter handling procedure at the Puy de Dôme, which was changed thereafter, demonstrating the need to work cleanly (Barry et al., 2021). Therefore, filters handled with non-powder-free gloves had to be disregarded. Moreover, filters containing parts of insects, which were sampled due to a leak in the WAI mesh, were excluded from the analysis.

2.3.5 The Leipzig Ice Nucleation Array (LINA)

LINA is based on a method described by Budke and Koop (2015). The filters were sampled as described in Sect. 2.3.2; however, only polycarbonate filters were analysed in LINA using washed suspensions. Of the resulting suspensions from the filter washing water, 90 droplets with a volume of 1 µL are pipetted onto a hydrophobic glass plate, which is placed on a Peltier element. Each droplet is contained in a separate compartment, which is covered by a second glass slide. Droplets are illuminated by a ring of light, which, together with a camera, is installed above. During the cooling process, typically done at 1 °C min⁻¹, a picture is taken every 6 s from above. Changes in the reflection of the light by the droplets related to freezing are automatically detected. A more detailed description can be found in Gong et al. (2019).

2.3.6 The LED-based Ice Nucleation Detection Apparatus (LINDA)

The LED-based Ice Nucleation Detection Apparatus (LINDA) is an immersion freezing detection device that allows automatic detection of freezing in closed tubes by light transmission and is described in detail by Stopelli et al. (2014). Quartz filters (15 cm diameter) were used for analysis with LINDA and taken with a high-volume sampler at the rooftop, with a sample flow of 500 L min⁻¹. The filters were stored in the freezer at −20 °C until analysis in the laboratory of LaMP close to the Puy de Dôme.

For analysis, four circular samples (1.2 cm diameter) were extracted from each filter and were washed in a 25 mL solution of 0.9 % NaCl for 20 min; then, 200 µL of the resulting solution was introduced to each of the 52 tubes. The array of tubes is placed in a cooling bath, with a Pt100 temperature probe at each corner of the array. A camera placed above the array detects the freezing of the tubes through the variation in the intensity of the transmitted light through the tubes. Error bars were calculated from freezing events from background filters and the NaCl solution.

INP concentration measurements from LINDA were already presented by Bras et al. (2022) to investigate the seasonal variability. Here, we focus on the comparison to other INP concentration measurements.

2.3.7 The Universidad Nacional Autónoma de México–Micro-Orifice Uniform Deposit Impactor–Droplet Freezing Technique (UNAM–MOUDI–DFT)

Aerosol particle collection was carried out by an inertial cascade impactor (MOUDI 100NR, MSP Corporation), which divides the particles according to their aerodynamic diameter in each of its eight stages (cut sizes of 0.18, 0.32, 0.56, 1.0, 1.8, 3.2, 5.6, and 10.0 µm). For this study, particles impacted on stages 2 to 7 were used. Hydrophobic glass coverslips (Hampton Research) were used as substrates in each of the eight stages. During PICNIC, the collection of particles was done in the laboratory via the WAI at a flow rate of 30 L min⁻¹. After particle collection, the samples were stored in 60 mm Petri dishes and refrigerated at ∼4 °C for transport to the laboratory in Mexico City, where the analysis using the droplet freezing technique (DFT) was performed. We note that storing samples for a longer transportation time might impact the INP concentration (e.g. Beall et al., 2020). Although we did our best to keep the samples below 0 °C by transporting them in a freezer with ice packs, it is very likely that the samples may have experienced temperatures slightly above 0 °C right before reaching their final destination.

The DFT, built at the Institute for Atmospheric Science and Climate Change at UNAM (Córdoba et al., 2021), is based on the design by Mason et al. (2015) and determines the concentration of INPs as a function of temperature and aerodynamic particle size via immersion freezing. Each substrate is isolated in a temperature-controlled cell. Supersaturated conditions with respect to water are generated to trigger cloud droplet formation on the aerosol particles deposited on the substrate. The typical size of the droplets is around 100 µm, and 30 to 40 droplets are formed in the study area (1.2 mm²). The experiment is monitored in real time with an optical microscope (Axiolab, Zeiss, Germany) with a 5× /0.12 magnification objective coupled to a video camera (MC500-W, JVLAB). Droplets are subsequently cooled down from 0 to −40 °C at a cooling rate of 10 °C min⁻¹. The temperature at which each droplet freezes is determined when the temperatures from the cold cell (monitored with a resistance temperature detector, RTD; ±0.1 °C uncertainty) and the videos are integrated. The INP concentration is derived from the following expression from Mason et al. (2015):

where [INPs(T)] is the INP concentration, N_u(T) is the unfrozen droplets (L⁻¹) at a certain temperature T (°C), N₀ is the total number of droplets analysed, A_deposit is the total area where the aerosol was deposited on the MOUDI hydrophobic glass coverslips (cm²), A_DFT is the area analysed by the DFT, V is the volume of air sampled by the MOUDI (L), f_nu is a correction factor (dimensionless) that takes into account changes in deposit inhomogeneity in a range between 0.25–0.10 mm in each of MOUDI sample, and f_ne is a correction factor that varies between 1.2 and 4.7 and that takes the uncertainty associated with the number of nucleation events in each experiment into account.

3.2.1 Investigation of INP differences using aerosol particle measurements

A wider spread between the methods based on filters collected at the rooftop and in the laboratory via the WAI is observed during many sampling intervals. In order to gain better insight into this deviation, the time series of the difference between the INP concentration measurements from the IS (rooftop) and INSEKT (laboratory) is investigated in relation to the wind velocity and the concentration of aerosol particles. Those freezing methods were selected as they are based on the same freezing analysis principle and both span a large range of T_nucleation. As seen in Fig. 10, the difference between the INP concentration measurements from the IS and the INSEKT at −10, −15, and −20 °C, given as the lognormal difference, is sometimes occurring during elevated wind velocities (Fig. 10b), which can decrease the transmission efficiency, especially of larger particles, as discussed earlier. No relation between the difference between IS and INSEKT is observed to the total particle number concentration and the particle number concentration 0.1–0.5 µm (Fig. 10c, d). Moreover, a higher ratio between IS and INSEKT is not observed during times of higher concentrations of particles between 0.5 and 2.5 µm and 1 and 2.5 µm (Fig. 10d), which would have been an indication for a generally higher concentration of larger particles in the ambient air and which might be preferentially lost in the inlet prior to the INSEKT filter samples. At the same time, all the aerosol particle concentration measurements (total, 0.1–0.5, 0.5–2.5, and 1–2.5 µm) are especially higher in the second period of the campaign, starting from 13 October, when a higher discrepancy between IS and INSEKT is observed. This might indicate that the aerosol population changed and could have caused this discrepancy, e.g. by an increased presence of larger particles that are not sampled at the WAI, or could have caused particle fragmentation in IS. This potential cause of the discrepancy depends on the assumption that especially the larger fraction of the aerosol particle population dominated the INP population. A study performed at the same location using a stage impactor for size-segregated measurements indeed revealed that INPs are mostly super-micrometre particles (Bras et al., 2022). It should be noted that the size distribution measurements were conducted at the WAI; thus, the interpretation of the presented time series of aerosol particles during those high-wind velocity times is limited. In order to precisely identify such an impact, more intensive measurements need to be conducted by, for example, having aerosol particle size distribution measurements at the rooftop and in the laboratory simultaneously.

Figure 10Time series of INP concentration differences between the IS and INSEKT at nucleation temperatures of −10, −15, and −20 °C (a); wind velocity (b); total particle concentration (c); and particle concentration in the size ranges 0.1–0.5, 0.5–2.5, and 1–2.5 µm (d).

Download

3.2.2 Comparison of INP concentrations using quartz fibre and polycarbonate filters

A subsample of the datasets was designed to test a possible influence of using different filter materials (quartz fibre versus polycarbonate filters). For this comparison, HERA and the standard sampler from TROPOS (Leibniz Institute for Tropospheric Research) were operated in parallel, using different filter materials. For the analysis, INDA and LINA, both operated at TROPOS, were used for evaluation. For the comparison shown here, HERA was equipped with polycarbonate filters (200 nm pore diameter), and the standard sampler with the quartz filter. Figure 11 shows results from sampling intervals between 9 (daytime) to the 11 (daytime) October. While both LINA and INDA can analyse particles collected with polycarbonate filters (creation of solution using the washing water), only INDA can analyse quartz fibre filter punches that are immersed in ultraclean water. No systematic difference between the INP concentrations using those different filter materials is observed, showing a good agreement between INDA and LINA, as previously reported (e.g. Knackstedt et al., 2018; Hartmann et al., 2019; Gong et al., 2020), which gives confidence that both materials can be used within the processing temperature ranges shown ($\ge -\mathrm{20}$ °C).

Moreover, quartz fibre filters and polycarbonate filters with a different pore size (800 nm) were used simultaneously in the TROPOS standard filter holder and in HERA for the analysis with INDA and LINA during some sampling intervals. Quartz fibre filters were used from the 14th nighttime (Fig. 6h) to the 16th daytime sample (Fig. 7a–c) and 800 nm polycarbonate filters for the sampling intervals from the 16th (nighttime) to the 18th (daytime; Fig. 7d–g). When comparing with the overall INP measurements from the other methods, there was no noticeable influence of using quartz fibre filters or polycarbonate filters with 800 nm pores, as compared to measurements using Nuclepore filters with a pore size of 200 nm. This shows that filters with a pore size of 800 nm and applied flow rate still have a sufficiently high collection efficiency for the majority of atmospheric INPs present during the PICNIC study. This is in agreement with Soo et al. (2016), who examined the collection efficiencies of a range of different filter materials and pore sizes for test particles with rather small sizes between 10 and 412 nm. They reported that the collection efficiency for polycarbonate filters with 800 nm pore sizes and the flow rates used here (>11 L min⁻¹) are above 97 % for all particles in the examined size range (10–412 nm).

Figure 11Comparison of different filter materials for parallel collected filters using INDA and LINA.

Download