The two measurement sites in Eri and HPB were equipped with a set of remote-sensing and in situ devices. The setup of the site in Eri is described in more detail in . Table provides a list of instruments that were used in this study.

Measurements were done in the framework of the CLOUDLAB and PolarCAP campaigns. Both campaigns took place side by side. The CLOUDLAB campaign involved a unique set of ground-based and airborne in situ cloud and precipitation sensors and remote-sensing instruments. During the two wintertime campaigns between 2022 and 2024, LACROS enhanced the remote-sensing capabilities of the CLOUDLAB campaign with a large number of ground-based equipment, such as a scanning 35 GHz and vertically-pointing 94 GHz cloud radar from TROPOS, Raman polarization lidar, Doppler lidar, ceilometer, micro rain radar, photometer, disdrometer, and microwave radiometer. In addition, INP sampling was conducted in Eri and additionally at the German Weather Service (Deutscher Wetterdienst, DWD) in Hohenpeißenberg, using a low volume aerosol filter sampler (DIGITEL, DPA14 LVS) with a PM₁₀ inlet at both sites (PM₁₀ refers to particles that pass through the size-selective air inlet, which has a separation efficiency of 50 % for an aerodynamic diameter of 10 µm). The filter magazines for the instruments at both sites were prepared simultaneously prior to shipment to the sites. This was done in a laminar flow hood at TROPOS to eliminate possible contamination on the filters. 47 mm polycarbonate filters with a pore size of 0.4 µm were used. A filter sample typically was collected for 12 h, with a flow rate of 25 L min⁻¹ at ambient conditions resulting in a typical sampling volume of approximately 18 000 STP L (standard conditions: 0 °C; 1013 hPa). During a Bise situation, the sampler was set to run continuously with an automatic filter change typically scheduled for 08:00 and 20:00 UTC see. We checked the weather forecast continuously and if we expected to have northeasterly wind conditions for at least four days, we started the measurements. Each sampled filter was stored in a separate Petri dish at -20 °C after sampling until off-line analysis was done at the TROPOS ice-laboratory. Field blank filters were also taken by inserting filters into the sampler, but without sucking air through them, and by subsequently treating these field blanks similar to sampled filters. During the campaign 21 and 17 samples were collected at HPB and Eri, respectively. The same type of sampler but with an inlet for total suspended particles collected particles on polycarbonate filters (pore size of of 0.8 µm) for INP analysis operationally at MEL. A filter is taken every third day for 24 h with a flow rate of about 22 L min⁻¹. A field blank is taken every 9 d. Filter handling and evaluation were the same as for filters collected at Eri and HPB.

Also, LACROS Leipzig Aerosol and Clouds Remote Observations System; data of the RPG94 Cloud radar, VISSS, 2DVD, HATPRO, and lidar PollyXT were used. In addition, windsondes were launched in Eri and radiosonde profile information for HPB was taken from for the Munich station. Surface temperature information for HPB was taken from the CDC Climate Data Center from DWD,.

Immersion freezing measurements were performed in the TROPOS ice-laboratory, using the two measurement devices Leipzig Ice Nucleation Array (LINA) and Ice Nucleation Droplet Array (INDA) . Both devices have been used in more than 20 studies in the past years, among them a comparison study done by , in which their results agreed well with those of other similar devices.

In short, INPs were washed off a filter sample by shaking the filter in 3 mL of ultra-pure water. For a LINA measurement, from this particle suspension, 90 droplets with a volume of 1 µL were then set onto a hydrophobic glass slide which itself sat on a Peltier element. On the glass slide, there was a grid which then again was covered with a second glass slide to separate the droplets from each other and the environment. The Peltier element cooled the droplets with a rate of 1 K min⁻¹. The setup was illuminated from above, and a picture was taken every 6 s (i.e., every 0.1 K) with a camera that was also installed above. Frozen droplets were discriminated from non-frozen droplets by a custom, Python-based image analysis algorithm. The result from such a measurement is a curve of frozen fractions fice (i.e., number of frozen droplets / total number of droplets) over temperature T.

The remaining suspension was diluted further by adding 3.5 mL ultra-pure water in order to create enough suspension to fill the wells of a PCR (Polymerase Chain Reaction) plate with 50 µL droplets to be used with the INDA device. PCR plates with 96 wells were used, and most of the time, two samples were examined simultaneously by filling 48 wells with one sample and the other 48 wells with the other (for MEL samples a full 96 well PCR plate per was used for each sample). After sealing the PCR tray with a clear adhesive film, it was immersed in an ethanol bath such that the liquid in the bath was above the liquid level in the wells. The bath was then cooled at 1 K min⁻¹. During cooling, the ethanol level was held constant with a custom-designed bath leveler similar to . A light in the bath illuminated the PCR plate from below, and a camera was installed above and coupled with the cryostat in order to take pictures in 0.1 K intervals. Again, frozen droplets were discriminated from non-frozen droplets with a custom, Python-based image analysis algorithm yielding fice(T).

Afterwards, the PCR plates were heated to 90 °C for 30 min to remove the ice nucleation ability of heat-labile INPs. These heat-labile INPs are a proxy for the fraction of proteinaceous biological INPs .

From all frozen fraction curves, INP concentrations were then calculated based on the assumption of a Poisson distribution of INPs via the examined droplets or wells, following . A more thorough description of the INP measurements can be found e.g., in , or .

HYSPLIT Hybrid Single-Particle Lagrangian Integrated Trajectory Model, is a model used to calculate trajectories, origin of air parcels, and dispersion. The model is freely available at . GDAS (Global Data Assimilation System, 1°-spatial resolution) meteorological data in terms of air pressure, temperature, humidity, wind speed, wind direction, or precipitation are required. The HYSPLIT model uses a Lagrangian approach (applied in the study) as well as a Eulerian approach. The Lagrangian approach is used with a moving frame of reference for the advection and diffusion calculations. The Eulerian approach uses a fixed three-dimensional grid as a frame of reference to compute pollutant air concentrations. The trajectories were calculated twice per measurement day at 02:00 and 14:00 UTC at 100, 250 and 1000 m a.g.l. (above ground level) as 72 h backward trajectories.

Figure a presents the temperature evolution during the cold measurement period (6–12 January 2024) as recorded at the field site of Eri and for HPB data from the DWD CDC portal . Throughout this period, temperatures remain consistently below freezing at ground level, with the exception of a short interval on 6 January 2024 in Eri. Especially on 8 and 9 January 2024, temperatures range between -5 and -10 °C. A notable temperature increase is observed at HPB on 10 January 2024, where temperatures increase sharply from -10 to 0 °C within a few hours.

Figure c displays vertical profiles of temperature for Eri for the same time period as Fig. a, showing the presence of a temperature inversion typically occurring between 100 and 1000 m a.g.l. Stratus clouds are located between 200 and 1500 m a.g.l. On 9 and 11 January, the inversion strength reaches nearly 10 K, resulting in temperatures around 0 °C in parts of the lower atmosphere at the inversion top.

In Fig. e the most remarkable change in temperatures happened in HPB (represented by radiosonde launches in München-Oberschleißheim) between 10 January 2024, 00:00 and 12:00 UTC. The inversion decreases in altitude by a few hundred meters which results in a temperature increase of around 10 K in HPB, as already mentioned above (Fig. a). Due to that, HPB is no longer situated in the PBL but above the local inversion (decoupled from the PBL) hereafter denoted as in the free troposphere.

During the warm measurement period (Fig. b), surface temperatures at both Eri and HPB remain consistently above freezing. Temperatures generally stay below 6 °C, except for HPB, where they occasionally increase to 10 °C, especially on 26 February 2024.

The vertical temperature profiles for the warm period are shown in Fig. d and f. Temperature inversions are present during this period as well. The surface temperatures range from 2 to 4 °C, decreasing to around 0 to -3 °C at the lower boundary of the inversion. The upper boundary of the inversion typically remains above freezing temperatures. The occurrence of the inversion leads to an inhibition of the exchange between the PBL and the free troposphere. Below the inversion the air should be well-mixed and therefore, we assume that the INP measurements at the ground are representative for the entire PBL.

Figure provides an overview of radar reflectivity at Eri during the two measurement periods. Figure a shows the cold period. Frontal systems produce vertically thick clouds on 6 and 7 January 2024, as well as in parts on 8 and 9 January 2024. Nevertheless, the supercooled low-level liquid Bise cloud layer remains persistent, even during these times (as it can be seen from the increased radar reflectivity between 0 and 1 km altitude). Especially from the radar depolarization ratio (not shown) it can be seen that it is a liquid cloud. In addition, the Holimo measurements Holographic Imager of Microscopic Objects in the frame of the CLOUDLAB campaign, see regularly confirmed this statement during the campaign. From noon on 8 January 2024 until midnight on 12 January 2024, the Bise cloud persists continuously, with only brief interruptions during the night from 9 to 10 January 2024. During this time, the cloud top slowly descends from approximately 1400 m to between 100 and 200 m a.g.l. INP filters were sampled throughout this period, although there were technical issues at both measurement sites, as indicated in the annotations above Fig. a and b (white gaps between the green lines).

Figure b shows the evolution of the profile of the radar reflectivity factor during the warm period. The Bise cloud formed on 27 February 2024 and persisted until the evening of 29 February 2024. During this warm Bise event, temperatures were mostly above freezing, both at the surface and within the cloud. Filter sampling was successfully conducted throughout the entire warm Bise period.

Figure , and Figs. S5, S6, and S7 in the Supplement show the backward trajectories ending at Eri at 14:00 UTC on 8, 9, 10, and 11 January 2024, respectively, for the cold Bise situation. Figures S8, S9, and S10 show the backward trajectories ending at Eriswil at 14:00 UTC on 27, 28 and 29 February 2024, respectively, for the warm Bise situation. In all cases it can be seen, that the backward trajectories move very close to HPB (especially during the cold Bise period) and especially on 8 January 2024 the trajectories are aligned with air masses blowing first over MEL, approximately 18 h later over HPB and another 12 h later over Eri. We also checked arbitrary trajectories in a few km lat/lon distance of HPB and Eri and found similar results for the trajectories, making it a robust result. Of course, individual trajectories themselves did not necessarily move directly from MEL via HPB to Eri, however, it can be assumed that the air mass is the same over all places coming roughly from the north-east. Along the trajectory there was only little precipitation.

Results from INP measurements are shown in Fig. , covering three different scenarios, each in a separate row. In each column, the leftmost panels Fig. a, d and g show a sketch of the respective situation, while all other panels show INP concentration spectra. INP spectra from Eri are always shown together with data from HPB taken 12 h earlier. During a Bise situation, the air mass moves towards Eri from northeasterly directions. We therefore assume that a comparable air mass also moved over MEL before. Therefore, INP measurements from MEL that correspond best in time are also added to the respective panels of Fig. . For MEL, there was no snow coverage on the ground at the time of the here shown measurements. Concerning Eri, as mentioned before artificial cloud seeding was done with a drone in supercooled Bise cloud layers. A flare was ignited at the drone, releasing silver-iodide that can act as an INP at temperatures below -5 °C . The flare was ignited inside the clouds upwind of the measurement site, so the plume of silver-iodide, the potential source of INPs, could have dispersed and reached the measurement site in theory. However, we want to note that we checked the entire dataset for signals from artificial seeding in the INP measurements. No evidence of cloud seeding particles on the ground was observed throughout the whole campaign.

Figure b and c each show the INP measurements at Eri and HPB for the situation when both sites are within the stratus cloud. The cloud top temperatures are below 0 °C and overall the temperatures in these clouds range between 0 and -10 °C. INP concentrations are very similar at Eri and HPB. Heat treatment causes a decrease in INP concentrations, indicating a fraction of heat-labile, probably biological INPs, however, not as pronounced as for the warm Bise situation in Fig. h and i. Two days earlier in MEL, the INP concentrations are much higher compared to the other two stations.

In the middle row, Fig. e and f show INP measurements for a time when Eri is located in the upper part of the Bise cloud within the PBL, equivalent to the measurements shown in Fig. b and c. HPB, however, is slightly above the height level of the temperature inversion and therefore within the free troposphere after the inversion has lowered in altitude (see Fig. ). The lowering of the inversion layer heights at HPB occurs at 10 January 2024 around 03:00 UTC. Therefore, HPB INP data in Fig. e stems from air collected from both the PBL and the free troposphere. In the time period from 10 January, 08:00 p.m. to 11 January, 08:00 a.m. shown in Fig. f the measurement at HPB was performed entirely within the free troposphere. Already in Fig. e, the INP spectrum from Eri is clearly below that from HPB. That, however, results from a lowering of the INP concentrations at Eri, visible when comparing Fig. c to Fig. e. For Fig. f, when HPB was entirely in the free troposphere, a clear increase in INP concentrations at HPB is seen. For Fig. e and f the fraction of heat-labile INPs is negligible at Eri and therewith even lower than before. But when HPB is in the free troposphere, there is a high heat-labile INP fraction. For MEL, which is always located in the PBL, INP spectra are always at the high end of observed INP concentrations. However, in Fig. f INP concentrations from HPB start to resemble those from MEL. Interestingly, these INP concentrations for HPB show strong similarity to the warm Bise situation in Fig. h and i.

The described results may indicate the presence of similar INP concentrations over a wide area, possibly emitted from snow-free regions upwind of HPB and even further upwind from MEL. Potentially these emissions were present sufficiently long to reach the free troposphere. In this context, it is interesting to note that MEL and a broad region north of MEL, extending over most of northern Germany and Poland, was not covered by clouds (see example from 8 January 2024 in the Supplement). Once a Bise cloud forms, INPs are removed quickly from the cloud, resulting in the low INP concentrations presented above. Of course, INP sources between HPB and Eri must also be taken into account. However, the region in between both sites was snow-covered and temperatures were below freezing which keeps the biological activity at a minimum. Therefore, additional INP sources are considered to be of minor importance.

In February, a warm Bise situation is observed. INP spectra are shown in Fig. h and i. Both Eri and HPB are within the stratus cloud and the cloud-top temperatures are not supercooled. INP concentrations at both locations are generally higher than during the cold Bise, and very similar to the respective values measured at MEL. The higher temperatures during the warm Bise do not allow for INP activation in clouds. Still, this similarity between the three locations is remarkable, as they are several hundred kilometers apart. Concerning heat-labile biological INPs that are active above -10 °C, we observe elevated INP concentrations in all unheated samples during the warm Bise situation, regardless of sampling location. This manifests itself as bump-like structures in the INP spectra. Similar bumps were observed for all Melpitz INP spectra shown in Fig. , and for the INP spectra from HBP while it was in the free troposphere during the cold Bise (Fig. d to f), but not for the cold Bise case, when Eri and HPB were within the PBL (Fig. a to c). In comparison to that INP spectra from Eri show that feature only in the warm Bise case (Fig. g to i). These elevated concentrations vanish for all heated samples and were hence comprised of proteinaceous, biological INPs. Calculating the heat-labile fraction in the temperature range relevant for biological INPs is hindered by the fact that all INP spectra for heated samples only start below -10 °C.

The loss of INPs upon heating above -10 °C was most pronounced. But when looking at Fig. , it can be seen that INP concentrations were also lowered at temperatures below -10 °C at all three stations and for all cases. This shows the presence of proteinaceous, biological INPs across the whole temperature range for which data for heated samples was obtained. In the temperature range below -10 °C, the relation between the fraction of heat-labile INPs and sampling location and time followed a different trend than above -10 °C. The smallest loss occurred for Eri during the cold Bise case while HPB was in free troposphere, with a decrease in INP concentrations by roughly 30 %–40 %. Higher losses were observed for all other cases at all three stations.

In summary, the presence of potentially biological INPs is noteworthy so early in the year, particularly for INPs active above -10 °C as they already were present in early January at MEL and at HPB when it was in the free troposphere.

Figure presents a zoomed view of the RPG94 cloud radar reflectivity measurements covering the time period from 7 to 12 January 2024 from the surface to 1.5 km above ground level. Temperature profiles from two radiosonde launches are also provided. In the lowest 300 m, temperatures range from -5 to -10 °C, with warmer air observed above the inversion on 10 and 12 January 2024. The Bise cloud was affected by natural cloud seeding (meaning that ice crystals generated by a higher cloud fell into the Bise cloud system) on 7 January which continues until approximately 8 January, 09:00 UTC see. Additionally, sedimenting ice crystals from upper-level cloud layers are visible in the evening of 9 January 2024.

The VISSS particle images (Fig. a1–a3) illustrate three periods of observations. The first period (Fig. a1, 8 January 2024, 03:23–03:59 UTC) shows ice crystals influenced by strong seeding from higher-altitude clouds, resulting in a diverse mixture of ice crystal shapes. Rimed particles, needles and aggregates are visible. The ice crystal number concentration (ICNC) at the surface observed by the 2DVD is around 10–100 L⁻¹. During the second period (Fig. a2, 8 January 2024, 13:42–14:42 UTC), heavily rimed dendrites are observed. These particles initially form at cloud-top temperatures slightly colder than -10 °C and become heavily rimed as they fell through the approximately 1 km thick cloud layer, so they appear almost spherical. The ICNC was around 0.1 L⁻¹ during this period. In the third period (Fig. a3, 11 January 2024, 13:00–14:57 UTC), ice crystals form at cloud-top temperatures around -5 °C, resulting in needle-shaped particles observed at ground level. The ICNC is around 0.01 L⁻¹ during this period.

The ICNC at Eri between 8 and 12 January 2024 measured with 2DVD is shown in Fig. c. In the first half of 8 January 2024, seeder-feeder events caused the highest values in the entire period of around 1 to 100 L⁻¹. Without seeding, the ICNC decreased from 1 to between 0.001 and 0.1 L⁻¹. In the evening of 9 January 2024 another seeding event caused increased ICNC of up to 10 L⁻¹. On 11 and 12 January 2024 the values were typically between 0.0001 and 1 L⁻¹. In general, a clear trend of decreasing ICNC during times without seeding is visible, from around 1 to 0.003. At the same time the Bise cloud thickness decreased from 1000 to 300–400 m and the temperatures in the cloud increased from around -10 to around -5 °C.

The evolution of the observed ice crystals presented in Fig. illustrates that light precipitation was recorded by VISSS and the 2DVD throughout the observation period of the cold Bise situation, indicating continuous availability of at least a few INPs. However, from the remote-sensing and in situ measurements at Eri it can be seen, that concentrations of ice crystals on ground were higher than the available INP concentrations at Eri. At this site, measurable INP concentrations are only recorded at temperatures below the cloud minimum temperatures of around -7 °C (compare with Fig. and the cold Bise INP measurements in Fig. b, c) that prevail during the period of observation. Specifically, mean ICNC were between 10-3 and 10-2 L⁻¹ from 11 through 12 January 2024, when cloud-minimum temperatures were about -7 °C. It is, however, remarkable that the INP observations of around 10-3 L⁻¹ at HPB, when this site was above the PBL (Fig. e, f), were on a similar order as the ICNC concentrations observed during the same period at Eri. This supports the hypothesis that INPs are entrained from the free troposphere via turbulence and afterwards immediately removed as they interact with the Bise cloud layer, leading to reduced availability of INPs downwind. However, it must be noted that an ICNC concentration which is higher than the observed INP concentration can in principle also be a result of secondary ice formation processes . Nevertheless, secondary ice formation processes generally lead to orders of magnitudes of increase in ICNC, which was, besides occasional peaks in the ICNC, not observed in the average ICNC values during the investigated time periods.