Methodology:

At East Beach hut, in Rothera, 30 Nuclepore Hydrophilic Membrane filters (see details below) were loaded in to filter holders for the magazine of the DIGITEL DPA14 LoVol automatic sampler. To avoid contamination, the filter holders and materials were cleaned with IPA and lint free wipes and the work was conducted inside a laminar flow hood. The DIGITEL DPA14 was set to change filters every 48 hours, and to sample at a flow rate of 15 Litres per minute. For each batch of 30 filters, at least 1 filter was not exposed to air and used as a handling blank to assess background contamination. Once the DIGITEL DPA14 had run through all of the filters, the magazine and filter holders were emptied in the laminar flow hood and each filter was packed in a separate, numbered plastic bag. The filters were stored and shipped to the British Antarctic Survey in Cambridge at -20C.

At the British Antarctic Survey the filters were put in 10 mL HPLC vials (P/N 055058) with 4 mL of purified water (18.2 MΩ cm at 25C, 0.22-μm-filtered) produced by a Milli-Q water purification system (Millipore, Watford, UK) and shaken for 20 minutes on an automatic shaker to create suspensions of purified water with particles which were washed off the filters. These were then put on an automatic stirrer and 1 μL droplets of the suspension were pipetted onto hydrophobic glass slides (see below) which had been pre-cleaned with IPA and lint-free paper. The hydrophobic glass slides were put on a Peltier cooling stage (details below) with squalene between the glass slide and the stage for thermal conduction. An in-house built plastic separator physically separated the droplets to prevent impacts from the Wegener-Bergeron-Findeisen effect. Another glass slide was put on top for this same reason, and the stack (slide-seperator with droplets-slide) was kept in place with an in-house built plastic holder. The Peltier stage was then set to cool at the constant rate of 1C/minute to -40C until all the droplets were frozen. By monitoring the droplets on the Peltier cooling stage with a camera, the temperature at which each droplet froze could be measured. The image taking, automatic identification of droplet freeze-up from the images and linking of the timing of each droplet freeze-up to the corresponding Peltier cold stage temperature, was done using in-house Python software (OpenCV and Keras packages)

Data collection:

Sample collection: DIGITEL DPA14 automatic filter sampler, Nuclepore Hydrophilic Membrane, 0.2 μm pore size, 47 mm track edged filters (10417012).

Sample analysis: For methodology see: Budke and Koop, 2015.

cooling: Linkam LTS120 with T95-LinkPad temperature controlled stage, Julabo F250 recirculating cooler.

set-up: Marienfeld (MARI0895040) hydrophobic glass slides, Sigma-Aldrich (S3626-100ML) squalene. Microsoft LifeCam webcam. In-house built droplets spacer, holder and lighting

Software: (picture taking and processing) in-house Python software

Data quality:

Each batch of 30 automatically collected filters from the DIGITEL DPA14 contained at least one filter which had not been exposed to ambient air and was therefore kept as a handling blank. The median temperature for the handling blanks was -26.8C and we therefore consider -26C as a reasonable cut-off above which we can distinguish ice nucleating activity on our filters.

We also ran Milli-Q blanks on each day before analysing filter samples to assess any background contamination in the lab or in the Milli-Q water system. We would not run any samples if the temperatures for this blank were above the handling blank temperature of -26C.

This dataset has not been corrected for background contamination, as this would lead to binning of the data and loss of resolution. Furthermore, background correction was not necessary for the publication related to this dataset.