Some of the most important particles in the atmosphere are also among the hardest to count. Ice-nucleating particles, the tiny specks of dust, pollen fragments, and biological molecules that trigger the formation of ice in clouds, shape rainfall, lightning, and the way clouds reflect sunlight. But a new laboratory study from the University of Leeds suggests that one of the most widely used methods for measuring these particles may be quietly losing a crucial fraction of them, simply because the molecules stick to the sampling filter and never come off.
The study, published in the journal Aerosol Research by Joseph Robinson, Benjamin Murray, and colleagues at the Institute for Climate and Atmospheric Science, focused on a technique known as the wash-off method. In this approach, air is drawn through a polycarbonate membrane filter that captures aerosol particles. Back in the laboratory, the filter is dropped into ultrapure water and agitated on a rotary mixer for half an hour, washing the collected particles into suspension. Droplets of that suspension are then pipetted onto a chilled surface and cooled at a controlled rate, and the temperatures at which individual droplets freeze reveal how many ice-nucleating particles the air contained. It is an elegant, sensitive, and inexpensive method, which is why it underpins a large share of the world’s atmospheric ice-nucleating particle datasets.
The Leeds team suspected a flaw. Proteins are notorious for adsorbing to surfaces, a phenomenon extensively documented in biophysics and membrane filtration. Once a protein adheres to a surface, the binding is often effectively irreversible, and a simple rinse with water will not remove it. Polycarbonate filters used in the wash-off technique are treated with polyvinylpyrrolidone to make them hydrophilic and reduce protein adsorption, but that treatment reduces rather than eliminates the problem. On hydrophilic surfaces, adsorption proceeds through hydrogen bonding between polar amino acid residues on the protein and polar groups on the surface. If ice-nucleating proteins, the remarkably efficient molecules produced by certain bacteria and fungi that can trigger freezing just below zero degrees Celsius, were sticking to the filter during the wash, they would vanish from the measurement entirely.
To test the idea, the researchers built a controlled experiment in a 0.729 cubic metre aluminium aerosol chamber. They aerosolised a series of well-characterised materials, including a commercial bacterial ice-nucleating product called Snomax, derived from the plant-pathogenic bacterium Pseudomonas syringae; two soil dusts, one from the Copper River delta in Alaska and one from agricultural land near Leeds; lichen and birch pollen filtrates; and a sea surface microlayer sample collected off Greenland. Each aerosol was sampled simultaneously through two parallel filter samplers, one loaded with a polycarbonate filter for the wash-off technique and one with a polytetrafluoroethylene filter for a more direct approach called the drop-on method, in which water droplets are pipetted straight onto the filter surface without any washing step. For some experiments, a third instrument, the Portable Ice Nucleation Experiment, or PINE, sampled the same air online, creating a miniature cloud inside an expansion chamber and counting the ice crystals optically.
The results were striking, and they depended entirely on the physical state of the ice-nucleating material. When the bacterial proteins were present as free macromolecules, isolated from larger particles by filtration before aerosolisation, the wash-off technique produced freezing spectra roughly nine degrees Celsius colder than the drop-on technique, corresponding to a dramatic underestimation of ice-nucleating activity. The PINE instrument agreed with the drop-on measurements, not the wash-off. Even more tellingly, when the researchers took punches of the washed polycarbonate filters and tested them directly in a multiwell freezing assay, they found highly active ice-nucleating material still clinging to the filter, even after a second, more vigorous wash. The proteins had adsorbed to the membrane and were simply never recovered.
The picture changed when the ice-nucleating proteins were anchored to larger particles. In dry-dispersed Snomax, where proteins ride on cell fragments, and in the two soil dusts, where fungal ice-nucleating proteins are internally mixed with mineral grains, the wash-off and drop-on techniques agreed reasonably well. The researchers explain this with simple physics: large particles experience enough hydrodynamic drag during the washing process to overcome the adhesive forces holding them to the filter, so they detach and enter the suspension. Small macromolecules, by contrast, experience far less drag and remain bound. The mixing state of the ice-nucleating material, whether it is free or attached to a carrier particle, therefore determines whether the wash-off method works or fails.
Not every biogenic macromolecule behaved the same way. Pollen filtrate, whose ice-nucleating entities are thought to be polysaccharides rather than proteins, showed good agreement between the two filter techniques, with both methods detecting activity within an order of magnitude below minus fourteen degrees Celsius. Yet even here, the washed-filter analysis revealed that a substantial amount of ice-nucleating material remained on the polycarbonate membrane after washing. Polysaccharides do adsorb to surfaces, a well-known cause of membrane fouling in water treatment, but their binding appears more reversible than that of proteins, and the steep freezing spectrum of pollen means that partial recovery is enough to produce a reliable measurement. The sea surface microlayer sample, a complex mixture of saccharides, amino acids, lipids, and proteins, showed the most erratic behaviour of all, with the wash-off results varying by more than an order of magnitude between experiments while the drop-on results varied by only about a factor of five, hinting at a combination of adsorption losses and aggregation in the suspension.
These laboratory findings cast new light on a nagging problem in the field literature. For years, researchers have reported discrepancies between measurement techniques that were hard to explain. An aircraft campaign over south-eastern England found that wash-off and drop-on measurements disagreed forty percent of the time, with wash-off often detecting up to an order of magnitude fewer ice-nucleating particles at temperatures warmer than minus twenty-two degrees. Wildfire plumes over the United States showed wash-off deficits of up to two orders of magnitude compared with an online continuous flow diffusion chamber. Ground-based measurements in Svalbard found condensation-on-filter results up to a factor of eight higher than wash-off at minus fifteen degrees, and a comparison in the eastern North Atlantic found the wash-off technique reporting over an order of magnitude fewer ice-nucleating particles per litre than PINE at minus twenty degrees. The Leeds experiments now suggest a unifying explanation: wherever free, unattached ice-nucleating macromolecules dominate the aerosol, the wash-off technique undercounts them.
The implications ripple outward into climate science. Global models of mixed-phase clouds rely on parameterisations of ice-nucleating activity built from field measurements, and many of those measurements were made with the wash-off technique. A widely used parameterisation for the ice-nucleating ability of sea spray aerosol, for instance, is based on wash-off data that may under-report the true activity. If atmospheric ice-nucleating particle concentrations in some regions are higher than current models assume, then projections of cloud phase, cloud lifetime, and radiative balance carry a hidden bias. The authors caution that no universal correction factor can rescue historical datasets, because the composition and mixing state of the sampled particles were rarely characterised. Instead, they recommend treating past wash-off measurements as a lower bound on ice-nucleating particle concentration unless independent evidence suggests otherwise.
The study does not condemn the wash-off technique outright. Wash-off suspensions remain valuable for diagnostic treatments, such as heating samples to identify the biological contribution to ice nucleation, and the smooth surface of polycarbonate filters lends itself to electron microscopy of individual particles. The authors suggest that future field campaigns should run the drop-on method, or another direct technique, in parallel with wash-off sampling. When the two agree, the wash-off data can be trusted; when they diverge, the disagreement itself becomes a signal, revealing the presence of free ice-nucleating macromolecules in the air mass. In the delicate business of counting the seeds of ice, it turns out that how you wash the filter matters as much as what the air contains.
Subject of Research: Measurement biases in atmospheric ice-nucleating particle quantification caused by adsorption of biogenic ice-nucleating macromolecules to sampling filters
Article Title: Challenges in measuring sticky biogenic ice-nucleating macromolecules
Article References: Robinson, J., Daily, M. I., Foster, P. B., Macklin, J. P., McQuaid, J. B., Tarn, M. D., & Murray, B. J. (2026). Challenges in measuring sticky biogenic ice-nucleating macromolecules. Aerosol Research, 4(2), 373-396. https://doi.org/10.5194/ar-4-373-2026
Image Credits: AI Generated
Keywords: ice-nucleating particles, biogenic aerosols, ice-nucleating proteins, wash-off technique, droplet freezing assay, polycarbonate filters, mixed-phase clouds, Snomax, pollen polysaccharides, sea surface microlayer, soil dust, aerosol measurement
News Source: Russell Cooper. (October 9, 2026). Sticky Proteins Make Cloud Ice Measurements Slip Through the Filter. Scienmag.
