The study, led by Theresa Mathes of the Technische Universität Berlin and the Leibniz Institute for Tropospheric Research, together with colleagues from the University of Leeds, Stockholm University, Uppsala University and TROPOS in Leipzig, was carried out as part of the ARTofMELT expedition, which sailed from Svalbard between 8 May and 15 June 2023. The campaign’s broader goal was to understand how atmospheric rivers, narrow corridors of warm, moist air, trigger the onset of sea ice melt. But the flux measurements collected along the way, published in the journal Atmospheric Chemistry and Physics, provide one of the longest continuous eddy covariance records of particle exchange ever obtained over the Arctic Ocean, and they carry implications that reach well beyond the expedition’s original purpose.

Aerosol particles matter enormously in the Arctic. They scatter and absorb sunlight directly, and they act as cloud condensation nuclei and ice nucleating particles, shaping the formation, brightness and lifetime of clouds. Because particle concentrations in the polar boundary layer are often very low, particularly in autumn, even small changes in the particle budget can ripple through cloud microphysics and the regional radiation balance. The Arctic is warming two to seven times faster than the global average since 1979, a phenomenon known as Arctic amplification, and shifts in the sources and sinks of aerosols are one of the feedbacks that models struggle to capture. Yet direct observations of turbulent particle fluxes over sea ice have been strikingly scarce, with most previous ship-based campaigns lasting only days to a couple of weeks.

The technical challenge of making these measurements from a moving ship is considerable. Eddy covariance, the gold-standard technique, requires capturing rapid fluctuations in vertical wind speed and particle concentration simultaneously, at sampling rates of around 10 hertz. The team mounted a sonic anemometer 20.3 meters above the surface on Oden’s foremast, alongside a heated inlet feeding a mixing-type condensation particle counter capable of detecting particles down to 7 nanometers. Because the instrument sampled air through a long tube, the researchers had to correct carefully for the time lag between wind and concentration fluctuations, for particle losses to the tubing walls, and for the motion and flow distortion of the ship itself, using motion sensors and computational fluid dynamics corrections developed for Oden. Ship exhaust contamination, an ever-present hazard on a vessel, was rigorously screened out, removing 17 percent of the data.

To interpret the fluxes, the team classified the surface upwind of the ship into three categories: closed ice, leads, and open water. This was done automatically using a machine-learning image segmentation algorithm, a random forest classifier, applied to hourly photographs taken by cameras on the ship’s decks, with the relevant camera chosen according to the prevailing wind direction. The approach allowed the researchers to attribute each 20-minute flux measurement to the type of surface it had sampled, something few previous studies have done systematically.

The headline result is a study in contrasts. Under low to moderate turbulence, closed ice behaved as expected: it was a net sink for particles, with a median deposition flux of −0.02 × 10⁶ particles per square meter per second, and deposition dominated roughly 70 percent of the intervals measured over ice. But when the wind picked up, the picture reversed. During eight high wind speed events identified during the campaign, five showed net particle emission, and during the longest of these, a storm on 24 to 26 May with winds reaching 19 meters per second, the team recorded net emission fluxes of up to 0.98 × 10⁶ particles per square meter per second, a statistically robust figure with a 95 percent confidence interval excluding zero. The inferred emission strength, calculated by subtracting a modeled deposition flux from the measured net flux, peaked at 1.57 × 10⁶ particles per square meter per second during a second storm at the end of May.

The most likely culprit is blowing snow. Strong winds lift snow grains from the surface, and because Arctic snow contains sea salt, sulfate, nitrate, organic matter and even black carbon, the sublimation of these wind-blown particles releases aerosols into the boundary layer. The particle size distributions measured during the storm events support this interpretation: elevated concentrations appeared across a broad range of diameters, roughly 30 to 300 nanometers, rather than showing the narrow nucleation-and-growth signature expected from new particle formation driven by precursor gases. The researchers also found that the strength of the particle flux correlated strongly with friction velocity, a measure of near-surface turbulence, with a correlation coefficient of 0.88, reinforcing the link between turbulent wind stress and particle exchange. The threshold wind speed for blowing snow is not fixed; it depends on temperature and snow properties, and at low temperatures mobilization could begin at winds as gentle as 6 meters per second.

Over leads and open water, the story was more nuanced. Both surfaces showed a mixture of emission and deposition events, with median fluxes close to zero, suggesting that while open ocean and cracks in the ice can act as particle sources, through mechanisms such as bubble bursting and sea spray, their net contribution during the campaign was modest. The researchers note that whitecaps in pack ice regions are smaller and less frequent than in the open ocean, which may explain the weak net emission observed. They also caution that open water periods were underrepresented in the filtered dataset because the ship was typically underway when open water was present, and surface classification was restricted to periods of slow drift.

What makes these findings consequential is the trajectory of the Arctic itself. Climate models project continued sea ice loss and a corresponding expansion of open water, which is expected to change the frequency and intensity of the particle exchange processes documented here. As the ice retreats, local particle sources, whether from leads, open water or wind-blown snow, are likely to play a growing role in the regional aerosol budget, with knock-on effects for cloud formation and the surface energy balance. Whether the net climatic effect is cooling, through brighter clouds reflecting more sunlight, or warming, through changes in cloud lifetime and phase, remains poorly constrained, and competing source and sink processes make the outcome genuinely uncertain.

By providing hard observational constraints on particle fluxes over three distinct surface types, the ARTofMELT dataset gives modelers something they have long lacked: direct numbers against which to test deposition schemes and emission parameterizations in the High Arctic. The team’s data are publicly available through the PANGAEA repository. As the researchers conclude, quantifying particle deposition on closed ice and emissions from leads and storm-whipped snowfields is a step toward reducing the stubborn uncertainties that still surround Arctic aerosol-cloud-climate interactions, in a region where the stakes of getting the feedbacks right could not be higher.

Subject of Research: Turbulent aerosol particle exchange between the atmosphere and sea ice, leads and open water in the High Arctic

Article Title: Turbulent atmospheric particle fluxes in the High Arctic: surface type dependence and variability

Article References: Mathes, T., Brooks, I., Murto, S., Tjernström, M., Kojoj, J., Zieger, P., Prytherch, J., Wehner, B., & Held, A. (2026). Turbulent atmospheric particle fluxes in the High Arctic: surface type dependence and variability. Atmospheric Chemistry and Physics, 26(19), 14507-14530. https://doi.org/10.5194/acp-26-14507-2026

Image Credits: AI Generated

DOI: 10.5194/acp-26-14507-2026

Keywords: Arctic, aerosol particles, eddy covariance, sea ice, blowing snow, particle fluxes, cloud condensation nuclei, Arctic amplification, leads, sea spray aerosol, ARTofMELT, dry deposition

A new study published in Atmospheric Chemistry and Physics by Aiden Jönsson of Stockholm University and colleagues tackles this identification problem head-on. The team built a comprehensive reference library of the fluorescence and optical morphology signatures of coarse aerosols from major sources — pollen, marine bacteria, dust, cellulose, and microplastics — measured in controlled laboratory experiments with a Multiparameter Bioaerosol Spectrometer, or MBS. They then trained machine learning models on these data and tested the resulting classification algorithm against a full year of field observations from the Zeppelin Observatory in Svalbard, using independent chemical tracer measurements as a reality check. The result is an open-source framework that improves bioaerosol detection while candidly exposing the limits of fluorescence-based identification.

The MBS is a single-particle ultraviolet light-induced fluorescence (UV-LIF) spectrometer developed to detect bioaerosols in ambient air. Each particle passes first through a low-power laser for sizing via Mie scattering theory, then through a high-power pulsed laser whose scattered light is recorded by two 512-pixel CMOS arrays positioned on either side of the beam. These arrays capture chords across the particle’s diffraction pattern, yielding statistical descriptors — mean signal, variance, skewness, kurtosis, peak count, peak width, and several asymmetry measures — that serve as proxies for particle shape and surface roughness. Finally, a xenon flash lamp emitting 280-nanometer light excites fluorescence, which is dispersed across eight detection channels spanning roughly 305 to 655 nanometers. The instrument can measure particles from 0.5 to about 20 micrometers at rates up to roughly 160 particles per second.

Fluorescence works as a biological marker because fluorophores such as the amino acid tryptophan and the coenzyme riboflavin are abundant in living material. But the technique has a well-known Achilles’ heel: many non-biological particles fluoresce too. Polycyclic aromatic hydrocarbons from biomass and fossil fuel combustion are highly fluorescent and can coat soot particles, causing them to glow. Plastic polymers fluoresce due to aromatic groups in their base materials or additives, or fluorophores produced during aging. The new laboratory characterization quantified these confounding similarities in detail — and found some striking examples.

Among the biological samples, bacteria cultured from Baltic seawater produced a distinctive fluorescence peak in the channel centered near 364 nanometers, consistent with tryptophan emission, and this feature proved to be a robust marker unique among the samples tested. Pollen fragments, by contrast, showed broader spectra peaking near 414 nanometers when dry, shifting toward 461 nanometers when wet-nebulized — evidence that moisture and aerosolization method can substantially alter pollen fluorescence, with implications for detecting pollen fragments produced by humid atmospheric processes such as thunderstorms. Most provocatively, freshly ground polyethylene and pure cellulose both produced fluorescence spectra closely resembling those of dry pollen fragments, meaning synthetic particles and plant debris can masquerade as bioaerosols in fluorescence-only measurements. Combustion particles from ship exhaust, observed during Arctic Ocean cruises, showed broad, intense fluorescence, and about 10 percent met the standard fluorescence-based criterion for biological particles.

To move beyond these ambiguities, the researchers built a supervised machine learning algorithm with three tasks: flag fluorescent particles that are likely combustion-derived interferents, classify fluorescent biological particles into broad subgroups (pollen fragments, bacteria, and fungal spores), and distinguish non-fluorescent dust-like from sea-spray-like particles. The interferent and dust tasks used logistic regression models, while the biological classification combined uniform manifold approximation and projection (UMAP), a dimensionality-reduction technique, with a k-nearest neighbors classifier operating in the transformed space. In testing, the pollution model identified combustion particles with 92 percent precision, and the dust model — trained only on particles larger than 2.5 micrometers, where dust and sea spray scattering signatures separate most cleanly — achieved 99 percent precision for both classes. Yet the multiclass model still misclassified roughly 12 percent of combustion particles as pollen fragments, reflecting genuine overlaps in both fluorescence and irregular morphology.

The most revealing test came when the algorithm was applied to real observations from the Zeppelin Observatory during 2020, part of the Ny-Ålesund Aerosol Cloud Experiment campaign. In its raw form, trained purely on laboratory data, the algorithm failed: its biological particle estimates correlated strongly with levoglucosan and equivalent black carbon, tracers of combustion, rather than with biological markers like arabitol and mannitol. The problem, the authors argue, is a domain gap — laboratory-generated aerosols differ from their real-world counterparts, and important particle classes such as biomass burning aerosols and plant debris were missing from the training data entirely.

The solution was domain adaptation, a transfer learning technique in which a model trained in one domain is adjusted for use in another. Using chemical tracer concentrations, the team constructed soft, continuous labels describing the degree of pollution versus biological influence for particles observed at Zeppelin, then further trained the models on these data with deliberately limited weight — capped at 20 percent of the laboratory training influence — to avoid overfitting to uncertain labels. The tuned algorithm transformed the results. It reproduced the previously published annual bioaerosol cycle, with near-zero biological particle concentrations in winter and early spring, a July peak of roughly 10 particles per liter, and a secondary autumn peak that coincided with elevated warm-activating ice-nucleating particles. In August, when biological tracer concentrations were highest, the machine learning approach detected fluorescent biological particle concentrations nearly two orders of magnitude higher than the older fluorescence-only decision tree method — levels consistent with independent offline analyses from Ny-Ålesund.

The study is equally instructive about what remains unresolved. The tuned classifier attributed most biological particles to the pollen fragment class, yet intact pollen counts in Svalbard are known to be very low, suggesting that plant debris or combustion particles with broad fluorescence spectra were likely misidentified as pollen-like. The dust model, before tuning, labeled about 80 percent of ambient non-fluorescent particles as dust — implausible given that total coarse particle concentrations correlated with sea salt tracers, not mineral dust tracers. Ambient sea spray particles, modified by atmospheric aging, apparently resemble laboratory dust more than freshly generated laboratory sea spray. These mismatches underscore that laboratory aerosolization methods can bias particle properties and that naturally aged aerosols may behave quite differently from their nascent counterparts.

The broader lesson is methodological humility paired with practical progress. No UV-LIF instrument configuration or classification scheme, the authors conclude, will fully eliminate non-biological interference, and no single method can establish definitive ground truth for bioaerosol concentrations. Parallel measurements of black carbon and source-specific chemical tracers remain essential for validating any fluorescence-based quantification. Still, the framework offers a flexible, expandable foundation: it explicitly incorporates morphology alongside fluorescence, accounts for a wider range of particle classes, and can be retrained as new characterization data arrive — particularly from biomass burning experiments, real plant debris, and studies of atmospheric aging. With the code and data openly available, the approach could be applied to reanalyze existing bioaerosol datasets worldwide, bringing climate models one step closer to accurately representing the tiny biological particles that punch far above their weight in Earth’s cloud system.

Subject of Research: Machine learning classification of coarse atmospheric aerosols using single-particle fluorescence and optical morphology

Article Title: Tracing biological, anthropogenic, and inorganic sources of coarse aerosols via single-particle fluorescence and optical morphology

Article References: Jönsson, A., Fu, J., Pereira Freitas, G., Crawford, I., Dagsson-Waldhauserová, P., Krejci, R., Tobo, Y., Yttri, K. E., & Zieger, P. (2026). Tracing biological, anthropogenic, and inorganic sources of coarse aerosols via single-particle fluorescence and optical morphology. Atmospheric Chemistry and Physics, 26(18), 13617-13643. https://doi.org/10.5194/acp-26-13617-2026

Image Credits: AI Generated

DOI: 10.5194/acp-26-13617-2026

Keywords: bioaerosols, coarse-mode aerosols, UV-LIF spectroscopy, fluorescence, machine learning, domain adaptation, mineral dust, sea spray aerosol, pollen fragments, Arctic, Zeppelin Observatory, ice nucleation

The research, published in National Science Review, was led by Prof. Wei Xu of the Institute of Urban Environment and Prof. Ru-Jin Huang of the Institute of Earth Environment, both of the Chinese Academy of Sciences, together with Prof. Jurgita Ovadnevaite of the University of Galway. The team reports the first size-resolved observational constraint on the number size distribution of marine primary organic aerosols, achieved through long-term aerosol mass spectrometry, hygroscopicity measurements and machine-learning-based source apportionment. The measurements were carried out at the Mace Head Atmospheric Research Station on the west coast of Ireland, a location chosen for its frequent exposure to clean marine air masses arriving from the open North Atlantic, far from continental pollution sources that could confound the signal.

The reason this measurement matters so much lies in a fundamental distinction between mass and number. Marine primary organic aerosols have been characterized in terms of their mass for roughly two decades, and models have generally inferred their number concentrations indirectly from those mass estimates. But cloud droplet number concentrations, and therefore the brightness or albedo of marine clouds, depend on how many particles are available in the atmosphere rather than on how much material they collectively contain. A population of many small particles can seed far more cloud droplets than an equal mass concentrated in a few large ones. Without direct number measurements, the models were essentially guessing at the particle population that governs cloud formation over much of the planet.

Isolating the organic fraction from the rest of the sea spray required an elegant physical trick. At 90 percent relative humidity, sea salt and sulphate particles absorb water readily and grow substantially, whereas marine primary organic aerosol particles are considerably less hygroscopic and swell far less. The researchers exploited this contrast using a humidified tandem differential mobility analyser, an instrument that classifies particles according to how much they grow in humid air. Combined with concurrent measurements of particle size, number and chemical composition, this approach allowed the team to resolve the mPOA number size distribution down to an unprecedented 35 nanometres, a size range directly relevant to cloud droplet activation.

What emerged from the data was a distribution with two distinct modes rather than one. The measured size distribution is bimodal, with a smaller mode centred near 60 nanometres and a broader mode spanning roughly 100 to 200 nanometres. By contrast, the parameterization embedded in several major climate models produces a unimodal distribution centred at 120 nanometres, with no representation of the smaller mode at all. That omission is not a minor detail. Particles near 60 nanometres are precisely the sizes that can activate into cloud droplets at the moderate supersaturations typical of marine boundary layer clouds, so an entire population of potential cloud condensation nuclei was effectively invisible to the models.

The quantitative consequences are substantial. For particles larger than 35 nanometres, the model parameterization underestimates the total mPOA number concentration by approximately a factor of three. In the 30 to 70 nanometre range, the underestimate grows to between fivefold and sevenfold. Translated into cloud condensation nuclei, the mPOA contribution is underestimated by about 16 percent at a supersaturation of 0.25 percent and by as much as 64 percent at 0.5 percent supersaturation, assuming a 30 percent reduction in surface tension caused by the organic material. Surface tension matters here because organic films on particles can lower the energy barrier for droplet formation, allowing smaller particles to activate than would otherwise be possible. The affected models include GEOS-Chem, UKESM and CMAQ, which infer mPOA number from mass, as well as E3SM and CESM2, which represent mPOA without any size-resolved description.

To gauge what this means for the climate system, the researchers combined satellite observations of cloud properties with simulations from the GEOS-Chem model. As an order-of-magnitude estimate, the additional particles correspond to a first aerosol indirect effect of approximately -0.03 to -0.04 watts per square metre when averaged over the global ocean. Over biologically productive waters, where mPOA emissions are highest because phytoplankton activity is most intense, the effect is substantially larger, reaching -0.1 to -0.2 watts per square metre once the surface tension reduction is taken into account. These are cooling effects, since more cloud droplets produce brighter, more reflective clouds that bounce more sunlight back to space. While the global average figure may appear modest, it is concentrated over some of the most climatically sensitive ocean regions on Earth.

The authors are careful to frame these radiative estimates as sensitivity calculations rather than definitive global assessments. The figures quantify the magnitude of the mPOA influence that would obtain if the size distribution derived at Mace Head were representative of the global ocean, which is an assumption the data cannot yet confirm. Because the observations come from a single site, regional differences in phytoplankton community composition, sea state and atmospheric ageing may modify the size distribution elsewhere. Even so, the environmental conditions sampled at Mace Head were broad: chlorophyll-a concentrations ranged from approximately 0.1 to 10 milligrams per cubic metre and wind speeds from 5 to 20 metres per second, encompassing a substantial portion of the parameter space that characterizes the global ocean, although not all of it.

The implications extend well beyond correcting a single parameterization. Aerosol-cloud interactions remain among the largest sources of uncertainty in climate projections, and the total aerosol forcing that has masked part of greenhouse warming depends critically on getting natural aerosol backgrounds right. An underestimate of a natural aerosol source does not simply introduce a local error; it propagates into estimates of anthropogenic forcing, because the radiative effect of human-emitted particles is calculated relative to a preindustrial baseline in which marine organics played a role. If models have been undercounting the natural cloud-seeding particles over the oceans, the historical trajectory of aerosol cooling, and by extension the sensitivity of the climate system to greenhouse gases, may need refinement in the marine sector.

The findings also carry practical weight for emerging proposals to brighten marine clouds deliberately as a means of cooling the planet. Marine cloud brightening schemes depend on an accurate characterization of the natural marine aerosol baseline, since the effectiveness of adding particles to low-lying ocean clouds hinges on how many particles are already there and how close the cloud is to saturation with condensation nuclei. A baseline underestimated by a factor of three fundamentally changes those calculations. With the first direct size-resolved measurements now in hand, climate modelers have a concrete observational target to correct their parameterizations against, and atmospheric scientists have a clear agenda: extend such measurements to other ocean basins, different plankton communities and varying sea states, to determine how universal the bimodal signature observed over the North Atlantic truly is. The ocean’s influence on clouds, it turns out, is richer and more finely structured than the models assumed.

Subject of Research: Size-resolved measurement of marine primary organic aerosols and their role as cloud condensation nuclei in climate models

Article Title: Marine organic particles that seed ocean clouds may be three times more abundant than climate models assume

Article References: Marine organic particles that seed ocean clouds may be three times more abundant than climate models assume. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: marine primary organic aerosol, sea spray aerosol, cloud condensation nuclei, aerosol-cloud interactions, climate models, Mace Head, North Atlantic, bimodal size distribution, aerosol indirect effect, marine cloud brightening, phytoplankton, surface tension