Dark microplastics trigger changes on snow metamorphism that depends on the snow initial density

Summary

Cryospheric regions are no exception to microplastic ubiquity. Still, microplastic's capacity to decrease snow albedo or advance snow melting, as light-absorbing impurities, remains unexplored. This study assesses the effect of dark microplastics on snow properties under realistic conditions. Six in situ experiments were conducted at the Central Pyrenees (Spain), exposing surface snow to different concentrations of dark micropellets for 4 h. Results were variable and dependent on snow initial conditions. In the experiments performed on recent, light snow (<250 kg m⁻³), increasing concentrations of microplastics yielded moderate decreases in albedo and high changes in snow specific surface area, reducing it by 11.4 m² kg⁻¹ as compared to blank samples, while snowmelt changes were <1%. On the contrary, in the experiments conducted on aged snow of high density (>450 kg m⁻³), high microplastic accumulation increased snowmelt 17% more than in the blanks. Further field studies are needed for a better understanding of the effect of microplastics on the global cryosphere.

Graphical abstract

Highlights

• •The effect of microplastics on snow metamorphism was assessed in field experiments
• •Results were variable and mostly dependent on snow initial conditions
• •The variable most affected by microplastics was the snow specific surface area
• •Toward the end of the snow season, microplastics enhanced snowmelt up to 17%

Earth sciences; Glacial processes; Environmental monitoring

Introduction

Microplastic particles (MP; <5 mm in any dimension) have already colonized every ecosystem on Earth, from the deepest part of the ocean¹ to the highest mountain.² Despite the elements of the cryosphere generally developing in remote mountainous and polar regions, they are no exception to microplastics' ubiquity. Back trajectories analysis in the Antarctic have shown that MPs, especially the smallest fraction (<250 μm) can be transported through the atmosphere at distances of up to 6000 km from their source,³ snowfall being considered an effective method of microplastic deposition.⁴ Apart from the Antarctic³ and the Artic,⁵ the presence of MP (at variable concentration levels) has been reported on glaciers in Iceland⁶ and the Alps,⁷ and seasonal snowpack in the Andes⁸ protected areas in Japan,⁹ the Himalayas,² National Parks in Western USA,¹⁰ northeastern China mountains,¹¹ and the Teide volcano in the Canary Islands.¹² Among the most abundant particles in these environments are polyesters (including PET, the most common type of polyester) and polyamide fibers.^2,5,12 However, the presence of fragments and pellets made of polyethylene (PE), ethylene-vinyl acetate, polyvinyl chloride (PVC), and (poly)urethane (PUR) is also common.^5,6,9,10,11 Reported particles display a wide range of physical characteristics in terms of sizes and colors, while the smallest fractions always prevail. Similarly, most of those studies reported a higher frequency of dark-colored MPs.^3,5,12

Any snow impurity, especially colored ones, can act as a light-absorbing particle (LAP), eventually increasing energy absorption, decreasing snow albedo, and triggering changes in snow metamorphism, as acknowledged for mineral dust and black carbon.^13,14,15,16 It has been recognized that the increase in black carbon deposition following the industrial revolution led to a reduction in the global Cryosphere, fostering the end of the Little Ice Age.¹⁵ Mid-latitude regions are particularly affected by the deposition of both black carbon and mineral dust, accumulating concentrations one to two orders of magnitude higher than those in the polar regions.¹⁷ The effect of different impurities in a given area should be further considered, since they may have a synergistic effect. For example, estimations of BC and mineral dust radiative forcing in the Tibetan Plateau have been estimated of around 43 W m⁻² for dust¹⁶ and between 2 and 20 W m⁻² and black carbon¹⁸ alone, while their combined effect reached a value of 145 W m^-2.16 Although less studied, the impact of microbial life on snow albedo has also been pointed.^13,16,19 Deep-learning emulator estimations have predicted variable albedo-reduction values among 180 surfaces, reaching reductions up to 0.13, 0.21, and 0.25 for red algae, mineral dust, and black carbon, respectively.²⁰

Several authors have suggested that MPs, which can partially contribute to carbonaceous aerosols, including black carbon, can exert radiative forcing and diminish the albedo of snow and ice surfaces.^4,21 Indeed, while the impact of MPs in cryospheric regions has been overlooked until recently, black carbon content in snow may have been overestimated, as well as their effects in the cryosphere, because the measurement techniques (optical absorption or pyrolysis-related) currently used to determine them don’t separate the signal of MPs from that of black carbon.²¹ According to Reynolds et al. (2024),²² the presence of many organic compound types common to tyres in snow suggested that atmospherically deposited black road-tyre-wear matter is among the LAPs that contributed the most to advance the onset and snow melt rate in the Colorado Rocky Mountains. Still, to our knowledge, the direct effect of MPs on snow metamorphism has not been assessed thus far. This work aimed to evaluate, for the first time, the effect of realistic concentrations of dark microplastic particles on snow properties. For this, in situ experiments were conducted at the Spanish Central Pyrenees employing a set of mini-lysimeters and parcels containing surface snow artificially aerosolized with different types and concentrations of MPs (Figure 1). We hypothesized that accumulations of dark MPs in the snow surface, similar to other dark impurities, would enhance light absorption and increase snow grain size, yielding a decrease in snow albedo and snow specific surface area (SSA), and an increase in liquid water content (LWC), eventually leading to accelerated snow melting. We expected these changes would be variable across the snow period, depending on snow characteristics (grain size, density, compaction) and meteorological conditions.

Figure 1.

Field experiments were conducted at the Spanish Central Pyrenees, nearby the Formigal sky resort

Images A-D show the overall experiment setup (A), where surface snow was added into mini-lysimeters, that had holes beneath them to allow liquid water passing through (B); the process of microplastic addition (C), from glass flasks; and surface snow parcels (D). In ech experiment, increasing concentrations of microplastics were added to the mini-lysimeters and snow parcels.

Results

Meteorological and initial snow conditions

Six field experiments were conducted under variable meteorological and initial snow conditions during the snow season 2023/2024 (from February to May, 2024). MP polymer type and concentrations used in each experiment are shown in Table 1, included in the STAR Methods section. Figure 2 shows the average daily values of air temperature, snow depth, and accumulated (rain or snowfall) precipitation from the 1^st of January to the 10^th of May 2024. The average temperature (during the observation period, around 4h) varied through the experimental dates from 2.3 °C in the experiment of March 6, up to 12.2 °C in the experiment conducted at the end of that month. Solar radiation was also quite variable, oscillating from average values of 337 W m⁻² in the first experiment (February 6) to 1240 W m⁻² in the last one (May 8). The snow depth was the highest in the experiment of April 2, with a value of ca. 1.80 m, reached after a week of almost continuous snowfall in the region. Fresh snow conditions yielded low-density snow values at the beginning of this experiment (180 kg m⁻³) and that from March 6 (210 kg m⁻³).

Table 1.

Dates and polymer types and concentrations used in each field experiment

Date (EXP ID)	Polymers and concentrations used
February, 6	PUR: C1, C2, C3 PET fibers (C4)
February, 20	PUR: C1, C2, C4
March, 6	PE: C2, C3, C6 (equivalent, in terms of mass, to PUR C1, C2 and C5, respectively) PUR: C3
March, 22	PUR: C2, C3, C6 PE-C6 (∼PUR-C5)
April, 2	PUR: C2 (2 replicates), C5 PE-C6 (∼PUR-C5)
May, 8	PUR C2 (2 replicates), C5 PE-C6 (∼PUR-C5)

Figure 2.

Time series displaying daily average values of temperature, snow depth and accumulated precipitation during the study period

The dashed lines correspond to each of the six experimental dates. Above of each line is indicated the value of initial snow density, measured in situ, except for April 2, which was retrieved from the Izas catchment,²³ due to unavailable data at Formigal on this date, and the high correlation observed between Izas and Formigal snow densities at other dates (not shown).

Effect of microplastics on snow properties

Snow variables were differently affected by the addition of microplastics, depending on the concentration used, but also on snow initial conditions. Broadband albedo displayed overall minor changes in terms of absolute values, oscillating from 0.71 with the highest MP concentration used across the whole period (PUR-C6, experiment of March 22, Table 1), and 0.93 in the blank of the experiment conducted on February 20 (Figure 3A; see also Table S1). Comparing the different treatments, MP yielded higher decreases in albedo in the first experiment (only with PET fibers) and the second experiment with all MP treatments used (and particularly with the highest concentration used, C4, resulting in a decrease from 0.93 to 0.76, or 0.17 units; Table S1), as compared to the blank. While minor changes in albedo values were observed with the addition of MP in the remaining experiments, still, looking at the albedo changing rate (Equation 1) clear differences are observed between the blanks and the higher MP concentrations used in most experiments (Figure 3B).

Figure 3.

Snow broadband albedo and specific surface area (SSA) measurements in the snow parcels

Values of broadband albedo (A), albedo changing rate (B), SSA (C) and SSA changing rate (D) were determined in the snow parcels after ∼4h of exposition to different concentrations of MP.

SSA was particularly sensitive to MP addition. In general, SSA decreased with higher MP additions in most of the experiments, but the pattern was clearer in the experiments of March 6 and April 2 (Figures 3C and S1), the two experiments performed under recent snowfall conditions. In the March 6 experiment, SSA decreased from ca. 18 m² kg⁻¹ in the blank (and 24 m² kg⁻¹ in C1) down to 6.5 m² kg⁻¹ with C5 (Table S2). Similarly, in the April experiment, SSA decreased from 13 m² kg⁻¹ in the blank to 4.1 ± 0.2 m² kg⁻¹ with C5. C2 and C5 replicates, used in the last two experiments, behaved overall similarly for albedo, SSA and SSA changing rate in both experiments, but C5 replicates displayed different for albedo changing rate, particularly in the experiment of April (Figure 3B). SSA changing rates (Figure 3D) were the highest in the last two experiments (April 2 and May 8), reaching values higher than 30%, related to higher initial SSA values and lower exposition times (closer to 3 than 4 h, in opposition to the remaining experiments, that lasted around 4 h). Still, the differences between treatments, particularly between the blanks and the highest MP concentrations used, were clear in all experiments but in the March 22. In this experiment, an unexpected increase in SSA was observed in C5 (and the estimated SSA rate was 0). This value could be explained by the particularly high wind speed average determined that day, which could have blown off some of the MP particles in this replicate.

Snow LWC increased even with the addition of low MP concentrations (C1 and C2) in the first and last experiments, and particularly with high MP concentrations (C4 and C5) in the experiments of February 20 and March 6 (Figures 4A and 4B). Unlike observed with SSA and, to a lesser extent, with albedo, C2 and C5 replicates used in the last experiment displayed quite different values for LWC, LWC changing rate and snow melting. In the case of C2, the first replicate showed a higher LWC value, while the opposite occurred with melting. LWC measurement in these treatments was repeated three times in the field, obtaining similar results. Therefore, differences in LWC values and melting might be related to slightly different dispersion of the particles in the replicate parcels, or partial removal of the particles by the wind in the first replicate—where slightly higher SSA values were determined. The percentage of snowmelt could not be measured in the second replicate (C2 and C5) in the experiment of April, because those mini-lysimeters were lost due to a gust of wind. Similarly, it could not be determined in the second replicate of C5 in May, because one of the mini-lysimeters could not be used. Melting after about 4 h of exposition to the different MP concentrations was overall small, with changes <5% (Figure 4C). Percentage of snowmelt was much higher in the experiments of March 22 and May 8, but values (between 34% and 65%) were similar in all treatments, including the blank. Therefore, it could not be attributed to the addition of MP. An exception was the highest MP addition (C5) in the experiment of May, which induced an increase of 65% melting, as compared to the 48% observed in the blank. In this experiment, linear models displayed a positive correlation between snowmelt and MP concentration (Figure S1D).

Figure 4.

Snow liquid water content (LWC) and snowmelt measurements

Values of LWC (A) and LWC changing rate (B) were determined in the snow parcels after ∼4h of exposition to different concentrations of MP. Snowmelt (C) was calculated by gravimetry in the mini-lysimeters, after subjecting surface snow to the different concentrations of MP during ∼4h.

Correlation among studied variables

The correlation between predictor and studied variables was initially visualized by means of a PCA biplot (Figure 5A). The first two dimensions of the PCA explained 77% of the variance. The plot shows a clear separation of samples according to the experimental date, that can be grouped in three main clusters of samples: The first two experiments (February 6 and 20) are located in the bottom left quadrant (negative values of dim 1 and dim 2), related to slightly higher exposition times to MPs, and lower solar radiation and wind velocity values during those dates. On the up left corner (positive values of dim 2, and negative of dim 1) are set the experiments of March 6 and April 2, those conducted with fresh snow of low density, and the highest values reported for albedo and SSA overall (Figure 3; not necessarily meaning higher changes among treatments in these experiments, although they were observed in SSA, as detailed in the previous section). These two variables were negatively correlated to MP concentration, the predictor variable that explained the least of the observed variability among samples. Melting and LWC followed almost the opposite sense than SSA, located on the right side of the plot (parallel to dimension 1), closer to the samples from the experiments of March 22 and May 8. The latter were also the experiments with initially higher snow density values recorded, as well as the highest temperature, triggering higher LWC and melting values, but not necessarily related to the addition of MP.

Figure 5.

PCA biplot and Spearman correlations between predictor and exploratory variables

The PCA biplot (A) displays the relationship between samples, predictor, and studied variables. Samples are colored according to the experiment date. Bigger dots indicate average values for samples from a given experiment. Spearman correlations (B) are colored as a function of the coefficient of correlation (from -1, dark blue, to 1, dark red) between predictor and studied variables. Asterisks indicate significant correlations (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p <0.001) among paired variables.

Similar results were found through Spearman pairwise comparisons among variables. MP concentration showed a positive but non-significant correlation with LWC and melting, and a negative correlation with SSA and albedo, the correlation with SSA being significant (Figure 5B). Still, SSA showed a higher negative correlation with temperature or solar radiation. Air temperature, and to a lesser extent, solar radiation, had the greatest effect on all the studied variables: negative impact on albedo and SSA, and positive impact on LWC and melting. Exposition time, being a bit longer (close to 4h) in the first three experiments, did not yield further decrease in SSA or increases in LWC and snowmelt, because of its negative correlation with temperature, solar radiation, or wind velocity. As expected, the correlation between albedo and SSA was positive (and significant), and so it was the correlation between LWC and melting, while albedo, and especially SSA, were negatively (and significantly) correlated to LWC and melting (the discussion section provides an interpretation of these results).

The Kruskal-Wallis test showed no differences in the studied variables between the different MP concentrations used. Instead, significant differences were found between experimental dates for the variables albedo (χ² = 14.025, p-value = 0.015), albedo changing rate (χ² = 16.483, p-value = 0.006), SSA (χ² = 21.276, p-value = 0.0007), SSA changing rate (χ² = 23.438, p-value = 0.0003), LWC (χ² = 21.713, p-value = 0.0002), LWC changing rate (χ² = 14.864, p-value = 0.0002), and melting (χ² = 21.069, p-value = 0.0050). Differences were further found between pairs, but are not fully detailed because they are out of the scope of this article, which intends to focus on the effect of microplastics on snow properties.

Discussion

There is already substantial research pointing to snow surface darkening from the deposition of LAPs, typically dark aerosols such as brown and black carbon and mineral dust, which accelerate snowmelt through enhanced absorption of solar irradiance.^13,14,24,25 The capacity of dust and carbon aerosols to decrease snow surface albedo has been demonstrated both by experimental (e.g.,^14,16,17) and modeling studies (e.g.,^15,18,24,26). These studies normally show the snow albedo feedback,^13,27 which can be summarized as a decrease in snow albedo that triggers an increase in LWC, a decrease in SSA (which implies an increase in snow grain size), which then amplifies the albedo decrease and a higher energy absorption. If the snowpack is close to zero °C (usually in older and denser snowpack), the energy absorption can be used to melt ice crystal, whereas if the snowpack has lower temperatures (usually in new and lighter snowpack), the energy absorption must be used to heat the snow crystals and increase snow grain size, with smaller increases observed in LWC and snowmelt.

While it has recently been pointed out that black carbon may have a common origin with black microplastics released from road-tyre wear, and accounting for the most important global source of atmospheric microplastics^28,29; no study so far has empirically evaluated the direct effect of microplastics on snow metamorphism and snowmelt. It is noteworthy to mention that the term microplastics includes a broad range of particles, from sizes of 1 μm up to 5 mm, different shapes, and virtually all possible color combinations, often making it difficult to assess their environmental and biological impacts. For an initial assessment of the potential climatic effect of microplastics on cryospheric regions, we decided to focus on dark microplastics, mostly in the form of pellets, that are easy to prepare and still representative of microplastic pollution in these regions.^3,5,9,11,12

Experiments were performed under diverse meteorological and snow conditions, since factors such as incoming solar radiation or snow age (i.e., time since the last snowfall, which determines initial density and SSA values) have well-known effects on snow reflectivity.^13,14,24 Empirical results from 6 experiments during the snow season period of 2023/2024 in the central Pyrenees provide the first insight into the potential climatic effects of microplastics deposited from the atmosphere—ranging from realistic to high deposition scenarios — in mid-latitude cryospheric regions. The scheme in Figure 6 summarizes the results obtained, which were mostly dependent on initial snow conditions, leading to three different scenarios. In the experiments performed under recent snow conditions (Snow “type I,″ fresh and light snow), when reflectance is the highest,¹³ increasing concentrations of microplastics enhanced absorption that was used to heat the snow crystals (e.g., in the experiment of March 6, snow temperature increased from −5°C to −2°C after particle exposition), leading to moderate changes on snow albedo and a high decrease on specific surface area – meaning an increase in snow grain size.^13,30 In the first two experiments, when (aged) snow initial density presented intermediate values between 250 and 450 kg m⁻³ (Snow “type II”), microplastics yielded a lesser decrease on SSA, probably because the grain size was already coarse, but a moderate decrease in snow albedo of up to 0.17 units with the highest concentration used. This change is greater than that estimated by Ginot et al. 2014 from dust and black carbon records from an ice core retrieved in the Himalayas,²⁴ and close to the maximum albedo reducing values estimated for mineral dust (0.21) and black carbon (0.25) using a deep-learning emulator.²⁰ A clear increase in LWC with increasing particle concentration was also observed in these experiments, in agreement with the snow albedo feedback theory.^13,27 However, the cold and compact conditions of those days hampered water from flowing through the snow grains, and only a small percentage of melting (<5%) took place in the minilysimeters containing the highest concentrations of microplastics.

Figure 6.

Overview of the changes yielded by dark microplastics on snow properties, under different snow conditions

Dark microplastics had different effects on snow surface properties, depending mostly on snow initial conditions. Snow type classification based on its density (kg m-3) was done accordingly to our results but agrees with previous classifications found in the literature (e.g. Muskett, 2012³¹).

Finally, we observed a third scenario. In the experiments performed on the warmer days (March 22 and May 8), toward the end of the snow season, an increase in LWC was observed in some of the samples (not necessarily the ones amended with the highest concentrations of microplastics), and effective melting was observed in all the samples in less than 4 h. The melting was even observed in the blank treatments and therefore was mostly attributed to wet snow conditions (density >450 kg m⁻³; snow “type III”) and meteorological conditions (temperature was particularly high on the 22^nd of March experiment, reaching 12.2 °C; and solar radiation was the highest on the 8^th of May, averaging 1240 W m⁻² during MP exposition time). Nonetheless, there was a clear positive relationship between snowmelt and increasing MP concentrations in the experiment of May 8, yielding a 65% melting with the highest concentration, 17% more than observed in the blanks. Révelliet et al. (2022)²⁶ reported that meteorological conditions were the main factor (accounting for ca. 80% of the total variance) affecting snow cover evolution in the Pyrenees. Still, those authors showed, by means of numerical simulations, that dust and black carbon deposition in the Pyrenees advanced snowmelt by 17 ± 6 days on average over the 1979–2018 period, and by 10–15 days the peak melt water runoff.

Our findings on the effect of microplastics on snow metamorphism are preliminary and do not yet support regional-scale estimates of their impact on radiative forcing or snowmelt timing. Because of the common origin of MP and black carbon, MP's efficiency to absorb solar radiation is expected to be similarly high. However, to our knowledge, there is currently no information providing quantitative data on MP optical properties or estimated radiative forcing in snow surfaces. First assessment of MP effective radiative forcing in the atmosphere boundary layer estimate values about −0.746 ± 0.553 mW m⁻²,³² slightly smaller to average values reported by IPCC AR5 for mineral dust.³³ Instead, black carbon radiative forcing is positive, estimated to be between 0 and 0.8 W m⁻² (average ∼0.4 W m⁻²).³³ However, radiative forcing simulations of MPs were performed on non-pigmented microplastics, and the authors acknowledged that pigmented microplastics (which normally account for more than 50% of all the analyzed particles, in every type of sample) may absorb more than they scatter in the visible spectrum, yielding a net positive radiative forcing and contributing to atmospheric warming.^21,32

Microplastics, in opposition to other LAPs such as mineral dust, constitute lower but constant inputs to mountain ranges, possibly accumulating toward the end of the snow period. Estimations of concentrations of MPs in surface, aged, snow from other mid-latitude regions, range from 0.04 to 0.4 g m⁻² in Western USA¹⁰ to 0.17 to 7.11 g m⁻² in northeastern China.¹¹ The concentrations used in our experiments are consistent with those values; the highest concentrations used here (5 g m⁻²) being lower than the highest concentration reported in northeastern China. These concentrations are expected to increase globally in the coming years – on one hand, because plastic concentration in the environment is expected to continue increasing, reflecting the exponential global plastic demand of the last decades³⁴; and on the other hand, because plastics break down over time, turning into secondary MP. Considering also the overall low-density of microplastics and their larger size, as compared to mineral dust, plus their low solubility in water, it could be expected that they do not percolate to lower layers, leading to a surface accumulation toward the end of the snow season, when the snowpack is thinner, isothermal, and of higher density. Indeed, Wen et al. (2024)¹¹ found significantly higher concentrations of MPs in aged than in fresh snow in northeastern China, and the same has been observed for black carbon.^16,18 Along with increasing temperature and solar radiation toward the end of the snow season in mid-latitudes, it is expected that microplastics lead to an advanced snowmelt, and this pattern will probably worsen in the coming years, if no further measures are implemented globally to reduce (micro-)plastic pollution.

Limitations of the study

In this study, microplastics were added to snow parcels simulating realistic atmospheric deposition fluxes (including here the deposition of particles with a size between 25 μm and 5 mm) between 100 (C1) and 1000 items m⁻² d⁻¹ (C2), as reported in mid-latitude regions across Europe.^35,36,37,38 We then used concentrations simulating higher fluxes up to 10,000 items m⁻² d⁻¹ (C3) and 100,000 items m⁻² d⁻¹ (C4), which are already realistic in other mid-latitude cryospheric regions,¹¹ and may become realistic in the study area in the near future (due to increasing MP fluxes, and because technical improvements will allow to account for the smaller size fractions). In any case, C3 and C4 were only used twice (C3) and once (C4), in the first experiments, and were substituted afterward for C5, that simulated a deposition event of 1000 items m⁻² d⁻¹ (C2), and the accumulation of those settled particles for a period of one month, representing realistic conditions, especially toward the end of the snow season, when snowfall may be almost absent, as observed in the area of study.

Probably the major limitation of the present work was the fact of assuming that all microplastics deposited in mid-latitude cryospheric areas from atmospheric fallout are black pellets or fragments. Dark-colored microplastics indeed account for most particles (50–80%) reported in cryospheric regions.^3,5,12 On the other hand, while a few studies have reported that fragments accounted for the majority of microparticles retrieved from snow samples,^6,10,11,39 most studies reporting MPs from atmospheric fallout or down in the snow have shown that microfibers account for the biggest fraction (usually >33%).^2,5,11,12,40 We attempted to do these experiments using both dark fragments and fibers, but at the end, PET fibers were only used in the first experiment because of the difficulty of retrieving microfibers from garments in a sufficiently small size (≤1 mm) and enough amount in terms of mass, for microfibers not to cluster together and disperse homogeneously through the snow parcels.

In this study, we have set the path to understand the potential effects of microplastics as LAPs, using different MP concentrations and under different (and real) snow conditions. Still, further research will need to consider different MP types (particles of different size, color, and shape), and a combination of them for a more accurate evaluation on the impact of microplastics in cryospheric regions worldwide. In addition, the experiments were conducted over a short period of time (4 h), while longer-term (e.g., 24 h to one week) experiments may better reflect the effect of microplastic accumulation under different conditions (probably leading to complete melting in less than 24 h toward the end of the snow season in mid-latitudes). Furthermore, more comprehensive studies aiming to better understand the effect of light-absorbing particles on snow properties and advanced snowmelt should focus on disentangling the separated and combined effects of the different snow impurities (black carbon, microplastics, mineral dust, and microorganisms) at different locations and under variable initial snow conditions.

Conclusions

The present work provides the first evidence on the capacity of dark microplastics to trigger changes in snow metamorphism, eventually triggering snow melting, even at simulated realistic atmospheric deposition fluxes. Results varied according to the microplastic concentrations used, but were mostly dependent on snow initial conditions, observing three different patterns as a function of initial snow density. For initial density values <250 kg m⁻³, microplastics had a predominant heating effect on snow grains, leading to decreases in SSA as large as 11.4 m² kg⁻¹ higher than in the control parcels. When the initial density ranged between 250 kg m⁻³ and 450 kg m⁻³, microplastic triggered the classic snow albedo feedback, meaning an increase in snow grain size and in liquid water content, up to 0.53 volumetric %, and a decrease in albedo of 0.17 units, in comparison to the controls. Moderate snowmelt was observed through the season, with higher values toward the end of the season, when the snow was wet, and the snowpack received higher energy inputs. In those experiments, initial snow density was >450 kg m⁻³, and an increase in melting of 17% above the controls was observed with the highest microplastic concentration used. We call for further field investigations on the effect of different types of microplastics on snow properties and snowmelt, and their combined effect with other snow impurities, particularly toward the end of the snow season in mid-latitudes, when impurities accumulate on a thinner snowpack.

Resource availability

Lead contact

Requests for further information and resources should be directed to the lead contact, Isabel Marín Beltrán (imbeltran@ualg.pt; isabel.marin.beltran@gmail.com).

Materials availability

This study did not generate new unique material.

Data and code availability

• •Data: The data supporting the findings of this study are available within the article and the supplemental information.
• •The code used to conduct statistical analysis and prepare the figures shown in this article is publicly available at Zenodo (https://zenodo.org/records/17976507).
• •Any additional information required is available upon request to the lead contact.

Acknowledgments

This study received Portuguese national funds from Foundation for Science and Technology (FCT) through projects UIDB/04326/2020, UIDP/04326/2020, LA/P/0101/2020, and 2023.00192.RESTART. IMB was awarded with an individual fellowship from the FCT Scientific Employment Stimulus call (CEECIND/03072/2017), and JB has an FPI predoctoral grant from the Spanish Ministry of Science and Innovation (PRE2022-103791). Further funding was received from the projects SNOWDUST (AEI, “Transición Ecológica y Digital” 2021; TED2021-130114B-I00), MARGISNOW (PID2021-124220OB-I00; Spanish Ministry of Science and Innovation), and VINILO (PROY_E31_24; Gobierno de Aragón). The authors are grateful to Eñaut Izaguirre and César Deschamps-Berger for their assistance with fieldwork campaigns. The Spanish meteorological agency network (AEMET) provided data on accumulated precipitation, and ARAMON-Formigal allowed free access to the experimental area.

Author contributions

I.M.B.: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, visualization, and writing - original draft; J.B.: conceptualization, methodology, investigation, data curation, and writing - review and editing; P.D.A.: investigation, data curation, and writing - review and editing; J.P.: conceptualization, funding acquisition, investigation, methodology, project administration, resources, and writing - review and editing; J.R.: resources, investigation, and writing - review and editing; J.I.L.M.: conceptualization, funding acquisition, investigation, methodology, project administration, resources, and writing - review and editing.

Declaration of interests

The authors declare no competing interests.

Declaration of generative AI and AI-assisted technologies in the writing process

The authors declare no particular use of AI either for the study planification, data treatment, preparation of results or writing purposes.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Chemicals
Low-density polyethylene (PE) black pellets, ∼300 μmm length	eBay GmbH	No new or registered material
Polyurethane (PUR) black fragments	This study: Fragments manually detached from a second-hand baby stroller	No new or registered material
Polyester (PET) blue fibers	This study: Fibers manually retrieved from an old backpack	No new or registered material
Ethanol (>99.8%)	Fisher Scientific	https://www.fishersci.co.uk/shop/products/2-5lt-ethanol-absolute-99-8-analysis-1/12478740
Software and algorithms
R studio (Version 1.0.143)	Open-source software	https://www.R-project.org/
IceCube SSA converter	Software provided by the manufacturer upon acquisition of the IceCube instrument (A2 Photonics Sensor; France)	https://a2photonicsensors.com/icecube-specific-surface-area-snow-measurement/
OMNIC Picta libraries	Software provided by the manufacturer upon acquisition of a Nicolet iN10MX μ-FTIR (Thermo Fisher Scientific, USA)	https://knowledge1.thermofisher.com/Molecular_Spectroscopy/Fourier_Transform_FT_Infrared_IR_and_Near_Infrared_NIR/FT-IR_and_FT-NIR_Knowledge/Omnic_Library_spectral_indexes
Deposited data
Raw and analyzed data	This paper; Zenodo repository	https://zenodo.org/records/17976507
Other
Canva	Online, graphic design platform	https://www.canva.com

Method details

Area of study and experimental setup

During the snow season of 2023/2024, six in situ experiments were conducted (from early February to early May 2024) at the Spanish Central Pyrenees, nearby the Formigal sky resort (Figure 1). The field experiments consisted of a set of mini-lysimeters (30 cm × 22 cm) and parcels (30 cm × 30 cm) containing surface snow aerosolised with different types and concentrations of MP (Table 1), following a similar procedure as in Bandrés et al. (2025).²⁵ The selected polymers used were low-density polyethylene (PE) and polyurethane (PUR) black pellets (ratio between length and width <1.5). PE pellets, acquired online (through eBay), had an initial diameter of ∼1 mm. They were crushed in the laboratory with a coffee mill, until reaching a median size of 300 μm (major axis; minor axis = 250 μm), as determined by the measurement of 20 particles under a Nikon SMZ1500 (Nikon Healthcare, Japan) magnifier.

To simulate a common, continuous output of MPs to the area of study, fragments were detached from extensively used tyres of a second-hand baby stroller, to simulate the release of MPs from road traffic-derived non-exhaust emissions. It has been estimated that about 1.3 million tons of tyre wear microparticles are generated on roads in Europe per year,⁴¹ accounting for between 22 and 43% of global tyre wear particles.^28,29 Those particles, especially the smallest fractions, have high transport efficiencies to remote regions.²² The particles used here had a median diameter of 450 μm (major axis, minor axis = 350 μm; as determined under the magnifier), aiming to simulate instead the deposition of tyre particles released from local roads, frequently visited during the snow period.⁴² Additionally, in the first experiment (February 6; Table 1), blue polyester fibers (major axis ≤1 mm; length:width >3; NOAA, 2022)⁴³ retrieved from an old backpack were also used. Because of the difficulty of preparing and adding fibers homogeneously through the snow parcels, they were not used in further experiments.

The chemical composition of commercial pellets, stroller tyre particles and fibers were determined by means of Fourier Transform Infrared spectroscopy coupled with microsocopy (μ-FTIR) using a Nicolet iN10MX μ-FTIR (Thermo Fisher Scientific, USA). Measurements were done on attenuated total reflectance mode, using a germanium tip. Spectra were collected in the middle infrared region (from 4000 to 675 cm⁻¹ wavenumbers), recording 16 scans at 4 cm⁻¹ spectral resolution. These analyses confirmed that the composition of commercial pellets was low-density PE, fibers were made of PET, and tyre main composition was PUR.

PE and PUR pellets were added to surface snow plots in different concentrations, starting with realistic atmospheric deposition fluxes of 100 MP items m⁻² d⁻¹ (C1) and 1000 items m⁻² d⁻¹ (C2). These values were then transformed into mass concentrations to be added to the lysimeters and snow parcels, considering the dimensions and density of each polymer. For comparison, MP atmospheric deposition fluxes ranging from 300 to 460 items m⁻² d⁻¹ have been reported in the French Pyrenees,³⁵ and thus concentrations C1 and C2 are of same order of magnitude to those already observed. Concentrations 10 (C3) and 100 times (C4) higher to C2 were also used in the first two experiments, when a minor effect of the impurities was expected due to snow conditions (compact and of high density). MP inputs in the region are relatively low as compared to, for example, Saharan dust events,¹⁴ but are continuous in time. That’s why, from the third experiment on, concentrations simulating the accumulation of fluxes equivalent to C2 during one-month (C5) or the whole winter period (120 days; C6), were used.

Since, in terms of mass, the concentration of PE-C2, PE-C3 and PE-C6 were on the same order of magnitude of PUR-C1, PUR-C2 and PUR-C5, respectively, and their effects on snow properties were similar, for visualization and statistical purposes, these samples were considered as simply “C1” (equivalent to surface concentration of 0.007 g m⁻²), “C2” (0.07 g m⁻²) or “C5” (1.24 g m⁻²). The intricate logistic of the experiments did not allow for having regular duplicate concentrations, but we set replicates of C2 and C5 in the last two experiments (considering PE-C6 equivalent to PUR-C5; Table 1).

In each experiment, a mini-lysimeter and equivalent snow parcel remained exempt of particle addition and were used as control blanks (“K”) to observe changes in the snowpack without MP presence. The mini-lysimeter and parcels containing surface snow were aerosolised with the different concentrations of MPs (depending on the experiment) and left for about 4 h to quantify changes in snow specific surface area (SSA, defined as the total area at the ice-air interface in a given snow sample per unit mass⁴⁴), snow liquid water content (LWC; defined as the amount of liquid water per unit volume³⁹), albedo,⁴⁵ and, ultimately, the total melted water after particle exposition. The latter was only measured in the mini-lysimeters, that had holes below (Figure 1B), allowing for the liquid water to pass to another white tray located below. Snow and water mass were calculated by gravimetry, weighing the pre-weighed trays at the start and endpoints of the experiment.

Analytical measurements

Just after the addition of MP particles on the snow parcels and mini-lysimeters, snow properties were determined on “clean snow” parcels (no addition of MP). A WISE (A2 Photonics Sensor; France) device was used to determine snow density and LWC (based on snow mass and permittivity), within a measurement range between 0 and 20 in terms of % of volume, and a typical uncertainty of ±1%. In the experiment of April 2, this instrument was not available. An IceCube (A2 Photonics Sensor; France) instrument was used to determine the voltage among snow grains (measurement uncertainty of ±10%) and obtain the SSA by means of the IceCube SSA converter software, provided by the manufacturer. The instrument shows accuracy for SSA values between 2 and 160 m² kg⁻¹ (melt-freeze crusts to fresh snow). Broadband snow surface albedo was retrieved from a ROX spectrometer (JB Hyperspectral Devices, Germany), with a spectral range covering from 340 to 1026 nm (nm). Only values from 400 to 900 nm were further considered in the calculations, though, since this is the optimal range of the instrument. Similarly, these properties were determined after about 4 h of exposition to MP in the experimental parcels, and in the mini-lysimeters in the case of albedo.

Meteorological data

At each experimental date, a portable automatic weather station (HOBO Onset U30) was set next to the experimental area to record the meteorological variables presumed to influence the snow energy and mass balance: air temperature and humidity, solar radiation and wind speed. Measurements were registered every 10 min, and the average value for the period while the experiments were conducted (about 4 h) was considered in further analysis, as detailed below. In addition, we retrieved data from snow depth measured every 10 min at the Izas Experimental Catchment,²³ located at an elevation of 2050 m a.s.l., and approximately 5 km apart from our experimental area. Accumulated precipitation from the equivalent period was obtained from the closest meteorological station from the Spanish meteorological agency network, close to the village of Formigal.

Quantification and statistical analysis

Broadband albedo (400–900 nm wavelengths) and SSA changing rates were calculated considering the initial value measured on clean snow, and the exposition time to particles in each experimental snow parcel, as indicated in Equation 1.

$$\left(Albedo,SSA\right)changingrate=\frac{\left(Albedo,SSA\right)t0-\left(Albedo,SSA\right)MP}{\left(Albedo,SSA\right)t0}x\frac{100}{expositiontime\left(h\right)}\text{,}$$ (Equation 1)

where “t0” is the measurement of the given variable in a clean snow parcel, determined at the start of each experiment, and “MP” indicates the value of the variable after the exposition time, in each experimental parcel amended with different MP concentrations. To calculate LWC rates, the equation used was the same as for albedo and SSA, but in this case, the initial value measured on the clean snow parcel was extracted to that on MP parcels, since LWC values normally increased after the addition of the particles.

A Principal Components analysis was conducted to visualize the effect of explanatory or predictor (MP concentration, exposition time, initial snow density, mean air temperature, mean solar radiation and mean wind velocity) variables on the studied variables albedo, SSA, LWC and snow melting. Before conducting the PCA, missing data were handled using the “missMDA” package⁴⁶ in R,⁴⁷ and the complete dataset was log-transformed afterward. The linear relationship between the studied variables and the interaction MP concentration∗Experimental date was assessed by means of general linear models. Still, since variable residuals did not usually come from normal distributions, even though after conducting variable transformation, non-parametric tests were performed afterwards. Spearman pairwise comparison was conducted among the above-mentioned predictor (except initial snow density, that was excluded from the analysis due to its clear correlation with air temperature) and studied variables, using the Bonferroni correction.⁴⁸ The Kruskal-Wallis⁴⁹ analysis was used to test if any of the studied variables (including both raw data and changing rates) responded differently to the MP concentrations used, and to test for statistical differences among experimental dates. If the test returned a p-value <0.05, the Wilcoxon pairwise pot-hoc test was used to test for differences between each pair of treatments. All statistical analysis and figures (except Figure 1) were computed in R studio (Version 1.0.143).⁴⁷ Figures created in R were later assembled in canva.com.

Published: January 29, 2026