Fresh Snowfall Microbiology and Chemistry are Driven by Geography in Storm-Tracked Events

Snowfall is a global phenomenon highly integrated with hydrology and ecology. The study of bioaerosols and their dependence on aeolian movement are largely constrained to either precipitation-independent analyses or in-silico models. Though snowpack and glacial microbiological studies have been conducted, little is known about the biological component of meteoric snow. Through culture-independent phylogenetic and geochemical analyses, we show that the geographical location at which snow precipitates determines snowfall9s geochemical and microbiological composition. Storm-tracking can be used as a valuable environmental indicator to trace down what factors are influencing bioaerosols. We estimate annual deposits of up to ~10 kg of bacterial/archaeal biomass per hectare along our study area of the eastern Front Range in Colorado. The dominant kinds of microbiota captured in seven snow events at two different locations, one urban, one rural, across the winter of 2016/2017 included phyla Proteobacteria, Bacteroidetes, Firmicutes and Acidobacteria, though a multitude of different kinds of organisms were found in both. Taxonomically, Bacteroidetes were more abundant in Golden (urban plain) snow while Proteobacteria were more common in Sunshine (rural mountain) samples. Chemically, Golden snowfall was positively correlated with some metals and anions. The work informs the 9everything is everywhere9 hypotheses of the microbial world and that atmospheric transport of microbiota is not only common, but is capable of disseminating vast amounts of microbiota of different physiologies and genetics that then affect ecosystems globally. Snowfall is a significant repository of microbiological material with strong implications for both ecosystem genetic flux and general bio-aerosol theory.


Introduction 62
Throughout temperate and polar regions of the world, snowfall is ubiquitous. 63 Specifically, along the Front Range of eastern Colorado, the mean annual snowfall from 2010-64 2016 is reported as 97.7 inches in Boulder, Colorado (1). Despite snow's prevalence, little is 65 known about the biological composition of this massive annual hydrological deposit; meteoric 66 snow is an essential part of yearly moisture input in Boulder, Colorado given that it represents 67 43% of the total precipitation (by liquid water volume) from 2010-2016 (1). Both snow and rain 68 require an initiation surface for atmospheric water to condense into a droplet of water or ice 69 they contact soil and surface receiving waters, however, remains unknown. Beyond the 93 community structure of precipitation microbiota, it is especially unknown how these structures 94 changes as a result of geography and atmospheric events. 95 Bioaerosol literature, in general, is best represented by reports of in-silico bioaerosol 96 movement, aerosol microbial composition and studies of canonical surface constituents such as 97 glaciers and snowpack (24-28). There exists little specific information, however, on the 98 microbiology of snowfall precipitation. The biological composition of snowfall-a massively 99 spatial meteorological event-should be well understood vis-à-vis magnitude and community 100 structure, especially given that precipitation is key to linking atmospheric and surface theory. An 101 aid to such study is the preponderance of weather information available that tracks storm events 102 in multiple-spectral analyses; these data further inform storm trajectories as well as hydrologic 103 source-water locations. 104 To this end, we sampled fresh snowfall throughout the entirety of the 2016-2017 snow 105 season along the Front Range of eastern Colorado. Our investigations include disparate sampling 106 sites (one rural mountain, and one urban plain) to help in the characterization of snowfall with 107 respect to both sampling location-within the same storm-and direction of origin of the storm. 108 We hypothesized that the biological content of fresh snowfall would be uniquely distinguishable 109 with respect to sampling location and that storm tracking could provide novel insight into other 110 factors that may affect biogeography. Here, we apply ecological investigative approaches 111 together with remote sensing based atmospheric storm analyses to assess the magnitude of 112 biological deposition, community structure, and abiotic solute associations in fresh snowfall.

Geochemistry: 155
Filtrate from melted snow was subsampled to two 15 mL aliquots for Ion 156

Chromatography (IC; major anions) and Inductively Coupled Plasma Optical Emission 157
Spectrometry (ICP-OES; major cations) analysis. ICP-OES samples were acidified with three 158 drops of trace metal grade 10% Nitric Acid. IC analysis was performed on a Dionex ICS-90 ion 159 chromatography system running an AS14A (4 × 250 mm) column, while ICP-OES was 160 . CC-BY-NC-ND 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under from each sample were pooled into one object and analyzed together; these pooled samples will 204 be referred to simply as 'samples' from here on. Samples were trimmed to contain only bacterial 205 . CC-BY-NC-ND 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted April 13, 2018. ; https://doi.org/10.1101/300772 doi: bioRxiv preprint and archaeal sequences, with chloroplast and mitochondrial sequences similarly excised. All 206 samples were rarefied to a depth of 3,391 sequences; 3,391 was chosen because it allowed 207 inclusion of all but one sample (12/07/16 Sunshine) while also retaining most of the sample 208 diversity ( Figure S1). Downstream sequence analyses were run with multiple random 209 subsampling sets and results were not found to change appreciably at our selected rarefaction 210 level of 3,391 bacterial / archaeal sequences. For reproducibility, a set random start seed was 211 used in the rarefaction process. Taxonomy assignments were called with a SILVA training file 212 through the DADA2 taxonomy assignment script. Phylogenetic trees were constructed with 213 sequence alignment via SINA (Silva Incremental Aligner; v1.2.11; 90% complementarity) (36) 214 in the QIIME package (37). OTU tables, taxonomy tables, phylogenetic trees, and sample meta-215 data were compiled in the data processing phyloseq object presented by the R package 216 'phyloseq' (38). 217 Assessment of eukaryotic sequences was done via the same method as above, with the 218 exception of merging forward and reverse Illumina reads. In a separate pipeline, all forward and 219 reverse reads were merged by concatenating the reads directly with a 10-Ns as the joiner. As in 220 the bacterial / archaeal analysis, sequences from technical replicates were pooled to one sample. 221 Bacterial and archaeal taxonomic assignments were removed from this dataset, leaving only 18S 222 rRNA eukaryotic sequences. The eukaryotic sequence dataset was rarefied to 529 sequences; few 223 eukaryotic sequences were returned in general, and rarefaction at 529 sequences therefore did not 224 dramatically reduce the number of sequences analyzed. Like the bacterial and archaeal analysis, 225 a set random start seed was used during rarefaction for reproducibility. Three samples were 226 removed from the eukaryotic analysis because they were below the sequence threshold (two 227 from Golden and one from Sunshine). 228 . CC-BY-NC-ND 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted April 13, 2018. ; https://doi.org/10.1101/300772 doi: bioRxiv preprint

Microscopy: 231
Imaged filter quadrants from all samples revealed patterns that distinguish samples taken 232 from Golden from those taken in Sunshine (Figure 1). There is substantially more material 233 captured by filtering snow from Golden than snow from Sunshine in the same storm. Any storm 234 localized to Golden generally yielded more material than storms localized to Sunshine.    Principal Component 1 suggests a high chemical variance between samples collected in Golden 251 . CC-BY-NC-ND 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted April 13, 2018. ; https://doi.org/10.1101/300772 doi: bioRxiv preprint relative to those collected in Sunshine. To assess the statistical differentiation of chemical data 252 with respect to sampling location (independent of the PCA), a Euclidean distance matrix was 253 constructed using the same samples and chemical data used in the chemical PCA analysis. An 254 Adonis test was run on the Euclidean distance matrix to inspect statistical differences between 255 Sunshine and Golden snowmelt chemistry; the two groups were found to be significantly 256 different (R 2 = 0.28, p = 0.002). NOAA HYSPLIT models were used to assess which direction the majority of storm 283 travel vectors came from for each sampling event. Most storms were of a SW origin while two 284 storms came from the NW and one storm was of a SE origin (a Colorado 'upslope' storm). We 285 verified from the HYSPLIT models that same-day samples from both geographic locations were 286 in fact from the same snow storm and could be assigned the same direction of origin (example: 287

Characterization of Snow Communities: 289
A community ordination chart organized by both sampling location and origin of storm is 290 presented. Principal Coordinate Analysis (PCoA) suggests strong clustering by sampling 291 location, as constructed by a weighted Unifrac distance matrix (Figure 4)

(39). An Adonis test 292
was run on the dissimilarity matrix used for Figure 4, and sampling locations were found to be 293 distinct in grouping (R 2 = 0.20, p = 0.033). Weighted Unifrac distance matrices were also tested 294 by Adonis for true-differences in samples taken from different storm trajectories; no significance 295 was detected. Snow filtering-process controls, DNA extraction controls, and no-template PCR 296 controls (NTCs), amplified during qPCR a minimum of two cycles after at least one filter 297 . CC-BY-NC-ND 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted April 13, 2018. ; https://doi.org/10.1101/300772 doi: bioRxiv preprint replicate from all samples, indicating that differences seen between samples are true biological 298 differences. Filtering-process controls, furthermore, were ordinated against true samples and 299 found to be distinct in clustering (data not shown). 300 Snowstorms from both sampling locations display a wide range of bacterial / archaeal 301 ( Figure S4A) and eukaryotic ( Figure S4B By use of SIMPER (vegan R package) (40), the taxa from each taxonomic rank most responsible 306 for the Bray-Curtis distance between samples were determined. Plotted in descending order of 307 contribution, we demonstrate how any taxonomic rank is sufficient to differentiate the two 308 sampling locations ( Figure 5). The top-ten most influential genera in differentiating sample 309 locations were pulled from the SIMPER analysis, and are shown in Table 1 along with their % 310 contribution to the total difference. The abundances of these genera are examined from both 311 sampling locations as well as the control from the filtering process to assess contamination; no 312 contamination is suspected ( Figure S6). For each of the top ten most influential genera (Table 1), 313 the OTU that had the highest abundance in each genus bin was searched with MegaBLAST (41) 314 for highly similar sequences within the 16S rRNA sequence database from NCBI. For each 315 sequence searched by MegaBLAST, the most highly similar sequence with information on 316 source was summarized: genera found more frequently in Golden were associated with soil, 317 horse dung, airborne dust, hard water, and clinical isolates (42-46); genera common to Sunshine 318 are reported in swimming pool water, clinical material, cultivated land, and may be spore 319 forming (47-51). Though Adonis showed no significant difference between samples with 320 . CC-BY-NC-ND 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted April 13, 2018. ; https://doi.org/10.1101/300772 doi: bioRxiv preprint different storm-origins via the weighted Unifrac distance, there are obvious taxonomic 321 differences between storm trajectories ( Figure S5). 322 Few eukaryotic sequences were recovered relative to the abundance of bacterial and 323 archaeal sequences found in snowfall; prior to rarefaction, 6.2% of the total number of sequences 324 recovered were of eukaryotic taxonomy. Of those eukaryotic sequences recovered and classified, 325 the two fungal phyla Ascomycota and Basidiomycota were clearly most common. Similar to the 326 results presented from bacterial and archaeal sequences, the two dominant eukaryotic phyla 327 display differential relative abundance with respect to both sampling location and the origin of 328 snowstorms ( Figures S4B, S5B). does not account for additional biomass that may have lysed during the thawing of our snow 358 samples prior to filtering and/or the organisms, e.g., nano-archaeota and bacteria (as well as 359 viruses) that pass through a 0.45-µm filter. Many eukaryotic sequences were also obtained from 360 our samples. Though we do not have qPCR data on eukarya in fresh snow, this additional 361 biomass (6.2% of all sequences)-known to be legitimate eukaryotic amplification by our 362 primers (32)-is likely to contribute substantially to the total biomass contained in snowfall. 363 Thus, 10 kg of biomass deposited hectare -1 year -1 is likely an underestimate. In addition, this 364 estimate ignores the mass of all abiotic particulate matter (Figure 1) trafficked as part of the same 365 . CC-BY-NC-ND 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted April 13, 2018. ; https://doi.org/10.1101/300772 doi: bioRxiv preprint snow event-a mass that is much more substantial than microbial elements alone and, similarly, 366 is deposited perennially over large swaths of land. 367 Herein we have chosen to use the total estimated biomass of annual snowfall over more 368 traditional measures like copy numbers of ssuRNA 16S because we believe that it is a more 369 tractable unit when assessing large-scale mass transport and deposition. It is our assessment that 370 the importance of our work is best demonstrated by the surprising bio-load of meteoric snow 371 (and its associated genetic bank)-and that whether total estimated cell count or total biomass is 372 the unit of discussion (both of which we report) is tangential to the overall finding that 373 microbiota and their genomes are a substantial component of fresh snowfall. 374 Since we show significant differences between snowfall communities with respect to 375 geography (Figure 4), we posit that snow depositions bear significant ecological consequences-376 especially given that we demonstrate the mass of any particular snow event to be non-trivial 377 (Figure 3). At all taxonomic ranks, we show that just a fraction of taxa are responsible for 378 differentiating geographic locations ( Figure 5). To the point of 'fingerprinting' snowstorms to 379 evaluate how different depositions may be affecting an area ecologically, it is important to 380 identify specific taxonomic groups that are responsible for separations on distance matrices and 381 ordinations (Table 1). We can infer from Table 1 that, even at the genus taxonomic rank, specific 382 taxa contribute heavily to geographic differences in fresh snowfall. Furthermore, these genera are 383 not contamination ( Figure S6), indicating that real biological differences can be identified 384 geographically in non-general terms. Of the influential genera, sequences that are more abundant 385 in Golden (searched by MegaBLAST-Results section) are highly similar to some sequences 386 that are associated with the human built environment / urban areas (i.e. horse dung and clinical 387 isolates). Influential sequences from Sunshine, similarly, are highly similar to what is found in 388 . CC-BY-NC-ND 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted April 13, 2018. ; https://doi.org/10.1101/300772 doi: bioRxiv preprint rural settings (genera isolated from cultivated land); one of the sequences more common in 389 Sunshine is also highly similar to known spore-formers, suggesting that resilience to colder 390 temperatures experienced in-transit to a higher elevation than Golden may have been beneficial 391 to survival. Though we caution too much extrapolation from the presence of just a few genera, 392 we think it is fair to suggest that demonstrated differences between Sunshine and Golden are the 393 result of environmental / physical differences between the two geographic sites. Our control data, 394 furthermore, strongly suggest that these differences are real, and not a product of contaminating 395 genera found in either sample group ( Figure S6). Not only do our data show that different 396 geographical regions receive distinct biological input, but they also indicate that storm 397 trajectories will be an essential metric to assess in future studies. Although we do not show 398 demonstrable significance in the microbiological differences between storm trajectories, storm-399 track is still likely a contributing component of the variance between samples ( Figure S5). Even 400 with strong evidence that geography drives differences in microbial communities, we note that 401 the two axes from our principal coordinate analysis represent less than 60% of the total variance 402 seen between samples. Our investigation of storm-track influence on community composition is 403 likely hindered by sample size; i.e. amongst the 10 samples profiled, only three of them were not 404 of a SW origin. Overall, our hypothesis that the biotic components of fresh snowfall are 405 distinguishable by geography is confirmed. To the point of meteorological information as 406 predictive data in future studies, storm tracks should not be ignored amongst the list of affecters 407 of microbial community composition. With ongoing collection of snow from more storms and 408 winters, storm tracks could be forensically investigated by an analysis of their surface-deposited 409 microbiota in snows, as well as the ices that result from snows in / on permanent snow fields and 410 glaciers. 411 . CC-BY-NC-ND 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted April 13, 2018. ; https://doi.org/10.1101/300772 doi: bioRxiv preprint Bowers et al. showed in 2011 (7) that air mass microbiological features were likely derived 412 from the land beneath. We extend this finding by showing that geography also appears to have 413 community effects on snowfall deposition. We, however, further posit that biological ice-414 nucleation in the atmosphere (53, 54) may be a selection event influential enough to differentiate 415 microbial communities even more than air masses do at any given geography. Thus, snowfall 416 provides definitive access to surface receiving systems whereas air masses may not be as capable 417 of moving large amounts of material between the atmosphere / lithosphere / hydrosphere. be considered an important part of the total air mass that affects cloud formation potential and 426 the precipitation nucleation process. Though it is known that high-temperature INs are 427 ubiquitous in precipitation (58), our work demonstrates that the vast majority of microbiota in 428 fresh snowfall are not of the high-temperature nucleating genera. More attention should be given 429 to total bacterial communities in precipitation due to the magnitude of their abundance, potential 430 for terrestrial impact, likely effects on cloud glaciation, and non-uniform patterns of deposition 431 in snowstorm events (Figure 4). We know that bacteria/archaea/eucarya can be viable in aerosols 432 (59), suggesting that the multitude of taxa that we find in snow may be contributing to 433 atmospheric chemistry in-situ around Earth. Given all of the bio-physical / chemical influences 434 . CC-BY-NC-ND 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted April 13, 2018. ; https://doi.org/10.1101/300772 doi: bioRxiv preprint on the atmosphere, bio-precipitation should be considered just as complex and variable-and 435 storm-track should be considered and essential component of this variance. Furthermore, the 436 importance of urban / rural areas receiving entirely different inputs in the same snowstorm 437 cannot be understated, especially given the large mass of these annual deposits. This could have 438 relevance to overall air quality and human health in such storm events. 439 Bio-aerosols are known components of human-associated dust particle traffic (12), and dust Broad questions on the impact of snowfall-deposited microbiology on both public health and 455 ecosystem function can be generated by the work we describe herein. With respect to public 456 health, in light of our work we believe that snowfall (and meteoric deposition of any kind -e.g., 457 . CC-BY-NC-ND 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted April 13, 2018. ; https://doi.org/10.1101/300772 doi: bioRxiv preprint rain, snow, ice, fog and perhaps even general humidity) can be considered a large transporter of 458 microbial biomass with a (to-date) not understood pathogenicity. In considering ecosystem 459 function, the impact of some ~10 kg of bacterial / archaeal biomass being deposited per hectare 460 in a year of snowfall clearly must include effects on receiving soil, rock endolith and/or near-461 surface water ecosystems. This has impacts on ecological soil formation, maintenance and health 462 of agricultural soils and water quality in both fresh and marine ecosystems. As climate change 463 continues to be a driver of large-scale environmental shifts, the intricacies of biological aerosol 464 transport and bio-cloud glaciation / snow nucleation are unlikely to escape disturbance. 465 The flux in genetic diversity brought about by fresh snowfall is in much need of greater 466 study. The amount of gDNA deposited by snow per hectare of land suggests that gene flow, via 467 known mechanisms of horizontal gene transfer, is likely widely seeded by continual deposition 468 from storm events. Ecosystem in-situ genetic diversity is thus likely to be in a state of higher flux 469 than previously acknowledged, with receiving ecosystems likely more capable of responding to 470 dynamic environmental conditions with updated 'genetic banks', literally always raining upon 471 them. As we argue that the biomass within meteoric snow is ecologically impactful, we 472 acknowledge that one may posit that what is deposited in snow would be just a fraction of the 473 biomass already existing at the soil / hydrologic interface. We point out that meteoric snow 474 would certainly be less consequential if it were uniformly distributed to all surface communities. 475 However, microbiota within snow are highly heterogeneous (even within the same storm) and 476 alternate genetic banks are deposited in different regions, some of which may be native to an 477 area already-but many are likely to be non-local. Within oceanic systems, McDaniels et al. 478 show that up to 47% of culturable bacteria are confirmed gene recipients by known mechanisms 479 of horizontal gene transfer (63); if one year of snowfall biomass represents even 0.1% of the 480 . CC-BY-NC-ND 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted April 13, 2018. ; https://doi.org/10.1101/300772 doi: bioRxiv preprint cells within the top 1 cm of receiving soil (assuming 1 × 10 9 cells cm -3 ), those cells would be in 481 sufficient quantity to impart new genetic material on the host community. If surface communities 482 demonstrated even a fraction of the horizontal gene transfer ability of oceanic communities, a 483 substantial portion of that entire 1 cm layer (and more) would be the beneficiary of new genetic 484 material-accelerating genome innovation (64). Bacteria / archaea genomic innovation and 485 resiliency has largely been driven by horizontal gene transfer (65) and we expect that 486 environmental systems comprised of these organisms are likely to depend upon this mechanism 487 for survival. Since lateral gene transfer increases genomic diversity at rates much higher than in-488 situ evolution alone (65), communities that receive higher fluxes of genetic material are likely to 489 be more resilient than those that receive very little. We demonstrate in our investigation that 490 geographic areas receive distinct biological input. This suggests that 'genetic banks' received by 491 surfaces during snow storms send different locales on alternate genetic uptake and evolutionary 492 pathways. Genetic transport by snowfall (and rain / ice / humidity) disallow microbial 493 communities at any given geography to remain isolated. In an Earth system that is increasingly 494 affected by climate change, the ability of local microbial communities to adapt will become 495 paramount; atmospheric microbial transfer and snow / rain / icing / humidity events should be 496 considered essential components in delivering novel genetic material to geographic areas. The 497 ability for soil / water receiving microbial systems to remain resilient will have demonstrable 498 effects on downstream ecological pathways such as macroscopic life and the mediation of soil / 499 water chemical equilibrium-including the management of contaminants. The 'everything is 500 everywhere' notion of microbial distribution across the globe can likely be better understood by 501 the interaction of Earth's compartments (lithosphere, hydrosphere and atmosphere) via transport. 502 . CC-BY-NC-ND 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted April 13, 2018. ; https://doi.org/10.1101/300772 doi: bioRxiv preprint On-going work with metagenomics, transcriptomics and single-cell genomics should better 503 inform the surface ecosystem-altering effects of the genetic potential of storm delivered events. 504 We suggest that bio-meteorology is likely to have impacts far beyond the Colorado Front 505 Range alone; everything that we describe here is globally relevant to the lithosphere / 506 hydrosphere / atmosphere interface in tropical, temperate and polar climates. Indeed, biology 507 likely needs to be better incorporated into meteorology and meteorologic modeling. Disparate 508 concepts ranging from ecological forest fire recovery to regional public health and pathogen 509 transport are likely to be affected by the geochemical and biological depositions that we have 510 described in snowfall (though these findings are likely also relevant to all meteoric waters 511 including rain, fog, icing events and humidity). Further, biometeorology could likely be better 512 informed and made capable of public-health related predictability with further microbiome 513 understanding of widespread and large-scale meterologic events. 514 515

Funding Information 516
A special thank you to the Zink Sunnyside Foundation for helping to fund this work. Funders did 517 not participate in research selection or execution, nor the decision to publish this work. 518 519 Acknowledgements 520 We would like to thank all members of the GEM (Geo-Environmental-Microbiology) Lab at 521 the Colorado School of Mines for helpful discussions. Specifically, thank you to Blake Stamps 522 and Gary Vanzin for feedback on bioinformatic techniques and experimental design. We also 523 would like to thank reviewers for helpful commentary that assisted in the improvement of our 524 work. 525 . CC-BY-NC-ND 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under  CC-BY-NC-ND 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted April 13, 2018. ; https://doi.org/10.1101/300772 doi: bioRxiv preprint   Figure S1: Rarefaction curves generated by rarefying bacterial / archaeal sequences at different 746 levels with random OTU sub-sampling. Each curve and color represents a different sample. A 747 . CC-BY-NC-ND 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted April 13, 2018.  Class-level assignments are listed first on the y-axis for each taxa, followed by a semi-colon and 767 its associated phylum. Taxonomy was assigned via the SILVA database by DADA2. These data 768 have been rarefied. 769 770 . CC-BY-NC-ND 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted April 13, 2018. ; https://doi.org/10.1101/300772 doi: bioRxiv preprint Figure S5: Relative abundances (more red = higher relative abundance) of bacterial / archaeal 771 (A) and eukaryotic (B) sequences recovered from snow samples pooled by Storm Origin. Class-772 level assignments are listed first on the y-axis for each taxa, followed by a semi-colon and its 773 associated phylum. Taxonomy was assigned via the SILVA database by DADA2. These data 774 have been rarefied. 775 776 Figure S6: Demonstration that the most influential genera listed in Table 1 are not 777 contamination. Numbers are relative abundances; more red means higher abundance. These data 778 have been rarefied the same as all other analyzed samples. 779 780 . CC-BY-NC-ND 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted April 13, 2018. ; https://doi.org/10.1101/300772 doi: bioRxiv preprint . CC-BY-NC-ND 4.0 International license a