Atmospheric rivers deliver Patagonian bioaerosol to the Antarctic Peninsula A 1,500 km latitudinal transect study of airborne microbial transport, with co-incident chemistry-transport-model evidence

Introduction

The atmosphere is a major route for global microbial dispersal. Bioaerosols—viable bacteria, fungal spores, viruses and biological remnants—circulate in the free troposphere and stratosphere^11,10, are exchanged between continents over thousand-kilometre scales^1,3,4,5, and impose seasonal and provenance-dependent biogeographic structure on remote aerial assemblages^2,8,14. Long-distance transport delivers crop pathogens^6,12, antibiotic-resistance determinants¹³, and even human pathogens⁵ across oceans, and is shaped by dust outbreaks^7,8, atmospheric circulation regimes^1,16, and host vehicle particles such as aeolian dust⁷. The Earth’s globalised aerobiome is increasingly recognised as a coherent ecological compartment with its own community-level dynamics¹⁵.

Antarctica is the most isolated terrestrial biome on the planet, ringed by the Antarctic Circumpolar Current (ACC) and the Polar Frontal Zone, and overlain by a circumpolar westerly storm-track that constrains meridional airflow^17,18. These oceanographic and atmospheric barriers have shaped a microbiota with deep evolutionary histories and substantial endemism^19,20, and the ACC fronts continue to imprint biogeographic structure on contemporary marine microbial communities^21,22,23. Yet isolation is not absolute: the Antarctic Polar Front acts as a partial barrier to bacterial dispersal while the ACC itself simultaneously connects sub-Antarctic provinces²¹, and storm-driven oceanic eddies and surface-wave transport carry temperate macroalgae across the Southern Ocean at rates much higher than once assumed²⁵. Antarctic terrestrial soils retain strongly regional bacterial communities⁵¹, but the open question is the rate and selectivity of atmospheric connectivity from temperate landmasses.

Studies of the Antarctic aerobiome have begun to disentangle the local vs long-range contributions. Boundary-layer aerosols over the Southern Ocean carry a predominantly marine bacterial signature with limited continental input, consistent with a pristine baseline state²⁶. A circumpolar survey of Antarctic air bacteria found low richness, high heterogeneity, and only ~4% of amplicon sequence variants shared with non-polar air, compatible with the polar air mass acting as a selective dispersal filter²⁷. Air arriving over isolated ice-free Antarctic soils does not fully explain extant soil community composition, suggesting strong selection during transit and at the receptor²⁸. Within the Antarctic Peninsula specifically, microbial assemblages in firn and snow cores have been used as tracers of warm air-mass advection from Patagonia²⁹, and air and fresh-snow fungi at maritime Antarctic sites are dominated by cosmopolitan taxa consistent with extra-polar deposition^31,32,33. Together these results sketch a picture in which long-range microbial input to Antarctica is real but episodic, and in which the most informative observations should target known atmospheric-transport events.

Atmospheric rivers (ARs)—long, narrow corridors of intense water-vapour transport from mid-latitudes—are now established as the dominant mechanism of poleward moisture delivery to the Antarctic Ice Sheet^37,40. Although rare (~3 days yr−¹ over the coast), ARs account for a substantial fraction of Antarctic precipitation, surface mass balance variability, and extreme melt events^37,41,50, and their frequency and intensity are projected to grow under continued warming^39,49. Recent work has extended the AR framework to aerosol transport, identifying polar aerosol atmospheric rivers—extreme poleward fluxes of dust, sea salt, black carbon, and organic carbon that may or may not co-occur with moisture-only ARs³⁸. South America—Patagonia in particular—is the dominant Southern-Hemisphere mineral-dust source^47,48, and Patagonia-to-Antarctica dust transport has intensified over the past two decades⁴⁶. Yet despite a now-mature atmospheric-science framework for these events, evidence for an associated microbial signal at the Antarctic receptor side remains thin—largely because dedicated, contemporaneous sampling of air on both ends of the corridor is rare. This study addresses that gap by sampling air, bulk soil, and rhizosphere along a 1,500 km latitudinal transect from Patagonia to the Antarctic Peninsula in austral summer 2022, with sampling at Risopatrón Base spanning a well-documented AR event on 2022-02-07/08, and by combining the resulting microbial inventories with reanalysis, dispersion-model back-trajectories, and a regional chemistry-transport simulation of the same event.

Methods (brief)

Sample design

A 1,500 km latitudinal transect from southern Patagonia (Puerto Natales, San Gregorio, Punta Arenas, Porvenir, Puerto Williams) to the Antarctic Peninsula (Risopatrón Base on Robert Island, South Shetlands; Yelcho Base on Doumer Island). Air sampled with Coriolis μ wet-cyclone samplers at 300 L min⁻¹ for 3 h (54 m³ per sample). Bulk soil and rhizosphere collected on-site; deposition, seawater, and buffer controls included at every site. Risopatrón sampled daily 2022-01-22 → 2022-02-14 (24 daily air samples, RA1–RA24, spanning before/during/after the AR). Yelcho sampled 2022-02-08 → 2022-02-28 (capturing during+after only).

Sequencing and bioinformatics

16S rRNA V4 (bacteria, primary axis), ITS1 (fungi), 18S rRNA (eukaryotes) on Illumina 2×150 bp / 2×250 bp. DADA2 v1.38 with truncLen=c(150,150), maxEE=c(2,2), pool=TRUE, seed 100. Taxonomy assigned against SILVA 138 (16S) and UNITE 10.1 (ITS). For 16S, the LRT dataset was merged with the Malard et al. 2022 Southern Ocean air reference²⁷ by re-running their FASTQs with identical DADA2 parameters to ensure byte-identical ASVs. Phylogeny: MAFFT —auto + FastTree 2.1.11 + midpoint root. Final merged 16S phyloseq: 368 samples × 38,665 ASVs (LRT 198 + Malard 170). Fungal phyloseq: 198 LRT samples × 11,403 ASVs (ITS1; justConcatenate=TRUE due to PE read overlap failure on long ITS1 amplicons).

Statistical analyses

Beta-diversity: weighted UniFrac with vegan::adonis2 (9999 permutations), with explicit alignment of sample_data row order to distance-matrix labels before each adonis2 call. Mantel test: vegan::mantel on community vs Haversine distance. Alpha: Chao1, Shannon (phyloseq::estimate_richness), Faith’s PD (picante::pd, include.root=TRUE). Within-sample phylogenetic structure: MPD via picante::ses.mpd; NRI = −ses.mpd$z; NTI = −ses.mntd$z, taxa.labels null, 999 perms. Between-sample turnover: βNTI (Stegen et al. 2013) on abundance-weighted βMNTD with 999 cophenetic tip-label shuffles in parallel. Stepwise weighted-UniFrac with cyclic-shift null (1999 perms). FAPROTAX 1.2.12 vendored locally for functional inference.

Contamination assessment (Fierer et al. 2025)

Following the recent low-biomass guidelines⁵², decontam v1.30 (prevalence method) was run with the five PBS/extraction buffer_control samples as the negative reference; sea_water_control was excluded (it’s a real environmental category; only 0.06% of its reads removed, sanity-confirming the exclusion). 408 ASVs were flagged at threshold 0.1 (1.2% of total). Top flagged genera Pseudomonas, Stenotrophomonas, Anaerobacillus, Empedobacter, Halomonas, Acinetobacter, Sphingomonas, Burkholderia, Brevundimonas are all on the canonical reagent-contaminant list of Fierer et al. 2025⁵². Headline analyses were re-run with these ASVs removed; all conclusions are robust to decontamination (only NTI drops out, as expected when removing dense reagent-ASV clusters).

Atmospheric / chemistry-transport evidence

ERA5 reanalysis for synoptic IVT and Z500. HYSPLIT (NOAA) 5-day back-trajectories at 50, 100, 500 m AGL arriving at each sampling site. CHIMERE regional chemistry-transport model output (nested d01/d02/d03 domains) for pDUST (Patagonian mineral dust tracer), pISOPA1 (terrestrial-vegetation isoprene-oxidation secondary organic aerosol), TracerG, pBCAR, NO₂, TSP at site-day resolution, plus full spatial fields for the AR window. On-site meteorology at Risopatrón Base from a Campbell CR1000X automatic weather station (wind, temperature, pressure).

Results

Figure 1. Global comparison: a distinct Antarctic-air community

The LRT air microbiome is compositionally distinct from globally curated air, soil, and marine references, and is internally structured by latitude. Puerto Williams sits in an intermediate position between Patagonian and Antarctic Peninsula air communities—the geographic bridge consistent with progressive turnover along the storm-track corridor.

Figure 2. LRT overview: phylum-level composition and Faith’s PD by medium

Air carries a phylogenetically narrow community relative to the on-site terrestrial reservoirs at the same sites—an expected signature of selective transport into the atmosphere and selective survival aloft, also visible in Antarctic Dry Valleys aerosols dominated by spore-forming Firmicutes³⁰.

Figure 3. Functional potential (FAPROTAX)

The functional inventory of Antarctic Air is biased toward an “atmospheric lifestyle”: methylotrophy/methanotrophy for one-carbon substrates available aloft, and sulfur-cycling pathways consistent with marine DMS-oxidation chemistry. These signatures are commonly seen in cold-desert and high-altitude microbial communities relying on atmospheric trace gases³⁵.

Figure 4. The atmospheric river itself: microbiology-independent evidence

This figure shows three independent windows on the same event: ground (A) → free troposphere (B) → modelled chemistry (C). The synoptic structure (B) is textbook AR^37,42,43; the on-site wind roses (A) confirm the arrival as a measurement (not a model), and the CHIMERE pISOPA1 (C) demonstrates that the AR delivered biogenic Patagonian terrestrial chemistry to the Peninsula—a signal that has its dust-side analogue in the increasing Patagonia→East-Antarctica dust deposition trend⁴⁶ and the Patagonian dust climatology of Gassó et al.⁴⁸

Figure 5. Does the bacterial community track the AR?

Whole-community PERMANOVA on weighted UniFrac is highly significant (R²=0.280, p=0.0014 for phase; R²=0.209, p=0.0024 for Before vs During+After). Faith’s PD jumps from 38.7 to 96.7 (~2.5×, p=0.024); MPD increases from 0.85 to 1.05 (p=0.012); NRI collapses from +10.4 to +1.0 (p=0.009)—the pre-AR Antarctic-air community is ~10 SDs more phylogenetically clustered than random, and the AR essentially erases this clustering. Between-community βNTI is significantly less negative across the AR boundary than within the Before window (Wilcoxon p=0.034), shifting the community from homogeneous-selection toward a stochastic regime: the AR delivers lineages broad enough to dilute the pre-AR phylogenetic-niche filter rather than installing a new one. Best-taxonomy source attribution (Patagonia broad / Malard SO Air / Risopatrón soil) refutes a marine origin (Malard SO carries near-zero prevalence of the AR-arrival genera) and supports a Patagonian terrestrial source for taxa such as Psychrobacter, Methylophilus, Empedobacter, Acinetobacter, Anaerobacillus. The AR signal is robust to decontamination: 0 of 168 representative AR-arrival ASVs are flagged by decontam at any threshold.

Discussion

Four independent lines of evidence converge on a single conclusion: the 2022-02-07/08 atmospheric river delivered a phylogenetically-broad, terrestrially-sourced Patagonian bioaerosol plume to the Antarctic Peninsula. The synoptic meteorology^37,41,42 (Figure 4B), the on-site wind/temperature signature at Risopatrón (Figure 4A), the modelled CHIMERE chemistry-transport delivery of terrestrial-vegetation biogenic SOA (Figure 4C) and Patagonian mineral dust^46,47,48, and the receptor-side bacterial community response (Figure 5) and a parallel fungal-ITS signal (Supplementary Figure S1) tell the same story through complementary lenses. This is, to our knowledge, the first study to pair a temporally-resolved Antarctic air microbiome time-series with co-incident chemistry-transport modelling of the same atmospheric event.

The phylogenetic-structure signal is particularly informative. Three complementary metrics—MPD, NRI, and βNTI—all show that pre-AR Antarctic-air community sits in a tight phylogenetic clique under strong homogeneous selection. The arriving Patagonian assemblage spans a much broader region of the bacterial tree: NRI collapses from +10 to ~0, MPD increases, and βNTI moves from the homogeneous-selection regime toward a stochastic regime. The interpretation is not that the AR installs a new selection regime opposing the pre-AR one, but that it releases the receptor community from its low-biomass selection filter by injecting a phylogenetically-broad temperate-source assemblage. This is consistent with prior work showing that strong selection acts on Antarctic-bound airborne microorganisms during long-range transport and at the receptor²⁸, and that connectivity across the Antarctic Polar Front is selective rather than uniform²¹. It complements the Uetake et al. finding²⁶ that Southern Ocean boundary-layer air is normally pristine and marine-dominated: the AR is precisely the kind of episodic mid-latitude intrusion that breaks the baseline.

Our results add direct biological evidence to the “polar aerosol atmospheric rivers” framework recently introduced by Lapere et al.³⁸—poleward extreme transport events that may or may not co-occur with moisture-only ARs and carry dust, sea-salt, black-carbon, and organic-carbon plumes. The 2022-02-07/08 case was a co-occurring AR + aerosol-river: ERA5 IVT, CHIMERE pDUST, and pISOPA1 spike together at Risopatrón. Our microbiology shows that the corresponding microbial fingerprint is also delivered, providing an in-situ ground-truth for the polar-aerosol-river concept. Under projected warming the Antarctic AR frequency is expected to rise^39,49 and Patagonian dust mobilisation has already intensified over the past two decades^46,47; together these trends predict a non-trivial future increase in the rate of viable bioaerosol delivery to Antarctica. For Antarctic conservation biogeography^17,25,51 this is the relevant time-scale: not the millennial residence-time of established endemics, but the decadal scale on which the present mix of episodic immigration events is set to accelerate.

Methodologically, we treated the recently-published Fierer et al. 2025⁵² contamination guidelines as a checkpoint rather than an afterthought. With 30 negative controls spanning buffer, deposition, and Coriolis blanks across the transect, decontam (prevalence method, buffer_control negative reference) flagged 408 reagent-contaminant ASVs (1.2% of the total). All six flagged headline genera (Pseudomonas, Stenotrophomonas, Sphingomonas, Burkholderia, Brevundimonas, Methylobacterium) match the canonical low-biomass reagent-contaminant list. None of the 168 AR-arrival representative ASVs are flagged at any threshold; the AR-event signal survives decontamination unchanged on the headline metrics (PERMANOVA, Faith’s PD, Shannon, MPD, NRI, βNTI), with only the unweighted NTI dropping out—an expected artefact of NTI being dominated by nearest-taxon distances within dense reagent-ASV clusters. We disclose this trade-off rather than relying on NTI for the manuscript’s headline claims.

Important caveats remain. Patagonian sampling for this dataset preceded the AR by ~6 weeks, so the strongest source-sink causal claim—contemporaneous source and receptor sampling during the AR—is not yet possible. For taxa present in both Patagonia and Risopatrón soil (most of the abundant AR-arrival genera) we cannot, from amplicon data alone, distinguish event-time atmospheric transport from local re-aerosolisation under high-wind AR conditions, or from previous AR-mediated deposition followed by local re-aerosolisation. Future campaigns should pair Patagonia + Peninsula sampling within an AR window, ideally with high-volume air sampling at intermediate elevations to resolve the vertical aerosol structure flagged by Lapere et al.³⁸, Gorodetskaya et al.⁴³, and Bryan et al.¹¹ The CHIMERE+ERA5+HYSPLIT evidence shown here is independent of the microbiology and provides the strongest available proxy for AR-event source attribution; combining it with explicit source-tracking and viability-resolving methods (e.g. RNA-based or qPCR-based quantification of viable cells) is a natural next step.

Supplementary material

Supplementary Figure S1. The fungal AR signal at Risopatrón

Data and code availability

Producing scripts, the methods/findings document, and the figure outputs are version-controlled in the project repository:

• 16S/figure_global_comparison.R, figure2_lrt_overview.R, figure3_faprotax_plot_byMedium.R, figure4_ar_event.R, figure5_ar_event_fungi.R, figure_atmospheric_synthesis_11MAY26.R
• 16S/decontam_controls_11MAY26.R, decontam_AR_besttax_check_11MAY26.R, decontam_robustness_11MAY26.R, decontam_phylo_metrics_11MAY26.R, decontam_betaNTI_11MAY26.R
• 16S/ar_patagonia_attribution_v3.R
• Supplementary tables: decontam_contaminants_11MAY26.tsv, decontam_AR_besttax_check_11MAY26.tsv, figure4_betaNTI_pairs_08MAY26.tsv
• The full methods + findings memo: methods_and_findings_11MAY26.docx

Sequencing reads will be deposited at NCBI SRA upon manuscript acceptance. Reference databases used: SILVA 138 (16S), UNITE 10.1 general FASTA release (ITS), PR2 v5 (18S).

References