Remote sensing reveals Antarctic green snow algae as important terrestrial carbon sink

Antarctic terrestrial ecosystems are extremely limited in extent, yet photosynthetic organisms significantly contribute to local ecology. Beyond vegetated ice-free ground, conspicuous green and red snow algal blooms occur on coastal snowfields and may represent a major primary producer influencing nutrient and carbon cycling. The Antarctic Peninsula has already warmed >1.5 °C over pre-industrial levels, with further warming projected, potentially expanding areas suitable for plant colonisation. However, most remote sensing estimates of vegetation exclude snow algae due to their spectral characteristics. The study aims to map and quantify the distribution, biomass, and carbon sink capacity of green snow algae across the Antarctic Peninsula and to assess environmental controls on their distribution and likely responses to climate warming.

Early Antarctic observations documented snow algae blooms and subsequent work showed diverse species and roles in nutrient and carbon cycling. Remote sensing of Antarctic terrestrial vegetation has focused on exposed ground, often excluding snow algae because classical vegetation indices are unsuitable. Northern Hemisphere studies using airborne hyperspectral imaging and satellite observations linked algae to bioalbedo effects and melt enhancement, especially in Greenland. Prior approaches for snow and ice algae have included hyperspectral detection and bioalbedo models, but multispectral satellite data present challenges due to limited spectral/spatial resolution, BRDF effects over snow, terrain complexity, low sun angles, cloud cover, and summer snowfall. These limitations motivate new methods tailored to detect chlorophyll absorption features of snow algae using Sentinel-2 multispectral data, complemented by in situ validation.

The study combined multi-year Sentinel-2A/B summer imagery (2017–2019) with two Antarctic field campaigns (Ryder Bay, Adelaide Island, 2017/18; Fildes Peninsula, King George Island, 2018/19). Field work: visible green-dominant blooms were sampled on 10 × 10 m grids. Hemispherical directional reflectance factors (HDRFs) were collected using a field spectrometer under consistent geometry, and snow was sampled for cell counts, dry biomass, nutrients, and C/N isotopes. Melted snow was analysed: cells were counted via ImageJ from brightfield microscopy (haemocytometer), dry mass measured by filtration and drying, nitrate and phosphate measured colorimetrically, and %C, %N, δ15N by elemental analysis and mass spectrometry. Thickness of the algal surface layer and snow density were measured to scale volumetric to areal concentrations. Net carbon exchange rate (NCER), ecosystem respiration (ER), and gross ecosystem photosynthesis (GEP) were measured using an IRGA with a clear chamber placed over blooms during light and dark periods, recording PAR, temperature, and CO2 fluxes. Remote sensing: Hyperspectral HDRFs were convolved to Sentinel-2 spectral response and a scaled integral index (IB4) computed, measuring the area of the chlorophyll-a absorption centered near 665–680 nm using Band 4 relative to Bands 3 and 5. A linear regression related IB4 to measured cell density (n=91), yielding r=0.85 (P<0.01) and a detection limit of ~4.4 × 10^3 cells ml^-1. Minimum detectable bloom area was empirically estimated at ~11 m^2 assuming central crossing of a pixel. Sentinel-2 processing applied the IB4 metric to imagery, with mask filters to remove false positives (other vegetation, crevasses, mixed pixels) and further filters requiring at least three adjacent positive pixels and biomass above the regression y-intercept. Pixel cell densities (cells ml^-1) were converted to areal densities (cells m^-2) using measured average algal layer thickness (9.05 mm) and snow density (0.58 ml melt cc^-1 snow). Biomass per pixel was computed using the average measured dry mass per green algal cell (2.4 × 10^-8 ± 2.2 × 10^-8 g), with blank corrections for non-algal mass. %C and %N were used to estimate elemental masses. Geospatial analyses used REMA DEM, RACMO 2 m temperature model, and penguin colony databases to relate blooms to elevation, temperature, slope, and proximity to marine fauna. Validation used published sightings, a SCAR visitor survey, and available images; agreement yielded a kappa score of 0.81 (n≈25).

• Detection and extent: 1,679 individual green snow algal blooms were identified. Total mapped area was ~1.95 × 10^6 m^2 (1.9 km^2), comprising ~1.9 × 10^4 Sentinel-2 pixels across ~2.7 × 10^5 km^2 of the Peninsula study area. Individual blooms averaged 1,043 m^2 (range: ~300 to 145,000 m^2). Validation achieved a kappa of 0.81 against independent observations. • Biomass and carbon: Estimated total dry biomass was ~1.3 × 10^3 tonnes for green blooms per season, corresponding to ~479 tonnes of carbon (based on measured %C). Areal biomass ranged from ~5 to 5,791 g dry mass m^-2 (mean ~58 g m^-2), in good agreement with in situ averages (~30 g m^-2). Propagated uncertainty on biomass estimates was approximately +564% / −5% using mean parameter values. • Cell concentrations: Remote-sensed pixel cell densities ranged 1.9 × 10^4 to 1.7 × 10^5 cells ml^-1 (mean ~4.2 × 10^4), consistent with in situ averages (~2.2 × 10^4 cells ml^-1). The IB4–cell density regression showed strong correlation (r=0.85, P<0.01) and a detection limit of ~4.4 × 10^3 cells ml^-1. • Distribution controls: Blooms occurred mainly on the western Peninsula between 62.3°S and 68.1°S, corresponding to regions with average summer air temperatures above 0 °C (liquid water availability). Blooms were predominantly at low elevations (mean 14.8 ± 9.0 m a.s.l.) on flat to moderate slopes (mean 14.5° ± 12.9°); no blooms >1,300 m^2 were found on slopes >30°. No consistent aspect effect was observed. • Marine nutrient influence: 49% of blooms were within 100 m of the sea; 60% were within 5 km of a penguin colony. Within 1 km of colonies, blooms tended to be larger (mean 1,257 vs 960 m^2; t=1.4; P<0.16) and had significantly higher mean cell concentrations (4.1 × 10^4 vs 3.7 × 10^4 cells ml^-1; t=6.4; P<0.01). Field biogeochemistry showed elevated nitrate and phosphate and enriched δ15N near marine fauna, indicating trophic nitrogen inputs. Estimated seasonal N demand to support observed green algal growth is ~71.7 tonnes (≈3.1 g bioavailable N m^-2 season^-1), implying continued nutrient resupply through melt mobilisation and/or direct inputs. • Carbon flux: To accumulate the observed biomass over an assumed 122-day season with 17 h day^-1 photosynthesis, blooms would require an average NCER of approximately −0.064 µmol CO2 m^-2 s^-1, closely matching in situ NCER (average ~−0.08 µmol CO2 m^-2 s^-1). ER averaged ~0.07 µmol CO2 m^-2 s^-1, indicating active heterotrophic respiration within the snowpack; nevertheless, green blooms acted as a net seasonal CO2 sink. • Latitudinal trends and exemplars: Average bloom area and maximum elevation increased toward lower latitudes (north). Large blooms adjacent to chinstrap and gentoo penguin colonies on Robert and Nelson Islands (South Shetlands) included the largest area (~145,000 m^2) and highest elevation (~99 m a.s.l.) observations; one site exhibited the highest observed biomass (~2.1 kg C m^-2). Biomass was highly skewed, with ~95% of biomass from ~0.05% of blooms (few very large blooms).

The study demonstrates that green snow algae form a measurable and regionally significant terrestrial carbon sink on the Antarctic Peninsula, governed primarily by the availability of liquid water (temperatures >0 °C) and nutrient inputs, especially from marine fauna. By explicitly mapping blooms and quantifying biomass, the work fills a remote sensing gap for Antarctic biota not captured by classical vegetation indices. The strong association with low-elevation coastal zones and with penguin colonies highlights the role of marine-derived nutrients in enhancing algal growth and biomass accumulation. Flux measurements and biomass-based estimates show that green snow algae are net CO2 sinks during summer despite concurrent heterotrophic respiration in the snowpack. Under projected warming, the 0 °C isotherm will move upward and southward, expanding potential habitat. Observed northward increases in bloom size and elevation, particularly on larger islands with extensive ablation zones and proximal bird/seal colonies, suggest that expansion on such landmasses will likely outweigh biomass losses from small low-lying islands that may lose seasonal snow cover. Consequently, a net increase in green snow algae extent and biomass is anticipated as warming proceeds. However, potential shifts toward fewer, larger blooms and the loss of blooms on small islands could reduce local terrestrial biodiversity, given the multi-species composition of green-dominant communities and limited knowledge of their dispersal and life histories.

This work provides the first Antarctic Peninsula-wide map and quantitative estimate of green snow algae distribution and biomass using Sentinel-2 imagery validated by field spectroscopy, cell counts, and biogeochemistry. Green snow algae cover ~1.95 × 10^6 m^2 and accumulate ~1.3 × 10^3 tonnes dry mass (~479 tonnes C) per season, acting as a net seasonal carbon sink. Their distribution is constrained by positive summer temperatures and strongly influenced by proximity to marine nutrient sources. Warming is projected to increase algal habitat on larger islands and the mainland, leading to a net biomass increase despite likely losses on small low-lying islands. Future research should incorporate red and orange snow algae into remote sensing estimates, employ higher-resolution and more frequent satellite observations with extensive ground truthing to reduce uncertainties, and conduct large-scale, season-long carbon flux measurements. Improved understanding of species composition, dispersal mechanisms, life cycles, and nutrient supply dynamics will refine predictions of ecosystem responses and biodiversity outcomes under climate change.

Remote sensing focused on green-dominant blooms; red/orange blooms were largely excluded due to spectral ambiguity from carotenoids and potential confusion with mineral dust. Multispectral Sentinel-2 data and BRDF effects necessitated stringent filtering that may have removed some smaller or mixed pixels and introduced commission/omission errors. Biomass estimates integrate chlorophyll signals across 10 m pixels, potentially overestimating visible area for subpixel blooms, while some blooms are obscured by overlying snow, leading to underestimation. Propagated uncertainties in biomass are high (+564%/−5%) due to variability in cell mass, layer thickness, density, and spectral relationships. Cloud cover, snowfall, terrain-induced illumination variability, and limited temporal sampling (seasonal snapshot) further constrain detection. Field validation was limited to two regions and may not capture full spatial variability in species composition, physiology, and nutrient regimes.