Sustained greening of the Antarctic Peninsula observed from satellites

The Antarctic Peninsula (AP) has warmed rapidly over the past ~60 years, at rates exceeding the global average, with the strongest trends in West Antarctica and the AP. Although a temporary pause in AP warming occurred from 1999–2014 due to natural variability, projections indicate continued warming (~0.34 °C per decade to 2100), increased precipitation, and longer growing seasons. Most assessments have focused on cryospheric responses (for example, widespread glacier mass loss), yet the anticipated expansion of ice-free area underscores the need to understand terrestrial biological change. Moss-dominated ecosystems are central to the development of vegetated ground, organic soil formation, and facilitation of higher plant colonization. While vertical accumulation in AP moss banks reflects multi-decadal warming, the extent, continuity, and controls on lateral spread remain unclear. This study investigates whether an AP-wide ‘greening’—analogous to trends observed across other cold-climate ecosystems—is underway by quantifying the spatial extent, rates, and direction (greening vs. browning) of vegetation change across 1986–2021 using satellite-derived vegetation indices.

Prior work documented AP warming and its climatic drivers, including a recent pause followed by continued anthropogenic warming. Terrestrial responses have been inferred from site-based observations since the 1960s and palaeoecological records showing increased moss productivity and vertical accumulation since ~1950. However, satellite-based mapping of AP vegetation extent existed primarily for a single year (2001), leaving multi-decadal trends unquantified. Global literature reports complex greening and browning patterns in Arctic moss-dominated systems and links greening to warming and changing moisture availability. On the AP, native vascular plants have expanded their ranges, and moss ecosystems are key to lateral spread of vegetation and soil development. Biosecurity concerns are rising due to tourism and research traffic, with non-native species introductions posing risks as climatic constraints relax. Remote sensing approaches using NDVI thresholds have previously been validated in the AP, and Tasseled Cap Greenness (TCG) offers complementary sensitivity to vegetation on bright substrates. Despite these insights, AP-wide, long-term, spatially explicit quantification of vegetation change had been lacking prior to this study.

The authors analyzed Landsat 4, 5, 7, and 8 Tier 2 top-of-atmosphere reflectance imagery from 1986–2021 via Google Earth Engine to quantify AP greening at 30 m resolution. To standardize seasonality and maximize vegetation detection, only March images were used (end of growing season; preliminary tests showed highest vegetated pixel returns in March). Images were filtered to cloud cover over land <40% and to USGS image quality value = 9. Landsat 4/5/7 imagery was radiometrically calibrated to Landsat 8 OLI using published coefficients; overlap analyses indicated Landsat ETM+ underestimated NDVI>0.2 area by ~9.58% and TCG area by ~5.68%, so NDVI areas from Landsat 4/5/7 were increased by 10% and TCG areas by 6% for consistency. Vegetation indices computed per image were NDVI (threshold >0.2 denoting ‘almost certain’ vegetation presence; additional interpretive thresholds at 0.05 and 0.1) and TCG (>0 as presence), with annual mosaics of maximum NDVI and TCG generated to capture maximum lateral extent each year. Cloud and snow were masked using the Fmask algorithm to ensure vegetation signals originated from unobscured land. To reduce erroneous high-latitude/high-altitude reflectance effects and constrain detection to plausible growth zones, land above 300 m a.s.l. was masked. The study area was tessellated into 5,000 km² hexagonal sample units covering all ice-free land (with a 300 m buffer to account for geolocation and data issues). For each hexagon and for the whole AP, annual area with NDVI>0.2 was computed, and space–time trend analysis using the Mann–Kendall test was performed (ArcMap 10.8.1 Space Time Cube tools), treating no-data years as zeros with a 1-year time step and distance interval equal to hexagon height. Observable land area per year (after masking) was quantified to assess potential confounding by data availability. Supplementary validation included very-high-resolution WorldView-2 (2 m) imagery around Robert Island (2013–2016) to examine local changes and coherence with Landsat-derived trends. Data and code are openly available; the GEE JavaScript extracts annual maximum NDVI>0.2 and TCG>0 extents and unmasked area.

• Statistically significant AP-wide greening from 1986 to 2021. Area with likely vegetation cover (NDVI>0.2) increased from 0.863 km² (1986) to 11.947 km² (2021). Mann–Kendall test: r = 2.31, P = 0.021 (NDVI); TCG corroborated the trend (TCG>0 area 1986–2021: τ = 2.54, P = 0.011). - Rate acceleration: mean rate 1986–2021 = 0.317 km² yr⁻¹; recent 2016–2021 window = 0.424 km² yr⁻¹, exceeding earlier windows (1986–2004: 0.291 km² yr⁻¹; 2004–2016: 0.310 km² yr⁻¹). - Spatial extent: positive increases are widespread across the western AP, from ~68.5° S to ~61° S (northern South Shetland Islands). - Observable land not the main driver: weak correlations between interannual variation in observable ice-free land and vegetated area (NDVI R² = 0.351; TCG R² = 0.272) indicate trends are not artefacts of data availability. - High-resolution validation: WorldView-2 imagery around Robert Island showed an 18.7% increase in vegetated area (2013–2016), coherent with Landsat-based trends; largest established patches showed slight browning while expansion occurred mainly in newly colonized areas. - Temporal and spatial complexity: Nonlinear time series reflect local greening and browning mosaics (e.g., browning on eastern Ardley Island), influenced by microtopography, moisture availability, wind exposure, orographic effects, interannual variability, and potential impacts of penguin colonies (short-term trampling/guano vs. longer-term N enrichment). - Possible climatic linkage: The post-2016 acceleration coincides with decreased Antarctic sea-ice extent and a strongly positive Southern Annular Mode, potentially fostering warmer, wetter conditions and increased moisture availability conducive to plant growth.

The findings provide AP-wide, multi-decadal evidence that moss-dominated terrestrial ecosystems are expanding in response to recent climate change, directly addressing the question of whether an Antarctic ‘greening’ is underway. The statistically significant increase in vegetation cover, its acceleration in recent years, and the broad spatial footprint indicate that warming and associated hydroclimatic changes are already reshaping AP terrestrial ecosystems. The coherence between NDVI- and TCG-based trends, alongside high-resolution validation, strengthens confidence that changes reflect real biological expansion rather than observation artefacts. Nevertheless, localized browning in some areas underscores the role of microclimatic controls (topography, wind, moisture), biological interactions (e.g., penguin colonies), and increasing climate variability and extremes. Overall, the results suggest increasing productivity, vigour, and lateral spread of moss communities, with implications for ecological connectivity, soil development, trophic interactions, and potential homogenization across AP biogeographic regions as ice-free area expands.

This study delivers the first AP-wide, annually resolved satellite assessment demonstrating sustained and accelerating greening of moss-dominated terrestrial ecosystems from 1986 to 2021. By leveraging a standardized, reproducible GEE workflow with cross-sensor calibration and rigorous masking, the authors quantify significant increases in vegetation extent, corroborated by complementary indices and high-resolution imagery. The results imply that continued warming will drive widespread terrestrial ecological change across the AP, including enhanced productivity and lateral expansion of vegetation, altered connectivity among biogeographic regions, and potential facilitation of higher plant colonization. Future research should prioritize long-term ground-based validation and ecohydrological monitoring, integration of multi-sensor high-resolution observations, refinement of vegetation detection thresholds across seasons, assessment of microclimate–topography interactions, and evaluation of biosecurity risks and species migration dynamics as climate constraints relax.

• Persistent cloud and snow cover constrain observations; although masked, interannual variation in observable land remains a source of uncertainty. - Use of Tier 2 top-of-atmosphere reflectance and radiometric calibration across sensors introduces residual cross-sensor differences; applied percentage adjustments are approximations, not full corrections. - Reliance on NDVI>0.2 as an ‘almost certain’ vegetation threshold may miss low-vigour or mixed pixels and could be influenced by a lengthening growing season pushing pixels over the threshold without substantial biomass change. - TCG is less sensitive in mixed pixels and to early colonization stages; differences between NDVI and TCG responses complicate interpretation. - Areas above 300 m a.s.l. were masked to avoid BRDF/topographic artefacts, potentially excluding legitimate high-altitude expansions. - Spatial tessellation and treatment of no-data years as zero may influence trend statistics. - Limited field validation across the AP; localized processes (e.g., penguin colonies, microtopography, moisture variability) and extreme events may drive local browning not fully captured by remote sensing alone.