Remote sub-Antarctic islands support globally significant seabird populations, yet extreme remoteness and logistical constraints limit conventional monitoring approaches. Australia's Heard Island and McDonald Islands (HIMI) exemplify these challenges, with restricted access, complex terrain, and limited opportunities for ground or helicopter-based surveys. Here, we demonstrate the application of an integrated drone and artificial intelligence (AI) workflow to enable comprehensive seabird monitoring under such constraints. Across two research voyages in 2025/26, we conducted ship- and terrestrial-based drone surveys targeting colonial seabirds. Surveys covered the majority of the ice-free coastal environment, accumulating more than 54 hours of flight time across 120 flights and with over 1,600 km flown. AI-assisted analysis was used for automated detection, classification, and counting of individuals, substantially increasing efficiency and consistency relative to manual methods. Notably, surveys resulted in the detection of several previously unrecorded penguin colonies and provided baseline data for McDonald Island where landings and helicopter operations were not possible. Our results demonstrate that drone–AI workflows provide a scalable, low-disturbance, and logistically flexible approach for monitoring colonial seabirds in remote and protected environments.

The sub-Antarctic Heard Island and McDonald Islands (HIMI) form one of Australia’s most remote territories, recognised for their exceptional ecological value through World Heritage listing and marine reserve status. These largely undisturbed ecosystems provide important habitat for diverse species, but their isolation limits opportunities to characterise biodiversity or to monitor ecosystem changes. During the 2025-2026 field season, the Australian Antarctic Program completed the first HIMI management and research voyages in two decades. As part of this campaign, we conducted an environmental DNA (eDNA) survey to improve understanding of native biodiversity, establish baseline data, and support future monitoring and marine reserve management.

Marine eDNA samples were collected across the HIMI region from the RSV Nuyina. Surface water was filtered at least three times daily, while passive sponge samplers were deployed alongside other scientific equipment to capture eDNA from deeper waters. At Heard Island, field teams collected terrestrial (freshwater) and coastal (seawater) eDNA samples using syringe filtration. This presentation reports preliminary eDNA metabarcoding results, and highlights some of the challenges of conducting eDNA surveys in such remote and extreme conditions.

High pathogenicity avian influenza (HPAI) has spread across the sub-Antarctic, causing significant wildlife impacts. We report its first detection in Australian territory at Heard Island and McDonald Islands, which support over one million breeding seabirds and seals. Drone and ground surveys (October 2025, January 2026), combined with viral genome analysis, confirmed infection with Influenza A H5N1 clade 2.3.4.4b at Heard Island. Drone surveys revealed mass mortality in southern elephant seals, with 8,573 pups (62%) recorded dead by the final surveys. Mortality reached up to 97% in some harems and increased at an average rate of 5.6% per day. Based on the average (76%) mortality in the final surveys, total estimated pup mortality at Heard Island was 13,359 (from a total population of 17,364 pups), though this may be an underestimate. HPAI was detected in six of nine species tested and led to elevated mortality in king and gentoo penguins. Phylogenetic analysis indicates the virus was introduced from Crozet Islands, with an estimated arrival around August 2025. These data show the continued easterly spread of HPAI around the sub-Antarctic, with severe but heterogeneous impacts across taxa. Our results demonstrate the value of drones for large-scale monitoring, underscoring the need for continued and enhanced HPAI surveillance across the Southern Ocean.

Australia’s sub-Antarctic Heard Island is one of the most intact ecosystems on Earth. Its World Heritage status is due to its “pristine and largely untouched ecosystems”. Yet the invasive grass Poa annua (annual bluegrass) is established on the island, being first detected in 1987. Heard Island has experienced rapid glacial retreat in recent decades, exposing new ground and creating new opportunities for its establishment. As climate continues to change and impact the islands’ glaciers and native vegetation, we need to determine the current distribution of P. annua and how it might change in the near future. The 2025/26 HIMI field expedition provided a rare opportunity to assess the current distribution of P. annua. We found P. annua has spread into new areas not previously detected in 2004, 2000 or 1987 across multiple sites. Future work using vegetation data, high resolution satellite imagery, digital elevation models and drone imagery collected during the field campaign will be integrated to develop a habitat suitability model to identify areas at risk of future establishment. This work will inform future targeted surveillance and support biosecurity planning for one of Australia's most significant World Heritage Areas.

Heard Island has lost approximately 22% of its glacier area since 1947, with retreat rates accelerating markedly over recent decades (Tielidze et al., 2025). The progressive exposure of new coastal terrain following glacial retreat creates opportunities for vegetation colonisation and change, yet the spatial extent, rate, and species composition of these changes remain poorly quantified. Here we analyse broadscale vegetation change in the Stephenson Glacier retreat zone of southeastern Heard Island using Landsat 7 ETM+ and Sentinel-2 MSI archives spanning the early 2000s to 2024. Cross-sensor harmonisation and multi-year compositing were used to minimise spectral inconsistencies, cloud contamination, and seasonal variability. We quantify three forms of change: vegetation establishment within newly exposed terrain, increases in satellite-derived greenness within established communities, and vegetation loss in coastal areas affected by erosion. The Stephenson Glacier margin is identified as a zone of rapid recent change.

This analysis forms the temporal backbone of a broader multi-scale monitoring framework for Heard Island. Future work will integrate a high-resolution satellite archive spanning 1991–2025 with ground-truth quadrats, repeat photography, and UAV multispectral, hyperspectral, and LiDAR surveys collected during the 2025–2026 HIMI campaign. These data will support investigation of species colonisation dynamics across recently deglaciated terrain. Together, these data offer an unprecedented opportunity to reconstruct vegetation change from landscape to community scale in a rapidly changing sub-Antarctic environment.

Antarctic fur seals (Arctocephalus gazella) were reclassified as Endangered by the IUCN in April 2026 following a global decline in breeding adults from ~2.2 million in 1999 to 944,000 in 2025. Population trajectories vary regionally: colonies in the Atlantic sector are declining sharply, whereas those in the Indian and Pacific Oceans are increasing. Heard Island, located on the Kerguelen Plateau within Australian territory, represents an important but infrequently surveyed breeding site, with no published census since 2000/01.

We report pup production from ground surveys conducted in 2003/04 and 2025/26, supplemented with drone surveys of an additional 7 breeding sites to improve island-wide coverage. In 2025/26, 1,610 pups were recorded across five sites, compared to 1,278 in 2003/04, with increases observed at all major colonies. The largest growth occurred at Fairchild Beach (+78%), with the highest proportional increases at Stephenson Lagoon–Spit Camp and Compton Spit. Estimated growth averaged ~1.9% per year since 2000/01, indicating a transition from rapid recolonisation to slower sustained growth. Ground and drone counts agreed within ~5%, supporting this combined approach. In contrast to global declines, the Heard Island population is demonstrably increasing and represents a critical stronghold within the Indian Ocean sector.

Remote and environmentally dynamic regions such as Antarctica and the sub-Antarctic present significant challenges for the management and conservation of cultural heritage sites. In many cases, in-situ preservation of these sites is not practical due to limited accessibility and visitation.
The former Australian station at Atlas Cove, Heard Island represents a site that is subject to extreme weather, abundant wildlife, potential volcanic activity, and is challenging to visit. This study presents the results of, and workflow for, developing high-fidelity digital twins of remote polar heritage sites through the integration of mobile Simultaneous Localization and Mapping (SLAM) LiDAR and emerging Gaussian splatting techniques.
Field data were acquired using handheld SLAM LiDAR to rapidly capture a structural model with <2 cm precision. The resulting point clouds were processed and combined with photogrammetric imagery to generate dense radiance fields using 3D Gaussian splatting, enabling photorealistic rendering at interactive frame rates. This hybrid approach addresses limitations of traditional photogrammetry in low-texture, while overcoming the computational overhead of conventional mesh-based models.
The resulting digital twin of the Atlas Cove station demonstrates high geometric accuracy and visual fidelity, preserving both structural detail and site context, including artefacts and environmental features. The workflow supports scalable data capture, and efficient processing, making it suitable for long-term monitoring, virtual access, and novel approaches to interpretation and transmission of heritage values.