CityUHK-led study reveals continued Antarctic ice loss this century, highlighting the importance of emissions reductions to mitigate coastal risks

An international research team led by City University of Hong Kong (CityUHK) has used a physics-informed artificial intelligence (AI) model to reveal that Antarctic ice loss will continue throughout this century, even if global warming is limited to the Paris Agreement's 1.5°C target. However, reducing greenhouse gas emissions can significantly limit the extent of ice loss. Under a high-emissions scenario, Antarctic ice loss could lead to an increase of up to 25.4 cm in global sea level by 2100. The study provides a clearer scientific basis for projecting sea-level rise and highlights the importance of rapid emissions reductions and green energy transitions for Hong Kong and other coastal regions worldwide.

The study was led by Professor Lin Yucheng, Assistant Professor in the School of Energy and Environment at CityUHK. The findings were published in the international scientific journal Nature Geoscience, in a paper titled "Committed Antarctic ice-sheet mass loss by the end of the 21st century".

Sea-level rise is one of the most visible signs of climate change, threatening low-lying coastal cities worldwide. On 24 September 2026, world leaders at the United Nations General Assembly adopted by consensus the UN Declaration on Sea Level Rise, reflecting global recognition of the threats that rising seas pose to coastal communities and sustainable development. More than 3 million people in Hong Kong and 40 million across the Greater Bay Area live in low-elevation coastal zones. Recent research found that approximately 70% of the extreme coastal flooding events Hong Kong has experienced since 2010 were attributable to human-caused climate change. Accurate projection of the rate and magnitude of future sea-level rise is therefore crucial for Hong Kong.

The Antarctic Ice Sheet is central to projections of future sea-level rise. Covering more than 14 million square kilometres, an area larger than China and India combined, it holds enough freshwater to raise the global mean sea level by 58 metres. However, scientists have struggled for decades to accurately project how the Antarctic Ice Sheet will evolve, as its future depends on multiple interacting factors, including human greenhouse gas emissions, changes in the atmosphere, ocean and sea ice, melting beneath ice shelves, and the movement of ice over the underlying bedrock. These interconnected processes create what scientists call "cascading uncertainty", which progressively widens the range of projections and leads research teams to reach different conclusions about whether Antarctica will gain or lose ice this century.

Traditional physics-based ice-sheet simulations require substantial computational resources, with a single simulation potentially taking days to complete. This makes it difficult for researchers to test every possible combination of physical assumptions. Professor Lin's team trained an AI model using an extensive archive of simulations accumulated through years of international research, enabling it to learn the relationships between various physical assumptions and simulation outcomes. Once trained, the model can project sea-level rise in a fraction of a second, a calculation that previously took days. This computational efficiency allowed the team to test millions of combinations of physical assumptions, compare the results against recent satellite observations, and retain only those consistent with the observed data, thereby improving the accuracy of predictions of the Antarctic Ice Sheet's future evolution.

The findings show that Antarctic ice loss is now very likely to continue throughout this century, even with strong greenhouse gas emissions reductions. Nevertheless, reducing emissions can significantly limit the extent of ice loss. Under a high-emissions scenario, rapid ocean warming around Antarctica could trigger a cascade of reinforcing effects: warmer ocean water melts ice shelves from below, causing them to thin; ice behind the shelves slides more rapidly over the underlying bedrock; and ice shelves that previously restrained the flow of ice into the ocean may break apart and lose their braking effect, allowing more ice to flow into the sea. Together, these processes could contribute up to 25.4 cm of global sea-level rise from Antarctica alone by 2100, enough to permanently flood the homes of more than 10 million people.

Professor Benjamin Horton, Dean of the School of Energy and Environment and Chair Professor of Earth Science at CityUHK, said, "Antarctica is no longer a distant problem. This study shows that ice loss during this century is now very likely, even with strong emissions reductions. But the scale of that loss is still in our hands: the emissions we avoid today will reduce the long-term threat to coastlines around the world."

Professor Lin said, "We have to transform our energy system to cut every emission we can, and we must do it quickly."

The new AI model also helps the research team identify the key factors determining the future evolution of the Antarctic Ice Sheet from more than 20 possible modelling choices. The study highlights three particularly important factors: how large ice shelves respond to a warming ocean, how ice slides over the underlying bedrock, and how the ocean and atmosphere warm across Antarctica. Professor Lin said, "We don't have unlimited resources to study every detail of an ice model, so we have to identify which details matter most."

The research team emphasised that the findings are far from the final word on Antarctica's future. Several physical processes are not yet represented in current simulations, including interactions between ice, ocean, atmosphere and solid Earth; the mechanisms of ice damage and failure; and the evolution of rivers and lakes beneath the ice sheet. As AI models can learn only from existing simulations, these omitted processes could substantially change scientists' understanding of Antarctica's future and affect whether coastal communities have sufficient time to prepare for rising seas. The study offers a faster approach to identifying the key factors influencing the future evolution of the Antarctic Ice Sheet.

Professor Horton added, "For Hong Kong and the rest of the Greater Bay Area, this research could not be more relevant. Professor Lin has combined physics, satellite observations and artificial intelligence to turn immensely complex Antarctic projections into clearer information for coastal decision-making. It is an outstanding example of the innovative, solutions-focused research being led by our young faculty."

The study was co-authored by researchers from CityUHK, the Commonwealth Scientific and Industrial Research Organisation (CSIRO), Victoria University of Wellington, Rutgers University, the University of New South Wales, the University of Tasmania and Durham University. It was supported by CityUHK and Project No. 36.9, "The Asia Sea-Level Knowledge Hub (ASK-Hub)", which is officially endorsed by the United Nations Educational, Scientific and Cultural Organization (UNESCO) under the UN Decade of Ocean Science for Sustainable Development (2021–2030).