Research conducted by a Cambridge institute sheds light on a data gap in the processing of satellite images of Antarctic ice and the corresponding predictions regarding their melting dynamics.

Global warming has exerted its formidable effects on polar regions first and foremost. Among these, the Antarctic Peninsula has experienced average summer temperatures rising by over 3°C since the 1970s. The Southern Ocean has absorbed up to 75% of the Earth’s excess heat and 40% of its carbon dioxide. «Warmer and more acidic oceans are impacting Antarctic ecosystems, with many penguin colonies shrinking and, in some cases, disappearing altogether» [source: Antarctic and Southern Ocean Coalition].

In a World Economic Forum article published on 15 November 2023 (“Antarctica is melting faster than ever. Here’s what we can do about it“), the authors highlight how, since 2016, the sea ice surrounding Antarctica has been diminishing at a rate of 150 billion tonnes per year.

«The primary cause is the warming oceans, which not only directly melt the ice sheet but also thin the floating ice shelves surrounding Antarctica that hold the ice sheet on the land. As the shelves weaken, more ice flows into the sea, raising sea levels», they observe.

A more recent study by Stanford University (“Heterogeneous Basal Thermal Conditions Underpinning the Adélie-George V Coast, East Antarctica” – Geophysical Research Letters, 19 January 2024), expresses concern for East Antarctica, which researchers have somewhat neglected, deeming it more stable compared to West Antarctica, which has been the focus of most recent attention.

«The Wilkes subglacial basin in East Antarctica is the size of California – explains the research team – and contains enough ice to raise global sea levels by over 3 metres. Few analyses have been conducted in this region, which holds an enormous volume of ice that has remained relatively stable until now. However, we are observing, for the first time, how close the temperature at the base of its ice sheet is to potential melting».

As part of a study on the floating ice shelves surrounding the Antarctic ice sheet, a group of British researchers focused on the meltwater resulting from surface ice melting during the austral summer, along with the associated snowmelt pools (slush) and lakes (stagnant water).
Since meltwater affects the stability of ice shelves, potentially causing fractures and collapses, it is crucial to monitor slush and lakes. Specifically, the former are more challenging to identify within satellite images, which are always rich in various types of data. Hence, the British researchers had the idea of training a machine learning algorithm to recognise them, with the aim of eventually mapping their presence across Antarctica.
In the future, having an AI system capable of providing a comprehensive view of the Antarctic ice melt process, including the surface ice melting during the austral summer, would result in more precise calculations of meltwater quantities, thereby enabling more accurate predictions of sea-level rise.

The global impacts of antarctic ice melt

The consequences of Antarctic ice melt (technically referred to as “fusion”) encompass a wide range of environmental and climatic impacts affecting the entire planet. The direct repercussion is sea-level rise, which threatens hundreds of millions of coastal inhabitants worldwide. Even a few centimetres of rise can lead to floods, erosion, and storm surges, causing significant damage to homes, infrastructure, and land [source: Intergovernmental Panel on Climate Change].

«A global average sea-level rise of one metre by 2100 would cause an increase in annual flood damage by two to three orders of magnitude, affecting up to 20.3% of global GDP. However, this might be an underestimate, as we cannot rule out a two-metre rise in global average sea level by 2100» estimates a study published in Nature four years ago (“Projections of global-scale extreme sea levels and resulting episodic coastal flooding over the 21st century”, 30 July 2020).

Analysts at the World Economic Forum also warn of the vicious cycle triggered by Antarctic ice melt, which is responsible for further increases in global temperatures. As the ice melts, larger expanses of open water are exposed, where «the darker water absorbs more solar radiation, trapping more heat within the Earth’s system».

Another dangerous mechanism set in motion by the melting of the Antarctic ice sheet, as noted by the WEF, is the dilution of seawater with freshwater, making it less dense. This risks «slowing down the great ‘conveyor belt’ that drives global ocean circulation, altering nutrient distribution, global weather patterns, and threatening food security».

Focus on meltwater from antarctic ice surface melting

A groundbreaking study by a group of researchers from the Scott Polar Research Institute, Cambridge (UK), published in Nature Geoscience on 27 June 2024 (“Substantial contribution of slush to meltwater area across Antarctic ice shelves”), highlights lesser-known but crucial aspects of Antarctic ice melting processes.

The team focused on slush pools – composed of mixed snow and water – found on the floating ice shelves of Antarctica. These pools form from surface ice melting during the austral summer (21 December to 20/21 March) when temperatures naturally rise.

There is considerable variability regarding the onset, duration, and extent of surface ice melt. What is certain is that meltwater is stored either as stagnant water (forming lakes on the ice shelves) or as slush, with significant implications for the dynamics affecting ice shelf fractures, the researchers from the English university note.

The fact that current climate models do not account for meltwater and its deposits is, according to the authors, a concerning limitation.

Why meltwater affects ice shelf stability

The risk associated with meltwater (in the form of lakes or slush) is its potential to infiltrate existing cracks in the ice shelves, widening them and causing fractures. In more fragile ice shelves, this can lead to collapse, resulting in ice release into the ocean and a subsequent rise in sea levels.

«Since slush is more solid than meltwater, it won’t cause hydrofractures in the same way that lake water does, but it is definitely something we need to consider when predicting whether and how Antarctic ice shelves might collapse», the team emphasises.

However, slush – along with stagnant lake water – impacts the melt rates of Antarctic ice: being less white than snow and ice, it absorbs more heat from the sun, leading to increased ice melt.

This additional melting is not factored into current climate models, which «could lead to underestimating predictions about ice sheet melt and the stability of Antarctic ice shelves», the authors comment.

Machine learning aids mapping of slush areas

Previous studies have utilised high-definition satellite images to analyse and monitor variations in surface meltwater on Antarctic ice shelves.

«But no study, » notes the Scott Polar Research Institute team, «has mapped meltwater areas across all Antarctic ice shelves over multiple seasons. Additionally, almost all studies to date have only mapped stagnant water, not slush. This is due to the difficulty in distinguishing slush from other surface types in satellite images, such as lakes, snow, and so-called blue ice, making it easily misidentifiable».

Hence the idea to train a machine learning algorithm using video data from NASA’s Landsat 8 satellite, which recorded monthly footage of slush and stagnant lake areas on 57 Antarctic ice shelves between 2013 and 2021.

The goal was to enable the algorithm to distinguish slush from non-slush areas and pinpoint its exact locations on the ice shelves. The ultimate aim is to use this machine learning model to quickly identify slush across all Antarctic ice.

Using their AI model, the Cambridge researchers first discovered, from satellite data covering 2013-2021, that in the peak of the austral summer (January), 57% of all meltwater on Antarctic ice shelves was contained in slush, with the remaining 43% in stagnant lakes. This unexpected proportion confirms the authors’ initial hypotheses. The issue, they highlight, is that «all these slush areas have never been mapped on a large scale across all major Antarctic ice shelves, meaning more than half of all surface meltwater has been overlooked until now».

Furthermore, the machine learning model revealed that slush and lakes contributed to the formation ofmeltwater at a rate 2.8 times higher than predicted by standard climate models, since slush absorbs more solar heat than ice or snow.

Beyond the austral summer, as the climate continues to warm, more surface meltwater forms on the ice shelves, increasing their instability, the risk of collapse, and consequently, sea level rise a chain reaction.

Immagine satellitare di accumuli di acqua stagnante e fanghiglia (in azzurro) sulla piattaforma di ghiaccio Tracy Tremenchus, affacciata sull'Oceano Antartico (dati video rilevati dal satellite Copernicus Sentinel modificati [2018], elaborati da Rebecca Dell).
Accumulations of stagnant water and slush (in blue) on the Tracy Tremenchus ice shelf, facing the Southern Ocean (video data recorded by the Copernicus Sentinel satellite modified [2018], processed by Rebecca Dell).

Glimpses of Futures

The illustrated work highlights a data gap in the processing of satellite images of Antarctic ice and the subsequent predictions regarding the dynamics of their melting. A gap that a mapping system of stagnant water areas and pools of snow mixed with water could potentially fill in the future.

Now, trying to anticipate possible future scenarios, let’s analyse, using the STEPS matrix, the impacts that the evolution of the described method could have from multiple perspectives.

S – SOCIAL: in a future scenario, having a system capable of providing a detailed and comprehensive view of the Antarctic ice melting process, including the phases marking the melting of surface ice during the austral summer, with the formation of lakes and pools on the floating ice platforms surrounding the ice sheet, would result in more accurate calculations of meltwater quantities for all Observers and relevant bodies. This would lead to more precise predictions of sea level rise. This is crucial for timely alerts and for implementing all possible measures in those geographical areas most at risk of flooding over time. NASA indicates that the Greenland and Antarctic ice sheets «store about two-thirds of all freshwater on Earth»: this figure illustrates the magnitude of the problem and the need to estimate it as accurately as possible, using the most advanced technologies.

T – TECHNOLOGICAL: the model developed by the study group is only an initial machine learning model for mapping slush and stagnant water areas on Antarctic platforms. In the future, to estimate even more rigorously the additional melting generated by surface melts of Antarctic ice – driven by both slush and stagnant water – and to make increasingly meticulous predictions about the melting of ice sheets and the stability of the platforms surrounding them, it will be necessary to develop artificial neural networks with deeper architecture. The goal is to discern more clearly, from satellite video data referring to Antarctic glacier melt in the coming years, the areas occupied by snow mixed with water pools.

E – ECONOMIC: a study published in “Scientific Reports” on 18 January 2024 (“Distribution of economic damages due to climate-driven sea-level rise across European regions and sectors“) modelled the potential economic impacts of sea level rise for 271 European regions by 2100, estimating a total cost of €872 billion for EU economies. The United Nations Office for Disaster Risk Reduction reports that floods cost the United States between $179.8 and $496 billion annually, whether from river flooding or coastal water level rise. In this scenario, a system like the one described, mapping stagnant water and slush areas on all Antarctic ice platforms and making predictions as closely aligned with reality as possible about the degree of water level rise in the future, could support activities to prevent events such as floods, erosion, and storm surges, which cause direct and indirect economic damage, affecting the lives of those who experience them firsthand and the economies of the affected countries.

P – POLITICAL: the European Environment Agency (EEA), in an article on extreme sea levels and coastal flooding published on 17 January 2024, confirms that sea levels have risen in numerous locations along European coasts. These increases, which the EEA describes as “ongoing” and unstoppable due to the global average sea level rise, will «amplify the frequency of historical extreme events, exposing most locations to critical conditions with even just a 10 cm rise in sea level». Reading these estimates, one wonders if they derive from climate models that also consider meltwater data and its deposits, aiming for more reliable predictions of sea level rise. It is noteworthy that the ML model developed by the authors, for the period between 2013 and 2021, helped detect a meltwater quantity 2.8 times greater than standard climate models predicted.

S – SUSTAINABILITY: the most severe consequence of Antarctic glacier melting is sea level rise, its core issue, around which a series of indirect effects revolve, impacting the planet’s environment and ecosystem. An artificial intelligence system that, in the future, helps us “measure” and quantify the phenomenon (resolving it is now impossible), based on data that captures the situation in its entirety, including the phases of surface ice melting during the austral summer and the formation of lakes and slush on floating platforms, would itself become “part” of the preventive action, through increasingly prompt and timely alerts and the implementation of physical measures to mitigate the risk of flooding and its effects.

Written by: