At Cambridge Bay in Canada’s Nunavut, researchers punctured the pack ice, pumped seawater from below and let it freeze on the surface. In mid-May, the ice in the treated areas was up to 32 centimeters thicker than in the control areas and, during the spring melt, it was even brighter. It is the most comprehensive field test yet of the idea of refreezing the Arctic. It works on some pieces of ice floe. The Arctic, however, is much larger.
The result comes from a study published in Earth’s Futureauthored by Edward Blanchard-Wrigglesworth of the University of Washington together with researchers from Real Ice, the Center for Climate Repair at the University of Cambridge and other partners. The experiment followed the ice for an entire season, from winter growth to spring melting: an important step for a technique that, until a few years ago, lived mainly within climate models.
They flooded a quarter of a square kilometer of ice floe
The test field measured one kilometer by one kilometre. Inside, the researchers set up three control areas, left undisturbed, and eight experimental areas. In total, 0.25 square kilometers of ice were artificially flooded, pumping seawater taken from beneath the ice floe onto the surface. Some areas were treated once, others twice, at different times in the winter of 2024-2025.
The mechanism takes advantage of a very unspectacular and rather useful feature of ice. The snow accumulated on top of the pack ice acts like a blanket and slows the loss of heat from the ocean. By bringing sea water to the surface during the winter, you put it directly in contact with the freezing air: the snow becomes soaked, new ice forms from above and the conditions that regulate growth at the base also change. Britain’s Advanced Research and Invention Agency, ARIA, compares the principle to that already used to build and strengthen ice roads.
In May, flooded areas were up to 32 centimeters thicker than control areas; those treated twice generally showed the greatest increase. During the melting season, the artificially thickened ice also appeared lighter and melted more slowly. To give a measure of the data, the authors observe that 30 centimeters corresponds roughly to the loss of thickness recorded in Cambridge Bay over half a century.
What the study does not demonstrate is equally important: we still do not know whether the same mechanism can be effectively repeated on hundreds of thousands of square kilometers, with what energy consumption, what ecological consequences and what behavior of the ice in conditions different from those of the Canadian bay.
Another experiment also showed where it can jam
Caution is not decorative. In April 2026 a second study published on Journal of Geophysical Research: Oceans described an independent experiment in Svalbard. There the artificial flooding of about 1,500 square meters produced 26 centimeters of additional ice and delayed the formation of the so-called rotten icethe ice now deteriorated by melting. The treated ice floe, however, disappeared in the summer substantially together with the reference ice floe.
Two tests, therefore, tell of a technique capable of thickening the ice. The duration of the effect depends on location, size of the area, snow, temperatures, salinity and season. Before you imagine a fleet of pumps busy remaking the Northern Hemisphere’s freezer, there’s a lot of physics to sort out.
The underwater robots also pass through Pisa
The centimeters gained on the experimental field are one thing. Repeating the same work on much larger portions of ice floe is a completely different matter, especially if each hole requires people and machinery on the surface. This is why Real Ice is working on autonomous underwater drones capable of moving under the ice, drilling it from below and pumping sea water to the surface. The stated goal is to run them on zero-emission energy and build a modular system that can, one day, work on a much larger scale. For now the tests remain small and the Arctic fleet still belongs to the projects.
It is Andrea Ceccolini, CEO of Real Ice, who places the emphasis on this passage. Preserving sea ice on a significant scale, he explains, requires scalable technologies, automation and systems capable of working reliably in one of the harshest environments on Earth. Real Ice therefore imagines combining autonomous robotics, climate science, knowledge from Inuit communities and monitoring ecological effects. A clarification that is anything but secondary: the organization itself maintains that any possible development will have to proceed gradually, season after season, on the basis of technical, environmental and social results.
And here Italy also comes in. The BioRobotics Institute of the Scuola Superiore Sant’Anna in Pisa is involved in the development of this technology: the project involves underwater robots capable of carrying a pump and a system similar to a snorkel, reaching the bottom of the pack ice and pushing the water upwards. In 2026 the Sant’Anna School also opened a research selection specifically dedicated to an “underwater robot for navigation in the Arctic environment and the protection of sea ice”. In short, before refreezing the Arctic, robots still need to be taught to work under it.
The new test is worth almost £10 million, but it is not a dress rehearsal
Since 2026, Real Ice has also participated in the Re-thickening Arctic Sea Ice project, RASI, led by the University of Cambridge and funded by ARIA. Here a clarification avoids losing a few million along the way: the overall financing of the project is approximately 9.9 million pounds in 42 months; the share approximately attributed to Real Ice is 3.5 million. The trials also involve Arctic Reflections and several university groups.
The new phase remains deliberately small: ARIA talks about experiments of less than one square kilometer, designed to measure thickness, salinity, temperature, reflectivity and possible effects on microbial life. The work at Cambridge Bay was agreed with the local community, elders and the Hunters and Trappers Organisation, with Inuit residents also involved in the research activities. Formal scientific results from the 2026 campaign are expected later: the agency indicated the end of the year as the likely horizon for publications.
The Arctic may become “ice-free,” but 2030 is not an expiration date
Here too we need to remove a bit of haste from the titles. Since 1979, September Arctic sea ice extent, when it reaches its annual minimum, has declined by about 13% per decade; the Cambridge Bay study summarizes the overall loss for the period at around 40%.
Saying that polar ice will “disappear by 2030”, however, brings together different things. The National Snow and Ice Data Center considers an Arctic with less than one million square kilometers of sea ice to be ice-free. A single day below that threshold could occur as early as the late 2020s and 2030s; for an average ice-free September the projections more likely point to the period within mid-century.
The urgency, therefore, remains. The shortcut a little less. ARIA itself writes that artificial ice thickening, even if it turns out to be safe and scalable, could at most buy time: without a rapid cut in emissions, no reasonable amount of water pumped over the ice pack can prevent the Arctic from continuing to warm.
For now we know that taking water from the ocean and freezing it on top of the ice can make it thicker and, at least under some conditions, slow its melting. Thirty-two centimeters over a quarter of a square kilometer experiment is an interesting scientific result. Between that and refreezing the Arctic there is still, literally, an ocean.
