Artificial intelligence (AI) has been used to aid in various scientific developments recently. Now scientists have employed AI to investigate the role of slush—water-soaked snow—on Antarctic ice shelves during the summer months.
This study aimed to better comprehend the impact of climate change on Antarctic ice shelves, particularly focusing on the role of slush in contrast to conventional surface meltwater.
According to a statement by researchers, slush comprises 57 percent of all meltwater on Antarctic ice shelves during peak summer, significantly more than previously recognized, influencing ice shelf stability and contributing to sea level rise.
AI model detected slush ‘with greater ease’
Dr Rebecca Dell, lead author from Cambridge’s Scott Polar Research Institute (SPRI) told Interesting Engineering that until now, slush has been very hard to detect from space using satellites.
“When traditional meltwater mapping methods are used it can be confused for cloud shadow, snow, and ponded water. This means we haven’t been able to map it on a large scale, and if we can’t map it, it is very difficult to understand and track!”
However, she added that AI was employed to train a classifier that could use many wavelengths of the electromagnetic spectrum.
“Most traditional meltwater mapping methods use the red and the blue bands only, but our AI approach used much more information. This meant that slush was detected with greater ease,” Dell told IE .
The classifier has an accuracy rate of 84 percent for ponded water and 82 percent for slush. This was calculated in a previous study , Dell said.
“This accuracy reflects the subjective nature of mapping slush itself, it is very hard to distinguish when slush becomes ponded water and vice versa.”
She said that the team’s previous study showed that expert scientists disagreed with one another on the classification of slush and ponded water as much as they did the computer. However, fieldwork data is required for validation.
43% Antarctic ice shelves’ slush in meltwater lakes
The statement noted that optical data from NASA’s Landsat 8 satellite was utilized by the Cambridge researchers, working in collaboration with researchers from the University of Colorado Boulder and the Delft University of Technology.
They trained a machine learning model to obtain monthly records of slush and meltwater lakes across 57 Antarctic ice shelves between 2013 and 2021.
The computer model determined that in the peak of the Antarctic summer in January, over half—57 percent of all meltwater on Antarctica’s ice shelves is held in slush. While the remaining 43 percent remained in meltwater lakes.
Dell explained to IE that slush is effectively saturated snow and firn (partially compacted snow).
“When an ice shelf melts, if there is pore space in the snow or firn, the meltwater will fill these spaces first. Once the snow/firn is fully saturated, any further meltwater has nowhere to go, and it will likely pond or form surface streams and rivers on the ice shelf.”
Other regional climate models usually do not account for the albedo effect—how much light a surface reflects in slush and ponded water.
The author further explained that since both slush and ponded water are less white, they absorb more of the sun’s energy than snow and ice, therefore resulting in elevated melting.
“Until we ran this study, we simply had no idea as to how much slush existed on Antarctica’s ice shelves!” Dell told IE .
“For it to account for over half of the surface meltwater on Antarctica’s ice shelves, but to have never been fully mapped across ice shelves on the continent before, was really surprising.”
The study was published in the journal Nature Geoscience .
Most Popular
Comments / 0