Researchers turn to artificial intelligence to model how snow cover is shrinking
In a leafy courtyard in the northern Italian city of Bolzano, children chase each other around as daycare workers look on, interns sip cappuccinos, and researchers hustle past on their way to the lab.
In the distance, pine-covered mountains rise in all directions like majestic gatekeepers. The famed Dolomites of the Italian Alps are breathtakingly beautiful, but also stark reminders of how climate change is making snowy peaks more unpredictable.
In July, 11 hikers were killed when record-high temperatures contributed to a massive chunk of the Marmolada Mountain Glacier breaking loose. The shrinkage of glaciers and a decline in snowfall also led to the drying out of the Po, Italy’s longest and most important river for agriculture and hydroelectrical power.
This week, as world leaders prepared to meet in Egypt for the COP27 climate change conference starting Sunday, a UN report warned glaciers around the globe, including the last one in Africa, will be gone by 2050.
Here in Bolzano, researchers with the private clean energy research group Eurac have pieced together a long-range picture of how snow cover around the world has already changed, using modelling and artificial intelligence.
Their study, published in Nature’s Scientific Reports, found that globally, it’s been decreasing over the past 38 years, with four per cent less mountain area covered with snow, and an average of 15 more snow-free days per year.
In the Rockies, the study found the number of days with no snow cover reached as many as 30 at certain times and areas, with a slight increase of snow in tiny micro-climates.
“The warming of the minimum temperature, as well as decreasing in winter precipitation and more [rain] … can make the melting phase faster,” said Claudia Notarnicola, the scientist with the Institute for Earth Observation at Eurac who led the study.
“The strongest effect we see is the anticipation of the melting season, [spring temperatures] coming earlier.”
From aluminum to clean energy development
Eurac’s work happens at a facility called the Nature of Innovation (Noi) Tech Park, which a century ago was the site of Italy’s burgeoning aluminum production, one of the most energy-intensive and polluting industries, launched by fascist dictator Benito Mussolini.
At its peak, the area produced a third of the country’s aluminum, until production petered out due to global competition and ended in the 1980s.
Today, converted factories, along with modern constructions, are part of the expanding hub for environmental innovation and research — housing everything from start-ups and clean energy labs to environmental agencies, a university campus and daycare.
“In this [region of] South Tyrol, nature has always had an important piece in our way of living and doing,” said Wolfram Sparber, head of renewable energy at Eurac, one of the main occupants of Noi. “The idea was to offer a place with a high work value, a nice place to be, a good combination of work-life balance.”
Sparber shows off a lab where scientists spend days in large, fridge-like rooms to test equipment and human response to extreme weather on mountain peaks as high as 9,000 metres.
In another building, solar panel testing is underway, with a technician focused on ferreting out malfunctions to increase efficiency. Eurac is involved in several large-scale European projects to develop high performance solar panels to help revive production in Europe after Chinese manufacturers undercut European manufacturing.
But — unusual for clean energy technology centres — Eurac also carries out climate change research, in a sleek, elevated glass structure nearby.
A global view extended back in time
The recent study is a follow-up to another by Notarnicola published in 2020 that looked at snowfall dating back two decades and showed evidence of a decline of snow cover in 78 per cent of mountain areas around the world.
What’s different this time is the researchers have used artificial intelligence (AI) to explore what was happening with snow cover in high altitudes before consistent satellite data became available in 2000.
The 2022 study used MODIS satellite data available from 2000 on and, employing artificial neural networks, modelled the data back to 1982.
“What Claudia has done here is really innovative,” said Chris Derksen, a research scientist in the Climate Research Division of Environment and Climate Change Canada.
“For climate studies what we really want is as many years as possible — 30 to 40.”
Derksen says mountain studies tend to be regional, with researchers in North America focusing on the Rockies or Sierra Nevada ranges, for instance, or in Switzerland, Austria or Italy, on the Alps.
“From a climate change perspective, the more we can look at the whole hemisphere, it just gives us a more powerful signal of how things are changing,” said Derksen of the need for global studies.
Ground data needed
Still, the use of MODIS satellites in snow study has its limitations, said John Pomeroy, Canada Research Chair in Water Resources and Climate Change at the University of Saskatoon.
He said the satellite data’s low resolution and inability to see through thick forests, thus missing the snow underneath it, can lead to errors. It can also mistake cloud cover for snow.
“I’m not disputing the findings,” said Pomeroy of Notarnicola’s study. “It’s useful to have a global analysis like this — how she tried to fill in the gaps and uncertainties with artificial neural networks was clever.
“But there are also issues with those in that they are trained for the past and are data driven, so can be driven in the wrong direction.”
Pomeroy isn’t against the use of satellite and AI technology combined, but he would like to see other checks involved, such as the on-the-ground study of snow carried out by field stations and mountain research sites, snow surveys and other data sets.
More observational data are becoming available around the world, with the setting up by Pomeroy and others of the Common Observational Period Experiment (COPE), a network of intensive observation sites of high mountain areas around the world.
Brian Menounos, a Canada Research Chair in Glacier Change at the University of Northern British Columbia, agrees direct observational data can only help enrich satellite data refined by complex mathematical models. Especially, he says, with the challenge in dealing with smaller, several-years-long climate trends of drought or increased precipitation, within the larger trend of global warming.
“We have to think about different time scales.… It’s really that decadal variability that’s much more difficult to predict and that will greatly influence availability of water,” he said. “And that’s really what we’ve seen throughout the globe.”