Predicting hidden storms in the Arctic
Norwegian Meteorological Institute experts piece together satellite and model data to predict fast-developing polar lows more accurately
Snowstorms may be the last thing on most people’s minds at the end of a hot European summer. Yet for Dr Roger Randriamampianina, a research professor at the Norwegian Meteorological Institute, preparing for winter blizzards is a year-round task.
By mid-autumn, the risk of polar lows over the Norwegian and Barents Seas is already climbing – short-lived but severe storms that can unleash hurricane-force winds, white-out blizzards, and trigger avalanches, and threaten lives and livelihoods.
Polar lows form when icy Arctic air sweeps across relatively warm seas, drawing up heat and moisture to fuel the storm. From above they can resemble miniature tropical cyclones, sometimes with a distinctive eye at the centre.
Yet it was only with the advent of modern weather satellites that they were recognised as a distinct phenomenon. Their remoteness, tendency to form in the long dark months of winter, and rapid development still make them some of the hardest storms to predict.
Though typically short-lived and less than a thousand kilometres across, polar lows can pose serious risks to coastal communities and transport. They can also be deadly at sea. In April 1952, a sudden and violent storm – now thought to have been a polar low – struck vessels north of Iceland, destroying five ships and claiming 79 lives. Forecasters had underestimated the winds, and a later warning never reached the crews.
“Polar lows are among the most hazardous weather systems we face in Norway,” says Randriamampianina, who also spent part of his career at the Hungarian Meteorological Service. “Forecasting them relies on numerical weather prediction, which combines computer models with streams of observations to create the best possible picture of the atmosphere.
“In the Arctic, where conventional measurements such as radiosonde and aircraft reports are scarce, polar-orbiting satellites like EUMETSAT’s Metop series are indispensable. Their successor, Metop Second Generation (Metop-SG), will build on that role with more advanced instruments and wider coverage, providing better observations of temperature, humidity and clouds – the kind of input we need to capture current conditions more reliably in our models.”
A Cloud RGB image from the Advanced Very High Resolution Radiometer on Metop-B showing a mature polar low over the Barents Sea on 26 April 2022. Image: EUMETSAT
A Cloud RGB image from the Advanced Very High Resolution Radiometer on Metop-B showing a mature polar low over the Barents Sea on 26 April 2022. Image: EUMETSAT
A combination of data from Metop’s Advanced Scatterometer and infrared images from the satellite’s Advanced Very High Resolution Radiometer enables meteorologists to detail storm force winds at the heart of polar lows, shown to the southwest in red over Svalbard’s Bear Island. Image: MET Norway/EUMETSAT/NOAA
A combination of data from Metop’s Advanced Scatterometer and infrared images from the satellite’s Advanced Very High Resolution Radiometer enables meteorologists to detail storm force winds at the heart of polar lows, shown to the southwest in red over Svalbard’s Bear Island. Image: MET Norway/EUMETSAT/NOAA
Metop Second Generation (Metop-SG) Image: EUMETSAT
Metop Second Generation (Metop-SG) Image: EUMETSAT
Timing and accuracy
Even when weather data are available, the challenge of merging observations from different sources, formats and wavelengths requires exceptional skill.
Randriamampianina grew up in Madagascar, where he once dreamed of becoming a ship’s captain or a pilot. However, his interest in maximising the impact of meteorological observations through data assimilation stretches back to his doctoral studies in the former Soviet Union, when he worked on constant-pressure balloon experiments tracked and simulated across the Northern Hemisphere – early tests of how new measurements could strengthen numerical weather prediction models through enhanced dropsonde observations.
Later, while collaborating with scientists at Météo-France, among other challenging tasks, he contributed to the ARPEGE global model’s radiation scheme, which simulates how energy from the Sun and Earth moves through the atmosphere. This work deepened his understanding of how radiance data from satellites could be represented and applied in forecasts.
Then, in 2011, Randriamampianina’s team published one of the first high-latitude demonstrations of how satellite radiance data could improve storm prediction. Working with international collaborators in the IPY-THORPEX project, they showed that assimilating observations from the Infrared Atmospheric Sounding Interferometer (IASI) on EUMETSAT’s Metop satellites into regional models substantially improved the prediction of polar lows.
In case studies, the results were striking. For a mature storm northwest of Lofoten in March 2008, the model initially overstated the wind speeds, but combining IASI radiances with dropsonde data – small sensor packages released from aircraft that record pressure, temperature and humidity as they fall – brought the storm’s circulation and pressure field into clearer focus, and the wind forecast closer to reality. For another, fast-developing storm later that month, the added satellite data improved both the predicted track and the intensity of the system
“It was among the first operational demonstrations of its kind in a high-latitude system, and was a fantastic moment for us,” Randriamampianina says. “The work deepened understanding of how radiance data from satellites could be represented and applied in forecasts. What makes data assimilation approaches so exciting is that we are bringing together data not just from satellites but many in situ and conventional methods as well.
“Together, these studies showed that data could be applied in new ways to provide earlier and more reliable predictions of where a polar low would hit, how strong it would be, and what communities should be prepared for.”
Towards better weather models
For Randriamampianina, this work marked a turning point. Until then, many centres relied on applying bias corrections or using pre-processed satellite products. His team was among the first to show that radiance data from IASI could be assimilated directly into regional models at high latitudes and used in operational forecasts.
That shift laid the foundation for much of his later work, focused on refining the initial conditions that feed weather models, often in partnership with EUMETSAT, its Satellite Application Facilities, and other expert groups.
“What I’ve learned is that the whole chain matters – from the observations we start with, to the forecast itself, and finally to how the warnings are delivered,” he says. “That is why it is so important to keep a close connection with both the providers of these data and the end users, such as weather forecasters.
“On one hand, we help forecasters to understand how and why different observations are used in the models, or why some are not included, so they can see what’s happening when things go wrong or when they work really well.
“On the other hand, forecasters take these data products and ensure that the science translates into information that is meaningful for the public – like why a breezy 5°C day can seem more like freezing, communicated as the ‘feels like’ temperature alongside the actual reading.”
Precipitation forecast from 7 August 2023. Image: Senorge.no
Precipitation forecast from 7 August 2023. Image: Senorge.no
New tools through international cooperation
With the launch of the EUMETSAT Polar System – Second Generation (EPS-SG), equipped with advanced infrared, microwave and imaging instruments, and the planned EPS-Sterna constellation of microsatellites carrying microwave sounders, new opportunities are emerging to collect the detailed temperature and humidity profiles needed to predict storms more accurately.
The challenge then falls to specialists like Randriamampianina to make sure these observations translate into better performance in weather models.
“When we first integrated IASI data into weather models in 2011, they ran at 10-kilometre resolution,” Randriamampianina says. “Now we are down to 2.5 or even 1.3 kilometres, and with Destination Earth – a European initiative to build high-quality weather and climate simulations of the planet called digital twins – we are pushing towards 500 metres.
“At that scale, you cannot treat satellite data as single points; you have to account for the full footprint – the actual area observed – to make them fit the model better. Together with international partners, we are also expanding ensemble methods, running many slightly different forecasts in parallel to make fuller use of observations. And through Copernicus initiatives such as the Arctic regional reanalysis, we are reconstructing past Arctic weather in fine detail.
“International cooperation helps us to use satellite data to their full potential and tackle challenges no single country could solve alone. And as Norway, like the rest of Europe, faces growing risks from extreme weather such as droughts and floods, these initiatives feed not just scientific advances but vital public services.”
Looking back, wherever he worked, the goal for Randriamampianina has stayed the same: turning streams of measurements into better forecasts.
“Every improvement, whether it is tracking a polar low or predicting floods, can help save lives and reduce damage,” he says. “That is what makes this work so rewarding.”
Built with Shorthand