People have been using the Arctic Ocean for over a century, mainly for shipping goods between the Atlantic and the Pacific Oceans. The northern passages come with dangers though – sea ice poses a significant threat to shipping and other marine activities. In the past decades, technological advances have made it safer to cross these waters, but the ocean and sea-ice conditions in the Arctic remain challenging. A recent study led by Anton Korosov at the Nansen Center is contributing to increase safe navigation in the Arctic.

U.S. and Canadian Coast Guard ships in the Arctic. Photo Credit: Patrick Kelley, U.S. Coast Guard. The photo is in the Public Domain (CC0 1.0). Source: "U.S. Geological Survey" on Flickr (click on image to get there)

One of the quite important tools to reduce the risk are Earth observation satellites. They can be used to look at the Earth’s surface with great detail and over massive distances – perfect for the remote Arctic. To produce good forecasts for the upcoming ocean and sea-ice conditions, detailed satellite images are essential. For example, the two satellites in the Sentinel-1 mission of the European Space Agency (ESA) provide these images regularly.

Today, many shipping companies use so-called sea-ice charts provided by national ice services of different countries to plan routes. Those are produced manually by skilled people, and they are based on satellite images. Another way to use satellite images is for the initialisation of sea-ice numerical models to produce automatic forecasts. The forecasts of sea-ice conditions rely on satellite images with the least amount of noise possible. In a photograph, noise is the random irregularity in colour or brightness, and on satellite images we see a similar phenomenon – disturbances in the image.

This noise is problematic both for visual interpretation of satellite images by ice experts, and for automatic algorithms for sea ice detection. As a result, the models may be initialized from inaccurate observations of sea ice, and the resulting forecasts will contain errors. It is possible to remove the noise: The European Space Agency provides an algorithm for data coming from Sentinel-1 to remove it mathematically.

But this algorithm has not been performing optimally, which is why Anton Korosov and Denis Demchev from NERSC, together with colleagues from ESA, ARESYS, and the Korea Polar Research Institute set out to improve it. They wanted to find the best possible way to clean up those satellite images to make sea-ice forecasts more reliable, and their results in a recent study show that they succeeded. Their new algorithm now significantly reduces the noise in images, letting sea-ice models produce more accurate forecasts! ESA has also improved their own noise correction software using the findings from this research – ultimately helping with safer navigation in Arctic waters.

Reference:

Korosov, A, Demchev, D, Miranda, N, Franceschi, N, and J-W Park. Thermal Denoising of Cross-Polarized Sentinel-1 Data in Interferometric and Extra Wide Swath Modes. IEEE Geoscience and Remote Sensing Society. 2021. https://ieeexplore.ieee.org/document/9627699

For those of you who want to hear more about the topic from the first author himself - Check out the latest Bjerkens Climate Podcast episode hosted by NERSC researcher Stephen Outten: How can we predict sea ice?

Sea ice covering the Arctic Ocean north of Svalbard, complicating navigation in the high latitudes. Photo: Espen Storheim/NERSC