Observed SIC for certain regions can vary substantially from one year to another. (a) SIC observed for September 2007, when the extent of the sea ice was lower than usual. (b) SIC observed for September 2013, when the extent of the sea ice was greater than usual. (c) Area where the observed SIC differed the most between September 2007 and September 2013 is highlighted (white), used as the ROI for gradient attribution.

Image:

Harald L. Joakimsen (UiT)

Observed SIC for certain regions can vary substantially from one year to another. (a) SIC observed for September 2007, when the extent of the sea ice was lower than usual. (b) SIC observed for September 2013, when the extent of the sea ice was greater than usual. (c) Area where the observed SIC differed the most between September 2007 and September 2013 is highlighted (white), used as the ROI for gradient attribution.

Collaborative paper with The Alan Turing Institute and the British Antarctic Survey accepted in IEEE Geoscience and Remote Sensing Letters

We are thrilled to announce that the paper "Interrogating Sea Ice Predictability With Gradients", a collaborative paper between Visual Intelligence, The Alan Turing Institute and the British Antarctic Survey, has been accepted in IEEE Geoscience and Remote Sensing Letters.

Go to event View publication

Collaborative paper with The Alan Turing Institute and the British Antarctic Survey accepted in IEEE Geoscience and Remote Sensing Letters

We are thrilled to announce that the paper "Interrogating Sea Ice Predictability With Gradients", a collaborative paper between Visual Intelligence, The Alan Turing Institute (AT) and the British Antarctic Survey (BAS), was accepted in the journal IEEE Geoscience and Remote Sensing Letters on February 14th 2024.

The paper focuses on interrogating the effect of the IceNet's, a state-of-the-art ice prediction model, input feature with a gradient-based analysis, which takes advantage of the developments within the deep learning literature to open the so-called "black box".

The authors' analysis focuses on the unusually large sea ice extent event in September 2013 and indicates that IceNet places a strong emphasis on previous observations of SIC, linear trends, and seasonal components when making predictions. They further identify which input features are most influential for the prediction and also which spatial location these measurements are particularly influential.

The authors of this particular work from the Visual Intelligence side are post doc. Luigi Luppino (UiT), PhD students Harald Lykke Joakimsen and Iver Martinsen (UiT), and Robert Jenssen (UiT). From the UK side, the authors are Scott Hosking (BAS/AT) and Andrew McDonald (BAS).

You may read the paper via IEEE Explore.

Abstract

Predicting sea ice concentration (SIC) is an important task in climate analysis. The recently proposed deep learning system IceNet is the state-of-the-art sea ice prediction model. IceNet takes high-dimensional climate simulations and observational data as input features and forecasts SIC for the next 6 months over a spatial grid over the northern hemisphere. The model has proven to be particularly good at predicting extreme sea ice events compared with previous dynamical models, but lacks interpretability. In the original IceNet paper, a permute-and-predict approach was taken for assessing feature importance. However, this approach is not capable of revealing whether a feature contributes positively or negatively to the final prediction, nor can it reveal the importance of features over the spatial grid of predictions. In this letter, we take steps to instead interrogate the effect of the IceNet input feature with a gradient-based analysis, taking advantage of developments within the deep learning literature to open the so-called black box. Our analysis focuses on the unusually large sea ice extent event in September 2013 and indicates that IceNet places a strong emphasis on previous observations of SIC, linear trends, and seasonal components when making predictions. In our analysis, we identify which input features are most influential for the prediction and also which spatial location these measurements are particularly influential.

Latest news

Meet Veronica and Qing, our new Associate Professors

February 19, 2026

We happily welcome Veronica Lachi and Qing Liu as new Associate Professors at SFI Visual Intelligence.

About this article

Collaborative paper with The Alan Turing Institute and the British Antarctic Survey accepted in IEEE Geoscience and Remote Sensing Letters

Collaborative paper with The Alan Turing Institute and the British Antarctic Survey accepted in IEEE Geoscience and Remote Sensing Letters

Abstract

About this article

Latest news

Meet Veronica and Qing, our new Associate Professors

My Research Stay at Visual Intelligence: Artur Radzivil

Successful PhD defense by Marius Aasan

Visual Intelligence at TEKDagen 2026

Centre Director Robert Jenssen meets Norwegian Minister of Research and Higher Education, Sigrun Aasland

My Researcher Stay at Visual Intelligence: Zheming Xu

NRK.no: Lara (16) startet med programmering som niåring

Visual Intelligence represented at Arctic Frontiers 2026

Petter Tømmeraas visits Visual Intelligence in Tromsø

Meet Keyne and Boye, our newest PhD Research Fellows

Successful PhD defense by Børge Solli Andreassen

NLDL 2026 was a great success!

Fruitful mentoring session with Professor Mihaela van der Schaar

NLDL 2026 featured on NRK; Norway's national public broadcasting and radio corporation

uit.no Toppforskere på kunstig intelligens samles i Tromsø

unn.no: Forskningsmillioner til pasientrettet KI-satsing

Meet Zijun, our newest Doctoral Research Fellow

Happy Holidays from SFI Visual Intelligence!

Visual Intelligence represented at EurIPS 2025

Anders Waldeland receives the Digital Trailblazer Award 2025