EGU2020-15481, updated on 12 Jun 2020
https://doi.org/10.5194/egusphere-egu2020-15481
EGU General Assembly 2020
© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

Deep learning for monthly Arctic sea ice concentration prediction

Tom Andersson1, Fruzsina Agocs1,2,3, Scott Hosking1, María Pérez-Ortiz4, Brooks Paige4,5, Chris Russell5,6, Andrew Elliott5, Stephen Law5,7, Jeremy Wilkinson1, Yevgeny Askenov8, David Schroeder9, Will Tebbutt10, Anita Faul1, and Emily Shuckburgh10
Tom Andersson et al.
  • 1British Antarctic Survey, Cambridge, UK (tomand@bas.ac.uk)
  • 2Astrophysics Group, Cavendish Laboratory, Cambridge, UK
  • 3Kavli Institute for Cosmology, Cambridge, UK
  • 4UCL, Centre for Artificial Intelligence, London, UK
  • 5Alan Turing Institute, London, UK
  • 6University of Surrey, Guildford, UK
  • 7UCL, Department of Geography, London, UK
  • 8National Oceanography Centre, Southampton, UK
  • 9University of Reading, Meteorology Department, Reading, UK
  • 10University of Cambridge, Cambridge, UK

Over recent decades, the Arctic has warmed faster than any region on Earth. The rapid decline in Arctic sea ice extent (SIE) is often highlighted as a key indicator of anthropogenic climate change. Changes in sea ice disrupt Arctic wildlife and indigenous communities, and influence weather patterns as far as the mid-latitudes. Furthermore, melting sea ice attenuates the albedo effect by replacing the white, reflective ice with dark, heat-absorbing melt ponds and open sea, increasing the Sun’s radiative heat input to the Arctic and amplifying global warming through a positive feedback loop. Thus, the reliable prediction of sea ice under a changing climate is of both regional and global importance. However, Arctic sea ice presents severe modelling challenges due to its complex coupled interactions with the ocean and atmosphere, leading to high levels of uncertainty in numerical sea ice forecasts.

Deep learning (a subset of machine learning) is a family of algorithms that use multiple nonlinear processing layers to extract increasingly high-level features from raw input data. Recent advances in deep learning techniques have enabled widespread success in diverse areas where significant volumes of data are available, such as image recognition, genetics, and online recommendation systems. Despite this success, and the presence of large climate datasets, applications of deep learning in climate science have been scarce until recent years. For example, few studies have posed the prediction of Arctic sea ice in a deep learning framework. We investigate the potential of a fully data-driven, neural network sea ice prediction system based on satellite observations of the Arctic. In particular, we use inputs of monthly-averaged sea ice concentration (SIC) maps since 1979 from the National Snow and Ice Data Centre, as well as climatological variables (such as surface pressure and temperature) from the European Centre for Medium-Range Weather Forecasts reanalysis (ERA5) dataset. Past deep learning-based Arctic sea ice prediction systems tend to overestimate sea ice in recent years - we investigate the potential to learn the non-stationarity induced by climate change with the inclusion of multi-decade global warming indicators (such as average Arctic air temperature). We train the networks to predict SIC maps one month into the future, evaluating network prediction uncertainty by ensembling independent networks with different random weight initialisations. Our model accounts for seasonal variations in the drivers of sea ice by controlling for the month of the year being predicted. We benchmark our prediction system against persistence, linear extrapolation and autoregressive models, as well as September minimum SIE predictions from submissions to the Sea Ice Prediction Network's Sea Ice Outlook. Performance is evaluated quantitatively using the root mean square error and qualitatively by analysing maps of prediction error and uncertainty.

How to cite: Andersson, T., Agocs, F., Hosking, S., Pérez-Ortiz, M., Paige, B., Russell, C., Elliott, A., Law, S., Wilkinson, J., Askenov, Y., Schroeder, D., Tebbutt, W., Faul, A., and Shuckburgh, E.: Deep learning for monthly Arctic sea ice concentration prediction, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-15481, https://doi.org/10.5194/egusphere-egu2020-15481, 2020

Comments on the presentation

AC: Author Comment | CC: Community Comment | Report abuse

Presentation version 5 – uploaded on 06 May 2020 , no comments
Affiliation correction
Presentation version 4 – uploaded on 06 May 2020
Affiliation correction
  • AC1: Response for Matti Kämäräinen, Tom Andersson, 06 May 2020

    Hi Matti, I don't think I can reply to you directly now that I have updated the poster with some minor changes. Many thanks for your comment and interesting ideas - I hadn't thought to include reanalysis variables from the stratosphere. I would be concerned about how well constrained the stratospheric values are to ground truth, because inputting a biased reanalysis variable can hinder the network's ability to learn a good input-output mapping, but it is definitely worth looking into. At the moment we only input patches centred on the grid cell for which a forecast is made. We do include a coarse 450 km resolution input for each variable, which increases the domain size to >2000km around the output grid cell. However, it is certainly possible to input data from a fixed remote location as well (such as SSTs in the North Atlantic), which could possibly improve predictive skill.

Presentation version 3 – uploaded on 06 May 2020 , no comments
Minor update
Presentation version 2 – uploaded on 06 May 2020 , no comments
Minor formatting changes
Presentation version 1 – uploaded on 06 May 2020
  • CC1: Comment on EGU2020-15481, Matti Kämäräinen, 06 May 2020

    Hi Tom. Nice and interesting work. If you want to improve the prediction accuracy, how about using stratospheric predictors and/or SSTs from Northern Hemisphere? I guess atmospheric forcing plays an important role in the monthly time scale, and increasing the predictor domain size to include Tropics could help, perhaps.