Project Description

Arctic warming has led to a dramatic reduction in sea ice coverage. As a result, marine traffic is increasing across the Arctic region and expanding the geographical reach of the Arctic blue economy. Increased marine traffic has the potential to impact marine mammal habitats, increase coastal erosion, alter sea ice growth and decay, and increase underwater noise amongst others. To ensure the predicted increase in marine traffic is managed sustainably, we need to predict how sea ice coverage will change into the future at the local scale of ports and coastal inlets which act as hubs for blue economic activity in the Arctic. This interdisciplinary PhD project will develop bespoke AI tools to predict future sea ice change across the Arctic and plan a sustainable future for marine shipping. The specific aims are as follows:

  • Develop an innovative multi-task deep learning model to forecast sea ice and marine shipping changes.
  • Use the model to understand drivers of seasonal sea ice growth and decay and its variability into the future.
  • Predict shipping traffic to inform a marine management plan that will help to deliver a sustainable future for the Arctic.

Building on recent advances in sea ice forecasting (Andersson et al., 2021), this project will develop bespoke data assimilation tools using deep learning (e.g. Convolutional Neural Networks) to combine multi-domain, cross-discipline data of sea ice (extent, concentration, drift) and ship movement. New high resolution data products (e.g. sea ice coverage) will be extracted from satellite image analysis to optimise the deep learning model at a local to regional scale. A deep learning transformer-based model in a probabilistic framework will then be developed which will be used to forecast sea ice conditions at hourly, daily, weekly, monthly and interannual timescales, superseding existing models such as IceNet which typically performs poorly at these timescales (Andersson et al., 2021). Finally, a route optimisation algorithm will be incorporated into the model (Wu et al., 2022) using Automatic identification System (AIS) data to understand existing patterns of ship traffic under different sea ice conditions.

The deep learning forecasting model will be used to assess future patterns of sea ice variability (growth/decay, drift and coverage) and the drivers of these changes, pushing forward our understanding of sea ice physics using data-driven techniques. The model will then be used to predict its impact on ship movement at key Arctic marine localities such as the Greenland Sea, Baffin Bay, and the Northwest Passage in the Canadian Arctic. The new understanding gained from this will be used to develop a marine management plan for the chosen localities, focusing on the potential implementation of sustainable policies such as: management schemes (vessel slowdown, voluntary avoidance), management tools (spatial vessel, monitoring, and outreach), speed restrictions, voluntary exclusion zones and voluntary speed reduction zones (McWhinnie et al., 2018). This will provide a new, data-driven perspective on Arctic marine management in the context of rapid climate change.

Essential Candidate Background:

  • Applications are encouraged from applicants with a degree (minimum 2:1) in: physical geography, environmental science, remote sensing, physics, computer science, marine science, GIS, and related disciplines. If your educational background differs from this, you are encouraged to contact the supervisor(s) beforehand.
  • Knowledge in one or more of the following would be beneficial: deep learning (e.g. CNNs), data science and statistics, satellite remote sensing processing/analysis, sea ice physics, marine transport geographies, sustainable shipping practices, and/or Arctic environmental change.

Desirable Candidate Background:

  • Some knowledge/experience in applying deep learning/AI algorithms to data (e.g. using TensorFlow, Sci-Kit Learn).
  • Programming experience (e.g. Python, Matlab) as well as knowledge and experience working with complex data sets (e.g. from remote sensing).
  • Demonstrable experience working in an interdisciplinary team is considered advantageous.


Photo by Hubert Neufeld on Unsplash


Primary Supervisor:

Dr William David Harcourt – University of Aberdeen, School of Geosciences / Interdisciplinary Centre for Data & Artificial Intelligence
Academic profile:

Secondary Supervisor:

Dr Georgios Leontidis – University of Aberdeen, School of Natural and Computing Sciences / Interdisciplinary Centre for Data and AI

Academic profile:


Brice Rea

Additional Supervisor:

Profile: Brice Rea
Institution: University of Aberdeen
Department/School: School of Geosciences

Matteo Spagnolo

Additional Supervisor:

Profile: Matteo Spagnolo
Institution: University of Aberdeen
Department/School: School of Geosciences

Additional Supervisor:

Lauren McWhinnie – Heriot-Watt University



  • Andersson, T.R., Hosking, J.S., Pérez-Ortiz, M., Paige, B., Elliot, A., Russell, C., Law, S., Jones, D.C., Wilkinson, J., Phillips, T., Byrne, J., Tietsche, S., Sarojini, B.B., Blanchard-Wrigglesworth, E., Aksenov, Y., Downie, R., Shuckburgh, E. (2021). Seasonal Arctic sea ice forecasting with probabilistic deep learning, Nat. Comms., 12, 5124.
  • McWhinnie, L.H., Halliday, W.D., Insley, S.J., Hilliard, C., Canessa, R.R. (2018). Vessel traffic in the Canadian Arctic: Management solutions for minimizing impacts on whales in a changing northern region, Ocean Coast. Manag., 160, 1-7.
  • Wu, A., Che, T., Li, X., Zhu, X. (2022). Routeview: An intelligent route planning system for ships sailing through Arctic ice zones based on big Earth data, Int. J. Digit. Earth, 15, 1588-1613.


  • earth-systems
  • environmental-management

View All Projects