Sea ice leads are narrow fractures formed as a result of shear and divergence in the ice. They play an important role in the energy fluxes between the surface and the atmosphere due to the difference in surface albedo and temperature between ice/snow and open water. In some circumstances, the turbulent heat flux over sea ice leads can be up to two orders of magnitude greater than that over sea ice.
Therefore, representing sea ice leads in weather and climate models is crucial to correctly simulate the energy exchange between the polar oceans and the atmosphere. In addition, it is much easier for icebreakers to navigate through sea ice leads than through packed ice. This means that information about leads can be valuable for ship crews when planning their itineraries.
Observations of sea ice leads are available from satellites. Still, most sea ice forecasting systems do not predict sea ice leads, even though Arctic sea ice typically drifts several kilometres per day. In this project we propose to develop a method to predict sea ice leads 24 hours in advance using machine learning techniques (probably convolutional neural networks).
The main dataset for this project (for ground truth and initial conditions) will be medium-resolution (about 1 km) sea ice lead observations from satellite sensors such as MODIS and VIIRS. In addition, other satellite observations of variables such as cloud cover, sea ice drift and concentration will be combined with weather forecasts to provide sufficient information to predict the evolution of sea ice leads.
We are looking for a motivated student with an interest in polar regions, sea ice, meteorology, machine learning and remote sensing.
Credit: https://cimss.ssec.wisc.edu/satellite-blog/archives/10229 | Click here for a bigger version.