Project for improving ice forecasting with artificial intelligence

Status: Active

The NRC is developing new tools and techniques using machine learning, a form of artificial intelligence (AI), to improve ice forecasting and make shipping safer in Northern waters.

Objectives

Using machine learning and ensemble techniques, more accurately forecast sea ice freeze-up and break-up within a similar-sized area as traditional models, but over a longer period of time

Deliverables

Develop a new set of tools to:
- better forecast sea ice freeze-up and break-up
- accurately predict ice conditions like concentration, lead openings, thicknesses and ridging
- forecast over a longer time period (10 to 30 days), including spatial details

Activities

Develop the new tools using machine learning, which is a type of artificial intelligence (AI) model that uses statistical techniques to give computer systems the ability to "learn" without specifically being programmed to do so.
Identify ways to include spatial details, which current models that forecast over a longer time typically do not.
Overcome challenges of traditional physics-based sea ice forecast models, such as:
- simultaneously capturing physical processes occurring at different scales
- complicated computing requirements as the spatial area of the model increases

Results

A forecasting system that can produce reliable mid-range forecasts of sea ice can be used routinely by the shipping industry and Government of Canada federal fleet operators to support their activities in the North.
This type of system could be used to help ensure safe operations, reduce risks to human life and the environment, improve route planning and scheduling, minimize fuel consumption and optimize costs.