Digital Twins of the Polar Regions

Developing AI and digital infrastructure for digital twinning of the polar regions

Start date
1 December, 2021
End date
30 November, 2030

Digital Twinning is next generation technology for data fusion and computer modelling enabling us to rapidly get answers to “what-if” questions. Digital Twins (DTs) are already in operation in industry and involve highly interoperable data pipelines, optimisation, and a mixture of knowledge-informed and data-driven probabilistic machine learning and artificial intelligence (AI). DTs are increasingly becoming a key component for research discovery, education, and aiding discussions at the board-level without having to wait weeks to months for results from traditional computer models. BAS scientists and engineers are developing Digital Twins of the Antarctic and Arctic natural environments, polar research bases, our research ship and automated vehicles. By collaborating with a wide range of international research institutes and businesses our aim is to support international efforts to develop a Digital Twin Earth.

AI for smart conservation

Climate change is an increasingly pervasive threat to global biodiversity. Animal populations in the rapidly changing Arctic are often seen as a litmus test for the response of wildlife to climate change, particularly those whose life histories are inextricably tied to seasonal sea ice, such as polar bears, walrus, and certain caribou populations. As Arctic species navigate their increasingly unpredictable and diminishing sea ice habitat, conservation practitioners require real-time and adaptive decision-relevant tools to limit the loss of wildlife.


Machine Learning for seasonal sea ice forecasting – IceNet

IceNet is a probabilistic, deep learning sea ice forecasting system developed by an international team and led by British Antarctic Survey and The Alan Turing Institute [Andersson et al., 2021]. IceNet has been trained on climate simulations and observational data to forecast the next 6 months of monthly-averaged sea ice concentration maps. IceNet advances the range of accurate sea ice forecasts, outperforming a state-of-the-art dynamical model in seasonal forecasts of summer sea ice, particularly for extreme sea ice events. This step-change in sea ice forecasting ability brings us closer to conservation tools that mitigate risks associated with rapid sea ice loss.

Autonomous Marine Operations Planning (AMOP)

The Autonomous Marine Operations Planning (AMOP) project is developing a suite of Artificial Intelligence (AI) methods that aim to optimise the efficiency of Antarctic field operations, while maximising science delivery. A core focus of this work is providing AI decision support for efficient management of the BAS marine fleet. This includes, but is not limited to, our research ship the RRS Sir David Attenborough (SDA) and our fleet of autonomous marine vehicles, such as Boaty McBoatface. The AMOP project is a key component of our efforts to combat climate change, aiming to minimise the use of fuel when delivering science in the polar regions. It is of paramount importance to meeting our goal ofNet Zero Carbon by 2040.

PolarRoute carbon Saving routing example
PolarRoute Example demonstrating route planning from a ship location to a destination either by the fastest or the carbon saving route. By taking the carbon saving route, which is 14 mins longer than the fastest route, you save 3.82% on carbon. This carbon saving is equivalent to: 1.7 times of a petrol-powered passenger vehicle driven for one year; 19,204 miles driven by an average petrol powered car; or, 0.94 a homes energy usage for one year.


Icesheet model emulation

We are developing a new approach to emulate the output of a process-based ice sheet model trained on a set of pre-computed ice sheet simulation runs. The aim of the emulator is to mimic how the process-based simulator would have responded, massively reducing computational cost. This allows the exploration of a much bigger range of environmental scenarios than would be possible using the simulator alone and, crucially, will enable probabilistic predictions of the ice sheet state, and sea level contribution, to be produced in an instantaneous manner.



Earth observation: automated object detection and tracking

Accurate measurements of sea ice extent, its concentration, and iceberg position are important for safe navigation of ships, maintaining fisheries and understanding ecosystem dynamics. These features have primarily been detected using microwave data; however the coarse, 6.25 km, spatial resolution of this microwave data precludes its use in decision-making tasks. Recent increases in the spatial coverage and temporal resolution of 40 m resolution Synthetic Aperture Radar (SAR) imagery, is opening new opportunities to rapidly detect ice sheet and iceberg position with greater positional accuracy. Our team are developing supervised and unsupervised machine learning techniques to automatically detect sea ice and iceberg position from Sentinel-1 SAR imagery.

SAR Sentinel image





.dl-sublanding { display:none; }