AI for Earth Observation

What AI for Earth observation does 

The project develops artificial intelligence (AI) tools to automatically detect environmental features from satellite imagery. These include icebergs, sea ice, ice floes, and wildlife. 

The team uses multimodal machine learning approaches, trained on large satellite datasets. Sources include Synthetic Aperture Radar (SAR), multispectral imagery from Copernicus, MODIS, and Planet, as well as Worldview 3 and other commercial products. 

Why this matters 

Satellite imagery is now available at higher resolution and coverage than ever before. AI techniques allow rapid and accurate identification of key environmental features at scale. 

This improves iceberg tracking, monitoring of sea ice, and navigation support for research vessels. It also strengthens climate records and supports safe and efficient polar operations. 

How the project works 

AI for Earth observation combines machine learning with a range of satellite data: 

• Supervised and semi-supervised learning: models are trained to detect environmental features with high positional accuracy. 
• Multimodal analysis: techniques integrate SAR and multispectral imagery to improve detection reliability. 
• Applications: outputs support in-ice navigation of the RRS Sir David Attenborough, iceberg tracking, and validation of sea ice concentration datasets. 

Science objectives 

The project aims to: 

• develop AI tools to detect icebergs, ice floes, sea ice, and wildlife from space 
• improve the accuracy and speed of satellite-based environmental monitoring 
• support navigation and operational planning for polar research vessels 
• strengthen climate datasets with validated observations 
• advance AI and machine learning methods for environmental science 

Who is involved 

The project is led by British Antarctic Survey scientists in collaboration with national and international partners.

Key collaborators include the Alan Turing Institute, European Space Agency, the US National Snow and Ice Data Center, and the Norwegian Ice Service. 

Science objectives 

The project aims to: 

• develop AI tools to detect icebergs, ice floes, sea ice, and wildlife from space 
• improve the accuracy and speed of satellite-based environmental monitoring 
• support navigation and operational planning for polar research vessels 
• strengthen climate datasets with validated observations 
• advance AI and machine learning methods for environmental science 

Recent publications:

Rogers, M.S., Fox, M., Fleming, A., van Zeeland, L., Wilkinson, J. and Hosking, J.S., 2024. Sea ice detection using concurrent multispectral and synthetic aperture radar imagery. Remote Sensing of Environment, 305, p.114073. https://doi.org/10.1016/j.rse.2024.114073 

Evans, B., Faul, A., Fleming, A., Vaughan, D.G. and Hosking, J.S., 2023. Unsupervised machine learning detection of iceberg populations within sea ice from dual-polarisation SAR imagery. Remote Sensing of Environment, 297, p.113780. https://doi.org/10.1016/j.rse.2023.113780