Nebula Public Library

The knowledge bank of ESA’s R&D programmes

STARCOP: Automated and self-improving follow-up verification of detrimental human activity from LEO

Programme
Discovery
Programme Reference
EISI_I-2022-00380
Start Date
End Date
Status
Closed
Country
United Kingdom
STARCOP: Automated and self-improving follow-up verification of detrimental human activity from LEO
Description

‘STARCOP’ is an initiative to use machine learning and multiple satellites with diverse detection capabilities to quickly detect methane leaks and send notifications in near real-time. In this project Trillium Technologies developed machine learning models to automatically detect methane in hyperspectral and multispectral imagery. Their hyperspectral model (HyperSTARCOP) is able to capture more than 90% of plumes in test data while reducing the false positive rate by 39% when compared to state-of-the-art models. In addition, the activity shows for the first time automatic multispectral approaches that are able to capture large plumes in Copernicus Sentinel-2 data and WorldView-3 data.

Executive summary