Automated greenhouse gas plume detection from satellite data using an unsupervised clustering algorithm
Ervelä, Elias (2024-03-05)
Automated greenhouse gas plume detection from satellite data using an unsupervised clustering algorithm
(05.03.2024)
Julkaisu on tekijänoikeussäännösten alainen. Teosta voi lukea ja tulostaa henkilökohtaista käyttöä varten. Käyttö kaupallisiin tarkoituksiin on kielletty.
avoin
Julkaisun pysyvä osoite on:
https://urn.fi/URN:NBN:fi-fe2024031311115
https://urn.fi/URN:NBN:fi-fe2024031311115
Tiivistelmä
A crucial part of tackling the problem of climate change is the monitoring of human-caused greenhouse gas emissions. To reach a global scale, greenhouse gas measuring satellites appear to be the best solution. The massive amounts of data produced by the satellites has increased the need for automated, efficient tools to extract knowledge from the data. Emissions from point sources, such as power plants, can produce distinct plumes that are visible from satellite data. Automated plume detection is key to identify and monitor the largest sources of human-caused greenhouse gas emissions.
This thesis presents a comprehensive literature review of existing plume detection methods. Moreover, a new unsupervised plume detection method, called SCEA (Spatial Clustering of Elevated-valued Areas), is introduced. Inspired by the DBSCAN algorithm, SCEA is a clustering algorithm that finds distinct high-valued areas in non-gridded data points.
The performance of the SCEA algorithm is evaluated with the simulated satellite data set of SMARTCARB in its ability to find point sources with co-located plumes in different noise scenarios. The SCEA algorithm reached a precision of 0.974, 0.884, and 0.661 in noise-free, low-noise, and high-noise scenarios, respectively. For point sources with annual emissions of 1000 tonnes, the SCEA reached a recall of 0.758, 0.660, and 0.548 for noise-free, low-noise, and high-noise scenarios, respectively.
This thesis presents a comprehensive literature review of existing plume detection methods. Moreover, a new unsupervised plume detection method, called SCEA (Spatial Clustering of Elevated-valued Areas), is introduced. Inspired by the DBSCAN algorithm, SCEA is a clustering algorithm that finds distinct high-valued areas in non-gridded data points.
The performance of the SCEA algorithm is evaluated with the simulated satellite data set of SMARTCARB in its ability to find point sources with co-located plumes in different noise scenarios. The SCEA algorithm reached a precision of 0.974, 0.884, and 0.661 in noise-free, low-noise, and high-noise scenarios, respectively. For point sources with annual emissions of 1000 tonnes, the SCEA reached a recall of 0.758, 0.660, and 0.548 for noise-free, low-noise, and high-noise scenarios, respectively.