A Caltech Library Service

Towards Accurate and Automated Detection and Quantification of Localized Methane Point Sources on a Global Scale


Jongramrungruang, Siraput (2022) Towards Accurate and Automated Detection and Quantification of Localized Methane Point Sources on a Global Scale. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/ab33-7a98.


Methane (CH4) is the second most important anthropogenic greenhouse gas with a significant impact on radiative forcing, tropospheric air quality, and stratospheric water vapor. Because methane has a much shorter lifetime compared to carbon dioxide (CO2), reduction in methane emission is deemed a key target for climate mitigation strategies in upcoming decades. One crucial step in emission reduction is determining the location and emission rate of localized methane sources. Remote-sensing instruments using absorption spectroscopy have emerged as one promising solution for measuring atmospheric CH4 concentration over large geographical areas. However, the identification and quantification of local point sources based on the observed methane column enhancement distribution has proven challenging due to uncertainties in the knowledge of local wind speed and retrieval errors arising from surface spectral interferences and instrument noise. In this thesis, it is shown how plume morphology based on a 2-D image of methane column enhancement can be used to quantify the source emission rate directly without relying on any ancillary data such as local wind speed measurements. Large eddies simulations (LES) are utilized to create realistic synthetic plume observations under various atmospheric conditions. Using this data, a deep learning model named MethaNet is trained to predict emission rates directly from 2-D methane plume images. The model achieves a level of performance for quantifying methane emission rates that is state-of-the-art for a method that does not rely on wind speed information. Obtaining methane column measurements with low precision error and bias is a key step for separating real plume enhancements from artefacts and enhancing the quantification performance. Here an instrument tradeoff analysis is presented to assess the effect of changing instrument specifications and retrieval parameters. It is shown how the retrieval errors can be mitigated with optimal spectral resolutions and a larger polynomial degree to approximate surface albedo variations in the retrieval process. The results in this thesis contribute towards building an enhanced monitoring system that can measure CH4 enhancement fields and determine methane sources accurately and efficiently at scale.

Item Type:Thesis (Dissertation (Ph.D.))
Subject Keywords:Methane, remote sensing, machine learning, greenhouse gas, methane emissions, gas detection, gas quantification
Degree Grantor:California Institute of Technology
Division:Geological and Planetary Sciences
Major Option:Environmental Science and Engineering
Minor Option:Computer Science
Thesis Availability:Public (worldwide access)
Research Advisor(s):
  • Frankenberg, Christian
Thesis Committee:
  • Wennberg, Paul O. (chair)
  • Frankenberg, Christian
  • Thompson, David R.
  • Ross, Zachary E.
Defense Date:20 July 2021
Funding AgencyGrant Number
Record Number:CaltechTHESIS:08092021-200722718
Persistent URL:
Related URLs:
URLURL TypeDescription adapted for Chapter 2.
Jongramrungruang, Siraput0000-0002-2477-2043
Default Usage Policy:No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:14319
Deposited By: Siraput Jongaramrungruang
Deposited On:25 Aug 2021 16:22
Last Modified:04 Aug 2022 21:17

Thesis Files

[img] PDF - Final Version
See Usage Policy.


Repository Staff Only: item control page