the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
GHGPSE-Net: A method towards spaceborne automated extraction of greenhouse-gas point sources using point-object-detection deep neural network
Abstract. Point sources account for a large portion of anthropogenic greenhouse gas (GHG) emissions. Timely detection, localization, and quantification of these emissions are critical for supporting carbon neutrality efforts. Spaceborne monitoring satellites can provide essential concentration data for identifying point sources. However, existing methods often require human intervention and typically detect plume masks instead of source locations, limiting their utility for regulatory applications. In this study, we present GHGPSE-Net, a deep learning method for greenhouse gas point source extraction. GHGPSE-Net simultaneously performs detection, localization, and quantification of emissions, eliminating the need for traditional segmentation steps. To train and evaluate the model, we construct synthetic datasets using an atmospheric transport model and validate its accuracy against radiosonde profiles and satellite observations. GHGPSE-Net demonstrates desirable performance in the simulation data across detection (F1-score of 0.96), subpixel-level localization and quantification (Pearson's correlation of 0.99, root mean square error of 89.9 tCO2 hr-1), tested on ideal instrument of 2 km × 2 km resolution with retrieval noise of 1.5 parts per million (ppm). The results also demonstrate considerable generalization of the proposed model when tested using two independent datasets. On the identified sources from OCO-3 spaceborne observations, GHGPSE-Net achieves a detection precision of 0.60, localization accuracy of 2.47 km, and a Pearson's R of 0.89 for quantification. The proposed method and datasets provide a valuable foundation for future research towards rapid and automated GHG point source extraction, offering critical data to support swift responses to abnormal emission events.
- Preprint
(7146 KB) - Metadata XML
-
Supplement
(14764 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2025-3631', Anonymous Referee #1, 28 Sep 2025
- AC3: 'Reply on RC1', Yiguo Pang, 20 Oct 2025
-
CEC1: 'No compliance with the policy of the journal', Juan Antonio Añel, 11 Oct 2025
Dear authors,
Unfortunately, after checking your manuscript, it has come to our attention that it does not comply with our "Code and Data Policy".
https://www.geoscientific-model-development.net/policies/code_and_data_policy.htmlFirst of all, for the WRF code you link a GitHub site. However, GitHub is not a suitable repository for scientific publication. GitHub itself instructs authors to use other long-term archival and publishing alternatives.
Also, for many of the data sources you cite webpages that are not repositories. It is necessary that you store all the data that you have used in repositories that we can accept, with a DOI. This includes the CarbonTracker, IGRA, OCO3, etc.
Therefore, the current situation with your manuscript is irregular. Please, publish the WRF code and all the data in one of the appropriate repositories and reply to this comment with the relevant information (link and a permanent identifier for it (e.g. DOI)) as soon as possible, as we can not accept manuscripts in Discussions that do not comply with our policy. Also, you must include a modified 'Code and Data Availability' section containing the information of the new repositories, which you must include in a potentially reviewed manuscript too.
I must note that if you do not fix this problem, we cannot accept your manuscript for publication in our journal.
Juan A. Añel
Geosci. Model Dev. Executive EditorCitation: https://doi.org/10.5194/egusphere-2025-3631-CEC1 -
AC1: 'Reply on CEC1', Yiguo Pang, 11 Oct 2025
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-3631/egusphere-2025-3631-AC1-supplement.pdf
-
CEC2: 'Reply on AC1', Juan Antonio Añel, 11 Oct 2025
Dear authors,
Unfortunately, your reply does not solve the problems pointed out. You must store all the WRF model, not only the code that you have developed or modified. Also, for the data sources, you continue linking webpages that are not valid repositories (e.g., ucar.edu, European Comisssion webpages, etc.).
Therefore, the problems remain and we can not consider your manuscript for publication in the journal.
Juan A. Añel
Geosci. Model Dev. Executive Editor
Citation: https://doi.org/10.5194/egusphere-2025-3631-CEC2 -
AC2: 'Reply on CEC2', Yiguo Pang, 16 Oct 2025
Dear Dr. Añel,
Thank you very much for your comments. I apologize for the delay of reply, as it takes some time to upload the software and data to Zenodo due to network issues.
The WRF-GHG model, including all source codes, configurations, emission files, initial conditions, and boundary conditions has now been uploaded to a publicly available Zenodo repository:
https://zenodo.org/records/17337441
Anyone can download, compile and reproduce the results.All processing scripts have been provided previously in another Zenodo repository:
https://zenodo.org/records/16751293External datasets that are not hosted on Zenodo or other acceptable repositories will be removed from the Data Availability section of the manuscript.
The OCO-3 measurements, IGRA profiles and other necessary external datasets are stored in the Zenodo repository.Please contact me if any further clarification is needed.
Best regards,
Yiguo PangCitation: https://doi.org/10.5194/egusphere-2025-3631-AC2
-
AC2: 'Reply on CEC2', Yiguo Pang, 16 Oct 2025
-
CEC2: 'Reply on AC1', Juan Antonio Añel, 11 Oct 2025
-
AC1: 'Reply on CEC1', Yiguo Pang, 11 Oct 2025
-
RC2: 'Comment on egusphere-2025-3631', Anonymous Referee #2, 14 Oct 2025
This work applies deep learning to satellite snapshots of CO2, performing point source detection, localisation and quantification. As the authors identify, most work in this area focuses on plume identification or source location, without performing quantification directly. The authors present an innovative solution, where a CNN is trained to identify sources as hotspots, which then get detected and quantified using a statistical Gaussian Kernel fitting method. The authors also introduce a greenhouse-specific augmentation method, varying the linear coefficients for each source, that presents a valuable contribution to the area of GHG+deep learning. The authors demonstrate their model on a wide range of observations and instruments, using also independent validation tests for verification. I believe that the work presents a significant contribution to the ML-remote sensing field, with some further clarifications on the method.
Clarifications required:
- The datasets used should be summarised more cohesively, as at the moment it is hard to follow 1) what regions, areas and time periods are covered for each dataset, 2) the number of samples in each dataset 3) whether the test datasets were augmented. The number of samples added in the augmentation needs to be specified.
- The architecture of the model needs to be outlined more clearly, including how it was tuned, the learning rate, loss functions etc.
- It is not clear to me if the resolution of the model was kept constant across the different instruments, or adapted to each instrument.
- Is it possible that the significantly lower skill on the SMARTCARB dataset is due to a different transport model?
- Would other inputs to the CNN, like the wind direction or the location of known sources, improve predictive skill?
- The Gaussian kernel fitting is demonstrated to improve when the kernel size is the same as the instrument pixel size, for the one instrument tested. Do you expect performance to improve for other instruments when the same process is applied? It is not clear what kernel size you applied for evaluation for each instrument.
Typos and minor corrections:
- Figure 5 shows, from the text, two days with large absolute errors. Clarify in the caption/figure labelling that these two days are not representative of the usual error distribution.
- Line 39 (“and The new generation…”)
- Line 151: areas (not area)
- L342
- Supplement S3.1, title: Definition (not defination)
Citation: https://doi.org/10.5194/egusphere-2025-3631-RC2 - AC4: 'Reply on RC2', Yiguo Pang, 20 Oct 2025
Data sets
Source code for "GHGPSE-Net: A method towards spaceborne automated extraction of greenhouse-gas point sources using point-object-detection deep neural network" Yiguo Pang https://zenodo.org/records/16751293
Model code and software
Source code for "GHGPSE-Net: A method towards spaceborne automated extraction of greenhouse-gas point sources using point-object-detection deep neural network" Yiguo Pang https://zenodo.org/records/16751293
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 1,821 | 70 | 28 | 1,919 | 32 | 15 | 15 |
- HTML: 1,821
- PDF: 70
- XML: 28
- Total: 1,919
- Supplement: 32
- BibTeX: 15
- EndNote: 15
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The manuscript "GHGPSE-Net: A method towards spaceborne automated extraction of greenhouse-gas point sources using point-object-detection deep neural network" presents a novel deep learning framework and a large dataset for detecting and quantifying greenhouse gas (GHG) point sources from satellite imagery. The authors are the first to introduce the point object detection approach in this domain, which is very valuable and insightful for the GHG point source monitoring community, as it has the potential to integrate and simplify the processing complexity largely. The authors also demonstrate the feasibility of the model by evaluating it on two datasets, including authentic satellite observations. Though I anticipate more evaluations may be required on the upcoming moderate-resolution carbon monitoring satellites (e.g., CO2M and TanSat-2) to fully explore the potential. This work marks an important step towards automated GHG point source monitoring and has the potential to make a significant contribution to the GHG remote sensing community.
Minor suggestions:
(1-1) The dataset construction process, including WRF-GHG simulation, XCO2 construction and data augmentation, involves multiple scenarios, especially it seems that the model is trained on the synthetic dataset and evaluated using independent datasets. It may be better clarified using a diagram.
(1-2) The authors summarized GHGPSE-Net in Figure 3. However, the overall methodology, including simulation, simulation evaluation, training dataset preparation, and deep learning evaluation, is quite complex and somewhat difficult to follow. The authors may consider summarizing the entire methodology in Section 2.
(1-3) In some related GHG plume detection studies, deep learning models usually require wind as an input. Does GHGPSE-Net not require the 2D wind field as input?
(1-4) According to the result (e.g., Table 3), it seems the "2 km × 2 km" in L10 should be 0.5 km × 0.5km.
Technical comments:
(2-1) Typo in L59 and L137.
(2-2) It should be "mean squared error" instead of "mean square error" in L10, L208, and L223.