pith. sign in

arxiv: 2605.24273 · v1 · pith:3N43SGZRnew · submitted 2026-05-22 · 💻 cs.CV · physics.ao-ph

Plume Segmentation from MethaneSAT with Cross-Sensor Transfer Learning and Physics-Informed Postprocessing

Manuel P\'erez-Carrasco , Maya Nasr , Zhan Zhang , Apisada Chulakadabba , Javier Roger , Raia Ottenheimer , S\'ebastien Roche , Maryann Sargent
show 11 more authors
Chris Chan Miller Daniel Varon Jack Warren Luis Guanter Kang Sun Jonathan Franklin Jia Chen Cecilia Garraffo Xiong Liu Ritesh Gautam Steven Wofsy
This is my paper

Pith reviewed 2026-06-30 15:16 UTC · model grok-4.3

classification 💻 cs.CV physics.ao-ph
keywords methane plume detectionMask R-CNNtransfer learningMethaneSATMethaneAIRsatellite imageryinstance segmentationphysics-informed postprocessing
0
0 comments X

The pith

Mask R-CNN fine-tuned from airborne data segments methane plumes in MethaneSAT images at 0.98 recall.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper presents a machine learning framework to detect and segment individual methane plumes in column-averaged dry-air mole fraction data from the MethaneSAT satellite. It tackles limited labeled satellite data by testing transfer strategies from the airborne MethaneAIR instrument and from synthetic plumes, then applies physics-informed post-processing to turn raw detections into two usable modes. The best approach uses Mask R-CNN with a ResNet-50 backbone pre-trained on MethaneAIR and fine-tuned on MethaneSAT, outperforming U-Net and other transfer options. This yields a baseline with instance-level precision of 0.60 and recall of 0.98, plus a high-sensitivity mode at precision 0.71 and recall 0.94, and a high-precision mode at precision 0.92 and recall 0.70. The work shows that reported precision numbers are lower bounds because some detections labeled false positives match real methane enhancements missed by the conservative wavelet ground truth.

Core claim

Mask R-CNN with ResNet-50 backbone fine-tuned from MethaneAIR pre-trained weights outperforms U-Net semantic segmentation and other cross-sensor transfer strategies on MethaneSAT data, delivering instance-level precision of 0.60 and recall of 0.98 at the baseline operating point. A physics-informed post-processing pipeline that applies morphological filtering, proximity-based merging, and a distribution-based classifier then converts these detections into a high-sensitivity mode (precision 0.71, recall 0.94) for comprehensive screening and a high-precision mode (precision 0.92, recall 0.70) for confident source attribution. Manual review of false positives against the wavelet-based labels in

What carries the argument

Mask R-CNN instance segmentation model with ResNet-50 backbone, cross-sensor transfer from MethaneAIR pre-trained weights, and physics-informed post-processing pipeline using morphological filtering, proximity merging, and distribution-based classification.

If this is right

  • The high-sensitivity mode supports broad emission screening across varied atmospheric and surface conditions.
  • The high-precision mode enables confident attribution of emissions to specific sources.
  • Transfer learning from airborne to satellite data overcomes scarcity of labeled MethaneSAT examples.
  • Instance segmentation outperforms semantic segmentation for separating individual plumes.
  • Reported precision values represent lower bounds on true detection performance.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same transfer strategy could be tested on other satellite methane sensors that lack large labeled datasets.
  • High-recall detections could feed into atmospheric transport models for improved total emission estimates.
  • Future work might replace conservative wavelet labels with multi-sensor or in-situ validation to raise measured precision.
  • The two-mode output suggests the framework can be tuned differently for screening versus regulatory attribution tasks.

Load-bearing premise

The wavelet-based ground truth labels accurately capture all real methane plumes even though they are known to be conservative and exclude some enhancements.

What would settle it

Independent aircraft or ground-based methane concentration measurements over a set of model-detected plumes that the wavelet labels marked as false positives, to determine how many are actually real enhancements.

Figures

Figures reproduced from arXiv: 2605.24273 by Apisada Chulakadabba, Cecilia Garraffo, Chris Chan Miller, Daniel Varon, Jack Warren, Javier Roger, Jia Chen, Jonathan Franklin, Kang Sun, Luis Guanter, Manuel P\'erez-Carrasco, Maryann Sargent, Maya Nasr, Raia Ottenheimer, Ritesh Gautam, S\'ebastien Roche, Steven Wofsy, Xiong Liu, Zhan Zhang.

Figure 1
Figure 1. Figure 1: Gridded column-averaged dry-air mole fractions of methane (XCH4) from the Permian Basin (a) MethaneAIR scene covering 120 km × 130 km at 10 m grid size and (b) MethaneSAT scene covering 220 km × 335 km at 45 m grid size over an oil and gas production region. Both L3 products are derived using the CO2 proxy method applied to vertical column densities of CH4 and CO2. Enhanced XCH4 values (yellow colors) indi… view at source ↗
Figure 2
Figure 2. Figure 2: Methane plume detection sample patches at the same patch size of 800×800 pixels in the Permian Basin. (a) MethaneAIR provides high-resolution (10 m/pixel) XCH4 retrievals (red contours). (b) MethaneSAT at coarser resolution (45 m/pixel). (c) Synthetic plume at MethaneSAT’s resolution using HRRR model. Note that at the same pixel dimensions, MethaneAIR patches cover approximately 8 km×8 km while MethaneSAT … view at source ↗
Figure 3
Figure 3. Figure 3: Overview of the proposed methane plume detection pipeline. (Step 1) Input XCH [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Mask R-CNN architecture used in this work. The backbone (ResNet-50, ViT, or MAE) extracts feature maps from the input XCH4 patch. A Region Proposal Network (RPN) generates candidate plume regions, which are resam￾pled to a fixed spatial size via RoIAlign and passed to three parallel heads: a classification head that predicts plume vs. background, a bounding box head that refines region coordi￾nates, and a … view at source ↗
Figure 5
Figure 5. Figure 5: Qualitative comparison of plume segmentation predictions on three MethaneAIR test scenes at patch size 448 [PITH_FULL_IMAGE:figures/full_fig_p012_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Qualitative comparison of plume segmentation predictions on three MethaneSAT test scenes at patch size 768 [PITH_FULL_IMAGE:figures/full_fig_p013_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Scene-level detection results for a representative MethaneSAT test scene. Left: input XCH4 map (ppb). Center: ground truth plume masks generated by the wavelet-based labeling method (red contours). Right: model predictions after the full post-processing pipeline (orange contours). Ground truth masks are intentionally conservative, restricted to within 5–10 km of each source to avoid biasing flux estimates … view at source ↗
Figure 8
Figure 8. Figure 8: Representative false positive categories identified during manual test set review. XCH [PITH_FULL_IMAGE:figures/full_fig_p016_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Probabilistic scene reconstruction for two MethaneSAT test scenes. Each row shows the confidence-weighted probability map Pˆ (left), the XCH4 concentration in ppb (center), and a zoom into the primary plume region with Pˆ overlaid on XCH4 (right). The dashed white box in the center panel marks the zoomed region. High-probability regions correspond to detected plume instances; background pixels are assigned… view at source ↗
Figure 10
Figure 10. Figure 10: Geographic distribution of MethaneAIR target basins [PITH_FULL_IMAGE:figures/full_fig_p027_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Mask R-CNN and U-Net performance on the MethaneAIR validation set as a function of learning rate for all four [PITH_FULL_IMAGE:figures/full_fig_p028_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Mask R-CNN (ResNet-50) performance on the [PITH_FULL_IMAGE:figures/full_fig_p028_12.png] view at source ↗
Figure 14
Figure 14. Figure 14: Instance-level detection metrics on the validation set [PITH_FULL_IMAGE:figures/full_fig_p029_14.png] view at source ↗
Figure 13
Figure 13. Figure 13: Mask R-CNN (ResNet-50) performance on the [PITH_FULL_IMAGE:figures/full_fig_p029_13.png] view at source ↗
Figure 15
Figure 15. Figure 15: Instance-level detection metrics on the validation set [PITH_FULL_IMAGE:figures/full_fig_p029_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: Instance-level detection metrics on the validation [PITH_FULL_IMAGE:figures/full_fig_p030_16.png] view at source ↗
Figure 17
Figure 17. Figure 17: Impact of post-processing stages on instance-level precision, recall, and F1 on the validation set. Four configurations [PITH_FULL_IMAGE:figures/full_fig_p031_17.png] view at source ↗
Figure 18
Figure 18. Figure 18: Mask R-CNN (ResNet-50) performance on the MethaneSAT validation set as a function of input patch size, shown [PITH_FULL_IMAGE:figures/full_fig_p032_18.png] view at source ↗
Figure 19
Figure 19. Figure 19: Performance vs. size filtering threshold applied on top of the high-sensitivity mode for the fine-tuning configuration on the MethaneSAT validation set. Precision, recall, and F1 are shown as a function of minimum mask area threshold (px2 ). Shaded bands indicate standard deviation across three cross-validation folds. F1 peaks at 1500 px2 before declining as the threshold encroaches on genuine small plume… view at source ↗
Figure 20
Figure 20. Figure 20: Pixel-level correspondence between predicted probability Pˆ and XCH4, pooled across all MethaneSAT test scenes (n = 40,665,022 pixels with P > ˆ 0). Bins are colored by log10 pixel count. Pearson r = 0.509 and Spearman ρ = 0.444 indicate a consistent positive tendency for higher XCH4 values to receive higher predicted probability within detected regions, while the concentration range of the dense low-prob… view at source ↗
read the original abstract

Automated detection and masking of individual methane plumes from satellite imagery is important for operational emission attribution and quantification. We present a machine learning framework for plume detection from MethaneSAT retrieved column-averaged dry-air mole fractions of methane. We address two core challenges: the scarcity of labeled MethaneSAT data and the need for inference reliability across diverse atmospheric and surface conditions. We first demonstrate that Mask R-CNN with a ResNet-50 backbone outperforms U-Net semantic segmentation on both MethaneAIR (an airborne version of MethaneSAT) and MethaneSAT data, with pixel-level F1 score gains of 10.49 and 5.48 respectively. To address MethaneSAT data scarcity, we evaluate three cross-sensor transfer strategies leveraging MethaneAIR flights and synthetic plumes. Mask R-CNN with ResNet-50 fine-tuned from MethaneAIR pre-trained weights is the most effective strategy, achieving instance-level precision of 0.60 and a near-perfect recall of 0.98 at the baseline operating point. A physics-informed post-processing pipeline converts detections into two operationally distinct modes. The first is a high-sensitivity mode that applies morphological filtering and proximity-based merging for comprehensive emission screening, achieving precision of 0.71 and recall of 0.94. The second is a high-precision mode that additionally applies a distribution-based classifier for confident source attribution, achieving precision of 0.92 and recall of 0.70. Manual review of detections classified as false positives against our wavelet-based ground truth labels reveals that a meaningful fraction of cases correspond to real methane enhancements excluded by conservative labeling criteria, indicating that precision values reported are lower bounds on true detection performance... Our data and code are available at: https://doi.org/10.7910/DVN/FR959H

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 0 minor

Summary. The manuscript presents a machine learning framework for instance-level segmentation of methane plumes in MethaneSAT column-averaged dry-air mole fraction imagery. It shows that Mask R-CNN with ResNet-50 backbone outperforms U-Net, demonstrates the effectiveness of cross-sensor transfer learning from MethaneAIR pre-trained weights over other strategies, and introduces a physics-informed post-processing pipeline that produces two operating modes (high-sensitivity with morphological filtering and proximity merging; high-precision with an added distribution-based classifier). Instance-level metrics are reported as 0.60/0.98 (baseline), 0.71/0.94 (high-sensitivity), and 0.92/0.70 (high-precision). The abstract states that wavelet-based ground-truth labels are conservative and that manual review of false positives indicates a meaningful fraction correspond to real enhancements, so reported precision values are lower bounds on true performance. Data and code are released.

Significance. If the empirical results and the lower-bound interpretation hold after clarification, the work supplies a concrete, transferable pipeline for operational methane plume detection that addresses data scarcity via cross-sensor transfer and offers distinct sensitivity/precision modes. The explicit release of data and code at the cited DOI is a clear strength that supports reproducibility and follow-on work in remote-sensing applications for greenhouse-gas monitoring.

major comments (1)
  1. [Abstract] Abstract: The claim that 'precision values reported are lower bounds on true detection performance' is grounded solely in the statement that 'a meaningful fraction' of false positives against the wavelet labels correspond to real enhancements upon manual review. No counts of reviewed cases, explicit decision criteria for identifying 'real methane enhancements,' or inter-reviewer consistency metrics are supplied. Because this interpretation directly affects the operational reading of the high-precision mode (0.92 precision), the absence of quantification is load-bearing for the central performance claims.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their constructive review and for recognizing the potential utility of the pipeline for operational methane monitoring. We address the single major comment below and commit to revisions that strengthen the presentation of performance claims.

read point-by-point responses

  1. Referee: The claim that 'precision values reported are lower bounds on true detection performance' is grounded solely in the statement that 'a meaningful fraction' of false positives against the wavelet labels correspond to real enhancements upon manual review. No counts of reviewed cases, explicit decision criteria for identifying 'real methane enhancements,' or inter-reviewer consistency metrics are supplied. Because this interpretation directly affects the operational reading of the high-precision mode (0.92 precision), the absence of quantification is load-bearing for the central performance claims.

    Authors: We agree that the lower-bound interpretation requires more rigorous support than is currently provided. The manual review was qualitative and intended only as supporting context for the conservative nature of the wavelet labels. In the revised manuscript we will remove the explicit 'lower bounds' phrasing from the abstract (and any corresponding claims in the main text) unless the original review records can supply the requested counts, decision criteria, and consistency metrics. If the latter is possible we will add a short supplementary note with those details; otherwise the claim will be qualified or omitted to ensure the reported metrics are presented without unsubstantiated qualification. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical metrics on held-out data with independent post-processing

full rationale

The reported instance-level precision and recall values are computed directly from model outputs evaluated against held-out MethaneSAT test imagery using standard detection metrics; they do not reduce by any equation to quantities defined in terms of the model's own fitted parameters or self-referential labels. The physics-informed post-processing (morphological filtering, proximity merging, distribution-based classifier) applies deterministic rules to detections rather than re-deriving performance from the training objective. The lower-bound interpretation of precision rests on an external manual review step whose details are not part of any derivation chain. No self-citations, uniqueness theorems, or ansatzes are invoked to justify core claims. The chain is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 0 invented entities

Abstract-only review provides limited visibility into internal model choices; the central performance claims rest on transferability between sensors and on the interpretation of ground-truth labels.

free parameters (1)
  • post-processing thresholds
    Morphological filtering parameters, proximity merging distance, and distribution-based classifier cutoff are selected to achieve the reported precision-recall operating points in each mode.
axioms (2)
  • domain assumption MethaneAIR airborne retrievals have sufficiently similar characteristics to MethaneSAT satellite retrievals for effective domain transfer
    Foundation for pre-training and fine-tuning strategy described in abstract.
  • domain assumption Wavelet-based labeling produces reliable but conservative ground truth that excludes some real plumes
    Underpins the claim that reported precision is a lower bound, based on manual review.

pith-pipeline@v0.9.1-grok · 5934 in / 1636 out tokens · 43004 ms · 2026-06-30T15:16:39.239335+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

95 extracted references · 9 canonical work pages · 6 internal anchors

  1. [1]

    Global methane tracker 2023

    IEA. Global methane tracker 2023. Technical report, International Energy Agency, Paris, 2023. Licence: CC BY 4.0 (report); CC BY NC SA 4.0 (Annex A)

    2023

  2. [2]

    Methane leaks from north american natural gas systems.Science, 343(6172):733–735, 2014

    Adam R Brandt, GA Heath, EA Kort, Francis O’Sullivan, Gabrielle P ´etron, Sarah M Jordaan, P Tans, Jennifer Wilcox, AM Gopstein, Doug Arent, et al. Methane leaks from north american natural gas systems.Science, 343(6172):733–735, 2014

    2014

  3. [3]

    Reconciling divergent estimates of oil and gas methane emissions.Proceedings of the National Academy of Sciences, 112(51):15597–15602, 2015

    Daniel Zavala-Araiza, David R Lyon, Ram ´on A Alvarez, Kenneth J Davis, Robert Harriss, Scott C Herndon, Anna Karion, Eric Adam Kort, Brian K Lamb, Xin Lan, et al. Reconciling divergent estimates of oil and gas methane emissions.Proceedings of the National Academy of Sciences, 112(51):15597–15602, 2015

    2015

  4. [4]

    Thorpe, David R

    Christian Frankenberg, Andrew K. Thorpe, David R. Thompson, Glynn Hulley, Eric Adam Kort, Nick Vance, Jakob Borchardt, Thomas Krings, Konstantin Gerilowski, Colm Sweeney, Stephen Conley, Brian D. Bue, An- drew D. Aubrey, Simon Hook, and Robert O. Green. Airborne methane remote measurements reveal heavy-tail flux distribution in four corners region.Proceed...

    2016

  5. [5]

    D. J. Varon, D. Jervis, J. McKeever, I. Spence, D. Gains, and D. J. Jacob. High-frequency monitoring of anoma- lous methane point sources with multispectral sentinel-2 satellite observations.Atmospheric Measurement Tech- niques, 14(4):2771–2785, 2021

    2021

  6. [6]

    Lyon, Benjamin Hmiel, Katherine A

    Mark Omara, Daniel Zavala-Araiza, David R. Lyon, Benjamin Hmiel, Katherine A. Roberts, and Steven P. Hamburg. Methane emissions from US low production oil and natural gas well sites.Nature Communications, 13(1):2085, 2022

    2085

  7. [7]

    J. P. Williams, M. Omara, A. Himmelberger, D. Zavala- Araiza, K. MacKay, J. Benmergui, M. Sargent, S. C. Wofsy, S. P. Hamburg, and R. Gautam. Small emission sources in aggregate disproportionately account for a large majority of total methane emissions from the us oil and gas sector.Atmospheric Chemistry and Physics, 25(3):1513–1532, 2025

    2025

  8. [8]

    Myhre, D

    G. Myhre, D. Shindell, F.-M. Br ´eon, W. Collins, J. Fu- glestvedt, J. Huang, D. Koch, J.-F. Lamarque, D. Lee, B. Mendoza, T. Nakajima, A. Robock, G. Stephens, T. Takemura, and H. Zhang.Anthropogenic and natural radiative forcing, pages 659–740. Cambridge University Press, Cambridge, UK, 2013

    2013

  9. [9]

    Highwood, and Keith P

    Maryam Etminan, Gunnar Myhre, Eleanor J. Highwood, and Keith P. Shine. Radiative forcing of carbon dioxide, methane, and nitrous oxide: A significant revision of the methane radiative forcing.Geophysical Research Letters, 43:12,614 – 12,623, 2016. SUBMITTED TO IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING 21

    2016

  10. [10]

    Global methane pledge: which countries are cutting emissions?World Economic Forum, August 2024

    World Economic Forum. Global methane pledge: which countries are cutting emissions?World Economic Forum, August 2024. Accessed: 2025-07-17

    2024

  11. [11]

    Oil & gas decarbonization charter

    COP28 UAE. Oil & gas decarbonization charter. https: //www.ogdc.org/, 2023. Launched at COP28, Dubai

    2023

  12. [12]

    D. J. Jacob, A. J. Turner, J. D. Maasakkers, J. Sheng, K. Sun, X. Liu, K. Chance, I. Aben, J. McKeever, and C. Frankenberg. Satellite observations of at- mospheric methane and their value for quantifying methane emissions.Atmospheric Chemistry and Physics, 16(22):14371–14396, 2016

    2016

  13. [13]

    Jervis, J

    D. Jervis, J. McKeever, B. O. A. Durak, J. J. Sloan, D. Gains, D. J. Varon, A. Ramier, M. Strupler, and E. Tar- rant. The ghgsat-d imaging spectrometer.Atmospheric Measurement Techniques, 14(3):2127–2140, 2021

    2021

  14. [14]

    Green, A

    R. Green, A. Thorpe, P. Brodrick, D. Chadwick, A. Lopez, C. Elder, C. Villanueva-Weeks, J. Fahlen, R. W. Coleman, D. Jensen, H. Bender, Q. Vinckier, C. Xi- ang, W. Olson-Duvall, S. Lundeen, and D. Thompson. EMIT L2B Estimated Methane Plume Complexes 60 m V001, 2023. Date Accessed: 2026-01-29

    2023

  15. [15]

    Tobias Storch, Hans-Peter Honold, Sabine Chabrillat, Martin Habermeyer, Paul Tucker, Maximilian Brell, An- dreas Ohndorf, Katrin Wirth, Matthias Betz, Michael Kuchler, Helmut M¨uhle, Emiliano Carmona, Simon Baur, Martin M ¨ucke, Sebastian L ¨ow, Daniel Schulze, Stef- fen Zimmermann, Christoph Lenzen, Sebastian Wiesner, Saika Aida, Ralph Kahle, Peter Willb...

    2023

  16. [16]

    Cogliati, F

    S. Cogliati, F. Sarti, L. Chiarantini, M. Cosi, R. Lorusso, E. Lopinto, F. Miglietta, L. Genesio, L. Guanter, A. Damm, S. P ´erez-L´opez, D. Scheffler, G. Tagliabue, C. Panigada, U. Rascher, T.P.F. Dowling, C. Giardino, and R. Colombo. The prisma imaging spectroscopy mission: overview and first performance analysis.Remote Sensing of Environment, 262:112499, 2021

    2021

  17. [17]

    R. M. Duren, J. Guido, J. Herner, S. Rao, R. O. Green, M. de Belloy, R. Schingler, D. R. Ardila, A. K. Thorpe, D. Cusworth, M. Falk, E. Scheehle, C. Vahlsing, T. Ide, G. Asner, and R. Lawrence. Carbon Mapper: global tracking of methane and CO2 point-sources. InAGU Fall Meeting Abstracts, volume 2020 ofAGU Fall Meeting Abstracts, pages A247–01, December 2020

    2020

  18. [18]

    Duren, D

    R. Duren, D. Cusworth, A. Ayasse, K. Howell, A. Di- amond, T. Scarpelli, J. Kim, K. O’neill, J. Lai-Norling, A. Thorpe, S. R. Zandbergen, L. Shaw, M. Keremed- jiev, J. Guido, P. Giuliano, M. Goldstein, R. Nallapu, G. Barentsen, D. R. Thompson, K. Roth, D. Jensen, M. Eastwood, F. Reuland, T. Adams, A. Brandt, E. A. Kort, J. Mason, and R. O. Green. The carb...

    2025

  19. [19]

    Multisatellite imaging of a gas well blowout enables quantification of total methane emissions.Geophysical Research Letters, 48(2):e2020GL090864, 2021

    Daniel H Cusworth, Riley M Duren, Andrew K Thorpe, Sudhanshu Pandey, Joannes D Maasakkers, Ilse Aben, Dylan Jervis, Daniel J Varon, Daniel J Jacob, Cyn- thia A Randles, et al. Multisatellite imaging of a gas well blowout enables quantification of total methane emissions.Geophysical Research Letters, 48(2):e2020GL090864, 2021

    2021

  20. [20]

    Zhang, E

    Z. Zhang, E. D. Sherwin, D. J. Varon, and A. R. Brandt. Detecting and quantifying methane emissions from oil and gas production: algorithm development with ground- truth calibration based on sentinel-2 satellite imagery.At- mospheric Measurement Techniques, 15(23):7155–7169, 2022

    2022

  21. [21]

    Cusworth, Daniel J

    Luis Guanter, Itziar Irakulis-Loitxate, Javier Gorro ˜no, Elena S ´anchez-Garc´ıa, Daniel H. Cusworth, Daniel J. Varon, Sergio Cogliati, and Roberto Colombo. Mapping methane point emissions with the PRISMA spaceborne imaging spectrometer.Remote Sensing of Environment, 265:112671, November 2021

    2021

  22. [22]

    Maasakkers, Daniel Zavala-Araiza, and Ilse Aben

    Itziar Irakulis-Loitxate, Luis Guanter, Joannes D. Maasakkers, Daniel Zavala-Araiza, and Ilse Aben. Satel- lites detect abatable super-emissions in one of the world’s largest methane hotspot regions.Environmental Sci- ence & Technology, 56(4):2143–2152, 2022. PMID: 35102741

    2022

  23. [23]

    Veefkind, I

    J.P. Veefkind, I. Aben, K. McMullan, H. F ¨orster, J. de Vries, G. Otter, J. Claas, H.J. Eskes, J.F. de Haan, Q. Kleipool, M. van Weele, O. Hasekamp, R. Hoogeveen, J. Landgraf, R. Snel, P. Tol, P. Ingmann, R. V oors, B. Kruizinga, R. Vink, H. Visser, and P.F. Levelt. Tropomi on the esa sentinel-5 precursor: A gmes mission for global observations of the at...

    2012

  24. [24]

    Masahiro Kasuya, Masakatsu Nakajima, and Takashi Hamazaki. Greenhouse gases observing satellite (gosat) program overview and its development status.Transac- tions of the Japan Society for Aeronautical and Space Sciences, Space Technology Japan, 7(ists26):To 4 5– To 4 10, 2009

    2009

  25. [25]

    Varon, Itziar Irakulis- Loitxate, Nicholas Balasus, and Daniel J

    Marc Watine-Guiu, Daniel J. Varon, Itziar Irakulis- Loitxate, Nicholas Balasus, and Daniel J. Jacob. Geo- stationary satellite observations of extreme and tran- sient methane emissions from oil and gas infrastruc- ture.Proceedings of the National Academy of Sciences, 120(52):e2310797120, 2023

    2023

  26. [26]

    R Rohrschneider, S

    R. R Rohrschneider, S. Wofsy, J. E. Franklin, J. Benmer- gui, J. C Soto, and Spencer B Davis. The methanesat mission.Small Satellite Conference

  27. [27]

    Frankenberg, U

    C. Frankenberg, U. Platt, and T. Wagner. Iterative max- imum a posteriori (imap)-doas for retrieval of strongly SUBMITTED TO IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING 22 absorbing trace gases: Model studies for ch 4 and co 2 retrieval from near infrared spectra of sciamachy onboard envisat.Atmospheric Chemistry and Physics, 5(1):9–22, 2005

    2005

  28. [28]

    Krings, K

    T. Krings, K. Gerilowski, M. Buchwitz, M. Reuter, A. Tretner, J. Erzinger, D. Heinze, U. Pfl ¨uger, J. P. Bur- rows, and H. Bovensmann. Mamap – a new spectrometer system for column-averaged methane and carbon dioxide observations from aircraft: retrieval algorithm and first inversions for point source emission rates.Atmospheric Measurement Techniques, 4(9...

    2011

  29. [29]

    Chan Miller, S

    C. Chan Miller, S. Roche, J. S. Wilzewski, X. Liu, K. Chance, A. H. Souri, E. Conway, B. Luo, J. Samra, J. Hawthorne, K. Sun, C. Staebell, A. Chulakadabba, M. Sargent, J. S. Benmergui, J. E. Franklin, B. C. Daube, Y . Li, J. L. Laughner, B. C. Baier, R. Gautam, M. Omara, and S. C. Wofsy. Methane retrieval from methaneair using the co 2 proxy approach: a d...

    2024

  30. [30]

    K. Sun, L. Zhu, K. Cady-Pereira, C. Chan Miller, K. Chance, L. Clarisse, P.-F. Coheur, G. Gonz ´alez Abad, G. Huang, X. Liu, M. Van Damme, K. Yang, and M. Zondlo. A physics-based approach to oversam- ple multi-satellite, multispecies observations to a com- mon grid.Atmospheric Measurement Techniques, 11(12):6679–6701, 2018

    2018

  31. [31]

    MethaneSAT loses contact with satellite

    Environmental Defense Fund. MethaneSAT loses contact with satellite. Press release, July 2025. Accessed: 2026- 03-03

    2025

  32. [32]

    E. K. Conway, A. H. Souri, J. Benmergui, K. Sun, X. Liu, C. Staebell, C. Chan Miller, J. Franklin, J. Samra, J. Wilzewski, S. Roche, B. Luo, A. Chulakadabba, M. Sargent, J. Hohl, B. Daube, I. Gordon, K. Chance, and S. Wofsy. Level0 to level1b processor for methaneair.At- mospheric Measurement Techniques, 17(4):1347–1362, 2024

    2024

  33. [33]

    Zhang, M

    Z. Zhang, M. Sargent, J. D. Warren, A. Chulakadabba, M. Russi, S. Ayvazov, J. Benmergui, M. Knapp, E. Kyzi- vat, C. C. Miller, S. Roche, B. Luo, D. J. Miller, M. Nasr, K. Sun, J. P. Williams, K. MacKay, M. Omara, L. Guan- ter, R. Gautam, J. Franklin, X. Liu, and S. C. Wofsy. Automatic methane plume masking based on wavelet transform image processing: Appl...

    2026

  34. [34]

    Thorpe, Geor- gios Matheou, and Christian Frankenberg

    Siraput Jongaramrungruang, Andrew K. Thorpe, Geor- gios Matheou, and Christian Frankenberg. Methanet – an ai-driven approach to quantifying methane point- source emission from high-resolution 2-d plume imagery. Remote Sensing of Environment, 269:112809, 2022

    2022

  35. [35]

    Automatic detection of methane emissions in multispectral satellite imagery using a vision transformer.Nature Communica- tions, 15(1):3801, 2024

    Bertrand Rouet-Leduc and Claudia Hulbert. Automatic detection of methane emissions in multispectral satellite imagery using a vision transformer.Nature Communica- tions, 15(1):3801, 2024

    2024

  36. [36]

    Semantic segmentation of methane plumes with hyperspectral machine learning models.Scientific Reports, 13(1), 2023

    V R ˚uˇziˇcka, G Mateo-Garcia, L G ´omez-Chova, A Vaughan, L Guanter, and A Markham. Semantic segmentation of methane plumes with hyperspectral machine learning models.Scientific Reports, 13(1), 2023

    2023

  37. [37]

    Towards Operational Automated Greenhouse Gas Plume Detection and Delineation

    Brian D. Bue, Jake H. Lee, Andrew K. Thorpe, Philip G. Brodrick, Daniel H. Cusworth, Alana K. Ayasse, Vassiliki Mancoridis, Anagha Satish, Shujun Xiong, and Riley M. Duren. Towards operational automated greenhouse gas plume detection.ArXiv, abs/2505.21806, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  38. [38]

    A machine learning approach to methane emissions mitigation in the oil and gas industry, 11 2020

    Jiayang Wang, Selvaprabu Nadarajah, Jingfan Wang, and Arvind Ravikumar. A machine learning approach to methane emissions mitigation in the oil and gas industry, 11 2020

    2020

  39. [39]

    B. J. Schuit, J. D. Maasakkers, P. Bijl, G. Mahapatra, A.- W. van den Berg, S. Pandey, A. Lorente, T. Borsdorff, S. Houweling, D. J. Varon, J. McKeever, D. Jervis, M. Girard, I. Irakulis-Loitxate, J. Gorro ˜no, L. Guanter, D. H. Cusworth, and I. Aben. Automated detection and monitoring of methane super-emitters using satellite data. Atmospheric Chemistry ...

    2023

  40. [40]

    Lee, Andrew K

    Vassiliki Mancoridis, Brian Bue, Jake H. Lee, Andrew K. Thorpe, Daniel Cusworth, Alana Ayasse, Philip G. Bro- drick, and Riley Duren. Multiplatform methane plume detection via model and domain adaptation.IEEE Transactions on Geoscience and Remote Sensing, 63:1– 12, 2025

    2025

  41. [41]

    Meqnet: Deep learning for methane point source emission quantification from sentinel-2 observations

    Di Xu, Philippa Mason, Jianguo Liu, and Yanghua Wang. Meqnet: Deep learning for methane point source emission quantification from sentinel-2 observations. In NeurIPS 2025 Workshop on Tackling Climate Change with Machine Learning, 2025

    2025

  42. [42]

    U-Net: Convolutional Networks for Biomedical Image Segmentation

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U-net: Convolutional networks for biomedical image segmentation.ArXiv, abs/1505.04597, 2015

    work page internal anchor Pith review Pith/arXiv arXiv 2015

  43. [43]

    J. H. Bruno, D. Jervis, D. J. Varon, and D. J. Ja- cob. U-plume: automated algorithm for plume de- tection and source quantification by satellite point- source imagers.Atmospheric Measurement Techniques, 17(9):2625–2636, 2024

    2024

  44. [44]

    Vaughan, G

    A. Vaughan, G. Mateo-Garc ´ıa, L. G ´omez-Chova, V . R˚uˇziˇcka, L. Guanter, and I. Irakulis-Loitxate. Ch4net: a deep learning model for monitoring methane super- emitters with sentinel-2 imagery.Atmospheric Measure- ment Techniques, 17(9):2583–2593, 2024

    2024

  45. [45]

    Ai for op- erational methane emitter monitoring from space.arXiv preprint arXiv:2408.04745, 2024

    Anna Vaughan, Gonzalo Mateo-Garcia, Itziar Irakulis- Loitxate, Marc Watine, Pablo Fernandez-Poblaciones, Richard E Turner, James Requeima, Javier Gorro ˜no, Cynthia Randles, Manfredi Caltagirone, et al. Ai for op- erational methane emitter monitoring from space.arXiv preprint arXiv:2408.04745, 2024

    work page arXiv 2024

  46. [46]

    Mask r-cnn

    Kaiming He, Georgia Gkioxari, Piotr Doll ´ar, and Ross Girshick. Mask r-cnn. InProceedings of the IEEE in- ternational conference on computer vision, pages 2961– 2969, 2017

    2017

  47. [47]

    Faster r-cnn: Towards real-time object detection with region proposal networks.Advances in neural information processing systems, 28, 2015

    Shaoqing Ren, Kaiming He, Ross Girshick, and Jian Sun. Faster r-cnn: Towards real-time object detection with region proposal networks.Advances in neural information processing systems, 28, 2015

    2015

  48. [48]

    Manjunath

    Satish Kumar, Carlos Torres, Oytun Ulutan, Alana Ayasse, Dar Roberts, and B.S. Manjunath. Deep remote SUBMITTED TO IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING 23 sensing methods for methane detection in overhead hy- perspectral imagery. In2020 IEEE Winter Conference on Applications of Computer Vision (WACV), pages 1765– 1774, 2020

    2020

  49. [49]

    Zhang, Shaoqing Ren, and Jian Sun

    Kaiming He, X. Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition.2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 770–778, 2015

    2016

  50. [50]

    Unlocking the potential: Multi-task deep learning for spaceborne quan- titative monitoring of fugitive methane plumes.arXiv preprint arXiv:2401.12870, 2024

    Guoxin Si, Shiliang Fu, and Wei Yao. Unlocking the potential: Multi-task deep learning for spaceborne quan- titative monitoring of fugitive methane plumes.arXiv preprint arXiv:2401.12870, 2024

    work page arXiv 2024

  51. [51]

    Chulakadabba, M

    A. Chulakadabba, M. Sargent, T. Lauvaux, J. S. Ben- mergui, J. E. Franklin, C. Chan Miller, J. S. Wilzewski, S. Roche, E. Conway, A. H. Souri, K. Sun, B. Luo, J. Hawthrone, J. Samra, B. C. Daube, X. Liu, K. Chance, Y . Li, R. Gautam, M. Omara, J. S. Rutherford, E. D. Sherwin, A. Brandt, and S. C. Wofsy. Methane point source quantification using methaneair...

    2023

  52. [52]

    Pittman, Jenna Samra, Jacob Hawthrone, Eleanor Walker, Apisada Chulakadabba, Ethan Manninen, Jacob Bushey, Bingkun Luo, David J

    Raia Ottenheimer, Laine Roper, Zhan Zhang, Maryann Sargent, Jack Warren, S ´ebastien Roche, Christopher Chan Miller, Ethan Kyzivat, Marvin Knapp, Joshua Benmergui, Jasna V . Pittman, Jenna Samra, Jacob Hawthrone, Eleanor Walker, Apisada Chulakadabba, Ethan Manninen, Jacob Bushey, Bingkun Luo, David J. Miller, Maya Nasr, Kang Sun, Jonathan Franklin, Xiong ...

    2026

  53. [53]

    High-resolution methane mapping with the enmap satellite imaging spectroscopy mission.IEEE Transactions on Geoscience and Remote Sensing, PP:1– 1, 01 2024

    Javier Roger, Itziar Irakulis-Loitxate, Adriana Valverde, Javier Gorro ˜no, Sabine Chabrillat, Max Brell, and Luis Guanter. High-resolution methane mapping with the enmap satellite imaging spectroscopy mission.IEEE Transactions on Geoscience and Remote Sensing, PP:1– 1, 01 2024

    2024

  54. [54]

    A. K. Thorpe, C. Frankenberg, and D. A. Roberts. Retrieval techniques for airborne imaging of methane concentrations using high spatial and moderate spectral resolution: application to aviris.Atmospheric Measure- ment Techniques, 7(2):491–506, 2014

    2014

  55. [55]

    D. H. Cusworth, D. J. Jacob, J.-X. Sheng, J. Benmergui, A. J. Turner, J. Brandman, L. White, and C. A. Randles. Detecting high-emitting methane sources in oil/gas fields using satellite observations.Atmospheric Chemistry and Physics, 18(23):16885–16896, 2018

    2018

  56. [56]

    Fast and accurate retrieval of methane concentration from imaging spec- trometer data using sparsity prior.IEEE Transactions on Geoscience and Remote Sensing, 58(9):6480–6492, 2020

    Markus D Foote, Philip E Dennison, Andrew K Thorpe, David R Thompson, Siraput Jongaramrungruang, Chris- tian Frankenberg, and Sarang C Joshi. Fast and accurate retrieval of methane concentration from imaging spec- trometer data using sparsity prior.IEEE Transactions on Geoscience and Remote Sensing, 58(9):6480–6492, 2020

    2020

  57. [57]

    Guanter, J

    L. Guanter, J. Warren, M. Omara, A. Chulakadabba, J. Roger, M. Sargent, J. E. Franklin, S. C. Wofsy, and R. Gautam. Detection and quantification of methane plumes with the methaneair airborne spectrometer.At- mospheric Measurement Techniques, 18(15):3857–3872, 2025

    2025

  58. [58]

    Guanter, J

    L. Guanter, J. Roger, J. Warren, M. Sargent, Z. Zhang, S. Roche, C. C. Miller, M. Steiner, H. Hadfield, M. Omara, J. Williams, K. MacKay, J. E. Franklin, B. Luo, S. C. Wofsy, S. P. Hamburg, and R. Gautam. Surveying methane point-source super-emissions across oil and gas basins with methanesat.Atmospheric Chem- istry and Physics, 26(4):2941–2963, 2026

    2026

  59. [59]

    S ´anchez-Garc´ıa, J

    E. S ´anchez-Garc´ıa, J. Gorro˜no, I. Irakulis-Loitxate, D. J. Varon, and L. Guanter. Mapping methane plumes at very high spatial resolution with the worldview-3 satellite.At- mospheric Measurement Techniques, 15(6):1657–1674, 2022

    2022

  60. [60]

    D. R. Thompson, A. K. Thorpe, C. Frankenberg, R. O. Green, R. Duren, L. Guanter, A. Hollstein, E. Middleton, L. Ong, and S. Ungar. Space-based remote imaging spectroscopy of the aliso canyon ch4 superemitter.Geo- physical Research Letters, 43(12):6571–6578, 2016

    2016

  61. [61]

    D. J. Varon, D. J. Jacob, J. McKeever, D. Jervis, B. O. A. Durak, Y . Xia, and Y . Huang. Quantifying methane point sources from fine-scale satellite observations of atmospheric methane plumes.Atmospheric Measurement Techniques, 11(10):5673–5686, 2018

    2018

  62. [62]

    J. D. Warren, M. Sargent, J. P. Williams, M. Omara, C. C. Miller, S. Roche, K. MacKay, E. Manninen, A. Chu- lakadabba, A. Himmelberger, J. Benmergui, Z. Zhang, L. Guanter, S. Wofsy, and R. Gautam. Sectoral contribu- tions of high-emitting methane point sources from major us onshore oil and gas producing basins using airborne measurements from methaneair.A...

    2025

  63. [63]

    U-net: Convolutional networks for biomedical image segmentation

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U-net: Convolutional networks for biomedical image segmentation. In Nassir Navab, Joachim Hornegger, William M. Wells, and Alejandro F. Frangi, editors, Medical Image Computing and Computer-Assisted Inter- vention – MICCAI 2015, pages 234–241, Cham, 2015. Springer International Publishing

    2015

  64. [64]

    Joyce, C

    P. Joyce, C. Ruiz Villena, Y . Huang, A. Webb, M. Gloor, F. H. Wagner, M. P. Chipperfield, R. Barrio Guill ´o, C. Wilson, and H. Boesch. Using a deep neural network to detect methane point sources and quantify emissions from prisma hyperspectral satellite images.Atmospheric Measurement Techniques, 16(10):2627–2640, 2023

    2023

  65. [65]

    Attmetnet: Attention-enhanced deep neural network for methane plume detection in sentinel-2 satellite imagery.arXiv preprint arXiv:2512.02751, 2025

    Rakib Ahsan, MD Shanto, Md Sultanul Arifin, and Tanzima Hashem. Attmetnet: Attention-enhanced deep neural network for methane plume detection in sentinel-2 satellite imagery.arXiv preprint arXiv:2512.02751, 2025

    work page arXiv 2025

  66. [66]

    An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, and Neil Houlsby. An image is worth 16x16 words: Transformers for image SUBMITTED TO IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING 24 recognition at scale.ArXiv,...

    work page internal anchor Pith review Pith/arXiv arXiv 2010

  67. [67]

    Segformer: Sim- ple and efficient design for semantic segmentation with transformers.Advances in neural information processing systems, 34:12077–12090, 2021

    Enze Xie, Wenhai Wang, Zhiding Yu, Anima Anandku- mar, Jose M Alvarez, and Ping Luo. Segformer: Sim- ple and efficient design for semantic segmentation with transformers.Advances in neural information processing systems, 34:12077–12090, 2021

    2021

  68. [68]

    Efficientvit: Memory efficient vision transformer with cascaded group attention

    Xinyu Liu, Houwen Peng, Ningxin Zheng, Yuqing Yang, Han Hu, and Yixuan Yuan. Efficientvit: Memory efficient vision transformer with cascaded group attention. InPro- ceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 14420–14430, 2023

    2023

  69. [69]

    R˚uˇziˇcka and A

    V . R˚uˇziˇcka and A. Markham. Hyperspectralvits: General hyperspectral models for on-board remote sensing.IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2025

    2025

  70. [70]

    Global monitoring of methane point sources using deep learning on hyperspectral radiance measurements from EMIT

    Vishal V Batchu, Michelangelo Conserva, Alex Wil- son, Anna M Michalak, Varun Gulshan, Philip G Bro- drick, Andrew K Thorpe, and Christopher V Arsdale. Global monitoring of methane point sources using deep learning on hyperspectral radiance measurements from emit.arXiv preprint arXiv:2604.10094, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  71. [71]

    Varon, and Fariba Mohammadimanesh

    Ali Radman, Masoud Mahdianpari, Daniel J. Varon, and Fariba Mohammadimanesh. S2metnet: A novel dataset and deep learning benchmark for methane point source quantification using sentinel-2 satellite imagery.Remote Sensing of Environment, 295:113708, 2023

    2023

  72. [72]

    Unpaired image-to-image translation using cycle- consistent adversarial networks

    Jun-Yan Zhu, Taesung Park, Phillip Isola, and Alexei A Efros. Unpaired image-to-image translation using cycle- consistent adversarial networks. InProceedings of the IEEE international conference on computer vision, pages 2223–2232, 2017

    2017

  73. [73]

    Curriculum learning

    Yoshua Bengio, J ´erˆome Louradour, Ronan Collobert, and Jason Weston. Curriculum learning. InProceedings of the 26th Annual International Conference on Machine Learning, ICML ’09, page 41–48, New York, NY , USA,

  74. [74]

    Association for Computing Machinery

  75. [75]

    How transferable are features in deep neural networks?Advances in neural information processing systems, 27, 2014

    Jason Yosinski, Jeff Clune, Yoshua Bengio, and Hod Lipson. How transferable are features in deep neural networks?Advances in neural information processing systems, 27, 2014

    2014

  76. [76]

    Deep learning for clouds and cloud shadow segmentation in methane satellite and airborne imaging spectroscopy.IEEE Trans- actions on Geoscience and Remote Sensing, pages 1–1, 2026

    Manuel P ´erez-Carrasco, Maya Nasr, Sebastien Roche, Chris Chan Miller, Zhan Zhang, Core Francisco Park, Eleanor Walker, Cecilia Garraffo, Douglas Finkbeiner, Sasha Ayvazov, Jonathan Franklin, Bingkun Luo, Xiong Liu, Ritesh Gautam, and Steven Wofsy. Deep learning for clouds and cloud shadow segmentation in methane satellite and airborne imaging spectrosco...

    2026

  77. [77]

    The methanesat mission

    Reuben R Rohrschneider, Steven Wofsy, Jonathan E Franklin, Joshua Benmergui, Juancarlos Soto, and Spencer B Davis. The methanesat mission. 2021

    2021

  78. [78]

    S. C. Wofsy and S. Hamburg. MethaneSAT - A New Observing Platform For High Resolution Measurements Of Methane and Carbon Dioxide. InAGU Fall Meeting Abstracts, volume 2019 ofAGU Fall Meeting Abstracts, pages A53F–02, December 2019

    2019

  79. [79]

    Ouerghi, T

    E. Ouerghi, T. Ehret, G. Facciolo, E. Meinhardt, R. Mar- ion, and J.-M. Morel. Tightening up methane plume source rate estimation in enmap and prisma images.At- mospheric Measurement Techniques, 18(18):4611–4629, 2025

    2025

  80. [80]

    Lauvaux, A

    T. Lauvaux, A. E. Schuh, M. Uliasz, S. Richardson, N. Miles, A. E. Andrews, C. Sweeney, L. I. Diaz, D. Mar- tins, P. B. Shepson, and K. J. Davis. Constraining the co2 budget of the corn belt: exploring uncertainties from the assumptions in a mesoscale inverse system.Atmospheric Chemistry and Physics, 12(1):337–354, January 2012

    2012

Showing first 80 references.