Remote Sensing Methane Emissions in the Bowen Region

We have previously discussed technologies that can be used for the remote sensing of methane emissions. In this article, we are going to focus on GeoSynergy’s activities in the Bowen region of Queensland. We will review some of our ongoing R&D activities, and highlight promising areas for future development.

1. Detection and Quantification of Single Methane Plumes in TROPOMI

Large methane emissions from point sources, such as coal mines, may be clearly visible in daily TROPOMI products (if not occluded by cloud). In this case, they can be identified by inspection, and quantified using a method such as the cross-sectional flux (CSF), as reported by Sadavarte et al. [1].

In the CSF method, we draw transects across a methane plume, along the major axis of the wind vector. Flux can be calculated, based on the assumption of a 2D plume, using a modified formula for volumetric flow rate. i.e. flow velocity times cross-sectional area. An example CSF calculation is shown below for an emission from Oaky Creek on 6 October 2019.

Figure 1: example cross-sectional flux calculation for a methane emission from Oaky Creek coal mine, in the Bowen region, on 6 October 2019. The pale blue box indicates the upwind area used to calculate background CH4. The pink line indicates the major axis of the wind vector. The small aqua box indicates the suggested point of origin. The TROPOMI CH4 data is shown in the bottom image, with a plume clearly visible.

The following figure shows some examples of quantified methane emissions, from super-emitting coal mines in the Bowen region. We have repeated, and independently verified, the flux values published by Sadavarte et al. for the methane emissions shown below, over the Bowen region.

Figure 2: methane plumes visible in daily TROPOMI CH4 products. ERA5 wind vectors also shown. Bottom right: table of corresponding flux values by Sadavarte et al. (independently repeated / verified by GeoSynergy) [1].

In TROPOMI products, methane plumes often have characteristic shapes. Specifically, we are looking for a high methane level above the point source, with a plume that extends away from the source, along the direction of prevailing wind (at the time of sensing), and then declines in strength along its axis. One of the limiting factors, with regard to plume detection, is the need for visual inspection of TROPOMI products. However, the detection of TROPOMI plumes is a task well-suited to AI. This technology opens up the possibility of both persistent monitoring of single sites, as well as ongoing global monitoring for new emissions.

Since our last article on TROPOMI, Schuit et al. [2] have published their AI for this purpose. They used a two-stage approach: a Convolutional Neural Network (CNN) was used to detect plume-like structures in TROPOMI CH4 data; a Support Vector Classifier (SVC) was then used to screen for artefacts amongst the potential plumes. The SVC incorporated additional information, such as wind vectors. Quantification was also be automated, using the Integrated Mass Enhancement (IME) method, whereby the emission rate (Q) is calculated from the observed methane enhancement in the plume, and the local wind [3]. An example detection is shown below:

Figure 3: Left: TROPOMI CH4. Middle: Class Activation Map which highlights the areas identified by the CNN as “plume-like structure”. Right: resultant plume mask, where “x” indicates the suggested origin [2].

AI screening of TROPOMI CH4 has considerable scope for further innovation, and this is an active area of interest for GeoSynergy. Some areas of active development include the use of richer imaging, and hybrid, datasets; and the incorporation of supplementary data at the CNN stage. We suggest that this is preferred over the use of a two-stage model, because it enables the CNN to develop more nuanced feature maps. In addition to wind vectors, supplementary data might include a “prior” based on site-specific emission characteristics, something that was excluded from the cited study.

2. Detection of Methane Emissions using Wind-rotated TROPOMI

Some of the challenges of using TROPOMI CH4 data are its relatively low spatial resolution (7x5.5 km), and its reliance on cloud free observations. TROPOMI data is often sparse. It is not uncommon for known plumes to be extremely difficult to detect within a single TROPOMI product, due to data sparsity and/or cloud occlusions.

However, as we have previously discussed, the wind rotation processing technique can enhance TROPOMI data, and allow point sources to be localized with more confidence [4]. These wind-rotated methane plumes are mean emissions, compiled over some period (months to years). However, mean methane emission rates can be still be calculated, based on the IME method discussed above [3, 4].

At GeoSynergy, we have been exploring how this technique can be developed, and applied to the detection of methane emissions from Australian coal mines. Although originally reported for periods of >=12 months [4], where TROPOMI data is sufficiently rich, the technique can be applied to monthly TROPOMI data.

The following examples show wind rotated methane plumes from coal mines in the Bowen region, over two different time windows.

Figure 4: Hail Creek coal mine in Bowen, Queensland. Left: Mean TROPOMI CH4, 2021. Right: plume enhancement resulting from wind-rotation, 2021.

Figure 5: Moranbah North coal mine in Bowen, Queensland. Left: Mean TROPOMI CH4, November 2020. Right: plume enhancement resulting from wind-rotation, November 2020.

With the requisite domain knowledge, both spatial (mine structure) and temporal (mine activity), we can attempt to interpret the cause of the observed emission. The following figure shows a wind-rotated methane plume at Moranbah North coal mine in October 2020, along with a map of the mine infrastructure. Relevant AOIs are highlighted.

Interestingly, inactive mines, such as Ravensworth underground coal mine owned by Glencore, are substantial methane emitters. The Australian Conservation Foundation recently found that Ravensworth reported 156 kt CO2-e to the CER in 2020 [5]. In this context, TROPOMI CH4 can be a useful tool for the ongoing monitoring of inactive mine sites. The following wind-rotated methane plume was detected over Ravensworth in September 2019.

Figure 7: wind-rotated methane plume was detected over Ravensworth in September 2019.

3. Detection and Quantification of Methane Emissions using Sentinel-2

TROPOMI provides global coverage, with daily revisit time, but it primarily detects emissions of >10 t/h, due to the large effective pixel size of the TROPOMI instrument [6]. Therefore, other techniques are also needed, to detect smaller point source emissions.

Methane point source monitoring using Sentinel-2 is an area of active development for GeoSynergy. Sentinel-2 methane enhancement has a minimum detection limit of 1–3 t/h [6]. For many point source emissions, this is significant. Consider, for example, that vents of large underground coal mines typically emit methane in the range of 1–10 t/h, so likely outside the range of TROPOMI detection [7].

With a spatial resolution of 20m in the methane sensitive bands (B11/B12), global coverage, and open source availability, Sentinel-2 is positioned to make an important contribution to point source methane monitoring. The high revisit time of Sentinel-2 is also suited to monitoring point sources, which are often intermittent in nature.

In our previous article we discussed Sentinel-2 processing for methane enhancement [8, 9, 10, 11], so here we will just focus on the active areas of R&D at GeoSyngery. This area has considerable scope for innovation and, as such, the study by Sherwin et al. is particularly useful [6]. In this study, the authors released a number of controlled methane pulses, which were timed to coincide with Sentinel-2 (and other), satellite overpasses. The controlled nature of these releases means that they can be used as a development sandbox. Some of these point source releases were also small, so we can begin to explore the limits of detection.

In the first instance, GeoSynergy has implemented, and evaluated, the method reported by Ehret et al. where we compute a background image from a time series of plume-free Sentinel2 products [12]. This optimized background image is then subtracted from our plume-positive image during the plume enhancement process. This process can be combined with other image processioning modalities, to improve enhancement. The correct combination of techniques becomes particularly when the target emission approaches the limit of detection. These may include Gaussian filtering [12], co-registration of Sentinel-2 images [12, 13], and also image normalization [14].

The below figure shows our implementation of an improved plume enhancement workflow, using some of the former modalities, for a gas leak in Kazakhstan.

Figure 8: GeoSynergy enhancement of an emission from a natural gas pipeline in Kazakhstan, 21 September 2020. B12/B11 ratio residual shown after plume enhancement processing. The suggested origin and wind vector and shown. This is a large emission, or approx magnitude 62.8 t/h CH4.

We have also been testing our plume enhancement metholody with much smaller emissions, from the Sherwin et al. dataset [6]. For example, the below figure demonstrates our ability to detect a methane plume of just 1.4 t/h. This emission is close to the limits of detection using current methods (it was actually missed by some participants in the Sherwin et al. study [6]).

Figure 9: GeoSynergy enhancement of the 3 November 2021 controlled point source CH4 emission by Sherwin et al. [6] quantified at 1.4 t/h. This approaches the current limit of detection for Sentinel-2, using current modalities.

We are currently investigating whether Sentinel-2 can be used for the persistent monitoring of emissions from coal mine ventilation shafts. As previously discussed, these emissions are often below the threshold for TROPOMI detection (1–10 t/h), but also likely an important contributor to anthropogenic warming. In the 2020 paper by Varon et al. [7], the authors detected and quantified wind-rotated mean GHGSat products over coal mine vents. We have discussed wind rotation in the context of TROPOMI, but we suggest that it should also be considered for Sentinel-2, where mean products may be able to push the lower limit of detection.

Figure 10: Time-averaged methane plumes from the San Juan, Appin, and Bulianta coal mine vents, as observed by GHGSat-D from August 2016 through December 2018 [7]

Lastly, analogous to the discussion above Re. TROPOMI data, the detection of Sentinel-2 methane emissions may be automated using AI modalities, and this is an area of active research for GeoSynergy. Indeed, in the context of emissions approaching the limits of detection, AI modalities may be particularly significant.

References

[1] Sadavarte, P., Pandey, S., Maasakkers, J.D., Lorente, A., Borsdorff, T., Denier van der Gon, H., Houweling, S. and Aben, I., 2021. Methane emissions from superemitting coal mines in Australia quantified using TROPOMI satellite observations. Environmental Science & Technology, 55(24), pp.16573-16580.

[2] Schuit, B.J., Maasakkers, J.D., Bijl, P., Mahapatra, G., Van den Berg, A.W., Pandey, S., Lorente, A., Borsdorff, T., Houweling, S., Varon, D.J. and McKeever, J., 2023. Automated detection and monitoring of methane super-emitters using satellite data. Atmospheric Chemistry and Physics Discussions, pp.1-47.

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

[4] Maasakkers, J.D., Varon, D.J., Elfarsdóttir, A., McKeever, J., Jervis, D., Mahapatra, G., Pandey, S., Lorente, A., Borsdorff, T., Foorthuis, L.R. and Schuit, B.J., 2022. Using satellites to uncover large methane emissions from landfills. Science advances, 8(31), p.eabn9683.

[5] Ember, Abandoned Mine Methane. https://ember-climate.org/insights/research/tackling-australias-coal-mine-methane-problem/

[6] Sherwin, E.D., Rutherford, J.S., Chen, Y., Aminfard, S., Kort, E.A., Jackson, R.B. and Brandt, A.R., 2023.  Single-blind validation of space-based point-source detection and quantification of onshore methane emissions. Scientific Reports, 13(1), p.3836.

[7] Varon, D.J., Jacob, D.J., Jervis, D. and McKeever, J., 2020. Quantifying time-averaged methane emissions from individual coal mine vents with GHGSat-D satellite observations. Environmental Science & Technology, 54(16), pp.10246-10253.

[8] Gorroño, J., Varon, D.J., Irakulis-Loitxate, I. and Guanter, L., 2023. Understanding the potential of Sentinel-2 for monitoring methane point emissions. Atmospheric Measurement Techniques, 16(1), pp.89-107.

[9] Varon, D.J., Jervis, D., McKeever, J., Spence, I., Gains, D. and Jacob, D.J., 2021. High-frequency monitoring of anomalous methane point sources with multispectral Sentinel-2 satellite observations. Atmospheric Measurement Techniques,14(4), pp.2771-2785.

[10] Sánchez-García, E., Gorroño, J., Irakulis-Loitxate, I., Varon, D.J. and Guanter, L., 2022. Mapping methane plumes at very high spatial resolution with the WorldView-3 satellite. Atmospheric Measurement Techniques, 15(6), pp.1657-1674.

[11] Mayer, B. and Kylling, A., 2005. The libRadtran software package for radiative transfer calculations-description and examples of use. Atmospheric Chemistry and Physics, 5(7), pp.1855-1877.

[12] Ehret, T., De Truchis, A., Mazzolini, M., Morel, J.M., D’aspremont, A., Lauvaux, T., Duren, R., Cusworth, D. and Facciolo, G., 2022. Global tracking and quantification of oil and gas methane emissions from recurrent sentinel-2 imagery. Environmental science & technology, 56(14), pp.10517-10529.

[13] How to Co-Register Temporal Stacks of Satellite Images, EOResearch.

https://medium.com/sentinel-hub/how-to-co-register-temporal-stacks-of-satellite-images-5167713b3e0b

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