A declining trend of methane emissions in the Los Angeles basin from 2015 to 2020

Even though there have been several published methane (CH₄) emission estimates for the Los Angeles (LA) basin using various methods, vast differences exist between bottom-up, and top-down CH₄ budgets. Two of these estimates are based on the bottom-up temporally invariant inventory for a single year, i.e. 2012 (e.g. Jeong et al 2013, Maasakkers et al 2016), and the remaining are the efforts that use CH₄ concentration measurements from in-situ sites (aka top-down), aircraft flights, or remote sensing instruments (e.g. Wunchst et al 2009, Hsu et al 2010, Wennberg et al 2012, Peischl et al 2013, Santoni et al 2014, Wecht et al 2014, Cui et al 2015, Jeong et al 2016, Wong et al 2016, Wunch et al 2016). The combined mean CH₄ emissions estimate of these studies is ∼43 000 kg h⁻¹, with a lower bound of ∼16 000 kg h⁻¹ and an upper bound of ∼70 000 kg h⁻¹ (figure 1). Overall, it has been estimated that CH₄ emissions of ∼43 000 kg h⁻¹ in the LA basin account for 20% of all CH₄ emissions in California (Jeong et al 2016). Note that the spatial domains of all the studies mentioned in figure 1 are inconsistent. For example, the estimates provided by Hsu et al (2010) only apply to LA County and not to the entire spatial domain of this study.

Zoom In Zoom Out Reset image size

Temporal trends in the estimates of CH₄ emissions across multiple studies remain indiscernible, especially when we account for uncertainty in emissions. This also applies to studies performed over a longer duration (e.g. Wong et al 2016, Wunch et al 2016). Conditionally, this requires a thorough assessment of the temporal change and uncertainty through longitudinal observations obtained over multiple years from either remote sensing or in-situ platforms. Out of these two platforms, only the latter provides high-fidelity measurements with increased frequency to detect relatively small temporal changes in the variability of CH₄ concentrations and emissions, albeit with reduced spatial coverage.

In this work, we assess temporal trends in the CH₄ emissions in the LA basin over five and half years. We use two methods to determine and confirm the declining trend in CH₄ emissions. We think a comprehensive approach to assessment, as presented in this work, is necessary for studies that focus on analyzing temporal change as they form the foundation for studying the impact of policies formulated for reducing greenhouse gas emissions.

Our objectives in this study are to (a) assess the temporal change in the CH₄ emissions in the LA basin and (b) identify regions in the LA basin that show a reduction in the CH₄ emissions.

Our study area in figure 2 spans the LA, Orange, San Bernardino and Riverside counties in California. This work uses the designation 'LA basin' to identify this area.

Zoom In Zoom Out Reset image size

We perform our inversions at 0.03° spatial resolution (1826 grid cells) and four-day temporal resolution for the period spanning from 27 January 2015 to 3 June 2020. Our emissions assessment is confined to the area constrained by the observations highlighted by the yellow outline in figure 2. The basis for this demarcation is described in Yadav et al (2019). The period for obtaining emissions estimates in this study was determined by the simultaneous availability of the measurements and the Jacobian from a coupled Weather Research Forecasting -Stochastic Time Inverted Lagrangian Model. The output from the WRF model for the LA basin was validated against data obtained from the Aircraft Communications, Addressing, and Reporting System and 42 surface observation sites data (see Yadav et al 2019 for details).

The location of the in-situ sites measuring surface CH₄ concentrations (see Verhulst et al 2017 for discussion on the measurement network and the precision and accuracy of measurements) used in this study is shown in figure 2, and the temporal duration of the measurements with data gaps is shown in figure S1. CH₄ enhancements (shown at monthly temporal resolution in figure S2) in this study are computed by subtracting the CH₄ concentrations at the SCI site from those at the remaining sites using the methodology described in Verhulst et al (2017). This background site is located on San Clemente Island, which is outside the spatial domain of this study. Hourly temporal variability (standard deviation) of the CH₄ concentrations and enhancements between 12:00 noon and 4:00 pm from 27 January 2015 until 22 April 2022 is used to assess trends in the temporal variability of CH₄ concentrations. For inverse estimation of the CH₄ emissions, we only use enhancements from 27 January 2015 to 3 June 2020 due to the reasons mentioned above.

We conducted a two-step analysis to fulfill the objectives of this study. First, we analyze the trend in the monthly temporal variability of the measured atmospheric CH₄ concentrations rather than the concentrations themselves to detect the decline in CH₄ emissions. In the second step, we quantify the magnitude of the reduction in CH₄ emissions and confirm the evidence obtained from the previous mode of analysis through inverse modeling. The first step in our assessment is an exploratory approach to detect the change, and the second confirmatory approach relies on modeling to quantify emissions resulting from this change.

3.1. First step analysis: analysis of CH₄ concentrations

3.2. Second step analysis: inverse estimation of emissions

4.1. Assessment of the CH₄ concentrations

We plot the mean monthly time series of the standard deviation of the CH₄ concentrations in figure 3. The slope (units: parts per billion per month) of the trendlines computed after removing the data gaps for all time series is <0, indicating a reduction in CH₄ emissions. However, compared to other sites, the pace of the decrease in the temporal variability was considerably higher for GRA and USC sites (see slopes in figure 3). For example, at the GRA site over 87 months (February 2015 to March 2022), this leads to a reduction in the temporal variability by 12.18 (−0.14 × 87 = −12.18; see equation in figure 3) parts per billion (ppb), which is about ∼1/2 of the peak temporal variability for 2021 (the last complete year for which data is available).

Zoom In Zoom Out Reset image size

4.2. Results: multiyear mean trend of CH₄ emissions and uncertainties

To reduce the short-term noise and for understanding the long-term trend in the mean CH₄ emissions, we plot seven four-day moving average of emissions (7 × 4 = 28 days) and uncertainties in figure 4. We chose this time period for averaging as it closely corresponds to the monthly duration of the assessment of the temporal variability of CH₄ concentrations. However, all our results are based on unsmoothed time series of grid-scale emissions (submitted as a data record with this manuscript).

Zoom In Zoom Out Reset image size

The trend in the four-day mean emissions confirms what was observed by assessing the temporal variability of CH₄ concentrations. The average CH₄ emissions declined by ∼2100 kg h⁻¹ from February 2015 to May 2020 (figure 4). We can also partition the mean basin-scale emissions by STAD regions. Most of the basin is covered by the STADs of the ONT, FUL, GRA, and CIT sites. This applies both with respect to the area (figure 2) as well as prior emissions (figure S3).

Assessment of emissions by STADs shows that most of the reduction in emissions happened in the STADs of ONT and GRA (note the slope of the trendline in figure 5). Other STADs, except for these two, do not show any substantial changes in CH₄ emissions. Due to the more extensive coverage of the STAD of ONT and GRA, the reduction of emissions in these regions had more significant impact on basin-scale CH₄ emissions.

Zoom In Zoom Out Reset image size

Figure 6 shows that the uncertainty on the estimated emissions declined over the time period of inversions; this is similar to the trend in the temporal variability of concentrations. This is consistent with a reduction in the seasonal amplitude (peak emissions minus lowest emissions in a year) of CH₄ emissions, which can also be seen in figure 3 (see panel titled 'All Towers'), figures 4, and 5 for the STAD of ONT and GRA sites.

Zoom In Zoom Out Reset image size

4.3. Results: reduction in emissions for the LA basin at aggregated scales

We only report aggregated results for five years, from 2015 to 2019. Furthermore, as our inversions start from February 2015, we use the mean of the ratio of January emissions to the emissions for the remaining months in the years 2016–2019 to compute total CH₄ emissions for January 2015.

Emissions at the annual scale declined in the basin from 2015. The difference between emissions for 2015 (272 ± 7.79 Gg; 1σ bound) and for years 2018 (244 ± 4.49 Gg; 1σ bound) and 2019 (251 ± 4.83 Gg; 1σ bound) was significant after accounting for >1σ uncertainty (figure 7). The emissions for 2018 and 2019 were significantly reduced even with respect to the emissions for 2017 (265 ± 5.03 Gg; 1σ bound). Emissions for 2018 were lower than those for 2019, though they were not outside each other's uncertainty bounds. However, emissions for 2018 were outside 2σ bounds with respect to 2016. For comparison, if we consider 2016 to be the baseline year as the last quarter of 2015 is influenced by the Aliso Canyon gas leak (duration: 18 October 2015 to 18 February 2016) then emissions of 2018 and 2019 were 10% and 7% lower, respectively. Note that emissions from the Aliso Canyon gas leak declined substantially from the start of the blowout on 23 October 2015, until 31 December 2015 (Conley et al 2016) and therefore, a comparison of 2018 and 2019 CH₄ emissions with emissions for 2016 is unlikely to introduce any significant bias in the evaluation.

Zoom In Zoom Out Reset image size

Assessment of the uncertainties themselves leads to an insight into the detectability limits of the network. Thus, with respect to 2016, a 7% annual reduction in the CH₄ emissions in 2019 remains detectable at 1σ, whereas ∼10% reduction in 2018 was detectable at the 2σ level. In terms of magnitude, this difference of 10% corresponds to ∼800 kg h⁻¹ or 11 Gg of CH₄ emissions per year. To reiterate, a detection of 800 kg h⁻¹ is applicable at an annual scale and not at an hourly temporal resolution.

At individual sites, we find declining emissions in the STADs of the GRA and ONT sites, but the rate of this decline is slowly plateauing (figure 8). These regions account for ∼45% of the emissions in the basin. The STADs for all other sites did not yield any significant trend in reduction (figure S6).

Zoom In Zoom Out Reset image size

With respect to the 1σ uncertainty bound in figure 8, the emissions from STAD for GRA and ONT for 2018 and 2019 were significantly different from the baseline emissions for 2016. If we disregard 2015 as uncommon due to the Aliso Canyon gas leak, then differences between 2015 emissions from February to December and those for 2019 for ONT were significant after accounting for 2σ uncertainty bound on emissions.

4.3.1. Discussion: inverse estimates of emissions

The time series of the average emissions in figures 4 and 5 show a declining trend. However, if we look at the uncertainty bounds, then we cannot claim that the mean estimate of emissions for any four days is any different from emissions for other periods, and this is also applicable for the STAD regions. This implies that there has been no reduction in the four-day mean emissions estimate. However, average estimates of the emissions have the property of removing the impact of point sources. In comparison, aggregated emissions, as reported in figures 7 and 8, avoid this problem and provide a complete picture of the trend in emissions and their uncertainty, albeit at a lower temporal resolution. The CH₄ emissions from the STAD for GRA and ONT significantly differed from the 2016 baseline emissions. Spatially, this is leading to a redistribution of CH₄ emissions, whereby the overall contribution of the LA basin emissions from the STADs of GRA and ONT is declining, leading to an increase in the contribution from other STAD regions.

It can be argued that reductions in CH₄ emissions are just an outcome of data gaps, atmospheric transport, or interannual variability. However, we do not think this is the case, as the trend in the suppression of the seasonal amplitude of the time series of mean emissions until May 2020 has been persistent for three years. This is also corroborated by the reduction in the temporal variability of CH₄ concentrations, as shown in figure 3.

A higher bar consisting of multiple modes of analysis is required to confirm reductions in emissions from atmospheric observations, as the question is not only about estimating emissions but also about assessing the impact of policies implemented to reduce emissions. In this work, we show through our rigorous two-fold analysis that CH₄ emissions in the LA basin are declining. However, the persistence of this trend has to be continuously monitored beyond the temporal duration of this study.

The inverse estimate of emissions and an exploratory analysis of the temporal variability showed that the seasonal amplitude of emissions has declined in the basin, leading to a reduction in annual CH₄ emissions. However, identifying individual sources of these reductions from inversions with the existing measurement network is not directly possible.

Reasonable pointers to the sources that led to the emissions reductions can be obtained by analyzing (a) the nature of the sources of emissions and (b) the regions where this reduction has happened. Indeed, a consistent reduction in the seasonal amplitude of the temporal variability in concentration and emissions can most likely only be achieved by reducing fugitive emissions, a source of which is natural gas infrastructure in the LA basin (He et al 2019). Spatially, most of the reduction in CH₄ emissions happened in the STAD regions of GRA and ONT that are dominated by CH₄ emitting infrastructure consisting of landfills, dairies, and wastewater treatment plants. The GRA STAD covers the Sunshine Canyon Landfill, the largest landfill in the LA basin, and its emissions have declined due to improved management practices (see Cusworth et al 2020). We can disregard dairies as potential sources of emissions reduction, as they are only responsible for minimal or zero CH₄ emissions in the LA basin (see Hsu et al 2010). We can also disregard wastewater treatment plants as sources of emissions reduction as they do not have a seasonal cycle of emissions whose amplitude can be reduced (see Daelman et al 2012 for variations in the CH₄ emissions from a wastewater treatment plant). As a result, landfills and natural gas infrastructure are the most likely plausible sources of emissions reductions in the LA basin after 2015.

Finally, this study also shows that the existing network of surface measurements of CH₄ concentrations in the LA basin can detect a 7%–10% reduction in CH₄ emissions on an annual scale. However, identifying significant changes (outside the uncertainty bounds) in the mean emissions estimate will require assessment over a decade or more. We think that the availability of more observations could reduce this period; however, we do not explore this in the current study. In the near term, the research should focus on identifying individual facilities (e.g. landfills) responsible for CH₄ emissions. This can help policymakers improve their oversight of the mitigation efforts undertaken by facilities to reduce CH₄ emissions.

© Jet Propulsion Laboratory, California Institute of Technology.

A portion of the research described in this paper was carried out at the Jet Propulsion Laboratory, California Institute of Technology, under contract 80NM0018D0004 with the National Aeronautics and Space Administration (NASA). Additional support was provided by the National Institute of Standards and Technology (NIST) Greenhouse Gas and Climate Science Measurements program. The authors also acknowledge support from NASA's Carbon Monitoring System program and the Prototype Methane Monitoring System for California project and Multi-tiered Carbon Monitoring System project. Certain commercial equipment, instruments, or materials are identified in this paper to specify the experimental procedure adequately. Such identification is not intended to imply recommendation or endorsement by the National Institute of Standards and Technology, nor is it intended to imply that the materials or equipment identified are necessarily the best available for the purpose.

Methane concentration data utilized in this study is already available from https://data.nist.gov/od/id/mds2-2388. This concentration data has also been submitted as part of this manuscript. Output of Weather Research and Forecasting and Stochastic Time-Inverted Lagrangian Transport Model is ∼50 terabytes in size and researchers interested in obtaining this data should provide authors a data repository to upload the model output.

The authors declare no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.