Decadal decrease in Los Angeles methane emissions is much smaller than bottom-up estimates

Methane, a powerful greenhouse gas, has a short atmospheric lifetime ( ~ 12 years), so that emissions reductions will have a rapid impact on climate forcing. In megacities such as Los Angeles (LA), natural gas (NG) leakage is the primary atmospheric methane source. The magnitudes and trends of fugitive NG emissions are largely unknown and need to be quantified to verify compliance with emission reduction targets. Here we use atmospheric remote sensing data to show that, in contrast to the observed global increase in methane emissions, LA area emissions decreased during 2011-2020 at a mean rate of (–1.57 ± 0.41) %/yr. However, the NG utility calculations indicate a much larger negative emissions trend of −5.8 %/yr. The large difference between top-down and bottom-up trends reflects the uncertainties in estimating the achieved emissions reductions. Actions taken in LA can be a blueprint for COP28 and future efforts to reduce methane emissions.


Supplementary Text 2: Background estimation for CLARS-FTS
The details of the background calibration are introduced in He et al. 2 .To derive an unbiased background XCH4 and XCO2 along the same path of the CLARS target mode, we combined the CLARS Spectralon retrievals and NOAA in situ flask dataset at Mt. Wilson (https://gml.noaa.gov/dv/site/site.php?code=MWO).The NOAA in situ flask dataset gives the background estimate using in situ flask-based sampling at Mt. Wilson next to the CLARS facility.
At night, the height of the boundary layer falls to far below the CLARS facility and the flask record is very likely to represent background conditions for the lower troposphere over the region, where there are no human activities.Therefore, to construct the background, we used the Spectralon measurements as the background for the atmosphere above the CLARS height, and the NOAA flask measurements at night as the background for the atmosphere below.Supplementary Figure 1 shows a comparison of the time series of CLARS observations for the surface target mode and the Spectralon mode, and the NOAA nighttime flask monthly averaged data.

Supplementary Figure S2. A comparison of CO2 bottom-up emission inventories from Hestia
(red; 2011-2015), CARB (green; 2011-2019), and ODIAC (blue; 2011-2019).CARB inventory values for the LA basin are scaled from the state's total emissions using the ratio of SOCAB population (14.6 million) to total California population (39 million).ODIAC is shifted upward by 3.5 TgCO2/month to match the Hestia annual estimate.After adjusting to the Hestia level, 10% uncertainty is assumed for these monthly estimates (similar to the uncertainty estimate of Hestia by Gurney et al. 5 ).For CO2 emissions in 2020, we used the 2019 value as the baseline and applied scale factors from Yadav et al. 6 to derive the drawdown of CO2 emissions in LA due to the COVID-19 pandemic lockdown.This extrapolation is applied to ODIAC and CARB inventories.
See text for details.Map data ©2019 Google observations.The color bar indicates the observation density, which is the number of measurement pairs.From the linear regression, as shown in the red dashed line, the intercept = 5.75 ppb and the slope = 6.38 ppb/ppm.
. NOAA in-situ nighttime and daytime flask measurements on Mt.Wilson for CH4 and CO2. .(a) scatter plot between CO2 excess and CH4 excess from NOAA MWO flask measurements from 2011 to 2020.Only data with positive excesses are used.The correlation coefficient is 0.82; (b) The monthly mean (in red) of excess ratio from NOAA MWO flask measurements from 2011 to 2020.The error bars represent the estimation uncertainty (one standard error) of the monthly values.We filtered the data by using CO2 excess larger than 5 ppm to exclude days with air not transported from the LA basin.
(ppb/ppm) Supplementary Figure S9.Correlation between natural gas consumption and CH4 emissions estimated using the (a) ODIAC and (b) CARB CO2 bottom-up inventories, respectively, and biogenic flux correction based on Miller et al. 4 .The error bars represent the estimation uncertainty (1s) of the monthly values.
based on Miller et al.4 .The error bars represent the estimation uncertainty (1s) of the monthly values.0.64±0.18Gg/month Supplementary FigureS12.Histograms of slopes from Monte Carlo simulations for uncertainty estimations of the emissions trend as shown in Figure3of the main text.The uncertainties for the slope in each case are estimated using the Monte Carlo method, which samples the monthly emissions using a normal distribution based on the mean and error and estimates of the slope.The method makes 10,000 simulations for the emissions time series and obtains the standard deviation of the slope samplings.in the basin, including western, central and eastern regions of the Los Angeles Basin.The background image in the upper panel is adopted from the Map data ©2019 Google.The error bars represent the estimation uncertainty (1s) of the monthly values.