Source Apportionment of Ambient Methane Enhancements in Los

Article pubs.acs.org/est

Cite This: Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Source Apportionment of Ambient Methane Enhancements in Los Angeles, California, To Evaluate Emission Inventory Estimates Toshihiro Kuwayama,*,† Jessica G. Charrier-Klobas,† Yanju Chen,† Nicholas M. Vizenor,‡ Donald R. Blake,‡ Thomas Pongetti,§ Stephen A. Conley,∥ Stanley P. Sander,§ Bart Croes,† and Jorn D. Herner† †

California Air Resources Board, 1001 I Street, Sacramento, California 95812, United States University of California at Irvine, 570 Rowland Hall, Irvine, California 92697, United States § NASA Jet Propulsion Laboratory, 4800 Oak Grove Drive, Pasadena, California 91109, United States ∥ Scientific Aviation, 3335 Airport Road Suite B, Boulder, Colorado 80301, United States Environ. Sci. Technol. Downloaded from pubs.acs.org by WEBSTER UNIV on 03/02/19. For personal use only.

‡

S Supporting Information *

ABSTRACT: Rapid increase in atmospheric methane (CH4) mixing ratios over the past century is attributable to the intensification of human activities. Information on spatially explicit source contributions is needed to develop efficient and cost-effective CH4 emission reduction and mitigation strategies to addresses near-term climate change. This study collected long-term ambient CH4 measurements at Mount Wilson Observatory (MWO) in Los Angeles, California, to estimate the annual CH4 emissions from the portion of Los Angeles County that is within the South Coast Air Basin (SCLA). The measurement-based CH4 emission estimates for SCLA ranged from 3.95 to 4.89 million metric tons (MMT) carbon dioxide equivalent (CO2e) per year between 2012 and 2016. Source apportionment of CH4, CO, CO2, and volatile organic compounds (VOCs) measurements were used to evaluate source categories that contributed to ambient CH4 mixing ratio enhancements (ΔCH4) at SCLA between 2014 and 2016. Results suggested ΔCH4 contributions of 56−79% from natural gas sources, 7−31% from landfills, and 4−15% from transportation sources. The SCLA-specific CH4 emission estimate made using a research grade gridded CH4 emission inventory suggested contributions of 47% from natural gas sources and 50% from landfills. Subsequent airborne measurements determined that CH4 emissions from two major CH4 sources in SCLA were significantly smaller in magnitude than previously thought. This study highlights the importance of studying the variabilities of CH4 emissions across California for policy makers and stakeholders alike.

■

INTRODUCTION Global Warming Solutions Act of 2006 (Assembly Bill 32 or AB 32) set a target for California to reduce greenhouse gas (GHG) emissions to 1990 levels by 2020, a reduction in emissions of approximately 30% below business-as-usual.1 CH4 is one of the seven major GHGs that are included in AB 32 and is classified as a short-lived GHG with an atmospheric lifetime of 12.4 years, with 100-year and 20-year Global Warming Potential (GWP) of 28 and 84, respectively.2 CH4 currently accounts for 9% of the California’s Statewide GHG Emission Inventory3 and is emitted primarily from agriculture, waste management, and the oil and gas (O&G) industry. Other sources of CH4 include, but are not limited to, transportation, electricity generation, commercial, and residential sources. The Statewide GHG Emission Inventory developed by the California Air Resources Board (CARB) is used to track GHG emissions in California and aids in the development of GHG emissions reduction and mitigation © XXXX American Chemical Society

strategies. CARB also manages the Statewide GHG Monitoring Network to track the changes in ambient CH4 mixing ratios (and other GHGs) throughout the state.4 Various other measurement and modeling techniques have been used to study the regional ΔCH4 and CH4 emission sources in California.5−8 Effective evaluation of CH4 emission reduction efforts requires accurate characterization of major CH4 sources, which includes comprehensive understanding of source distribution, persistence, and magnitude. Since CH4 emissions can vary significantly from source to source and region to region, it becomes imperative that spatially explicit GHG emissions and mitigation challenges are evaluated throughout the state.5,9−12 Received: October 5, 2018 Revised: February 16, 2019 Accepted: February 19, 2019

A

DOI: 10.1021/acs.est.8b02307 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology

derived from SCLA. One-hour samples were collected seven consecutive days each month, four times a day, every two hours, starting at 11 AM using a 32-channel autosampler (ATEC). Midnight samples were also collected to approximate the incoming background mixing ratios in the lower free troposphere that subsided during the evening hours. This number was used to calculate the daily mixing ratio enhancements observed at MWO. The list of sampling days and times is available in Table S1 of the Supporting Information. The whole-air canister samples were quantified using three gas chromatography systems (HP 6890, Agilent Technologies) that integrated electron-capture detectors, flame ionization detectors, and a quadrupole mass spectrometer. Similar sample collection and analysis methods were used during a previous canister-based air sampling campaign conducted at MWO in 2007.16,17 Analytical accuracy was within 5%, with precision confident within 9% for all hydrocarbons. Of the 74 compounds measured during the campaign, 21 were utilized for source apportionment analyses using the PMF model 5.0 based on their chemical relevance and detection thresholds (Table S2 in the Supporting Information). The PMF model is a multivariate receptor model that can be used to quantitatively determine the contribution of air pollution concentrations or mixing ratios in a data matrix through identification of mathematically unique signatures, or chemical profiles. The fundamental approach used in PMF is mass balance described as

The current study evaluates the regional CH4 emissions from SCLA using long-term ambient CH4 and carbon monoxide (CO) measurements collected between 2012 and 2016 at MWO. In addition, inventory-based CH4 emission estimates for SCLA were developed by scaling the statewide CH4 emission inventory using surrogate variables and by evaluating spatially comprehensive, gridded CH4 emission inventory constructed for research purposes. The comparison of measurement-based and inventory-based CH4 emission estimates was performed as a follow-up to a previous study that resulted in the refinement of the Statewide GHG Emission Inventory through reallocation of CH4 emissions between various sources.13 Furthermore, the sources in the CH4 emission inventory were evaluated against the results from a source apportionment model, U.S. Environmental Protection Agency (U.S. EPA) PMF model version 5.0, that used a suite of CH4, CO, CO2, and VOC measurements collected between 2014 and 2016 at MWO. The model-resolved chemical factor profiles were assigned to various sources that exist in SCLA: landfills, petroleum gas and industrial processes, cold-start fossil fuel combustion, petroleum refining residuals, hotrunning fossil fuel combustion, natural gas, and vegetation (or biogenic sources). This study is the first to compare multiyear CH4 source apportionment results to a SCLAspecific CH4 emission inventory. Additional airborne CH4 measurements were used to evaluate two major CH4 sources in SCLA to support the source apportionment results. This manuscript will help refine future CH4 emission inventories and develop efficient and cost-effective CH4 emission reduction and mitigation strategies to meet California’s GHG reduction goals.

p

xij =

∑ gikfkj k=1

■

+ eij

(1)

where x is input data matrix (i.e., ambient measurements) with i number of samples and j number of chemical compounds or species, p is the total number of factors (i.e., sources) contributing to the input data matrix (assigned by the user), g is the contribution of each factor, f is the chemical profile (i.e., source signature) of each factor, and e is the residuals of the input data matrix that were not resolved within the model run. In this study, the optimum number of resolvable factors was determined by successive PMF model runs that evaluated the convergence indicators, rotational ambiguity, and model residuals according to the PMF guidelines provided by U.S. EPA.18 Bootstrapping and displacement tests were used to assess the robustness of the results by ensuring that >80% of the factors were appropriately mapped during the error estimation. The converging run with the lowest Q value was selected for further analysis. The Q value is an object function used to describe the performance of the model based on e and the uncertainty matrix of the input data, u: Ä É ÅÅ xij − ∑ p g f ÑÑÑ n m Å ÅÅ k = 1 ik kj Ñ ÑÑ Q = ∑ ∑ ÅÅÅ ÑÑ ÅÅ ÑÑ u ij i=1 j=1 Å ÑÖ (2) Ç

METHODS CH4, CO, and other gaseous air pollutants were measured at MWO located on top of the San Gabriel mountain range (34°13′21′′ N, 118°3′42′′ W) that overlooks the Los Angeles Basin. Its unique geological location and meteorological conditions allow measurement of well-mixed urban air pollution that typically travels toward the monitoring site from regions between southeast (SE) and west-southwest (WSW).13−15 The upslope air flow, rapid atmospheric mixing from the growth of boundary layer height, and relatively consistent meteorological pattern between 10 AM and 6 PM in SCLA makes this monitoring site ideal for long-term measurements of urban emissions.13,16 CH4 and CO were continuously measured using Cavity Ring-Down Spectroscopy Model G1301/G2201-i (Picarro, Inc.) and Off-Axis Integrated Cavity Output Spectroscopy Model 913-0015 (Los Gatos Research, Inc.), respectively, from May 2012 to December 2016. The measurements from these high-precision, fastresponse analyzers were averaged over each hour and were corrected using drift check gases (Scott-Marrin, Inc.) measured five times a day to develop quantitative confidence in the small deviations of atmospheric CH4 and CO mixing ratios observed over multiple years. The two drift check gases consisted of CH4 and CO mixing ratios above and below typical ambient levels observed at MWO (1.8 and 3.0 ppm for CH4 and 100 and 500 ppb for CO). These gases were also evaluated using calibration gases certified by National Oceanic and Atmospheric Administration’s Earth System Research Laboratory (NOAA ESRL). Two-liter whole-air canister samples were collected at MWO between July 2014 and April 2016 to measure VOCs that were

The variability in the Q value during the displacement test was evaluated to determine if p should be reduced prior to further analyses.18 Constrained model runs were not performed due to lack of time specific activities data for each of the resolvable factors. The source assignment for the resolved factors relied on the evaluation of potential sources affecting the measurements at the receptor site. The results were limited by the user’s interpretation of g and f. Additional guidelines and quality control strategies were also followed.19−22 B

DOI: 10.1021/acs.est.8b02307 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology

Figure 1. IB2 CH4 emissions map for SCLA. Colors in each of the 4 × 4 km grid cells represent CH4 emissions in MMT CO2e/year. Each grid cell represents the sum of all CH4 emissions from all sources identified in the IB2 approach.

approach and was used for further comparison with the measurement-based source apportionment results. Geospatial and activity data was not available on a year-by-year basis; therefore, the IB2 approach assumed that the spatial distribution of CH4 sources and its emissions remained relatively consistent over the study period. The IB2 approach identified annual SCLA-specific CH4 emission estimates of 3.96, 3.96, 4.02, 4.04, and 4.07 MMT CO2e/year from 2012 to 2016, respectively. These values compared within 4−6% of the CH4 emission estimates using the IB1 approach, which was considered marginal given the uncertainty of the underlying data. Unlike IB1, the IB2 approach utilized CARB’s 2017 Edition Statewide GHG Emission Inventory based on availability of the data. This inventory only provided information from 2000 to 2015; therefore, the data was linearly extrapolated for the 2016 CH4 emission estimate. Comparison between the two editions of the Statewide GHG Emission Inventory resulted in minor differences. Figure 1 shows IB2 CH4 emissions aggregated in 4 × 4 km grid cells within SCLA. Detailed information on both IB1 and IB2 approaches is available in the Supporting Information. Measurement-Based CH4 Emission Estimation. CO emissions primarily derive from mobile sources in SCLA and are spatially well-distributed throughout the region. Due to this homogeneity, CO had been used as a surrogate to characterize the well-mixed air mass that traveled to MWO from the urban lowlands.13 This study utilized multiple linear regression models to describe the statistical relationships between ambient CH4 and CO mixing ratios (CH4:CO) as a metric to calculate CH4 emissions that derived from SCLA. Monthly CH4:CO orthogonal regression slopes were calculated using the data from real-time analyzers and whole-air canister samples collected from 2012 to 2016 and 2014 to 2016, respectively. The data sets were analyzed independently to increase confidence in the results. The robustness of the monthly linear regression models was investigated through repetitive randomized-subsampling tests. The subsampling

Airborne measurements of CH4 mixing ratios were conducted by Scientific Aviation in the fourth quarter of 2017 and the second quarter of 2018.23,24 The research aircraft was equipped with flight-ready Cavity Ring-Down Spectroscopy Model G2401-m (Picarro, Inc.) along with threedimensional Global Positioning System (GPS) and meteorological sensors. The research aircraft conducted three-dimensional spirals around targeted CH4 sources throughout the atmospheric boundary layer for integrated mass balance calculations to determine area-wide CH4 emission fluxes.25

■

RESULTS

Inventory-Based CH4 Emission Estimation. SCLAspecific inventory-based CH4 emissions were estimated using two different approaches to gain confidence in the results. The first approach relied on a comprehensive scaling of CARB’s Statewide CH4 Emission Inventory using surrogates such as population, vehicle-miles-traveled, and land-use as a strategy to apportion sectoral emissions to subregions within the State.26−28 Based on the first inventory-based (IB1) approach, CH4 contributed to approximately 4.10, 4.18, 4.20, 4.23, and 4.25 MMT CO2e/year in SCLA from 2012 to 2016, respectively. One unit of MMT CO2e/year is equivalent to 40 gigagrams (Gg) CH4/year assuming GWP of 25. Note that CARB’s 2018 Edition Statewide GHG Emission Inventory was used as a basis of this calculation.3 The second inventory-based (IB2) approach utilized a more involved gridded CH4 emission inventory that was developed for research purposes using CH4 emission factors from CARB’s Statewide GHG Emission Inventory29 integrated with geospatial and activity data retrieved from CARB, CalRecycle, Division of Oil, Gas, and Geothermal Resources (DOGGR), State Water Board, U.S. Geological Survey (USGS), University of California (UC), U.S. Energy Information Administration (EIA), and U.S. Department of Agriculture (USDA). The IB2 approach provided additional source sector- and regionspecific information that was more precise than the IB1 C

DOI: 10.1021/acs.est.8b02307 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology

Figure 2. Comparison of annual SCLA-specific inventory- and measurement-based CH4 emission estimates from 2012 to 2016. IB1 was calculated through comprehensive scaling of the Statewide CH4 Emission Inventory; IB2 was calculated using the gridded CH4 emission inventory developed for research purposes; MB1 was calculated using the real-time measurements; and MB2 was calculated using whole-air canister sample data. The box-whiskers represent the 25th quartile (darker box), the 75th quartile (lighter box), and the minima and the maxima (whiskers) of the monthly CH4 emission estimates aggregated over each year. The black dots represent annual averages.

time analyzers (MB1) resulted in annual SCLA-specific CH4 emission estimates of 4.89, 4.74, 4.56, 4.39, and 4.07 MMT CO2e/year from 2012 to 2016, respectively. The second measurement-based approach using whole-air canister samples (MB2) resulted in CH4 emission estimates of 4.45, 4.15, and 3.95 MMT CO2e/year from 2014 to 2016, respectively. Outliers identified during the Aliso Canyon CH4 leak incident in late 2015 were removed from this analysis.33,34 The accuracy of this calculation relies on the quality of the CO Emission Inventory and the characteristic representativeness of the SCLA air described by the perceived relationships between CH4 and CO. Assumptions used in the calculation of CH4 emission estimate are available in the Supporting Information. Comparison between Inventory- and MeasurementBased CH4 Emission Estimates. Figure 2 compares the annual SCLA-specific inventory- and measurement-based CH4 emission estimates described in the previous sections. The annual CH4 emission estimates had standard deviation that ranged from 0.5 to 0.9 MMT CO2e/year for MB1 and 0.7 to 1.0 MMT CO2e/year for MB2. In addition to the seasonality, higher variabilities seen in MB2 estimates resulted from the smaller number of daily samples used to calculate the monthly regressions. Comparison of all annual CH4 emission estimates (IB1, IB2, MB1, and MB2) showed good agreement with annual averages that fell within 10 ± 7% of each other, an improvement since the study by Hsu et al. in 2007/2008.13 The deviation is similar to the lower end of the range reported by Wong et al. for the Los Angeles region (IB estimates lower than MB estimates by 18−61%).35 Measurement-based CH4 emission estimates showed a decreasing trend between 2012 and 2016, which was consistent between both MB1 and MB2. Despite drought conditions, the annual reduction of approximately 0.2 ± 0.1 MMT CO2e in MB1 signified changing source characteristics that were not captured by the inventory-based CH4 emission estimates. The gap between MB and IB annual average CH4 emission estimates gradually improved from 2012 to 2016, with the largest difference of approximately 23% in 2012 and smallest difference of less than 1% in 2016. PMF Factor Profiles. Figure 3 represents the PMF factor profiles that were derived from modeling the whole-air canister sample data set with 21 chemical compounds and over 350

threshold criterion was described by the coefficient of variation, [cv = σ/μ] < 15%, where σ is the standard deviation and μ is the average of the monthly regression slopes generated through subsampling. Sample sets that satisfied the threshold criterion were used in further analyses. Statistical assessment of the ambient data is available in the Supporting Information. The coefficient of determination, R2, of monthly CH4:CO that passed the threshold criterion was 0.83 ± 0.08 for data from real-time analyzers and 0.84 ± 0.08 for data from whole-air canister samples (Figures S1 and S2 in the Supporting Information). The monthly CH4:CO orthogonal regression slopes were aggregated into annual averages to estimate the SCLA-specific measurement-based CH4 emissions using the following equation: ji MWC H4 zyz = Ṁ CO(MMT/year) × RS × jjj z × (GWPCH4 ) j MWCO zz k {

CH4(MMT CO2 e/year)

(3)

where Ṁ CO is the total mass of CO emitted into the atmosphere within SCLA in MMT per year, RS is the annual average CH4:CO orthogonal regression slope, MWCH4 is the molecular weight of CH4 in gram per mole, and GWPCH4 is the 100-year GWP of CH4. In this analysis, 100-year GWP of 25 was used to keep consistency with CARB’s Statewide GHG Emission Inventory methodology and current international GHG inventory practices.30 The inventory-based Ṁ CO is derived from CARB’s Emission Inventory that describes the estimated amount of air pollutants emitted into the atmosphere from specific source categories in certain geographical area.31 In this particular manuscript, Ṁ CO from Los Angeles County was split into South Coast Air Basin, which excluded the contributions from the Mojave Desert to be consistent with the analysis that utilized time-and-meteorology constrained data from MWO that targeted sources located in the southern side of the San Gabriel Mountain range. Documentation on CARB’s Emission Inventory is available online.32 CO emissions used to estimate the CH4 emissions ranged from 970 to 1279 tons CO/day between 2012 and 2016. The first measurement-based approach using the realD

DOI: 10.1021/acs.est.8b02307 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology

Figure 3. PMF factor profiles produced from whole-air canister samples collected at MWO. Profiles were assigned as (a) Factor 1 − Landfills, (b) Factor 2 − Petroleum gas and industrial processes, (c) Factor 3 − Cold-start fossil f uel combustion, (d) Factor 4 − Petroleum refining residuals, (e) Factor 5 − Hot-running fossil f uel combustion, (f) Factor 6 − Natural gas, and (g) Factor 7 − Biogenic.

the suspect for CH4 and CO2 contributions in this factor. However, the molar ratio between CH4 and CO2 digressed away from landfill gas (LFG) characteristics, and the contributions from CO2 and aromatics were larger than expected. This may indicate influences from associated power generation and on-site landfill activities (e.g., off-road engines, garbage truck traffic, waste incineration). Factor 2 was assigned as petroleum gas and industrial processes based on the heavy contributions from alkanes, specifically propane and butane,

data points that passed quality control checks. Factor 1 was assigned as landfills based on CH4, CO2, ethylbenzene, and m/ p- and o-xylene as well as mix of other VOCs that are typically derived from urban waste.36−41 This factor is independent from petroleum related natural gas sources since there are no strong links to lighter alkanes (C2−C5) such as ethane, consistent with findings by Peischl et al. (Figure S3 in the Supporting Information).5 Anaerobic decomposition of organic waste within the complex landfill environment was E

DOI: 10.1021/acs.est.8b02307 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology Table 1. Average Monthly ΔCH4 and PMF Source Contributions from July 2014 to April 2016 month-year Jul-14 Aug-14 Sep-14 Oct-14 Nov-14 Dec-14 Jan-15 Feb-15 Mar-15 Apr-15 May-15 Jun-15 Jul-15 Aug-15 Sep-15 Oct-15a Nov-15a Dec-15a Jan-16a Feb-16a Mar-16 Apr-16

Factor 1

Factor 2

Factor 3

Factor 5

Factor 6

Factor 7

average monthly ΔCH4 (ppb)

landfills

petroleum gas and industrial processes

FFCS

FFHR

natural gas

biogenic

± ± ± ± ± ± ± ± ±

34 33 45 43 38 22 18 69 31

± ± ± ± ± ± ± ± ± ± ± ±

20 61 38 37 53 97 55 43 47 30 35 26

17% 19% 23% 19% 26% 25% 31% 23% 21% 17% 25% 13% 7% 11% 9% 16% 11% 15% 11% 13% 22%