Direct and Indirect Measurements and Modeling of ... - ACS Publications

Direct and Indirect Measurements and Modeling of...

0 downloads 60 Views 604KB Size

This is an open access article published under an ACS AuthorChoice License, which permits copying and redistribution of the article or any adaptations for non-commercial purposes.


Direct and Indirect Measurements and Modeling of Methane Emissions in Indianapolis, Indiana Brian K. Lamb,*,† Maria O. L. Cambaliza,‡,○ Kenneth J. Davis,§ Steven L. Edburg,† Thomas W. Ferrara,∥ Cody Floerchinger,⊥ Alexie M. F. Heimburger,‡ Scott Herndon,⊥ Thomas Lauvaux,§ Tegan Lavoie,‡ David R. Lyon,# Natasha Miles,§ Kuldeep R. Prasad,∇ Scott Richardson,§ Joseph Robert Roscioli,⊥ Olivia E. Salmon,‡ Paul B. Shepson,‡ Brian H. Stirm,□ and James Whetstone∇ †

Laboratory for Atmospheric Research, Washington State University, Pullman, Washington 99164, United States Departments of Chemistry, and Earth, Atmospheric and Planetary Sciences, Purdue University, West Lafayette, Indiana 47907, United States § Department of Meteorology, The Pennsylvania State University, University Park, Pennsylvania 16802, United States ∥ GHD, Niagara Falls, New York 14304, United States ⊥ Aerodyne Research, Inc., Billerica, Massachusetts 01821, United States # Environmental Defense Fund, Austin, Texas 78701, United States ∇ National Institute for Standards and Technology, Gaithersburg, Maryland 20899, United States □ School of Aviation & Transportation Technology, Purdue University, West Lafayette, Indiana 47907, United States ‡

S Supporting Information *

ABSTRACT: This paper describes process-based estimation of CH4 emissions from sources in Indianapolis, IN and compares these with atmospheric inferences of whole city emissions. Emissions from the natural gas distribution system were estimated from measurements at metering and regulating stations and from pipeline leaks. Tracer methods and inverse plume modeling were used to estimate emissions from the major landfill and wastewater treatment plant. These direct source measurements informed the compilation of a methane emission inventory for the city equal to 29 Gg/yr (5% to 95% confidence limits, 15 to 54 Gg/yr). Emission estimates for the whole city based on an aircraft mass balance method and from inverse modeling of CH4 tower observations were 41 ± 12 Gg/yr and 81 ± 11 Gg/yr, respectively. Footprint modeling using 11 days of ethane/methane tower data indicated that landfills, wastewater treatment, wetlands, and other biological sources contribute 48% while natural gas usage and other fossil fuel sources contribute 52% of the city total. With the biogenic CH4 emissions omitted, the top-down estimates are 3.5−6.9 times the nonbiogenic city inventory. Mobile mapping of CH4 concentrations showed low level enhancement of CH4 throughout the city reflecting diffuse natural gas leakage and downstream usage as possible sources for the missing residual in the inventory.


To address this need, there has been renewed interest in determining how much CH4 is lost from each sector of the natural gas supply chain.4,5 Recent work has shown that estimates of CH4 emissions from top-down methods appear to be larger than CH4 inventories built from bottom-up emission factors (EF) and activity factors (AF),6−8 although a new comprehensive set of measurements and inventory development for the Barnett Shale basin yielded close agreement between top-down and bottom-up estimates.9 In urban distribution systems, this comparison is complicated by the

Natural gas consumption has increased by 40% from 1990 to 2014 in the U.S.1 This rapid increase has occurred due to the lower cost of natural gas resulting from the widespread introduction of horizontal drilling and hydraulic fracturing technologies. However, methane (CH4), the main component of natural gas, is a powerful short-lived greenhouse gas, and the emission of CH4 associated with the natural gas supply chain can offset the climate benefits of reduced CO2 emissions relative to other fossil fuels.2,3 Therefore, an accurate estimate of the CH4 lost to the atmosphere from the natural gas infrastructure and usage is needed to understand the climate impacts of using natural gas as an energy source and to identify viable opportunities for overall reductions in CH4 emissions. © XXXX American Chemical Society

Received: March 11, 2016 Revised: June 30, 2016 Accepted: July 19, 2016


DOI: 10.1021/acs.est.6b01198 Environ. Sci. Technol. XXXX, XXX, XXX−XXX


Environmental Science & Technology

leaks and high-flow sampling of individual components within M&R facilities,11 (2) ground-based mobile sampling for inverse plume modeling for specific sources, and (3) tracer ratio measurements for selected sources.11 The results from the gas distribution sampling were used along with data from the national sampling program to develop Indianapolis specific emission factors for use in the inventory. The results from inverse plume modeling, including tracer releases for calibration purposes, for the landfill, wastewater treatment facility, and compressor station were used to assess the uncertainty in inventory estimates for these sources. To obtain top-down estimates for the metropolitan area and for the Southside landfill, we used results from a series of aircraft mass balance flights methods,12 and tower-based ambient CH4 measurements and inverse-modeling for whole city emissions.13 The tower-based modeling was supplemented during February, 2015 with 11 days of sampling from a central city tower for methane and ethane levels. Inverse footprint modeling was completed using these data to estimate the relative contributions of biological, natural gas usage, and other sources. Because of the short duration of these measurements from a single tower, these contribution estimates can only be considered preliminary. Mobile mapping of ambient CH4 concentrations during repeated traverses throughout the city on 46 different days over several years was also conducted to help identify significant sources and to quantify CH4 enhancements above background levels. Details for each of these methods are described in the SI and results from the application of these methods are presented in the following sections.

presence of landfills, wastewater treatment plants, industrial sources, and natural gas emissions downstream of customer meters. The emission factors for recent urban inventories are largely based upon a comprehensive study conducted in the 1990s by the Gas Research Institute and EPA.10 The Environmental Defense Fund (EDF), with industry partners, recently sponsored a national sampling program to directly measure emissions from pipeline leaks and metering and regulating (M&R) stations in 13 local gas distribution companies (LDCs) across the U.S.11 During this national sampling study, measurements were conducted in Indianapolis, IN which is also the site for the Indianapolis Flux Experiment (INFLUX) ( The INFLUX study, sponsored by the National Institute of Standards and Technology (NIST), is focused on determining urban greenhouse gas emissions using atmospheric budget methods including both an aircraft mass balance approach12 and tower-based atmospheric inversions.13 Collaboration was established between the INFLUX program and the EDF distribution sampling study to address both topdown and bottom-up methods for measuring CH4 emissions. In this paper, we report results from this collaboration with an emphasis on compilation of bottom-up methods and comparison to top-down methods applied to the same sets of sources. The overall goal is to improve our understanding of CH4 emissions from urban areas and to address potential discrepancies between top-down and bottom-up emission estimates.

2.0. MATERIALS AND METHODS The metropolitan area of Indianapolis (see map, Supporting Information (SI) Figure S1) includes urban areas of Marion and parts of seven adjacent counties. For this work, we are focused on the higher density population areas (shaded in SI Figure S1) with an estimated population of 1.46 million in an area of approximately 1855 km2 (U.S. Census Bureau, http:// (see SI Section S1.0) Major CH4 sources include the Southside Landfill and two wastewater treatment plants. There is also a city steam plant and a large power plant along with a number of interstate highways that encircle and cross the city. For the natural gas system, there are two transmission distribution transfer stations (TDTS, also called city gates) where gas is transferred from the interstate gas transmission system to the local distribution system. The TDTS in the northwest corner of the city is located adjacent to a transmission pipeline compressor station. The local distribution system encompasses 6521 km of main pipelines and 269 134 services connecting customers to the main pipeline system14 (see SI Table S1). The pipeline system consists almost entirely of cathodically protected steel (49%) and plastic main pipelines (51%) with only 24 km of cast iron, while the services are 90% plastic. There are 36 M&R facilities within the city. There are also 11 compressed natural gas (CNG) refueling stations within the city. Natural gas consumption varies from 28 Gg/month in the summer to 140 Gg/month during the winter,1 although the system is always pressurized. We used a variety of data sources to compile a metropolitan area emission inventory and we used direct and indirect measurements for specific sources to either supply data for the inventory or to assess the uncertainty in current inventory estimates (see SI 5.0). The bottom-up emission measurement approaches used here for Indianapolis included (1) direct emission measurements using surface enclosures for pipeline

3.0. RESULTS AND DISCUSSION 3.1. Measured Methane Emissions from Pipeline Leaks and M&R facilities. Leak rates from a variety of distribution infrastructure components, including 14 underground pipeline leaks and 23 M&R stations, were measured in Indianapolis during the June 2013 field period. The pipeline leaks measured were randomly selected from an existing list of 115 leaks identified by the LDC during standard leak survey activities. For each leak, a surface enclosure was used with a high flow sampling unit11,15 to measure the leak rate (see SI S2.2). The measured leak rates ranged from 0.013 g/min to 22.3 g/min with a median value of 0.34 g/min and a mean value of 2.4 g/min. The total emission rate for these 14 leaks was 34.2 g/min. The distribution of leak rates was highly skewed with one-third of the leaks accounting for 95% of the total measured emissions. The range of leak rates was similar in Indianapolis to the range in the national sampling program (see SI Figure S12). However, the largest two leaks measured in Indianapolis were the second and third largest leaks measured among the national sample population and, as a result, the median emission rate for the Indianapolis measurements equaled 0.34 g/min vs a median of 0.06 g/min in the national data set. This difference between the Indianapolis and national data was taken into account via a Monte Carlo random sampling scheme16 to draw data from the Indianapolis and national sample populations to construct an inventory estimate for Indianapolis (Table 1 and see SI S5.0). For M&R facilities, the emissions were measured by first screening all components within a facility to identify leaks, measuring each leaking component with a high flow sampler,11,15 and then summing all the measured emissions within the facility for a total emission rate (see SI S2.1). The B

DOI: 10.1021/acs.est.6b01198 Environ. Sci. Technol. XXXX, XXX, XXX−XXX


Environmental Science & Technology

also released from the wastewater treatment facility and the Southside Landfill as a basis for testing the inverse plume modeling approach and using the model results to estimate emissions for these large sources (see SI Section S2.4). We conducted two tracer tests on a leak from a pipeline running underneath a 4 lane bridge in the city. This leak was not identified by the LDC survey records. The results, shown in SI Figure S2, yield a CH4 emission rate of 22.3 g/min. This is significant since it is the largest leak identified in Indianapolis, and it is the second largest leak in the overall national sampling leak database.11 The leak was not accessible and was not measured with the high flow enclosure system. We also conducted tracer tests on several days at an M&R station, and high flow sampling was conducted at the facility on two of these days, although not at the same time as the tracer tests. For this M&R station, the measurement data were used with a plume inversion model (SI S2.4) to estimate the location and strength of the tracer and methane sources. The inversion analysis predicted a tracer emission rate of 1.9 g/min, essentially equal to the actual release rate of 1.8 g/min. In turn, the plume inversion model applied to the CH 4 measurements predicted a CH4 emission rate of 49.9 g/min which is higher than that obtained from the high flow sampling (37.8 g/min), but the difference (24%) is within the experimental uncertainties of the two methods (see SI Sections S2.3 and S2.4). These measurements help confirm the use of emission factors for M&R stations from the national sampling study for the Indianapolis inventory (Table 1), but also highlight the temporal variability that can exist for these sources. Mobile measurements with multiple vehicles were also made around the TDTS and transmission compressor station in the northwestern corner of the city on three different days. The data showed the presence of a distinct plume that was resolved by the measurements on the various roads. Our plume modeling results gave a methane emission rate of 1260 g/ min from the TDTS and compressor station. The high flow sampling results for the distribution side of the TDTS yielded a very small emission rate of 3.7 g/min so that most of the emissions can be attributed to the compressor station. For comparison, the emission rate reported in the EPA GHG Reporting Program17 for the compressor station was 923 g/min which is within 26% of the inverse modeling result and within the experimental uncertainties of the inverse modeling method. For this source, the GHG Reporting Program results were used in the city inventory (Table 1 and SI 5.0). Nighttime mobile measurements were made with two vehicles downwind of the Southport Waste Water Treatment Plant (WWTP) on June 19−20, 2013. These measurements included CH4 as well as tracer in one of the vehicles. Tracer was released at a single point near the entrance to the plant for purposes of inverse model evaluation. This single tracer release point was not sufficient for typical tracer ratio applications. The predicted tracer release rate from the plume inversion model was 6.4 g/min, which compared very favorably with the actual release rate of 6.7 g/min. In turn, application of the inverse plume model for the CH4 data downwind of the WWTP produced a methane emission rate of 1720 g/min. This is a factor of 3 less than obtained from the GHG Reporting Program17 which reported an estimated emission rate of 5,644 g/min. Details of the operational status of the WWTP were not available, but changes in plant operations and throughput could affect the emission rate.17−19 In this case, we used the GHG

Table 1. Summary of Indianapolis Methane Emission Inventorya indianapolis inventory sectors pipeline leaks (emission factors derived from direct leak measurements in Indianapolis and nationally) M&R (national emission factors) meters (EPA GHG emission factors) uncombusted (assumed 0.3% inefficiency) dig-ins/blow-downs (LDC data with national GHG emission factors) transmission and storage (GHG Reporting program) CNG fleets (assume 2% loss of fuel use) major point sources (GHG Reporting Program) power plants (air permits) Vehicles (EPA NEI 2011) biological sources (see SI 3.6) landfills (GHG reporting program) wastewater (GHG reporting program) livestock (see SI 3.6) wetlands Total


upper confidence limit Gg/yr

































0.487 0.594

0.730 0.891

1.7% 2.1%

8.0% 9.7%



% of inventory

% of nonbiogenic emissions





1.31 0.43 28.9

3.93 1.30 53.8

4.5% 1.5% 100.0%



Further details are given in SI section 3.6. Uncertainty limits: estimated as the 95% confidence limit for pipeline leaks, M&R, dig-ins, blow-downs; estimated as a factor of 3 for meters, combustion, and transmission and storage and enteric fermentation; and estimated as a factor of 2 for point sources, power plants, vehicles, and landfills. (see SI S3.3).

emissions per facility ranged from 0.003 g/min to 25 g/min, and three stations had no detectable emissions. The median emission rate was 0.033 g/min, the mean emission rate was 1.37 g/min and the total emissions measured for all M&R facilities equaled 34.5 g/min. The largest emitting facility was an M&R station with an inlet pressure of approximately 500 psi and with 9 pneumatic devices. Two TDTS facilities were measured, but in both cases only the distribution side of the facility was accessible. As described in the following section, inverse plume modeling was used at a site with a TDTS and a transmission compressor station to obtain an emission rate for the entire complex (1260 g/min., see below). For M&R stations and for the compressor station, the measured emission rates were similar to those from the national sampling study and so the emission factors for M&R stations in Indianapolis were taken from the national sampling results (Table 1 and see also SI 5.0). 3.2. Tracer Ratio and Plume Measurements with Inverse Plume Modeling. During the June 2013 field study, tracer ratio experiments were conducted at several M&R facilities and underground pipeline leaks. Tracer was released at each facility, and tracer and CH4 concentrations were measured with an instrumented van along downwind traverses. The CH4 emission rate was calculated using the measured tracer release rate and the ratio of CH4/tracer concentrations.11 Tracer was C

DOI: 10.1021/acs.est.6b01198 Environ. Sci. Technol. XXXX, XXX, XXX−XXX


Environmental Science & Technology

Figure 1. City emission estimates for individual aircraft flight mass balance measurements and sequential 5-day period tower inversion estimates using two different a priori estimates. Closed box shows mean and 5−95% confidence limits for aircraft, tower, and city inventory estimates.

Reporting Program emissions, but set the uncertainty to a factor of 3 (Table 1 and SI 5.0). Nighttime mobile measurements were conducted downwind of the large Southside Landfill on June 21, 2013 under SSE winds at approximately 3 m/s. Although tracer was released during these tests, the available roadways for transecting the plume were too far downwind and the tracer concentrations were too near the detection limit to be used for inverse modeling. However, the CH4 plume was clearly detected in two vehicles operating downwind of the landfill. CH4 enhancements were found to be as high as 30 ppm above the background value during nighttime conditions. For the landfill, the inversion plume analysis predicted a CH4 emission rate of 30.3 kg/min (15.9 Gg/yr) based on the mobile data. As described below, emission estimates were also obtained from the aircraft mass balance flights (14.5 ± 7.2 Gg/yr) and from the EPA GHG Reporting Program17 (15.0 Gg/yr). The plume inversion estimate is essentially equal to both the aircraft and GHG estimates. For the inventory, we used the GHG Reporting Program estimates (Table 1 and SI 5.0). 3.3. Whole City and Landfill Emissions from Aircraft Mass Balance Methods. The data from five aircraft flights during 2011 for whole-city and landfill emission estimates have been presented by Cambaliza et al.12 We have recalculated the emissions for these five flights and nine more recent flights in 2014 using a different method for assigning background levels (see SI 3.1), and we have incorporated data from an initial set of 8 flights in 2008−200920 as shown in Figure 1. The range of results, shown in Figure 1, yields a mean of 41.3 Gg/yr for the whole city total, with 95% confidence limits of ±10.6 Gg/yr (see SI S3.3 for details regarding the confidence limit calculations). The aircraft flights occurred during three different periods and included both winter and nonwinter seasons as shown in Table 2. For three of the periods, the mean emission rate was similar, but for the winter flights in 2008−2009, the mean emission rate from three flights was much higher. In all periods, there was considerable variability in the city emission estimate from flight to flight. This variability reflects the uncertainties in

Table 2. Comparison of City Inventory with Aircraft Mass Balance and Tower Based Inverse Modeling Emission Estimates inventory total biological emissions total emissions from fossil fuel sources total city emissions aircraft mass balance city emissions March−April, 2008 November−January 2008−09 April−July, 2011 November−December, 2014 all flights tower based city emissions nonwinter, 2012−2013 winter, 2012−2013 all tower periods source contributions from ethane/methane inverse modeling* biological sources (48%) natural gas usage (43%) other sources (9%) a

CH4 Gg/yr

95% upper confidence limita CH4 Gg/yr

22.8 6.1

38.0 15.8

28.9 CH4 Gg/yr

84.5 76.9 80.9 CH4 Gg/yr using aircraft results

53.8 95% confidence limit CH4 Gg/yr 10.7−45.4 51.8−129.8 23.3−52.6 17.7 to 50.3 30.7−51.9 95% confidence limits CH4 Gg/ yr 75.4 to 93.7 64.8 to 88.9 73.5 to 88.2 CH4 Gg/yr using tower results

19.8 17.8 3.7

38.8 34.8 7.6

28.1 90.8 37.9 34.0 41.3 CH4 Gg/yr

See SI S3.4.

the aircraft mass balance method, and it may also be related to different source footprints captured during each flight. For all of the flights, there were CH4 enhancements observed in the downwind horizontal transects. Cambaliza et al.12 showed that a portion of the enhancements could be attributed to the Southside Landfill and that the mean CH4 contribution from the landfill to the citywide emission rate was estimated to be ∼35%. Analyses of the 2014 flights showed a similar contribution of the landfill to the city total. Landfill CH4 D

DOI: 10.1021/acs.est.6b01198 Environ. Sci. Technol. XXXX, XXX, XXX−XXX


Environmental Science & Technology emissions are known to be sensitive to changes in ambient atmospheric pressure21,22 as well as seasonal temperatures, and this may explain some of the variability shown in Figure 1. 3.4. Inverse Modeling Using Ambient Methane Observations From the INFLUX Tower Observing Network. For this work, a CH4 inversion system13 assimilated atmospheric CH4 mixing ratios, measured continuously from September, 2012 through April, 2013 at tower sites 1, 2, and 10 ( and SI Figure S1 and S3.2). The results from the inversions are presented in Figure 1 in comparison to the aircraft results. Two a priori cases were calculated. The first used an initial city emission rate of 68 Gg/yr (based on the 2011 aircraft flight results), and the second used half this rate. In both cases, the spatial distribution of the a priori was based on point sources estimates for the landfill and WWTP, and the remainder was distributed proportional to the HESTIA CO2 emissions from natural gas combustion.23 For these two different inversion model initializations, the posterior whole-city emission rates (81.9 and 79.8 Gg/yr) agreed to within a few percent (see also SI Table S2). These results for two very different a priori estimates indicate that the inversion is robust with respect to the initialization process. The inverse modeling results also show that during the nonwinter periods (September−October, March−April) the overall emission rate is 84.5 ± 9.2 Gg CH4/yr while in the winter (November−February) the mean emission rate is 76.9 ± 12.0 Gg/yr. Thus, the inverse modeling results also suggest that emissions are not significantly different in winter than in fall and spring. There is still considerable variability from 1 week to the next, and this variability is due in part to the limited number of towers (3) generating different tower footprints over the city for the different 5-day periods. During any specific period, the wind direction changes may not capture emissions from all of the city. 3.5. Ethane/Methane Measurements and Source Attribution. To help determine the relative contributions of natural gas usage and biological sources, we measured ambient ethane (C2H6), CH4, CO, and CO2 concentrations from tower 11 located in north central Indianapolis from February 26 through March 9, 2015 (see SI S3.4). We also obtained natural gas composition data from the LDC (see SI Table S3) which showed a large range in the molar ratio of C2H6/CH4 (1.8− 5.8%, mean = 3.5%) over an 8 month period, but the ratio was equal to 4.2% for the period during the tower sampling. As shown in Figure 2 (see also SI Figure S8), the ethane/methane correlations exhibited considerable variability with characteristic ratios of several distinct plume events. An analysis of these events (see SI S3.4 and Table S4) showed mainly two cases. With winds from the southwest, very low C2/C1 ratios were measured reflecting biological (landfill) sources. With northwest winds, C2/C1 ratios of 4.6% reflected natural gas emissions from the transmission compressor station located in the northwestern part of the city. Overall, the ratio was 3.65%. Simple mixing models (seeSI S3.4 and Table S6) do not account for the heterogeneous contributions of large point sources, such as the landfill or wastewater treatment plant. To address this deficiency, we conducted inverse footprint calculations (SI S3.4) using the tower data and obtained a source distribution of 48% biological, 43% natural gas usage, and 9% other urban fossil fuel sources. Given the short duration of the ethane/methane sampling period and the use of a single tower, it is difficult to assign uncertainties to these source distributions. The estimate of 48% for biological sources is

Figure 2. Ambient ethane vs methane concentrations (with background subtracted) measured at Tower 11 during February 26 through March 9, 2015. The ratios represented by the three tails are (1) natural gas usage, 4.6%, (2) combustion or mixed sources, 1.7%, and (3) biological landfill and wastewater treatment