Comparison of Field Measurements to Methane Emissions Models at

Jul 25, 2016 - In 2012, U.S. landfills were estimated to emit 103 Tg (million metric tons) CO2 equivalents (CO2e), making landfills the third and four...
0 downloads 14 Views 2MB Size
Article pubs.acs.org/est

Comparison of Field Measurements to Methane Emissions Models at a New Landfill Florentino B. De la Cruz,*,† Roger B. Green,‡ Gary R. Hater,‡ Jeffrey P. Chanton,§ Eben D. Thoma,∥ Tierney A. Harvey,⊥ and Morton A. Barlaz† †

Department of Civil, Construction, and Environmental Engineering, Campus Box 7908, North Carolina State University, Raleigh, North Carolina 27695-7908, United States ‡ Waste Management, Inc., 2956 Montana Avenue, Cincinnati, Ohio 45211, United States § Department of Earth, Ocean, and Atmospheric Science, Florida State University, Tallahassee, Florida 32306, United States ∥ U.S. EPA, Office of Research and Development, National Risk Management Research Laboratory, Research Triangle Park, North Carolina 27711, United States ⊥ Department of Engineering and Physics, University of Central Oklahoma, Edmond, Oklahoma 73034, United States S Supporting Information *

ABSTRACT: Estimates of methane emissions from landfills rely primarily on models due to both technical and economic limitations. While models are easy to implement, there is uncertainty due to the use of parameters that are difficult to validate. The objective of this research was to compare modeled emissions using several greenhouse gas (GHG) emissions reporting protocols including: (1) Intergovernmental Panel on Climate Change (IPCC); (2) U.S. Environmental Protection Agency Greenhouse Gas Reporting Program (EPA GHGRP); (3) California Air Resources Board (CARB); and (4) Solid Waste Industry for Climate Solutions (SWICS), with measured emissions data collected over three calendar years from a young landfill with no gas collection system. By working with whole landfill measurements of fugitive methane emissions and methane oxidation, the collection efficiency could be set to zero, thus eliminating one source of parameter uncertainty. The models consistently overestimated annual methane emissions by a factor ranging from 4−31. Varying input parameters over reasonable ranges reduced this range to 1.3−8. Waste age at the studied landfill was less than four years and the results suggest the need for measurements at additional landfills to evaluate the accuracy of the tested models to young landfills.



INTRODUCTION Landfills are the dominant disposal alternative for municipal solid waste (MSW) in the U.S. Approximately 54% of the 251 million metric tons of MSW generated in the U.S. was estimated to be landfilled in 2012.1 When organic matter is buried in a landfill, it is converted to short chain fatty acids and ultimately to CH4 and CO2 in a series of biochemical reactions.2 Methane represents a source of energy and when recovered, methane can offset the use of nonrenewable fossil fuels. In 2014, an estimated 645 landfill methane recovery projects were in operation in the U.S., and the U.S. Environmental Protection Agency (EPA) estimates that energy could be recovered at an additional 440 landfills.3 While landfills represent an energy source, they also represent a source of greenhouse gas (GHG) emissions as some methane is released prior to gas collection system installation, and some methane is not captured due to imperfect gas collection. In 2012, U.S. landfills were estimated to emit 103 Tg (million metric tons) CO2 equivalents (CO2e), making landfills the © XXXX American Chemical Society

third and fourth largest sources of anthropogenic methane in the U.S. and globally, respectively.4,5 With the exception of the California Landfill Methane Inventory Model (CALMIM),6 which models methane emissions directly, landfill methane emissions models estimate emissions by reducing methane production by the fraction of the methane that is collected and the fraction of uncollected methane that is oxidized in the landfill cover soil. This approach can lead to considerable error due to uncertainty in methane production, the time-varying collection efficiency and the fraction oxidized. Estimates of methane oxidation vary widely. For many years, the U.S. EPA GHG Inventory assumed that 10% of uncollected methane was oxidized, a value that was recently updated to Received: January 26, 2016 Revised: May 23, 2016 Accepted: July 25, 2016

A

DOI: 10.1021/acs.est.6b00415 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology

Figure 1. Estimation of methane production rate and oxidation rate from measured emissions and oxidation. The x-axis represents the year from start of waste burial. (A) Measured methane emissions; (B) Calculated methane production using eq 9, where both emissions and oxidation data are available; (C) Measured methane oxidation factor (OX) and mass of methane oxidized. Outlier OX at around year 3.5 is not presented. Dashed lines in A and B represent 95% prediction interval band.

between 10 and 35%, depending on the site-specific flux of methane passing through the soil cover.4,7,8 Chanton et al.9 have summarized a number of field measurements of methane oxidation and reported a mean of approximately 37% oxidation of the methane passing through a soil cover. A number of studies have measured methane emissions directly. Some work has been conducted using static chambers10 leading to questions on whether the spatial coverage of the chambers is representative. In other studies, optical remote sensing has been used to measure whole landfill methane emissions.11−13 Some of these measurements have been made in concert with data on gas collection, making it possible to estimate a collection efficiency, while in other cases, the measured data were not sufficient to calculate a collection efficiency.14 Collection efficiencies have been reported to vary from approximately 20 to 90%, a range that is not surprising as the collection efficiency will vary with the landfill cover type and well density, both of which vary temporally.10−17 The collection efficiency will increase as more of the landfill is under gas collection and with installation of an intermediate or final cover. There is, however, uncertainty associated with whole landfill emissions measurements due to both methodological issues and the fact that measurements tend to occur over short periods of time.18 The concept of a temporally averaged collection efficiency over the entire decomposition period has been introduced.19,20 While this approach is useful in life-cycle studies where the functional unit is typically 1 Mg of MSW as

disposed, this approach is not applicable to annual emissions estimates where the focus is on an entire landfill containing wastes of varying ages. The objective of this research was to compare modeled emissions using several GHG emissions reporting protocols to measured data from a landfill without a gas collection system in place. By working with data from a landfill with no gas collection system, the collection efficiency is zero, thus eliminating one source of model input parameter uncertainty. The study site is described in the following section, followed by a description of each of the models and the assumed default parameters. Measured methane oxidation and emissions are then presented and compared to model predictions.



MODELING APPROACH AND DATA ANALYSIS Site Description. The studied landfill is located in the Southeastern U.S. in a region that recorded average annual precipitation of 1176 mm (2010−2014 data for Carrolton, GA)21 and a mean temperature of 4 and 26 °C in January and July, respectively. The landfill has a capacity of 26 million metric tons (1 Mg = 1 t) and started receiving 264−330 thousand Mg of waste per year in 2010. The waste footprint and total waste in place was 145, 282 m2 and 1.16 million Mg in December 2013, respectively. The waste deposited in the landfill is estimated to consist of 84−100% MSW, 0.2−4.4% construction and demolition (C&D) waste, and 0−11.5% inert materials. B

DOI: 10.1021/acs.est.6b00415 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology Table 1. Input Parameters Used in Different Models to Estimate GHG Emissions As Applied to Study Landfill GHG reporting protocol EPA GHGRP bulkc

EPA GHGRP modified bulkc MSW C&D EPA GHGRP compositionc food waste garden waste paper wood and straw textiles diapers sewage sludge IPCC bulkd

IPCC compositiond food waste garden waste paper wood and straw textiles diapers sewage sludge CARBe SWICS a

k

(yr−1)a

DOC (Mg C/Mg wet MSW or component)a

DOCFa

L0 (m3 CH4/Mg wet MSW or component)b

0.057

0.2

0.5

0.057 0.04

0.31 0.08

0.15 0.2 0.4 0.43 0.24 0.24 0.05

0.5 0.5 0.5 0.5 0.5 0.5 0.5

0.09

0.19

0.5

0.15 0.2 0.4 0.43 0.24 0.24 0.05 0.19 0.19

comment

93

flux dependent

k, DOC and DOCF derived from Table HH-1 (Table 1). OX is flux dependent and adapted from (table HH-4 of 40 CFR 98; Table S1)7,8

143

flux dependent

as for EPA GHGRP bulk

79

flux dependent

waste composition given in Table 2

89

0.1

IPCC North America default (Tier 2)27 as the country-specific parameters are already reflected in the EPA GHGRP.

79

0.1

waste composition given in Table 2

100 100

0.1 0.35

bulk MSW DOC and DOCF calculated from composition

0.5 0.5

0.185 0.1 0.06 0.03 0.06 0.1 0.185

0.185 0.1 0.06 0.03 0.06 0.1 0.185 0.057 0.057

OXa

0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.54 0.54

k, Decay rate; DOC, degradable organic carbon; DOCF, fraction of DOC dissimilated; OX, oxidation factor. bFor bulk option: ⎡ m 3CH ⎤ ⎡ Mg C ⎤ ⎛ 1mole ⎞ ⎛ 106g ⎞ ⎛ 22.4L ⎞ ⎛ 1m 3 ⎞ 4 ⎟ × ⎜ ⎟⎟ × ⎜ ⎥ = DOC⎢ ⎥ × DOCF × F × ⎜ L 0⎢ ⎟ × ⎜⎜ ⎟ ⎣ Mg MSW ⎦ ⎝ 12g C ⎠ ⎝ 1Mg ⎠ ⎝ 1mole ⎠ ⎝ 1000L ⎠ ⎣ Mg ⎦

⎡ m3CH4 ⎤ ⎡ Mg component ⎤ × Wi⎢⎣ Mg wet MSW ⎥⎦. cFOD parameters adapted from refs 7 and 8. dFOD parameters For composition/multiphase option: L0 ∑ L0i⎢ Mg component ⎣ ⎦⎥ adapted from ref 27. eFOD parameters adapted from ref 29.

CH4(emission)t = a × t b

Field-Measured Methane Emissions, CH4 oxidation and CH4 Mass Balance. Whole-facility methane emissions were measured by mobile tracer correlation (TC) using a vehicle-mounted cavity ring down spectrometer. In TC, both methane and a metered release of acetylene tracer (for this study) are measured in a path around the landfill perpendicular to the plume direction.11,18,22,23 Data processing has been described previously.18 Briefly, the TC data were assessed and data rejection was based on the following criteria: (1) visual inspection; (2) signal-to-noise ratio (SNR) < 3; (3) tracer not released or fully developed; (4) incomplete transect; or (5) transect executed within 500 m of the release centroid for landfill applications. The measurement data are presented in Table S1 of the Supporting Information (SI). To compare TC measurements to model predictions, it was necessary to integrate the daily emissions measurements over a year (Figure 1). First, a nonlinear model was fitted to the emissions data as shown in eq 1:

(1)

where CH4(emission)t (Mg/yr) is the measured instantaneous CH4 emission at any time t; a and b are fitting parameters; and t (yr) is the time of initial waste burial. eq 1 was integrated to calculate cumulative annual emissions as is required for comparison to the emissions models (eq 2): CH4(emission)t =

a × t b+1 +c b+1

(2)

where c is the integration constant. Nonlinear curve fitting and data analysis were done using the nonlinear least-squares (nls) function of R.24 The prediction interval was estimated by the Delta method.25 A given function g(X) with estimator X, and true value μ, can be estimated by using Taylor series approximation truncated up to the second term: g (X ) ≈ g (μ) + ∇g (μ)T ·(X − μ) C

(3)

DOI: 10.1021/acs.est.6b00415 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology where ∇g(μ)T is the transpose of the gradient vector. The variance was calculated using eq 4. T

var(g (X )) ≈ ∇g (μ) ·cov(X ) ·∇g (μ)

OX T = 1 −

(4)

∑ (y − y ̅ )2 n−2

(5)

where ŷ0 is the predicted emissions at (x0, y0); tcrit is the critical t distribution value at α = 0.05 and degrees of freedom (df) = n − 2; y ̅ is the mean of the measured emissions. An estimate of the range of annual emissions was calculated as the integral over a year of the lower and upper 95% prediction interval estimates of emissions. The fraction of CH4 production from the landfill that is oxidized (oxidation factor − OX) was measured using a stable carbon isotope approach and the data used for this analysis are presented in SI Table S2. These data represent the average of 3−8 replicate samples collected sequentially. Briefly, gaseous plume samples were collected 1−2 km downwind to measure fugitive CH4. Methane composition and δ13C were measured in the samples and were corrected for the background CH4 to estimate OX.26 Both CH4 emissions and OX were measured between June 2011 and November 2013 during which time a landfill gas collection system was not in place. Thus, methane emissions were controlled by methane production and oxidation during this period. Measurements were conducted when landfill personnel and equipment were available and weather conditions were favorable. These measurements may not completely reflect seasonal variations in emissions and oxidation. Since both methane emissions and methane oxidation were measured, methane production could be calculated as described by eq 6 to eq 9. An estimate of CH4 production, CH4(production), was calculated by mass balance using measured emissions and daily OX data as follows with all units in Mg/yr:

(6)

paper and paperboard glass metals plastics rubber and leather textiles wood other food waste yard trimmings misc. inorganic waste

14.8 5.1 8.8 17.9 3.8 6.8 8.4 2.0 21.3 8.8 2.4

0.4 0 0 0 0 0.24 0.43 0.24 0.15 0.2 0

The amount of C&D waste was given in the site description and the studied landfill did not receive biosolids. Neither C&D waste nor biosolids are included in the U.S. EPA’s definition of MSW, so the emissions associated with the disposal of these components were only considered in the composition options of the EPA GHGRP and IPCC models. US EPA Greenhouse Gas Reporting Program (EPA GHGRP). Mandatory reporting of GHG emissions from landfills started in 2010.7 The protocol specifies that methane production be calculated using equation HH-1 of the rule as in eq 11.7,8

(8)

CH4(emission) (1 − OX)

DOCb (Mg/Mg wet MSW)

(7)

Substituting eq 8 into eq 7 results in CH4(production) =

composition (% wet basis)

This composition does not include C&D waste, industrial waste and certain other wastes that are disposed in landfills.28 bDOC, degradable organic carbon. Data adapted from ref 27.

and, CH4(oxidation) = OX × CH4(production)

material

a

Without a gas collection system, CH4(collection) = 0, thus, CH4(production) = CH4(oxidation) + CH4(emission)

(10)

Table 2. U. S. Average Composition for Discarded Waste in 2009.a

CH4(production) = CH4(collection) + CH4(oxidation) + CH4(emission)

CH4(production)T

where CH4(emission)T and CH4(production)T are the CH4 emissions and production for year T with all values in Mg/yr. Methane Emissions Models. Seven implementations from four different models were used to estimate CH4 emissions as summarized in Table 1. As shown in Table 1, some models were implemented with more than one set of inputs. For example, the EPA GHGRP and IPCC models allow for the use of either bulk MSW as one material with one set of properties, or for a user-specified waste composition with waste component-specific properties, and both model implementations were explored. In all cases, methane production is described by a first order decay, and the different model formulations are mathematically equivalent.27 Each model was used as intended to compare emissions predictions based on the model guidance for default parameters. All models assume that waste is deposited at the beginning of the year with an average lag of six months and that the landfill is strictly anaerobic so that all decomposition occurs under anaerobic conditions. Each GHG reporting protocol is described in this section. Site-specific waste composition for the study landfill was not available. Therefore, the 2009 U.S. national average waste composition was used whenever a protocol allowed for an input waste composition (Table 2).28

Applying eq 4 to emissions data, the prediction interval, y95%PI about point (x0, y0) was calculated: y95%PI = y0̂ ± tcrit var(g (X )) +

CH4(emission)T

(9)

Both CH4 emissions and OX were directly measured and used to estimate methane production. Note that the OX in eq 8 and (9) represents daily measured OX as a fraction rather than the annual fraction oxidized (OX) which is an input parameter in the emissions models as described in the following section. Finally, the weighted average OX over a year (OXT), which is also needed as a model input, was calculated from eq 10:

T−1

∑ Wx × MCF × DOC × DOCF × F ×

CH4(production) =

x=s

× (e D

−k(T − x − 1)

−e−k(T − x))

16 12 (11)

DOI: 10.1021/acs.est.6b00415 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology

fraction of organic carbon dissimilated for component, i, respectively. The default OX in the CARB method is 0.1, which is from the IPCC default for well-managed landfills (Table 1). The CARB protocol uses the 2007 U.S. EPA national average MSW composition for 2008 onward (SI Table S4). Solid Waste Industry for Climate Solutions (SWICS). An FOD calculation was also implemented in the Solid Waste Industry for Climate Solutions (SWICS) protocol with default parameters presented in Table 1. SWICS is similar to CARB except that OX is calculated as the weighted average for n cover types and m coverage as shown in eq 14.

where CH4(production) = modeled methane generation in reporting year T (Mg CH4); x = year in which waste was disposed (yr); k = decay rate (yr−1); S = start of calculation (yr); T = reporting year to which emissions are calculated (yr); Wx = quantity of waste disposed in year x (Mg wet weight); MCF = methane correction factor that accounts for portions of landfill that are aerobic; DOC = degradable organic carbon (Mg C/Mg wet waste); DOCF = fraction of DOC dissimilated (i.e., the fraction of carbon that is mineralized to CO2 and CH4); F = fraction of CH4 in landfill gas; 16/12 = the molecular weight ratio that converts a mass of C to a mass of CH4. For this study, the GHGRP protocol was implemented three ways based on the waste classification: (1) bulk waste; (2) modified bulk waste; (3) composition waste. The parameters used for each model implementation are presented in Table 1. In the bulk waste option, the waste is assumed to be decomposing as a single stream, and k is a weighted average of the decomposition of different components. In the modified bulk option, MSW and C&D waste are modeled as two waste streams. In the composition option, values for k and DOC are specified in the GHGRP for each waste component, and CH4 production from each component is calculated. As discussed above, CH4(emission) is equivalent to CH4 (production) adjusted for methane oxidation, for a landfill with no gas collection. The OX used in model simulations was assigned based on the fluxdependent regulatory guidance given in Table S3,7,8 where methane flux (MF) was calculated from the modeled methane production (eq 11) divided by the surface area of the landfill at the beginning of the reporting year. In all cases, the owner reported that the landfill had at least 24″ of cover soil over the majority of the landfill so that OX is dependent only on flux and not on other conditions in the GHG reporting rule (Table S3).7,8 This methane oxidation schedule was followed for all implementations of the GHGRP. IPCC Methodology. The IPCC methodology27 follows three tiers to estimate emissions from solid waste disposal systems (SWDSs) with the opportunity to use country− or site−specific data. For this study, two implementations of the IPCC model were analyzed, including (1) bulk MSW and (2) waste composition (Table 1). The k, DOC and DOCF are the same as for the EPA GHGRP, but methane oxidation is different. Under the composition option, the waste categories in the U.S. were reassigned to match the categories in IPCC.27 In all cases, OX was assumed to equal 0.1 for a well-managed landfill.27 California Air Resources Board (CARB). The State of California implemented greenhouse gas reporting as part of its state GHG inventory.29 The CARB approach uses bulk first order decay (FOD) calculations but the DOC of bulk MSW (DOCMSW) was derived from waste composition data with n components, as in eq 12.

m

OX =

∑ WIPi × DOCi i

dDOCm(t ) = (D(t − Δ) − k DDOCm(t ))dt

CH4(production)(i) = q × [a′ × DDOCma(i) − b′ × DDOCmd(i − 1) + c′ × DDOCmd(i)]

(16)

where q = MCF × F × 16/12 a′ = 1 − e − k

b′ =

(12)

1 −k(1 −Δ) (e − e −k ) − Δ × e −k k

c′ = 1 − Δ −

n

× DOCi

(15)

where dDOCm = change in DDOCm at time t; D(t−Δ) = disposal rate function at which the disposed carbon remains inert for a period (lag time, Δ) before decomposition commences; DDOCm(t) = DDOCm available at time t for decay. The solution to eq 15 has been derived previously for methane production for year i.27,30

1 (1 − e−k(1 −Δ)) k

The Δ was varied from 0−12 months, and then the modeled and measured emissions were compared. In this calculation, parameters for bulk MSW are the same as the parameters used in the EPA GHGRP protocol (k = 0.057 yr−1, DOC = 0.2, DOCF = 0.5, OX = 0.1).

∑i WIPi × DOC × DOCFi n ∑i WIPi

(14)

i

where AFi = the fraction of the landfill area covered by cover j (daily, intermediate, final and biocover); OXi = fraction of methane oxidized in cover type i (organic = 0.38; clay = 0.22; sand = 0.55; other = 0.30; and unknown = 0.35). The default OX for the SWICS model is 0.35 and was used in this case due to insufficient information on soil types and the extent of daily and intermediate cover. Effect of Lag Time. The assumption that waste is deposited at one time at the beginning of the year (instantaneous deposition) implies that the waste would have an average lag time of approximately 6 months.27 This assumption does not reflect the actual process of daily waste burial. To evaluate the effect of lag time on modeled methane emissions, the FOD equation that incorporates a lag time for a continuous deposition of waste was used (eq 15).27,30

DOCF for bulk MSW is calculated using eq 13, DOCF(MSW) =

∑ ∑ AFj × OX ji j

n

DOCMSW =

n

(13)

Where WIPi (Mg/Mg wet waste), DOCi, and DOCFi are the waste in place, fraction degradable organic carbon, and the E

DOI: 10.1021/acs.est.6b00415 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology

Figure 2. Comparison of modeled methane production and emissions to measurements. (A) Production; (B) Emissions; (C) Oxidation. Error bars for measured values represent 95% prediction interval estimates as discussed in the text. The model values are point estimates.



RESULTS AND DISCUSSION

which is consistent with a limit on the oxidizing capacity of the cover due to oxygen limitation or mass transfer constraints.9,31 Comparison of Modeled and Measured CH4 Emission, Production and Oxidation. Annual CH4 emissions from the study landfill were estimated using seven different model implementations. These estimates are compared to the integrated values of measured emissions in Figure 2. Figure 2A shows that in all cases, estimated CH4 production overestimates the measured CH4 production by an average factor of 16, 7, and 4 in 2011, 2012, and 2013, respectively. For all years, the EPA GHGRP bulk option CH4 production estimate was the lowest while the EPA GHGRP modified bulk was the highest. Methane emissions estimated using the different GHG reporting protocols are presented in Figure 2B along with the measured emissions calculated using instantaneous emissions data (Figure 1A). On average, the modeled methane emissions were higher than the measured emissions by a factor of 31, 10, and 4 for 2011, 2012, and 2013, respectively. When modeled emissions are compared to the upper estimate of measured emissions by using the upper band of the prediction interval in Figure 1A, modeled methane emissions are still higher than the

CH4 Mass Balance. Measured CH 4 emissions and oxidation, and calculated methane production as a function of time are presented in Figure 1. To estimate annual emissions from the measured data, a nonlinear power model was fitted to the TC emissions data by nonlinear regression (eq 1). Using this fit, integration of the curve over time yields an estimate of annual CH4 emissions (eq 2, Figure 1A). Methane production was calculated at each time point for which emissions and oxidation data were available, and then integration was conducted to obtain annual CH4 production (Figure 1B, eq 9). Annual methane oxidation was calculated by using eq 10, the results of which are presented in Figure 2C. As expected, CH4 production increased with time (Figure 1B). As decomposition of organic matter progresses and more waste is disposed, gas production increases, resulting in an increase in CH4 flux through the cover. Consequently, the OX decreases as oxygen likely limits the fraction of methane that can be oxidized. Interestingly, the mass of CH4 that was oxidized remained relatively constant (∼150 Mg/yr, Figure 1C) F

DOI: 10.1021/acs.est.6b00415 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology

Lincoln, NE) and a 3-D sonic anemometer (Windmaster Pro, model 1352, Gill Instruments Ltd., Lymington, England), with other equipment and data analysis procedures detailed elswhere.32,33 The EC systems was installed at 2.36 m height near the center of the landfill on April 25, 2012 and operated until May 8, 2013.33 Results showed an average flux obtained by EC of 5.39 μmol/ (m2-s1) which is within about 2% of the value obtained by TC (5.29 μmol/ (m2-s1)). While not proving that either method is accurate, it is promising when two independent methods show close agreement. The TC method has been evaluated in several studies. For example, the method was evaluated in a controlled release study in an open field. Over four releases that ranged from 1−3 g CH4/s, the TC method was shown to overestimate the controlled CH4 emission rate by 4−19%, suggesting that the method is not biased low.34 A comparison of TC and the vertical radial plume mapping (VRPM) approach described in the U.S EPA’s Other Test Method (OTM-10) was conducted at two California landfills.35,36 The two methods generally agreed within a factor of 2 and there was not a trend suggesting that one method was consistently higher.36 The discrepancy observed for the study site is consistent with other literature. For example, Monster et al.22 showed that for 15 Danish landfills, model predicted emissions were on average, a factor of 5 greater than the measured emissions. Similarly, Green et al.,37 conducted emissions measurements at four closed U.S. landfills and reported that model predicted emissions were greater than measurements by a factor ranging from two to seven. It should be noted, however, that in both of these studies,22,37 the annual emissions estimates were based on extrapolation from short-term emissions measurement data (order of days to a few months). This extrapolation could result in an error due to temporal variations in emission rate. A strength of the data set used here is that there was sufficient data to calculate an annual emission rate from measurements at multiple time points. To illustrate the significance of this issue, if annual emissions were to have been extrapolated from any single day’s emissions measurement, then annual emissions would have ranged from 101 to 722 Mg. In contrast, our time integrated estimate of measured emissions was 311 Mg. To determine if the measured emissions can be modeled using input parameters that are within ranges reported in the literature, simulations were run using values that are plausible and would reduce predicted emissions. The k, DOC, OX, and Δ were all evaluated to determine whether changes over ranges consistent with the literature could explain discrepancies between measured and modeled emissions. To evaluate the effect of Δ on CH4 production and emissions, the first order model was run assuming a continuous deposition rate for each year at k = 0.057 yr−1, DOC = 0.2, DOCF = 0.5 and OX = 0.1. The Δ was varied from 0 to 12 mo as the measurement data suggests that gas production commences within one year of waste placement. The behavior of the tested model as a function of lag time is compared with measured emissions in Figure 3. As expected, methane production and emissions decrease as lag time increases. However, the modeled production and emissions both remain much greater than the measured quantities. The FOD model was also run using the lowest literature value for k (0.02 yr−1), the maximum average value of OX suggested in the literature (0.37),38 and a Δ of 12 mo. In addition, the default DOC was reduced by 25%. Figure 4 shows

measured emissions by a factor of 9, 5, and 3 for 2011, 2012 and 2013, respectively. The higher modeled emissions as compared with measured emissions are consistent with the overpredictions of methane production presented in Figure 2A. For all years, the modeled emissions were highest in the EPA GHGRP modified bulk, while lowest emissions were observed with the SWICS protocol because of the use of the highest OX (0.35) (Table 1). The mass of methane oxidized that was estimated using the different GHG reporting protocols is higher than the measured values (Figure 2C) by a factor of 5, 2, and 2 for 2011, 2012, and 2013, respectively. It is difficult to compare the measured daily OX and the annual average OX used in GHG reporting protocols because the measured OX is instantaneous and shown to be a function of methane flux (Figure 1C). As presented in Figure 1C, the measured values of OX range from 0.05 to 0.6, which are both above and below the various model defaults. To compare measured and modeled OX, an annual average OXT was calculated by mass balance using eq 10, and the result is presented in Table 3. The OXT calculated from Table 3. Comparison of Methane Oxidation Factor (OX) Specified by the Different Reporting Protocols and OX Calculated from Mass Balance Using Measured Instantaneous Emissions and Oxidation protocol

2011

2012

2013

IPCC bulk IPCC composition EPA GHGRP bulk EPA GHGRP modified bulk EPA GHGRP composition CARB SWICS mass balance (OXT)a

0.1 0.1 0.25 0.25 0.25 0.1 0.35 0.57

0.1 0.1 0.25 0.1 0.25 0.1 0.35 0.40

0.1 0.1 0.25 0.1 0.25 0.1 0.35 0.26

a

OXT is the average annual fraction of produced methane that is oxidized. OXT was calculated as the difference between annual production and annual emissions (eq 10)]. Annual methane production was calculated by integration of instantaneous production which was calculated from instantaneous methane emissions and oxidation measurements.

mass balance is highest during the first year of decomposition and gradually decreases. This trend is consistent with the mechanistic basis of methane oxidation in which OX decreases with increasing flux as the oxidation process is limited by O2. The OXT calculated from mass balance is higher than the OX used in different models for the 2011 and 2012, but within the range of values (0.1−0.35) used in different models for 2013 (Table 3). The measured OX used in this study was derived from plume measurements including all uncollected methane and methane that is released through cracks and fissures in the cover. Evaluation of Discrepancy between Measured Emissions and Model Predictions. There are two potential explanations for the divergence between the measured emissions and model predictions; (1) the measurement technique is not accurate, or (2) the model parametrization is not correct. Both possibilities are evaluated here. To compare the TC measurements, a parallel measurement using an eddy covariance (EC) method was conducted at the study landfill. The EC system consisted of a LI-7700 CH4 and a LI-7500A CO2/H2O gas analyzer (LI-COR Biosciences, G

DOI: 10.1021/acs.est.6b00415 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology

that with these input parameters, modeled emissions are approximately a factor of 4.1, 1.4, and 0.7 of the measured emissions for 2011, 2012, and 2013 respectively. While modeled and measured emissions are comparable in 2012 and 2013, the inputs selected are extreme, particularly, the low value of k for a region that received 118 cm of precipitation annually. If a k of 0.04 yr−1 is used, then predictions and measurements are not as well aligned, with modeled emissions greater by a factor ranging from 1.3−8 (Figure 4). Evaluation of Variability among Model Predictions. As illustrated in Figure 2, there is variability among model predictions. Marked differences in modeled CH4 production were observed under different calculation options of the EPA GHGRP and IPCC models. These differences can be attributed to differences in k, DOC and DOCF that are related to waste composition. When compared to EPA GHGRP bulk, CH4 production in the GHGRP modified bulk is 44−51% greater (Figure 2A) because in the modified bulk model implementation, the waste stream is divided into MSW and C&D, each of which is assigned a k and DOC (Table 1). The MSW fraction is assigned the same k as the bulk option but the DOC increases from 0.2 to 0.31 Mg C/Mg. Thus, even when a small amount of C&D is present in the waste stream, the higher DOC assigned to the MSW results in a higher estimate of methane production. Similarly, when compared to GHGRP bulk, the composition option is 10−15% greater (Figure 2A). Under IPCC guidelines, CH4 production for the IPCC model option utilizing waste component-specific data is 21− 22% less than the IPCC bulk option. Since comparable parameters were used, comparable methane production was calculated using the CARB and SWICS protocols. To determine the extent to which the OX contributes to the difference in the modeled emissions, emissions were recalculated assuming an OX of 0.1 in all models while other parameters were maintained at their default value. As expected, predicted emissions increased for the EPA GHGRP and SWICS that have an OX > 0.1 as the specified default parameter. The coefficient of variation (CV = standard deviation/mean ×100%) of modeled methane emissions decreased from ∼23% for the default case to ∼16% when a uniform OX = 0.1 was used for all models. Not surprisingly, the use of a uniform OX decreased variation in model predictions. The methane production potential is a function of the DOC and DOCF and is a default in each model. The methane production potential is represented either by one value in the models treating MSW as one constituent, or as the sum of the product of waste composition and component-specific methane production potentials in the GHGRP Composition and IPCC Composition models. The overall methane production potential for each model implementation is presented in Table 1 and varies from 79−143 m3 CH4/wet Mg. When all bulk MSW models were run with the same methane production potential of 93 m3 CH4/wet Mg (i.e., DOC = 0.2 and DOCF = 0.5, Figure S1), the modeled emissions calculated using each guidance’ recommended OX and k vary by approximately 16% (excluding the IPCC bulk, which has the highest gas production because it uses a default k = 0.09 yr−1). Implications. For the studied landfill, model predictions exceed measured emissions by a factor of 4−31. The comparisons in this study are for the relatively early years of a landfill and cannot be used to infer discrepancies throughout the landfill’s life, although other studies have reported similar trends for older and closed landfills. In addition, the results

Figure 3. Effect of lag on modeled (A) CH4 production; and (B) CH4 emission (k = 0.057 yr−1, DOC = 0.2, DOCF = 0.5, OX = 0.1).

Figure 4. Comparison of modeled and measured CH4 production when varying model parameters are used (DOCF = 0.5, OX = 0.37, Δ = 12 mo).

H

DOI: 10.1021/acs.est.6b00415 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology

(9) Chanton, J.; Abichou, T.; Langford, C.; Hater, G.; Green, R.; Goldsmith, D.; Swan, N. Landfill Methane Oxidation Across Climate Types in the U.S. Environ. Sci. Technol. 2011, 45 (1), 313−319. (10) Mosher, B. W.; Czepiel, P. M.; Harriss, R. C.; Shorter, J. H.; Kolb, C. E.; McManus, J. B.; Allwine, E.; Lamb, B. K. Methane Emissions at Nine Landfill Sites in the Northeastern United States. Environ. Sci. Technol. 1999, 33 (12), 2088−2094. (11) Galle, B.; Samuelsson, J.; Svensson, B. H.; Borjesson, G. Measurements of Methane Emissions from Landfills Using a Time Correlation Tracer Method Based on FTIR Absorption Spectroscopy. Environ. Sci. Technol. 2001, 35 (1), 21−25. (12) Green, R. B.; Chanton, J. P.; Hater, G. R.; Swan, N.; Goldsmith, C. D. Estimates of Methane Emissions from Western Landfills Using OTM-10. In Proceedings of the Solid Waste Association of North America (SWANA) Landfill Symposium, Dallas, TX, 2011. (13) Quantifying Methane Abatement efficiency at Three Municipal Solid Waste Landfills. http://nepis.epa.gov/Adobe/PDF/P100DGTB. pdf. (accessed September 21, 2015). (14) Goldsmith, C. D.; Chanton, J.; Abichou, T.; Swan, N.; Green, R.; Hater, G. Methane emissions from 20 landfills across the United States using vertical radial plume mapping. J. Air Waste Manage. Assoc. 2012, 62 (2), 183−197. (15) Borjesson, G.; Samuelsson, J.; Chanton, J.; Adolfsson, R.; Galle, B.; Svensson, B. H. A national landfill methane budget for Sweden based on field measurements, and an evaluation of IPCC models. Tellus, Ser. B 2009, 61 (2), 424−435. (16) Borjesson, G.; Samuelsson, J.; Chanton, J. P. Methane Oxidation in Swedish Landfills Quantified with the Stable Carbon Isotope Technique in Combination with an Optical Method for Emitted Methane. Environ. Sci. Technol. 2007, 41 (19), 6684−6690. (17) Lohila, A.; Laurila, T.; Tuovinen, J.; Aurela, M.; Hatakka, J.; Thum, T.; Pihlatie, M.; Rinne, J.; Vesala, T. Micrometeorological Measurements of Methane and Carbon Dioxide Fluxes at a Municipal Landfill. Environ. Sci. Technol. 2007, 41 (8), 2717−2722. (18) Foster-Wittig, T. A.; Thoma, E. D.; Green, R. B.; Hater, G. R.; Swan, N. D.; Chanton, J. P. Development of a mobile tracer correlation method for assessment of air emissions from landfills and other area sources. Atmos. Environ. 2015, 102 (0), 323−330. (19) Barlaz, M. A.; Chanton, J. P.; Green, R. B. Controls on landfill gas collection efficiency: Instantaneous and lifetime performance. J. Air Waste Manage. Assoc. 2009, 59, 1399−1404. (20) Levis, J. W.; Barlaz, M. A. Is Biodegradability a Desirable Attribute for Discarded Solid Waste? Perspectives from a National Landfill Greenhouse Gas Inventory Model. Environ. Sci. Technol. 2011, 45 (13), 5470−5476. (21) National Weather Service Forecast Office. NOWData - NOAA Online Weather Data http://w2.weather.gov/climate/xmacis. php?wfo=ffc. (accessed November 4, 2015). (22) Mønster, J.; Samuelsson, J.; Kjeldsen, P.; Scheutz, C. Quantification of methane emissions from 15 Danish landfills using the mobile tracer dispersion method. Waste Manage. 2015, 35 (0), 177−186. (23) Scheutz, C.; Samuelsson, J.; Fredenslund, A. M.; Kjeldsen, P. Quantification of multiple methane emission sources at landfills using a double tracer technique. Waste Manage. 2011, 31 (5), 1009−1017. (24) R Core Team. R. A Language and Environment for Statistical Computing: R Foundation for Statistical Computing: Vienna, Austria, 2013; http://www.R-project.org/. (25) Cramér, H. Mathematical Methods of Statistics. Princeton University Press: Princeton, NJ, 1946. (26) Chanton, J. P.; Rutkowski, C. M.; Mosher, B. Quantifying Methane Oxidation from Landfills Using Stable Isotope Analysis of Downwind Plumes. Environ. Sci. Technol. 1999, 33 (21), 3755−3760. (27) IPCC Intergovernmental Panel on Climate Change. 2006 IPCC Guidelines for National Greenhouse Gas Inventories, Chapter 3; IPCC National Greenhouse Gas Inventory Program; IPCC, WMO: Geneva, Switzerland, 2006; Vol. 5.

reflect a comparison at one landfill and this analysis should be repeated at more young landfills before concluding whether the models evaluated are valid for young landfills. Currently, landfills above 2.5 million Mg in capacity are required to install a gas collection and control system within 5 years of waste burial and one area where these requirements could be more restrictive would be to require the installation of a gas collection and control system sooner after waste burial. The results for the studied landfill indicate that the benefits of earlier gas collection will be less than the benefits predicted from the models evaluated.



ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.est.6b00415.



A summary of the emissions and oxidation data (PDF)

AUTHOR INFORMATION

Corresponding Author

*Phone: (919) 513-4421; fax: (919) 515-7908; e-mail: [email protected]. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS Partial funding for this research was provided by Waste Management, Inc. We thank Dr. Jason Osborne of the Department of Statistics, NCSU for the assistance on statistical data analyses. The views expressed in this paper are those of the author(s) and do not necessarily represent the views or policies of the U.S. Environmental Protection Agency.



REFERENCES

(1) Municipal Solid Waste Generation, Recycling, and Disposal in the United States: Facts and Figures for 2012. http://www.Epa.Gov/ osw/nonhaz/municipal/pubs/2012mswfs.Pdf (accessed September 21, 2015). (2) Barlaz, M. A. Forest products decomposition in municipal solid waste landfills. Waste Manage. 2006, 26 (4), 321−333. (3) Landfill Methane Outreach Program. http://www3.epa.gov/ lmop/basic-info/index.html. (accessed September 21, 2015). (4) Inventory of U.S. Greenhouse Gas Emissions and Sinks: 1990−2012, EPA 430-R-14-003; U.S. Environmental Protection Agency: Washington, DC, 2014. (5) Denman, K. L.; Brasseux, G.; Chidthaisong, A.; Ciais, P.; Cox, P. M.; Dickinson, R. E.; Hauglustaine, D.; Heinze, C.; Holland, E.; Jacob, D.; Lohmann, U.; Ramachandran, S.; da Silva Dias, P. L.; Wofsy, S. C.; Zhang, X. Couplings between changes in the climate system and biogeochemistry. In Climate Change 2007: The Physical Science Basis Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change; Solomon, S., Qin, D., Manning, M., et al., Eds.; Cambridge University Press: Cambridge, U.K., 2007. (6) Spokas, K.; Bogner, J.; Chanton, J. A process-based inventory model for landfill CH4 emissions inclusive of seasonal soil microclimate and CH4 oxidation. J. Geophys. Res. 2011, 116, G04017. (7) Mandatory Greenhouse Gas Reporting. Code of Federal Regulations, Part 98, Title 40, 2009; Fed. Regist. 2009, 74, 56374. (8) Revisions to the Greenhouse Gas Reporting Rule and Final Confidentiality Determinations for New or Substantially Revised Data Elements; Final Rule. Code of Federal Regulations, Part 98, Title 40, 2013; Fed. Regist. 2013, 78, 71968. I

DOI: 10.1021/acs.est.6b00415 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology (28) Municipal Solid Waste in the United States: Facts and Figures for 2009. http://www.epa.gov/wastes/nonhaz/municipal/pubs/ msw2009rpt.pdf (accessed July 15, 2014). (29) California’s 2000−2009 Greenhouse Gas Emissions Inventory Technical Support Document. http://www.arb.ca.gov/cc/inventory/ doc/methods_00-09/ghg_inventory_00-09_technical_support_ document.pdf (accessed September 21, 2015). (30) Pingoud, K.; Wagner, F. Methane Emissions from Landfills and Carbon Dynamics of Harvested Wood Products: The First-Order Decay Revisited. Mitigation Adapt. Strat. Global Change 2006, 11 (5− 6), 961−978. (31) Chanton, J.; Abichou, T.; Langford, C.; Spokas, K.; Hater, G.; Green, R.; Goldsmith, D.; Barlaz, M. A. Observations on the methane oxidation capacity of landfill soils. Waste Manage. 2011, 31 (5), 914− 925. (32) Xu, L.; Lin, X.; Amen, J.; McDermitt, D.; Welding, K. Impact of Changes in Barometric Pressure on Landfill Methane Emission. Global Biogeochem. Cycles 2014, 28 (7), 679−695. (33) Li, J.; Green, R. B.; Magnusson, D. A.; Amen, J.; Thoma, E. D.; Foster-Wittig, T. A.; McDermitt, D. K.; Xu, L.; Burba, G. Using Eddy Covariance to Quantify Methane Emissions from a Dynamic Heterogeneous Area. In Proceedings of the Air and Waste Management Conference and Exhibition, Raleigh, NC, June 22−25, 2015. (34) Babilotte, A. Field Comparison of Methods for Assessment of Methane Fugitive Emissions from landfills. 2011. http://erefdn.org/ publications/uploads/FugitiveEmissions_FinalReport.pdf (accessed September 21, 2015). (35) Thoma, E.; Green, R.; Hater, G.; Goldsmith, C.; Swan, N.; Chase, M.; Hashmonay, R. Development of EPA OTM 10 for Landfill Applications. J. Environ. Eng. 2010, 136 (8), 769−776. (36) Green, R. B., Hater, G. R., Thoma, E. D., DeWees, J., Rella, C. W., Crosson, E. R., Goldsmith, C. D., Swan, N. Methane Emissions Measured at Two California Landfills by OTM-10 and an Acetylene Tracer Method. In Proceedings of the Global Waste Management Symposium, San Antonio, TX, October 3−6, 2010. (37) Green, R. B.; Swan, N. D.; Thoma, E. D.; Footer, T. L.; Chanton, J.; Hater, G. R. Measured and Modeled Methane Emissions at Closed MSW Landfills without Gas Collection. In Proceedings of the Global Waste Management Symposium, Phoenix, AZ, September 30October 3, 2012. (38) Chanton, J. P.; Powelson, D. K.; Green, R. B. Methane Oxidation in Landfill Cover Soils, is a 10% Default Value Reasonable? J. Environ. Qual. 2009, 38 (2), 654−663.

J

DOI: 10.1021/acs.est.6b00415 Environ. Sci. Technol. XXXX, XXX, XXX−XXX