United States
Environmental Protection
Agency
Air and Energy Engineering
Research Laboratory
Research Triangle Park NC 27711
Research and Development
EPA/600/SR-92/037 April 1992
ifirEPA Project Summary
Development of an Empirical
Model of Methane
Emissions from Landfills
Rebecca L. Peer, David L. Epperson, Darcy L. Campbell,
and Patricia von Brook
The U.S. Environmental Protection
Agency's (EPA's) Air and Energy Engi-
neering Research Laboratory (AEERL)
began a research program in 1990 with
the goal of improving global landfill
methane (CH4) emissions estimates. Part
of this program is a field study to gather
information that can be used to develop
an empirical model of CH4 emissions.
The field study is the subject of this
report.
Twenty-one U.S. landfills with gas
recovery systems were included in the
study. Site-specific information includes
average CH4 recovery rate, landfill size,
tons of refuse (refuse mass), average
age of the refuse, and climate. A corre-
lation analysis showed that refuse mass
was positively linearly correlated with
landfill depth, volume, area, and well
depth. Regression of the CH4 recovery
rate on depth, refuse mass, and volume
was significant, but depth was the best
predictive variable (R2 = 0.53). Refuse
mass was nearly as good (R2 = 0.50).
None of the climate variables—precipi-
tation, average temperature, dewpoint—
were correlated with the CH4 recovery
rate or with CH4 recovery per metric ton
of refuse. Much of the variability in CH4
recovery remains unexplained, and is
likely due to between-site differences in
landfill construction, operation, and
refuse composition. A model for global
landfill emissions estimation is pro-
posed.
This Project Summary was devel-
oped by EPA's Air and Energy Engi-
neering Research Laboratory, Research
Triangle Park, NC, to announce key find-
ings of the research project that is fully
documented in a separate report of the
same title (see Project Report ordering
information at back). -
Introduction
The U.S. Environmental Protection
Agency's (EPA's) Air and Energy Engi-
neering Research Laboratory (AEERL)
began a research program in 1990 with
the goal of improving global landfill meth-
ane (CH.,) emissions estimates. A review
of currently available models and data iden-
tified several theoretical models and labo-
ratory experiments used to estimate CH4
production in individual landfills. However,
adapting these methodologies for gbbal
estimates posed several problems, the
worst being that site-specific data would
be needed for every country. The few glo-
bal emissions methodologies that were
found were reasonable, but were ham-
pered by a paucity of data. In particular,
reliable refuse generation rates and waste
composition data were not available for
many countries. In addition, many landfill
experts believe that climate (particularly as
it affects moisture input to the landfill) has
an effect on CH4 generation rates. No cur-
rently available model incorporates climate
as a controlling variable.
In order to accurately estimate CH4 gen-
eration in landfills on a global basis, a
model is needed that is responsive to a
wide range of climates, types of waste,
and landfill practices. Understanding the
effects of climate on CH4 production is
especially important to climate modelers
who are studying feedback effects of glo-
Printed on Recycled Paper
-------
bal climate change. Therefore, AEERL ini-
tiated a field study to gather data to:
Identify key variables that affect CH4
generation; and
• Develop an empirical model of CH4
generation based on those variables.
The results of a field study of 21 U.S.
landfills are presented. The program was
limited to acquisition of CH4 data gathered
by on-s'rte monitors. Furthermore, no other
sampling or testing was planned. Data ac-
quisition was confined to historical records
kept at individual sites.
The objectives of the study were to:
• Develop a statistical model of annual
landfill CH4 emissions as a function
of climate, refuse mass and age, and
other physical characteristics (if war-
ranted);
• Compare the performance of the sta-
tistical model to a deterministic kinet-
ics-based model of landfill CH4
production; and,
• Develop a simple model that can be
used to estimate global CH4 emis-
sions from landfills.
It is important to note that CH4 recovery is
being used as a surrogate for CH4 emis-
sions in this study, thus affording the
potential to both underestimate and over-
estimate emissions. The method may un-
derestimate if gas recovery is not 100%
efficient; some CH4 may still be lost
through the cap or by lateral gas migra-
tion out of the landfill. On the other hand,
the method may overestimate if gas re-
covery circumvents the reoxidation of CH4
by methanotrophs, methanogens. and sul-
fate-reducing bacteria. Given that strong
arguments can be made for both cases
and no quantitative data exist for either,
the approach used in this study is to as-
sume that both cases are true but the net
effect is zero. If data that refute this as-
sumption become available, the model will
be adjusted.
Data Summaries and Statistical
Analyses
Table 1 shows the average CH4 recov-
ery rate for each landfill, as well as other
summary statistics. The number of mea-
surements available varied a great deal
between sites. Table 2 summarizes the
landfill statistics used in the analysis and
model development.
Climate data were obtained from the
Southeast Regional Climate Center for a
cooperative National Weather Service
(NWS) station nearest each landfill. The
monthly average temperature and total rain-
fall values were summed and converted to
average annual temperature and total an-
nual rainfall values for each year. The
Table 1. Summary Statistics for Methane Recovery Rates Grouped by Measurement Type
Measurements
Methane Recovery (nf/min)
Landfill Type Number Average
Standard Minimum Maximum
Median Deviation Value Value Range
1
2
3
4
5
6
7
8
9
10
11
12
13
16
17
20
21
22
23
24
25
daily
daily
daily
daily
daily
daily
monthly
daily
daily
monthly
monthly
daily
daily
monthly
minute
monthly
daily
daily
daily
daily
daily
194
302
314
85
209
37
12
626
15
6
12
232
11
15
13
12
11
51
202
333
331
55.3
18.0
40.0
98.4
24.8
16.7
9.7
11.7
7.7
29.3
11.3
8.0
10.4
16.0
13.8
35.0
27.4
33.2
2.2
17.7
20.2
55.3
18.2
40.3
98.7
24.9
16.8
10.2
12.4
7.0
30.5
11.7
7.7
11.0
16.6
13.9
35.1
26.5
31.8
2.3
17.9
20.8
2.12
1.19
2.32
1.33
1.70
2.07
2.01
2.46
1.42
3.34
1.22
1.02
1.50
4.13
1.50
4.75
2.94
7.84
0.51
2.18
2.80
48.0
12.3
30.2
93.3
20.5
12.6
4.0
0.5
5.7
23.4
9.1
5.4
7.8
7.4
10.1
26.5
24.5
21.6
0.3
3.1
1.4
61.4
20.5
44.6
101.5
27.9
22.6
12.0
17.1
10.5
32.4
12.7
10.4
11.7
21.6
16.5
41.3
32.9
60.0
2.9
22.1
24.4
13.4
8.2
14.4
8.2
7.4
10.0
8.0
16.6
4.8
9.0
3.6
5.0
3.9
14.2
6.4
14.3
8.4
38.4
2.6
19.0
23.0
annual temperature and rainfall values for
the years of refuse acceptance were then
averaged for comparison to landfill data for
each landfill.
In addition to the daily weather data,
the 30-year averages of annual mean tem-
perature, mean dewpoint temperature, and
total rainfall were obtained for the, NWS
stations. These 30-year averages of tem-
perature and rainfall are referred to as the
'normal' values.
Dewpoint temperature was included in
this analysis because it is a readily avail-
able variable that provides a better mea-
sure of moisture availability than either
temperature or precipitation. Better com-
posite variables could be chosen (such as
actual evapotranspiration), but calculating
these values was beyond the scope of this
project.
Based on the preliminary data analy-
ses, a linear model appeared to be suffi-
cient to model CH4 recovery rate. The SAS
regression procedure (PROC REG) was
used to generate regression statistics for
various models. Two general models were
used—one to predict CH4 recovery rate,
the other to predict CH4 recovery rate per
unit mass. Selection of variables for the
regression models was based on the re-
sults of. the correlation and scatter plots
summaries discussed above. In addition,
the data distribution of potential regression
variables was examined for normality. Al-
though most variables were not normally
distributed, the distributions were not so
far off as to warrant data transformations.
Table 3 shows the results of several
linear regression models. For most of the
models that use a single landfill parameter,
the intercept term was found to be insig-
nificant. From the regression model results
shown in Table 3, landfill depth appears to
be the best predictor of CH4 emissions (P
= 0.0002, R2 = 0.53). However, refuse
mass is very nearly as good (P - 0.0003,
R2 = 0.50). The best model used both
depth and mass as predictive variables (P
= 0.0001, R2 = 0.65). Because waste pro-
duction data are much more widely avail-
able than landfill depth data on a global
basis, the no-intercept regression of CH4
recovery on refuse mass is the better model
choice. This model is:
CH4 = 4.52W
where:
CH4 = methane flow rate (mVmin); and
W = mass of refuse (106 Mg).
Figure 1 shows the regression line for
CH4 recovery rate as a function of refuse
mass. The 95-% confidence interval of the
regression line is shown by the dashed
lines.
No other variables were found to have
any effect on CH4 production. In particular,
no functional model was found linking CH4
production to climate variables. This does
not mean that climate is not important.
Given the unexplained variability in the
CH4-versus-tons regression, some aspect
of climate may actually play a controlling
role. However, as shown in this study, site-
specific factors and difficulties in accu-
rately quantifying key parameters confound
the problem. ~"
-------
1
3
C3
HJH
lit
B=
11
21!!-
S^S
S.?a
t
•«c
01
I
.5
ti
c\i
-
3d o' co •» -: co c> ri K co' CM' N;
•^ rt
« »- CS <0 F O
T* CM CM co
CO CO C^ C\j CO 00 l*"x 00 \J* OS (£5 *•••
O Cft v-
K irf c\i
K Ki Ol 1C jp- K K tx
co' o' ci oi uj N« o' c> K CM' ci oo' K uj ci w co c\i o' ui c\j
-------
Table 3. Landfill Regression Summary
Regression Mode!' Prob > F
bO
b1
b2
Comments
methane * depth
methane = depth
methane « 1& Mg
methane « 1 V Mg
methane = volume
methane « volume
methane = we/te
methane = we/fc
methane « cfeptft + 70s M<7
methane « depth •*• 70s My
methane^ 10s Mg
+ moan rain
methane* 10s Mg
+ mean temp
methane = 10s Mg
+ dowpoint 30
mothane/Mg^ mean rain
methane/Mg = mean temp
methane/Mg « dewpoint 30
0.0002
0.0001
0.0003
0.0001
0.0011
0.0001
0.0701
0.0001
0.0001
0.0001
0.0011
0.0015
0.0009
0.7688
0.7607
0.6127
0.53
—
0.50
—
0.44
—
0.16
—
0.65
—
0.53
0.52
0.54
0.00
0.01
0.01
-1.09
—
1.89
—
7.38
—
6.87
—
-5.96
—
-10.31
-4.67
-2.98
4.48
6.64
4.61
9.13E-1 —
8.84E-1 —
4.27 —
4.52 —
1.37E-6 —
1.73E-6 —
3.08E-1 —
4.07E-1 —
2.36 0.18
2.056 0.15
4.32 1.22E-1
4.11 5.61E-1
3.97 9.49E-1
6.19E-3 —
-4.21E-2 —
7.03E-2 —
intercept not significant
no intercept in model
intercept not significant
no intercept in model
intercept not significant
no intercept in model
model fit & wells borderline;
intercept not significant
no intercept in model
intercept not significant
no intercept in model
intercept & mean rain not
significant
intercept & mean temp not
significant
intercept & dewpoint 30 not
significant
poor model fit; mean rain
not significant
poor model fit; mean temp not
significant
poor model fit; dewpoint 30
not significant
Methane
variable 1 + b2 • variable 2.
In order to validate the statistical model,
Hs performance was compared to that of
the U.S. EPA's Landfill Air Emissions Esti-
mation Model, which is a deterministic com-
puter model that was developed for
regulatory purposes. Assuming that the
refuse has been accepted at the same
annual rate over time (i.e., all submasses
are of the same size), the model equation
Is as follows:
QCH4
where:
QCH4 .
R (exp{-kc) - exp(-kt)}
CH4 generation rate at time t,
fttyr
LO - potential CH, generation capac-
ity of the refuse, ffVMg refuse
R - average annual refuse accep-
tance rate during active life,
Mg/yr
k m CH4 generation rate constant,
1/yr
c - time since landfill closure, year
(c - 0 for an active landfill)
t - time since initial refuse place-
ment, year
The Landfill Model methodology is
based on the Scholl Canyon model, which
Is a first order decay equation. Because
site-specific characteristics are required as
model input, the Landfill Model is impracti-
cal for use on a global scale.
The relative performances of the mod-
els were compared using a ratio of pre-
dicted CH4 emissions to actual CH4
recovered. The mean value of the ratio for
all 21 landfills provides a measure of the
model's relative accuracy.
The results of the Landfill Model and
regression model comparisons are shown
in Table 4. As shown by the mean of the
ratios, the Landfill Model with LO of 50 m3/
Mg tends to underpredict (ratio less than
1). When LQ is set to 162 m3/Mg, the
model, on average, is very accurate (the
mean ratio of 1.07 approximates 1); the,
model default LO (298 m3/Mg) tends to
overestimate CH4 (mean ratio = 1.97). The
regression model's mean ratio of 1.39 falls
between Landfill Model runs 1 and 3.
The regression model performs rea-
sonably well compared to the Landfill
Model. One particular advantage of using
a statistical model is that only one variable
is required. Furthermore, it is relatively
easy to add new observations and further
refine the model, as only average CH4
recovery and refuse mass are required.
The confidence limits of the regression
coefficient can be used to bound estimated
CH4 emissions. The upper and lower 95-%
confidence limits are 6.52 and 2.52 m3
CH4/Mg refuse, respectively.
Summary and Conclusions
This research program had as its goal
the development of an empirical model of
CH4 emissions from landfills, ft was suc-
cessful in meeting its major objectives, but
much remains to be learned. The main
objective—developing a- model for global
emissions—was achieved. The strengths
and successes of this program include:
Development of a model that accu-
rately reflects real world variability of
landfill CH4 recovery;
The model is very simple and easily
adapted to global emissions estima-
tion;
The uncertainty associated with CH4
recovery was quantified; and,
The program was cost-effective, al-
lowing maximization of sample size.
The weaknesses of this approach are:
• The model is not mechanistic, and is
therefore limited in its usefulness.
Between-site variability is high, and
much of the variability remains unex-
plained by the model.
• Recovery is used as a surrogate for
emissions. The validity of this substi-
tution is unknown.
-------
100
80-
(S! eo
I
1
g 40
20-
U
1 1
0 5 10 15
Tons of Refuse (millions of metric tons)
Figure 1. Methane recovery regression with 95% confidence interval of regression coefficient.
A factor for estimating landfill CH4 emis-
sions can be proposed based on the CH4
per refuse mass regression model. The
intercept was not significant, so the sim-
pler model (with the line forced through the
origin) can be used. The slope for this line
is 4.52 m3 CH4 per min/108 Mg of refuse;
this factor can be used to estimate annual
CH, emissions by multiplying it by the total
refuse landfilled each year.
On a global basis, this factor may over-
estimate CH4 production for many coun-
tries. The composition of wastes from
less-developed countries in particular is
lower in paper and therefore less likely to
produce CH,. Also, global landfilling prac-
tices vary much more than those of the
sample population of U.S. sites. On the
other hand, if waste decays more slowly
than assumed in this study (20 years),
then this factor underestimates CH4 per
ton of refuse.
Despite these concerns, the CH4 po-
tential factor developed in this study should
yield more reasonable estimates of global
landfill CH4 emissions than are currently
available because the factor is based on
actual landfill data rather than theoretical
models. By careful consideration of all the
mitigating effects, some of which are dis-
cussed in this report, this simple model
can be used to quantify and reduce some
of the uncertainty in global estimates.
Table 4.
Comparison of Model Performances
Landfill Air Emissions Estimation Model
Site Number
1
2
3
4
5
6
7
8
9
10
11
12
13
16
17
20
21
22
23
24
25
Mean
Standard Deviation
Run 1
Pred./Actual
0.16
0.48
0.28
0.22
0.58
0.24
0.46
0.37
0.36
0.25
0.23
0.54
0.16
0.33
0.49
0.41
0.15
0.19
1.74
0.54
0.82
0.43
0.34
Run 2
Pred/Actual
0.40
1.21
0.71
0.55
1.44
0.60
1.16
0.93
0.90
0.64
0.57
1.34
0.39
0.82
1.23
1.02
0.36
0.47
4.35
1.34
2.06
1.07
0.85
Run 3
Pred./Actual
0.73
2.23
1.31
1.01
2.66
1.10
2.14
1.71
1.67
1.17
1.05
2.47
0.72
1.52
2.26
1.88
0.67
0.87
8.00
2.46
3.79
1.97
1.56
Regression
Pred/Actual
0.52
1.55
0.83
0.62
1.95
0.73
1.50
1.15
1.15
0.83
0.85
1.72
0.57
1.02
1.73
1.24
0.47
0.57
6.32
1.60
2.34
1.39
1.24
•&U.S. GOVERNMENT PRINTING OFFICE: 1992 - 648-080/40227
-------
-------
-------
RL Peer, D.L Epperson, D.L Campbell, and P. von Brook are with Radian Corp.,
Research Triangle Park NC 27709.
Susan A. Thorneloe is the EPA Project Officer, (see below).
The complete report, entitled "Development of an Empirical Model of Methane Emis-
sbns from Landfills," (Order No. PB92-152875/AS; Cost: $26.00, subject to change)
will be available only from:
National Technical Information Service
5285 Port Royal Road
Springfield, VA 22161
Telephone: 703-487-4650
The EPA Project Officer can be contacted at:
Air and Energy Engineering Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park NC 27711
United States
Environmental Protection
Agency
Center for Environmental Research
Information
Cincinnati, OH 45268
BULK RATE
POSTAGE & FEES PAID
EPA PERMIT NO. G-35
Official Business
Penalty for Private Use $300
EPA/600/SR-92/037
------- |