United States
Environmental Protection
Agency
Air and Energy Engineering
Research Laboratory
Research Triangle Park NC 2771
Research and Development
EPA/600/S7-86/055 Apr. 1987
Project Summary
Estimation of Uncertainty for the
1980 NAPAP Emissions Inventory
John E. Langstaff and Janice K. Wagner
The National Acid Precipitation As-
sessment Program (NAPAP) has
charged its Task Group on Emissions
and Controls with developing compre-
hensive and accurate inventories of
natural and anthropogenic emissions of
substances thought to be important in
acid deposition processes. This respon-
sibility includes quantifying the degree
of uncertainty associated with those
estimates. This report presents the
methods, assumptions, and results of
an effort to develop quantitative esti-
mates of emissions uncertainties for
the 1980 NAPAP Emissions Inventory.
Conventional statistical concepts were
applied to sample uncertainty data. The
shortcomings of these data and ques-
tions concerning the proper application
of statistical methods to the problem at
hand were deferred for later study.
Results of this study consist of lessons
learned and problems identified, and
fall into three categories: uncertainty
relationships at various levels of emis-
sions disaggregation, statistical meth-
odology questions, and gaps in input
data. The computed uncertainty values
are illustrative in nature and included
for completeness; these values may be
used only to identify uncertainty trends
and relationships between the various
pollutants at increasing levels of ag-
gregation and allocation.
This Project Summary was developed
by EPA's Air and Energy Engineering
Research Laboratory, Research Triangle
Park, NC, to announce key findings of
the research project that Is fully docu-
mented In a separate report of the same
title (see Project Report ordering In-
formation at back).
Introduction
The National Acid Precipitation Assess-
ment Program (NAPAP) has charged its
Task Group on Emissions and Controls
with developing comprehensive and ac-
curate inventories of natural and an-
thropogenic emissions of substances
believed to be important in acid deposition
processes. The Task Group is developing
estimates of past, present, and future
emissions with adequate geographic,
temporal, and source resolution to sup-
port the research requirements of NAPAP.
The Task Group is also responsible for
the quantification of the uncertainty as-
sociated with those estimates.
This report presents the methods, as-
sumptions, and results of a study to
develop quantitative estimates of emis-
sions uncertainties for the 1980 NAPAP
Emissions Inventory. Conventional statis-
tical concepts, adapted to the emissions
inventory by Brookhaven National Labora-
tory in previous work, were applied to
illustrative uncertainty estimates supplied
by a team of emissions inventory devel-
opers. The shortcomings of these data
and questions concerning the proper ap-
plication of statistical methods to the
problem at hand were deferred for later
study.
Approach
The calculation of emissions uncer-
tainty for individual sources and source
aggregation is performed by a computer
program. Annual emissions estimates for
each process are allocated temporally,
spatially, and by pollutant subspecies.
This allocation is carried out by merging
process- and pollutant-specific allocation
-------
factors to the individual emission records
in the file. After these allocation factors
have been assigned, the system calculates
emissions by pollutant for a specified
averaging time, source classification, or
geographical region. The emissions un-
certainty is calculated similarly. Uncer-
tainty estimates, expressed as percent
error values, are assigned to the emission
records on a process- and pollutant-
specific basis. By use of the error propa-
gation formulas for products and sums, it
has been assumed that emissions uncer-
tainty can be computed directly for the
various levels of source aggregation.
The initial set of uncertainty input
values was provided by a panel of emis-
sions inventory developers who were
assembled at a 1980 NAPAP Emissions
Inventory Uncertainty Workshop in May
1985 at the Air and Energy Engineering
Research Laboratory, U.S. EPA, Research
Triangle Park, NC. Details of the results of
this Workshop are included in report
Appendix A. At the Workshop, the panel
was asked to provide consensus estimates
of the 90% confidence interval surround-
ing nominal mean values for generalized
classes of the variables used to calculate
emissions at various levels of species,
temporal, and spatial resolution.
Although members of this panel were
experienced in developing emissions in-
ventories and the methods for estimating
emissions, they could not be considered
experts in the calculation of emissions
and associated uncertainty for all source
types They were generally not experi-
enced with the actual emitting processes
under consideration or with the typical
variability of emissions from those
processes
This study was intended as a first step
toward quantifying uncertainty in the
1980 NAPAP Emissions Inventory. Ob-
vious shortcomings in the data set and
doubts about validity of the methods were
deferred to later study. Key assumptions
in the uncertainty calculations are:
• Emission calculation parameters are
independent; i.e., they do notcovary.
• Emission factors represent true mean
values.
• All estimates are unr ased.
• The emission para leters can be
treated as random variables which
are approximately normally dis-
tributed.
• No coding or transcription errors are
present.
• The data are complete; no emissions
data, emissions sources, or emis-
sions source categories are missing.
Results from the Workshop were later
extended to other pollutants, and other-
wise interpreted, by contractor personnel
and the EPA Project Officer. These ex-
tensions were performed to illustrate
uncertainty results for all emission
estimates in the inventory. For these
reasons, the results of the uncertainty
calculations should be used only to identify
uncertainty relationships between the
various pollutants at increasing levels of
aggregation and allocation.
Results
Based upon the uncertainty estimates
of the panel, the contractor, and the EPA
Project Officer, and the methodology
described in report Section 3, illustrative
emissions uncertainty values have been
calculated for sulfur dioxide (S02), oxides
of nitrogen (NOJ, speciated oxides of
nitrogen (S-NOJ, carbon monoxide (CO),
volatile organic compounds (VOC), speci-
ated volatile organic compounds (S-VOC),
total suspended particulates (TSP),
speciated total suspended particulates
(S-TSP), lead (Pb), ammonia (NH3),
hydrogen chloride (HCI), hydrogen fluo-
ride (HF), and sulfate (SO4). (See Tables 1
and 2.) The percent uncertainties show
an expected decrease as aggregation level
increases, and an increase in uncertainty
with disaggregation to finer levels of
detail. Note that these phenomena result
inevitably from the combination of as-
sumptions about the character of the
uncertainty input data and the workings
of the error propagation formulas on data
of that character; the particular values
provided by the panel are not important
in this effect.
The values of uncertainty estimates for
national levels of aggregation appear to
be unreasonably small The emission in-
ventory data do not warrant this degree
of confidence. Therefore, the assump-
tions, data, and methodology must be
reexammed for reasonableness. Prelimi-
nary indications are that the methodology
must be expanded to incorporate other
sources of error (e.g., bias, coding, and
emission errors). Other elements of the
uncertainty estimates are also counter-
intuitive and should be corrected by
improved assumptions, data, and metho-
dology Thus, the uncertainty estimates
presented are only illustrative of the
results achieved from applying the sample
data set to the proposed methodology.
Despite the limitations noted above,
this experiment represents a helpful step
in learning how best to approach quantifi-
cation of uncertainty in the NAPAP Emis-
sions Inventories. Lessons learned here
will guide further efforts toward an im-
proved level of scientific validity and
credibility. Investigation of the assump-
tions, methodology, and required input
data is continuing.
-------
Tab/a 1. Annual/Hourly Illustrative Uncertainty of the 1980 NAPAP Emissions Inventory
Percent Error"
Nation
point & area
Region average?
point & area
State average
point & area
County Average
point & area
point
area
Grid average
point & area
area
SCC average
point
Annual
Hourly
S02 NOxSNOxb CO VOCSVOCC TSP STSP" Pb NH3 HCI HF SO4 S03 NOtSNO,b CO VOC SVOCC TSP STSP* Pb NH3 HCI HF S04
11111
42344
3524191
9 14 7 14 6 8 5
2 1 2 3 2
4 6 4 5 4 13 3
84596 9 12 17 13 19 19 21 9
64 5 88 14 16 25 13 26 13 13 12 14 8 9 17 12 18 22 31 21 35 34 34 21
16 16 19 29 32 54 73 110 54 60 57 34 62
24 27 32 31 36 79 30 72 33 74 24 19 34
16 16 19 30 33 56 79 120 56 60 94 100 73
25 18 21 30 31 52 74 11O 45 50 25 2O 34
20 19 22 33 37 60 82 120 59 63 97 110 75
32 29 31 51 48 68 10O 140 82 82 94 74 88
48 48 51 66 68 100 61 95 80 100 65 61 68
32 29 31 52 49 70 110 150 83 82 130 140 99
48 30 32 49 46 64 99 130 68 67 68 64 61
33 31 33 54 53 74 110 150 86 84 130 140 1OO
39 45 54 45 52 120 47 120 88 1OO 99 99 89 74 78 83 78 100 160 99 150 130 140 140 140 130
" With 90% confidence.
*Speciated N0r
cSpeciatedVOC.
d Speciated TSP.
"EPA Region.
Up to two significant digits have been reported to facilitate comparisons, and does not imply a corresponding level of precision.
Table 2. Daily/Seasonal Illustrative Uncertainty of the 1980 NAPAP Emissions Inventory
Percent Error"
Nation
point & area
Region average0
point & area
State average
point & area
County average
point & area
point
area
Grid average
point & area
area
SCC average
point
Daily
Seasonal
S02 NOxSNOxb CO VOCSVOCC TSPSTSP* Pb NH3 HCI HF S04 S02 NOxSNOltb CO VOCSVOCC TSP STSP* Pb NH3 HCI HF S04
11121 2 35241 10 2
53354 7 9 14 8 14 9 10 6
8 5 6 9 8 14 17 26 14 27 17 18 13
11111
3 5 2 4 1 10 1
42344 7 9 14 7 14 6 8 5
6 4 5 8 8 14 16 25 13 26 14 14 12
21 19 22 33 34 56 77 120 58 63 63 42 65 17 16 19 30 32 55 74 110 55 61 58 35 62
31 33 38 37 41 83 35 75 42 78 32 28 40
21 20 22 33 36 58 82 120 59 63 98 110 77
25 28 33 32 36 80 31 73 35 75 25 2O 35
17 16 20 31 34 56 79 120 56 61 94 100 74
32 21 23 33 33 53 77 110 49 52 34 29 38 26 18 21 30 31 52 74 110 46 50 27 22 34
23 22 25 36 39 61 86 130 63 65 100 110 78 20 20 22 33 37 60 82 120 59 63 98 11O 75
50 55 62 55 61 130 56 120 94 110 110 100 95 41 47 55 47 53 120 49 120 89 100 10O 100 90
' With 9O% confidence.
*SpeciatedN0r
c Speciated VOC.
''Speciated TSP.
'EPA Region.
Up to two significant digits have been reported to facilitate comparisons, and does not imply a corresponding level of precision
-------
J. Langstaff and J. Wagner are with Alliance Technologies Corporation. Bedford.
MA 01730.
J. David Mobley is the EPA Project Officer (see below)
The complete report, entitled "Estimation of Uncertainty for the 1980 NAPAP
Emissions Inventory," (Order No. PB 87-145 397/AS; Cost: $13.95, 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
Official Business
Penalty for Private Use $300
EPA/600/S7-86/055
0000329 PS
------- |