nonroad emissions model uncertainty analysis for the state of … · 2015-09-10 · nonroad...
TRANSCRIPT
NONROAD Emissions Model Uncertainty NONROAD Emissions Model Uncertainty Analysis for the State of GeorgiaAnalysis for the State of Georgia
Rosa ChiAlper Unal, Di Tian, Armistead (Ted) Russell
School of Civil & Environmental EngineeringGeorgia Institute of Technology
International Emission Inventory Conference, "Working for Clean Air in Clearwater," , Clearwater, FL
7-10 June 2004
Georgia Institute of Technology
Georgia Institute of Technology
0 80 16040 Miles±
LegendCounty Boundaries
AUGUSTA
COLUMBUS
ATLANTA
MACON
• Areas at or near non-attainment means Air Quality Modeling!!
• Understanding emissions uncertainty leads to better understanding of AQM limitations, errors, uncertainties…
Motivation
Fall line Air Quality Study
Georgia Institute of Technology
• Assess urban and regional air pollution, identify the sources of pollutants and pollutant precursors, and recommend solutions to the Augusta, Macon, and Columbus metropolitan areas of Georgia.
Why Nonroad?VOC Emissions
• Significant source of emissions in polluted areas
• EPA NONROAD model is user-friendly, easy to manipulate good for uncertainty analysis!!
NOx Emissions
Georgia Institute of TechnologySource: Fall-line Air Quality Study August 2000 Inventory for 11 counties
Objective
• Quantify uncertainty of emissions from the NONROAD model using a bottom-up, emissions-based approach for the state of Georgia
Georgia Institute of Technology
EPA NONROAD Emissions Model
• Includes over 260 off-road equipment type in source categories: Construction, Lawn & Garden, Industrial, Agricultural, etc.
• Does NOT include: Aircraft, Locomotives, Commercial Marine
Georgia Institute of TechnologyPhotos from www.freefoto.com
NONROAD Emissions Calculations
• EMISSIONS = (Population) x (Rated Power) x (Load Factor) x (Activity) x (Emission Factor)
• Complications:– Geographic allocations– Age distribution– Deterioration– Growth– Scrappage
Georgia Institute of Technology
Emission Factors
Deterioration
ActivityPopulation
Growth
EF Corrections
Geographic Allocation
Conversion Factors
Scrappage
Age Distribution
Emissions
HC Conversions
PM Fractions
Rated Power
Load Factor
Hours per Year
Seasonal/ Monthly
Corrections
Activity Corrections
Weekend/ Weekday
Corrections
Temperature RVP Fuel SulfurOxygenate
Average Life
Georgia Institute of Technology
Methods
• Use 1999 average summer day scenario for Georgia
• Sensitivity analysis of NONROAD
• Expert elicitation of uncertainty of major NONROAD
input parameters
• Monte Carlo uncertainty analysis of NONROAD
emissions output uncertainty
Georgia Institute of Technology
Sensitivity Analysis• Determine sensitivity of NONROAD emissions to
input parameters by: base
base
EP
PPEES ×
−−
=)()(
%90%110
%90%110
Georgia Institute of Technology
-50%
0%
50%
100%
150%
RV
P
Sul
fur
Tem
pera
ture
Pop
ulat
ion
Act
ivity
Load
Fac
tor
Use
ful L
ife
Em
issi
onFa
ctor
**
200%
250% VOC Exhaust
VOC EvapVOC Total
NOxCO
PM2.5CO2
SOx
Emission Factor Bootstrap Analysis• Diesel engine emission test certification data available in
NONROAD documentation for model years 1996 +
• AuvTool (Frey, Zheng; NC State) empirical distribution fits and
bootstrap analysis
• Shows uncertainties due to random sampling errors, but not
errors of representativeness
Model YearAverage -5% 5% -21% 21% -4% 3% -16% 16% -10% 10%Maximum -7% 7% -49% 46% -6% 5% -20% 23% -18% 17%Minimum -2% 2% -4% 4% -2% 2% -8% 8% -4% 4%
HP 95% Confidence
Interval
HC 95% Confidence
Interval
NOx 95% Confidence
Interval
CO 95% Confidence
Interval
PM 95% Confidence
Interval
Georgia Institute of Technology
Expert Elicitation• Focus on experts in nonroad emissions modeling, not
air quality modeling
• Experts self-rate based on experience
• Experts give opinions of uncertainties of NONROAD
input parameters
• Expert opinions aggregated as average of answers
weighted by self-rating
Georgia Institute of Technology
Expert Elicitation
Input Parameter
Most Uncertain 95% Confidence Interval
Least Uncertain 95% Confidence Interval
Population +68%, -25% +29%, -23%Geographic Allocation +194%, -50% +10%, -10%Activity +65%, -38% +22%, -22%Load Factor +37%, -41% +23%, -21%Emission Factor +96%, -29% +16%, -14%
Category
Georgia Institute of Technology
Monte Carlo SimulationGenerate random inputs
PopulationActivity
Load FactorEmission Factor
Geographic Allocation
Run NONROAD n times
Calculate statistics on outputs
AverageStandard Deviation
Skew
Georgia Institute of Technology
Georgia Institute of Technology
Monte Carlo Simulation
20.0020.5021.0021.5022.0022.5023.0023.5024.0024.5025.00
0 1000 2000 3000
# Runs
Tons
Per
Day
AvePMBasePM
4.00
5.00
6.00
7.00
8.00
9.00
10.00
0 500 1000 1500 2000 2500 3000
# Runs
Tons
Per
Day Std
DevofPM
Monte Carlo SimulationGeorgia 1999 Summer Day Statewide NONROAD Output Results
InputDistribution THC NOx CO PMNormal 204 205 2581 24Triangle 220 214 2804 26Uniform 227 222 2931 28InputDistribution THC NOx CO PMNormal 24 29 29 30Triangle 23 27 28 28Uniform 26 33 32 33
Average (Tons Per Day)
Standard Deviation as % of Average (%)
Georgia Institute of Technology
Monte Carlo SimulationUncertainty of NONROAD Output Results for State and Counties
Maximum Minimum AverageTHC 35 19 23 24NOx 33 20 29 29CO 38 19 28 29PM 37 22 31 30
County Emissions State Emissions
Standard Deviation for Emissions as % of Average
Georgia Institute of Technology
• Uncertainty differences between counties and pollutants driven by dominant equipment types
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
THC CO
% o
f Tot
al N
ON
RO
AD
Em
issi
ons
AC Refrigeration
Pleasure Craft Outboard
Pleasure Craft Inboard
Recreational Equipment
Railroad Equipment
Industrial Equipment
Lawn & Garden Equipment
Logging Equipment
Lawn & Garden Equipment
Recreational Equipment
Agricultural Equipment
Construction Equipment
Commericial Equipment
Airport Ground Support Equipment
Monte Carlo Simulation
NOx PM
Georgia Institute of Technology
Monte Carlo SimulationPM % Change From Base Case
0 80 16040 Miles±
LegendPM % CHANGE
-80 to -60
-60 to -30
-30 to 0
0 to 30
30 to 72
AUGUSTA
COLUMBUS
ATLANTA
MACON
NOx % Change From Base Case
0 80 16040 Miles±
LegendNOx % CHANGE
-84 to -60
-60 to -30
-30 to 0
0 to 30
30 to 72
AUGUSTA
COLUMBUS
ATLANTA
MACON
Georgia Institute of Technology
Monte Carlo Simulation
0 80 16040 Miles±
LegendPM SD % of Ave
22 to 24
24 to 28
28 to 32
32 to 36
36 to 37
AUGUSTA
COLUMBUS
ATLANTA
MACON
1999 NONROAD PM Uncertainty1999 NONROAD NOx Uncertainty
0 80 16040 Miles±
LegendNOx SD % of Ave
19 to 21
21 to 24
24 to 27
27 to 30
30 to 33
AUGUSTA
COLUMBUS
ATLANTA
MACON
Georgia Institute of Technology
Conclusions
Georgia Institute of Technology
• Uncertainties of NONROAD emissions as calculated in this study were between 24% and 30% (standard deviation as % of mean) for the state of Georgia for THC, NOx, CO, and PM
• Uncertainties of input parameters were often positively skewed uncertainties of output emissions also positively skewed
• Using the Monte Carlo approach in this study resulted in approximately 10% increase of calculated emissions for all studied pollutants.
• Uncertainty analysis of NONROAD model does NOT account for all possible uncertainties underestimation?
On-going and Future Work
• Improvement of distributions of NONROAD input parameters and sampling methods
• Sensitivity analysis of CMAQ model to NONROAD emissions uncertainties in Georgia
• Uncertainty of aircraft, locomotive, commercial marine emissions rest of emissions inventory
• Uncertainty analysis of CMAQ model results
Georgia Institute of Technology
Acknowledgements
Much thanks to…• NONROAD Uncertainty Experts Panel: Ms. Archana
Agrawal (CARB), Dr. Christopher Frey (NC State), Dr. Christian Lindhjem (Environ), Mr. Scott Southwick (GA EPD), and Ms. Kirstin Thesing(Pechan)
• EPA, Georgia Depart of Natural Resources (DNR), and Georgia Power for funding this study
Georgia Institute of Technology