the impact of trajectory starting heights on the mura trajectory source apportionment (tsa) method

ARTICLE IN PRESS

1352-2310/$ - se

doi:10.1016/j.at

�Correspondfax: +1530 752

E-mail addr

Atmospheric Environment 41 (2007) 7022–7036

www.elsevier.com/locate/atmosenv

The impact of trajectory starting heights on the MURAtrajectory source apportionment (TSA) method

Stephanie Lee�, Lowell Ashbaugh

Crocker Nuclear Laboratory, University of California, One Shields Ave, Davis, CA 95616, USA

Received 18 December 2006; received in revised form 2 April 2007; accepted 3 May 2007

Abstract

Trajectory source apportionment (TSA) methods have been used in many research projects to attempt to identify the

sources of pollution. Hybrid Single Particle Lagrangian Integrated Trajectories (HYSPLIT) is a popular model for use in

various TSA methods. One of the options in this model is to choose a starting height. Very little research is available to

assist a user in making this choice. This paper evaluates starting heights of 10, 50, 100, 250, and 500m on the accuracy of

the Multi-Receptor (MURA) method using artificial sources for three different simulations. It was found that using

ensembles of trajectories in the MURA method appear to average out most of the biases found from different trajectory

starting heights up to the 500m tested.

r 2007 Elsevier Ltd. All rights reserved.

Keywords: Trajectory analysis; HYSPLIT; Source apportionment; MURA; Trajectory starting height

1. Introduction

Trajectory source apportionment (TSA) methodshave been used in many research studies to identifysources of pollution. These methods use calculatedback trajectories and various statistics to identifythose regions that impact the receptor. Some methodsare more accurate than others as shown in Lee andAshbaugh’s (2007a, b) recent work. The originalresearch on this method used the AtmosphericTransport and Dispersion (ATAD) model to calcu-late trajectories (Heffter, 1980). ATAD calculatedaverage winds within the mixed boundary layer, anddid not allow the user to choose the starting height.

e front matter r 2007 Elsevier Ltd. All rights reserved

mosenv.2007.05.005

ing author. Tel.: +1 916 421 6482;

4107.

ess: [email protected] (S. Lee).

More recently, the Hybrid Single Particle LagrangianIntegrated Trajectories (HYSPLIT) model has beenwidely used to calculate back trajectories. One of thechoices for the model is the trajectory starting height(HYSPLIT_4, 2003). Little information is availableto help a new user decide on this choice. Someresearchers use a height of 500m with the logic thatthey want to start well within the boundary layer(Xu et al., 2006; Chen et al., 2002). Others use aheight of 10m to start their trajectories within thereceptor sampling zone (Lee and Ashbaugh,2007a, b). Gebhart et al. (2005) found that there weresome directional biases in trajectories started atdifferent heights from Big Bend National Park. Theyalso found that the trajectories started higher aboveground tended to move faster and went back fartherthan those started at lower heights (Gebhart et al.,2005).

.

www.elsevier.com/locate/atmosenv

dx.doi.org/10.1016/j.atmosenv.2007.05.005

mailto:[email protected]

ARTICLE IN PRESSS. Lee, L. Ashbaugh / Atmospheric Environment 41 (2007) 7022–7036 7023

This paper tests the impact of starting heights onthe accuracy of the Multi-Receptor (MURA)method developed by Lee and Ashbaugh(2007a, b). Back trajectory starting heights of 10,50, 100, 250, and 500m (AGL) were used togetherwith an artificial concentration field to test howdifferent starting heights impact the accuracy of theMURA method.

2. Method

2.1. Data

Daily artificial SO2 concentrations were generatedfor the year 2002 using the HYSPLIT model version4 (HYSPLIT_4, 2003) with wind fields from theEDAS_80km model (ARL, 2004). The model wasused in the puff mode and incorporates drydeposition but does not account for wet depositionor chemical transformation. In the puff mode, puffsof pollution released from a designated sourceexpand until they exceed the grid cell and then thepuff splits into multiple puffs each with their ownshare of the pollution (HYSPLIT_4, 2003). Multiplesources with unique emission rates and stack heightscan be designated and the pollution from eachsource can be tracked separately. Plume rise is nottaken into account as the temperature of theemission cannot be entered. The resulting SO2

concentration field, then, represents an inert tracerwith dry deposition for removal. This selection wasmade deliberately to minimize the effects ofchemical transformations. The same three simula-tions (Table 1) used in Lee and Ashbaugh (2007b)comparison are used here to test the effect of

Table 1

Source locations and emission strengths for three simulations

Simulation Source Sou

Simulation One Mohave Power Plant 38,

Navajo Power Plant 63,

Phelps Dodge Copper Smelter 7

Carbon Power Plants I & II 230,

Simulation Two Western Resources Inc. 58,

TVA Paradise Plant 181,

Mill Creek Generating 45,

Cincinnati Gas & Electric, Miami Fort 78,

Bowen Steam Electric 140,

Duke Power Marshall Plant 74,

Simulation Three SS1 60,

SS2 60,

trajectory starting height on the accuracy of theMURA method. Two simulations use actual sourcelocations and emissions (Airdata, 2005). The thirdsimulation uses completely artificial sources. Dailyconcentrations were obtained at each IMPROVEsite from the calculated concentration field, and thecontribution from each source was likewise ob-tained. The concentrations at each site were assessedto determine how much was contributed by eachsource.

HYSPLIT_4 (HYSPLIT_4, 2003) was then usedto compute back trajectories from each IMPROVEsite using data from the EDAS meteorologicalmodel (ARL, 2004). Back trajectory calculationsdiffer from the dispersion calculations as the backtrajectory model follows a single air particle back-wards in time whereas the dispersion model followspuffs that expand with time (HYSPLIT_4, 2003).Vertical motion was used in the default mode. Foreach simulation, five different analyses were con-ducted using trajectories initiated every 6 h at aheight of 10, 50, 100, 250, and 500m (AGL) and runbackwards 5 days. Trajectories that arrived at thereceptor during a measurement period at or abovethe 80th percentile of simulated concentrations aredesignated high-incident trajectories.

2.2. MURA trajectory source apportionment (TSA)

method

Lee and Ashbaugh’s (2007a) MURA methodconsists of two parts. The first step is to identifysource regions that contribute to high concentra-tions at sampling sites (receptors). This is referred toas the high incident trajectories (HIT) calculation

rce strength (tons yr�1) Location Stack height (m)

639 35.1N, 114.6W 153

878 36.9N, 111.4W 236

819 33.4N, 100.6W 10

000 26.8N, 100.6W 153

247 39.3N, 96.1W 180

065 37.3N, 86.9W 182

051 38.1N, 85.9W 180

086 39.1N, 84.8W 180

179 34.1N, 84.9W 304

538 35.6N, 80.9W 180

000 35.5N, 86.0W 180

000 45.0N, 77.0W 180

ARTICLE IN PRESSS. Lee, L. Ashbaugh / Atmospheric Environment 41 (2007) 7022–70367024

(Lee and Ashbaugh, 2007a):

HITij ¼ Hij=Tij ,

where Hij is the number of HIT passing through thegrid cell, Tij the total number of trajectories passingthrough the grid cell, i, j the cell designation(coordinates).

The HIT parameter is filtered for significanceusing a binomial test with a 99% confidence level.Those HIT levels not significantly exceeding 20%are set to zero. This value was chosen because thehigh incident days are defined as those days at orabove the 80th percentile, which in turn means thatwe are looking at those concentration values thatreside in the highest 20%. This filter eliminatesthose grid cells whose HIT values are not statisti-cally significant. The remaining cells are contoured.These contoured areas are designated potentialsource regions (PSRs). Large contoured areas canbe left as they are and designated as one PSR ormay be divided into multiple PSRs using saddleareas as boundaries. Saddle areas occur in a PSRwhen the HIT values decrease and then increase.

The second part of the MURA method identifiesthe relationship between each PSR and eachreceptor by calculating the fraction of each recep-tor’s trajectories that pass through each PSR forboth the sample days and high incident days. So, forthe sample days the fraction of trajectories passingthrough PSR S and terminating at receptor R is

Table 2

Receptors used in Simulation One

Site code Location L

BADL Badlands, SD 43

BAND Bandelier, NM 35

BIBE Big Bend, TX 29

BRCA Bryce Canyon, UT 37

CANY Canyonlands, UT 38

CHIR Chiricahua, AZ 32

GICL Gila, NM 33

GRBA Great Basin, NV 39

GRCA Grand Canyon, AR 35

GRSA Great Sand Dunes, CO 37

GUMO Guadalupe Mountains, TX 31

MEVE Mesa Verde, CO 37

MOZI Mount Zirkel, CO 40

PEFO Petrified Forest, AZ 35

ROMO Rocky Mountain, CO 40

SAGO San Gorgonio, CA 34

TONT Tonto, AZ 33

UPBU Upper Buffalo, AR 35

WEMI Weminuche, CO 37

given by

Rrs ¼ Srs=Sr,

where Srs is the number of sample day trajectoriesending at receptor R that passed through PSR S,and Sr the total number of sample day trajectoriesending at receptor R.

For high incident days, the fraction of trajectoriespassing through PSR S and terminating at receptorR is given by

R�rs ¼ Hrs=Hr,

where Hrs is the number of high incident trajectoriesending at receptor R that passed through PSR S, Hr

the total number of high incident trajectories endingat receptor R.

High R and R* values indicate high fractions ofthe sample day and HIT passing through that PSRfor the receptor in question. A high R�rs value mayindicate that PSR S is potentially an importantsource to receptor R.

In the MURA TSA method, the first passidentifies PSRs that ‘‘might be important’’. SincePSRs are created from all the receptors, somereceptors have more high incident trajectories thatpassed through an individual PSR than other re-ceptors and some PSRs have only a few trajectoriestraversing it from each receptor allowing it to passthe binomial test. Lee and Ashbaugh (2007a) foundempirically that if a PSR did not have at least 10%

atitude Longitude Elevation (m)

.7435 �101.9412 736

.77967 �106.2664 1988

.30277 �103.178 1067

.6184 �112.1736 2481

.45873 �109.8209 1798

.00893 �109.3891 1555

.22043 �108.2351 1776

.00518 �114.2161 2066

.97311 �111.9841 2267

.72491 �105.5186 2498

.83301 �104.8094 1672

.19842 �108.4907 2172

.53833 �106.6765 3243

.078 �109.7683 1766

.27833 �105.5457 2760

.1924 �116.9013 1726

.64935 �111.1088 775

.82587 �93.20291 723

.65935 �107.7998 2750

ARTICLE IN PRESS

Fig. 1. HIT contours with the numbers designating the PSRs for Simulation One. Starting heights are: (a) 10m, (b) 50m, (c) 100m,

(d) 250m, and (e) 500m.

S. Lee, L. Ashbaugh / Atmospheric Environment 41 (2007) 7022–7036 7025

ARTICLE IN PRESS

Table 3

Rrs and R�rs values for Simulation One for 10m starting height

PSR BADL BAND BIBE BRCA CANY CHIR GICL

Rrs R�rs Rrs R�rs Rrs R�rs Rrs R�rs Rrs R�rs Rrs R�rs Rrs R�rs

1 0.04 0.13 0.69 0.75 0.15 0.05 0.75 0.97 0.56 0.95 0.51 0.41 0.62 0.51

2 0 0.01 0.19 0.21 0.03 0.03 0.02 0.01 0.07 0.09 0.35 0.51 0.40 0.34

3 0.08 0.26 0.04 0.03 0.12 0.15 0 0 0.01 0.01 0.03 0.06 0.03 0.044 0.06 0.22 0.16 0.28 0.81 0.91 0 0.01 0.01 0.01 0.13 0.36 0.16 0.42

PSR GRBA GRCA GRSA GUMO MEVE MOZI PEFO


1 0.58 0.98 1 1 0.71 0.93 0.28 0.11 0.98 1 0.37 0.75 0.88 0.98

4 0 0.01 0.01 0.01 0.08 0.14 0.57 0.90 0.02 0.01 0.02 0.01 0.04 0.05

PSR ROMO SAGO TONT UPBU WEMI

Rrs R�rs Rrs R�rs Rrs R�rs Rrs R�rs Rrs R�rs

1 0.35 0.64 0.48 0.79 0.85 0.79 0.02 0.05 0.79 0.99

2 0.02 0.02 0.01 0.04 0.54 0.88 0 0.01 0.14 0.183 0.05 0.04 0 0 0 0 0.16 0.28 0.03 04 0.04 0.02 0 0 0.03 0.01 0.12 0.26 0.04 0.028 0.01 0 0 0 0 0 0.24 0.33 0 0

Bold face indicates ‘‘primary’’ PSRs for each receptor.

Table 4




1 0.03 0.11 0.63 0.77 0.12 0.01 0.75 0.98 0.61 0.98 0.27 0.18 0.46 0.43

2 0 0.01 0.18 0.18 0.04 0.01 0.02 0.01 0.06 0.08 0.36 0.55 0.39 0.29

3 0.08 0.25 0.02 0.02 0.11 0.16 0 0 0 0 0.02 0.03 0.02 0.014 0.05 0.19 0.13 0.26 0.77 0.90 0 0.01 0.01 0 0.13 0.36 0.14 0.34

14 0.29 0.55 0.02 0.02 0.02 0.03 0 0 0.01 0.01 0 0.01 0.01 0



1 0.54 0.97 1.00 1.00 0.53 0.74 0.16 0.10 0.97 1.00 0.37 0.77 0.84 0.98

4 0 0 0.01 0.01 0.08 0.15 0.53 0.88 0.03 0.01 0.02 0.02 0.04 0.055 0 0 0 0 0.02 0.05 0.11 0.30 0.01 0 0 0 0.02 0.0215 0 0 0 0 0.46 0.70 0.02 0.02 0.01 0.01 0 0 0.01 0.02



1 0.38 0.68 0.38 0.73 0.70 0.58 0.01 0.05 0.77 0.95

2 0.02 0.01 0.01 0.04 0.50 0.85 0 0.01 0.12 0.173 0.04 0.04 0 0 0 0 0.17 0.32 0.02 04 0.04 0.01 0 0 0.04 0.02 0.20 0.31 0.04 0.048 0.01 0 0 0 0 0 0.21 0.28 0 09 0 0 0 0 0 0 0.15 0.24 0 0


S. Lee, L. Ashbaugh / Atmospheric Environment 41 (2007) 7022–70367026


of at least one receptor’s high incident trajectoriestraversing it, it was unimportant in terms of locatingsources. The second pass identifies the ‘‘primary’’PSRs that have the top R�rs values for each receptor.Lee and Ashbaugh (2007a, b) found in theircomparison of the MURA method with Ashbaughet al.’s (1985) conditional probability method andthe Lee and Ashbaugh (2007b) SIRA method thateach method missed some sources by one or twogrid cells and that the accuracy of all three methodsgreatly improved with the addition of a one-degreebuffer around the designated PSR. If a sourceresided outside the PSR but within the one-degreebuffer, it was considered a source for that PSR.Thus that buffer will be used in this analysis of thetrajectory starting heights.

To summarize, three simulations tested whethertrajectory starting heights from 10 to 500m affectedthe accuracy of the MURA TSA method. This wasdone by comparing the R�rs values for each receptor/PSR combination to the artificial concentration

Table 5


PSR BADL BAND BIBE BRC

Rrs R�rs Rrs R�rs Rrs R�rs Rrs

1 0.05 0.16 0.60 0.74 0.13 0.06 0.83

2 0 0.01 0.21 0.21 0.04 0.02 0.03

3 0.10 0.36 0.04 0.04 0.13 0.14 0

4 0.06 0.23 0.13 0.25 0.74 0.90 0

11 0.20 0.43 0.02 0.02 0.01 0 0

PSR GRBA GRCA GRSA GUM

Rrs R�rs R�rs Rrs R�rs Rrs

1 0.53 0.97 1.00 1.00 0.53 0.75 0.15

2 0 0.01 0.07 0.10 0.17 0.17 0.09

3 0 0 0 0 0.06 0.03 0.18

4 0 0 0 0 0.08 0.14 0.52

12 0 0 0 0 0.30 0.45 0.02

PSR ROMO SAGO TON

Rrs R�rs Rrs R�rs Rrs

1 0.38 0.69 0.50 0.81 0.64

2 0.03 0.05 0.02 0.06 0.57

3 0.05 0.03 0 0 0

4 0.03 0 0 0 0.04

9 0 0 0 0 0


fields. The R�rs value is the fraction of a receptor’shigh incident trajectories that have passed throughthe PSR. The artificial concentration field is ameasure of the percent contribution of eachartificial source to each receptor. A method wasconsidered successful for a receptor if it couldidentify the correct sources in the correct order.

3. Results

The geographical area of the continental UnitedStates, southern Canada, and Northern Mexico isdivided into 0.51� 0.51 grid cells. Only trajectoriesresiding in the area bounded by 15–601N latitudeand 30–1501W longitude are considered in thisanalysis.

3.1. Simulation One

Table 1 lists the location and emissions (Airdata,2005) of the point sources used to create the

A CANY CHIR GICL

R�rs Rrs R�rs Rrs R�rs Rrs R�rs

0.98 0.69 0.97 0.29 0.19 0.45 0.44

0.07 0.08 0.13 0.36 0.55 0.42 0.30

0 0 0 0.03 0.03 0.02 0.01

0.01 0.01 0 0.13 0.31 0.14 0.31

0 0.01 0 0 0.01 0.01 0

O MEVE MOZI PEFO

R�rs Rrs R�rs Rrs R�rs Rrs R�rs

0.07 0.97 1.00 0.38 0.75 0.82 0.97

0.05 0.19 0.30 0.05 0.05 0.32 0.14

0.33 0.03 0 0.02 0.01 0.01 0

0.86 0.03 0 0.02 0.02 0.05 0.05

0 0.01 0 0 0 0 0

T UPBU WEMI

R�rs Rrs R�rs Rrs R�rs

0.56 0.02 0.06 0.74 0.96

0.90 0 0.01 0.14 0.19

0 0.17 0.33 0.02 0

0.03 0.17 0.33 0.04 0.03

0 0.42 0.57 0 0

ARTICLE IN PRESS

Table 7




1 0.05 0.18 0.58 0.73 0.12 0.06 0.96 0.98 0.68 0.98 0.27 0.24 0.46 0.52

2 0.01 0.03 0.17 0.15 0.04 0.02 0.02 0.01 0.04 0.05 0.36 0.52 0.33 0.19

4 0.07 0.23 0.13 0.24 0.74 0.86 0 0 0.01 0 0.14 0.31 0.17 0.35

10 0 0.01 0 0 0.17 0.25 0 0.02 0 0 0.01 0.06 0.01 0.05

16 0.18 0.41 0.02 0 0.02 0.01 0 0 0 0 0 0 0.01 0.02



1 0.53 0.98 1.00 1.00 0.49 0.75 0.19 0.10 0.89 1.00 0.39 0.83 0.81 0.98

4 0 0 0 0 0.10 0.14 0.57 0.90 0.03 0.01 0.02 0 0.06 0.05



1 0.39 0.75 0.74 0.94 0.65 0.56 0.05 0.08 0.68 0.94

2 0.02 0.04 0.01 0.05 0.48 0.83 0.01 0.02 0.12 0.17

3 0.01 0.03 0 0 0 0 0.36 0.61 0 0

4 004 0 0 0 0.05 0.01 0.22 0.47 0.04 0.04

7 0 0 0 0 0 0 0.17 0.24 0 0.01


Table 6




1 0.03 0.14 0.59 0.74 0.12 0.05 0.96 1.00 0.69 0.98 0.32 0.28 0.49 0.53

2 0 0.01 0.22 0.25 0.05 0.02 0.03 0.08 0.08 0.13 0.38 0.57 0.38 0.29

3.1 0.06 0.21 0.04 0.05 0.08 0.10 0 0 0 0 0.02 0.02 0.02 0.014 0.07 0.25 0.12 0.22 0.72 0.88 0.01 0.02 0.01 0.01 0.13 0.29 0.15 0.32

11 0.25 0.52 0.02 0.01 0.01 0 0 0 0.01 0 0.01 0.02 0.01 0



1 0.57 0.98 1.00 1.00 0.51 0.76 0.17 0.09 0.91 1.00 0.38 0.84 0.84 0.98

4 0 0 0 0 0.08 0.14 0.55 0.90 0.03 0.01 0.03 0.01 0.04 0.04



1 0.39 0.71 0.39 0.75 0.74 0.64 0.02 0.04 0.69 0.92

2 0.05 0.08 0.03 0.11 0.63 0.90 0 0 0.15 0.213 0.01 0.03 0 0 0 0 0.18 0.33 0 04 0.04 0.01 0 0.01 0.05 0.03 0.24 0.40 0.04 0.067 0 0 0 0 0 0 0.45 0.53 0 0




artificial SO2 concentration field for 2002. Theconcentrations were evaluated at all IMPROVEsites, and 19 of them (Table 2) were chosen foranalysis based on the highest average 80th percentilevalues in the artificial data. The MURA methodwas applied to the 19 IMPROVE sampling recep-tors for each back trajectory starting height.Trajectories associated with high artificial SO2

concentrations were identified for each receptorand the HIT parameters were calculated. Contouredareas were designated as PSRs and these PSRs wereused in the Rrs and R�rs calculations to determinehow often each PSR affected each receptor on highincident days and throughout the year.

Table 9

Receptors used in Simulations Two and Three

Site code Location

BRIG Brigantine, NJ

DOSO Dolly Sods Wilderness, WV

GRSM Great Smokey Mountains, TN

MACA Mammoth Cave, KY

ROMA Cape Romain, SC

SHEN Shenandoah, VA

SIPS Sipsy Wilderness, AL

WASH Washington, DC

Table 8

Percent contribution of each source to each receptor for

Simulation One

Site Percent contribution

Mohave Navajo Phelps

Dodge

Carbon

I&II

BADL 22 48 0 30

BAND 15 72 4 9

BIBE 0 1 0 99

BRCA 27 71 0 2

CANY 17 81 0 2

CHIR 13 24 19 44

GICL 14 50 5 31

GRBA 84 14 0 2

GRCA 8 92 0 0

GRSA 14 76 2 8

GUMO 6 12 3 79

MEVE 10 85 1 4

MOZI 26 66 1 7

PEFO 10 87 1 2

ROMO 23 66 2 9

SAGO 77 18 1 4

TONT 0 3 97 0

UPBU 6 13 2 79

WEMI 13 78 1 8

Fig. 1 shows the HIT contours for all five startingheights. As can be seen in this figure, there is littledifference between each of the graphs. PSR 1through 8 share similar areas for all heights andPSR designations above 8 over the various heightsdo not necessarily share similar areas. PSR 1contains the Mohave and Navajo Power Plantsources, PSR 2 contains the Phelps Dodge source,and PSR 4 has the Carbon I & II (Carbon) powerplants. These three PSRs consistently identify thesources no matter which trajectory starting height isused. The remaining PSRs are not as consistentamong the different starting heights. In other words,they sometimes appear depending on the trajectorystarting height, but do not identify sources at anytrajectory starting height.

Tables 3–7 show the Rrs and R�rs for each startingheight and Table 8 lists the artificial source contribu-tions to each receptor. These values are used todetermine whether the MURA method can correctlyidentify the sources in the right order for each receptor.The 10, 50, and 500m starting heights work equallywell, identifying the correct sources in the correct orderfor 16 of the 19 receptors. It misses the Navajo andMohave sources for BADL and UPBU and identifiesthe sources in the wrong order for CHIR. The 100 and250m starting heights are only slightly less accurate.They correctly identify the sources for the same 16receptors except GUMO, which does not identify itsmuch smaller Navajo source, though it does consis-tently identify its primary Carbon source through allthe starting heights. In this simulation, any startingheight up to and including 500m would be acceptableand would give almost the same results.

3.2. Simulation Two

The next simulation is in the eastern UnitedStates. Table 1 lists the location and emissions

Latitude Longitude Elevation (m)

39.465 �74.4492 5

39.1053 �79.4261 1182

35.6334 �83.9416 811

37.1318 �86.1479 235

32.941 �79.6572 5

38.5229 �78.4348 1079

34.3433 �87.3388 286

38.8762 �77.0344 15

ARTICLE IN PRESS

Fig. 2. HIT contours with the numbers designating the PSRs for Simulation Two. Starting heights are: (a) 10m, (b) 50m, (c) 100m,

(d) 250m, and (e) 500m.


ARTICLE IN PRESS

Table 10

Rrs and R�rs values for Simulation Two for 10m

PSR BRIG DOSO GRSM MACA ROMA SHEN SIPS WASH

Rrs R�rs Rrs R�rs Rrs R�rs Rrs R�rs Rrs R�rs Rrs R�rs Rrs R�rs Rrs R�rs

1 0.17 0.35 0.72 1.00 1.00 1.00 0.87 0.94 0.40 0.63 0.70 0.94 0.72 0.97 0.42 0.71

2 0.03 0.11 0.09 0.20 0.14 0.10 0.09 0.06 0.54 0.85 0.07 0.07 0.13 0.13 0.06 0.06


Table 12




1 0.20 0.38 0.72 1.00 0.66 0.38 0.37 0.17 0.47 0.74 0.78 0.97 0.30 0.46 0.35 0.65

1.1 0.04 0.14 0.23 0.47 0.76 0.91 0.38 0.20 0.16 0.38 0.20 0.45 0.64 0.95 0.09 0.19

1.2 0.10 0.24 0.23 0.22 0.33 0.36 0.65 0.80 0.17 0.31 0.19 0.27 0.44 0.65 0.17 0.39


Table 11




1 0.19 0.39 0.73 1 1 1 0.56 0.29 0.52 0.77 0.81 0.96 0.70 0.97 0.38 0.71

1.1 0.09 0.25 0.21 0.20 0.36 0.42 0.66 0.81 0.17 0.31 0.21 0.29 0.45 0.65 0.18 0.34


Table 13




1 0.27 0.49 0.77 1.00 1.00 1.00 1.00 1.00 0.39 0.76 0.75 0.98 0.75 0.99 0.40 0.73

2 0.04 0.06 0.09 0.27 0.21 0.16 0.10 0.04 0.33 0.63 0.09 0.16 0.16 0.17 0.06 0.08



(Airdata, 2005) of the point sources used to createthe artificial SO2 concentration field for 2002. EightIMPROVE receptors (Table 9) were chosen foranalysis based on the highest average 80th percentilevalues in the artificial data.

Fig. 2 shows the HIT contours for all five startingheights. Again there are minor differences in thegraphs for the different starting heights. PSRs 1 and2 share similar areas over the various startingheights and PSR designations over 2 do notnecessarily share similar areas over the different

starting heights. PSR 1 contains the TVA, MillCreek, Duke, and Bowen sources, and sometimescontains the Miami Fort source.

Tables 10–14 show the Rrs and R�rs for eachstarting height and Table 15 lists the artificial sourcecontributions to each receptor. In this case, thehigher the trajectory starting height, the morespread out the HIT contours. Because of this, thereis more of a difference in the accuracy of theMURA method than there was in Simulation One.Several of the sources in this simulation are

ARTICLE IN PRESS

Table 14




1 0.32 0.58 0.84 1.00 1.00 1.00 1.00 1.00 0.56 0.89 0.79 0.98 0.72 1.00 0.44 0.82


Table 15


Simulation Two


WRI TVA Mill Creek Miami Fort Duke Bowen

BRIG 2 37 12 16 15 18

DOSO 1 36 10 13 16 24

GRSM 0 11 1 2 0 86

MACA 0 81 10 2 0 7

ROMA 1 11 3 7 34 44

SHEN 1 45 9 13 17 15

SIPS 0 52 6 5 3 34

WASH 1 39 11 20 10 19


relatively close to each other. Thus, when the HITcontours cover more area, they encompass more ofthe sources. This is especially important for theMiami Fort source for the trajectory startingheights of 10 and 50m. In both of these cases, theMiami Fort source is one and a half degrees fromPSR 1. At the 100m height it is one degree fromPSR 1 and for the remaining 250 and 500m startingheight, it is a half degree from PSR 1. Thus for the250 and 500m height, the MURA method correctlyidentifies the sources in the correct order for allreceptors. The trajectory starting height of 10mcorrectly identifies sources for only five of the eightreceptors because it does not identify the MiamiFort source for BRIG, DOSO, and WASH.However, the Miami Fort source only contributesbetween 10% and 12% to these three receptors,making it a relatively minor source. The trajectorystarting heights of 50 and 100m identify the correctsources in the correct order for only three and onereceptors, respectively. For the 50m starting height,the MURA method identifies the correct sources inthe wrong order for five of the eight receptors, andmisses the Miami Fort source for the same receptorsas the 10m starting height. For the 100m startingheight, the correct sources were identified in thewrong order for six of the eight receptors, and itfailed to identify the Bowen source for WASH. The

250 and 500m would be the best choices for startingheights for this simulation, followed closely by the10m starting height.

3.3. Simulation Three

The third simulation is in eastern part of theUnited States, and uses two artificial sources of equalstrength placed so that one source was located inTennessee and one in Canada. Table 1 lists thelocation and emissions of the two point sources usedto create the artificial SO2 concentration field for2002. The same eight IMPROVE receptors (Table 9)used in Simulation Two were used in this analysis.

Fig. 3 shows the HIT contours for all five startingheights. PSRs 1, 3, 4, and 5 share similar areasacross the trajectory starting heights. PSR designa-tions over 5 do not necessarily share similar areasacross the starting heights. PSR 5 contains the SS1source and PSR 1 contains the SS2 source.

Tables 16–20 show the Rrs and R�rs values for eachstarting height and Table 21 lists the artificial sourcecontributions to each receptor. As in SimulationTwo, the higher the trajectory starting height, themore spread out the HIT contours. But in thissimulation that does not affect the accuracy of theMURA method. All starting heights perform fairlywell with the 10m starting height correctly identify-ing sources for seven of the eight receptors, the 50,100, and 250m starting heights correctly identifyingsources for all eight, and the 500m starting heightcorrectly identifying sources for six of the eightreceptors. In the case of the 10m starting height forDOSO, PSR 1 and 5 have R�rs values of 0.38 and0.37, respectively, suggesting that both sourcesaffect DOSO fairly equally. But the SS1 sourcecontributes 75% of DOSO’s high artificial SO2

concentrations where the SS2 source contributes25%. So both sources are correctly identified but inthe wrong order. In the case of the 500m startingheight for GRSM and SIPS, the MURA methodfails to identify a source for either receptor. Thus inthis simulation, trajectory starting heights up to

ARTICLE IN PRESS

Fig. 3. HIT contours with the numbers designating the PSRs for Simulation Three. Starting heights are: (a) 10m, (b) 50m, (c) 100m,

(d) 250m, and (e) 500m.


ARTICLE IN PRESS

Table 16

Rrs and R�rs values for Simulation Three for 10m



1 0.45 0.65 0.18 0.38 0.07 0.05 0.03 0.03 0.14 0.09 0.22 0.32 0.03 0.02 0.31 0.48

3 0.02 0.03 0.17 0.26 0.21 0.19 0.06 0.10 0.13 0.22 0.13 0.24 0.07 0.10 0.05 0.08

4 0.04 0.05 0.32 0.50 0.17 0.05 0.07 0.08 0.11 0.17 0.24 0.46 0.04 0.07 0.13 0.19

5 0.07 0.11 0.30 0.37 0.89 1.00 0.59 0.92 0.20 0.40 0.25 0.41 0.63 0.90 0.17 0.26


Table 17




1 0.28 0.43 0.12 0.27 0.05 0.02 0.02 0.02 0.06 0.02 0.12 0.18 0.02 0.03 0.18 0.36

1.1 0.16 0.30 0.07 0.13 0.03 0.04 0.02 0.01 0.02 0.02 0.08 0.15 0.01 0.02 0.12 0.22

5 0.06 0.10 0.28 0.38 0.92 1.00 0.53 0.89 0.22 0.46 0.24 0.41 0.70 0.97 0.13 0.23

8 0.09 0.17 0.41 0.54 0.42 0.26 0.14 0.20 0.33 0.41 0.32 0.54 0.18 0.24 0.16 0.22

10 0.03 0.06 0.07 0.05 0.15 0.17 0.09 0.23 0.33 0.48 0.08 0.09 0.14 0.10 0.05 0.07


Table 18




1 0.32 0.51 0.13 0.32 0.05 0.04 0.02 0.02 0.08 0.04 0.14 0.21 0.02 0.02 0.22 0.38

4 0.08 0.08 0.42 0.59 0.14 0.04 0.07 0.08 0.12 0.17 0.29 0.50 0.06 0.13 0.14 0.18

5 0.07 0.17 0.35 0.49 0.94 1.00 0.51 0.90 0.34 0.57 0.27 0.48 0.72 0.93 0.13 0.23

7 0.22 0.34 0.05 0.13 0.03 0.03 0 0 0.05 0.03 0.07 0.07 0.02 0.02 0.10 0.17

8 0.05 0.08 0.03 0.04 0.08 0.10 0.04 0.08 0.25 0.36 0.07 0.10 0.08 0.13 0.07 0.09

9 0.02 0.04 0.03 0.03 0.08 0.08 0.06 0.13 0.46 0.65 0.04 0.06 0.08 0.06 0.02 0.03


Table 19




1 0.31 0.50 0.13 0.26 0.04 0.03 0.03 0.02 0.08 0.04 0.14 0.20 0.03 0.05 0.19 0.35

4 0.09 0.14 0.41 0.57 0.28 0.10 0.19 0.22 0.13 0.18 0.30 0.50 0.08 0.19 0.16 0.23

5 0.07 0.10 0.25 0.38 0.84 0.98 0.50 0.88 0.19 0.41 0.20 0.39 0.62 0.91 0.11 0.22

6 0.27 0.40 0.06 0.11 0.02 0.02 0 0 0.06 0.03 0.06 0.07 0.02 0.06 0.09 0.15

7 0.03 0.09 0.14 0.20 0.31 0.24 0.14 0.36 0.21 0.39 0.12 0.19 0.16 0.16 0.06 0.10

8 0.02 0.05 0.04 0.04 0.09 0.05 0.05 0.08 0.17 0.33 0.04 0.03 0.07 0.07 0.03 0.04



ARTICLE IN PRESS

Table 20




1 0.30 0.46 0.11 0.25 0.03 0.01 0.03 0.03 0.06 0.02 0.14 0.21 0.02 0.03 0.20 0.38

4 0.04 0.04 0.23 0.32 0.15 0.05 0.07 0.13 0.08 0.15 0.18 0.30 0.05 0.10 0.08 0.16

5 0.13 0.22 0.32 0.45 0 0 0.73 0.96 0.33 0.60 0.26 0.43 0 0 0.16 0.29

5.1 0.05 0.07 0.10 0.15 0.27 0.15 0.11 0.25 0.20 0.40 0.09 0.13 0.12 0.17 0.05 0.07

8 0.35 0.46 0.05 0.13 0.03 0.03 0.01 0.02 0.05 0.03 0.07 0.08 0.02 0.04 0.13 0.17

9 0.03 0.05 0.04 0.03 0.08 0.07 0.04 0.11 0.23 0.39 0.03 0.03 0.05 0.05 0.03 0.03

10 0.02 0.03 0.05 0.10 0.06 0.17 0.08 0.15 0.02 0.03 0.04 0.11 0.17 0.06 0.03 0.04


Table 21


Simulation Three


SS1 SS2

BRIG 24 76

DOSO 75 25

GRSM 100 0

MACA 100 0

ROMA 95 5

SHEN 72 28

SIPS 100 0

WASH 49 51


250m would be appropriate, with the best resultscoming from starting heights of 50–250m.

4. Conclusions

Overall, using ensembles of trajectories in theMURA method appears to average out most of thebiases found from starting back trajectories atdifferent heights up to 500m. There still are somebiases left that result in mostly minor differences inthe accuracy of the identification of sources ofpollution. Simulation One showed that startingheights of 10, 50, and 500m correctly identifiedsources for more receptors (by one) than the 100and 250m starting heights. But it was shown thatthe source missed for that receptor was relativelyminor compared to the other source for thisreceptor. Simulation Two showed that the startingheights of 250 and 500m worked well for allreceptors, with the 10m working well for five ofthe eight. The three receptors (BRIG, DOSO, and

WASH) that did not correctly identify all thesources at the 10m starting height missed only aminor source. The 50m starting height workedcorrectly for three of the eight receptors. For theremaining five it identified the sources in the wrongorder. The 100m starting height was accurate forone receptor, identified the correct sources in thewrong order for six, and missed a fairly substantialsource for WASH. Simulation Three used the samereceptors as Simulation Two, but with completelydifferent artificial sources. In this simulation allstarting heights up to 250m worked well, but the500m starting height completely missed the sourcesfor GRSM and SIPS.

In summary, any trajectory starting height up to500m would be appropriate for the region tested inSimulation One. The 10, 250, and 500m startingheights were appropriate for the sources used inSimulation Two. For Simulation Three startingheights up to 250m worked well. Analyses of theseknown sources using the other trajectory startingheights resulted in variable performance dependingon the source layout. Since the purpose of thismethod is to identify unknown sources, the accu-racy of the method should not depend on the layoutof the sources. Overall, trajectory starting heights of10 and 250m performed more consistently in oursimulations. Other regions or meteorological driversnot tested by these simulations may have differentresults.

References

AirData, 2005. Access to Air Pollution Data, /http://www.epa.

gov/air/data/S (accessed June 2005).

ARL, 2004. Gridded Meteorological Data Archives, /http://www.arl.noaa.gov/ss/transport/archives.htmlS (accessed 2004).

http://www.epa.gov/air/data/

http://www.epa.gov/air/data/

http://www.arl.noaa.gov/ss/transport/archives.html

http://www.arl.noaa.gov/ss/transport/archives.html

ARTICLE IN PRESSS. Lee, L. Ashbaugh / Atmospheric Environment 41 (2007) 7022–70367036

Ashbaugh, L.L., Malm, W.C., Sadeh, W.Z., 1985. A residence time

probability analysis of sulfur concentrations at Grand Canyon

National Park. Atmospheric Environment 19 (8), 1263–1270.

Chen, L.-W.A., Doddridge, B.G., Dickerson, R.R., Chow, J.C.,

Henry, R.C., 2002. Origins of fine aerosol mass in the

Baltimore–Washington corridor: implications from observa-

tion, factor analysis, and ensemble air parcel back trajectories.

Atmospheric Environment 36, 4541–4554.

Gebhart, K.A., Schichtel, B.A., Barna, M.G., 2005. Directional

biases in back trajectories caused by model and input data.

Journal of Air & WasteManagement Association 55, 1649–1662.

Heffter, J.L., 1980. Air Resources Laboratories Atmospheric

Transport and Dispersion model (ARL-ATAD). Technical

Memo ERL-ARL-81.

HYSPLIT_4, 2003. NOAA ARL HYSPLITModel, /http://www.

arl.noaa.gov/ready/hysplit4.htmlS (accessed 2003).

Lee, S., Ashbaugh, L., 2007a. Comparison of multi-receptor

and single-receptor trajectory source apportionment (TSA)

methods using artificial sources. Atmospheric Environment

41, 1119–1127.

Lee, S., Ashbaugh, L., 2007b. Comparison of the MURA and an

improved single-receptor (SIRA) trajectory source apportion-

ment (TSA) method using artificial sources. Atmospheric

Environment in press.

Xu, J., DuBois, D., Pitchford, M., Green, M., Etyemezian, V.,

2006. Attribution of sulfate aerosols in Federal Class I areas

of the western United States based on trajectory regression

analysis. Atmospheric Environment 40, 3433–3447.

http://www.arl.noaa.gov/ready/hysplit4.html

http://www.arl.noaa.gov/ready/hysplit4.html

the impact of trajectory starting heights on the mura trajectory source apportionment (tsa) method

Documents