Science of the Total Environment 533 (2015) 156–164

Contents lists available at ScienceDirect

Science of the Total Environment

j ourna l homepage: www.e lsev ie r .com/ locate /sc i totenv

Air quality impacts and health-benefit valuation of a low-emissiontechnology for rail yard locomotives in Atlanta Georgia

Boris Galvis a,b,⁎, Michael Bergin a,c, James Boylan d, Yan Huang d, Michelle Bergin c, Armistead G. Russell a

a Georgia Institute of Technology, Atlanta, GA, United Statesb Universidad de La Salle, Bogotá, Colombiac Duke University, Durham, NC, United Statesd Environmental Protection Division— Air Protection Branch — Georgia Department of Natural Resources, Atlanta, GA, United States

H I G H L I G H T S G R A P H I C A L A B S T R A C T

• We quantify the impact on local airquality from two rail yards in Atlantabefore and after conversion to a lowemission technology.

• We calculate the avoided incidence inhealth impacts and the economic valuesaved by the reduction of fine particu-late coming from the rail yards

• We evaluate the cost benefit of the con-version.

⁎ Corresponding author at: Programa de Ingeniería AmE-mail address: bgalvis@unisalle.edu.co (B. Galvis).

http://dx.doi.org/10.1016/j.scitotenv.2015.06.0640048-9697/© 2015 Elsevier B.V. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 18 March 2015Received in revised form 4 June 2015Accepted 17 June 2015Available online xxxx

Editor: D. Barcelo

Keywords:Atlanta, GeorgiaRail yard locomotive emissionsHealth impactsBlack carbonDispersion modelingCost benefit

One of the largest rail yard facilities in the Southeastern US, the Inman and Tilford yards, is located in the north-western section of Atlanta, Georgia alongside other industries, schools, businesses, and dwellings. It is a signifi-cant source of fine particulate (PM2.5) and black carbon (BC) (Galvis, Bergin, & Russell, 2013). We calculate2011 PM2.5 and BC emissions from the rail yards and primary industrial and on-road mobile sources in thearea and determine their impact on local air quality using Gaussian dispersion modeling. We determine thechange in PM2.5 and BC concentrations that could be accomplishedbyupgrading traditional switcher locomotivesused in these rail yards to a lower emitting technology and evaluate the health benefits for comparisonwith up-grade costs.Emissions from the rail yards were estimated using reported fuel consumption data (GAEPD, 2012b) and emis-sion factors previouslymeasured in the rail yards (Galvis et al., 2013). Model evaluation against 2011monitoringdata found agreement between measured and simulated concentrations. Model outputs indicate that the line-haul and switcher activities are responsible for increments in annual average concentrations of approximately0.5 ± 0.03 μg/m3 (39%) and 0.7 ± 0.04 μg/m3 (56%) of BC, and for 1.0 ± 0.1 μg/m3 (7%) and 1.6 ± 0.2 μg/m3

(14%) of PM2.5 at two monitoring sites located north and south of the rail yards respectively. Upgrading theswitcher locomotives at the yards with a lower emitting technology in this case “mother slug” units could

biental y Sanitaria , Universidad de La Salle. Carrera 2a No. 10–70., Bogotá, Colombia.

157B. Galvis et al. / Science of the Total Environment 533 (2015) 156–164

decrease PM2.5 and BC emissions by about 9 and 3 t/year respectively. This will lower annual average PM2.5 con-centrations between 0.3 ± 0.1 μg/m3 and 0.6 ± 0.1 μg/m3 and BC concentrations between 0.1 ± 0.02 μg/m3 and0.2 ± 0.03 μg/m3 at monitoring sites north and south of the rail yards respectively, and would facilitate PM2.5

NAAQS attainment in the area. We estimate that health benefits of approximately 20 million dollars per yearcould be gained.

© 2015 Elsevier B.V. All rights reserved.

1. Introduction

Ambient particulatematter affects public health.Worldwide, it is re-sponsible for approximately 3 million deaths a year (Lim et al., 2012). Itcauses lung cancer (Hystad et al., 2012; Loomis et al., 2013; López-Cimaet al., 2011; Nafstad et al., 2003; Pope et al., 2011; Raaschou-Nielsenet al., 2013; Raaschou-Nielsen et al., 2011; Sax et al., 2013) and cardio-vascular and respiratory diseases (Laden et al., 2006; Nachman andParker, 2012; Pope et al., 2002). It is among the 10 main risk factorsthat impact human health globally, having a greater impact than otherfactors such as high body-mass index, a diet high in sodium or unim-proved sanitation (Lim et al., 2012). Emissions from on-road and non-road diesel engines (e.g., locomotives, marine vessels, heavy-dutyequipment, etc.) are among the main sources of ambient particulatematter inmany urban areas. Diesel emissionsmeasured as diesel partic-ulate matter (DPM) represent around 6% of the nationwide emissionsinventory of particles with an aerodynamic diameter of 2.5 μm or less(PM2.5) in the U.S., ranging from 10% to 40% in urban areas (EPA,2002; Simon et al., 2008).

Many actions have been taken to lower diesel emissions in the last30 years. Improved fuel quality, stringent emission regulations, new ex-haust treatments, and more efficient engine technologies have signifi-cantly decreased nitrogen oxides and particulate emissions from on-road diesels (Hricko et al., 2014; McDonald et al., 2013; Parris, 2003).However, inventory trends in the U.S. show that non-road diesel emis-sions have increased during the same period (Simon et al., 2008). Railyard emissions are expected to have increased due to the use of switch-er locomotives to reassemble freight cars. Switchers are typically oldermodels operating at low-power duty cycles (EPA, 2008, 2011,2011-09-21; Miller et al., 2006) along with the fast growing economicactivity of the rail industry (Chester and Ryerson, 2014; Hricko et al.,2014; Laurits and Christensen Associates, 2009; Leachman and Jula,2012; Rowangould, 2013; Spychalski and Swan, 2004). Currently, theindustry is reducing emissions from yards across the nation, with thesupport of the USDepartment of transportation's CongestionMitigationand Air Quality Improvement Program (CMAQ) and other federal, stateand private funds. Measures taken to reduce emissions involveupgrading switcher locomotives used in rail yards with cleaner alterna-tives for low-speed applications. One approach is replacing the tradi-tional switchers for “mother-slug” locomotives. In a “mother-slug”switcher, a conventional diesel locomotive called “mother” sends theexcess power generated by its diesel electric engine at low speeds to a“slug” which is a locomotive with only traction motors but no enginenor electric generator. The slug contains a large block of ballast to pro-vide sufficient weight for traction. A mother-slug switcher replacestwo conventional switchers and can save approximately 33% of thefuel consumed, meeting EPA tier II/III emission standards (NS, 2011).However, the fuel consumption reduction has not been widely docu-mented. The Georgia Environmental Protection Division (GAEPD)along with the Georgia rail industry is currently pressing forward witha project to replace older switcher locomotives operating in the ‘urbancore’ of Atlanta. This area is currently in non-attainment of the PM2.5 Na-tional Ambient Air Quality Standard (NAAQS). Funding has beenawarded by the Georgia Department of Transportation to the GAEPDthrough the CMAQ (2009) program to employ lower emission technol-ogies at the Inman and Tilford yards.

Changes in air quality resulting from the implementation of emis-sion reduction measures at rail yards have seldom been quantified

(Jaffe et al., 2014). The same is true for their associated health benefits.These tasks have been hindered by significant uncertainties of rail yardemissions and insufficient air quality monitoring data near thesesources. Estimates of emissions from rail yards are typically highly un-certain due to inadequacies in available emission factors and activity in-dicators (Kean et al., 2000). Generic emission factors normally usedmayfail to effectively represent operating conditions, technologies, and yardfleet mix (Galvis et al., 2013), and often activity indicators may not de-scribe the characteristics of a given rail yard (Gould and Niemeier,2009). Previous work carried out by Sierra Research (2011) comparedmodeled diesel particulate matter (DPM) and nitrogen oxides (NOx)ground-level concentrations to measured upwind–downwind concen-tration differences of BC, elemental carbon (EC), organic carbon (OC),PM2.5, and NOx measured at four monitoring stations operated duringthe Roseville Rail yard Air Monitoring Project (RRAMP) in California.Gaussian dispersion models were used to assess the impact of railyard emissions on local air quality. Models were run with rural andurban dispersion coefficients and two different meteorological datasets. In all cases, both measurements and models, found reductions inDPM and NOx impacts over the four-year period of the RRAMP study.Reductions observed were mostly attributed to the decrease of emis-sions at the rail yard over that period. Comparisons of the measuredPM2.5 and NOx concentrationswith simulated DPMandNOx concentra-tions predicted by the models were not in good agreement (Campbelland Fujita, 2009). Feinberg et al. (2011), estimated impacts of theCSXT Rougemere rail yard in Dearborn, Michigan on local air qualityusing a Gaussian atmospheric dispersion model without evaluation ofmodeling results. They developed a bottom-up temporally and spatiallyallocated PM2.5 emissions inventory before and after a retrofit of theswitchers in the yard. Results of the inventory estimated a reductionin PM2.5 emissions from 2007 to 2008, attributed to the retrofits and re-ductions in the sulfur content of the diesel fuel.

Health risk assessments for several rail yards have been carried outby the California Air Resources Board (CARB, 2011). They used emissioninventories and air quality modeling results previously prepared for therail yards, to characterize potential cancer and non-cancer risks associ-ated with exposure to DPM. They estimated impacted areas and ex-posed population associated with different cancer risk levels fordifferent exposure durations. They also reported near-source cancerrisks. GAEPD (2006) assessed benefits of avoidedmortality andmorbid-ity of several emissions control strategies including reducing 10% ofemissions of ground level anthropogenic primary carbon PM2.5 (ECandOC) throughout the state of Georgia. EC is one of themain emissionsfrom rail yard areas. They used the Community Multiscale Air QualityModel to estimate changes in ambient air pollution levels and the Envi-ronmental Benefits Mapping and Analysis Program (BenMAP) (ABT, A,2012) to assess the health benefits and concluded that ground level con-trols of primary carbon significantly reduced exposure and have thehighest health benefits of all the strategies evaluated, saving 223milliondollars annually.

The objectives of this research are to estimate the impact on local airquality of PM2.5 and BC emissions from Tilford and Inman rail yards inAtlanta, GA, and to assess the reduction on the PM2.5 and BC concentra-tions that could be accomplished by converting the switcher locomo-tives at the rail yards to mother-slugs. Emissions from the rail yardsare estimated using available fuel consumption data and emission fac-tors measured for the rail yards (Galvis et al., 2013). First, a 2011 basecase is simulated, and results are compared to measurements of BC

158 B. Galvis et al. / Science of the Total Environment 533 (2015) 156–164

and PM2.5 made at monitoring sites near the rail yards over the sameyear. A scenario simulating that all the switchers at both yards aresubstituted by mother-slug locomotives was evaluated. The change inlocal PM2.5 concentrations between the base and controlled scenariowas used to determine health benefits using BenMAP.

2. Material and methods

2.1. Study location

The Inman and Tilford rail yard complex is located in Norwest At-lanta, Georgia inside the I-285 perimeter freeway (Fig. 1). Inman isoperated by Norfolk Southern (NS) and Tilford by CSX Transporta-tion (CSXT). Descriptions of the rail yard complex can be found inprevious works (GAEPD, 2009a; Galvis et al., 2013). Marietta BlvdNW (~15,000 annual average daily traffic [AADT] approximately)and Bolton Rd (~18,000 AADT) run alongside northwest and north-east of the rail yards, respectively. Marietta Rd NW (~2000 AADT)approximately separates the Inman intermodal section from the ar-rival section of Tilford yard.

During 2011, BC and PM2.5 concentrations were monitored at theFire Station 8 (FS) (33.80176 N,-84.43559 W) ASACA network site(Butler et al., 2003), and at the Dixie Driveline & Spring Co. (DX)(33.79080 N,-84.44026 W) (Fig. 1). PM2.5 measurements were madewith Tapered Element Oscillating Microbalances [TEOMs] (model1400ab; R&P Thermo Scientific, Franklin, MA). BC measurements weremade with Multi-Angle Absorption Photometers [MAAPs] (model5012; Thermo Scientific, Franklin, MA). A description of the monitoringsites and measurements can be found elsewhere (Galvis et al., 2013).These monitoring data, along with PM2.5 concentrations measured by

Fig. 1.Model domain. Layout of the rail yards in gray. Major industrial sources include A)GeneraBuildingMaterials, Inc., D) Cobb County R.L. Suttonwater reclamation facility, E) Atlanta R.M. ClMetals Co. Major streets included in the model are shown. Interstate highways are shown for gJefferson Street (JS).

GAEPD (2013) using a Federal Reference Method sampler (FRM) at FSwere used to evaluate modeling results. Other monitors, includingGAEPD, SEARCH (Hansen et al., 2003), and ASACA (Butler et al., 2003)were used to determine background concentrations.

2.2. Dispersion modeling

Emission impacts from Inman and Tilford rail yards, thenearby smallHowells yard, major surface streets and 8 industrial sources wereassessed using an atmospheric Gaussian dispersion model, theAmerican Meteorological Society/Environmental Protection AgencyRegulatory Model (AERMOD) (EPA version 14134) (EPA, 2012b). Themodel domain was set to cover a 15 km by 12 km area centered at theFS. A 500-meter spaced gridded receptor network was defined in themodel and discrete receptors were set at FS and DX sites. Gridded anddiscrete receptors were assigned terrain elevations using Digital Eleva-tion Model data (USGS, 2012). AERMOD was applied using the urbanoption to account for the urban heat island effect. The roughness lengthparameter was set to 1 m. AERMET (EPA version 14134) was used topreprocess 2011 upper air meteorological data at 12Z GMT from thePeachtree City, GA NWS station and from hourly surface observationsat the Atlanta Hartsfield Airport, GA NWS station. AERSURFACE (EPAversion 13016) with the NCLD92 dataset was used to estimate landuse characteristics and micrometeorological parameters (i.e., albedo,Bowen ratio and surface roughness) (Table S1).

2.2.1. Sources

2.2.1.1. Mobile sources. The Inman and Tilford rail yard complex, thesmall Howells Yard, and the on-road mobile sources on Marietta Blvd,

l Shale Brick Inc. plant, B) Georgia Power CompanyMcDonough-Atkinson plant, C) Lafargeaytonwater reclamation facility, F) Ennis Paint, Inc., G)Mead Packaging Co., and H) Centraleographic reference. Monitoring sites, denoted (o), are Fire station 8 (FS), Dixie (DX), and

Table 1Fuel consumption, emission factors, and emissions from Inman and Tilford rail yards.

Unit Tilford Inman

Base case. traditional switcher locomotivesSwitchers fuel usagea gal/year

(m3/year)600,000(2270)

1,007,000(3810)

BC emission factorb g/gal(g/m3)

2.4 ± 0.2(634 ± 53)

3.1 ± 0.2(819 ± 53)

PM2.5 emission factorb g/gal(g/m3)

4.8 ± 0.6(1268 ± 159)

7.2 ± 0.8(1,9024 ± 211)

Line-haul + switcher BC emissions t/year 3.3 ± 0.3 7.8 ± 0.5Line-haul + switcher PM2.5 emissions t/year 6.6 ± 1.0 18.1 ± 2.0

Scenario. Mother-slug switchersMother-slug switchers fuel usage gal/year

(m3/year)475,000(1800)

560,000(2220)

Mother-slug switchers PM2.5 and BC emission factorsc g/gal(g/m3)

2.9 ± 0.4(766 ± 106)

1.6 ± 0.4(423 ± 106)

Line-haul + mother-slug switchers PM2.5 emission t/year 5.14 ± 0.6 11.8. ± 1.4Line-haul + mother-slug switchers BC emission t/year 3.2 ± 0.3 5.5 ± 0.5

a (GAEPD, 2012b).b (Galvis et al., 2013).c (EPA, 2010a).

159B. Galvis et al. / Science of the Total Environment 533 (2015) 156–164

Marietta Rd, and Bolton Rd (Fig. 1) were defined in the model as multi-ple volume sources. Inman and Tilford yards were defined each as twovolume sources (Inman-A, Inman-B, Tilford-A, and Tilford-B) whileHowells was treated as a single volume source. Emissions from linehaul and switcher operations were split, but used the same source pa-rameters (Table S2). The release height and initial vertical coordinatefor rail yard sources was set to 4.6 m, which is an estimated averageheight of the diesel locomotive engines in the rail yards (Table S2).The initial lateral coordinates (Table S2) were estimated from the railyards' width and length (EPA, 1995; GAEPD, 2012a). Bolton Rd andMar-ietta Rd are represented in the model as three volume sources each.Marietta Blvd is represented as a total of 27 volume sources, corre-sponding to eleven 50m, ten 120m, four 300m, and two ~1500m seg-ments. Relatively fine segments are defined close to FS and coarsefurther away. On-road emission release heights and initial vertical coor-dinates are set to 2.4 m, an estimated average height of vehicles in thearea. The initial lateral coordinates were calculated from each segmentwidth and length (Table S2).

2.2.1.2. Industrial sources. Emissions from themajor industrial sources inthe domain were modeled. Seven facilities were modeled as pointsources, where the stack information was obtained from IntegratedAir Information Platform (IAIP) or fromAeromatic Information RetrievalSystem (AIRS) (Table S3). Ennis Paint wasmodeled as a volume source,with parameters estimated following GAEPD (2012a) and EPA (1995).Central Metals Co is simulated as three point sources with 1/3, 1/2,and 1/6 of total emissions, respectively.

Table 2Emissions from major industrial and mobile sources at the modeling domain.

PM2.5

emissionsBCemissions

[t/year] [t/year]

Bolton Rd 0.3 0.1Marietta Blvd 1.2 0.4Marietta Rd 0.4 0.1General Shale Brick Inc. plant 24.9 0.5Cobb County R.L. Sutton water reclamation facility 36.6 0.7Atlanta R.M. Clayton water reclamation facility 9.5 0.2Mead Packaging Co. 19.1 0.4Central Metals Co. 7.3 0.1

2.2.2. EmissionsEmissions from the rail yards (Table 1) were calculated by multiply-

ing PM2.5 and BC rail yard specific emission factors (REF) measured in aprevious study (Galvis et al., 2013) by the estimated 2011 fuel con-sumption in the modeling domain (GAEPD, 2012b). The fuel consump-tion in the domain was calculated separately for switchers (SFCD) andline-haul locomotives (LHFCD). The SFCD for Inman yard was obtainedfrom GAEPD (2012b).We used the result of the adjusted tonnage meth-od, which is based on link-level line-haul tonnage data and yard andfleet specific information provided by NS. The Tilford switcher yardSFCD was obtained by multiplying the number of switchers in the yard(10), by the system average switcher fuel consumption obtained fromGAEPD (2012b) updated to 2011. Fuel usage for mother-slug switcherswas calculated as 67% of 2011 SFCD. This was done to account for fuelsavings of the new technology.

LHFCDwas obtained for each rail yard by dividing the gross tonmiles(GTM) transported in the modeling domain (GD) by the system-widefuel combustion efficiency (η) as follows:

LHFCD gal½ � ¼ GD GTM½ �η GTM=gal½ � ð1Þ

where GD was calculated as the GTM transported in the county (GC)times the ratio of the track miles in the modeling domain (TD) to thetrack miles in the county (TC) as follows:

GD GTM½ � ¼ GC GTM½ � � TD miles½ �TC miles½ � ð2Þ

Table 3Model evaluation metrics.

PM2,5 BC

FS DX FS DX

Average ratio 1,2 1,1 1,0 1,2Fractional bias 0,0 −0,1 −0,3 0,0Agreement index 0,5 0,4 0,7 0,4Root mean square error (μg/m3) 5,7 6,7 0,8 1,1Average bias (μg/m3) 0,2 −1,2 −0,4 0,0Average error (μg/m3) 4,7 5,4 0,6 0,8Normalized bias 0,0 −0,1 −0,2 0,0Average fractional bias 8% −5% −15% −2%Normalized average error 39% 41% 37% 61%Fractional average error 40% 42% 40% 54%

Fig. 2. Source apportionment for BC and PM2.5 at FS and DX.

160 B. Galvis et al. / Science of the Total Environment 533 (2015) 156–164

η was calculated by dividing the GTM transported system-wide (GS) bythe system-wide fuel consumption (FCS), as follows:

η GTM=gal½ � ¼ GS GTM½ �FCs gal½ � : ð3Þ

GC, TD, and TC were provided for each rail yard by NS and CSXT com-panies GAEPD (2009a). GS and FCS are data contained in NS (2012) andCSXT (2012) Class I Railroad R-1 Annual Report to the Surface Transpor-tation Board (Table S4). Line-haul fuel usage was 779,000 gal/year(2950 m3/year) and 1,500,000 gal/year (5680 m3/year) for Tilford andInman respectively. These values were used to calculate total rail yardemissions.

Two types of emission factors were reported by (Galvis et al., 2013),one for the mix of sources inside the rail yards, (i.e. trucks, cranes andlocomotives) and another for switcher and line haul locomotives. A spe-cific emission factor was reported for each of the rail yards. In this workwe applied the emission factor for switcher and line haul locomotives toestimate rail yard emissions, given that fuel consumption from trucksand other sources inside the intermodal rail yards was not available,and our focus is on assessing switcher emissions. This could underesti-mate rail yard emissions.

Emissions from rail yard sources were split in proportion to theirsize. Both Inman-A and Inman-B are assumed to each produce half ofthe switcher and line-haul emissions from the Inman yard, whileTilford-A and Tilford-B are assumed to produce two-thirds and one-third of the of the switcher and line-haul emissions of the Tilford yardrespectively, based on approximate physical size of each. Emissions ofmother-slug switchers were calculated using PM2.5 estimates of fuelconsumption and emission factors reported previously (EPA, 2010a;GAEPD, 2009b). Uncertainties in emission factors were considered inour emission inventory.

The on-road mobile emissions from Bolton Road (between JamesJackson Parkway and Marietta Blvd), Marietta Rd, and Marietta Blvdwere obtained from Atlanta Regional Commission link-based VehicleMiles Traveled (VMT) database for 2010 (ARC, 2011) (Table 2).Marietta

Blvd is a four-lane arterial road with high volume of heavy-duty truckstransporting goods to and from the rail yard; therefore, its emissions areconsiderably larger than Bolton Rd andMarietta Rd which are two-laneminor collector roads. The emissions for each segment of the roadswereset to be proportional to its length relative to the total length of the road(Table S5). BC emissions are a proportion to PM2.5 emissions calculatedusing ratios reported by EPA (2012a) and traffic splits between dieseland gasoline vehicles (ARCADIS, 2005) (Table S5).

For industrial sources, PM2.5 emission rateswere estimated based oninformation contained in the CERR emission inventory and the GAEPDpermitting database. Whenever PM2.5 emissions were not available,PM10 emissions or PM emissions were modeled (Table S6). BC emis-sions are proportional to PM2.5 emissions calculated using ratios report-ed for each type of industrial activity by EPA (2012a) (Table 2).

2.2.3. Background concentrationsBackground concentrations were obtained frommonitoring data re-

ported by the Southeastern Aerosol Research and Characterization Net-work (Hansen et al., 2003) at Jefferson Street (JS) (33.777627°N,-84.416672°W), which is situated well away from the rail yards andthe other major sources being modeled (Fig. 1). They measure PM2.5

with a TEOM and BC with an Aethalometer. Wavelet analysis(Daubechies, 1992) was used to separate the low frequency compo-nents of five minute average PM2.5 and BC concentrations. A linear re-gression between local minima of the low frequency componentsproduced five-minute background concentrations that were averagedby hour, by day of theweek, and bymonth. Background annual averageconcentrations input to AERMOD in this application are approximately9.9 μg/m3 of PM2.5 and 0.52 μg/m3 of BC.

2.3. Health impacts

BenMAPwas used to assess the avoided health impacts and their as-sociated economic value. The reduction in PM2.5 concentrations

Fig. 3. Spatial distribution of annual average concentrations of a) BC, b) PM2.5 from all sources in the domain and c) PM2.5 from the rail yards. Units of the isolines are μg/m3. Industrialsources include (A) General Shale Brick Inc. plant, (B) Georgia Power CompanyMcDonough-Atkinson plant, (C) Lafarge BuildingMaterials, Inc., (D) Cobb County R.L. Sutton water recla-mation facility, (E) Atlanta R.M. Clayton water reclamation facility, (F) Ennis Paint, Inc., (G) Mead Packaging Co., and (H) Central Metals Co.

161B. Galvis et al. / Science of the Total Environment 533 (2015) 156–164

accomplished by changes to switcher locomotives at both rail yardsalong with population calculated for the model domain using datafrom the Atlanta census county division (Census, 2010) were used asmain inputs. BenMAP calculates health related benefits usingconcentration-response (C-R) functions. C-R functions (Table S7) relatea change in the concentration of a pollutantwith a relative change in theincidence of a health endpoint. Next BenMAP calculates the economicvalue of avoided health effects multiplying the incidence in health ef-fects by a monetary value of the health effect. We used the EPA-default options for PM health impact assessments to obtain incidenceand valuation results (EPA, 2010b). We used the value of statistical life(VSL) recommended by the EPA Science Advisory Board (EPA, 2010b)to calculate the health benefits of avoided mortality.

3. Results and discussion

3.1. Model evaluation

Model results reproduce detailed observed data with low error andbias, especially for BC (Table 3) providing confidence that rail yard

emissions are being successfully simulated (comprehensive evaluationmetrics are shown in supplemental materials). Model results as com-pared to measured PM2.5 and BC show average ratios close or equal to1with very low fractional bias (Table 3). Annual average concentrationsestimated with AERMOD at FS and DX are within 8% and 20% of mea-sured PM2.5 and BC concentrations, respectively. In summary, modelevaluation metrics show that PM2.5 concentrations are slightlyunderestimated at both sites, whereas BC concentrations areunderestimated at FS but well represented at DX. Discrepancies are at-tributed, in part, to AERMOD limitations when reproducing concentra-tions close to the sources (Holmes and Morawska, 2006) and touncertainty in on-road mobile source emissions, especially at FS forthis site is influenced by a major road. Further evaluation of model re-sults is shown in the supplemental material.

3.2. Source apportionment

Apportionment of BC and PM2.5 from AERMOD results indicates thatthe line-haul and switcher activities in the rail yards are the most im-portant source of BC in the domain. They account for approximately

Fig. 4. Spatial distribution of annual average PM2.5 reduction by replacing switcher locomotives with new mother-slug switchers. Units of the isolines are μg/m3.

Table 5Annual reductions in health costs and premature mortality valuation by implementationof mother-slug switchers.

162 B. Galvis et al. / Science of the Total Environment 533 (2015) 156–164

0.5 ± 0.03 μg/m3 (39%) and 0.7 ± 0.04 μg/m3 (56%) of BC at FS and DXrespectively, and for approximately 1 ± 0.1 μg/m3 (7%) and 1.6 ±0.2 μg/m3 (14%) of PM2.5 at FS and DX respectively (Fig. 2). Calculationsindicate a greater impact on PM2.5 at DX and FS came from the Inmanyard. Approximately 5% and 13% of PM2.5 at FS and DX respectivelyare apportioned to Inman yard, whereas 2% and 1.5% of PM2.5 at FSand DX respectively are attributed to Tilford yard. Line-haul activitiesat both yardswere found to have slightly higher impacts than switchers,accounting for roughly 4% and 9% of PM2.5 at FS and DX respectively.Switchers at both yards were responsible for roughly 4% and 5.5% ofPM2.5 at FS and DX respectively.

Table 4Annual avoided incidence by implementation of mother-slug switchers.

Health endpoint | Age group Mean reductionin incidence± standard deviation

Mortality, all cause | 30–99 1 ± 0.1Mortality, all cause | 25–99 2.1 ± 0.5Mortality, all cause | infants 0.01 ± 0.01Emergency room visits, asthma | 0–99 0.7 ± 0.2HA, all respiratory | 65–99 0.2 ± 0.03HA, asthma | 0–17 0.02 ± 0.01HA, chronic lung disease | 18–64 0.08 ± 0.01HA, all cardiovascular (less myocardial infarctions) | 65–99 0.3 ± 0.03HA, all cardiovascular (less myocardial infarctions) | 18–64 0.2 ± 0.03Work loss days | 18–64 166 ± 12Minor restricted activity days | 18–64 966 ± 85Acute bronchitis | 8–12 1.4 ± 0.8Lower respiratory symptoms | 7–14 18 ± 5Upper respiratory symptoms | 9–11 25 ± 10Asthma exacerbation, cough | 6–18 340 ± 164Asthma exacerbation, shortness of breath | 6–18 121 ± 128Asthma exacerbation, wheeze | 6–18 40 ± 16

HA: hospital admissions.

3.3. Air quality impact evaluation

The spatial distributions of BC correspond to the rail yard layoutwhereas distributions of PM2.5 also correspond to the location of the in-dustrial sources (Fig. 3a and b). BC concentrations of approximately1 μg/m3 outline the rail yards up to 2 km from the center of the complex(Fig. 3a).

Endpoint | Valuation method | Age group Mean yearly benefits ±standard deviation [$]

Mortality | VSL, based on 26 value of life studies | 0–99 19,900,000±14,400,000Hospital admissions, respiratory | COI: med costs +wage loss | 65–99

4700±3300

Hospital admissions, respiratory | COI: med costs +wage loss | 0–64

500±150

Hospital admissions, cardiovascular | COI: med costs +wage loss | 65–99

6200±3100

Hospital admissions, cardiovascular | COI: med costs +wage loss | 18–64

7700±1800

Acute respiratory symptoms | WTP: 1 day, CVstudies | 18–99

66,000±17,000

Lower respiratory symptoms | WTP: 1 day, CVstudies | 0–17

370±170

Upper respiratory symptoms | WTP: 1 day, CVstudies | 0–17

800±500

Work loss days | Median daily wage, county-specific |18–65

31,500±2300

Asthma exacerbation | WTP: bad asthma day | 18–99 4900±6400Emergency room visits, respiratory | COI | 0–99 240±150Acute bronchitis | WTP: 6 day illness, CVstudies | 0–17

670±500

Total 20,000,000±14,400,000

VSL: value of statistical life, COI: cost of illness, WTP: willingness to pay, CV:cardiovascular.

163B. Galvis et al. / Science of the Total Environment 533 (2015) 156–164

The spatial distribution of PM2.5 annual average concentrations overthe domain shows hot spots of 2 to 5 μg/m3 above background, at thecenter of the rail yard complex and near specific sources, east of (D),of (F), and of (G) (Fig. 3b). Elevated impacts of PM2.5 from the railyards are located largely to the northeast of the domain. Annual averagePM2.5 impacts from the rail yards are about 1 μg/m3 up to 1 km north-east from the center of the complex (Fig. 3c).

Reductions of PM2.5 concentrations by replacing switchers withmother-slug units (Fig. 4) are 0.3 ± 0.1 μg/m3 and 0.6 ± 0.1 μg/m3 atFS and DX respectively (i.e. 3% and 5% of total PM2.5 concentration ateach site). PM2.5 reductions of about 1 μg/m3 are located over the railyards and extendmostly toward the northeast of the domain. PM2.5 im-pacts from the switcher locomotives at the rail yards are reduced ap-proximately by 35%. BC from the rail yards would be reduced byapproximately 23% if mother-slug switchers were implemented. BCconcentrations will be reduced by 0.1 ± 0.02 μg/m3 and 0.2 ±0.03 μg/m3 at FS and DX respectively, when the conversions take effect.

3.4. Health incidence and valuation

BenMAP was used to calculate the avoided incidence in health im-pacts and the economic value saved by the reduction in primary PM2.5

concentrations. Annual avoided incidence results (Table 4) are basedon estimates of reduced exposure to PM2.5 of the population in themodel domain. Results show approximately three avoided cases of pre-mature mortality in the 25–99 age group per year and less than oneavoided case for infants. Minor restricted activity days have the highestincidencewith approximately 1200 cases. Reductions in asthma exacer-bation and work loss days are also important.

Economic value is assigned by BenMAP based on specific cost factorsfor each health endpoint. Cost factors correspond to research compiledin BenMAP (ABT, A, 2012). Reductions in primary PM2.5 concentrationsdue to retrofitting switcher locomotives at Inman and Tilford rail yardssave approximately $20 million in annual avoided health costs(Table 5). Avoided mortality accounts for 99% of the savings.

3.5. Cost–benefit

Funding for retrofitting switcher locomotives awarded throughCMAQ and matched by industry are expected to amount to 3 disburse-ments, each of 17.143 million dollars per fiscal year (GAEPD, 2009b).The retrofitted switcher locomotives will remain in service for at least10 years. With a discount rate of 0.75% (Federal discount rate for April2013 when this calculations were made), the resulting positive netpresent value (NPV) of replacing switcher locomotives at Inman andTilford yards with mother slug sets is approximately $140 million dol-lars. This result does not take in to account additional pollutants orother factors such as fuel savings or maintenance costs that could affectthe cash flows of the project.

4. Conclusions

Replacing traditional switchers withmother-slug sets would reducePM2.5 and BC emissions by 7.8 ± 0.9 and 2.4 ± 0.6 t/year. This can leadto a decrease in concentrations of primary PM2.5 and BC of approximate-ly 38% and 29%. Greater reductions can be achieved over the rail yardsand to the northeast of the domain. Reductions in PM2.5 concentrationsdue to converting switcher locomotives to lower emission technologiesmight save approximately $20 million in annual avoided health costsand premature mortality. The measure has a positive NPV of about$140 million dollars through the ten-year period implementation.

AbbreviationsAERMOD American Meteorological Society/Environmental Protection

Agency Regulatory Model

AADT Annual average daily trafficBenMAP Environmental Benefits Mapping and Analysis ProgramBC black carbonCMAQ CongestionMitigation and Air Quality Improvement ProgramDPM diesel particulate matterEC elemental carbonGAEPD Georgia Environmental Protection DivisionMAAP Multi-Angle Absorption PhotometerMother-slug a conventional diesel locomotive called “mother” coupled

to a locomotive with only traction motors or “slug”.NPV net present valueOC organic carbonPM2.5 fine particulate matterTEOM tapered element oscillating microbalanceVMT vehicle miles traveled

Acknowledgments

This work was funded by the Congestion Mitigation and Air QualityImprovement (CMAQ) Program through Georgia DOT (GDOT ProjectNumber: CSCMQ-0009-00{186) P.I. # 0009186), and Georgia DNR.Funding also was provided by Georgia Power. We specially thank GilGrodzinsky from GAEPD for his collaboration in finding data. Furtherthanks to Universidad de La Salle, LASPAU, and COLCIENCIAS for provid-ing a fellowship to B. Galvis.

References

ABT, A, 2012. BenMAP environmental benefits mapping and analysis program — user'smanual Retrieved Jan, 2013, from http://www.epa.gov/air/benmap/models/BenMAPManualOct2012.pdf (Oct).

ARC, 2011. Atlanta Regional Commission link-based vehicle miles traveled data RetrievedNov, 2012, from http://www.atlantaregional.com/.

ARCADIS, 2005. Bolton/Moores Mill Transportation and Circulation Study Retrieved May,2013, from https://dl.dropboxusercontent.com/u/18763207/bmt%20final%20report.pdf.

Butler, A.J., Andrew,M.S., Russell, A.G., 2003. Daily sampling of PM2.5 in Atlanta: results ofthe first year of the assessment of spatial aerosol composition in Atlanta study.J. Geophys. Res. Atmos. 108 (D7), 8415. http://dx.doi.org/10.1029/2002JD002234.

Campbell, D., Fujita, E.M., 2009. Roseville Rail Yard Air Monitoring Project (RRAMP). FinalReport Summary of Data QA and Trend Analysis. Desert Research Institute, Reno, NV(Retrieved Oct, 2012, from http://www.placer.ca.gov/Departments/Air/~/media/apc/documents/UP/2009/December/RRAMPFinalUpdate121009.ashx).

CARB, 2011. Railyard health risk assessments and mitigation measures Retrieved Apr,2013, from http://www.arb.ca.gov/railyard/hra/hra.htm.

Census, B., 2010. Population and housing census Retrieved Apr, 2013, from http://www.census.gov/popfinder/.

Chester, M.V., Ryerson, M.S., 2014. Grand challenges for high-speed rail environmental as-sessment in the United States. Transp. Res. A Policy Pract. 61 (0), 15–26. http://dx.doi.org/10.1016/j.tra.2013.12.007.

CMAQ, 2009. EPD railroad related emissions reduction project in Atlanta GA20090013 Re-trieved Apr, 2013, from https://fhwaapps.fhwa.dot.gov/cmaq_pub/View/default.aspx?id=GA20090013.

CSXT, 2012. Class I railroad annual report Retrieved Dec, 2012, from. http://www.stb.dot.gov/econdata.nsf/f039526076cc0f8e8525660b006870c9/8775aca41ef66efb852579db004c9dee?OpenDocument.

Daubechies, I., 1992. Ten Lectures on Wavelets.EPA, 1995. User's guide for the Industrial Source Complex (ISC3) dispersion models Vol1

Retrieved Oct, 2012, from http://www.epa.gov/scram001/userg/regmod/isc3v1.pdf.EPA, 2002. Health assessment document for diesel engine exhaust Retrieved from. http://

www.epa.gov/ttn/atw/dieselfinal.pdf.EPA, 2008. Regulatory impact analysis: control of emissions of air pollution from locomo-

tive engines and marine compression ignition engines less than 30 liters per cylinderRetrieved NOV 2011, from nepis.epa.gov/Exe/ZyPURL.cgi?Dockey=P10024CN.TXT.

EPA, 2010a. Exhaust and crankcase emission factors for nonroad engine modeling com-pression ignition Retrieved Jan, 2013, from http://www.epa.gov/otaq/models/nonrdmdl/nonrdmdl2010/420r10018.pdf.

EPA, 2010b. Valuing mortality risk reductions for environmental policy Retrieved Mar,2013, from http://yosemite.epa.gov/ee/epa/eerm.nsf/vwAN/EE-0563-1.pdf/$file/EE-0563-1.pdf.

EPA, 2011. Rail and locomotives | clean ports USA | US EPA from http://www.epa.gov/cleanschoolbus/ports-rail.htm 2011–09-21.

EPA, 2012a. Report to Congress on black carbon Retrieved Dec, 2012, from http://www.epa.gov/blackcarbon/2012report/fullreport.pdf.

EPA, 2012b. User's guide for the Ams/Epa regulatory model — AERMOD Retrieved Dec,2012, from http://www.epa.gov/ttn/scram/dispersion_prefrec.htm#aermod.

164 B. Galvis et al. / Science of the Total Environment 533 (2015) 156–164

Feinberg, S., Yadav, V., Heiken, J., Turner, J., 2011. Midwest rail study: modeled near-fieldimpacts of emissions of fine particulate matter from railyard activities. Transp. Res.Rec. 2261 (−1), 106–114. http://dx.doi.org/10.3141/2261-12.

GAEPD, 2006. CMAQ/BenMAP-based health-benefits analysis in support of the GeorgiaSIPs for O3 and PM2.5. Paper presented at the Community Modeling and Analysis Sys-tem. CMAS, Chapel Hill, NC (http://www.cmascenter.org/conference/2006/abstracts/marmur_session5.pdf).

GAEPD, 2009a. Dispersion modeling to support the Atlanta PM2.5 SIP local area analysisRetrieved Aug, 2012, from. http://www.gaepd.org/Files_PDF/plans/sip/Appendix_M_PM2.5_Dispersion_Modeling_at_FireSta8.pdf.

GAEPD, 2009b. Voluntary reduction of emissions from railyards in metro Atlanta regionRetrieved Oct, 2012, from. http://www.georgiaair.org/airpermit/downloads/planningsupport/regdev/sips_and_revisions/atlanta_8hr_ozone/appendixt.pdf.

GAEPD, 2012a. Guideline for modeling PM10 ambient concentration in areas impacted byquarry operation producing crushed stone Retrieved Feb, 2013, from. http://www.georgiaair.org/airpermit/downloads/sspp/modeling/quarryguideline_august2012.pdf.

GAEPD, 2012b. Railroad emission inventories. Locomotive emission inventories for theUnited States from ERTAC rail Retrieved Mar, 2013, from. http://www.georgiaair.org/airpermit/downloads/planningsupport/regdev/locomotives/railyard_switcher_semap.xls.

GAEPD, 2013. Ambient monitoring program air quality database Retrieved Feb, 2013,from. http://www.georgiaair.org/amp/amp_query.html.

Galvis, B., Bergin, M., Russell, A.G., 2013. Fuel-based fine particulate and black carbonemission factors from a railyard area in Atlanta. J. Air Waste Manage. Assoc. (63),618–628.

Gould, G., Niemeier, D.A., 2009. Review of regional locomotive emission modeling and theconstraints posed by activity data Retrieved Jan, 2013, from. http://www.escholarship.org/uc/item/3gn498w6.

Hansen, D.A., Edgerton, E.S., Hartsell, B.E., Jansen, J.J., Kandasamy, N., Hidy, G.M.,Blanchard, C.L., 2003. The southeastern aerosol research and characterization study:part 1—overview. J. Air Waste Manage. Assoc. 53 (12), 1460–1471. http://dx.doi.org/10.1080/10473289.2003.10466318.

Holmes, N.S., Morawska, L., 2006. A review of dispersion modelling and its application tothe dispersion of particles: an overview of different dispersion models available.Atmos. Environ. 40 (30), 5902–5928. http://dx.doi.org/10.1016/j.atmosenv.2006.06.003.

Hricko, A., Rowland, G., Eckel, S., Logan, A., Taher, M., Wilson, J., 2014. Global trade, localimpacts: lessons from California on health impacts and environmental justice con-cerns for residents living near freight rail yards. Int. J. Environ. Res. Public Health 11(2), 1914–1941.

Hystad, P., Demers, P.A., Johnson, K.C., Brook, J., van Donkelaar, A., Lamsal, L., …, Brauer,M., 2012. Spatiotemporal air pollution exposure assessment for a Canadianpopulation-based lung cancer case–control study. Environ. Heal. 11, 22. http://dx.doi.org/10.1289/ehp.1002353 (10.1289/ehp.0901623 http://dx.doi.org/10.1186/1476-069X-11-22).

Jaffe, D.A., Hof, G., Malashanka, S., Putz, J., Thayer, J., J. L. F.,…, Pierce, J.R., 2014. Diesel par-ticulate matter emission factors and air quality implications from in-service rail inWashington State, USA. Atmos. Pollut. Res. 5, 344–351. http://dx.doi.org/10.5094/APR.2014.040.

Kean, A.J., Sawyer, R.F., Harley, R.A., 2000. A fuel-based assessment of off-road diesel en-gine emissions. J. Air Waste Manag. Assoc. 50 (11), 1929–1939.

Laden, F., Schwartz, J., Speizer, F.E., Dockery, D.W., 2006. Reduction in fine particulate airpollution and mortality. Am. J. Respir. Crit. Care Med. 173 (6), 667–672. http://dx.doi.org/10.1164/rccm.200503-443OC.

Laurits, R., Christensen Associates, I., 2009. Description of the U.S. freight railroad industryRetrieved February, 2012, from. http://www.stb.dot.gov/stb/docs/CompetitionStudy/Volume%201.pdf.

Leachman, R.C., Jula, P., 2012. Estimating flow times for containerized imports from Asiato the United States through the Western rail network. Transportation ResearchPart E 48 (1), 296–309. http://dx.doi.org/10.1016/j.tre.2011.07.002.

Lim, S.S., Vos, T., Flaxman, A.D., Danaei, G., Shibuya, K., Adair-Rohani, H., …, Ezzati, M.,2012. A comparative risk assessment of burden of disease and injury attributable to67 risk factors and risk factor clusters in 21 regions, 1990?2010: a systematic analysisfor the Global Burden of Disease Study 2010. The Lancet 380 (9859), 2224–2260.

Loomis, D., Grosse, Y., Lauby-Secretan, B., Ghissassi, F.E., Bouvard, V., Benbrahim-Tallaa, L.,…,Straif, K., 2013. The carcinogenicity of outdoor air pollution. Lancet Oncol. 14 (13),1262–1263.

López-Cima, M.F., García-Pérez, J., Pérez-Gómez, B., Aragonés, N., López-Abente, G.,Tardón, A., Pollán, M., 2011. Lung cancer risk and pollution in an industrial regionof Northern Spain: a hospital-based case–control study. Int. J. Health Geogr. 10, 10.http://dx.doi.org/10.1186/1476-072X-10-10.

McDonald, B.C., Gentner, D.R., Goldstein, A.H., Harley, R.A., 2013. Long-term trends inmotor vehicle emissions in U.S. urban areas. Environ. Sci. Technol. 47 (17),10022–10031. http://dx.doi.org/10.1021/es401034z.

Miller, A.R., Peters, J., Smith, B.E., Velev, O.A., 2006. Analysis of fuel cell hybrid locomotives.J. Power Sources 157 (2), 855–861. http://dx.doi.org/10.1016/j.jpowsour.2005.12.051.

Nachman, K.E., Parker, J.D., 2012. Exposures to fine particulate air pollution and respirato-ry outcomes in adults using two national datasets: a cross-sectional study. Environ.Heal. 11, 25. http://dx.doi.org/10.1186/1476-069x-11-25.

Nafstad, P., Haheim, L.L., Oftedal, B., Gram, F., 2003. Lung cancer and air pollution: a27 year follow up of 16 209 Norwegian men. Thorax 58 (12), 1071–1076.

NS, 2011. Biz NS continuing a railroad tradition Retrieved May, 2013, from. http://www.nscorp.com/nscorphtml/bizns/bzns1211/NovDecBizNS_WEB.pdf.

NS, 2012. Class I railroad annual report Retrieved Nov, 2012, from. http://www.stb.dot.gov/econdata.nsf/f039526076cc0f8e8525660b006870c9/d945e6d27425d1f6852579db004cfdbd?OpenDocument.

Parris, T.M., 2003. Trends in motor vehicle fuel economy and emissions. Environment 45(9), 3-3.

Pope, I.C., Burnett, R.T., Thun, M.J., 2002. Lung cancer, cardiopulmonary mortality, andlong-term exposure to fine particulate air pollution. JAMA 287 (9), 1132–1141.http://dx.doi.org/10.1001/jama.287.9.1132.

Pope III, C.A., Burnett, R.T., Turner, M.C., Cohen, A., Krewski, D., Jerrett, M., …, Thun, M.J.,2011. Lung cancer and cardiovascular disease mortality associated with ambient airpollution and cigarette smoke: shape of the exposure–response relationships. Envi-ron. Health Perspect. 119 (11), 1616–1621.

Raaschou-Nielsen, O., Andersen, Z.J., Hvidberg, M., Jensen, S.S., Ketzel, M., Sørensen, M.,…,Tjønneland, A., 2011. Lung cancer incidence and long-term exposure to air pollutionfrom traffic. Environ. Health Perspect. 119 (6), 860–865.

Raaschou-Nielsen, O., Andersen, Z.J., Beelen, R., Samoli, E.g., Stafoggia, M., Weinmayr, G., …,Hoek, G., 2013. Air pollution and lung cancer incidence in 17 European cohorts: pro-spective analyses from the European Study of Cohorts for Air Pollution Effects (ESCAPE).Lancet Oncol. 14 (9), 813–822. http://dx.doi.org/10.1016/S1470-2045(13)70279-1.

Rowangould, G., 2013. Public financing of private freight rail infrastructure to reducehighway congestion: a case study of public policy and decision making in theUnited States. Transp. Res. A Policy Pract. 57 (0), 25–36. http://dx.doi.org/10.1016/j.tra.2013.09.007.

Sax, S.N., Zu, K., Goodman, J.E., 2013. Air pollution and lung cancer in Europe. LancetOncol. 14 (11), e439–e440. http://dx.doi.org/10.1016/S1470-2045(13)70438-8.

Sierra Research, I., 2011. Modeling evaluation study for the Union Pacific J.R. Davis (Rose-ville) rail yard Retrieved Jan, 2013, from. http://www.placer.ca.gov/departments/air/~/media/apc/documents/UP/2011/RailYardMonitoringReport101311.ashx.

Simon, H., Allen, D.T., Wittig, A.E., 2008. Fine particulate matter emissions inventories:comparisons of emissions estimates with observations from recent field programs.J. Air Waste Manage. Assoc. 58, 320–343.

Spychalski, J.C., Swan, P.F., 2004. US rail freight performance under downsized regulation.Util. Policy 12 (3), 165–179. http://dx.doi.org/10.1016/j.jup.2004.04.002.

USGS, 2012. The national map viewer and download platform Retrieved Oct, 2012, from.http://nationalmap.gov/viewer.html.