ARTICLE IN PRESS

0301-4797/$ - se

doi:10.1016/j.je

�CorrespondE-mail addr

Journal of Environmental Management 83 (2007) 213–227

www.elsevier.com/locate/jenvman

A framework for assessing the impact of land use policy on communityexposure to air toxics

Melvin R. Willis�, Arturo A. Keller

Donald Bren School of Environmental Science and Management, University of California, Santa Barbara, CA 93106, USA

Received 28 April 2005; received in revised form 21 January 2006; accepted 15 March 2006

Available online 13 July 2006

Abstract

Our research focuses on the linkage between land use planning policy and the spatial pattern of exposure to air toxics emissions. Our

objective is to develop a modeling framework for assessment of the community health risk implications of land use policy. The modeling

framework is not intended to be a regulatory tool for small-scale land use decisions, but a long-range planning tool to assess the

community health risk implications of alternative land use scenarios at a regional or subregional scale. This paper describes the

development and application of an air toxic source model for generating aggregate emission factors for industrial and commercial zoning

districts as a function of permitted uses. To address the uncertainty of estimating air toxics emission rates for planned general land use or

zoning districts, the source model uses an emissions probability mass function that weights each incremental permitted land use activity

by the likelihood of occurrence. We thus reduce the uncertainty involved in planning for development with no prior knowledge of the

specific industries that may locate within the land use district. These air toxics emission factors can then be used to estimate pollutant

atmospheric mass flux from land use zoning districts, which can then be input to air dispersion and human health risk assessment models

to simulate the spatial pattern of air toxics exposure risk. The model database was constructed using the California Air Toxics Inventory,

1997 US Economic Census, and land assessment records from several California counties. The database contains information on more

than 200 air toxics at the 2-digit Standard Industrial Classification (SIC) level. We present a case study to illustrate application of the

model. LUAIRTOX, the interactive spreadsheet model that applies our methodology to the California data, is available at http://

www2.bren.ucsb.edu/�mwillis/LUAIRTOX.htm.

r 2006 Elsevier Ltd. All rights reserved.

Keywords: Air toxics; Risk assessment; Air pollution; City planning; Regional planning; Strategic environmental assessment

1. Introduction

For the last three decades, public concern aboutenvironmental pollution has become a significant factorin the US political process. The growth of cities and thepopulation exodus to the suburbs concentrated a signifi-cant portion of the population in urban communities inproximity to sources of environmental pollutants.Although the problem of environmental pollution iscomplex, with a myriad of interactions, it is ultimatelyrelated to land use.

e front matter r 2006 Elsevier Ltd. All rights reserved.

nvman.2006.03.011

ing author. Tel.: +1805 6761240; fax: +1 805 6761240.

ess: mwillis@bren.ucsb.edu (M.R. Willis).

Although the concept of incorporating industrial per-formance standards into zoning ordinances received con-siderable early attention by land use planners, its practicewas not widely implemented through the 1970s (Loch-moeller et al., 1975). In the 1970s and early 1980s, thefederal government passed a number of far-reachingenvironmental laws designed to reduce pollutant emissionsto air, water, and land, principally aimed at industrialsources. Over the last three decades, the major influence ofthese federal laws on municipal land use policy was toincrease the importance of environmental planning—thepractice of evaluating the potential impacts of proposedpublic and private activities intended to achieve socio-economic objectives. However, the stringent industrialperformance standards that were adopted as the result of

ARTICLE IN PRESSM.R. Willis, A.A. Keller / Journal of Environmental Management 83 (2007) 213–227214

these federal laws are enforced by state agencies; munici-palities were given no role in the enforcement of thesestandards nor any mandate to incorporate their provisionsinto their zoning ordinances. The result is that municipalland use policy planning in the US focuses on the siting ofindustrial facilities, but generally defers consideration andenforcement of industrial performance standards to stateand federal agencies.

Although the city general plan, supported by its zoningordinance, is the principal policy instrument affecting thespatial pattern of development within a jurisdiction, theweak or lack of direct consideration of industrial perfor-mance in the planning process calls into question theeffectiveness of land use planning and zoning at protectingresidents from exposure to toxic industrial pollutants(Schwab, 1993). Due to the rapidly expanding role of thefederal government during the 1970s and 1980s in the fieldof environmental protection, the local government plan-ning process generally deferred decisions about theappropriate level of industrial emissions to state andfederal regulatory agencies and still primarily relied onthe old concept of perceived general industrial character-istics in the development of its land use planning policiesand controls.

A significant problem in metropolitan areas is chronicpopulation exposure to air toxics1—very small amounts oftoxic chemicals emitted to the atmosphere by industrial,commercial, and transportation sources—resulting inincreased risk of cancer and noncarcinogenic healthhazards. The degree of long-term exposure to air toxicsvaries considerably among individuals within an urbanarea, but is directly related to the proximity of residence toemission sources. Because the siting of both emissionsource and receptor is guided by land use policy, plannersneed to consider the community health risk implications oflong-range development plans.

A major urban air monitoring and evaluation studyfound that air toxics exposure to residents of the LosAngeles air basin resulted in an individual excess cancerrisk of 1.40� 10�3, or about 1400 excess cancer cases permillion population in the Los Angeles basin, assuming alifetime of exposure (South Coast Air Quality ManagementDistrict, 2000). However, the study also found significantspatial variation in the level of individual carcinogenic riskexposure, with the highest risk generally correlated tomajor industrial areas or transportation corridors. Becausethe type, intensity, and location of development is guidedby local land use policy, the Los Angeles air basin studysuggests that land use planning can affect the spatialpattern of air toxics exposure, but has been ineffective

1Air toxics or hazardous air pollutants (HAPs) is the designation given

by the 1990 US Clean Air Act amendments to 188 air pollutants having

the potential to cause adverse environmental and health effects, including

cancer, neurological damage, and birth defects. The California Air

Resources Board uses the designation toxic air contaminant (TAC) for

244 substances known or suspected to be emitted in California having

potential adverse health effects, which includes all EPA-listed HAPs.

overall at protecting residents from significant carcinogenicrisk.Despite the relationship between land use policy and air

toxics risk exposure, review of the literature and availablecase studies found little evidence that community healthrisk assessment and management is a significant compo-nent of the local government land use planning process,and no evidence of any attempt to quantify the health riskimplications of long-range land use planning decisions. Themost comprehensive survey of the risk assessment processin local government is a report by the International City/County Management Association (ICMA, 1997) thatconcluded, ‘‘The definition and process of risk manage-ment still remains a broad, scientific, unclear, andpolitically charged concept for many local governments’’.Using information obtained through telephone andwritten questionnaires from local government officialsinvolved in risk assessment and management, the ICMAfound that risk assessment and management of industrialfacilities was primarily performed by state agencies,and that the role and degree of involvement by localgovernment in this process varied significantly, but wasmainly indirect participation. The 33 case studiesdocumented by the ICMA report all involved activitiesby local governments to assess the human or ecological riskfrom past industrial operations—a reactive process; noexamples were provided of local government takingproactive or preventative actions to protect the publicfrom future industrial development risk, such as throughlong-range land use policy (i.e., the general plan). TheICMA report concluded with the following finding andrecommendation:

In a period of increasing state and local flexibility,diminishing resources, and a growing demandfor enhanced community involvement, local officialsneed to have better access to human healthand environmental risk and decision-making tools,to help their communities better understand theirrisks, to set priorities based on their risks, and tomanage risks in the most effective and inclusive waypossible.

Review of the literature suggests that an effective airtoxics risk assessment framework for land use planning canbe accomplished by linking a geographic informationsystem (GIS) with a risk exposure simulation model. Dentet al. (2000) found that ‘‘the marriage of (GIS andsimulation models) is optimal because it leverages thepredictive capacity of modeling with the data management,analysis, and display capabilities of GIS.’’In Europe, individual risk and societal risk metrics have

been used in land use planning to minimize populationexposure to the effects of an accident (Laheij et al.,2000). Spadoni et al. (2000) describe a GIS-basedmethodology used in Europe to conduct quantified arearisk analysis for land use planning. Sengupta et al. (1996)used a GIS to assess the exposure and population

ARTICLE IN PRESSM.R. Willis, A.A. Keller / Journal of Environmental Management 83 (2007) 213–227 215

health risk of the Greater Bombay region from atmo-spheric pollutants. The GIS was used to create interpolatedmaps of three air pollutants that were then overlain onpopulation density maps to determine the level of popula-tion exposure. Dent et al. (2000) describe a technique bywhich an air dispersion model was coupled with a GIS topredict the pattern of TAC exposure, identify thepotentially exposed population, and estimate the healthrisk burden on that population.

For air toxics exposure simulation, a commonly usedfate and transport model is the EPA’s Industrial SourceComplex model (ISC3), which is a steady-state Gaussianplume model that can be used to assess pollutantconcentrations from a wide variety of sources associatedwith an industrial complex (US Environmental ProtectionAgency, 1995b). This model can account for settling anddry deposition of particles; downwash; point, area, line andvolume sources; plume rise as a function of downwinddistance; and separation of point sources and limitedterrain adjustment. ISC3 operates in both long- and short-term modes.

Review of the literature suggests that the field of landuse planning could benefit from extending the use ofGIS-based linked simulation models into the area ofcommunity health risk management. The major limitationof previous studies to land use planning has been theirfocus on individual facilities; long-range land use policyplanning, however, generally relies on broad categories ofpotential uses. Given the inherent uncertainty in long-rangedevelopment plans about the type and scale of individualland uses, how should the mass flux or volatilizationrates of all pollutants of concern be estimated for aplanning area?

The focus of our research is the relationship of land usepolicy decisions to community health risk from HAPexposure. We believe that long-range land use policydecisions significantly influence the spatial pattern ofpopulation exposure to air toxics, but municipal land useplanning has not explicitly considered the communityhealth risk implications of these policy decisions. This canbe explained by the lack of a suitable framework forassessing the community health risk implications ofalternative land use patterns resulting from the uncertain-ties inherent in the traditional reliance of the planningprocess on broad categories of land use. This paperprovides a modeling framework and methodology forassessing the impact of land use policy on communityexposure to air toxics. The following sections describe asource model for estimating aggregate air toxics emissionrates for broad general categories of industrial andcommercial land use, given local policies about permitteduses in each zone. A case study is then presented in Section3 to demonstrate how the air toxics emission factors sourcemodel can be used in land use policy analysis to simulatethe potential spatial pattern of air toxics exposure andassociated carcinogenic and noncarcinogenic risk in thestudy area.

2. Model development

2.1. Land use emission factors

Long-range planning practice typically relies on land useand zoning designations to guide the spatial configurationand use of private land. The principal regulation of landuse by local government is normally exercised through itsgeneral plan and zoning ordinance—related but distinctdocuments. The general plan is a ‘‘blueprint’’ for the futuredevelopment of the community because it functions as thelocal government’s statement of policy about the types ofdevelopment that will be allowed, the spatial relationshipamong land uses, and the general pattern of futuredevelopment. The zoning ordinance regulates developmentin the present by setting specific standards for allowableuses, lot size, and buildable area (Fulton, 1999).The major characteristic of the practice of zoning is the

division of a municipality into zones or districts withuniform regulation throughout the district, but withdifferent regulations for each district about the type, scale,and intensity of economic activities (Curtin, 1997). Theonly facilities allowed to locate within a zoning districtmust be consistent with one of the subset of permittedclasses of use. For example, heavy industrial zonesgenerally allow only uses involving the processing of rawmaterials (e.g., produce, wood, minerals, etc.) into finishedgoods or intermediate products. Specific uses wouldinclude petrochemical plants, paper mills, steel mills, andcanneries. Light industrial zones typically allow only usesthat assemble intermediate products into finished goods.These uses would include manufacturers of electrical orelectronic equipment, metal fabricating, and wood, paper,or plastic goods.Our research objective was a land use emission factors

source model for estimating the mass volatilization rate orflux of individual air toxics as a function of allowable landuse policy for individual zoning districts, normalized byarea (Q0p;z). The standard reference for air contaminantemission factors is the US EPA’s AP-42 publication (USEnvironmental Protection Agency, 1995a). Use of AP-42

and other emission factor source models generally relies onknowledge about a specific source, including type (point,area, volume, line), and activity level (e.g., energy use,resource input, product output, etc.). The major limitationof these source models for application to land use policyanalysis is their basis on knowledge about specificeconomic activities associated with individual properties,which is impractical because land use policy usuallypermits a broad range of potential future land uses. Forexample, AP-42 input parameters include system type (e.g.,boiler, heater, internal combustion, etc.), fuel type,materials processed, emission control device, and systemactivity or operating levels. However, land use policy onlyspecifies the location and areal extent of broad categoriesof future land use within which a wide range of potentialeconomic activities are allowed, thus necessitating an

ARTICLE IN PRESSM.R. Willis, A.A. Keller / Journal of Environmental Management 83 (2007) 213–227216

emission factors source model that takes into account thisvariation and uncertainty in potential use. Instead of one-dimensional point sources, land use policy analysisemploys zones or districts, two-dimensional spatial unitsthat are area sources of emissions. Each zoning district iscomprised of individual recorded land parcels (i.e.,assessor’s parcels), the smallest spatial unit employed incommunity or regional land use planning. Therefore, anarea source employed in land use policy analysis would bethe aggregate of unit parcels within an area designated by aspecific land use or zoning classification.

Rather than normalizing emission rates by a highlyuncertain future activity level, the normalization variablefor land use emission factors is the area of the parcel ofland on which the source is located. In our modelframework, the normalized pollutant mass volatilizationrate for an individual facility (q0p,f) is defined as

q0p;f ¼qp;f

Af

, (1)

where qp,f is the mass volatilization rate of pollutant p fromfacility f (g s�1), and Af the total area (m2) within theproperty line of the parcel on which the emitting facility islocated. A legal parcel of record is generally referred to asan assessor’s parcel, or simply the ‘‘area within the fence.’’The area of the legal parcel of record is used as thenormalization variable for pollutant mass flux fromindividual facilities because it is the smallest spatial unitused in land use policy, and zoning district boundaries aregenerally drawn to coincide with property lines. However,parcel size may vary significantly between facilities of asimilar type and scale because of differences in locallandscaping and parking requirements and the need toreserve space for potential future expansion, thus introdu-cing a source of variation into the model.

Fig. 1. Venn diagram illustrating a land use inventory.

2.2. Defining land use policy

To predict the normalized mass flux of air toxics as afunction of land use policy, the basic framework of oursource model is

Q0p;z ¼ f ðCÞ, (2)

where Q0p,z is the normalized mass flux of pollutant p

(g s�1m�2) from zoning district z, and C is a dimensionlessproxy variable for land use policy. If ui is the set of allpotential land uses in a region and zj is the set of zoningdistricts specified by land use policy, the land use policyvariable can be defined by an i� j matrix of binaryvariables:

W ¼

c11 � � � c1j

..

. . .. ..

.

ci1 � � � cij

26664

37775, (3)

where ui is permitted in zj only if cij ¼ 1; otherwise, ifcij ¼ 0, ui is not permitted.Permitted land use policy for individual zoning districts

can also be defined by Wj, the vector of binary variables forzj. The only economic activities allowed to locate withinzone zj must be classified as a use consistent with one of thesubset of permitted uses defined by Wj.

2.3. Source model framework

Our emission factors source model uses a probabilisticframework to explain the functional relationship betweenthe normalized air toxics emission rate and land use policysuch that

EðQ0p;zÞ ¼Xi

u¼1

cuzq0p;u

� �f Q0p;z

q0p;u

� �� �� �, (4)

where E(Q0p,z) is the expected value of the normalizedemission rate (g s�1m�2) of pollutant p from zoning districtz, and the functional relationship is the sum of averagenormalized emissions of pollutant p from each land usecategory u (q0p;z),weighted by f Q0p;z

ðq0p;uÞ, the emissionsprobability mass function associated with the likelihoodthat pollutant p is emitted by an economic activityclassified as use u. The binary variable Cuz specifieswhether ui is included in the weighted sum for zj.Fig. 1 illustrates how the emissions probability mass

function is derived using an emissions inventory and countsof facilities for a region. In this example, Fig. 1 does notdepict the spatial relationship of the individual polygons,but is a Venn diagram representing the total inventory ofmanufacturing facilities in region < and the relativedifference in facility counts when grouped into 20 subsetsof land use categories u1;...;20

� �. The size of each box labeled

ui in Fig. 1 represents Nui, the census of facilities in a land

use category ui. The total number of facilities in region < isPNui¼ N. For example, the land use category with the

ARTICLE IN PRESSM.R. Willis, A.A. Keller / Journal of Environmental Management 83 (2007) 213–227 217

largest number of facilities in Fig. 1 is u15; the smallestnumber of facilities is in land use category u9.

Fig. 1 also shows that the 20 land use subsets aregrouped into three zones according toCuz. Although Zone2 allows the most types of uses, the largest number offacilities in the example is permitted in Zone 1. The totalnumber of facilities in each zone is Nz ¼

Piu¼1cuzNu,

where Cuz is the binary variable that specifies whether ui ispermitted in zj.

The stippled portions of the boxes in Fig. 1 representfacilities that emit pollutant p (i.e., qp40). The portionsoutside stippled areas represent facilities that do not emitthe pollutant (i.e., qp ¼ 0). In this simplified example, fourpollutants are emitted; for example, p4 is emitted by thelargest number of facilities, clustered into six of the 20 usesubsets found in the region. Facilities from four use subsets(u6, u7, u11, and u16) do not emit any of the four pollutants,while a few facilities from two use subsets (u13, u18) emitboth p3 and p4, represented by the overlapping stippledareas.

The emissions probability mass function term in Eq. (4)that is used in our model to weight the normalized sum ofemission rates is

f Q0p;zðq0p;uÞ ¼ PfugPfpjug, (5)

where P{u} is the probability that a facility is classified asuse u, and P{p|u} is the conditional probability that thefacility emits pollutant p, given classification as use u. Thisis illustrated in Fig. 1 by the portions overlapped by thestippled areas, the joint occurrence of two attributes of afacility—land use classification (ui) and emission of aspecific pollutant (qp40). Assuming a random distributionamong each subset of facilities in Fig. 1, the probabilitiesare

Pfug ¼Nu

Nz

¼NuPi

u¼1cuzNu

, (6)

Pfpjug ¼Np;u

Nu

. (7)

Substituting into Eq. (4), our emission factors sourcemodel becomes

EðQ0p;zÞ ¼Xi

u¼1

cuzq0p;u

� �P uf gP pju

� �� �� �, (8)

EðQ0p;zÞ ¼Xi

u¼1

ðcuzq0p;uÞ

NuPiu¼1cuzNu

!Np;u

Nu

� !, (9)

EðQ0p;zÞ ¼Xi

u¼1

ðcuzq0p;uÞ

Np;uPiu¼1cuzNu

! !. (10)

Examination of Eq. (10) and Fig. 1 indicates thatvariation of C, the proxy variable for land use policy,modifies the set boundaries, which has the effect ofchanging the probability mass function and, therefore,the weighting assigned to each component of the summa-

tion of normalized emission rates. Using Eq. (10), it ispossible to derive unique air toxic emission factors forindividual zoning districts as a function of local permittedland use policy (W), given an emissions inventory, and acensus of facilities.

2.4. Emissions data

The largest available source of data on toxic chemicalreleases to the environment is the US EPA’s Toxic ReleaseInventory (TRI), which contains reported annual releaserates by individual facilities of approximately 600 desig-nated chemicals to air, water, and land (US EnvironmentalProtection Agency, 2001). The value of this database foremission factor source model development is somewhatdiminished because the TRI reporting requirement islimited to facilities with 10 or more full-time employeesthat generate more than 10,000 lbs year�1 of any designatedchemical or 25,000 lbs year�1 cumulative of all designatedchemicals. This reporting limitation thus excludes the manysmall toxic pollutant generators—printers, dry cleaners,and gas stations—that comprise the fabric of urban landuse.The California Air Toxics Inventory (CATI) is a much

more comprehensive database of toxic chemical releasesthat is an integral component of its statewide air toxicsidentification and control program, though it is limitedgeographically to California and to only air pollutants(California Air Resources Board (CARB), 2002). Emissioninventories are developed by each air pollution controldistrict and submitted to the CARB for inclusion in CATI.Inventory updates are required every 4 years, but shorterintervals may occur. The data used for this study comprisesthe 1996 and 1998 database years, but these data may bederived from slightly earlier emission inventory years.CARB provided a copy of the entire CATI for ourresearch, yielding data on annual air toxic emission ratesfor approximately 600 chemicals from more than 9500facilities throughout California. The data included name,address, and use Standard Industrial Classification (SIC)for each facility. We believe this data can be used as a basisfor estimating emissions in other regions of the US, andthat our approach can serve as a methodology for similarstudies around the world, adjusted for local conditions.The 1997, US Economic Census provided counts of all

manufacturing facilities in California classified by use (USCensus Bureau, 2000). This is the most recent census yearfor industrial facilities and closely coincided with the datesof records in the CATI.A major problem encountered with both the TRI and

CATI was the lack of information about facility parcelsizes necessary to normalize emission rates by area.Although parcel information is available from countyassessors, public access to this data varies widely in formand medium among the 58 counties in California.Although a few counties provide access to parcel informa-tion through the Internet (e.g., Los Angeles County), the

ARTICLE IN PRESSM.R. Willis, A.A. Keller / Journal of Environmental Management 83 (2007) 213–227218

majority requires public retrieval of information by directsearch of paper records at each county courthouse. Becauseof the lack of a centralized repository for parcel data, timeand funding constraints precluded obtaining this data forall facilities in California reporting releases of air toxics.Instead, parcel records were selectively accessed to obtain arepresentative sample of parcel sizes from CATI-reportingfacilities in each land use category. Mining and agriculturaluses were excluded because they are generally rural landuse activities and not typically found in urban areas, thusreducing the total statewide inventory sampled to 6977facilities. To minimize travel costs, the sample was drawnfrom the largest and most industrialized counties innorthern and southern California. The result was a sampleof 1051 facilities in California reporting releases of airtoxics, representing 15% of the total statewide inventory.

The CATI facilities report emissions in lbs year�1, whichwe converted g s�1 for use as input parameters in anatmospheric fate and transfer model. Parcel data obtainedfrom county assessors was similarly converted from acresto m2. The unit for normalized emission rates in ourdatabase is therefore g s�1m�2.

2.5. Land use emission factors source model

Derivation of the probabilistic source model to estimateair toxics emission factors for a region requires aninventory of emitting facilities, categorized by type ofpollutant and land use type; facility emission rates of eachpollutant; parcel sizes for each emitting facility to normal-ize emission rates by area; and counts of facilities,categorized by land use type. This section describes theinteractive spreadsheet model that we created fromCalifornia emissions data for estimating air toxic emissionfactors as a function of local permitted land use policy.

After obtaining the sample of facility parcel sizes, thegeometric mean normalized emission rate for each pollu-tant (q0p;u) was estimated for each land use category at the2-digit SIC level, as well as the upper (98%) and lower(2%) confidence intervals for the mean. Although CATIreported facility use type at the 4-digit SIC level, weaggregated the data at the 2-digit level because parsing thesample at greater detail resulted in confidence intervals forthe predicted emission factors that were two or more ordersof magnitude in width—too impractical for application tosimulation modeling.

LUAIRTOX is an interactive spreadsheet, based onCalifornia emissions and census data, which applies Eq.(10) to the California emissions and census data to estimateemission factors as a function of local land use policy forindividual zoning districts. Land use policy is input toLUAIRTOX using a vector of binary variables (Cz) tospecify permitted uses from among approximately 60categories of manufacturing at the 2-digit SIC level. Eachunique Cz vector results in a model realization thatestimates (EðQ0p;zÞ), the expected value of emission ratesfor a zoning district, normalized by area, for over 200

individual air toxics. The model realization also includesthe upper (98%) and lower (2%) confidence limits for theexpected values. The spreadsheet model and instructionsfor use are available at http://www2.bren.ucsb.edu/�mwillis/LUAIRTOX.htm.Variation of the binary vector Cz adds or deletes land

use categories from the set of permitted uses within a givenzoning district, which has the effect of changing theweighting applied to the normalized emission rates for allpermitted uses in that district. The LUAIRTOX model canbe used to explore the potential health risk implications ofallowing certain types of land use activities within a zoningdistrict by exploring the sensitivity of the aggregateemission factor to changes of the permissibility ofindividual uses.Because of sample size limitations, Cz in the LUAIR-

TOX spreadsheet model currently can only be applied atthe 2-digit SIC level. This level of land use categorizationwill limit the practical value to land use planners, whogenerally require more detail to differentiate amongeconomic activities when setting permitted land use policy.Expansion of the LUAIRTOX model from the 2- to the3-digit SIC level for manufacturing uses would increase thedimensions of the Cz binary vector from 61 to about 325,yielding far better differentiation among uses, but sig-nificantly increasing the width of emission factor con-fidence intervals so as to be impractical for simulationmodeling. The problems associated with the tradeoffbetween model detail and precision can be overcome inthe future by increasing the size of the database to providelarge enough sample sizes at the 3-digit SIC level.The major source of model error results from Af, the

normalization variable. Existing emission inventories, suchas the TRI or CATI databases, do not provide thisinformation, nor is there a centralized repository ofindividual facility parcel data, thus necessitating obtainingthe data directly from local assessor records. Theprocedure is to match facility street address from anemissions inventory with the address associated with aspecific parcel in assessor’s records, which also includesparcel size. Data error can be introduced if the facilityaddress matches only a single parcel of land of a largeindustrial facility encompassing multiple adjacent parcels,or when the address for a small facility, such a dry cleaner,matches a single large parcel of land, such as a shoppingcenter, on which multiple businesses are located.We attempted to avoid the first source of error by

examining surrounding parcels for similar owner name. Ifclassified by the assessor with the same or closely relatedland use code as the parcel with the matched address, theywere added to the total parcel area. The second source oferror was much more difficult to screen and was limited todiscarding parcel area data that intuitively seemed toolarge for the type of business.Another potential source of error is introduced by using

current assessor data for parcel sizes to normalizeemissions data from previous years. The most currently

ARTICLE IN PRESSM.R. Willis, A.A. Keller / Journal of Environmental Management 83 (2007) 213–227 219

available TRI data is 2–3 years old, and the CATI datathat we used to develop the LUAIRTOX model is from1996 and 1998. Changes in facility parcel area in theintervening years would result in normalization error,although we suspect this is a minor source of error.

2.6. Application to environmental risk assessment of land

use policy

The risk assessment process generally followed in the UShas four basic steps: hazard identification, exposureassessment, toxicity assessment, and risk characterization(National Research Council, 1983). For air toxics assess-ment (see Fig. 2), the objective of the hazard identificationstep is to identify sources of hazardous emissions atindividual sites or areas and estimate (Qp), the mass flux orvolatilization rates of individual contaminants of concern.During the exposure assessment step, Qp is input to anatmospheric fate and transport model to predict thespatially variable ground level contaminant concentrations(Cp,x,y) to which receptor populations are exposed. Basedon predicted exposure, individual health risk is character-ized by estimating the rate of chemical intake and dosage atreceptor locations (Ip,x,y), and then comparing these tocarcinogenic and noncarcinogenic toxicity thresholds foreach chemical of concern.

The measure of toxicity of a noncarcinogen is itsreference dose (RfD), which is the maximum amount thatthe human body can absorb without experiencing chronichealth effects, normalized for time and body weight (Watts,1997). Because of the assumption of a threshold dose belowwhich no adverse effects are expected, chronic noncarcino-genic chemical risk is characterized as a dimensionlesshazard quotient (HQp), which is the ratio of average dailychemical intake or exposure level to RfDp. For multiplenoncarcinogenic pollutants, cumulative risk is character-ized by a hazard index (HI) for each organ of the bodyaffected, which is calculated by summing HQp from allsubstances that affect the same organ such that

HIj ¼X

j

HQp;j, (11)

Fig. 2. Risk assess

where p is each chemical of concern and j is the organ ofthe body affected. HQp’s and hazard indices are organ-specific and should not be combined if affecting differenttoxicological endpoints, otherwise the risk will be over-estimated and the potential effects on the receptorpopulation will be overstated.For carcinogens, which are assumed to be hazardous at

all levels of exposure, common toxicity metrics for ingestedcarcinogens at low exposures are slope factor (SF) or cancer

potency factor (CPF); these measure the probability ofcancer per unit dose, assuming a lifetime of exposure.Because no threshold level of exposure is assumed,carcinogenic risk of pollutant p is characterized by theincremental probability of an individual member of theimpacted population developing cancer over a lifetime ofchronic daily exposure to the chemical (CRp). For multiplecarcinogens, the individual cancer risk from simultaneousexposure is assumed to be the sum of the individual cancerrisks from each chemical of concern (Asante-Duah, 1996):

CR ¼X

p

Xj

CRp;j, (12)

where p is the carcinogen of concern and j the pathway orroute of exposure. For this study, only the inhalationpathway was considered. Multiplication of CR by 106

converts the metric to a cancer risk index (CRI) represent-ing the number of excess cases of cancer per million people.

2.7. Assessing land use policy

The LUAIRTOX emission factors source model can becoupled with a number of atmospheric fate and transportmodels, for example the US EPA’s Industrial SourceComplex-Short Term (ISCST3), for risk assessment ofalternative spatial patterns of land development. ISCST3 isa steady-state Gaussian plume model used to assess toxicchemical concentrations in the air from a wide variety ofindustrial sources, including point, line, area and volumesources (US Environmental Protection Agency, 1995b).Our source model can be used to estimate air toxicsemissions for properties designated for future industrialdevelopment, which could then be input to the ISCST3

ment process.

ARTICLE IN PRESS

126 Freeway101 Freeway

33 F

reew

ay

0 4 81 2 6Kilometers

Legend

Facilities

Parks

Schools

Residential

City Limits

Area of Interest

RoadsFreewayMajor roadLocal Road

Ventura, California

N

Fig. 3. Existing land use in study area.

M.R. Willis, A.A. Keller / Journal of Environmental Management 83 (2007) 213–227220

model as area sources. The LUAIRTOX source model canalso be used with more advanced fate and transport modelsprovided other input data are available.

2.8. Case study

To demonstrate application of the land use emissionfactors source model to the risk assessment modelingframework, a case study was developed. The city of SanBuenaventura (Ventura), California, was used to evaluatethe modeling framework for assessing the potentialcommunity health risk implications of long-range landuse policy on community air toxics exposure. Ventura islocated on the coast of California, approximately 75 kmnorthwest of the Los Angeles civic center. It is a medium-sized (approximately 100,000 persons) urban communitywith a mixture of old and new industrial, commercial, andresidential uses in a pattern typical of communitiesundergoing rapid development after World War II.Ventura is located on the Oxnard coastal plain at thelower reach of the east-west trending Santa Clara Rivervalley, bounded by the Pacific Ocean on the west andcoastal foothills on the north. In 2002, the incorporatedcity limits encompassed an area of approximately 5470 ha,which lies within an approximately 21,600 ha area ofinterest that defines the city’s long-range land use planningarea. A portion of the city extends up the north-southtrending Ventura River valley. With prevailing westerlywinds and a Mediterranean climate, local meteorologicalconditions are typically land-sea breezes up or down theriver valleys.

Fig. 3 is a map of the study area land use including themajor sources of air toxics emissions and the location ofpotentially exposed residential areas, schools and parks.The primary sources of air toxics emissions in the Venturaenvirons are a mixture of businesses and on-road vehicletraffic along the network of roads and freeways. Othersources of air toxics emissions are off-road mobile traffic(i.e., rail, aircraft, ships), agriculture, and mineral produc-tion. Because of small annual release rates and/or distancefrom receptor populations, these latter sources wereassumed to be minor contributors to study area ambientconcentrations of air toxics and were ignored in this study.Fig. 4 is the city’s general plan for development within itsarea of interest. The goal of this plan is to guidedevelopment within the city to provide sufficient jobs andhousing to accommodate projected population growththrough the year 2010. Ventura is currently undertakingthe process of updating this plan to extend the planninghorizon to 2020. The results of our modeling methodologycan be used to guide these planning decisions.

Based on information obtained from the CARB, 86commercial, industrial, and institutional facilities withinVentura’s area of interest are listed as stationary sourcesreporting annual releases of air toxics (CARB, 2002). Thisinventory is dominated by numerous small automobile-related sources (gasoline service stations and repair

facilities) typical of most communities. The remainingstationary sources are representative of common light andheavy industrial uses. Two hospitals are included in theemissions inventory. The only stationary sources notrepresentative of most communities are two boat buildingand repair facilities that result from the coastal location ofthe city.Mobile source emissions are not directly reported, but

were estimated for each road segment in the study areausing emission factors derived from statewide annual totalsfor on-road mobile air toxics emissions and total vehiclemiles traveled. Use of these statewide averages assumes nospatial variation in the mix of vehicles and operating speedsthroughout the state roadway network and any deviationof emission rates from the mean on local road segments israndom. These local roadway emission estimates alsoassume a constant hourly and daily traffic flow. Bothassumptions grossly oversimplify actual traffic conditionsalong road segments and would be flawed for use inpredicting acute (i.e., short-term) exposure to air toxics,but are acceptable at a regional scale over a large networkfor predicting chronic (i.e., long-term) population expo-sure. Assuming that traffic—and therefore emissions—onall major road segments will grow in proportion to total

ARTICLE IN PRESS

Ventura 2010 Land Use Plan

Kilometers

City Limits

Sphere of InfluenceArea of InterestTransitional Residential

Res - SFRes - Planned

Planned CommercialPlanned Manufacturing

Professional Office

Park

Mixed Development

Manufacturing - LightManufacturingInstitutionalCommercial - HarborHospitalRes-DowntownDowntown CoreRes - Core

Res - MF

Res - Mobile Home Park

Commercial - GeneralCommercial - Light

Agricultural Use

MAPLUSE

Legend

0 0.5 1 2 3 4

N

Fig. 4. Planned land use in study area.

M.R. Willis, A.A. Keller / Journal of Environmental Management 83 (2007) 213–227 221

study area population, mobile source emission rates forfuture conditions were projected by increasing existingmobile source emission rates by the assumed populationgrowth percentage for the planning period.

To evaluate the utility of the LUAIRTOX source modelfor assessing the human health risk implications of land usepolicy, it was linked to ISCST3 to predict the spatialpattern of air toxics risk exposure for the study area underexisting and future land use scenarios. For existing land use(Scenario 1), the LUAIRTOX emission factors sourcemodel was evaluated by first running ISCST3 using existingfacilities and reported rates as point sources (Scenario 1A)and then comparing these results to the spatial pattern ofrisk exposure predicted by ISCST3 using existing devel-oped areas zoned for industrial, commercial, and officeuses are polygon area sources (Scenario 1B). For futureconditions (Scenario 2), ISCST3 was run assuming buildout of the study area according to Ventura’s 2010 land useplan (see Fig. 4).

In general, the ISCST3 model’s regulatory defaultsettings were used for all computer runs. These defaultvalues are usually conservative estimates that were selectedby regulatory agencies to provide upper bound risk values.Following the methodology outlined in Fig. 2, an overallnoncarcinogenic HI and carcinogenic risk index (CRI)were estimated at each of almost 13,000 receptor locationsin the study area.

Baseline land use, street, and population data wereobtained from the City of Ventura’s Community Devel-opment Department in the form of a CD with data layersin ArcView format. Preprocessed meteorological data forthe Ventura area to provide hourly stability class, winddirection, wind speed, temperature, and mixing height wereobtained from the Ventura County Air Pollution ControlDistrict’s web site (http://www.vcapcd.org/air_toxics.htm#metdata), which uses 1991–1993 as representativeyears for the study area. Because of the size of the studyarea, only 1 year of meteorological data was used to reducemodel run time and 1991 was assumed to be therepresentative year.ISCST3 has the capability to take into account the

effects of terrain in the modeling of emissions dispersion,but requires a digital elevation model (DEM) for sourceand receptor base elevations. For the study area, 1:24,000-scale DEM terrain data at 30-m resolution were importedinto the ISCST3 model after downloading in seamlessraster format from the US Geological Survey’s NationalElevation Dataset (http://edcnts14.cr.usgs.gov/Website/store/viewer.htm).

3. Results

Using emission factors generated by the LUAIRTOXmodel and total area zoned for industrial, commercial, and

ARTICLE IN PRESSM.R. Willis, A.A. Keller / Journal of Environmental Management 83 (2007) 213–227222

office use as source terms, the geometric mean rate of totalpredicted toxic air emissions in the study area is 1.55 g s�1,with 98% confidence limits of 1.04 and 7.40 g s�1. The totalactual rate of total toxic air emissions reported by the 86facilities within the study is 1.46 g s�1, which falls within theconfidence limits of the predicted rate. However, thepredicted and reported emissions of individual chemicalsare only weakly correlated (R ¼ 0:55 for the mean values).To illustrate, Table 1 presents a comparison of the reportedto predicted stationary source emissions in the study areafor a select list of chemicals (N ¼ 151). These resultssuggest that the emission factors source model is a goodpredictor of total stationary source air toxic emissions forthe study area, but should not be relied upon to predictindividual chemicals. This indicates that the LUAIRTOXsource model should be a useful tool for land use policyanalysis by providing a conservative estimate of total air

Table 1

Comparison of predicted to reported stationary source air toxic emissions in

Chemical Qp(g s�1)

Reported

CAS Air toxic

106,990 1,3-Butadiene 4.12E�04

123,911 1,4-Dioxane 2.43E�03

540,841 2,2,4-Trimethylpentane 1.48E�06

75,070 Acetaldehyde 1.98E�03

75,058 Acetonitrile

7,440,360 Antimony compounds

7,440,382 Arsenic compounds

71,432 Benzene 5.43E�03

7,440,439 Cadmium compounds

56,235 Carbon tetrachloride 3.26E�07

108,907 Chlorobenzene

67,663 Chloroform 8.65E�06

7,440,473 Chromium compounds

110,827 Cyclohexane 7.40E�07

25,321,226 Dichlorobenzenes 5.66E�05

100,414 Ethyl benzene 2.98E�02

74,851 Ethylene

106,934 Ethylene dibromide

50,000 Formaldehyde 1.48E�02

9911 Gasoline organic gas

18,540,299 Hexavalent chromium

7,647,010 Hydrochloric acid 5.34E�03

7,783,064 Hydrogen sulfide 9.82E�03

7,439,921 Lead compounds

7,439,965 Manganese compounds

7,439,976 Mercury compounds

78,933 Methyl ethyl ketone

108,101 Methyl isobutyl ketone 5.21E�02

75,092 Methylene chloride 2.06E�02

7,440,020 Nickel compounds

1151 PAHs

127,184 Perchloroethylene 1.67E�02

100,425 Styrene 8.38E�02

7,664,939 Sulfuric acid

108,883 Toluene 2.28E�01

79,016 Trichloroethylene 5.84E�04

75,014 Vinyl chloride 5.09E�07

toxic emissions using the area of industrial and commer-cially zoned land as the basis.When the reported and predicted emissions are weighted

by Tp, their toxicity value (for noncarcinogenic inhalationexposure, Tp ¼ REL�1; for carcinogenic exposure,Tp ¼ CPF), both the total carcinogenic and noncarcino-genic weighted reported emissions for the study area areapproximately an order of magnitude less than the totalweighted predicted emissions. For the carcinogenic-weighted comparison, the total reported emissions were3.44E-03 g s�1, which was significantly less than the rangeof the predicted weighted total emission rate of 1.88E-02–1.47E-01 g s�1. Similarly, for the noncarcinogenic-weighted comparison, the total reported emissions were1.75E-02 g s�1, which was also significantly less than therange of the predicted weighted total emission rate of5.24E-02–7.90E-01 g s�1. This suggests that use of the

study area for a select list of chemicals (N ¼ 151)

Predicted

LCL (2%) Mean (geometric) UCL (98%)

1.23E�05 1.25E�05 7.56E�04

2.24E�03 2.82E�03 1.87E�02

1.24E�02 1.25E�02 1.34E�02

2.21E�08 2.21E�08 2.21E�08

8.07E�06 8.07E�06 8.07E�06

3.10E�05 3.40E�05 6.52E�04

2.71E�03 4.37E�03 1.72E�02

4.20E�05 5.55E�05 3.72E�04

4.11E�05 8.67E�05 1.70E�04

4.59E�04 4.59E�04 4.70E�04

3.02E�03 3.10E�03 1.19E�02

1.75E�05 1.77E�05 2.04E�03

4.96E�04 6.54E�04 3.20E�03

1.34E�05 1.34E�05 3.33E�03

5.84E�02 6.00E�02 5.97E�01

1.63E�03 1.63E�03 1.63E�03

5.68E�07 5.68E�07 1.27E�06

2.26E�03 3.47E�03 5.49E�02

2.23E�05 2.23E�05 2.23E�05

8.61E�06 1.11E�05 8.26E�05

1.06E�02 1.11E�02 6.01E�02

3.88E�03 5.00E�03 2.55E�02

1.81E�04 2.20E�04 1.03E�03

3.30E�04 3.65E�04 4.22E�03

1.43E�05 1.57E�05 2.65E�04

1.90E�02 2.23E�02 6.18E�02

7.99E�03 1.06E�02 7.59E�02

1.89E�02 2.51E�02 1.27E�01

1.12E�04 1.44E�04 1.10E�03

2.55E�04 2.65E�04 5.17E�04

3.51E�02 6.24E�02 1.78E+00

3.76E�02 4.93E�02 3.13E�01

9.00E�07 9.00E�07 9.00E�07

4.01E�02 8.04E�02 2.26E�01

1.54E�03 1.63E�03 5.19E�03

1.25E�05 1.35E�05 2.81E�04

ARTICLE IN PRESSM.R. Willis, A.A. Keller / Journal of Environmental Management 83 (2007) 213–227 223

LUAIRTOX source model to provide air toxics emissionestimates for land use policy analysis will result inconservative risk assessments.

Figs. 5 and 6 compare the predicted spatial pattern ofcarcinogenic and noncarcinogenic risk exposure in thestudy area under Scenarios 1A and B. As can be seen inFig. 5, all of the city’s planning area is estimated to beexposed to a slight (CRIo100) carcinogenic risk fromstationary source emissions under both scenarios, withsignificant portions of the urbanized area exposed tomoderate risk (CRIo1000). The major difference betweenthe two scenarios is the location and extent of moderateto high-predicted carcinogenic risk. For Scenario 1A(Fig. 5A), modeling of reported emissions predicts that

CARCINOGENIC RISKStationary Sources

Scenario 1A

Scenario 1B

0 3 6 91.5

(A)

(B)

Fig. 5. Comparison of spatial pattern of p

the areas of moderate carcinogenic risk are primarily intwo locations—a large portion of the mid-town and asmaller area in the western portion of the city along theVentura River. Modeling of emissions under Scenario 1B(Fig. 5B) using the emission factors derived with LUAIR-TOX similarly predicts a moderately exposed area in thewestern portion of the city, but an area of moderateexposure in the central portion of the city much reduced inextent. Other differences between the two scenarios are twosmall areas of high exposure in the central and westernportions of the city under Scenario 1A but not 1B; and asmall area of moderate risk in the eastern portion ofthe city predicted under Scenario 1B, but not underScenario 1A. For noncarcinogenic risk (Fig. 6), most of

12

N

Kilometers

CRI

< 10

10 - 100

100 - 1,000

1,000 - 10,000

> 10,000

PlanArea

CityLimits

Facilities

redicted CRI for Scenarios 1A and B.

ARTICLE IN PRESS

HI

< 1

1 -10

10 - 25

25 - 50

50 - 75

75 - 100

100 - 125

> 125

PlanArea

CityLimits

Facilities

NONCARCINOGENIC RISKStationary Sources

Scenario 1A

Scenario 1B

0 3 6 9 121.5Kilometers

(A)

(B)N

Fig. 6. Comparison of spatial pattern of predicted HI for Scenarios 1A and B.

M.R. Willis, A.A. Keller / Journal of Environmental Management 83 (2007) 213–227224

the urbanized portion of the city is exposed to a slighthazard (HI41) from stationary sources, but the extent ofthe impacted area is much smaller when modeled using theLUAIRTOX emission estimates under Scenario 1B. UnderScenario 1A (Fig. 6A), several small pockets are predictedto be exposed to moderate to high noncarcinogenic risk,but Scenario 1B (Fig. 6B) modeling predicts a single smallarea of moderate noncarcinogenic risk in the south centralportion of the city.

Fig. 7 more clearly shows the differences between theextent and magnitude of the predicted indices for bothcarcinogenic and noncarcinogenic risk under Scenarios 1Aand B. These maps were created by overlaying the grids of

predicted CRI and HI indices for both scenarios and thensubtracting the values of Scenario 1A from 1B at eachreceptor location. As can be seen, use of the meanLUAIRTOX emission estimates tends to underestimatecarcinogenic risk in the central portion of the city, butoverestimate risk in the western and eastern portions of thecity. Risk underestimation tends to occur where actualfacilities emit a disproportionate share of carcinogens.Overestimation of risk occurs in locations of large areas ofindustrially zoned land, but with few reporting facilities,or with facilities that report significantly lower amounts ofcarcinogenic emissions. For noncarcinogenic risk, use ofthe LUAIRTOX emission estimates (UCL) under Scenario

ARTICLE IN PRESS

DIFFERENCE IN PREDICTED RISK

Carcinogenic Risk

Noncarcinogenic Risk

0 3 6 9 121.5Kilometers

CRI

< -10,000

-10,000 to -1,000

-1,000 to - 100

-100 to -10

-10 to + 10

10 to 100

100 to 1,000

1000 to 10,000

> 10,000

PlanArea

CityLimits

HI

< -100

-100 to -75

-75 to -50

-50 to -25

-25 to -15

-15 to -5

-5 to -1

-1 to +1

1 to 5

5 to 15

> 15

CityLimits

PlanArea

(A)

(B)N

Fig. 7. Comparison of spatial pattern of the differences in predicted risk indices between Scenarios 1A and B.

M.R. Willis, A.A. Keller / Journal of Environmental Management 83 (2007) 213–227 225

1B tends to under predict risk compared to Scenario 1A inlarge portions of the central and downtown core areas ofthe city and over predict risk at two locations in thewestern and south central portions of the city. Theexplanation for the differences in modeling results issimilar to that for differences in predicted carcinogenicrisk values.

Fig. 8 compares the difference in spatial distribution ofthe predicted stationary source CRI and noncarcinogenicHI for future land use conditions (Scenario 2) to existingconditions (Scenario 1B). For carcinogenic risk, most ofthe older residential areas in the city’s west end will bepotentially impacted by increases of 1–2 orders of

magnitude. Similarly, residential areas adjacent to theindustrial area in the east end of the city will be impactedby increased carcinogenic risk of the same orders ofmagnitude, but over a smaller area spatially. A largeportion of the central area of the city will be impacted bysignificant increases in carcinogenic risk, but the areasaffected are primarily industrial and commercial. Thepredicted increases in noncarcinogenic risk are muchsmaller spatially, primarily confined to a few locations inthe west end of the city and a few small pockets in thecentral and west end, where hazard indices are predicted toincrease from 1 to 6. The results of the air toxics riskassessment for Scenario 2 suggest that build out of the

ARTICLE IN PRESS

0 3 6 9 121.5Kilometers

DIFFERENCE IN PREDICTED RISK

Carcinogenic Risk

Noncarcinogenic Risk

CRI

< -100

-100 to -10

-10 to +10

10 to 100

> 100

PlanArea

CityLimits

HI

< -5

-5 to -3

-1 to -3

-1 to + 1

1 to 3

3 to 5

> 5

PlanArea

CityLimits

(A)

(B) N

Fig. 8. Comparison of spatial pattern of the differences in predicted risk indices between Scenarios 2 and 1B.

M.R. Willis, A.A. Keller / Journal of Environmental Management 83 (2007) 213–227226

study area according to the city’s adopted land use policieswill result in significantly increased cancer burden withinthe study area.

4. Conclusions

Land use policy has the potential to significantly affectlevels of community exposure to air toxics. The majorlimitation of risk assessment methodology to land useplanning has been its focus on individual facilities; long-range land use policy planning generally relies on broadcategories of potential uses or zoning districts. A major

question not addressed by previous research is how toevaluate the potential health risk implications of land usepolicy decisions, given the inherent uncertainty about thetype and scale of future land uses. Our land use emissionfactors source model and risk assessment methodologyaddresses this question and is intended to provide a link toextend risk assessment and management to long-range landuse policy analysis. The emission factors source model usesa probability mass function to weight land uses differentlyand can be used to explore the community health riskimplications of varying the permitted uses within a zoningdistrict.

ARTICLE IN PRESSM.R. Willis, A.A. Keller / Journal of Environmental Management 83 (2007) 213–227 227

The results of the case study suggest risk modeling usingLUAIRTOX-derived air toxics emission factors canprovide a reasonably accurate prediction of the generalpattern of community health risk exposure, but it is muchless spatially accurate than using reported emissions. Thecase study results suggest the source model is a goodpredictor of total air toxics emissions, but is less accurate atpredicting emissions of individual chemicals within acommunity, which is likely the result of the large variationin emissions between individual facilities at the 2-digit SIClevel. When the predicted spatial pattern of communityhealth risk using land use emission factors generated by oursource model was compared to the risk pattern usingreported emissions from individual facilities in the studyarea, the results suggest that zoning designations canaccurately predict the broad pattern of air toxics riskexposure in a community or region, but its accuracy as aspatial model diminishes at the subarea level.

No ambient air toxics measurements were available tocompare the ISC dispersion model estimates in the casestudy. Future evaluations of this approach should includevalidation of the results with measured data. However, themethodology described in this paper is not intended to beused as a regulatory tool for individual or small-scale landuse decisions, but a long-range planning tool to assessalternative land use scenarios at a regional or subregionalscale. The methodology would be used to assess thecommunity health risk implications of the alternative landuse development policies by using the LUAIRTOX sourcemodel to estimate emission factors for each land usecategory or zoning district, inputting these source termsinto an atmospheric dispersion model, linked to anepidemiological model, and then comparing the spatialpatterns of risk exposure in the planning area underdifferent scenarios. Using the results, risk managementstrategies could then be developed for the adopted land usepolicy to minimize the potential risk burden from air toxicsexposure.

Acknowledgements

Portions of this work were completed with funding fromthe University of California Toxic Substances Researchand Training Program.

References

Asante-Duah, D., 1996. Management of Contaminated Site Problems.

Boca Raton, FL, CRC Press.

California Air Resources Board (CARB), 2002. California Emission

Inventory Data. California Air Resources Board.

Curtin, J.D., 1997. Curtin’s California land use and planning law. Solano

Press Books, Point Arena, CA.

Dent, A., Fowler, D., et al., 2000. Using GIS to study the health impact of

air emissions. Drug and Chemical Toxicology 23 (1), 161–178.

Fulton, W., 1999. Guide to California Planning. Solano Press Books,

Point Arena, CA.

International City/County Management Association (ICMA), 1997. Risk

assessment: the role of local government. International City/County

Management Association.

Laheij, G.M.H., Post, J.G., et al., 2000. Standard methods for land-use

planning to determine the effects on societal risk. Journal of

Hazardous Materials 71 (1-3 SI), 269–282.

Lochmoeller, D.C., Muncy, D.A., et al., 1975. Industrial Development

Handbook. Urban land Institute, Washington, DC.

National Research Council, 1983. Risk Assessment in the Federal

Government. National Academy Press, Washington, DC.

Schwab, J., 1993. Industrial Performance Standards for a New Century.

American Planning Association, Chicago, p. 68.

Sengupta, S., Patil, R.S., et al., 1996. Assessment of population exposure

and risk zones due to air pollution using the geographical information

system. Computers, Environment and Urban Systems 20 (3), 191–199.

South Coast Air Quality Management District, 2000. MATES-II: Multi-

ple Air Toxics Exposure Study in the South Coast Basin. South Coast

Air Quality Management District, Diamond Bar, CA.

Spadoni, G., Egidi, D., et al., 2000. Through ARIPAR-GIS the quantified

area risk analysis supports land-use planning activities. Journal of

Hazardous Materials 71 (1-3 SI), 423–437.

US Census Bureau, 2000. Comparative Statistics: 1997 Economic Census.

US Census Bureau, Washington, DC, p. 491.

US Environmental Protection Agency, 1995a. Compilation of Air

Pollutant Emission Factors, AP-42, fifth Ed., volume I: Stationary

Point and Area Sources. Washington, DC.

US Environmental Protection Agency, 1995b. User’s Guide for the

Industrial Source Complex (ISC3) Models. Volume I—User’s Instruc-

tions. US Environmental Protection Agency, Research Triangle Park,

NC.

US Environmental Protection Agency, 2001. Toxics Release Inventory:

Community Right-to-Know.

Watts, R.J., 1997. Hazardous Wastes: Sources, Pathways, Receptors.

Wiley, New York.