application of a gis-based modeling system for effective ......merical model, and a geographical...

ENVIRONMENTAL ENGINEERING SCIENCEVolume 19, Number 5, 2002© Mary Ann Liebert, Inc.

Application of a GIS-Based Modeling System for EffectiveManagement of Petroleum-Contaminated Sites

Z. Chen,1* G.H. Huang,2 A. Chakma,3 and J. Li4

1Environmental Systems Engineering Program2Energy and Environment Program

University of ReginaRegina, Saskatchewan S4S 0A2, Canada

3Department of Chemical EngineeringUniversity of Waterloo

Waterloo, Ontario N2L 3G1 Canada4Department of Civil Engineering

Ryerson UniversityToronto, Ontario N2L 3G1 Canada

ABSTRACT

A GIS-aided simulation (GISSIM) system is developed for effective management of petroleum-contami-nated sites in this study. The GISSIM contains two components: an advanced three-dimensional (3D) nu-merical model, and a geographical information system (GIS). The modeling component undertakes sim-ulation for the fate of contaminants in subsurface unsaturated and saturated zones. The GIS component isused in three areas throughout the system development and implementation process: (1) managing spatialand nonspatial databases; (2) linking inputs, the model, and outputs; and (3) providing an interface be-tween the GISSIM and its users. The system is applied to a North American case study. Concentrationsof benzene, toluene, and xylenes in groundwater under a petroleum-contaminated site are dynamicallysimulated. Conditions of the contamination in different time stages under a variety of remediation sce-narios are predicted. Reasonable outputs have been obtained and presented graphically. Implications ofthe modeling outputs have been analyzed based on the local environmental regulations. They providequantitative and scientific bases for further assessment of site-contamination impacts and risks, as well asdecisions of practical remediation actions. GISSIM is useful for both industrial and government sectorsto make informed decisions on waste management, pollution control, site remediation, and environmen-tal impact assessment.

Key words: petroleum waste; contamination; numerical simulation; soil; groundwater; geographical in-formation system; decision support

291

*Corresponding author: Environmental Engineering Program, University of Regina, Regina, Saskatchewan S4S 0A2 Canada.Phone: 306-337-2277; Fax: 306-585-4855; E-mail: [email protected]

Dow

nloa

ded

by U

NIV

ER

SIT

Y O

F W

AT

ER

LO

O f

rom

onl

ine.

liebe

rtpu

b.co

m a

t 10/

05/1

7. F

or p

erso

nal u

se o

nly.

INTRODUCTION

PROBLEMS DUE TO THE CONTAMINANT LEAKAGE and spillfrom pipelines and storage tanks in petroleum in-

dustries have received a significant amount of attentionin the past decades (Dowd, 1984; Newton, 1991). Thenumber of underground storage tanks for petroleum prod-ucts in the North America was estimated to be between1.5 and 2 millions (Predpall et al. 1984). Tejada (1984)reported that as many as 23% of storage tanks leak. Soiland groundwater in thousands of sites have been con-taminated by petroleum-derived contaminants. There-fore, an in-depth analysis on impacts associated with theexposed chemicals is desired for effective site manage-ment.

To evaluate the impacts, it is important to gain an in-sight into the fate of contaminants underground. The fateof leaked petroleum products (i.e., nonaqueous phase liq-uids, NAPLs) in the subsurface is related to a number ofphysical, chemical, and biologic processes. Upon releaseto the environment, NAPLs will migrate downward un-der the force of gravity. When significant amounts ofNAPLs are released, they will transport downward untilthey encounter a physical barrier or are affected by buoy-ancy forces near the groundwater table. Once the capil-lary is reached, NAPLs may move laterally as a contin-uous, free phase layer along the upper boundary ofwater-saturated zone due to gravity and capillary forces.The convection, dispersion, diffusion, adsorption,

volatilization, and biodegradation may govern transportof a variety of petroleum-derived chemicals in thegroundwater system.

Previously, a number of studies have been undertakenfor simulating the fate of NAPLs in soil and groundwa-ter. For example, Huyakorn and Pinder (1983) discussedseveral formulations and solution methods for multiphaseflow analysis. Abriola and Pinder (1985a, 1985b) pro-posed a comprehensive approach to simulate the simul-taneous transport of a chemical contaminant in threephysical forms: as a nonaqueous phase, as a solute com-ponent of a water phase, and as a mobile fraction of agas phase. Kaluarachchi and Parker (1989) formulated afinite element model for simulating the multiphase flowof organic contaminants. Katyal et al. (1991) used a two-dimensional finite element program to simulate multi-phase and multicomponent transport of NAPLs in sub-surface with an assumption of the first-order decay. Ingeneral, most of the recent modeling efforts were basedon multiphase, multicomponent analyses, such as MOVER,BIOF&T, and UTCHEM (Freeze et al., 1995; Katyal,1997), which can effectively reflect complexities in sub-surface systems. However, extensive applications of thedeveloped models to practical problems were limited dueto the ineffectiveness in presentation and management oftheir inputs/outputs and the lack of dynamic and interac-tive systems for depicting the related spatial informationgraphically.

Fortunately, important progress has been made during

292 CHEN ET AL.

FIG. 1. Structure of the GISSIM system.

Dow

nloa

ded

by U

NIV

ER

SIT

Y O

F W

AT

ER

LO

O f

rom

onl

ine.

liebe

rtpu

b.co

m a

t 10/

05/1

7. F

or p

erso

nal u

se o

nly.

the last decade with regard to the use of a geographic in-formation system (GIS) to deal with spatial data and pres-ent them graphically (Mattikall, 1994; Hiscock et al.,1995; Wilkinson, 1996). With its powerful capacity ofpresenting landscape features by means of geospatiallyreferenced data, GIS has been widely used in environ-mental modeling and risk assessment. For example,Schenk et al. (1993) studied the integration of a 3Dgroundwater model with a multidimensional GIS system.Batelaan et al. (1993) incorporated a groundwater modelwithin a GIS system. However, most of the previous stud-ies were limited within the scope of interactive and vi-sual representation of modeling results (Stein et al., 1995;Bober et al., 1995). Only a few of them went beyond re-search and produced user-friendly software packages(Ehlers et al., 1989; Lovertt et al., 1997).

This research provides an extended integration of asubsurface model and GIS with a focus on the followingtwo objectives:

1. To develop a GIS-based simulation (GISSIM) systemfor petroleum waste management. This system willcontain an advanced 3D numerical model for simu-lating the fate of contaminants in soil and groundwa-ter, as well as a data management system based on

GIS for supporting the simulation process and pre-senting the modeling inputs and outputs graphically.

2. To apply the developed system to a North Americancase study. Concentrations of benzene, toluene, andxylenes in groundwater under a petroleum-contami-nated site will be dynamically simulated. The impactof the contamination in different time intervals undera variety of remediation scenarios will be predictedthrough interactive handling the modeling database ina desktop GIS environment.

THE GIS-BASED SIMULATION (GISSIM) SYSTEM

Multicomponent transport model

One of the key components in the GISSIM system isa multicomponent simulation model. It can be used forsimulating the transport and biodegradation of multiplecontaminants in subsurface media under various site con-ditions and remediation scenarios. A typical subsurfacetransport media has five regions: (1) voids filled with air,(2) mobile water located inside the larger interaggregatepores or fractures, (3) immobile water located mainly inthe intraaggregate pores or in the porous media sur-

GIS-SUPPORTED MODELING OF PETROLEUM-CONTAMINATED SITES 293

ENVIRON ENG SCI, VOL. 19, NO. 5, 2002

FIG. 2. The study site.

Dow

nloa

ded

by U

NIV

ER

SIT

Y O

F W

AT

ER

LO

O f

rom

onl

ine.

liebe

rtpu

b.co

m a

t 10/

05/1

7. F

or p

erso

nal u

se o

nly.

rounding fractures, (4) a dynamic soil region in equilib-rium with the mobile phase, and (5) a stagnant soil re-gion where mass transfer is diffusion limited. The gen-eral transport equation can be expressed as (VanGenuchten and Wierenga 1976):

(umCwm)/t 1 (uimCwim)/t 1 (f rPwm)/t

1 [(1 2 f)rPwim]/t 5 (umDijCwm/xj)/xi (1)2 (qiCwm)/xi 2 qsCws

where Cwim is the concentration of contaminant w in im-mobile water [ML23]; Cwm is the concentration of con-taminant w in mobile water [ML23]; Cws is the concen-tration of contaminant w in injected fluid [ML23]; Dij isthe hydrodynamic dispersion tensor corresponding to adirection defined by i, j [L2 T21]; f is the fraction of sorp-

tion site that is in direct contact with mobile liquid; i isthe direction i in a Cartesian coordinate system; j is thedirection j in a Cartesian coordinate system; Pwim is theadsorbed phase concentration of contaminant w in im-mobile phase [MM21]; Pwm is the adsorbed phase con-centration of contaminant w in mobile phase [MM21]; qi

is the Darcy velocity in direction i [LT21]; qs is the volumetric flow rate of fluid injection (or withdrawal)per unit volume of porous medium [L3 L23 T21]; t is thetime [T]; xi is the distance in direction i [L]; xj is the dis-tance in direction j [L]; r is the soil bulk density [ML23];uim is the fraction of soil filled with immobile water; andum is the fraction of soil filled with mobile water.

The value of Dij is defined as follows (Bear, 1972):

umDij 5 dLuqudij 1 (dL 2 dT) qiqj/uqu 1 umtDcdij (2)

294 CHEN ET AL.

FIG. 3. Site location of three USTs displayed by GISSIM. D, UST-1; u , UST-2; s , UST-3; ( , Monotoring well.

Dow

nloa

ded

by U

NIV

ER

SIT

Y O

F W

AT

ER

LO

O f

rom

onl

ine.

liebe

rtpu

b.co

m a

t 10/

05/1

7. F

or p

erso

nal u

se o

nly.

where dL is the longitudinal dispersivity [L]; dT is thetransverse dispersivity [L]; Dc is the molecular diffusioncoefficient [L2T21]; uqu is the absolute value of Darcy ve-locity [LT21], uqu is the (uqiu2 1 uqju2)1/2; qj is the Darcyvelocity in direction j [LT21]; dij is the Kronecker delta;and t is the tortuosity.

Using the following continuity equation for water flow:

2 (qi)/xi 5 (um)/t 2 qs (3)

and assuming a linear sorption isotherm (Pm 5 KdCm andPim 5 Kd Cim), Equation (1) can be written as:

Cwm/t[um 1 f rkd] 1 Cwim/t [uim 1 (1 2 f) rkd]5 (umDijCwm/xj)/xi 2 qiCwm/xi (4)2 qs(Cws 2 Cwm)

The concentrations of contaminants in mobile and im-mobile phases have the following relation (Kaluarachchiand Parker, 1990):

Cwim/t [uim 1 (1 2 f)rkd] 5 X (Cwm 2 Cwim) (5)

where X is a mass transfer coefficient for diffusive mass ex-change between the mobile and immobile phases [T21].

Incorporating decay losses lwm and contaminant load-ing from a hydrocarbon source to the mobile phase Hw

in Equations (4) and (5), yields (Kaluarachchi and Parker,1990):

[um 1 f rkd] Cwm/t1 [uim 1 (1 2 f ) rkd] Cwim/t5 (umDijCwm/xj)/xi 2 qiCwm/xi (6)2 qs(Cws 2 Cwm) 2 lwm 1 Hw

[uim 1 (1 2 f)rkd] Cwim/t 5 X (Cwm 2 Cwim) 2lwm (7)

More detailed formulation and solution process forthe multiphase and multicomponent transport model inporous media were provided by Kaluarachchi andParker (1990), Katyal and Parker (1992), and Katyal(1997).

Generally, the above model has the capability to modelcomplex heterogeneous and anisotropic hydrogeology.Through this modeling system, concentrations of up tofive contaminants in a 3D domain can be predicted at aprescribed time horizon. The modeling outputs will pro-vide important bases for decisions of site remediation ac-tions.



FIG. 4. Observed BTX concentrations at groundwater table layer in 1997.

Dow

nloa

ded

by U

NIV

ER

SIT

Y O

F W

AT

ER

LO

O f

rom

onl

ine.

liebe

rtpu

b.co

m a

t 10/

05/1

7. F

or p

erso

nal u

se o

nly.

296 CHEN ET AL.

FIG

. 5.

Con

stru

ctio

n of

a G

IS d

atab

ase.

Dow

nloa

ded

by U

NIV

ER

SIT

Y O

F W

AT

ER

LO

O f

rom

onl

ine.

liebe

rtpu

b.co

m a

t 10/

05/1

7. F

or p

erso

nal u

se o

nly.



GIS system

A desktop GIS is chosen as a basic tool throughout themodeling processes due to its ability to clearly exposecomplex environmental conditions in the subsurface. Ingeneral, GIS is used in three areas throughout the systemdevelopment and implementation process: (1) managingspatial and nonspatial databases; (2) linking inputs,model, and outputs; and (3) providing an interface be-tween the GISSIM and its users. The goal is to couple aGIS with the model to allow smooth communication be-tween the modeling system and its users. Figure 1 showsstructure of the proposed GISSIM system. Microsoft Ex-cel and Access are used for the spreadsheet and relationaldatabase components. Special user-defined Graphic UserInterfaces are created through programming MicrosoftVisual Basic and the Avenue language, which providetools for preparation of model input data set, data retrievaland inquiries, visual presentation of modeling results, andsite spatial characterization.

Database management. A GIS is used to enter data,compare data from different sources and in different for-mats, assess data availability and quality (e.g., accuracyand scale), and identify potential data errors (Grossmann

and Eberhardt, 1993). A centralized database with a struc-ture that allows efficient storage and retrieval of spatiallyreferenced time-series data is developed before the sim-ulation process. The database assures the integrity of dataand makes it possible to analyze related information com-prehensively.

Generally, spatially referenced physical data for thegroundwater model can be divided into two categories,one value per location (or one-for-one), and many valuesper location (or many-for-one). Examples of the one-for-one data include area of a polygon, and elevation of apoint. Time-series data belong to the many-for-one cat-egory, such as temporal variations of pollutant concen-trations and groundwater flows. To store and retrievethese two types of data efficiently during a modeling pro-cess, different data structures and retrieval methods arerequired. It is assumed that features of a location (e.g.,monitoring wells, and pollution sources) can be uniquelyidentified through a location identifier (ID).

The one-for-one data can be stored as location attrib-utes attached to a GIS coverage. In this type of database,the location ID is used as a key for data storage and re-trieval. The many-for-one data can be stored in separatedatabases with one file for each attribute at each physi-cal location. In this type of databases, a file name should

FIG. 6. Database preparation.

Dow

nloa

ded

by U

NIV

ER

SIT

Y O

F W

AT

ER

LO

O f

rom

onl

ine.

liebe

rtpu

b.co

m a

t 10/

05/1

7. F

or p

erso

nal u

se o

nly.

298 CHEN ET AL.

reflect both the location ID and the name of physical at-tribute it stores. In this study, a number of locations areconsidered, with each of them referring to a group of at-tribute data. This leads to a large amount of files corre-sponding to inputs and outputs of the simulation model.To enhance management of these files, a communicationchannel between Microsoft Access database and the Arc-View GIS database is built through Visual Basic pro-gramming.

Model interface. This study considers both hydrogeo-logic characteristics and contaminant features in evalu-ating the impact of contamination sources on subsurfacewater quality. The proposed GISSIM system links digi-

tal data, model, and computational outputs together. Itsinterface generates proper input data files automaticallyfor the model, serving as a bridge between the input dataand the simulation model. The GISSIM’s retrieval andpresentation function allows users to retrieve and analyzeinformation of site conditions and modeling results dy-namically. A geographic “hot link” is created as a hy-permedia function to communicate spatially referencedinformation into various types of data. For example, acontaminant’s concentration at a spatial location can beretrieved through clicking on that location at the screenreferring to a subsurface layer. In addition, the model in-terface connects the simulation results to a graphic dis-play supported by GIS.

FIG. 7. The grid system. D, UST-1; u , UST-2; s , UST-3.

Dow

nloa

ded

by U

NIV

ER

SIT

Y O

F W

AT

ER

LO

O f

rom

onl

ine.

liebe

rtpu

b.co

m a

t 10/

05/1

7. F

or p

erso

nal u

se o

nly.

User interface. The user interface makes the imple-mentation of the GISSIM a robust and user-friendlyprocess. This component provides a two-way commu-nication between the system and its users (Faust et al.,1991). On the one hand, the user may interactively de-lineate an area of concern, identify contaminationsources to be considered, add additional data, or spec-ify a particular planning objective. On the other hand,the system can explain to the user about each step inthe modeling process and display results from runningthe simulation model. The system can also provide itsuser an evaluation of the quality of data, accuracy ofthe result, and level of uncertainty. If the user is notsatisfied with the results from available data, the sys-tem can recommend to what data are needed to improvethe modeling performance.

The display of data is the final stage in the modelingprocess, providing a link between the data and the user.The most powerful medium for this communication is thegraphics, usually in the form of maps, charts, or tables.In this study, ArcView is used for managing and dis-playing modeling results. The user can use ArcView’sstatistical and spatial query functions to select output in-formation. Avenue scripts were developed to create a cus-tomized user interface. Distribution of pollutant concen-trations in the subsurface at a prescribed time can thusbe displayed dynamically through contours and 3D sur-faces within the GISSIM system.

CASE STUDY

Overview of the study site

The study site is located in western Saskatchewan,Canada (Fig. 2). It was operated as a natural gas pro-cessing plant from mid-1960s to early 1990s. The plantwas utilized to remove naphtha condensate from the nat-ural gas stream prior to transport to a regional transmis-sion line. Throughout the history of the site, the conden-sate was disposed of in three perforated undergroundstorage tanks (USTs) (Fig. 3), which then leaked into thesoil and finally into the groundwater following seepage.Two contaminant concentrated zones were formed in thesubsurface capillary fringe (at the interface between un-saturated and saturated zones).

The site is bounded in all directions by agriculturalland. There are several farm residences located withina 1.5-km radius of the site all with domestic waterwells. In recent years, agricultural land in the southeasthas been used for livestock husbandry (for cows andhorses).

Groundwater was encountered between 5 and 10 m be-low the surface. The general groundwater flow direction

is towards the south, with the gradient of the water tablebeing slightly from the northeast to the southwest. Thegroundwater table is located predominantly within a clay-till soil layer.

Figure 4 shows the observed concentrations of benzene,toluene, and xylene at the study site. Two highly concen-trated zones exist around underground storage tank 2(UST-2) and under UST-3. The highest benzene concen-tration (7.58 mg/L) was observed at a monitoring well lo-cated at the east of UST-3. The peak concentrations oftoluene and xylene (5.92 and 1.87 mg/L, respectively) wereobserved at a well located at 20 m east of UST-2. If thelocal Non-Potable Groundwater Quality Guidelines are ap-plicable to the study site, the peak concentration of ben-zene will be over four times the regulated 1.9 mg/L, andthat of toluene will be over three times the regulated 1.9mg/L. In comparison, the xylene concentrations are gen-erally lower than the regulated values.

The site conditions and modeling outputs can be con-veniently represented through the developed GISSIMsystem. For instance, facilities and emission sources onthe site can be presented graphically through the GIS.The observed and predicted contaminant concentrations,as well as information about the groundwater table, canbe displayed as 2D contour lines or 3D surfaces. Thestratigraphy in the subsurface can be visualized with suf-ficient data on the distribution of soil types at differentelevations and cross-sections. In addition, many model-ing parameters at a spatial location, such as hydraulicconductivity, soil density, water saturation level, and airpressure, are grouped into a many-for-one data categoryin the GIS. The above information can be retrieved byclicking the related features (point, lines, and polygons)in the GISSIM system.

GISSIM modeling process

The GISSIM modeling process involves the followingfour steps:

Step 1: Digitization and GIS database development.This step is to provide a GIS database of pollution-re-lated parameters at the site. The GIS database addressedhere is a special component hosting all the informationrelated to site contamination and hydrological conditionssuch as soil types, pollution source distributions, conta-mination levels, and surrounding environmental features(e.g., residential zones, river, and lakes). The databasehas two components: ARC coverages for geographic fea-tures and unique feature IDs; and Microsoft Access ta-bles for attribute items, with the feature ID as a uniquekey. The connection between the two is built on the Data-



Dow

nloa

ded

by U

NIV

ER

SIT

Y O

F W

AT

ER

LO

O f

rom

onl

ine.

liebe

rtpu

b.co

m a

t 10/

05/1

7. F

or p

erso

nal u

se o

nly.

300 CHEN ET AL.

FIG. 9. Predicted toluene concentrations.

FIG. 8. Predicted benzene concentrations.

Dow

nloa

ded

by U

NIV

ER

SIT

Y O

F W

AT

ER

LO

O f

rom

onl

ine.

liebe

rtpu

b.co

m a

t 10/

05/1

7. F

or p

erso

nal u

se o

nly.

base Integrator in ARC/INFO and ARCVIEW3. Figure5 shows the details of the construction of a spatial GISdatabase. The database is useful for further modelingwork. For example, user could build up the input data forthe multicomponent transport simulator by using the in-terface as shown in Fig. 6.

Step 2: Grid system development. The multicompo-nent transport model is implemented through a compu-tational grid system. Normally, 2D grid systems can bedirectly generated in the GIS. Because this study re-quires a 3D mesh, an external mesh editor is launchedto create and edit 3D finite element meshes at 16 3 17-m resolution. This editor allows generation of irregularquadrilateral meshes in two or three dimensions, as wellas analysis of soil and groundwater properties in thestudy domain. Figure 7 shows the top slice of the gen-erated 3D mesh.

Step 3: Contaminant transport modeling. The multi-component transport simulator can be launched throughthe proposed event-driven interface. After the modelinginputs are prepared in the GIS database, the simulator canbe used for predicting the fate of benzene, toluene, andxylenes (BTXs) in the aquifer.

Step 4: Results representation. The simulation outputsare transformed into a GIS acceptable format. They can bepresented in several ways, such as statistical charts, con-tour lines, and 3D surfaces. Figures 8 to 10 present the sim-ulated concentration of BTXs through contour lines.

Model evaluation and testing

Calibration and verification are important steps forthe modeling study with a focus on groundwater. Theywere undertaken using data obtained during 1993 to1997 from five monitoring wells. Information of siteconditions and contamination sources in 1993 was usedfor estimating model parameters. The monitoring dataof 1997 were then used to verify the model. The ab-solute errors between simulated and observed benzeneconcentrations range from 0.023 to 0.532 mg/L, with amean relative error of 44.4%. The mean relative errorsfor toluene and xylenes are 71.0% and 56.2%, respec-tively. The results indicated that existence of the threecontaminants in most of the monitoring wells is wellsimulated by the model. A few exceptions exist due tolack of monitoring data (for model calibration), poten-tial sampling errors, low soil permeability, and sub-surface stratification. For example, simulated concen-



FIG. 10. Predicted xylene concentrations.

Dow

nloa

ded

by U

NIV

ER

SIT

Y O

F W

AT

ER

LO

O f

rom

onl

ine.

liebe

rtpu

b.co

m a

t 10/

05/1

7. F

or p

erso

nal u

se o

nly.

tration of toluene (4.37 mg/L) in a monitoring wellclose to UST-2 is remarkably different from observeddata (2.48 mg/L). This observed value is associatedwith many uncertainties, which significantly increasedthe average error of simulating the fate and transportof toluene. If this observation is considered as an out-lier, the mean relative error for toluene will be reducedto 57%. In general, the above verification results indi-cate that the developed model can simulate the fate ofBTX in groundwater with reasonable errors.

RESULTS

After the model is well calibrated and verified, it canthen be used for dynamically simulating the fate of con-taminants under a number of remediation scenarios. Fig-ures 8 to 10 show distributions of benzene, toluene, andxylenes concentrations 10 years later when no remedia-tion action is undertaken. It is indicated that the peak ben-zene concentration will still be 7.40 mg/L 10 years later,which is higher than the regulated level of 1.90 mg/L inthe local Non-Potable Groundwater Quality Guidelines.This peak occurs at about 30 m west of UST-2 and 30 mnortheast of UST-3. The results also demonstrate that,due to the large contamination area and the specific strati-graphic and hydraulic characteristics of the site, BTX inthe unsaturated zone will be continuously released togroundwater if no remediation action is undertaken. It isthus recommended that the on-site groundwater not beused for irrigation or drinking water, even 10 years laterunder this scenario.

The peak toluene concentration 10 years later will be6.46 mg/L, which is still higher than the regulated levelof 5.90 mg/L. In comparison, the peak xylenes concen-tration (2.41 mg/L) is lower than the regulated upper limitat 5.60 mg/L. If a remediation action with 60% efficiencyis undertaken, all pollution problems at the site will becompletely resolved 10 years later. The peak concentra-tions of benzene, toluene, and xylenes 10 years laterwould be reducecd to 1.49, 2.10, and 0.46 mg/L, re-spectively, which are lower than the regulated values.

Potential impacts on communities

The impact information is incorporated within the GIS-SIM system, through systematic links to the GIS data-base, the modeling outputs, and the related environmen-tal guidelines. The farm residences to the north andnortheast of the study site withdraw groundwater for do-mestic use. Fortunately, the on-site groundwater flow di-rection is generally to the south. No complaint has yetbeen received from the farmers regarding the quality ofgroundwater in the north. However, due to the high dis-

persivity and volatility of BTX, substantial impacts mayarise in the future.

There is a large area of crop and grazing land in thesouth and west of the study area. The subsurface area un-der these lands is seriously contaminated, based on thein situ investigation and the model output. Therefore, in-teractions between the contaminated subsurface and theabove-ground vegetation may lead to health impacts onhumans and livestock.

CONCLUSIONS

A GIS-based simulation (GISSIM) system has beenproposed for petroleum waste management. It containsan advanced 3D numerical model for simulating the fateof contaminants in soil and groundwater and a GIS forsupporting the simulation process and presenting spatial-temporal information graphically. Through incorporationof GIS within the modeling framework, an object-ori-ented simulation environment with an open software ar-chitecture is provided. It can be used by environmentalengineers for supporting decisions of site remediation andrisk management. The GISSIM is available for use on ashareware or a contract basis.

The developed GISSIM system has been applied to aNorth American case study. Concentrations of benzene,toluene, and xylenes in groundwater under a petroleum-contaminated site are dynamically simulated. Conditionsof the contamination in different time stages under a va-riety of remediation scenarios are prediced and graphi-cally presented. Generally, it is indicated that benzene isthe most important contaminant at the site. If no reme-diation action is undertaken, serious benzene contamina-tion problem will still exist at the site even 10 years later.

Implications of the modeling outputs have been ana-lyzed based on the local guidelines for nonpotablegroundwater quality. They provide quantitative and sci-entific bases for further assessment of site-contaminationimpacts and risks, as well as decisions of practical re-mediation actions.

ACKNOWLEDGMENT

This research was supported by the Natural Sciencesand Engineering Research Council of Canada (NSERC).

REFERENCES

ABRIOLA, L.M., and PINDER, G.F. (1985a). A multiphaseapproach to the modeling of porous media contamination byorganic compounds. 1. Equation development. Water Resour.Res. 21, 11–18.

302 CHEN ET AL.D

ownl

oade

d by

UN

IVE

RSI

TY

OF

WA

TE

RL

OO

fro

m o

nlin

e.lie

bert

pub.

com

at 1

0/05

/17.

For

per

sona

l use

onl

y.

ABRIOLA, L.M., and PINDER, G.F. (1985b). A multiphaseapproach to the modeling of porous media contamination byorganic compounds, 2. Numerical simulation, Water Resour.Res. 21, 11–18.

BATELAAN, O., DE SMEDT, F., and OTERO, M.N. (1993).Development and application of a groundwater model inte-grated in the GIS GRASS. IAHS Publicat. 211, 581–592.

BEAR, J. (1972). Dynamics of Fluids in Porous Media. NewYork: Elsevier.

BOBER, M.L., WOOD, D., and MCBRIDGE, R.A. (1996). Useof digital analysis and GIS to assess regional soil compactionrisk. Photogram. Eng. Remote Sensing 62, 1397–1407.

DOWD, R.M. (1984). Leaking underground storage tanks. En-viron. Sci. Technol. 18, 10–15.

EHLERS, M., EDVARDS, G., and BEDARD, Y. (1989). In-tegration of remote sensing with geographic information sys-tems: A necessary evolution. Photogram. Eng. Remote Sens-ing 55, 1619–1627.

FAUST, N.L., ANDERSON, W.H., and STAR, J.L. (1991). Ge-ographic information systems and remote sensing futurecomputing environment. Photogram. Eng. Remote Sensing57, 655–668.

FREEZE, G.A., FOUNTAIN, J.C., POPE, G.A., and JACK-SON, P.E. (1995). Numerical simulation of surfactant-en-hanced remediation using UTCHEM. AIChE Symp. Ser. 91,68–71.

GROSSMAN, W.D., and EBERHARDT, S. (1993). Geo-graphical information systems and dynamic modelling. InM.M. Fischer and P. Nijkamp, Eds., Geographic Informa-tion Systems, Spatial Modelling, and Policy Evaluation.Berlin: Springer-Verlag.

HISCOCK, K.M., LOVERTT, A.A., and PARFITT, J.P.(1995). Groundwater vulnerability assessment: Two casestudies using GIS methology. Q. J. Eng. Geol. 28, 179–188.

HUYAKORN, P.S., and PINDER, G.F. (1983). ComputationalMethods in Subsurface Flow. New York: Academic Press.

KALUARACHCHI, J.J., and PARKER, J.C. (1989). An effi-cient finite element method for modeling multiphase flow inporous media. Water Resour. Res. 25, 43–54.

KALUARACHCHI, J.J., and PARKER, J.C. (1990). Modeling

multicomponent organic chemical transport in three-fluid-phase porous media. J. Contam. Hydrol. 5, 349–374.

KATYAL, A.K. (1997). BIOF&T Flow and Transport in theSaturated and Unsaturated Zones in 2 or 3 Dimensions:Technical Document & User Guide. Blacksburg, VA: DraperAden Environmental Modeling, Inc.

KATYAL, A.K., and PARKER, J.C. (1992). An adaptive so-lution domain algorithm for solving multiphase flow equa-tions. Comput. Geosci. 18, 1–9.

KATYAL, A.K., KALUARACHCHI, J.J., and PARKER, J.C.(1991). MOFAT: A Two-Dimensional Finite Element Pro-gram for Multiphase and Multicomponent Transport, Pro-gram Document and User’s Guide. Washington, DC: US.EPA/600/2-91/020.

LOVERTT, A.A., PARTFITT, J.P., and BRAINARD, J.S.(1997). Using GIS in risk analysis: A case study of hazardouswaste transport. Risk Anal. 17, 625–632.

MATTTIKALL, N.M. (1994). An integrated GIS’s approach toland cover change assessment. Int. J. Remote Sensing 2,1204–1206.

NEWTON, J. (1991). Investigating leaking underground stor-age tanks. Pollut. Eng. 23, 80–83.

PREDPALL, D.F., ROGERS, W., and LAMONT, A. (1984).An underground tank spill prevention program. Conferenceand Exposition on Petroleum Hydrocarbon and OrganicChemicals in Ground Water, National Water Well Associa-tion, Worthington, OH.

SCHENK, J., KIRK, K., and POETRE, E. (1993). Integrationof three-dimensional groundwater modeling techniques withmulti-dimensional GIS. IAHS Publicat. 211, 243–254.

STEIN, A., STARITSKY, I., and VAN GROENIGEN, J.W.(1995). Interactive GIS for environmental risk assessment.Int. J. Geograph. Informat. Systems 9, 509–515.

TEJADA, S. (1984). Underground tanks contaminate ground-water. EPA J. 10, 20–22.

VAN GENUCHTEN, M.TH., and WIERENGA, P.J. (1976).Mass transfer studies in sorbing media 1. Analytical solu-tions. Soil Sci. Soc. Am. J. 40, 473–480.

WILKINSON, G.G. (1996). A review of current issues in theintegration of GIS and remote sensing data. Int. J. Geograph.Informat. Systems 10, 85–101.



Dow

nloa

ded

by U

NIV

ER

SIT

Y O

F W

AT

ER

LO

O f

rom

onl

ine.

liebe

rtpu

b.co

m a

t 10/

05/1

7. F

or p

erso

nal u

se o

nly.

application of a gis-based modeling system for effective ......merical model, and a geographical...

Documents