Vous êtes sur la page 1sur 8

Using the Analytic Hierarchy Process for Selecting

an Appropriate Fate and Transport Model for Risk-Based


Decision Making at Hazardous Waste Sites
Venkatesh Uddameri1

Abstract: The management and regulation of hazardous waste sites are increasingly being carried out by assessing risks via different
Downloaded from ascelibrary.org by PENN STATE UNIV on 08/11/14. Copyright ASCE. For personal use only; all rights reserved.

exposure pathways. The selection of an appropriate multimedia fate and transport for carrying out screening level risk assessment is a
challenging task given the paucity of data and the need for modeling rigor. The utility of a multicriteria decision-making technique, called
the analytic hierarchy process 共AHP兲, is explored in this study to select an appropriate model that satisfies the needs of the stakeholders
in an optimal manner. The criteria for model selection focused on various issues ranging from scientific credibility to data availability and
end-user requirements, and were compiled from a survey of modeling practices and needs of various agencies of the California/
Environmental Protection Agency that was carried out previously. Seven different public-domain multimedia modeling codes were
evaluated against these criteria using AHP. The study indicated that the highest ranked model 共CalTOX兲 only satisfied about 31% of the
needs of the stakeholders for the site conditions assumed. This result indicates that there is a wide scope for further improvement in all
the models that are considered here. The local priority matrix obtained for each criteria was seen to be useful to identify limitations of
individual models. Certain improvements are suggested to enhance the available public-domain multimedia models to meet stakeholder
needs. AHP is seen as a consistent, flexible, and intuitive approach to assess necessary model complexity and is recommended for use in
similar policy and regulatory applications.
DOI: 10.1061/共ASCE兲1090-025X共2003兲7:2共139兲
CE Database subject headings: Hazardous waste; Waste sites; Models; Decision making; Matrix methods.

Introduction vironmental models to develop risk-based cleanup standards 共e.g.,


DTSC 1994; ASTM 1995; USEPA 1996; OEHHA 1998兲. Several
Risk-based corrective action protocols are increasingly being used different multimedia modeling schemes of varying levels of com-
to regulate and manage hazardous waste sites all over the world plexity have also been developed over the last two decades
today. These approaches 共e.g., ASTM 1998兲 assess the nature and 共Mackay 1979; Cohen et al. 1989; Laniak et al. 1997兲 that can
extent of risk to human health and other ecological receptors via potentially be used for developing risk-based cleanup standards.
different exposure pathways due to the release of contaminants. However, the application of multimedia models to develop
The necessary cleanup goals and remedial action are then devel- risk-based corrective action guidelines is by no means an easy
oped to ensure that the risks posed by these contaminants along task. Most multimedia modeling schemes require extensive
exposure pathways are below generally acceptable levels 共USEPA amounts of site-specific data that are not readily available at most
1996兲. sites, especially during the early stages of site investigation 共e.g.,
Mathematical models are commonly used to characterize the Daugherty 1993兲. In the absence of site-specific data, the model
partition behavior of contaminants into various environmental inputs are inherently uncertain and this parametric imprecision
compartments and obtain estimates for contaminant concentra- gets propagated through the model and contributes to uncertain-
tions via various exposure pathways. As most organic contami- ties in the output. In addition, almost all multimedia modeling
nants exhibit the ability to partition into more than one environ- schemes have undergone limited field-scale evaluation and test-
mental compartment, the application of risk-based corrective ing, and as such their conceptual adequacy to characterize the fate
action protocols to these contaminants must be carried out in a and transport has not been sufficiently tested 共e.g., Cowan et al.
multimedia framework. There has been growing recognition in 1995兲. These conceptual inadequacies translate into uncertainties
recent times among regulatory agencies to utilize multimedia en- in regulatory policies and guidelines that are developed using
these models 共NRC 1999兲. Risk-based corrective action protocols
1
Department of Environmental Engineering, MSC 213, Texas A&M call for a close collaboration between various stakeholders in-
Univ., Kingsville, Kingsville, TX 78363. E-mail: vuddameri@tamuk.edu volved in the cleanup process. Hence, the success of these cor-
Note. Discussion open until September 1, 2003. Separate discussions rective action guidelines critically hinges upon all parties under-
must be submitted for individual papers. To extend the closing date by standing the advantages and limitations of multimedia modeling
one month, a written request must be filed with the ASCE Managing
schemes 共Uddameri 2002兲. From this standpoint, easy to learn
Editor. The manuscript for this paper was submitted for review and pos-
sible publication on December 12, 2002; approved on December 12, and use, transparency in model documentation, availability of the
2002. This paper is part of the Practice Periodical of Hazardous, Toxic, model in the public domain, and friendly graphical user interfaces
and Radioactive Waste Management, Vol. 7, No. 2, April 1, 2003. may all be useful features that will assist in the model selection
©ASCE, ISSN 1090-025X/2003/2-139–146/$18.00. process.

PRACTICE PERIODICAL OF HAZARDOUS, TOXIC, AND RADIOACTIVE WASTE MANAGEMENT © ASCE / APRIL 2003 / 139

Pract. Period. Hazard. Toxic Radioact. Waste Manage. 2003.7:139-146.


2. Development of a Judgmental Matrix: Here the elements of a
particular level are compared in pair-wise fashion with a
specific element of an upper level. First, criteria are com-
pared pairwise to the goal. A judgmental matrix 共A兲 is devel-
oped using the comparisons. Each entry of the judgmental
matrix (a i j ) is formed by comparing row element A i with
column element A j . Pair-wise comparisons of two criteria
C i and C j are made by questions of, ‘‘of the two criteria C i
and C j which one is more important to achieve the goal 共G兲
and by how much?’’ A point scale is typically used to trans-
form verbal judgments into numerical quantities 共Saaty
2000兲. Entries a i j are governed by the following rules: a i j
⬎0 and ⫽1/a ji . Hence, the judgmental matrix is a positive
reciprocal pair-wise comparison matrix 共e.g., in Table 2兲.
Downloaded from ascelibrary.org by PENN STATE UNIV on 08/11/14. Copyright ASCE. For personal use only; all rights reserved.

3. Local Priorities and Consistency of Comparisons: This step


entails obtaining local priorities of the criteria and evaluating
the consistency of judgments. Local evaluation is estimated
by finding the principal eigenvector (␭ max) of the matrix 共A兲.
Fig. 1. Schematic of pathways considered during model selection
The consistency in comparison is evaluated using a measure
process
called the consistency ratio 共CR兲, defined as
CI
CR⫽
RI
As can be seen from the above discussion, the development where
and selection of an appropriate multimedia modeling software for ␭ max⫺n
regulatory applications are confounded by multiple and often- CI⫽
times conflicting objectives. As an example, incorporating all op- n⫺1
erative processes may render the model more scientifically cred- where n is the number of alternatives being compared, and
ible, but will also increase the number of necessary input RI is the consistency index of a random reciprocal matrix
parameters 共data requirements兲 and associated parametric uncer- from a nine-point scale, with the reciprocals forced 共Saaty
tainties in addition to making the model cumbersome to follow. 1980兲. Tabulations for RI are readily available in the litera-
Therefore the selection of the final model from a set of alterna- ture 共Saaty 1980; Hanna 1996; Ramanathan 2001兲. In gen-
tives invariably involves a trade-off among objectives governing eral, if the computed values of the CR are greater than 0.1,
the model selection process. Multiobjective decision-making ap- the judgmental matrix is considered to be unreliable and
proaches provide a convenient set of mathematical tools to iden- must be elicited again.
tify an optimal alternative given a set of competing objectives 4. Aggregation of Local Priorities: The local priorities at each
共e.g., Clemens 1997兲. One such multiobjective decision-making level are aggregated to obtain the final priorities of the alter-
technique, the analytic hierarchy process 共AHP兲, is employed in native. The aggregation is based on the principle of hierar-
this study to demonstrate how it can be employed to identify an chic composition 共Saaty 2000兲 and is given by
appropriate multimedia modeling scheme for use with a risk- final priority of alternative A 1
based decision-making process. AHP was selected for this appli-
cation because it has been used extensively in many selection
applications 共e.g., Lin and Yang 1996; Ramanathan 2001; Lai ⫽ 兺i 关 local priority of A 1 with respect to criteria C i
et al. 2002兲 and is known to provide reliable results. However, to
the best of the author’s knowledge, this study is the first applica- ⫻local priority of C i with respect to the goal 共 G 兲兴
tion of AHP pertaining to the selection of a multimedia fate and
transport model for risk-based decision making. AHP is a useful tool for ranking and selecting models from an
available pool of formulations under a given set of criteria. It
should be noted here that other multicriteria methodologies such
Methodology as multiobjective value theory 共Kirkwood 1996兲 might also be
used for assessing appropriate model complexity. However, AHP
The analytic hierarchy process is an intuitive, pair-wise compari- is often the preferred approach since it can combine both quali-
son method for analyzing alternatives that was proposed by Saaty tative and quantitative elements, aggregate the opinions of mul-
共1980兲 and has since been refined and applied extensively 共Saaty tiple stakeholders and experts, and provide mechanisms by which
1986, 1987, 2000; Lootsma 1993; Elkarmi and Mustafa 1993; to check for inconsistencies in the formulation 共Ramanathan
Ramanathan 1999; and others兲. The application of AHP entails 1999兲. Rank reversal is cited as a limitation of AHP 共Belton and
the following four steps: Gear 1983兲, however AHP in the form here has been used exten-
1. Structuring the Decision Problem into a Hierarchical sively and has proved useful in many situations 共see, e.g., Lin and
Model: This includes the decomposition of the decision Yang 1996; Ramanathan 2001; Lai and Wong 2002兲; as such,
problem into elements of different hierarchical states. The reversal of AHP ranking is not addressed in this study. Rank re-
top level corresponds to the focus of the problem or the final versal may become an issue when two modeling codes based on
goal 共G兲, the intermediate levels correspond to criteria and the same mathematical formulations 共e.g., two codes for level-III
subcriteria, and the lowest level corresponds to ‘‘decision formulations兲 are being simultaneously considered. Alternatives
alternatives’’ 共Fig. 1兲. such as the concept of absolute measurement 共Saaty 1987兲 and

140 / PRACTICE PERIODICAL OF HAZARDOUS, TOXIC, AND RADIOACTIVE WASTE MANAGEMENT © ASCE / APRIL 2003

Pract. Period. Hazard. Toxic Radioact. Waste Manage. 2003.7:139-146.


lators of CAL/EPA to the modeling practices and needs survey
共Uddameri and Young 1999a兲 and they are presented in Table 2.
According to Table 2, scientific credibility and regulatory issues
were seen to be relatively more important factors followed by
data availability, end-user needs, and characterization of the ap-
plication domain for assessment of screening levels.
While some multimedia modeling codes are available com-
mercially 共e.g., MEPAS兲 and have been used in risk-assessment
studies 共see, e.g., Laniak et al. 1997兲, the study focused exclu-
sively on codes that are available in the public domain since the
ease in availability of the modeling packages allows for greater
Fig. 2. Evaluation objective, criteria, and alternatives considered in clarity of the regulatory management process and eliminates any
this study potential bias 共Daugherty 1993兲 共D. Siegel, personal communica-
tion, OEHAA, CAL/EPA, 1999兲. For the same reasons, certain
Downloaded from ascelibrary.org by PENN STATE UNIV on 08/11/14. Copyright ASCE. For personal use only; all rights reserved.

multiplicative AHP 共Lootsma 1999兲 have been proposed in the legacy multimedia models like UTM-TOX 共Paterson et al. 1984兲
literature as possible remedies and can be used as necessary. were also eliminated from consideration. Seven different ‘‘public-
domain’’ multimedia modeling schemes were identified 共Fig. 2兲
and evaluated using 20 different subcriteria that were categorized
Application under the five evaluation criteria listed above 共Fig. 3兲. Brief de-
scriptions of the models employed in this study are presented in
The overall objective of this study is to identify an appropriate the Appendix. The criteria and subcriteria were used to carry out
multimedia fate and transport model from an available pool of pair-wise comparisons of the alternatives in an objective fashion.
public-domain multimedia modeling schemes for use in risk- Comparison of the models was limited to fate and transport mod-
based corrective action schemes. The study focuses on screening ules and how the models computed exposure concentrations in
level type models that are suitable to carry out preliminary assess- each environmental compartment.
ments. These models are often used to obtain an average exposure In some situations, the models provided additional information
concentration within each environmental compartment that is then such as carcinogenic and noncarcinogenic risks using additional
coupled with toxicological algorithms to assess risks posed along toxicological parameters. These computations and the associated
different pathways 共e.g., ASTM 1995兲. data requirements were not considered in this evaluation to ensure
A generic conceptual model that describes various contami- all models were assessed on a common platform. The relative
nant pathways typically encountered at a hazardous waste site significance of one model over another was evaluated by assess-
共Fig. 1兲 was used as the basis for model selection. A survey on ing the extent to which the models under consideration satisfied
fate and transport modeling practices, needs and recommended the specified subcriteria. The assignment of weights was based on
criteria for evaluating multimedia, and media specific models in the author’s experience with these models and on previous evalu-
different California/Environmental Protection Agency 共CAL/ ation studies 共e.g., Uddameri and Young 1999b, 2000兲. The pair-
EPA兲 agencies was carried out as part of a previous multimedia wise comparison matrix for each criterion is presented in Table 3
model identification, integration, and development study 共Ud- and the final weight matrix 共priorities兲 along with aggregate
dameri and Young 1999a兲. The results obtained from this survey weights is presented in Table 4. As can be seen from Tables 3 and
were broken down to structure the decision problem into five 4, no single model demonstrates clear superiority across all crite-
hierarchical criteria 共Fig. 2兲. ria.
A nine-point pair-wise comparison scale was developed to fa- CalTOX and level-III models are considered more scientifi-
cilitate the comparison process and it is presented in Table 1. cally credible because they have been developed from first prin-
ciples, conserve mass, and have undergone some sort of evalua-
tion or testing 共e.g., Maddalena et al. 1995; Severnsin et al.
Results and Discussion 1996兲. While MULTIMED and PC-GEMS also provide mecha-
nisms to simulate multiple pathways of interest that are pertinent
The relative importance of the criteria 共judgmental matrix兲 used to hazardous waste sites, global mass balance may not be
in the study was ascertained from answers provided by the regu- achieved by these models since transport equations for individual
compartments are not linked by appropriate boundary conditions.
Level-I and level-II fugacity models rate high from the standpoint
Table 1. Semantic Scale Used in AHP 共Saaty 1980兲 of data requirements because they are capable of producing an
Intensity Definition Remarks estimate for average output concentrations in different compart-
ments using relatively little input data. MULTIMED and PC-
1 Equal importance Ai is as important as A j
GEMS are seen to be better suited when spatial variability within
3 Weak importance Ai is slightly favored over A j
the domain is to be explicitly modeled since the other modeling
5 Essential importance Ai is moderately favored over A j
schemes are box formulations and only provide an average con-
7 Demonstrated Ai is strongly favored over A j
centration within the compartment. Among the box formulations,
importance
CalTOX is placed over level-III, level-II, and level-I models be-
9 Absolute importance A i is most strongly preferred
cause it includes more exposure pathways pertinent to hazardous
over A j
waste sites and also accounts for transient loadings in some com-
2, 4, 6, 8 Intermediate When a compromise is needed
partments. CalTOX and level-III models rate highly from the
values between judgments
standpoint of end-user requirements since they both run under the
are used
Windows environment and possess intuitive user interfaces and

PRACTICE PERIODICAL OF HAZARDOUS, TOXIC, AND RADIOACTIVE WASTE MANAGEMENT © ASCE / APRIL 2003 / 141

Pract. Period. Hazard. Toxic Radioact. Waste Manage. 2003.7:139-146.


Table 2. Pair-wise Matrix for Comparing Factors to Obtain Weights
Scientific credibility Data requirements Application domain End-user needs Regulatory issues
Scientific credibility 1.000 4.000 7.000 6.000 2.000
Data requirements 0.250 1.000 4.000 4.000 0.250
Application domain 0.143 0.250 1.000 0.250 0.125
End-user needs 0.167 0.250 4.000 1.000 0.143
Regulatory issues 0.500 4.000 8.000 7.000 1.000

also provide substantial outputs that characterize the transport and models with respect to application domains 共e.g., Wania 1996;
fate of contaminants within the system. Although level-I and Ares et al. 1998; Woodfine et al. 2002兲. However, unlike
level-II models possess user interfaces similar to that of the level- regional-scale applications cited herein, the geographic domain
III model, they do not compare as well as level-III with respect to must capture local scale variability in and around the hazardous
Downloaded from ascelibrary.org by PENN STATE UNIV on 08/11/14. Copyright ASCE. For personal use only; all rights reserved.

this criterion because these models provide less output 共informa- waste site. Similarly, the availability of public-domain level-III
tion兲 about the system than the level-III model. CalTOX, level-III, computer codes that incorporate vegetation and groundwater
and the preliminary endangerment assessment 共PEA兲 tool devel- pathways 共Cousins and Mackay 2001兲, along with a database for
oped by the CAL/EPA Department of Toxic Substances and Con- plant–soil and plant–air partition coefficients, will increase the
trol 共DTSC兲 rate high in terms of regulatory issues since they suitability of the level-III model for hazardous waste site applica-
primarily have been endorsed by regulatory agencies 共e.g., DTSC tions. The inclusion of a separate nonaqueous phase will enable
1994; USEPA 1998兲 in addition to having other benefits such as these models to capture hazardous waste site characteristics in a
extensive documentation, especially pertaining to model limita- more realistic fashion 共MacFarlane and Mackay 1998兲. In addi-
tions 共Mackay and Patterson 1990; DTSC 1993; McKone et al. tion field-scale evaluations will enhance the scientific credibility
1996兲 and tamper-proof computer codes. of all the models 共Cowan et al. 1995兲. Currently PC-GEMS is
The results of the aggregate matrix presented in Table 4 indi- undergoing significant revamping including coupling with GIS
cate that if we consider all the criteria used in this study simulta- 共USEPA 2002兲 and this newer model 共IGEMS兲 may satisfy sev-
neously, the CalTOX model is the most suitable for obtaining eral requirements identified in this study. Although not evaluated
screening-level exposure concentrations in various compartments rigorously, the evaluation criteria selected and the pair-wise com-
within a multimedia framework followed by the level-III multi- parison process carried out in this study appear to favor a hybrid
media model and the MULTIMED model. However, the CalTOX compartmental-hybrid multimedia modeling framework that was
model has an aggregate score of 31%, indicating that all multi- proposed recently in the literature 共Cohen and Cooter 2002兲 since
media modeling schemes and associated computer codes can be they are able to provide spatial resolution with fewer input pa-
improved substantially to satisfy the needs of regulatory person- rameters. The availability of such modeling codes in the public
nel and other stakeholders. The local priority matrices 共Tables 3 domain may be useful for effective site management in years to
and 4兲 can be used to identify limitations of the model with regard come.
to specific criterion and as such are useful for future developmen- It should be borne in mind that the objective of this study is to
tal work related to refinement of these multimedia codes. As ex- illustrate the utility of AHP to the multimedia model selection
amples, running box models in Monte Carlo simulation mode and process and, as such, the weights obtained here represent the
coupling of the level-III or CalTOX model with a geographic views of one set of stakeholders and modelers and the ranking
information system 共GIS兲 would reduce the limitations of these obtained is indicative of their preferences and are not to be inter-

Fig. 3. Subcriteria used to carry out pair-wise comparison of alternatives

142 / PRACTICE PERIODICAL OF HAZARDOUS, TOXIC, AND RADIOACTIVE WASTE MANAGEMENT © ASCE / APRIL 2003

Pract. Period. Hazard. Toxic Radioact. Waste Manage. 2003.7:139-146.


Table 3. Local Priorities Matrix for Different Criteria
Model Level-I Level-II Level-III CalTOX MULTIMED DTSC-PEA PC-GEMS
Scientific credibility criterion ␭ max⫽7.72; CR⫽0.09
Level-I 1.00 0.33 0.13 0.11 0.13 2.00 0.17
Level-II 3.00 1.00 0.17 0.14 0.25 3.00 0.25
Level-III 8.00 6.00 1.00 0.25 2.00 8.00 4.00
CalTOX 9.00 7.00 4.00 1.00 2.00 9.00 3.00
MULTIMED 8.00 4.00 0.50 0.50 1.00 8.00 4.00
DTSC-PEA 0.50 0.33 0.13 0.11 0.13 1.00 0.14
PC-GEMS 6.00 4.00 0.25 0.33 0.25 7.00 1.00

Data requirement criterion ␭ max⫽7.78; CR⫽0.10


Level-I 1.00 2.00 5.00 6.00 9.00 5.00 9.00
Downloaded from ascelibrary.org by PENN STATE UNIV on 08/11/14. Copyright ASCE. For personal use only; all rights reserved.

Level-II 0.50 1.00 3.00 4.00 8.00 3.00 8.00


Level-III 0.20 0.33 1.00 2.00 4.00 0.50 6.00
CalTOX 0.17 0.25 0.50 1.00 2.00 0.25 7.00
MULTIMED 0.11 0.13 0.25 0.50 1.00 0.33 8.00
DTSC-PEA 0.20 0.33 2.00 4.00 3.00 1.00 8.00
PC-GEMS 0.11 0.13 0.17 0.14 0.13 0.13 1.00

Application domain criterion ␭ max⫽7.53; CR⫽0.07


Level-I 1.00 1.00 1.00 0.20 0.17 0.50 0.14
Level-II 1.00 1.00 1.00 0.25 0.17 0.50 0.14
Level-III 1.00 1.00 1.00 0.33 0.17 0.50 0.14
CalTOX 5.00 4.00 3.00 1.00 0.20 3.00 0.17
MULTIMED 6.00 6.00 6.00 5.00 1.00 6.00 0.50
DTSC-PEA 2.00 2.00 2.00 1.00 0.17 1.00 0.17
PC-GEMS 7.00 7.00 7.00 6.00 2.00 6.00 1.00

End-user needs criterion ␭ max⫽7.82; CR⫽0.10


Level-I 1.00 0.50 0.33 0.25 6.00 7.00 6.00
Level-II 2.00 1.00 0.50 0.33 7.00 8.00 7.00
Level-III 3.00 2.00 1.00 0.50 7.00 8.00 7.00
CalTOX 4.00 3.00 2.00 1.00 8.00 9.00 8.00
MULTIMED 0.17 0.14 0.14 0.13 1.00 6.00 1.00
DTSC-PEA 0.14 0.13 0.13 0.11 0.17 1.00 0.17
PC-GEMS 0.17 0.14 0.14 0.13 1.00 6.00 1.00

Regulatory issues criterion ␭ max⫽7.69; CR⫽0.09


Level-I 1.00 0.33 0.20 0.14 0.50 0.33 0.20
Level-II 3.00 1.00 0.25 0.20 0.33 0.33 0.33
Level-III 5.00 4.00 1.00 0.33 2.00 2.00 2.00
CalTOX 7.00 5.00 3.00 1.00 3.00 6.00 3.00
MULTIMED 2.00 3.00 0.50 0.33 1.00 2.00 1.00
DTSC-PEA 3.00 3.00 0.50 0.17 0.33 1.00 4.00
PC-GEMS 5.00 3.00 0.50 0.33 1.00 0.25 1.00

preted in any absolute sense. In addition, the application focused Summary and Conclusions
on screening-level risk assessments; supplementary criteria re-
lated to site heterogeneity may be necessary when model selec- The growing recognition that most contaminants partition and
tion for site-specific risk assessment has to be performed. Also, exist in more than one environmental compartment has prompted
additional input from a variety of stakeholders and multiple ex- regulators to assess exposure and risk to human health and eco-
perts needs to be considered when region and site-specific policy logical receptors via different exposure routes. Multimedia mod-
decisions with regard to the selection and use of multimedia en- els are necessary to obtain estimates for exposure concentrations
vironmental models are to be made. Given the availability of such along various pathways. While different multimedia modeling
data, AHP can be readily used to rank and evaluate alternative schemes have been proposed in the literature over the last two
multimedia models. However, it may be advantageous to famil- decades and several software codes are readily available in the
iarize the stakeholders with the mechanics of the AHP process public domain, the selection of an appropriate model for regula-
prior to its application to avoid any inconsistencies in the appli- tory use has proved to be a daunting task given the often con-
cation 共Lai et al. 2002兲 and this illustrative case study may also be founding and sometimes conflicting needs and preferences of
useful for such familiarization. various stakeholders and decision makers. In this study, the utility

PRACTICE PERIODICAL OF HAZARDOUS, TOXIC, AND RADIOACTIVE WASTE MANAGEMENT © ASCE / APRIL 2003 / 143

Pract. Period. Hazard. Toxic Radioact. Waste Manage. 2003.7:139-146.


Table 4. Aggregate Matrix of Local Priorities
Scientific credibility Data requirement Application domain End-user needs Regulatory issues Final weight
Level-I 0.03 0.38 0.04 0.13 0.04 0.09
Level-II 0.05 0.24 0.04 0.18 0.06 0.09
Level-III 0.23 0.10 0.04 0.24 0.18 0.19
CalTOX 0.35 0.07 0.13 0.35 0.37 0.31
MULTIMED 0.20 0.05 0.29 0.05 0.12 0.14
DTSC-PEA 0.02 0.13 0.07 0.02 0.13 0.07
PC-GEMS 0.12 0.02 0.38 0.05 0.11 0.11
Total 1.00 1.00 1.00 1.00 1.00 1.00
Downloaded from ascelibrary.org by PENN STATE UNIV on 08/11/14. Copyright ASCE. For personal use only; all rights reserved.

of a multicriteria decision-making technique, that is, the analytic become instantaneously distributed to an equilibrium condition.
hierarchy process, was demonstrated for the selection of an ap- The level-II model can be downloaded from http://www.trentu.ca/
propriate multimedia model. Five different criteria for model se- cemc/models.html
lection were developed from modeling practices and need survey. Level-III Fugacity Model: The level-III fugacity model was
Seven different public-domain multimedia models were compared described by Mackay and Patterson 共1990兲 and it represents a
in pair-wise fashion against the criteria developed using 20 dif- significant improvement over level-I and level-II formulations.
ferent subcriteria. AHP was used to rank the alternative models The phases are not assumed to be in equilibrium with one another
based on their performance versus these criteria. and as such intermedia transport can be simulated. The flow in
The results of the study indicate that, among the given models, various compartments can also be simulated and assumed to be
the CalTOX and level-III fugacity models were best suited for under steady-state conditions; degradation reactions assuming
carrying out screening-level risk assessments. However all the first-order kinetics are also simulated. The level-III model repre-
models only satisfied less than 31% of the needs of the stakehold- sents a significant conceptual improvement over level-I and
ers under consideration, indicating a large area for improvement. level-II models and is a more realistic representation of many
Based on the assessment of weight matrices obtained as part of environmental systems. The level-III model can be downloaded
the AHP process, coupling of CalTOX models with GIS and the from http://www.trentu.ca/cemc/models.html
addition of vegetative and groundwater compartments to the CalTOX: The CalTOX model was developed by McKone of
level-III fugacity modeling code and performing field-scale veri- Lawrence Berkeley National Laboratory for the DTSC of the
fication of these models are suggested as some possible improve- CAL/EPA. It is a hybrid level-III/level IV fugacity formulation
ments. that has been implemented on a spreadsheet and can be run in
Monte Carlo simulation mode using spreadsheet add-in software
like Crystal Ball and @RISK. This model predicts the time-
Acknowledgments dependent concentrations of a chemical in the seven environmen-
tal compartments of air, water, three soil layers, sediment, and
The stimulus for this work came from numerous discussions with plants at a hazardous waste site. The model is well documented
regulatory personnel in CAL/EPA especially the Office of Envi- and is described in a series of technical and user manuals. The
ronmental Health Hazard Assessment, whose cooperation is model and pertinent documentation are available from http://
gratefully acknowledged. The thoughtful comments and sugges- www.cwo.com/⬃herd1/caltox.htm
tions of two anonymous reviewers and Dr. Ni-Bin Chang have MULTIMED: simulates the fate and transport of contaminants
greatly helped improve the readability of the manuscript. that leach from a waste disposal facility into the multimedia en-
vironment. Release to either air or soil including the unsaturated
and saturated zones and possible interception of the subsurface
Appendix contaminant plume by a surface stream are included in the model.
The different compartments of the model are not tightly inte-
The seven different screening-level models used in this study are grated and the model was developed by externally linking various
briefly described below. The readers are encouraged to refer to the analytical and semianalytical solutions in various compartments.
technical documentation of the models for additional details. The model runs under MS-DOS however, pre- and post-
Level-I Fugacity Model: The level-I fugacity model was de- processors are available to assist users. The model can be down-
scribed by Mackay 共1979兲 and it is based on an idealized descrip- loaded from http://www.epa.gov/ceampubl/mmedia/multim2/
tion of no flow and equilibrium amongst phases. In other words, index.htm
all the phases possess equal fugacity and the mass input to the DTSC-PEA: The PEA tool was developed by the DTSC of the
system is instantaneously partitioned amongst phases according to CAL/EPA to determine whether current or past waste manage-
their specified volume and fugacity capacity. A version of the ment practices have caused the release of hazardous substances
level-I model can be downloaded from http://www.trentu.ca/ that pose a threat to human health and ecological resources. The
cemc/models.html threat to public health and/or the environment is assessed in a
Level-II Fugacity Model: The level-II fugacity model was de- multimedia framework using screening-level equations that have
scribed by Mackay 共1979兲 and it represents a conceptual improve- been developed using idealized single-media fate and transport
ment over level-I. While the phases are all still assumed to be in models that assume uniform flow and equilibrium partitioning.
equilibrium with one another, steady-state flow and degradation The equations are designed to provide conservative estimates
of the contaminants can also be simulated. The medium receiving using available field data and unlike fugacity models are not
the emission is unimportant because the chemical is assumed to mathematically linked by appropriate boundary conditions. The

144 / PRACTICE PERIODICAL OF HAZARDOUS, TOXIC, AND RADIOACTIVE WASTE MANAGEMENT © ASCE / APRIL 2003

Pract. Period. Hazard. Toxic Radioact. Waste Manage. 2003.7:139-146.


model equations can be easily implemented on a spreadsheet and Lai, V. S., Wong, B. K., et al. 共2002兲. ‘‘Group decision making in a
further details can be found by consulting the DTSC guidance multiple criteria environment: A case using the AHP in software se-
manual 共DTSC 1994兲. lection.’’ Eur. J. Oper. Res., 137, 134 –144.
PC-GEMS: PC-GEMS is a personal computer version of Laniak, G. F., Droppo, J. G., et al. 共1997兲. ‘‘An overview of a multimedia
benchmarking analysis for three risk assessment models: RESRAD,
EPA’s graphical exposure modeling system 共GEMS兲. It is a col-
MMSOILS, and MEPAS.’’ Risk Anal, 17共2兲, 203–214.
lection of several media-specific and multimedia models 共includ-
Lin, Z.-C., and Yang, C.-B. 共1996兲. ‘‘Evaluation of machine selection by
ing MULTIMED兲. The PC-GEMS system provides a wrapper the AHP method.’’ J. Mater. Process. Technol., 57共3– 4兲, 253–258.
framework in which many media-specific models can be imple- Lootsma, F. A. 共1993兲. ‘‘Scale sensitivity in the multiplicative AHP and
mented. An external linkage between a vadose zone and a subsur- SMART.’’ J. Multi-Criteria Decision Anal., 2, 87–110.
face model is provided in PC-GEMS. The modeling system pro- Lootsma, F. A. 共1999兲. Multi-criteria decision analysis via ratio and dif-
vides tools and programs to simulate the spatiotemporal ference, Kluwer, Dordrecht.
distribution of contaminants in air, water, soil, and aquifer sys- MacFarlane, S., and Mackay, D. 共1998兲. ‘‘A fugacity-based screening
tems. However, PC-GEMS is not a true multimedia modeling model to assess contamination and remediation of the subsurface con-
system since these compartments are not tightly integrated by taining non-aqueous phase liquids.’’ J. Soil Contaminat, 7, 17– 46.
Downloaded from ascelibrary.org by PENN STATE UNIV on 08/11/14. Copyright ASCE. For personal use only; all rights reserved.

appropriate boundary conditions. PC-GEMS is a public domain Mackay, D. 共1979兲. ‘‘Finding fugacity feasible.’’ Environ. Sci. Technol.,
code and can be obtained from the U.S. EPA on a CD-ROM 13, 1218 –1223.
Mackay, D., and Patterson, S. 共1990兲. ‘‘Evaluating the multimedia fate of
共http://www.epa.gov/opptintr/cahp/actlocal/pcgems.html兲.
organic chemicals: A level III fugacity model.’’ Environ. Sci. Technol.,
25, 427– 436.
References Maddalena, R. L., McKone, T. E., et al. 共1995兲. ‘‘Comparison of multi-
American Society for Testing and Materials 共ASTM兲. 共1995兲. ‘‘Standard media transport and transformation models: Regional fugacity model
guide for risk based corrective action applied at petroleum release vs. CalTOX.’’ Chemosphere, 30, 869– 889.
sites.’’ American Society for Testing and Materials, West Consho- McKone, T. E., Hall, D. et al. 共1996兲. ‘‘Modifications of CalTOX to
hocken, Pa., 761– 811. assess the potential health impacts of hazardous wastes landfills
American Society for Testing and Materials 共ASTM兲. 共1998兲. ‘‘Standard 共draft兲.’’ Office of Scientific Affairs, Department of Toxic Substances
guide for risk based corrective action.’’ American Society for Testing and Control, Sacramento, Calif.
and Materials, West Conshohocken, Pa., 1130–1230. National Regulatory Commission 共NRC兲. 共1999兲. Environmental cleanup
Ares, J., Miglierina, A. M., et al. 共1998兲. ‘‘CLFUG: A GIS-scalable at navy facilities—Risk based methods, National Academy, Washing-
model for pesticide fate in the soil-groundwater system based on ton, D.C.
clearance and fugacity paradigms.’’ Environ. Model. Assess., 3, 95– Office of Environmental Health Hazard Assessment 共OEHHA兲. 共1998兲.
105. California’s emerging environmental challenger: A workshop to iden-
Belton, V., and Gear, T. 共1983兲. ‘‘On the shortcoming of Saaty’s method tify future issues for CAL/EPA. California Environmental Protection
of analytic hierarchies.’’ Omega, 11, 228 –230. Agency, Sacramento, Calif.
Clemens, R. T. 共1997兲. Making hard decisions—Introduction to decision Paterson, M. R., et al. 共1984兲. An user’s manual for UTM-TOX and uni-
science, Duxbury, Belmont, Calif. fied transport model, Oak Ridge National Laboratories, Oak Ridge,
Cohen, Y. 共1989兲. ‘‘Multimedia and intermedia transport modeling con- Tenn.
cepts in environmental modeling.’’ Intermedia pollutant transport— Ramanathan, R. 共1999兲. ‘‘Selection of appropriate greenhouse gas mita-
Modeling and measurement, Y. C. D. T. Allen and I. R. Kaplan, eds., gation options.’’ Global Environmental Change: Human and Policy
Plenum, New York, 3–16. Dimensions, 9, 203–210.
Cohen, Y., and Cooter, E. 共2002兲. ‘‘Multimedia environmental distribu- Ramanathan, R. 共2001兲. ‘‘A note on the use of the analytic hierarchy
tion of toxics 共Mend-Tox兲. 1. Hybrid compartmental-spatial modeling process for environmental impact assessment.’’ J. Environ. Manage.,
framework.’’ Pract. Period. Hazard., Toxic, Radioact. Waste Manage., 63, 27–35.
6共2兲, 70– 86. Saaty, T. L. 共1980兲. The analytic hierarchy process: Planning, priority
Cousins, I. T., and Mackay, D. 共2001兲. ‘‘Strategies for including vegeta- setting, and resource allocation, McGraw-Hill, New York.
tion compartments in multimedia models.’’ Chemosphere, 44, 643– Saaty, T. L. 共1986兲. ‘‘Axiomatic foundations of the analytic hierarchy
654. process.’’ Manage. Sci., 18, 259–268.
Cowan, C. E., Mackay, D., et al. 共1995兲. ‘‘The multi-media fate model: A Saaty, T. L. 共1987兲. ‘‘Rank generation, preservation and reversal in the
vital tool for predicting the fate of chemicals.’’ Society of Environ- analytic hierarchy process.’’ Decision Sci., 18, 157–177.
mental Toxicology and Chemistry, Pensacola, Fla. Saaty, T. L. 共2000兲. Fundamentals of decision making and priority theory
Daugherty, S. J. 共1993兲. ‘‘Hydrocarbon contamination in the subsurface: with the analytic hierarchy process, RWS, Pittsburgh.
Realities, issues and prospects.’’ Hydrocarbon contaminated soils and Severinsen, M., Andersen, M. B., et al. 共1996兲. ‘‘A regional chemical fate
groundwater, P. T. Kostecki and E. J. Calabrese, eds., Lewis, Chelsea, and exposure model suitable for Denmark and its coastal sea.’’
3–18. Chemosphere, 32, 2159–2175.
Department of Toxic Substances and Control 共DTSC兲. 共1993兲. Uddameri, V. 共2002兲. ‘‘Knowledge management to support fate and trans-
‘‘CALTOX, a multimedia total exposure model for hazardous waste port modeling efforts in risk-based decision-making frameworks—
sites.’’ Department of Toxic Substances and Control, Sacramento, Salient issues and paradigm development.’’ Clean Technologies and
Calif. Environmental Policy, 4, 140–150.
Department of Toxic Substances and Control 共DTSC兲. 共1994兲. ‘‘Prelimi- Uddameri, V., and Young, T. M. 共1999a兲. ‘‘Survey of Cal/EPA fate and
nary endangerment assessment—Guidance manual.’’ Department of transport modeling practices and needs and recommended criteria for
Toxic Substances and Control, CAL/EPA, Sacramento, Calif. evaluating media specific models for integration 共draft report to Office
Elkarni, F., and Mustafa, I. 共1993兲. ‘‘Increasing the utilization of solar of Environmental Health Hazard Assessment兲.’’ Univ. of California,
energy utilization in Jordan–Analytic hierarchy process.’’ Energy Davis, Davis, Calif.
Policy, 21, 978 –984. Uddameri, V., and Young, T. M. 共1999b兲. ‘‘An evaluation of various
Hanna, M. 共1996兲. Introduction to management science—Mastering multimedia and single media fate and transport models.’’ Task 4 Re-
quantitative approaches, South-Western College Publishing, Cincin- port draft submitted to Office of Health Hazard Assessment, CAL/
nati. EPA, Univ. of California, Davis, Davis, Calif.
Kirkwood, C. W. 共1996兲. Strategic decision making—Multiobjective de- Uddameri, V., and Young, T. M. 共2000兲. ‘‘A critical review of fate and
cision analysis with spreadsheets, Duxbury, Belmont, Mass. transport models and algorithms used in screening level risk assess-

PRACTICE PERIODICAL OF HAZARDOUS, TOXIC, AND RADIOACTIVE WASTE MANAGEMENT © ASCE / APRIL 2003 / 145

Pract. Period. Hazard. Toxic Radioact. Waste Manage. 2003.7:139-146.


ment procedures.’’ Task 4a Report draft submitted to Office of Health U.S. Environmental Protection Agency 共USEPA兲. 共2002兲. ‘‘Integrated
Hazard Assessment, CAL/EPA, Univ. of California, Davis, Davis, Geographical Exposure Modeling System 共IGMS兲.’’ Information
Calif. available online at http://www.epa.gov/opptintr/exposure/docs/ gem-
U.S. Environmental Protection Agency 共USEPA兲. 共1996兲. ‘‘Soil screening s.htm, U.S. EPA.
guidance: User’s guide.’’ Office of Emergency and Remedial Re- Wania, F. 共1996兲. ‘‘Spatial variability in compartmental fate modeling—
sponse, Washington, D.C. Linking fugacity models to GIS.’’ Environ. Sci. Pollut. Res., 3, 39– 46.
U.S. Environmental Protection Agency 共USEPA兲. 共1998兲. ‘‘A multimedia Woodfine, D., MacLeod, M., et al. 共2002兲. ‘‘A regionally segmented na-
strategy for priority persistent, bioaccumulative, and toxic 共PBT兲 pol- tional scale multimedia contaminant fate model for Canada with GIS
lutants.’’ U.S. Environmental Protection Agency, Washington, D.C. input and display.’’ Environ. Pollut., 119, 341–345.
Downloaded from ascelibrary.org by PENN STATE UNIV on 08/11/14. Copyright ASCE. For personal use only; all rights reserved.

146 / PRACTICE PERIODICAL OF HAZARDOUS, TOXIC, AND RADIOACTIVE WASTE MANAGEMENT © ASCE / APRIL 2003

Pract. Period. Hazard. Toxic Radioact. Waste Manage. 2003.7:139-146.

Vous aimerez peut-être aussi