Chapter 20

Groundwater Flow and Solute-Transport

Modeling

20.1 Introduction

Numerical modeling has become an integral component of most water-resources

investigations. The primary objective of groundwater modeling is the prediction of
aquifer responses to current and future groundwater use and other hydrologic
changes. An additional benefit of groundwater modeling is that it can provide
insights on the properties of hydrologic systems through the calibration process.
Two basic types of modeling have wide applications for groundwater resources
modeling: flow and solute transport. Groundwater flow models simulate the
hydraulic behavior of the aquifer such as the flow of water, changes in volume of
water in storage, and changes in aquifer water levels or heads (pressure). Solute-
transport models simulate the fate and transport of dissolved constituents in
groundwater, which include salts and contaminants. Additional common types of
modeling include heat-flow and integrated surface water-groundwater interactions.
Computer modeling has to a large degree supplanted the previously used
analytical (equation-based) modeling for the following reasons:
• Computer modeling techniques provide much greater flexibility and capability
to simulate complex geologic systems.
• The quality of groundwater modeling codes, particularly their flexibility and
ability to simulate natural systems, has progressively improved.
• The historical exponential increase in computation power allows for complex
models to be quickly executed on personal computers.
• Pre-processing and post-processing (graphical user interface or GUI) software
packages (e.g., Visual ModflowTM) have greatly simplified the construction and
running of models, analysis and interpretation of the modeling results, and the
preparation of high-quality graphic outputs, such as contour maps and three-
dimensional animations.

R. Maliva and T. Missimer, Arid Lands Water Evaluation and Management, 475
Environmental Science and Engineering, DOI: 10.1007/978-3-642-29104-3_20,
 Springer-Verlag Berlin Heidelberg 2012
476 20 Groundwater Flow and Solute-Transport Modeling

Groundwater modeling is a major subdiscipline onto itself and there is a

voluminous literature of journal articles and technical papers on all aspects of the
subject, but few dedicated textbooks on applied numerical groundwater modeling.
‘‘Applied Groundwater Modeling, Simulation of Flow and Advective Transport’’
by Anderson and Woessner (1992) provides a good introduction and overview of
groundwater modeling techniques and is still the most widely used text. Hill and
Tiedeman (2007) provide an overview of model calibration and Zheng and Bennett
(2002) provide a good introduction to contaminant transport modeling.
Arguably, it has become too easy now to perform groundwater modeling
(Maliva and Missimer 2010). Groundwater modeling software packages have
become so user friendly that they allow groundwater flow and solute-transport
models to be developed by groundwater professionals who lack both formal
modeling training and a familiarity with the mathematical and hydrogeological
assumptions and limitations inherent in the model codes. It has become readily
possible to develop and run models and generate impressive graphic outputs,
which may be fundamentally wrong. Groundwater models are also increasingly
being developed by well-trained modelers in corporate ‘‘modeling centers,’’ who
lack the critical understanding of local hydrogeology. Groundwater modeling is
best performed as team activity, in which an experienced modeler teams with
people having local hydrogeological expertise to produce a superior product.

20.2 Groundwater Modeling for Arid and Semiarid Lands

Groundwater modeling procedures for arid and semiarid lands are basically similar
to those used for more humid climates. However, arid and semiarid regions pose
some greater challenges caused by the great spatial and temporal variation in
precipitation, evapotranspiration, surface water flow, and groundwater recharge,
which are the primary inputs of the water-balance. Arid lands aquifers are also
stressed to a greater degree than aquifers in humid regions (Lloyd 2004), with
stress considered as the ratio of annual abstractions to recharge. Accurate esti-
mation of aquifer recharge and discharge is particularly difficult to perform
because of the inherent difficulties in making such measurements. In addition,
current hydrological conditions that may not be representative of prior wetter
conditions during which most recharge occurred. Large, regional aquifers may not
be in hydraulic equilibrium with current natural recharge and discharge, much less
recent well abstractions.
Hydrologic data are commonly sparse in arid lands with a low density of
boreholes both spatially and with depth (Lloyd 2007). Data are particularly sparse
for the deep sections of aquifers. Hydraulic data are often obtained from single
well tests performed on partially-penetrating wells. Hydraulic data obtained from
shallow wells (e.g., storativity, average hydraulic conductivity) are often applied to
the entire aquifer thickness, not considering likely changes with depth or the issue
of partial penetration corrections (Lloyd 2007). The absence of data on aquifers
20.2 Groundwater Modeling for Arid and Semiarid Lands 477

‘‘pales into insignificance when characteristics of intervening ‘non-aquifer’ units

are considered!’’ (Lloyd 2007, p. 43). Data are usually not available on the vertical
hydraulic conductivity of semi-confining units, which have a major influence on
inter-aquifer flows.
Evaluation of large-scale flow systems also poses technical challenges. Multi-
aquifer flow systems are often present within groundwater basins, with evaluation
of cross-boundary flows between aquifers being extremely difficult, if not
impossible, to evaluate, based on the available data. Assessment of regional
aquifers needs to consider local faulting, folding, and facies changes, which can
have a considerable influence on groundwater flow, particularly by causing dis-
continuities within aquifers (Lloyd 2007). The sequence stratigraphy of sedi-
mentary aquifers, which greatly affects the vertical and horizontal distribution of
hydraulic conductivity, rarely has been characterized (Maliva and Missimer 2010).
Interbasin groundwater flow (intermediate and regional flow system) may exist
to varying degrees in both arid and humid regions. However, interbasin flow may
have a proportionally greater effect on to basin water budgets in arid lands (and
thus needs to be considered in groundwater models) because of low recharge rates.
Groundwater models in arid lands can provide a broad understanding of large-
scale groundwater flow, but there are often insufficient data to accurately simulate
local groundwater flow (Lloyd 2007).

20.3 Limitations of Groundwater Modeling

The most fundamental limitation on groundwater modeling is that it does not result
in unique solutions. Even if a model result is consistent with present and past
observational data, there is no guarantee that the model will perform at an equal
level when used to predict the future (Oreskes et al. 1994), particularly if future
input conditions are well outside the range of historic conditions. The results of
predictive simulations should therefore be viewed as best estimates. There can be a
false perception that as a numerical technique, modeling inherently produces more
scientific, objective, and reliable results. Perhaps the most effective use of pre-
dictive models is as a tool under continuous refinement in which model predictions
are regularly compared to hydrologic monitoring data and the model is updated
and recalibrated as necessary.
Although numerical groundwater models can incorporate considerable hydro-
geologic detail, they are still simplifications of complex hydrogeological systems in
which there will never be sufficient data to fully characterize the aquifer system in
question. There will always be some uncertainty associated with modeling results.
The pertinent question is whether or not a model output adequately represents real
world conditions for the purposes for which the model is to be used. Models do not
have to be perfect to be useful from a practical water management perspective.
The quality or accuracy of the output from groundwater models is dependent on
the quality of the data incorporated into the model. The computer software adage
478 20 Groundwater Flow and Solute-Transport Modeling

‘‘garbage in, garbage out’’ is certainly applicable to groundwater modeling.

Groundwater modeling has very limited value in the absence of quality hydro-
geologic data and accurate conceptual models. There is a movement towards more
stochastic (probability-based) groundwater modeling, particularly in the academic
community. However, detailed and high-quality field data are still necessary to
provide the geostatistical parameters needed for stochastic modeling. Without
accurate field data, the principle of non-uniqueness can render any groundwater
modeling results to be inconclusive and of questionable value. Hydrogeological
data are critical for constraining model construction and narrowing the envelope of
possible solutions, and thus uncertainty.
The uncertainty associated with modeling results can be evaluated in part
through sensitivity analyses. Large-scale models developed for groundwater
management should not viewed as an end product, but rather as a work in progress,
with progressive refinement and improvement anticipated as more data become
available. Numerical modeling is an invaluable tool for groundwater management,
but it can be abused when it is used to make decisions beyond the capabilities of
the models (Missimer et al. 2010). Problems lie not in the uncertainty associated
with models, but rather in the failure to understand and consider the uncertainty
inherent in models when using them to make decisions.
An additional source of uncertainty, which may be important in some arid and
semiarid lands, lay in the input data. For example, the values of key input variables,
such as groundwater pumping rates, may have a large uncertainty if there is inade-
quate monitoring data. A model may thus provide a good representation of aquifer
behavior, but predictive simulations using the model may have large errors if, for
example, current and future groundwater pumping rates are substantially greater than
expected. Discrepancies between observed and measured data may thus be indicative
of problems associated with the input and monitoring (observational) data.

20.4 Groundwater Model Development

The development of a groundwater model includes several stages or processes,

which may or may not be performed as discrete project phases. For example, an
experienced modeler may intuitively know which model code is most appropriate
for a given project. A general outline of the model development process, which is
similar to the model protocol proposed by Anderson and Woessner (1992),
is herein presented.

20.4.1 Statement of Problem and Modeling Objectives

The specific reason for performing modeling and the desired model output should
be established. The modeling objectives should also considered, such as the
20.4 Groundwater Model Development 479

desired accuracy of simulated drawdowns, and whether or not they are realistically
obtainable. Consideration should also be given to the economic and technical
resources that are available for a project, which will dictate what is possible.

20.4.2 Conceptual Model

Bredehoeft (2005, p. 38) described the role of the conceptual model as

Every model has as its foundation a conceptual model. The conceptual model is the basic
idea, or construct, of how the system or process operates; it forms the basic idea for the
model.

The conceptual model may include such factors as the model boundaries and
boundary conditions, aquifers included and their interrelationship, recharge and
discharge sources and rates, predominant type of groundwater flow (matrix versus
conduit), aquifer heterogeneity, and active geochemical processes. The later stage
of conceptual model development involves consideration of the number and
geometric array of the cells (or elements) in either finite-difference or finite ele-
ment models.
If the conceptual model does not accurately represent the basic hydrogeologic
conditions of the study area, then the results of any modeling based on the con-
ceptual model will be questionable. Numerical modeling is the method by which
the appropriateness of the prevailing concept is tested. Conceptual and numerical
model development should be viewed as an iterative process in which the con-
ceptual model is continuously reformulated and updated (Bredehoeft 2005). Poeter
(2007) claimed that the uncertainty associated with conceptual models is typically
larger than measures of uncertainty that can be calculated for any individual model
(i.e., uncertainty associated with estimates of parameter values).
The term ‘surprise’ is defined as (Bredehoeft 2005, p. 38)
The collection of new information that renders one’s original conceptual model invalid.

In other words ‘surprise’ results in a complete paradigm shift. Bredehoeft

(2005) proposed that surprises are more frequent (perhaps 20–30% of models) than
is commonly realized. Ideally, alternative conceptual-mathematic models should
be constructed and tested to determine the best fit and to fully capture the
uncertainty associated with a modelling investigation (Poeter 2007). However, this
idea is typically not put into practice (Bredehoeft 2005), because of the great
amount of effort it would require. Instead, the problems associated with incorrect
conceptual models can be reduced by (Bredehoeft 2005)
• collecting as much data as feasible and necessary using all applicable methods and
• remaining open to fact that the conceptual model can change dramatically as
more information is collected.
480 20 Groundwater Flow and Solute-Transport Modeling

Block-Centered Finite- Finite-Element Grid

Difference Grid
No flow
cells

Model
boundary

River

River
cells

Model
node

Fig. 20.1 Diagram of finite-difference and finite-element discretization. The finite-element

discretization has greater flexibility to model irregularly shaped features. The finite-difference
discretization can be used to simulate finer scale features by reducing the grid size

20.4.3 Choice of Model Code

Some of the commonly used groundwater models are discussed in Sect. 20.5. The
model code used for an investigation should be amenable to meeting the modeling
objectives. Model choice issues include:
• finite difference versus finite-element discretization,
• whether or not surface water-groundwater interaction needs to be simulated,
• whether or not simulation of flow through the unsaturated zone is needed,
• whether or not solute-transport (either density dependent or independent) needs
to be simulated,
• model acceptability and defensibility, and
• open-source versus proprietary code.
Two main types of numerical models are used for three-dimensional modeling
of groundwater flow in aquifers, which are the finite-difference and finite-element
types. The model types differ in the manner of discretization or subdividing the
aquifer into a three-dimensional grid. The finite-difference method subdivides the
aquifer into regular rectangular grid blocks. In block-centered, finite-difference
models, such as MODFLOW, hydraulic heads are calculated at nodes located in
the center of the grid blocks (Fig. 20.1). The finite-element method uses an
irregular arrangement of node points commonly arranged is triangular shapes
(Fig. 20.1). The finite-element grid has greater flexibility to simulate irregularly
shaped features, such as rivers and wells, which can be placed on nodes. Most
geological and hydrogeological features do not have an orthogonal geometry, so
the flexibility of the finite-element grid is advantageous. However, finite-difference
codes are mathematically simpler and more stable.
20.4 Groundwater Model Development 481

The model codes used for an investigation must meet the technical require-
ments of the project. The code must also meet the requirements of the system
owner, regulatory agencies, and other groups or professionals that may use or
review the model. Accessibility of a model to future potential users is an important
consideration, where a model is to be used as a long-term water management tool.
From an owner’s perspective, there is great value in having people other than the
model developer be able to review a model and in the future, update it and perform
simulations as needed. Therefore, widely used and available model codes should
be used where possible.
An important distinction occurs between public-domain, open-source codes,
and proprietary software in which the source code is a trade secret. As a general
principle, public domain software is preferred over a proprietary software package,
unless the latter have features or capabilities necessary for a project that are not
readily available in public domain software. The critical advantages of public
domain software include the following:
• The source codes for public-domain modeling programs are available for
independent review and analysis, which is not possible for proprietary software
where the source code is kept secret. Proprietary software is in essence a ‘‘black
box’’ in which the users must accept their validity and accuracy largely as a leap
of faith.
• Public domain software programs, such as MODFLOW, benefit from refine-
ments and additions that are possible from the pool of users.
• Proprietary software programs may not be admissible in legal settings if the
source code is not released.
• Public domain software packages may have greater acceptance, particularly in
regulatory settings.
• Public domain software is free, although proprietary pre- and post-processing
(graphic user interface) software is often privately developed and commercially
distributed.
Proprietary model codes may not be acceptable in some circumstances. In the
United States, modeling results are commonly ruled inadmissible in various court
settings during litigation unless the model code is also provided. Usually the
developer of the proprietary model code will not release the model code because it
is a trade secret, which would make simulations run with the code inadmissible.
Many water management and regulatory agencies also prefer or require that
modeling results submitted in support of permits must be performed using com-
monly used public domain codes, so as to allow for their independent review. As a
general principle, transparency is desirable in science and open-source, public-
domain codes have a clear advantage in this respect. However, some projects may
have modeling needs that can be best met with advanced proprietary software. If
the capabilities of a proprietary code are needed, then preference should be given
to codes that are widely commercially available. Sometimes, it is necessary to use
a proprietary code to assess a problem and then re-model the problem using an
open-source so that it can be used in courtroom testimony or in water management
482 20 Groundwater Flow and Solute-Transport Modeling

decisions. This method is fraught with problems because the assumptions between
the models may not be similar and the results may not be exactly the same.

20.4.4 Model Development and the Population of the Model Grid

Values for the various physical, hydraulic, and chemical parameters must be
assigned to all points or cells in the model grid. A wide variety of different types of
data are collected as part of hydrogeological investigations. A challenge exists
concerning how to best integrate and incorporate data from diverse sources, such
as lithologic logs, borehole geophysical logs, surface geophysics, packer (drill
stem) tests, remote sensing, and aquifer testing into groundwater models. Much of
the hydrogeological data is obtained from point sources (e.g., wells) and must be
interpolated and extrapolated in order to populate model grids. GIS has become a
powerful tool for the population of model grids because its capabilities to store and
process georeferenced data (Chap. 19).
In the absence of site-specific data, data from regional models may have to be
used or best estimates based on aquifer lithologies. A variety of methods have been
developed in the oil and gas industry to predict the three-dimensional distribution of
rock and sediment types, and thus hydraulic parameters, based on data from a limited
number of wells. Sedimentologic (facies) and sequence stratigraphic analyses
provide a means for projecting the distribution of sediment and rock types based on
the concept that these sediments were deposited in a non-random, predictable
manner. Stochastic methods also exist for populating a grid based on limited data.
Storage of geologic and hydrogeologic data in three dimensions allows interpo-
lation of three-dimensional (3D) hydrostratigraphic units, designation of measured
or interpolated properties to the units, volumetric calculations, morphology analysis,
representation of subsurface structures, parameter flux (i.e., groundwater flow,
chemical diffusion) between units, and interpolation of hydrologic properties
(National Research Council 2008). Three-dimensional data management and visu-
alization software are currently commercially available (e.g., HydroGeoAnalystTM).
Much more sophisticated workflow software have been developed for the oil
and gas industry, which can incorporate the wide variety of data collected from
reservoir investigations and development activities into models that simulate
complex reservoirs. For example, the Petrel workflow developed for the oil and
gas industry offers a set of aquifer characterization, modeling, and simulation
programs that can help develop the most efficient and cost-effective solutions to
managed aquifer recharge and other water management problems (Herrmann
2006). Petrel allows hydrogeologists, geochemists, geologists, and water manag-
ers, to share data and results within the same environment.
The Petrel workflow system allows the integration of the following disciplines:
• geophysical interpretations,
• surface imaging and mapping,
20.4 Groundwater Model Development 483

Water Elevation (m)

4

0 Computed
Observed

0 200 400 600

Time (days)

Fig. 20.2 Example of calibration of a groundwater flow model to water elevation data from an
observation well (Lee County, Florida). Model captures the seasonal variation in water levels

• geological log interpretation and well correlation,

• complex faulting and fracture modeling,
• facies and geophysical modeling,
• aquifer testing results,
• uncertainty analysis,
• surface and subsurface interaction,
• up-scaling processes, and
• flow and mass transport simulation.
The ultimate data management goal should be to have all of the hydrogeologic
data for a region in a single central GIS-based database (Sect. 18.3.5), from which
modelers could readily extract the data needed for modeling investigations. The data
would also include input data (e.g., pumping and rainfall records) and observational
data for model calibration (e.g., monitoring well locations and historic data).

20.4.5 Model Calibration

Model calibration is the process by which the values of model inputs (hydraulic
parameters) are adjusted so that the model output matches observed data
(Fig. 20.2). The observed data, referred to as calibration targets, may include
• water levels (pressures) in wells (static or in time-series),
• surface-water body elevations and flows (static or in time-series), and
• water quality data (e.g., salinity) measured in wells or recovered water (static or
in time-series).
The limitation of non-calibrated simulations is that their accuracy can only be
as good as the values assigned to the hydraulic parameters. The results of non-
calibrated simulations thus have potential large, unquantifiable uncertainties
(errors).
484 20 Groundwater Flow and Solute-Transport Modeling

Confidence in a model to provide an adequate representation of the system

under consideration generally improves with an increase in the number of cali-
bration targets and their geographic distribution, a longer duration of the cali-
bration period, and a greater range of conditions or stresses during the calibration
period. For example, it is preferable that the model calibration period includes both
wetter and drier than average periods.
Model calibration is evaluated by analysis of the residuals (errors), which are
the difference between simulated and observed values. The main calibration sta-
tistics used in groundwater modelling are mean error (ME), mean absolute error
(MAE), root mean square error (RSME), and weighted least square residual
(WLSR).
Mean error, mean absolute error, and root mean square error are defined with
respect to heads as:

1X n
ME ¼ ðhm  hs Þi ð20:1Þ
n i¼1

1X n
MAE ¼ jhm  hs ji ð20:2Þ
n i¼1
" #0:5
1X n
RMSE ¼ ðhm  hs Þ2i ð20:3Þ
n i¼1

where,
n = number of calibration targets
hm = measured head (m, ft)
shs = simulated head (m, ft)
Mean error is sensitive to the directions of the error (whether simulated values
are greater or less than the measured values) and large errors between target
calibrations may cancel out if they are in opposite directions. Calibration statistics
commonly provide measures of the average error, as all targets are given an equal
weight. The weighted least square residual (WLSR) includes a weighting factor
(xi), which reflects the expected measurement error (i.e., observation data that
may have great errors or variance are given a lower weight) or proportional
importance allocated to target ‘‘i’’ and is defined as

WLSR ¼ xi ðhm  hs Þ2i ð20:4Þ

The objective of the calibration procedure is to minimize the residuals. Model
calibration has been historically performed manually using a systematic trial and
error process. The values for a model parameter in all or a spatial zone of the model
grid are adjusted either upwards or downwards in increments to try to obtain a better
fit (lower residual) to the calibration targets. An understanding of the hydrogeologic
system in question and the response of models to changes in the various parameters
allows the process to be performed efficiently. A key component of the calibration
20.4 Groundwater Model Development 485

process is knowledge regarding the real range of the hydraulic parameters within
which calibration should lie. Calibration using unrealistic hydraulic parameters
outside of the realistic range of variation may lead to models that are well-calibrated
but poorly representative of real conditions. In an area where there are a few high
quality aquifer performance tests (APTs) with a corresponding high reliability of
the hydraulic data, these points can be held nearly constant in the calibration process
and the other data varied to meet the calibration targets.
Calibration of groundwater flow models should also be performed using data
other than just aquifer heads (if such data are available). For example, environmental
isotope data can provide constraints on boundary conditions (Herczeg and Leaney
2011). Groundwater age data can also be used for model calibration, but the
limitations inherent in groundwater age data must be considered (Sanford 2011).
Software is available that automatically performs model calibration. The PEST
parameter estimation and optimization software program is increasingly being
used for model calibration (Doherty 2005). The PEST program takes control of the
model and automatically performs as many model runs as it needs, while adjusting
the parameters of the model until the weighted least squares residual is reduced to
a minimum value. Much has been written on the relative merits of manual versus
automated calibration. Automated calibration (like manual calibration) is the
determination of parameter values that produce the best model fit with observa-
tions for a given conceptual model (Poeter and Hill 1997). The major advantages
are that it expedites that process of adjusting parameters, and thus allows for more
calibration runs to be performed in a given time period, and that it may reveal
issues overlooked in manual calibration (Poeter and Hill 1997).
A criticism of manual calibration based on a limited number of homogenous
parameters is that it results in a great degree of simplification (parsimony). Hunt
et al. (2007) presented the advantages of automated calibration through regularized
inversion, which can maximize insights gained from field data while also intro-
ducing some spatial variability to improve fit.
However, because of the non-uniqueness of model calibration, the best fit
solution may not necessarily result in the most accurate predictive model, par-
ticularly where the predictive simulation scenario differs significantly from the
calibration model conditions. Adjustments in model parameters during model
calibration should be hydrogeologically reasonable. If the best fit obtained by
automated calibration requires the use of hydrogeologically unreasonable values
for parameters, then the results may indicate problems with the conceptual model
or perhaps the observational data (e.g., pumping information from wells in the
model area) (Poeter and Hill 1997). A key point is that automated calibration,
when properly employed, is not a quick and easy method, but still requires a high
level of technical sophistication and local hydrogeological understanding by the
modeler and team of other experts working on the project.
An important issue in groundwater modeling is that simulated heads are largely a
function of hydraulic conductivity and recharge rates (Anderson and Woessner
1992). A decrease in hydraulic conductivity and increase in recharge can both act to
increase simulation heads. There is thus no unique solution to the model calibration.
486 20 Groundwater Flow and Solute-Transport Modeling

Recharge rates are poorly known, especially in arid and semiarid regions. The range
of possible recharge rates in arid and semiarid ranges can be narrowed by inde-
pendent estimates of the recharge rate obtained from groundwater ages. For
example, Zhu (2000) calibrated simulated 14C ages from a numerical flow and
transport model against observed 14C age, in addition to a calibration against
aquifer heads. Such an approach narrows the envelope of potential recharge rate
and hydraulic conductivity values. The age-correction of observed 14C values
allowed for the use of a non-reactive transport model, which needed to simulate
only advection-dispersion/diffusion and radioactive decay (Zhu 2000).
A sensitivity analysis is a critical component of the model calibration process as
it can provide approximate error bars for the model results. Sensitivity analyses are
performed by running predictive simulations in which the values of hydraulic
parameters are adjusted by fixed amounts (e.g., upwards and downwards by 50%).
The goal of the sensitivity analysis is to quantify the effects of the inherent
uncertainty in the values of hydraulic parameters on model results. If the results of
the sensitivity analysis show only a minor change in predictive simulations results,
then a high degree of confidence can be placed on the model. Conversely, if a
modest adjustment in the value of a hydraulic parameter that has a high degree of
uncertainty causes a substantial change in simulation results, then a large uncer-
tainty (error bars) exists for the simulation results.

20.4.6 Model Verification, Post Audits, and Refinement

The purpose of model verification is to establish greater confidence in a calibrated

model by performing additional simulations using new stress conditions and
observational data (Anderson and Woessner 1992). For example, a model could be
calibrated for a 2-year period, and then a verification simulation could be per-
formed on the data for a third year. If there is a good match between the simulated
and observed data for the third year, then greater confidence may be placed in the
ability of the model to adequately represent real conditions.
The term verification means the establishment of truth by demonstration. It is
not possible to verify or validate any numerical model in the sense that the veracity
of a model is established (Oreskes et al. 1994). Instead the burden on a modeler is
to ‘‘demonstrate the degree of correspondence between the model and the material
world it seeks to represent and to delineate the limits of that correspondence’’
(Oreskes et al. 1994, p. 644).
The model verification process, as the term is widely used in the modeling
community, should be an ongoing process. Post audits should be performed in
which predictions are compared to what actual happened (i.e., good-quality
monitoring data) in order to confirm that the model does indeed provide an
acceptable representation of real world conditions and to obtain insights into future
possible model enhancements (Anderson and Woessner 1992; Lloyd 2004).
If there is a significant discrepancy between predicted and observed data, then the
20.4 Groundwater Model Development 487

model should be refined and recalibrated. Models should also be revaluated and
refined, as necessary, as new hydrogeological data become available, particularly
if it does not support the conceptual model of the study area.

20.4.7 Predictive Simulations

The end product of groundwater model development should be a model that is well
calibrated against available monitoring data and, as such, can accurately predict
future system performance. Predictive simulations can then be performed to assess
issues of water management concern, such as changes in aquifer water levels as the
result of the continuation of current pumping rates or as a result of changes in
future water use (new groundwater development).

20.5 Groundwater Flow Modeling Codes

The choice of software for a groundwater modeling project depends upon:

• model code capabilities versus project needs,
• accessibility of the model code (private versus public domain), and
• acceptability of the model to the system owner, regulatory agencies, courts, and
the potential user community.
An additional factor that comes into play during model code selection is pro-
vincialism. In the United States, most groundwater modeling is performed using
software codes developed by or for the U.S. Geological Survey, U.S. Environ-
mental Protection Agency, or other government agencies. These codes are nearly
universally accepted, are usually publically available, and have undergone
extensive review and improvements because they are in the public domain. The
U.S. Geological Survey MODFLOW code is very widely used for groundwater
flow modeling and has great regulatory acceptability. The public domain codes
have been packaged into proprietary commercial software products (e.g., Visual
MODFLOW). The U.S. government agency codes are less commonly used in
Europe and elsewhere in preference for locally developed software. There may
also be an element of inertia in the choice of codes as a modeler or project owner
may be more familiar with or normally prefer the use one particular code.
Model codes vary in their discretization (finite-difference versus finite-element)
and their capabilities concerning simulation of solute transport and heat flow.
Solute-transport models are required to simulate the fate and transport of con-
taminants and the interface between fresh and saline waters. A density-dependent
solute transport model is needed for simulations of systems in which there are
significant differences in fluid density such as those resulting from differences in
salinity or temperature. Reactive solute-transport codes can simulate geochemical
processes during groundwater flow, such as mineral dissolution and precipitation.
488 20 Groundwater Flow and Solute-Transport Modeling

20.5.1 MODFLOW

The U.S. Geological Survey MODFLOW code (McDonald and Harbaugh 1988) is
the industry standard in the United States and is widely used elsewhere. MOD-
FLOW has evolved into a family of codes, which were built upon the basic
MODFLOW groundwater flow code. McDonald and Harbaugh (1988) provide an
excellent description of the model and its underlying mathematics. Domenico and
Schwarz (1998) and Fetter (2001) provide a summary of the MODFLOW family
codes, which continues to grow. Codes that have been developed upon or utilize
the outputs of MODFLOW include MODPATH (Pollock 1994; particle tracking),
MT3DMS (Zheng and Wang 1999, solute transport), SEAWAT (Guo and
Langevin 2002, density dependent solute-transport), PHT3D (Prommer et al. 2003,
reactive solute transport), and SEAWAT Version 4, which includes heat transport
(Langevin et al. 2008)
MODFLOW is a three-dimensional, finite-difference, ground-water flow model.
The modular design and public-domain status of MODFLOW has allowed new
capabilities to be added to the original model. The great advantage of the modular
design is that new features or processes can be added to the model with minor
modification of the existing core code. The original code included packages
(modules) to simulate basic features and processes, such as rivers, recharge,
evapotranspiration, drains, and wells.
The MODFLOW code is widely used in both academia and industry, and has
undergone a great deal of review. New or improved modules have been developed
to better simulate a diversity of hydrogeological processes. The community using
MODFLOW is very large, which has allowed numerous improvements to be made
to the model. There are international MODFLOW conferences at which new or
expanded uses of the model are discussed.

20.5.2 MicroFEM

MicroFEM is a three-dimensional, finite-element program for steady-state and

transient groundwater flow modeling (Hemker and de Boer 2009). The Microsoft
Windows version of MicroFEM includes a graphic user interface and packages for
grid generation, data input, presentation of results, and model optimization.

20.6 Groundwater Model Codes for Solute Transport

A number of solute-transport codes have been developed that have or can be used
for the simulation of contaminant transport, managed aquifer recharge, saline-
water intrusion, and other processes of water management concern. Differences
20.6 Groundwater Model Codes for Solute Transport 489

between codes include whether finite-difference or finite-element discretization is

used, and whether or not the models simulate density-dependent flow and heat
transport. Solute-transport codes incorporate groundwater flow codes (or are built
upon groundwater flow codes such as MODFLOW) and can therefore be used to
model groundwater flow alone. Some of the more commonly used solute transport
models are herein summarized.

20.6.1 MT3DMS

MT3DMS (Zheng and Wang 1999) is a modular three-dimensional multispecies

transport model for simulation of advection, dispersion and chemical reactions of
contaminants in groundwater systems. The MT3DMS code is developed for use
with a block-centered, finite-difference flow model such as MODFLOW, and is
based on the assumption that changes in the solute concentration do not signifi-
cantly impact groundwater flow (flow is not density dependent). The MT3DMS
code has a modular structure similar to MODFLOW, which allows for the sim-
ulation of advection, dispersion/diffusion, source/sink mixing, and chemical
reactions. The modular design allows for other transport processes or reactions to
be added to the model readily without having to modify the existing code. The
combination of MODFLOW and MT3DMS can be used to simulate density-
independent solute transport.

20.6.2 SEAWAT

SEAWAT (Guo and Langevin 2002; Langevin et al. 2008) is a coupled version of
the MODFLOW and MT3DMS codes designed to simulate three-dimensional,
variable-density, saturated groundwater flow. The latest version (Version 4) of
SEAWAT added the capability to simulate heat transport. SEAWAT is now
widely used in the United States to simulate density-dependent solute-transport in
coastal settings. A SEAWAT model can be developed using existing MODFLOW
models for the initial flow component. SEAWAT is suitable for the simulation of
ASR systems in aquifers containing brackish or saline waters and for the simu-
lation of saline-water intrusion.

20.6.3 SUTRA

SUTRA (Saturated Unsaturated Transport Code; Voss and Provost 2002) was
originally a two-dimensional computer program developed by the U.S. Geological
Survey that simulates fluid movement and the transport of either energy or
490 20 Groundwater Flow and Solute-Transport Modeling

dissolved substances in a saturated or unsaturated subsurface environment. The

code was later modified for three-dimensional simulations. SUTRA simulations
are based a hybridization of the finite-element and integrated finite-difference
methods. The discretization consists of quadrilateral elements in two dimensional
simulations and hexahedral elements in three-dimensional simulations. The hybrid
method preserves the geometric flexibility of the finite-difference method, while
taking advantage of the finite-difference efficiency (Voss and Provost 2002). The
SUTRA codes can simulate density-dependent flow and are thus suitable for
simulating systems involving saline-water intrusion and ASR systems using
brackish-water aquifers.

20.6.4 HST3D

HST3D (Heat- and Solute-Transport; (Kipp 1987, 1997) is a three-dimensional

code developed by the U.S. Geological Survey for the simulation of groundwater
flow and associated heat and solute transport. HST3D is a finite-difference, vari-
able-density code that can be used for the analysis of subsurface waste injection,
landfill leaching, saltwater intrusion, freshwater recharge and recovery, radioac-
tive-waste disposal, hot-water geothermal systems, and subsurface-energy storage.

20.6.5 FEFLOW

FEFLOW (Finite Element subsurface FLOW system; Diersch 1998) is a three-

dimensional, finite-element simulation package that can be used to model
groundwater flow and solute transport in porous media under saturated and
unsaturated conditions. FEFLOW can be used in two or three-dimensional modes,
and for steady state and transient simulations. FEFLOW can simulate coupled-
density dependent flow and is thus capable of simulating saline-water intrusion and
ASR systems using brackish-water aquifers. FEFLOW is proprietary software and
is commercially available in a package that includes pre- and post-processing
software.

20.6.6 ECLIPSEÒ

ECLIPSE is proprietary reservoir simulation software developed by Schlumberger

Limited for the oil and gas industry, and has been the benchmark core in reservoir
simulations for over 25 years. ECLIPSE is a finite-volume simulator that allows
for the modeling of multi-component (water, oil, and gas) and multi-phase (liquid,
gas, gas in solution) systems. ECLIPSE can be used to model individual wells and
20.6 Groundwater Model Codes for Solute Transport 491

near well-bore conditions, dual porosity and permeability systems, chemical

reactions, and complex reservoir geologies. The simulation capability of ECLIPSE
greatly exceeds the needs of most groundwater investigations. The utility of
ECLIPSE is greatly enhanced when combined with the Petrel* workflow software,
which allows for the integration of geological, geophysical, and hydrogeological
data into reservoir or aquifer models (Sect. 15.2). The great capability of ECLIPSE
has applications for the simulation of projects involving geologically complex
aquifers as found in fractured, dual-porosity systems. ECLIPSE also has the
capability of simulating subsurface carbon sequestration systems.

20.6.7 Hydrus

Hydrus is a Microsoft Windows based finite-element for simulating two- and

three-dimensional movement of water, heat, and multiple solutes in variably sat-
urated media (Radcliff and Šimůnek 2010). The program may be used to simulate
water and solute-transport in unsaturated, partially saturated, or fully saturated
media. Hydrus is well suited for the simulation of vadose zone processes and can
also simulate seepage face boundaries. A Windows-based graphical user interface
is used for grid design and editing, parameter allocation (data inputs), model
execution, and visualization and analysis of results. Hydrus is distributed in five
different versions (levels) of different sophistication and price, which allows users
to acquire the segments of software that are most appropriate for their needs.

20.6.8 PHT3D

The PHT3D code (Prommer et al. 2003) is a general purpose reactive 3D multi-
component model that combines the widely established MODFLOW/MT3DMS
(Zheng and Wang 1999) with a batch-type geochemical model PHREEQC-2
(Parkhurst and Apello 1999) to compute a wide range of biogeochemical reactions.
The PHT3D code has the advantage of being based on widely used, public domain
models that are all widely tested and documented.

20.7 Simulation of Compaction and Land Subsidence

Land subsidence caused by aquifer depressurization resulting from groundwater

pumping is major problem in some regions (Chap. 15). Compaction can also alter
the hydraulic properties of aquifers and confining units. Water released during
compaction of confining and semiconfining units may be a significant source of
non-renewable groundwater.
492 20 Groundwater Flow and Solute-Transport Modeling

The release of water and land subsidence can be simulated using packages
developed for the MODFLOW code. The interbedded storage package (ISB1)
(Leake and Prudic 1991) simulates storage changes from both elastic and inelastic
compaction of compressible fine-grained beds that are within or adjacent to an
aquifer (i.e., interbeds). Elastic and inelastic compaction is assumed to be pro-
portional to the change in head. The ISB1 package assumes instantaneous equi-
librium of the heads in the interbeds with heads in the surrounding aquifers. This
assumption is valid for thin interbeds with very short time constants (Hoffmann
et al. 2003a).
The Subsidence and Aquifer-System Compaction (SUB) Package simulates
elastic (recoverable) compaction and expansion, and inelastic (permanent) com-
paction of compressible fine-grained beds (interbeds) within the aquifers (Hoff-
mann et al. 2003b). The SUB package differs from the ISB1 package in that it can
be used to simulate the delayed release of water from storage or uptake of water
into storage in the interbeds. The SUB Package supersedes the Interbed Storage
Package (IBS1) for MODFLOW, which assumes that water is released from or
taken into storage with changes in head in the aquifer within a single model time
step. If the time-dependent drainage is turned off, the SUB Package gives results
identical to those from IBS1.
The Subsidence and Aquifer-System Compaction package for Water-Table
Aquifers (SWT) (unconfined aquifers) simulates vertical compaction in models of
regional groundwater flow. The program simulates groundwater storage changes
and compaction in discontinuous interbeds or in extensive confining units,
accounting for stress-dependent changes in storage properties (Leake and Gallo-
way 2007). Geostatic stress can be treated as a function of the water-table altitude,
and compaction is a function of computed changes in effective stress at the bottom
of a model layer. Thickness of compressible sediments in an unconfined model
layer can vary in proportion to saturated thickness.
An important modeling issue is the treatment of the delayed dissipation of
overpressures in thick interbeds. If short-term responses of hydraulic heads to
changes in pumping rate are to be simulated, then the presence of delayed drainage
in interbeds must be accounted for in the model (Hoffmann et al. 2003a). More
water will initially be required to be derived from storage in the aquifer to supply
the pumping wells, which would result in larger drawdowns than would occur
without delayed drainage from the interbeds. Delayed drainage effects will have a
small impact on long-term model runs unless the interbeds are of great thickness.

20.8 Integrated Surface Water-Groundwater Modeling

An objective in water resources management is to develop more holistic models

that incorporate the interaction of surface water and groundwater. Major modeling
goals are (1) to more accurately simulate aquifer recharge from climatological data
(particularly precipitation and temperature), (2) to simulate the effects of
20.8 Integrated Surface Water-Groundwater Modeling 493

groundwater withdrawals on surface water environments and wetlands, and (3) to

predict the effects of land and water use changes on surface water flow and
groundwater levels. Incorporation of recharge into groundwater flow models has
long been a challenge. Often, recharge is treated as the residual in the water
balance, i.e., the model is used to determine recharge through the calibration
process, rather than recharge being a primary input variable.
The linkage between surface water and groundwater with respect to downward
flow (recharge) can be addressed in two main manners. The historically more
common method is to not implicitly deal directly with the vadose zone and surface
waters in the model, but rather just apply a net recharge as a boundary condition to
the upper layer of the model (i.e., water table). An average recharge rate can be
estimated using a variety of methods (Chap. 11). Recharge from channel trans-
mission losses can be estimated from rainfall-runoff modeling.
A technically more sophisticated approach is integrated surface water—
groundwater modeling, which combines a surface water flow model with a
groundwater flow model. Two general approaches have been taken to develop
integrated models that simulate surface-water and ground-water interactions
(Markstrom et al. 2008). Fully integrated approaches simultaneously solve surface
and subsurface flow equations. The alternative is the ‘‘coupled regions’’ approach,
which partitions the surface and subsurface systems into separate regions with the
governing equations that describe flow in each region integrated (coupled) using
iterative methods (Markstrom et al. 2008). For example, separate equations can be
used to simulate runoff and infiltration, horizontal and vertical flow in the soil zone,
vertical gravity driven flow in the vadose zone, and three-dimensional flow in the
saturated zone (Markstrom 2008). For water resources investigations of arid envi-
ronments, the surface water model could provide recharge rates (e.g., from trans-
mission losses in wadis), which would be input into the groundwater flow model.
Both fully integrated and coupled models are capable of simulating coupled
surface-water and ground-water systems. There is no inherent overall advantage of
one method over the other. The choice of model codes should depend on project-
specific needs, computation efficiency (run times), availability of technical
expertise, cost, and whether or not the code is open source. As a general rule,
widely used open-source codes are preferred because of their transparency and
they benefit from testing and improvements (upgrades) developed by a large pool
of users. However, proprietary codes may have specific features needed for a given
project, and for this reason, may be the best option.
The U.S. Department of Agriculture SWAT code has been integrated with
MODFLOW (Sophocleous et al. 1999; Sophocleous and Perkins 2000; Kim et al.
2008; Chung et al. 2010). Conceptually, recharge rate calculated using SWAT is
input into MODFLOW and discharge from MODFLOW is input into SWAT.
Recharge elements input from SWAT to MODFLOW are stream transmission
losses, percolation beyond the root zone, and pond seepage. Subroutines were
developed to represent spatial heterogeneity within subbasins in SWAT and to
associate the geographic domain in SWAT with the MODFLOW finite-difference
grid.
494 20 Groundwater Flow and Solute-Transport Modeling

The GSFLOW code developed by the U.S. Geological Survey (Markstrom et al.
2008) is a good general example of a coupled regions integrated surface-water and
groundwater model. GSFLOW combines the USGS Precipitation-Runoff Model-
ing System (PRMS) surface water model (Leavesley et al. 1983, 2005) with the
2005 version of the USGS three-dimensional modular groundwater flow model
MODFLOW (Harbaugh 2005). GSFLOW simulates flow within and between three
regions; the top of the canopy to the base of the soil zone (PRMS), streams and
lakes (PRMS), and subsurface flow below the soil zone (MODFLOW).
The MIKE SHE/MIKE 11 codes, developed by DHI, is an example of a pro-
prietary fully-integrated surface-water and groundwater modeling system. It
includes a full suite of pre- and post-processing tools, plus a flexible mix of
advanced and simple solution techniques for each of the hydrologic processes
(DHI Software 2007). MIKE SHE/MIKE 11 allows for the simulation of the major
processes in the hydrologic cycle and includes process models for evapotranspi-
ration, overland flow, unsaturated flow, groundwater flow, and channel flow and
their interactions. Each of these processes can be represented at different levels of
spatial distribution and complexity.
Although modeling systems are now available to simulate integrated surface-
water and groundwater systems, a pertinent question is the value of such modeling
for actual water resources management in arid and semiarid lands, and in other
areas. There are an increasing number of ever more sophisticated tools for eval-
uating water resources. The key issue is determining whether the benefits they
provide in terms of improved water management is commensurate with the cost of
their deployment.
The fundamental challenge and limitation associated with integrated surface-
water and groundwater modeling is the huge amount of data that are required,
which is typically greatly beyond that actually available with any certainty for
study areas. For example, the large uncertainty inherent in parameters such as
actual evapotranspiration rates, stream bed conductance, and unsaturated soil
hydraulic conductivities in general, are incorporated into models. Soil zones often
have a high degree of heterogeneity with respect to their properties, which is a
particular difficult problem for the simulation of complex areas, such a mountain
front systems. Climatological data may also be sparse. The large numbers of
parameters (i.e., degrees of freedom) compounds the non-uniqueness associated
with model simulations in general. That an integrated surface-water and ground-
water model is well calibrated to limited observation data (e.g., a downstream
gauge) does not imply that it is an accurate tool for predicting future runoff and
aquifer recharge. The issue of convergence in an integrated surface-water/
groundwater model is critical because of the time difference between surface-water
flows and corresponding groundwater flow rates. If the model fails to converge, the
predictions can be very inaccurate, particularly for surface-water discharge rates
and stages.
Distributed parameter models, in which values have to be assigned for
numerous variables contributing to the surface water budget, are particularly
problematic because of the invariably huge data deficiency in project areas relative
20.8 Integrated Surface Water-Groundwater Modeling 495

to the specific data requirements. The common practice of using general or

regional default values for parameters may introduce large potential errors if these
values are not in fact locally appropriate. This is a very important issue in that
there is little advantage in using a distributed parameter model with great data
requirements if most of the values used are, at best, educated guesses. Use of
lumped-parameter models is preferred in areas with large data deficiencies. For
example, it may be more accurate to simulate wadi recharge in a groundwater
model using an empirical rainfall-runoff-recharge relationship function, than by
creating a detailed integrated surface-water/groundwater model for which the
distribution of the values of most of the parameters are very poorly constrained.
Integrated surface-water/groundwater models are therefore most appropriately
used in studies of semi-arid or arid lands for which an exceptional quantity of
hydrologic data are available. It is more appropriate to use a less complex model
when first evaluating a given area that has a poor data base. This initial model will
allow the identification of data deficiencies and could provide some initial pre-
dictions based on the error ranges caused by the paucity of data. As additional data
are obtained, the modeling should be upgraded to an integrated model that may
then provide a more comprehensive and accurate set of simulations. A key con-
clusion is that the sophistication of the model code used for water resources
management in a given area should evolve with the ability to input the appropriate
quantity and quality of hydrologic data.

References

Anderson, M. P., & Woessner, M. W. (1992). Applied groundwater modeling, simulation of flow
and advective transport. San Diego: Academic Press.
Bredehoeft, J. (2005). The conceptualization model problem—surprise? Hydrogeology Journal,
13, 37–46.
Chung, H.-M., Kim, N.-W., Lee, J., & Sophocleous, M. (2010). Assessing distributed groundwater
recharge using integrated surface water-groundwater modeling. Hydrogeology Journal, 18,
1253–1264.
DHI Software (2007). MIKE SHE uses manual (Vol. 1). User Guide.
Diersch H. J. (1998). FEFLOW: Interactive, graphics-based finite-element simulation system for
modeling groundwater flow, contaminant mass and heat transport processes. In Getting
started; user’s manual; reference manual, version 4.6. Berlin, Germany; WASY, Institute for
Water Resources Planning and System Research Ltd.
Doherty, J. (2005). PEST: model-independent parameter estimation (5th ed.). Brisbane,
Australia: Watermark Numerical Computing.
Domenico, P. A., & Schwartz, F. W. (1998). Physical and chemical hydrogeology (2nd ed.). New
York: Wiley.
Fetter, C. W. (2001). Applied hydrogeology (4th ed.). New Jersey: Englewood Cliffs Prentice Hall.
Guo, W., & Langevin, C. D. (2002). User’s guide to SEAWAT: A computer program for
simulation of three-dimensional variable-density ground-water flow. U.S. Geological Survey
Open-File Report, 01-434.
Harbaugh, A. W. (2005). MODFLOW-2005, the U.S. Geological survey modular ground-water
model–the ground-water flow process. U.S. Geological Survey Techniques and Methods, 6-A16
496 20 Groundwater Flow and Solute-Transport Modeling

Hemker, C. J., & de Boer, R. G. (2009). MicroFem (Version 4.10.02).

Herczeg, A. L., & Leaney, F. W. (2011). Review: Environmental tracers in arid-zone hydrology.
Hydrogeology Journal, 19(1), 17–30.
Herrmann, R. (2006). ASR well field optimization un unconfined aquifers in the Middle East. In
Recharge systems for protecting and enhancing groundwater resources, proceedings of the
5th international symposium on management of aquifer recharge (pp. 109–114). Berlin,
Germany, Paris: UNESCO. 11–16 June 2005.
Hill, M. C., & Tiedeman, C. R. (2007). Effective groundwater model calibration: With analysis of
data, sensitivities, predictions, and uncertainty. Hoboken, NJ: Wiley-Interscience.
Hoffmann, J., Galloway, D. L., & Zebker, H. A. (2003a). Inverse modeling of interbed storage
parameters using land subsidence observations. Antelope Valley, California: Water Resources
Research, 39(2), 1031. doi:10.1029/2001WR001252,2003.
Hoffmann, J., Leake, S. A., Galloway, D. L., & Wilson, A. M. (2003b). MODFLOW-2000
ground-water model–user guide to the subsidence and aquifer-system compaction (SUB)
package. U.S. Geological Survey Open-File Report 03-233.
Hunt, R. J., Doherty, J., & Tonkin, M. J. (2007). Are models too simple? Arguments for Increased
Parameterization: Ground Water, 45, 254–262.
Kim, N. W., Chung, I. M., Won, Y. S., & Arnold, J. G. (2008). Development and application of
the integrated SWAT-MODFLOW model. Journal of Hydrology, 356, 1–16.
Kipp, K. L., Jr. (1987). HST: A computer code for simulation of heat and solute transport in
three-dimensional ground-water flow systems. U.S. Geological Survey Water-Resources
Investigations Report 86-4095.
Kipp, K. L., Jr. (1997). Guide to the revised heat and solute transport simulator; HST3D (Version
2). U.S. Geological Survey Water-Resources Investigations Report 97-4157.
Langevin, C. D., Thorne, D. T., Jr., Dausman, A. M., Sukop, M. C., & Weixing, G. (2008).
SEAWAT version 4: a computer program for simulation of multi-species solute and heat
transport. U.S. Geological Survey Techniques and Methods Book 6, Chapter A22.
Leake, S. A., & Galloway, D. L. (2007). MODFLOW ground-water model—user guide to the
subsidence and aquifer-system compaction package (sub-wt) for water-table aquifers. U.S.
Geological Survey, Techniques and Methods 6-A23.
Leake, S. A., & Prudic, D. E. (1991). Documentation of a computer program to simulate aquifer-
system compaction using the modular finite-difference ground-water flow model. U.S.
Geological Survey Techniques of Water—Resources Investigations, book 6, chapter. A2.
Leavesley, G. H., Lichty, R. W., Troutman, B. M., & Saindon, L. G. (1983). Precipitation-runoff
modeling system: user’s manual. U.S. Geological Survey Water-Resources Investigations
Report 83-4238.
Leavesley, G. H., Markstrom, S. L., Viger, R. J., & Hay, L. E. (2005). USGS modular modeling
system (MMS)—precipitation-runoff modeling system (PRMS) NMS-PRMS. In V. Singh &
D. Frevert (Eds.), Watershed models (pp. 159–177). Boca Raton, FL: CRC Press.
Lloyd, J. W. (2004). Do we adequately understand the potential value of our regional aquifers?
Proceedings, International Conference on Water Resources in Arid Environments.
Lloyd, J. W. (2007). The difficulties of regional groundwater resources assessments in arid
regions. Arabian Journal for Science and Engineering, 32(1C), 35–47.
Maliva, R. G., & Missimer, T. M. (2010). Aquifer storage and recovery and managed aquifer
recharge using wells: Planning, hydrogeology, design, and operation. Houston, Schlumberger
Water Services, Methods in Water Resources Evaluation Series.
Markstrom, S. L., Niswonger, R. G., Regan, R. S., Prudic, D. E., & Barlow, P. M. (2008).
GSFLOW—couple ground-water and surface-water flow model based in the integration of the
Precipitation-Runoff Modeling System (PRMS) and the Modular Ground-Water Flow Model
(MODFLOW-2005). U.S. Geological Survey Techniques and Methods 6-D1.
McDonald, M. G., & Harbaugh, A. W. (1988). A modular three-dimensional finite-difference
ground water flow model. U.S. Geological Survey Techniques of Water-Resources
Investigation Report 06-A1.
References 497

Missimer, T. M., Maliva, R. G., & Guo, W. (2010). Sustainability and the management of water
resources in Florida. In F. Bloetcher (Ed.), Sustainable water resources: A compendium of
issues and trends (pp. 153–161). Denver: American Water Works Association.
National Research Council (2008). Prospects for managed underground storage of recoverable
water. Washington, DC: National Academy Press.
Oreskes, N., Shrader-Frechette, K., & Belitz, M. (1994). Verification, validation, and
confirmation of numerical models in the Earth Sciences. Science, 263, 641–646.
Parkhurst, D. L., & Appelo, C. A. J. (1999). PHREEQC (Version 2)—A computer program for
speciation, batch reaction, one-dimensional transport, and inverse geochemical calculations.
U.S. Geological Survey, Water-Resources Investigations Report 99-42549.
Poeter, E. (2007). ‘‘All models are wrong, how do we know which are useful?’’—looking back at
the Darcy lecture tour. Ground Water, 45(4), 390–391.
Poeter, E., & Hill, M. C. (1997). Inverse models: A necessary next step in ground-water
modeling. Ground Water, 35(2), 250–260.
Pollock, D. W. (1994). User’s guide for MODPATH/MODPATH-PLOT Version 2: A particle
tracking post-processing package for MODFLOW, the U.S. Geological Survey finite-
difference ground-water flow model. U.S. Geological Survey Open-File Report 94-464.
Prommer, H., Barry, D. A., & Zheng, C. (2003). MODFLOW/MT3DMS based reactive multi-
component transport modeling. Ground Water, 41, 347–257.
Radcliff, D., & Šimůnek, J. (2010). Soil physics with HYDRUS: Modeling and applications. Boca
Raton, FL: CRC Press.
Sanford, W. (2011). Calibration of models using groundwater age. Hydrogeology Journal, 19,
13–16.
Sophocleous, M. A., Koelliker, J. K., Govindaraju, R. S., Birdie, T., Ramireddygari, S. R., &
Perkins, S. P. (1999). Integrated numerical modeling for basin-wide water management:
The case of the Rattlesnake Creek basin in south-central Kansas. Journal of Hydrology, 214,
179–196.
Sophocleous, M. A., & Perkins, S. P. (2000). Methodology and application of combined
watershed and ground-water models in Kansas. Journal of Hydrology, 236, 185–201.
Voss, C. I., & Provost, A. M. (2002). SUTRA, A model for saturated-unsaturated variable-density
ground-water flow with solute or energy transport. U.S. Geological Survey Water-Resources
Investigations Report 02-4231.
Zheng, C., & Bennett, G. D. (2002). Applied contaminant transport modeling (2nd ed.). New York:
Wiley.
Zheng, C., & Wang, P. P. (1999). MT3DMS: A modular three-dimensional multi-species model
for simulation of advection, dispersion and chemical reactions of contaminants in ground
water systems: Documentation and user’s guide. U.S. Army Engineer Research and
Development Center, Vicksburg, Mississippi, report SERDP-99-1.
Zhu, C., & Murphy, W. M. (2000). On radioactive dating of ground water. Ground Water, 38,
802–804.