Issue Paper/

Using Hydraulic Head Measurements in

Variable-Density Ground Water Flow Analyses
by Vincent Post1, Henk Kooi2, and Craig Simmons3

Abstract
The use of hydraulic head measurements in ground water of variable density is considerably more compli-
cated than for the case of constant-density ground water. A theoretical framework for dealing with these com-
plications does exist in the current literature but suffers from a lack of awareness among many hydrogeologists.
When corrections for density variations are ignored or not properly taken into account, misinterpretation of both
ground water flow direction and magnitude may result. This paper summarizes the existing theoretical framework
and provides practical guidelines for the interpretation of head measurements in variable-density ground water
systems. It will be argued that, provided that the proper corrections are taken into account, fresh water heads can
be used to analyze both horizontal and vertical flow components. To avoid potential confusion, it is recommended
that the use of the so-called environmental water head, which was initially introduced to facilitate the analysis of
vertical ground water flow, be abandoned in favor of properly computed fresh water head analyses. The presented
methodology provides a framework for determining quantitatively when variable-density effects on ground water
flow need to be taken into account or can be justifiably neglected. Therefore, we recommend that it should
become part of all hydrogeologic analyses in which density effects are suspected to play a role.

Introduction nature of Darcy’s law. There are uncertainties in flow esti-

Using hydraulic head observations to infer ground mates, which arise from insufficient knowledge of
water flow directions and flow rates is a basic skill of hydraulic conductivity and heterogeneity, complications
every hydrogeologist. It is an application of Darcy’s law: due to anisotropy, or large well spacing or screen length,
all that is required are estimates of hydraulic conductiv- but otherwise, it is rather straightforward.
ity, K, and of the hydraulic gradient, rh, or components What is less well known is the fact that the classical
thereof. The practicality and convenience of the afore- form of Darcy’s law, cast in terms of hydraulic head, and
mentioned field method obviously stems from the simple hence, the intuitive field method, does not apply to
ground water of variable density. Density variations can
result from differences in temperature or pressure but
1Corresponding author: Department of Hydrology and more often are caused by differences in solute concentra-
Geo-Environmental Science, Faculty of Earth and Life Sciences, tion. Variable density is particularly relevant in coastal
Vrije Universiteit, De Boelelaan 1085, 1081 HV Amsterdam, areas, in sedimentary basins, and where dense contami-
The Netherlands; 131 (0) 20 598 7402; fax 131 (0) 20 598 9940; nant plumes are present. The theory of ground water flow
[email protected]
2Department of Hydrology and Geo-Environmental Science, in variable density systems is considerably more compli-
Faculty of Earth and Life Sciences, Vrije Universiteit, Amsterdam, cated than under density invariant conditions, but there
The Netherlands. are still practical methods for dealing with it in combina-
3School of Chemistry, Physics and Earth Sciences, Flinders
tion with field data.
University, Adelaide, South Australia, Australia. No simple guidelines currently exist that allow hy-
Received October 2006, accepted March 2007.
Copyright ª 2007 The Author(s) drogeologists to easily and robustly determine a priori
Journal compilation ª 2007 National Ground Water Association. whether a hydrogeologic analysis should be treated in
doi: 10.1111/j.1745-6584.2007.00339.x a density-dependent or density-independent manner. The
664 Vol. 45, No. 6—GROUND WATER—November-December 2007 (pages 664–671)
analyses presented in this paper should be conducted variable-density systems, we intentionally refrain from
where this uncertainty arises and in order to examine the presenting an analysis of the more complicated effects of
potential consequences of ignoring density-dependent anisotropy, heterogeneity, and dipping aquifers. Readers
flow in hydrogeologic analyses. are referred to the works of Bachu (1995) and Bachu and
Despite its great importance for hydrogeologists, there Michael (2002) for a detailed discussion of these topics.
are few, if any, textbooks that explain how to correctly
use field data to assess ground water flow under variable-
density conditions. A theoretical treatment was presented Fundamentals of Variable-Density Flow
by Lusczynski (1961), and although it was reiterated
in some other publications (DeWiest 1967; Bear 1972; Darcy’s Law
Oberlander 1989; St. Germain 2001), the proposed meth- The well-known short-hand notation of the differen-
odology did not become part of the mainstream literature. tial equation form of Darcy’s law is as follows:
Excellent treatises, including examples, were presented by
Van der Eem (1992) and Juster (1995), but these appear q ¼ 2 Krh
~ ð1Þ
not to have reached a wide international audience.
Van Dam (1977) noted that ‘‘. there is much un- In terms of physics, this equation relates three quan-
acquaintedness and misunderstanding about the theory to tities. ~
q denotes specific discharge (volume of fluid per
be applied .’’ and Custodio (1987) stated that the vari- unit cross-sectional area of porous medium per unit time,
ous concepts of water heads of variable density ‘‘. do m3/m2/s), also referred to as the Darcy velocity. rh is the
not solve the problem in a clear way.’’ In a recent article, driving force of ground water flow per unit weight of
Simmons (2005) cautioned about the potential abuse of ground water (dimensionless). K is the hydraulic conduc-
the fresh water head, noting, ‘‘Possibly one of the simplest tivity, a proportionality coefficient that describes the ease
analysis approaches used in variable density flow is the by which fluid flows through a porous medium per unit
concept of ‘equivalent freshwater head’ but this is often flow rate (m/s). As stated, we neglect directional depen-
too simple or even erroneous, especially where vertical dency or anisotropy of the latter quantity in this manu-
flow is of interest.’’ According to experience of the script and, therefore, treat K as a scalar quantity, which
authors, confusion and misconceptions about the proper can be expressed by a singular numerical value at each
ways of dealing with density variations in flow calcu- point in the porous medium. Under these conditions, the
lations still abound, which forms the prime motivation for three flow components are the following:
the present paper. The most common misconception @h
observed by the authors (even in research papers and text- qx ¼ 2 K ð1aÞ
@x
books) is the notion that converting measured heads to
fresh water heads suffices to analyze flow patterns and
rates in variable-density ground water systems. This erro- @h
qy ¼ 2 K ð1bÞ
neous approach is not only caused by a lack of attention @y
paid to variable-density flow calculations in textbooks
but is probably also linked to the fact that several main-
@h
stream numerical codes that simulate variable-density qz ¼ 2 K ð1cÞ
flow solve ground water equations written in terms of @z
fresh water head. Equation 1 is a simplified form of the more general
The theoretical framework as outlined in this paper physical law for fluid flow in a porous medium, which
is based on the existing literature. The objective here is to also applies to variable-density fluids (Bear 1972):
increase awareness among professionals in the field and
thus avoid misinterpretation of head data in variable- k
q¼ 2
~ ðrP 2 q~
gÞ ð2Þ
density settings. The paper reinforces the appropriate l
methodology that should be applied in variable-density
ground water analyses and discusses common problems with components:
that may be encountered along the way. Moreover, it will
k @P
be argued that the use of the so-called environmental qx ¼ 2 ð2aÞ
water head, as initially introduced by Lusczynski (1961), l @x
is best avoided. We develop ‘‘four golden rules’’ that we
believe will provide useful and much needed guidance in k @P
the proper application of these concepts. qy ¼ 2 ð2bÞ
l @y
Emphasis is placed on situations where density is
influenced by solute concentration since they are encoun-


tered by the vast majority of hydrogeologists. However, k @P
examples and procedures are readily transferable to con- qz ¼ 2 1 qg ð2cÞ
l @z
ditions where temperature or pressure is the prime control
on density variability by using the appropriate density where k is intrinsic permeability (m2), a property of the
contrasts. Since the aim here is to present the simplest set porous medium; l is dynamic viscosity (kg/m/s) of the
of practical rules for dealing with head measurements in ground water; P is fluid pressure (kg/m/s2); q (kg/m3) is
V. Post et al. GROUND WATER 45, no. 6: 664–671 665
fluid density; and ~ g is the gravitational acceleration point i. For stagnant water conditions in the well, hp,i is
(m/s2). Equation 2 explicitly shows the two basic driving related to the pressure of the ground water at the well
forces for ground water flow: 2rP is the force per unit screen Pi by the following:
volume of ground water due to spatial differences in pore
water pressure, and q~ g is the gravity force per unit Pi
hp;i ¼ ð4Þ
volume experienced by the ground water. Both forces, qi g
naturally, also act on ground water of uniform density.
where qi (kg/m3) is the density of the water in the piezo-
However, in Equation 1, these two forces are lumped into
meter tube, i.e., of the ground water surrounding the well
the single, convenient expression of gradient of hydraulic
screen. It follows that, in a system where q varies spa-
head, where the individual driving forces are rendered
tially, values of hp,i do not correctly represent spatial var-
invisible. For ground water of variable density such
iations of P. In other words, the same pressure can
a ‘‘gradient form’’ does not exist. This fundamental dis-
correspond to different values of hp,i, depending on
tinction is the main reason why quantification of ground
ground water density. Lusczynski (1961), therefore, used
water flow from field data, which normally occurs in the
the term point water head for hi to indicate that the values
form of head measurements, requires a special treatment.
are uniquely linked to the ambient ground water density
at the well screen.
Head and Pressure Formulation
To eliminate the ambiguity between hp,i and Pi, hp,i
Equation 2 shows that for variable-density flow calcu-
can be normalized using a reference density. That is, the
lations, ground water pressure P and density q should be
water column in each observation well is replaced by an
known, rather than hydraulic head h. Pressure, however, is
(imaginary) equivalent column of water of equal density
not often used in everyday life, and hydrogeologists are
for all the wells (Figure 1b). Any value of q can be used
more familiar with the concept of head. A number of key
for this purpose (Van der Eem 1992), but fresh water is
relationships among these quantities are therefore sum-
used most often, which gives rise to the definition of fresh
marized here, which will subsequently be exploited to cast
water head:
Equations 2a through 2c in terms of head.
Hydraulic head h (m) is obtained by measuring the Pi
level of the water-air interface in a ground water observa- hf;i ¼ zi 1 ð5Þ
qf g
tion well, where levels refer to a common datum, often
mean sea level. Two contributions to h are distinguished, where qf is fresh water density. Fresh water head can be
and indicated in Figure 1a: readily calculated from point water head measurements
using the following:
hi ¼ zi 1 hp;i ð3Þ
qi q 2 qf
where zi (elevation head) represents the (mean) level of hf;i ¼ hi 2 i zi ð6Þ
qf qf
the well screen, and hp,i (pressure head) is the length of
the water column in the well relative to zi. The subscript i Fresh water head is always larger than or equal to
is added to indicate that these values are measured at point water head.

Figure 1. Schematic representation of head definitions in variable-density ground water systems (modified from Lusczynski
1961). Lightest shading corresponds to fresh water and darker shading represents increasing salinity.

666 V. Post et al. GROUND WATER 45, no. 6: 664–671

 From Equations 2a and 2b follows that horizontal That is, instead of point water or fresh water, the well is
flow components (qx and qy) should be calculated from thought to be filled with ‘‘environmental’’ water. With this
the corresponding horizontal components of the pressure assumption and in the absence of vertical ground water
gradient. Alternatively, the horizontal component of the flow, the water level in the well will coincide with the
head gradient can be used, provided the heads refer to the water table since the water pressure is hydrostatic both
same density. Rearranging and differentiating Equation 5 inside and outside the well. If there is vertical flow, the
with respect to x and y and inserting the result into Equa- water pressure outside the well will differ from the hydro-
tions 2a and 2b gives the following: static pressure. In Lusczynski’s (1961) definition of the
environmental water head, this difference is expressed as
kqf g lf @hf @hf a column of fresh water, which is a measure for the devia-
qx ¼ 2 ¼ 2 Kf ð7aÞ
lf l @x @x tion of he,i from the water table. The concept is ingenious
but unfortunately becomes nonintuitive when high-
density water is present all the way up to the water table.
kqf g lf @hf @hf DeWiest (1967) introduced the ‘‘true environmental
qy ¼ 2 ¼ 2 Kf ð7bÞ
lf l @y @y head’’ in which environmental water head is related to
pressure according to the following:
Kf is the fresh water hydraulic conductivity. It is assumed
here that salinity variations have a negligible effect on l pi
he;i ¼ zi 1 ð9Þ
so that lf/l ’ 1 in Equations 7a and 7b, which is a very qe g
good approximation for most practical applications.
Moreover, the difference between Kf and field-measured where qe is the average density of the water between zi
values of hydraulic conductivity, which are for ambient and he,i inside the well. This definition is not very practi-
values of l and q, is much smaller than the uncertainty cal, however, since he,i and qe are interdependent (Juster
associated with this parameter. Hence, no special correc- 1995). Moreover, qe is easily confused with the average
tions to existing hydraulic conductivity information are density of the water outside the well qa (defined later on
normally required. in this paper) used by Lusczynski (1961) in his original
Equation 2c shows that evaluation of the vertical definition of environmental water head.
flow component is different from the horizontal compo-
nents in that a term involving local ground water density
is needed. Similar to the horizontal flow components, the Application and Interpretation Procedure
vertical component can also be cast in terms of fresh In this section, the procedure for the interpretation
water head by rearranging and differentiating Equation 5 of head measurements in variable-density ground water will
and inserting the result into Equation 2c: be outlined. These will be illustrated with examples and the
implications of the necessary assumptions will be discussed.
" !#
kqf g lf @hf q 2 qf
qz ¼ 2 1 Horizontal Flow Component
lf l @z qf
" !# When calculating horizontal flow, it is crucially
@hf q 2 qf important that the fresh water head gradient in Equations
¼ 2 Kf 1 ð7cÞ
@z qf 7a and 7b (or pressure gradient in Equations 2a and 2b) is
evaluated using fresh water heads at the same depth
q 2 qf because, in contrast to uniform density ground water,
in which the term , which represents the relative
qf fresh water head may vary with depth, even for hydro-
density contrast, accounts for the buoyancy effect on
static (i.e., no vertical flow) conditions. Thus, when meas-
the vertical flow. Equation 7c is used in several well-
urements are taken from piezometers with screens at
known variable-density flow and transport codes (e.g.,
different depths, fresh water heads need to be calculated
MOCDENSE, SEAWAT).
at a suitable reference depth. A common approach is to
Lusczynski (1961) introduced the concept of envi-
assume hydrostatic conditions between the well screen
ronmental water head (he,i) in order to calculate vertical
and the reference depth. The pressure at the reference
flow with the convenient and familiar classical form of
depth (zr) then becomes as follows:
Darcy’s law:
Z zr
kqf g lf @he;i @he;i P r ¼ Pi 2 g qdz ¼ Pi 2 qa gðzr 2 zi Þ ð10Þ
qz ¼ 2 ¼ 2 Kf ð8Þ zi
lf l @z @z
with
The buoyancy effect on the vertical flow is taken
Z zr
into account in the definition of the environmental water 1
head. In an appendix to his paper, Lusczynski (1961) qa ¼ qdz ð11Þ
zr 2 zi zi
demonstrated the validity of this approach. Figure 1c illus-
trates that environmental water head is obtained when the qa denotes the average water density between meas-
observation well is filled with stagnant water in which the urement point zi and the reference level zr. The corre-
variations of density are identical to those encountered sponding fresh water head at zr (hf,r) is then obtained
along the vertical in the ground water just outside the well. from Equation 5:
V. Post et al. GROUND WATER 45, no. 6: 664–671 667
 uncertainty can be demonstrated for the data of example
Pr q q
hf;r ¼ zr 1 ¼ zr 1 i ðhi 2 zi Þ 2 a ðzr 2 zi Þ ð12Þ 1. Above, the average of the densities of the water at the
qf g qf qf two screens was used for qa, in order to obtain a fresh
water head value for the 50-m-deep well at zr ¼ 240 m.
Using field data, the horizontal component of flow is
This average density can be thought to correspond to a
obtained from the following:
linear vertical density profile in the depth range between

hf the two screens as shown in Figure 2. Evidently, other
qx ¼ 2 K f ð13Þ density profiles and corresponding values of qa are possi-

x
ble. In the absence of additional constraints, two ‘‘end-
and analogously for qy. member’’ density profiles can be constructed, indicated
by ‘‘min’’ and ‘‘max’’ in Figure 2, where values of qa
Example 1. Horizontal Flow correspond to the density values listed for both piez-
Consider two piezometers some distance
x apart ometers (Figure 2). Minimum, maximum, and mean val-
that have their well screens located in the same aquifer ues for hf,r are listed in Table 2. Results show that the
(Figure 2). Screen depths and the measured point water head difference between the two piezometers varies with
heads and densities are listed in Table 1. Because the the assumed average density by about 40%, which im-
screen depths differ 10 m, fresh water heads have to be plies a similar uncertainty in the magnitude of the flow
calculated for a single reference depth using Equation 12. component. Strictly speaking, the uncertainty may still
Results for zr ¼ 240 m (the depth of piezometer 1) and have been underestimated with the adopted approach
qa ¼ 1005 kg/m3 (average of the densities at the two because the density at 240-m depth at piezometer 2 need
screens) are given in Table 2 under the heading ‘‘mean.’’ not be equal to the density at piezometer 1 if lateral varia-
Comparison of the values with the point water heads in tions in density between the two wells occur. Therefore,
Table 1 shows that the horizontal gradients and hence the alternative methods of uncertainty assessment are pos-
suggested flow directions for the two head types are sible. The present example does demonstrate, however,
opposite. Clearly, the gradient of hf,r should be combined the importance of conducting such an assessment to check
with a value for Kf to arrive at a proper estimate of hori- to what extent inferred flow conditions are significant.
zontal flow.
It is important to realize that several of the steps
outlined in example 1 involve assumptions and that each Vertical Flow Component
of these assumptions introduces uncertainty in the final The second term in large brackets in Equation 7c is
flow estimate. A potential source of uncertainty specific essential to correctly describe variable-density flow. For
for variable-density flow estimation is the required esti- example, under hydrostatic conditions (qz ¼ 0) in a saline
mate of average density qa between screen and reference aquifer of sea water concentration (q ¼ 1025 kg/m3), the
depth. A simple approach to (approximately) quantify this density excess ratio is (q 2 qf)/qf ¼ 0.025, and fresh

Figure 2. Left: Piezometers used in example problems. Darker shading represents increasing salinity. Note that in the exam-
ples, only the density at the well screens is known and not the true density distribution in the aquifer. Right: vertical density
distributions considered in the calculations of hf,r and he,i.

668 V. Post et al. GROUND WATER 45, no. 6: 664–671

 Table 1 Table 3
Well Screen Depth, Point Water Head, Density, and Fresh Water Head Gradient, Buoyancy Term,
Calculated Fresh Water Heads of Example 1 and Vertical Component of Specific Discharge
of Example 2 for qa ¼ 1004 (Minimum), qa ¼ 1005
Screen Depth hi q hf,i (Mean), and qa ¼ 1006 kg/m3 (Maximum)
Piezometer (m) (m) (kg/m3) (m)
Minimum Mean Maximum
1 240 1.25 1004 1.42
2 250 1.20 1006 1.51
hf
9 3 1023 9 3 1023 9 3 1023

z
qa 2 qf
qf 0.004 0.005 0.006
qz (m/d) 0.05 0.04 0.03
water head decreases with depth according to @hf/@z ¼
20.025. Thus, both terms cancel each other in Equation
13 to correctly describe the zero flow condition. Ignoring
the buoyancy term, however, would yield a markedly Example 3. Vertical Flow: Environmental Water Head
erroneous flow estimate. Formulation
Again, consider the same piezometers as in example
Example 2. Vertical Flow: Fresh Water Head Formulation 2. Application of Equation 15 assuming zr ¼ 0 m (to
For the calculation of vertical flow, consider the make sure reference depth is above the domain of ‘‘non-
same piezometers as in example 1. In this example, how- fresh water’’ for minimum, mean, and average density
ever, their well screens are not separated laterally but distributions) results in the values listed in Table 4. Val-
are in the same vertical (
x ¼ 0). To evaluate the ues of qz are calculated with the finite-difference form of
vertical flow component, Equation 7c must be cast in finite- Equation 8:
difference form:

he;i
" !# qz ¼ 2 Kf ð16Þ

hf q a 2 qf
z
qz ¼ 2 Kf 1 ð14Þ

z qf Calculated values of qz are the same as in example 2,
as they should be.
where
hf ¼ hf,2 2 hf,1 and
z ¼ z2 2 z1 are the differ- The expression for he,i in Equation 15 is identical to
ence in fresh water head and elevation head of the piez- that of hf,r in Equation 12. The values of hf,r listed in
ometers, respectively, and qa is the average density of the Table 2 indeed yield the vertical gradients in environmen-
ground water between the screens, defined analogously to tal water head listed in Table 4. Values of he,i and hf,r do
Equation 11. The calculated values of qz for a value of differ, however, because different reference levels were
Kf ¼ 10 m/d are listed in Table 3. As before, uncertainty used (0 and 240 m, respectively). A constant difference
arises from the unknown average density between the in head is of no consequence, however, when gradients
point measurements (Figure 2). are calculated.
For coastal settings where fresh overlies saline The equivalence of the environmental water head
ground water, Lusczynski (1961) inferred the following: approach and the fresh water head approach to calculate
qi q vertical flow can furthermore be demonstrated in the fol-
he;i ¼ zr 1 ðhi 2 zi Þ 2 a ðzr 2 zi Þ ð15Þ lowing way. Equation 15 is written in terms of fresh water
qf qf
head using Equations 5 and 10:
where zr denotes an arbitrary reference level above which
qa 2 qf
ground water is fresh and where qa is the average density he;i ¼ hf;r ¼ hf;i 2 ðzr 2 zi Þ ð17Þ
of water between zr and screen depth zi. The latter is cal- qf
culated with Equation 11.
At two well screens in the same vertical, this gives
the following:
Table 2 qa;1 2 qf
Fresh Water Heads of Piezometers at he;1 ¼ hf;1 2 ðzr 2 z1 Þ ð18aÞ
qf
Reference Depth zr ¼ 240 m of Example 1 for
qa ¼ 1004 (Minimum), qa ¼ 1005 (Mean), and
qa ¼ 1006 kg/m3 (Maximum) qa;2 2 qf
he;2 ¼ hf;2 2 ðzr 2 z2 Þ ð18bÞ
qf
hf,r
Note that the average densities are different for each
Minimum Mean Maximum screen because they are evaluated over different intervals
Piezometer zr (m) (m) (m) (m)
and that:
1 240 1.42 1.42 1.42 Z zr
2 240 1.47 1.46 1.45 qa;1 ðzr 2 z1 Þ ¼ qdz ð19aÞ
z1

V. Post et al. GROUND WATER 45, no. 6: 664–671 669

 Z zr
qa;2 ðzr 2 z2 Þ ¼ qdz ¼ qa;1 ðzr 2 z1 Þ 1 qa ðz1 2 z2 Þ Table 4
z2 Environmental Water Heads of Piezometers at
ð19bÞ zi ¼ 240 and zi ¼ 250 m, Environmental
Water Head Gradient and Vertical Component of
where qa is the average density between the two screen Specific Discharge of Example 2 for Different
depths. Taking the difference (
he,i ¼ he,2 2 he,1) yields Values of qa
the following:
Minimum Mean Maximum
q 2 qf he,240 1.42 1.34 1.26

he;i ¼ hf;2 2 hf;1 1 a ðz2 2 z1 Þ
qf he,250 1.47 1.38 1.29
q 2 qf
he
5 3 1023 4 3 1023 3 3 1023
¼
hf;i 1 a
z ð20Þ
z
qf qz (m/d) 0.05 0.04 0.03

Dividing by
z gives the term in large brackets in

Equation 14.
hf deviates up to several meters from the ‘‘measured’’
value of hi ¼ 0. Note that in flat coastal areas (e.g., deltas,
Discussion and Conclusions sedimentary basins), head differences that drive ground
water flow are typically on the order of decimeters.
The previous discussion demonstrates that very sub-
The line in Figure 3 represents the precision of care-
tle density variations can have a major impact on the
fully taken head measurements (0.02 m) and can be used
flow field and necessitate employment of the methodol-
to assess when density variations start to become impor-
ogy outlined earlier. A detailed description of the
tant. Slight deviations from fresh water densities may
full density field in the area under investigation is costly
already lead to corrections exceeding the precision of the
and difficult to obtain and will usually not be available.
head measurements at depths of several tens of meters or
At minimum though, all head measurements should be
more. Assessments of this type should always be the first
carried out in conjunction with measurements of elec-
step in any hydrogeologic study to determine if variable-
trical conductivity. The density can then be estimated
density effects need to be taken into account. If these
from simple relationships between density and salinity
effects cannot be ruled out a priori, they should be quan-
available in the literature (Reilly and Goodman 1985;
tified by means of the analyses presented in this paper. In
Holzbecher 1998).
order to justify the choice of a conceptual hydrogeologic
So far, it has not been assessed at what density con-
framework in an explicit rather than implicit manner,
trasts the aforementioned corrections become significant
such checks should be applied more routinely than they
and need to be taken into account. To this aim, the value
appear to be at present.
of hf,i (Equation 6) for hi ¼ 0 is contoured in Figure 3 as
It was already discussed previously that the required
a function of qi and zi. It provides a measure of the degree
estimate of the average density qa between screen and
of misinterpretation of fluid pressure at the well screen
reference depth is a significant source of uncertainty in
when solely relying on point water heads for typical con-
the flow calculations. Uncertainties obviously tend to in-
ditions in coastal aquifers. As can be seen from the figure,
crease considerably for larger vertical distances between
screen depth and reference level. As a general rule, refer-
ence depth should, therefore, be chosen within the depth
range of the employed well screens. Contour maps of
fresh water head in extensive aquifers, even when referred
to the same vertical level, should therefore be regarded
with suspicion and subsequently used with caution. The
requisite of using a single horizontal reference level fur-
ther implies that isohyps maps for tilted aquifers cannot
be constructed over large distances in the dip direction.
Other potential sources of uncertainty in these analy-
ses are (1) estimates of hydraulic conductivity, Kf, which
is beyond the scope of the present paper because it is
common also to uniform density ground water flow; (2)
the assumption of hydrostatic conditions between screen
and reference depth; and (3) finite-length screens. Addi-
tional complexity (and uncertainty) will occur in systems
where there is substantial heterogeneity, anisotropy, and
Figure 3. Contour plot of the value of hf,i (in m) according complex geometries associated with, for example, sloping
to Equation 6 for hi ¼ 0 m as a function of qi and zi. The aquifer configurations.
dashed line represents the minimum error associated with
head measurements (0.02 m). The third source of uncertainty noted previously,
finite length of well screens, stems not only from
670 V. Post et al. GROUND WATER 45, no. 6: 664–671
ambiguity in assigning a single value depth to the screen 4. Provide an assessment of the uncertainty associated with
but also from additional uncertainty regarding the mean- the estimated ground water flow components.
ing of the density of water obtained from the well due to
uncertain mixing conditions along the well screen.
Assessment of the impact of these uncertainties appears Acknowledgment
nontrivial but at least suggests that variable-density The authors would like to thank the reviewers Christian
ground water flow assessment using data from wells with Langevin and Ann Mulligan for their thoughtful and con-
long screens should be avoided. structive comments, which helped improve the manuscript.
Although the environmental water head approach has
the nicety of a simpler and more familiar expression of
the flow Equation 16, there is no true advantage over References
Equation 14. Whereas in the latter approach density cor- Bachu, S. 1995. Flow of variable-density formation water in
deep sloping aquifers: Review of methods of representation
rections are applied to the gradient component, in the
with case studies. Journal of Hydrology 164, no. 1–4: 19–38.
environmental water head approach, similar corrections Bachu, S., and K. Michael. 2002. Flow of variable-density for-
are incorporated in the calculation of environmental water mation water in deep sloping aquifers: Minimizing the
heads before the gradient operator is applied. Environ- error in representation and analysis when using hydraulic-
mental water heads may seem more practical because head distributions. Journal of Hydrology 259, no. 1: 49–65.
Bear, J. 1972. Dynamics of Fluids in Porous Media. New York:
they are more readily rendered in the form of isohyps
Dover Publications.
maps or vertical cross sections. However, such maps are Custodio, E. 1987. Salt-fresh water interrelationships under nat-
hardly useful and may even be considered ‘‘dangerous’’ ural conditions. In Groundwater Problems in Coastal
because they easily cause mis- or overinterpretation and Areas, Studies and Reports in Hydrology, vol. 45, ed.
do not allow visualization of the often large uncertainties. E. Custodio, 14–96. Paris, France: UNESCO.
DeWiest, R.J.M. 1967. Geohydrology. New York: John Wiley.
Furthermore, the choice of reference depth in settings
Holzbecher, E.O. 1998. Modeling Density-Driven Flow in
where salt water overlies fresh water or where saline sur- Porous Media. Berlin and Heidelberg, Germany: Springer.
face water is present, such as in estuarine and offshore Juster, T.C. 1995. Circulation of saline and hypersaline ground-
ground water hydrology, is nonintuitive, not described in water in carbonate mud: Mechanisms, rates, and an exam-
literature, and, therefore, less practical for nonspecialists. ple from Florida Bay. Ph.D. diss., Department of Geology,
University of South Florida-Tampa.
It can therefore easily be argued that the fresh water head
Lusczynski, N.J. 1961. Head and flow of ground water of vari-
approach with the appropriate correction for negative able density. Journal of Geophysical Research 66, no. 12:
buoyancy should be endorsed as the preferred approach in 4247–4256.
variable-density analyses because it only requires making Oberlander, P.L. 1989. Fluid density and gravitational variations
assumptions about the water density variations between in deep boreholes and their effect on fluid potential.
Ground Water 27, no. 3: 341–350.
well screens.
Reilly, T.E., and A.S. Goodman. 1985. Quantitative analysis of
The procedures and guidelines set forward in this saltwater-freshwater relationships in groundwater systems.
paper can be summarized as a set of four golden rules A historical perspective. Journal of Hydrology 80, no. 1–2:
that should be adhered to in order to correctly infer 125–160.
ground water flow (directions and magnitudes) of Simmons, C.T. 2005. Variable density groundwater flow: From
current challenges to future possibilities. Hydrogeology
variable-density ground water, namely:
Journal 13, no. 1: 116–119.
1. Collect fluid density information with all head measure- St. Germain, D.J. 2001. Salt water intrusion in coastal aquifers.
The Professional Geologist 38, no. 12: 3–7.
ments.
Van Dam, J.C. 1977. Determination of horizontal and vertical
2. Calculate horizontal flow components from fresh water groundwater flow from piezometric levels observed in
heads referenced to the same elevation. groundwater of varied densities. Delft Progress Report 3,
3. Calculate vertical flow components from the gradient of 19–34.
Van der Eem, J.P. 1992. Rekenen aan de stroming van zoet,
fresh water head with an appropriate correction for (nega-
brak en zout grondwater [Calculation of the flow of fresh,
tive) buoyancy of the ground water between the two mea- brackish and salt groundwater] [In Dutch]. KIWA-mededeling
surement depths. 121, Nieuwegein, The Netherlands.

V. Post et al. GROUND WATER 45, no. 6: 664–671 671