43 - Analytical Model of Salinity Risk From Groundwater Discharge in Semi-Arid, Lowland Floodplains

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

HYDROLOGICAL PROCESSES

Hydrol. Process. 23, 3428– 3439 (2009)


Published online 16 September 2009 in Wiley InterScience
(www.interscience.wiley.com) DOI: 10.1002/hyp.7447

Analytical model of salinity risk from groundwater discharge


in semi-arid, lowland floodplains
Kate L. Holland,* Ian D. Jolly, Ian C. Overton and Glen R. Walker
CSIRO Land and Water, PMB 2, Glen Osmond, South Australia 5064, Australia

Abstract:
River regulation and irrigated agricultural developments have increased saline groundwater discharge to semi-arid floodplain
environments, resulting in soil salinization and vegetation dieback. There is a need for rapid assessment tools to identify areas of
floodplain vegetation at risk from salinization. This article describes the development and testing of a simple, one-dimensional
analytical model that distributes floodplain groundwater discharge as seepage at the break of slope, evapotranspiration across
the floodplain and as groundwater flow into and out of the river. The analytical model provided comparable estimates
of groundwater discharge to MODFLOW (R2 D 0Ð998) over a broad range of floodplain scenarios. Using regional scale
geographic information system (GIS) data along an 85-km long reach of the lower River Murray in South Australia, the
model predicted the location of most (66%) of the unhealthy trees affected by seepage and soil salinization near the edge of
the river valley and more than 75% of the unhealthy trees across the broader floodplain. The model predicted 98% of the
observed variance in measured kilometre-by-kilometre river salt loads, identifying ‘hotspots’ of salt input to the river. It can
therefore be used as a rapid assessment tool to assess the potential salinity risk to the floodplain vegetation and river from
irrigation developments and river management. This approach can be applied more broadly, although the impact of wetlands
and oxbows, which may act to intercept groundwater within the floodplain, is not taken into account. Copyright  2009 John
Wiley & Sons, Ltd.

KEY WORDS groundwater; modelling; salinity; irrigation; floodplain; discharge; River Murray

Received 9 June 2009; Accepted 17 July 2009

INTRODUCTION Because the groundwater is naturally saline, the increased


evapotranspiration and seepage results in floodplain salin-
In the lower River Murray in South Australia, there are
ization, which consequently affects the health of flood-
natural inflows of saline regional groundwater to the
plain vegetation. These problems are further exacerbated
floodplains, wetlands and river channel. However, these
by less frequent removal of the accumulated salt in the
have greatly increased as a result of the development
floodplain by large floods, the frequency and duration
of large areas of irrigated horticulture (predominantly
of which has been greatly reduced by upstream storages
citrus, grapes, vegetables, stone fruits and nut crops)
and irrigation diversions (Jolly, 1996). Although a num-
immediately adjacent to the river and its floodplain.
ber of other factors such as clearing and coppicing of
High rates of recharge occur in these irrigation areas,
trees, grazing of livestock and feral animals, insect attack
which leads to the formation of groundwater mounds
and parasite infestation can contribute to floodplain veg-
with elevations of up to 20 m above those of the river
etation health decline, these are generally second-order
and the groundwater levels within the floodplain. These
effects in the lower River Murray compared with those
groundwater mounds cause significant increases in the
of salinization (Jolly, 1996).
inflow of groundwater to the river valley.
Estimates derived from Department for Environ-
When combined with river regulation by flow control
ment and Heritage (South Australia) floodplain vege-
weirs/locks that hold the river at higher average levels
tation floristic and tree health mapping in 2002–2003
than under natural conditions, the increased groundwater
(Smith and Kenny, 2005) suggest that approximately
inflows to the river valley have led to raised water tables
40% (40 000 ha) of the floodplain in South Australia
beneath the floodplain. In some areas, the inflows are now
is severely degraded by soil salinization (Walker et al.,
so high that the water table at the break in slope of the
2005), and this is expected to increase in the future.
river valley has reached the floodplain surface causing
groundwater seepage to occur. In response, the South Australian River Murray Salinity
The raised water tables beneath the floodplain have led Strategy (DWR, 2001a) has proposed several manage-
to increased rates of evapotranspiration of groundwater, ment options to ameliorate floodplain salinization, includ-
and in the case of seepage areas, waterlogging as well. ing groundwater lowering, improved flooding and the
establishment of floodplain protection zones to ensure
new irrigation development will not affect floodplains of
* Correspondence to: Kate L. Holland, CSIRO Land and Water, PMB 2, high conservation value. There is a need for decision sup-
Glen Osmond SA 5064, Australia. E-mail: kate.holland@csiro.au port tools that use published regional scale information

Copyright  2009 John Wiley & Sons, Ltd.


FLOODPLAIN SALINITY RISK ASSESSMENT 3429

to assist in the development and implementation of these 3. Loss of groundwater through evapotranspiration, ET
options. These decision support tools should identify [L T1 ] is calculated using the simple linear func-
where floodplain salinization can be minimized by salt tion described by Equations (2) and (3); where a is
interception schemes or floodplain irrigation protection the maximum rate of groundwater discharge by evapo-
zones and where floodplain groundwater levels can be transpiration at the floodplain surface on an areal basis
lowered to protect valuable floodplain communities from [L T1 ]. This maximum groundwater evapotranspira-
salinization. tion rate declines to zero at zext , which corresponds to
the average vegetation rooting depth; hf is the height
of the floodplain above river level [L]; and zext is
ANALYTICAL MODEL DEVELOPMENT the evapotranspiration extinction depth [L], or rooting
depth, below which there is no evapotranspiration.
The mathematical development of the model is summa-
rized below and complete derivations can be found in a[h  hf  zext ]
ET D h > hf  zext , 2
Holland et al. (2004). The definitions and units of vari- zext
ables used in the model are described in Table I. ET D 0 h  hf  zext ; and 3
4. A sharp cliff and a flat floodplain characterize the shape
Conceptual model of the floodplain and regional aquifers
of the river valley.
The floodplains have been conceptualized in a simple
cross-sectional model, comprising a surface layer of Governing equations
Coonambidgal Clay overlying a layer of Monoman Sands Under the above assumptions, the generalizsed flow
(Figure 1). This is based on drilling records (AWE, Equation (1) can be simplified by considering groundwa-
1999) and modelling studies around the Bookpurnong ter flow in terms of changes in the slope of the water
and Loxton Irrigation Areas (Armstrong et al., 1999; table between the edge of the floodplain and the river.
Barnett et al., 2002). The highland area is represented by
the unconfined Upper Loxton (Pliocene) Sands aquifer. 1. When the groundwater level is below the extinction
Both the highland and floodplain are underlain by the depth, groundwater flow is controlled by the hydraulic
Lower Loxton Sands, a relatively impermeable clayey gradient between the edge of the floodplain and the
sand formation. river and therefore the slope of the water table is
The arrows in Figure 1 represent groundwater flow constant, i.e.
directions and potential groundwater discharge sites.
Groundwater flow in the Upper Loxton Sands is a com- d2 h
Kb h  hf  zext  4
bination of irrigation recharge and regional groundwater dx 2
flow. Groundwater flow into the floodplain is discharged 2. When the groundwater level is above the extinction
as either seepage at the break of slope if the groundwater depth, groundwater flow is controlled by the hydraulic
level is above the floodplain surface, evapotranspiration gradient and evapotranspiration across the floodplain
through the floodplain when the water table is within and therefore the slope of the water table changes with
the evapotranspiration extinction depth (vegetation root- evapotranspiration across the floodplain, i.e.
ing depth), or as base flow to or from the river.
d2 h azext  hf C h
Kb 2
D h > hf  zext . 5
Model assumptions dx zext
The following assumptions are made: Non-dimensional equations
The above equations can be written in non-dimensional
1. Groundwater flow within the floodplain is one-dimen- form to minimize the number of variables, and to
sional, under steady-state conditions with no recharge, put into perspective the balance between the different
and the floodplain aquifer is homogenous and isotropic; floodplain groundwater discharge processes. To do this,
and we substitute the following:
2. Groundwater flow under the floodplain is defined by
Darcy’s Law (Equation 1), where Q is the discharge h Ł x zext
hŁ D , x D and zŁ D 1  .
of groundwater through a unit width of floodplain [L2 hf L hf
T1 ]; K is the horizontal hydraulic conductivity of the
The non-dimensional variables hŁ and x Ł normalize
aquifer [L T1 ]; b is the aquifer thickness [L]; and
water table elevation, location between the edge of the
dh/dx is the groundwater hydraulic gradient, where h
floodplain and the river to values between 0 and 1,
is the height of the groundwater above river level [L]
respectively (Figure 2). The non-dimensional variable zŁ
and x is the distance from the edge of the river valley
expresses evapotranspiration extinction depth in relation
[L] to a maximum distance L at the river [L];
to the river level, with the extinction depth above river
dh level when 0 < zŁ < 1 and below river level when 1 <
Q D Kb ; 1 zŁ < 0.
dx

Copyright  2009 John Wiley & Sons, Ltd. Hydrol. Process. 23, 3428– 3439 (2009)
DOI: 10.1002/hyp
3430 K. L. HOLLAND ET AL.

Table I. Definition and dimensions of model parameters

Parameter Definition Dimensions

Q Discharge of groundwater through a unit width of floodplain L2 T1


K Horizontal hydraulic conductivity of the aquifer L T1
b Floodplain aquifer thickness L
dh/dx Groundwater hydraulic gradient dimensionless
h Height of the groundwater above river level L
x Distance from the edge of the river valley L
L Maximum distance from the edge of the river valley to the river L
ET Evapotranspiration L T1
a Maximum rate of groundwater discharge by evapotranspiration L T1
hf Height of the floodplain surface above river level L
zext Evapotranspiration extinction depth below which there is no evapotranspiration L
hŁ Height of the water table above river level relative to the evapotranspiration dimensionless
extinction depth
xŁ Distance from the edge of the river valley relative to the maximum distance from dimensionless
the edge of the river valley to the river (0–1)
zŁ Evapotranspiration extinction depth relative to river level (1 < zŁ < 1) dimensionless
WŁ Ratio between the volume of groundwater that the floodplain aquifer can transmit dimensionless
and the volume of groundwater that can be discharged as evapotranspiration
Qin Volume of groundwater discharge entering the floodplain aquifer L2 T1
qin Volume of groundwater discharge to the floodplain aquifer relative to the volume of dimensionless
groundwater that the floodplain aquifer can transmit
Qs Volume of groundwater discharged as seepage at the edge of the floodplain L2 T1
qs Volume of groundwater discharged as seepage relative to the volume of groundwater dimensionless
that the floodplain aquifer can transmit
QET Volume of groundwater discharged as evapotranspiration through the floodplain L2 T1
surface
qET Volume of groundwater discharged as evapotranspiration relative to the volume of dimensionless
groundwater that the floodplain aquifer can transmit
Qr Volume of groundwater discharged to the river as base flow L2 T1
qr Volume of groundwater discharged to the river as base flow relative to the volume dimensionless
of groundwater that the floodplain aquifer can transmit
Ł
xcrit Point on the floodplain where the floodplain groundwater level equals the dimensionless
evapotranspiration extinction depth
h0Ł Height of the water table above river level at the break of slope where x Ł D 0 dimensionless
relative to the evapotranspiration extinction depth
Wedge Width of the floodplain at the edge of the river valley L
Cgw Floodplain groundwater salinities M L3

In non-dimensional form, the generalized flow Equa- Boundary conditions


tions (4) and (5) become: The river represents a constant head boundary condi-
tion, therefore:
1. Groundwater level below the extinction depth hŁ D 0 at x Ł D 1 8

d2 h2 At the edge of the floodplain (x Ł D 0) when there is


D 0 hŁ  z Ł 6
dx Ł2 no seepage, i.e. the water table is below the floodplain
surface, the boundary condition is given by:
2. Groundwater level above the extinction depth, i.e.
within the rooting zone dhŁ
D qin when hŁ < 1, 9
dx Ł
d2 hŁ hŁ  z Ł
D hŁ > z Ł 7 where the non-dimensional variable, qin , is the ground-
dx Ł2 WŁ
water discharge to the floodplain normalized with respect
Kbhf to the volume of groundwater that the floodplain aquifer
where WŁ D 1  zŁ . Qin L ), and Q is defined as the
can transmit (i.e. qin D Kbh
aL 2 f
in
volume of groundwater discharge entering the floodplain
WŁ is the ratio between the volume of groundwater
 aquifer [L2 T1 ].
Kbhf When the volume of groundwater entering the flood-
that the floodplain aquifer can transmit L and
plain is greater than the floodplain aquifer can transmit,
the volume of groundwater
  that can be discharged as seepage occurs so that the water table is at the flood-
evapotranspiration aL .
1  zŁ plain surface. When seepage occurs, the dimensionless

Copyright  2009 John Wiley & Sons, Ltd. Hydrol. Process. 23, 3428– 3439 (2009)
DOI: 10.1002/hyp
FLOODPLAIN SALINITY RISK ASSESSMENT 3431

Figure 1. Conceptual model of groundwater inputs to the floodplain and potential groundwater discharge pathways within the floodplain. Groundwater
entering the river valley can be discharged as either seepage at the break of slope, evapotranspiration through the floodplain, or as base flow to the
river

2. Evapotranspiration across the floodplain:


QET L
qET D ; and 14
Kbhf
3. Base flow to the river (x Ł D 1):
Qr L dhŁ
qr D D  Ł when hŁ D 0, 15
Kbhf dx

where Qs is the volume of seepage at the edge of


the floodplain, QET is the volume of groundwater dis-
charged by evapotranspiration, and Qr is the volume of
Figure 2. Schematic diagram of the floodplain cross-section showing groundwater discharged to the river as base flow. The
Ł occurs where the floodplain groundwater
dimensions (xŁ , hŁ and zŁ ). xcrit non-dimensional groundwater discharges express seepage
level equals the evapotranspiration extinction depth (i.e. hŁ D zŁ )
(qs ,), evapotranspiration (qET ) and base flow (qr ) relative
to the volumeof groundwater
 that the floodplain aquifer
discharge at the edge of the floodplain (x Ł D 0) is defined Kbhf
as: can transmit L .
qin D 1 when hŁ ½ 1, 10
Floodplain scenarios
where qin is the maximum possible dimensionless dis- There are several scenarios that may arise within the
charge through the floodplain aquifer, defined thus: floodplain: (i) groundwater level completely below the
dhŁ rooting depth across the entire floodplain; (ii) ground-
qin D  when hŁ D 1 at x Ł D 0. 11 water level completely within the rooting depth across
dx Ł
the entire floodplain but no seepage occurs; (iii) ground-
When seepage does not occur, the dimensionless dis- water level completely within the rooting depth across
charge at the edge of the floodplain is defined as: the entire floodplain and seepage occurs; (iv) groundwa-
ter level below the rooting depth at the river but above
Qin L
qin D when hŁ < 1. 12 the rooting depth at the floodplain edge but no seepage
Kbhf occurs; (v) groundwater level below the rooting depth
We are interested in determining the proportion of at the river but above the rooting depth at the flood-
groundwater discharged as seepage, evapotranspiration plain edge and seepage occurs; and (vi) groundwater level
and base flow. We define the possible discharges through above the rooting depth at the river but below the rooting
a unit width of floodplain as: depth at the floodplain edge. Mathematically, we con-
sider two general situations that are used individually or
1. Seepage at the edge of the floodplain (x Ł D 0): in combination, depending on the particular scenario, to
solve for the river and seepage discharges. The evap-
Qs L otranspiration discharge is determined from the differ-
qs D when hŁ ½ 1; 13
Kbhf ence between these and the discharge into the floodplain.

Copyright  2009 John Wiley & Sons, Ltd. Hydrol. Process. 23, 3428– 3439 (2009)
DOI: 10.1002/hyp
3432 K. L. HOLLAND ET AL.

The following generalized solutions are used to solve The height of the water table at the edge of the floodplain
Ł
Equations
 (4) hD Ax C B for hŁ zŁ and (5) hŁ D zŁ C can be calculated from:
Ł Ł  
C cos h px Ł C D sin h px Ł for hŁ > zŁ , respec- p 1
W W qin WŁ sin h p  zŁ
tively. Values A, B, C and D are solved by substituting W Ł
h0Ł D   C zŁ . 22
the boundary conditions for each scenario. 1
cos h p

(a) River level below rooting depth (zŁ > 0). We define
Ł Ł Therefore, when h0Ł D 1, i.e. the water table is at the
a point on the floodplain xcrit 0 < xcrit < 1 where the
floodplain surface, qin can be calculated from zŁ and WŁ :
floodplain groundwater level equals the evapotranspira-
 
tion extinction depth (i.e. hŁ D zŁ ). At any point between p 1
Ł
xcrit and the river, the groundwater level is below the
Ł
qin W sin h p  zŁ
W Ł
extinction depth, i.e.: 1z D Ł   . 23
1
cos h p
hŁ D qr 1  x Ł  when xcrit
Ł
< x Ł < 1. 16 WŁ
When seepage does not occur, qr can be calculated
Ł
Substituting hŁ D zŁ and x Ł D xcrit into (16) allows us to from (21) with qin defined by (12). When seepage occurs,
Ł
quantify qr for known values of zŁ and xcrit , i.e.: qin can be determined from (23), and then qr can be
calculated from (21).

qr D Ł . 17 Check of the analytical model solutions against an
1  xcrit 
equivalent numerical model
Using the generalized solution for (5), qr can be calcu- The analytical model was compared to the numerical
lated iteratively from values of qin , zŁ and WŁ , i.e.: groundwater flow model, MODFLOW, over a wide
  range of possible scenarios. Both models simulated a
qr  zŁ simple cross-sectional area of highland and floodplain
qin D qr cos h p . 18
qr WŁ (Figure 1). Steady-state modelling scenarios representing
up and downstream of a weir; uncleared, cleared and
To determine whether seepage occurs, the non-dimen- irrigated highland recharge rates; five floodplain widths
sional height of the water table at the edge of the (L); and two maximum floodplain evapotranspiration
floodplain is defined as hŁ D h0Ł at x Ł D 0, and can be rates (a) were used. Details of the model configurations
calculated using the generalized solution for (5) from: and parameterizations can be found in Holland et al.
  (2004).
Ł
p qr  zŁ Analytical model estimates of floodplain groundwater
Ł
h0 D qr W sin h p C zŁ . 19
qr WŁ discharge patterns were compared to those predicted
by MODFLOW for each of the 30 scenarios. The
Therefore, when h0Ł D 1, i.e. the water table is at the analytical model estimates of seepage, evapotranspiration
floodplain surface, the discharge to the river can be and base flow discharge per unit width of aquifer [L3
calculated from zŁ and WŁ : T1 ] were consistent with those of MODFLOW over
  a broad range of floodplain physical and hydraulic
p Ł
Ł qr  z parameters. The regression was statistically significant
1  z D qr WŁ sin h p . 20 (n D 90, R2 D 0Ð998, P < 0Ð001) and described by the
qr WŁ
following equation:
Ł
When seepage does not occur, the value of xcrit can
MODFLOW D 0Ð99Analytical C 0Ð23. 24
be determined when (17) is substituted into (18) and qin
Ł
is defined by (12). This value of xcrit is then used to Seepage was predicted in six scenarios, characterized
calculate qr from (17). When seepage occurs, qr can by large groundwater inputs from irrigation, wide flood-
be determined from (20), and then qin can be calculated plains and low evapotranspiration rates.
from (18). The MODFLOW model was sensitive to starting
heads, despite steady-state conditions. Generally, initial
(b) River level within rooting depth (zŁ < hŁ ). Using conditions are not important for steady-state simulations;
the generalized solution for (5), the discharge to the river, however, they can be important in certain non-linear
qr , is determined from: situations where discharge, transmissivity or saturation
  are a function of head (McDonald and Harbaugh, 1988).
p 1
Ł Ł
qin W C z sin h p A traditional sensitivity analysis to model parameters
qr D  WŁ .
 21 was not performed for the MODFLOW model. How-
p 1 ever, the analytical model provides the means for analysis
Ł
W cos h p
WŁ of the sensitivity of both models to final groundwater

Copyright  2009 John Wiley & Sons, Ltd. Hydrol. Process. 23, 3428– 3439 (2009)
DOI: 10.1002/hyp
FLOODPLAIN SALINITY RISK ASSESSMENT 3433

discharge estimates. This analysis was carried out by


comparing the discharge to the river relative to the dis-
charge entering the floodplain normalized with respect to
the evapotranspiration extinction depth. This showed that
both models were sensitive to situations where discharge
to the floodplain was greater than aquifer transmissiv-
ity normalized with respect to evapotranspiration extinc-
tion depth (i.e. high discharge into the floodplain relative
to floodplain aquifer transmissivity) and where potential
evapotranspiration rates greatly exceed floodplain aquifer
transmissivity (i.e. wide floodplains, high discharge into
the floodplain and river levels within the evapotranspira-
tion extinction depth). Conversely, they were insensitive
when the floodplain aquifer transmissivity was greater
than the potential evapotranspiration rate (i.e. groundwa-
ter below the evapotranspiration extinction depth) and
so all discharges were transmitted to the river and no
seepage or evapotranspiration occurred.

APPLICATION TO LOWER RIVER MURRAY


FLOODPLAINS
Figure 3. Study area between Locks 4 and 3 in the Riverland region of
Site description the River Murray

The study area (Figure 3) represents 85 km of River


Murray floodplains in South Australia, located upstream measured groundwater depths in regional bores. High-
of Overland Corner between Locks 4 and 3 (flow con- land groundwater gradients ranged between 0Ð0003 and
trol weirs, 516 and 431 km upstream of the River Mur- 0Ð0364. Reported regional hydraulic conductivity values
ray mouth, respectively). The River Murray valley is ranged between 1 and 5 m d1 (Barnett 1991; Doble
incised into the Pliocene Sands in this region, and com- et al., 2006), a value of 2 m d1 was used in this study.
prises 21 600 ha of floodplains. There are approximately The total discharge to each floodplain division was cal-
7700 ha of highland irrigated horticulture along the edge culated from the discharge per unit width (Qin ) and the
of the river valley in the study area. The geology of the width of the floodplain at the edge of the river valley
River Murray valley in this region consists of an upper (Wedge ).
layer of heavy Coonambidgal clay, underlain by Mono- The distance between the edge of the river valley
man Sands. and the river (L) was calculated from an average of
the two division sides. A 100 GL d1 flood extends
Methodology across most of the floodplain and is a good approximation
of the elevation at the break of slope and the average
The floodplain was divided into 447 divisions of floodplain surface. The difference between the elevation
approximately 250 m width based on floodplain ground- of a 100 GL d1 flood (modelled floodplain surface) and
water flow paths between the edge of the river valley and the river elevation at entitlement flow (hf ) was calculated
the river. Within the study area, the following simulations from GIS elevation data (Overton et al., 1999) for each
allowed the model to be tested and validated against field division.
data: Evapotranspiration parameters were set to constant val-
ues based on measured lower River Murray floodplain
1. A range of possible river regulation effects from soil- and vegetation-limited groundwater discharge rates.
immediately upstream of Lock 3 to downstream of A maximum evapotranspiration rate (a) of 0Ð1 mm d1
Lock 4; was used. This value lies between the soil limited ground-
2. Irrigated and non-irrigated highland areas; water discharge rate of 0Ð02–0Ð03 mm d1 measured by
3. Floodplain widths up to 5700 m and Jolly et al. (1993) and the vegetation-limited groundwater
4. Measured river salt loads at kilometre intervals and discharge rate of 0Ð03–2Ð0 mm d1 measured by Thor-
detailed floodplain tree health maps. burn et al. (1993). An evapotranspiration extinction depth
(zext ) of 2 m was used as this represents a combina-
Model input parameters. Groundwater inflows were tion of the shallow (0Ð5–1Ð5 m) soil limited and deeper
calculated from highland groundwater gradients, and (1Ð5–5 m) vegetation-limited evapotranspiration extinc-
regional aquifer hydraulic conductivity and thickness. tion depths (Doble et al., 2006).
Highland groundwater gradients were estimated from The horizontal floodplain hydraulic conductivity (K)
groundwater contours, which were interpolated from was set to a constant value of 10 m d1 . It is know to vary

Copyright  2009 John Wiley & Sons, Ltd. Hydrol. Process. 23, 3428– 3439 (2009)
DOI: 10.1002/hyp
3434 K. L. HOLLAND ET AL.

between 10 and 35 m d1 , however, the value used by aquifer cannot transmit incoming groundwater and evapo-
Doble et al. (2006) to model the Bookpurnong floodplain transpiration occurs, causing the floodplain water table
was adopted in this study. Floodplain aquifer thickness to rise to within 2 m (zext ) of the surface. Evapotrans-
(b) was set to a constant value of 7 m following Doble piration predictions were compared to tree health in the
et al. (2006). Floodplain groundwater salinities (Cgw ) middle of the floodplain, excluding a buffer within 100 m
were interpolated from measured values in highland bores of the river and 100 m from the edge of the river valley.
along the edge of the river valley (MDBC, 2005). These These buffer zones represent areas where floodplain tree
values represent the long-term spatial distribution of health is influenced by proximity to the river or edge of
natural salt loads to the river valley. Modelled salinities the river valley, rather than salinization associated with
range from 10 650 to 42 450 mg L1 . groundwater discharge by evapotranspiration.
Ideally, model estimates of groundwater discharge
would be compared to measured, field validated ground- River salt loads. Model estimates of base flow were
water discharge. However, field measurements of seep- converted from the volume of groundwater discharged
age, groundwater losses by evapotranspiration and base to the river per division (m3 d1 ) to tons of salt using
flow groundwater discharge are not available or easily interpolated floodplain groundwater salinities (mg L1 ).
measured at a regional scale. Instead, surrogate measures These salt loads were assigned to the nearest river
of groundwater discharge: observed seepage areas, tree kilometre for each division as tons of salt entering the
health and river salt loads, were used to evaluate model river per kilometre per day for each river kilometre.
predictions. Modelled salt loads were compared to measured ‘Run
Groundwater inflows from interpolated contours were of River’ salt loads determined from river flow rates and
calibrated against known conditions to assist in determin- in-stream salinity measurements taken over 5 years at
ing accurate groundwater inflows. In many areas, ground- low flows (DWR, unpublished data, 2001b; Porter, 2001).
water depths were missing, so groundwater inflows were An average of five ‘Run of River’ salt loads was used
estimated from irrigation history and by matching the to estimate long-term salt accessions. The variability of
observed patterns and magnitude of salt loads. These cur- salt loads measured depended on recent (6–18 months)
rent groundwater inflows represent the best estimate for flow history, particularly in areas with large permanently
the study area. inundated wetlands upstream of Lock 3.

Floodplain vegetation health. Floodplain vegetation Sensitivity analysis. Model sensitivity to input param-
health was mapped in approximately 40% of the study eters was tested by varying groundwater inflows to
area (PPK, 1997, 1998; AWE, 2000). This GIS coverage the river valley. Equation (12) shows that changes to
includes approximately 3500 ha of trees and approxi- groundwater inflows (Qin ) are proportional to changes
mately 5000 ha of other vegetation, including shrubs, to other important model parameters: aquifer hydraulic
ground covers and wetland species. conductivity (K), aquifer thickness (b) and height of
The vegetation mapping was completed over all of the the floodplain surface above river level (hf ). It is also
floodplain to the south of the river and several flood- inversely proportional to the length of the groundwater
plains on the northern side of the river. Predominant flow path (L), which can be accurately estimated in the
tree communities in the study area include black box GIS model. Therefore, model sensitivity was tested by
(Eucalyptus largiflorens, F. Muell.), red gum (E. camald- varying groundwater inflows, which is the same as vary-
ulensis, Dehnh.) and cooba (Acacia stenophylla, A.Cunn. ing K, or b or hf by the same amount.
ex Benth.). Tree communities were mapped as healthy
(>75% canopy cover), poor health (25–75% canopy
cover) and dead (<25% canopy cover) from ground sur- RESULTS
veys and aerial photography. Because of the complexity
of floodplain vegetation associations, only tree health was Model results for different groundwater inflow scenarios
used in this analysis. were compared to observed seepage areas, mapped tree
Areas where seepage was observed, including seep- health and measured river salt loads between Locks 4 and
age faces and areas of emergent aquatic vegetation were 3 in the lower River Murray.
mapped from ground surveys and aerial photography.
Observed seepage areas were compared to predicted seep- Seepage
age areas for a range of groundwater inflow scenarios. Seepage was observed in 68 floodplain divisions near
Seepage assessments of tree health were limited to within irrigation areas over a total distance of 17Ð6 km at the
100 m of the edge of the river valley. Areas where the edge of the river valley. Seepage predictions were sen-
floodplain was <100 m wide were considered separately sitive to changes in groundwater inflows, hydraulic con-
due to possible bank recharge influences on vegetation ductivity, aquifer thickness and height of the floodplain
health. surface above river level. The number of divisions where
Ł
The parameter xcrit was used to assess when vegetation seepage was observed and predicted rose with increas-
was at risk of salinization associated with groundwater ing inflows, as did erroneous predictions where seepage
Ł
discharge. For divisions where xcrit > 0, the floodplain was modelled, but not observed (Table II). Increasing

Copyright  2009 John Wiley & Sons, Ltd. Hydrol. Process. 23, 3428– 3439 (2009)
DOI: 10.1002/hyp
FLOODPLAIN SALINITY RISK ASSESSMENT 3435

Table II. Number of divisions where seepage was observed and


modelled for different groundwater inflows

Groundwater inflow scenarios

ð0Ð5 ð1Ð0 ð1Ð5 ð2Ð0

Observed and modelled 1 33 46 51


Observed, not modelled 67 35 22 17
Not observed, modelled 0 14 28 67

Table III. Seepage as a predictor of area of trees classed as good


health, poor health or dead within 100 m of the highland edge of
the floodplain

Good Poor Dead

Observed seepage area (ha)


Not observed 87 18 11
Observed 25 12 15
Total area 112 30 26
Groundwater inflow scenario ð1Ð0
Not modelled 105 24 14
Modelled 7 6 12
Groundwater inflow scenario ð1Ð5
Not modelled 97 20 13
Modelled 15 10 13
Groundwater inflow scenario ð2Ð0
Not modelled 78 19 8
Modelled 34 11 18

groundwater inflows by 50% (ð1Ð5) increased the num-


ber of divisions correctly predicted by the model (from
33 to 46), but doubled the number of divisions where
seepage was erroneously predicted (from 14 to 28). This
shows that modelled seepage predictions are sensitive to
estimates of groundwater inflows, hydraulic conductiv-
ity, aquifer thickness and height of the floodplain surface Figure 4. Evapotranspiration and seepage predictions (a) and observed
above river level. Observed seepage may also be asso- tree health (b) near the Bookpurnong Irrigation Area, showing that
ciated with aquifer thinning at the break of slope due to most of the unhealthy trees were in areas where seepage and/or
evapotranspiration were predicted
deposition of thick clay layers, i.e. a reduction in aquifer
transmissivity at the break of slope. The model cannot
account for these spatial variations. trees where seepage was observed and predicted were
Seepage estimates were compared to tree health within unhealthy.
100 m of the edge of the river valley. Seepage was
not observed or predicted in any divisions where the Evapotranspiration
floodplain was <100 m wide. In areas where seep- Floodplain divisions were grouped into those where the
age was observed, approximately half of the treed area floodplain aquifer could transmit all incoming groundwa-
was classed as unhealthy (either poor health or dead, ter discharge without evapotranspiration (ET D 0) and
Table III). Similarly, most of the unhealthy trees (66%) where evapotranspiration was predicted to occur (ET
were in areas where the model predicted seepage under >0). More than 75% of the unhealthy trees in the study
current groundwater inflows (ð1Ð0). This was not statisti- area were in divisions where evapotranspiration was pre-
cally significant [one-way analysis of variance (ANOVA), dicted for the ð1Ð0, ð1Ð5 and ð2Ð0 groundwater inflow
P D 0Ð13]. scenarios. The area of unhealthy or dead trees in divi-
Seepage predictions matched observed seepage areas sions where evapotranspiration was predicted was three
where groundwater inflows were well characterized, e.g. times greater than the area of unhealthy or dead trees
Bookpurnong Irrigation Area downstream of Lock 4 in divisions where no evapotranspiration was predicted
(Figure 4). Within 100 m of the edge of the river valley, for the ð1Ð0, ð1Ð5 and ð2Ð0 groundwater inflow scenar-
seepage was predicted in four of the five divisions where ios (Table IV). This was statistically significant (one-way
it was observed. In this area, more than 97% of the ANOVA, P D 0Ð03).

Copyright  2009 John Wiley & Sons, Ltd. Hydrol. Process. 23, 3428– 3439 (2009)
DOI: 10.1002/hyp
3436 K. L. HOLLAND ET AL.

Table IV. Evapotranspiration (ET) as a predictor of area of trees


classed as good health, poor health or dead in the middle of the
floodplain (>100 m from the highland and river’s edge)

Good Poor Dead

Observed tree health (ha)


Total area 1689 323 532
Groundwater inflow scenario ð0Ð5
ET D 0 1040 155 186
ET >0 648 168 346
Groundwater inflow scenario ð1Ð0
ET D 0 841 87 121
ET >0 847 236 411
Groundwater inflow scenario ð1Ð5
ET D 0 816 80 120
ET >0 872 243 412
Groundwater inflow scenario ð2Ð0
ET D 0 802 78 119
ET >0 886 245 413

Almost half (45%) of the area of unhealthy trees in


divisions where evapotranspiration was predicted under
current groundwater inflows were on Pyap floodplain
(Figure 5). Pyap floodplain lies mid-reach (river km
480–466) near a small irrigation area and includes
approximately 290 ha of unhealthy and approximately
480 ha of healthy trees. The results from Pyap flood-
plain show that the model predicts where the greatest
probability of poor tree health occurs at a sub-regional
scale. However, the model does not predict the small-
scale variations in tree health associated with differences
in flooding frequency, soils and salinity.

River salt loads


River salt loads measured over 5 years (1987, 1991,
1997, 1998 and 2001) between Locks 4 and 3 were Figure 5. Evapotranspiration predictions (a) and observed tree health
used to refine groundwater inflow estimates during initial (b) on Pyap floodplain, showing that most of the unhealthy trees were
testing. Average measured total salt loads over the reach observed in areas where evapotranspiration was predicted
were 205Ð0 š 14Ð0 SE tonnes d1 (DWR, unpublished
data, 2001b; Porter, 2001). This is comparable to model 95% of the variance in measured salt loads. However,
values under current groundwater inflows (Table V). model predictions of floodplain attenuation vary with
The ability of the model to describe groundwater dis- groundwater inflow scenario (i.e. ð0Ð5 and ð1Ð0 < 30%,
charge patterns was also tested against common assump- ð1Ð5–30% and ð2Ð0 > 30%; Table V). In addition, this
tions of no attenuation and 30% attenuation of inflows simple approximation does not indicate where saliniza-
by the floodplain (e.g. Barnett et al., 2002). The assump- tion associated with seepage or high rates of evapotran-
tion of 30% attenuation by the floodplain explained spiration occurs, or how these patterns change with the
magnitude of groundwater inflows.
Table V. Total river salt loads (tonnes d1 ) between Locks 4
The cumulative discharge of salt to the river (tonnes
and 3 predicted by the model and assuming either 0 or 30% d1 ) under current groundwater inflows was plotted
attenuation by the floodplain against distance from the mouth of the river (river km)
between Lock 4 (516 km) and Lock 3 (431 km). The
Groundwater inflow scenarios regression of modelled and measured salt loads was
Salt loads
(tonnes d1 ) ð0Ð5 ð1Ð0 ð1Ð5 ð2Ð0 statistically significant (n D 85, R2 D 0Ð979, P < 0Ð001;
Figure 6)
Measured 205Ð0 š 14Ð0
Modelled 120Ð7 219Ð0 297Ð1 364Ð5 Modelled D 0Ð987Measured C 9Ð45, 25
0% Attenuation 141Ð3 251Ð2 423Ð9 565Ð2
30% Attenuation 98Ð9 175Ð8 296Ð7 395Ð6 indicating good prediction of the patterns of salt input.
Salt loads were highest near the large irrigation areas

Copyright  2009 John Wiley & Sons, Ltd. Hydrol. Process. 23, 3428– 3439 (2009)
DOI: 10.1002/hyp
FLOODPLAIN SALINITY RISK ASSESSMENT 3437

of the river valley. The model does not simulate pos-


sible thinning of the floodplain aquifer associated with
geomorphologic features.
Evapotranspiration predictions were relatively insensi-
tive to groundwater inflows, being similar for groundwa-
ter inflow scenarios at or above current conditions (ð1Ð0,
ð1Ð5 and ð2Ð0). In these three scenarios, divisions where
evapotranspiration was predicted contained three times as
much area of unhealthy trees.
Salt loads were sensitive to estimates of groundwater
inflows and groundwater salinity. The model estimates of
current salt loads exceeded measured salt loads in down-
stream portions of the study area (river km 460–431).
This part of the study area is characterized by perma-
nently inundated wetlands and greater variability of the
kilometre-by-kilometre river salinity data (Figure 6). It is
likely that these wetlands act as salt stores, intercepting
groundwater discharge and thereby minimizing salt loads
in the river channel during low flows.
Figure 6. Cumulative salt loads from Lock 4 (516 km) to Lock 3
(431 km) showing measured and modelled salt loads under current
groundwater inflows (R2 D 0Ð979) DISCUSSION
The scenarios used to compare the analytical and MOD-
FLOW models are representative of the conditions expe-
(river km 516–480), with smaller peaks near smaller irri-
rienced in the lower River Murray in South Australia.
gation areas (river kms 460, 450 and 440). There is some
They are based on previously reported regional values
off set between the locations of peak inputs, with model for the area and represent a broad range of likely mod-
predictions 1–5 km upstream of observed salt loads. This elling scenarios. Therefore, the analytical model can also
variability may be due to the in-stream sampling tech- be applied to other scenarios where the basic conceptual
nique, where salt loads are not measured until the dense, model of a gaining stream holds, but recharge rates and
saline groundwater mixes with river water. This mixing aquifer properties differ.
may occur several kilometres downstream of groundwa- The analytical model also gives an indication of the
ter inflows. It may also arise from variations in recent sensitivity of groundwater modelling to the selection of
flow history or the dynamic nature of the groundwater modelling parameters. It indicates that modelling param-
flow system. The steady-state conditions assumed by the eters should be chosen carefully for scenarios where
model do not attempt to simulate this temporal variability discharge to the floodplain is greater than the volume
between measurement periods, but provide an estimate of of groundwater that the floodplain aquifer can transmit
long-term conditions. normalized with respect to evapotranspiration extinction
depth and where potential evapotranspiration rates greatly
Sensitivity analysis exceed the volume of groundwater that the floodplain
aquifer can transmit. Modelling parameters with these
Surrogate measures of groundwater discharge were conditions result in significant proportions of groundwa-
used to assess model sensitivity to parameterization. The ter being discharged as seepage and evapotranspiration,
sensitivity analysis showed that calculation of groundwa- affecting base flow estimates.
ter discharge entering the river valley (Qin ), particularly When the analytical model was applied to floodplains
near irrigation areas was critical. Those floodplains where in the lower River Murray, the accuracy of seepage pre-
groundwater contours were well documented (particu- dictions was strongly dependent on the estimates of cur-
larly near large irrigation areas) were predicted well. In rent groundwater inflows and aquifer thickness. Higher
contrast, estimates of Qin near smaller irrigation areas groundwater discharge to the river valley resulted in
where regional bores were absent relied on estimates from greater success in predicting seepage areas, but increased
irrigation area history and observations of floodplain veg- errors, i.e. seepage modelled, but not observed. Model
etation and salt loads. predictions of seepage areas encompassed most of the
Seepage predictions suggested that groundwater unhealthy trees (66%) under current groundwater inflows
inflows, hydraulic conductivity or floodplain aquifer (ð1Ð0); however, this was not statistically significant.
thickness was over-estimated by 50–100%. The presence It is suggested that thinning of the floodplain aquifer
of backwaters near many of the observed seepage areas might contribute to observed but not predicted seepage in
and bore logs (AWE, 1999) in some of these areas sug- areas near backwaters under current groundwater inflows.
gest that seepage often occurs when groundwater flow is Thick clay lenses were observed in bore logs from seep-
impeded by a thickening of the clay layer near the edge age areas near backwaters in the study area. Another

Copyright  2009 John Wiley & Sons, Ltd. Hydrol. Process. 23, 3428– 3439 (2009)
DOI: 10.1002/hyp
3438 K. L. HOLLAND ET AL.

source of error arises from the model assumption that groundwater being discharged as seepage. This is an
the floodplain surface is horizontal and flat. If the flood- area of future research, and has the potential to refine
plain is sloping then the model will not accurately predict the model’s estimates of seepage, salt loads and evapo-
whether or not seepage will occur at the break of slope. transpiration discharge through the floodplains. Airborne
Similarly, the model is unable to predict seepage areas geophysics could play an important role in mapping sub-
which, in part, may be due to topographical low points in surface geological features within the floodplains, as has
the vicinity of the break in slope, i.e. ephemeral wetlands been shown for the highland areas of the lower River
and oxbows. Murray.
More than 75% of the unhealthy trees in the study
area were in divisions where evapotranspiration was pre-
dicted. These divisions contained a significantly greater CONCLUSIONS
area of unhealthy or dead trees compared to divisions
where no evapotranspiration was predicted. The model A simple, one-dimensional, cross-sectional analytical
predicts evapotranspiration when floodplain water levels model of a low land river floodplain was comparable to
are within the extinction depth (zext ). This occurs when MODFLOW numerical estimates of patterns of ground-
groundwater inflows exceed the volume of groundwa- water discharge. The models predicted groundwater dis-
ter that the floodplain aquifer can transmit without dis- charge as seepage at the break of slope, evapotranspira-
charge by evapotranspiration. In both situations, soil lim- tion through the floodplain and base flow to the river.
ited groundwater discharge becomes a significant compo- However, the analytical model computing requirements
nent of the water balance. This suggests that salinization were lower and the required data inputs could be obtained
associated with groundwater discharge by evapotranspira- from published regional scale information.
tion is the principal process driving floodplain vegetation The model was applied to a study area compris-
health (along with salinization and waterlogging due to ing 85 km of floodplains in the lower River Murray.
seepage) despite the complex interactions between flood- The model’s predictions provided good correlations with
plain geomorphology, flooding and land use history on observed seepage areas, vegetation health data and mea-
vegetation health. This result indicates the usefulness of sured salt loads to the river. Model predictions of seep-
this modelling approach for predicting floodplain vegeta- age areas were sensitive to estimates of groundwater
tion responses to management. inflows. Observations of floodplain geomorphology and
The overall pattern and magnitude of modelled salt poor prediction of some seepage areas suggested that
load predictions are comparable to the measured values, seepage might be associated with floodplain aquifer thin-
which is more accurate than simply assuming that 30% ning near backwaters and local topographical low points.
of groundwater inflows are attenuated by the floodplain. This requires further investigation and conceptualization.
This means that the model can be used to predict future Airborne geophysics could play an important role in
salt loads from new irrigation developments and can mapping sub-surface geological features within the flood-
also be used as a planning tool to identify the impact plains.
of proposed developments on floodplain and wetland More than 75% of the unhealthy trees in the study area
health and river salinity. However, accurate estimates of were in divisions where evapotranspiration was predicted.
groundwater inflows are crucial. This suggests that salinization associated with ground-
In areas upstream from locks, where the river level water discharge by evapotranspiration is the principal
is within the evapotranspiration extinction depth, high process driving floodplain vegetation health, despite the
evapotranspiration rates across the floodplain are pre- complex interactions between floodplain geomorphology,
dicted. Not surprisingly, in these scenarios the model is flooding and land use history on vegetation health.
sensitive to small changes in parameterization and con- The model has shown that with good groundwater
ceptualization. Small changes to model parameters (hf , inflow data, it is capable of predicting kilometre-by-
zext , L, and a) can change model predictions from small kilometre salt loads to identify ‘hotspots’ of salt input
discharges to the river, to large discharges from the river to the river. Model predictions of salt loads are more
into the floodplain. Therefore, model predictions need accurate under a range of groundwater inflow scenarios
to be tested against independently measured field data than simply assuming that 30% of groundwater inflows
(observed seepage, tree health and river salt loads) to are attenuated by the floodplain.
ensure that accurate predictions are made. The model highlights the physical relationships that
The model presented in this article does not attempt exist between floodplain variables in the calculation of
to simulate floodplain wetlands, backwaters or oxbows, interception of groundwater by the floodplain. The model
which occur as the river meanders across the floodplain. is sensitive to parameterization when the river level or
This is because the role that these water bodies play floodplain water level is within the evapotranspiration
in intercepting saline groundwater flowing towards the extinction depth. This is typical of conditions upstream
river is still unknown. The analysis of seepage predic- of weirs and near irrigation areas.
tions under current groundwater inflows also highlighted The approach described in this article can be applied
the need to determine the role that aquifer thinning near more broadly, although the impact of wetlands and
backwaters and local topographical low points play in oxbows, which may act to intercept groundwater within

Copyright  2009 John Wiley & Sons, Ltd. Hydrol. Process. 23, 3428– 3439 (2009)
DOI: 10.1002/hyp
FLOODPLAIN SALINITY RISK ASSESSMENT 3439

the floodplain, is not taken into account. The model DWR. 2001a. South Australian River Murray Salinity Strategy 2001-
2015 . Department for Water Resources, Government of South
provides a tool to address the management of floodplain Australia: Adelaide.
and wetland health and river salinity. It can also be used Holland KL, Overton IC, Jolly ID, Walker GR. 2004. An Analyt-
to identify environmental protection zones for a range of ical Model to Predict Regional Groundwater Discharge Pat-
terns on the Floodplains of a Semi-Arid Lowland River , CSIRO
groundwater inflow scenarios. Land and Water Technical Report 06/04. CSIRO: Adelaide.
http://www.clw.csiro.au/publications/technical2004/tr6-04.pdf.
Jolly ID. 1996. The effects of river management on the hydrology
and hydroecology of arid and semi-arid floodplains. In Floodplain
ACKNOWLEDGEMENTS
Processes, Anderson MG, Walling DE, Bates PD (eds). John Wiley
& Sons: New York; 577–609.
The first author was supported by an Australian Post- Jolly ID, Walker GR, Thorburn PJ. 1993. Salt accumulation in semi-
graduate Award scholarship and Centre for Ground- arid floodplain soils with implications for forest health. Journal of
water Studies bursary at the time of the project, Hydrology 150: 589– 614.
McDonald MC, Harbaugh AW. 1988. A Modular Three-Dimensional
under the supervision of Glen Walker, Steve Tyerman, Finite-Difference Ground-Water Flow Model , Chapter A1, Book
Craig Simmons, Lisa Mensforth and Andrew Telfer. 6, United States Geological Survey Technical Water Resources
Land and Water Australia (Project CWS8-Guidelines Investestigation: Washington, DC.
MDBC. 2005. Basin Salinity Management Strategy, SIMRAT v2Ð0Ð1
for Managing Groundwater for Vegetation Health in Final Report. Murray-Darling Basin Commission: Canberra.
Saline Areas), the River Murray Catchment Water Overton IC, Newman B, Erdmann B, Sykora N, Slegers S. 1999.
Management Board, and the South Australian Salin- Modelling floodplain inundation under natural and regulated flows in
the Lower River Murray. Proceedings of the 2nd Australian Stream
ity Mapping and Management Support Project provided Management Conference, Adelaide, February, 1999.
funding towards this project. Craig Simmons, Neville Porter B. 2001. Run of River Salinity Surveys. A method of measuring
Robinson, Anthony Barr and David Rassam provided salt load accessions to the River Murray on a kilometre by
kilometre basis. Proceedings of the Murray Darling Basin Groundwater
useful suggestions and comments that improved this Workshop, Victor Harbour, September, 2001.
article. PPK. 1997. Assessment of the Impact of the Loxton Irrigation
District on Floodplain Health and Implications for Future Options,
PPK Environment and Infrastructure Report No. 27J121A 97–542,
Adelaide.
REFERENCES PPK. 1998. Assessment of the Impact of the Bookpurnong/Lock 4
Armstrong D, Yan W, Barnett SR. 1999. Loxton Irrigation Area— Irrigation District on Floodplain Health and Implications for Future
Groundwater Modelling of Groundwater/River Interaction. Department Options, PPK Environment and Infrastructure Report No. 27K055A
of Primary Industries and Resources: Adelaide. 98– 422, Adelaide.
AWE 1999. Clarks Floodplain Investigations. Prepared for the Loxton Smith F, Kenny S. 2005. Floristic vegetation and tree health mapping,
to Bookpurnong Local Action Planning Committee. Australian Water River Murray floodplain, South Australia. Department for Environment
Environments Report No. 98Ð031-2, Adelaide. and Heritage Internal Report, Adelaide.
AWE. 2000. Pyap to Overland Corner Floodplain Assessment Report. Thorburn PJ, Hatton TJ, Walker GR. 1993. Combining measurements of
Prepared for the Loxton to Bookpurnong Local Action Planning transpiration and stable isotopes of water to determine groundwater
Committee, Australian Water Environments Report No. 98027, discharge from forests. Journal of Hydrology 150: 563– 587.
Adelaide. Walker GR, Doble RC, Mech T, Lavis T, Bluml M, MacEwan R, Sten-
Barnett SR. SA Department of Mines and Energy. 1991. Renmark son M, Wang E, Jolly ID, Miles M, McEwan KM, Bryan B, Ward J,
Hydrogeological Map (1 : 250000 Scale), Bureau of Mineral Resources, Rassam D, Connor J, Smith C, Munday TJ, Nancarrow B, Williams S.
Geology and Geophysics: Canberra. 2005. Lower Murray Landscape Futures Phase One Report . CSIRO
Barnett SR, Yan Y, Watkins NR, Woods JA, Hyde KM. 2002. Murray Land and Water Final Year 1 Technical Report to the Centre
Darling Basin Salinity Audit: Groundwater Modelling to Predict Future for Natural Resource Management, the Victorian NAP Office and
Salt Loads to the River Murray in South Australia. Department for the CSIRO Water for a Healthy Country Flagship Program, Ade-
Water Resources Report DWR 2001/017, Adelaide. laide. http://www.clw.csiro.au/publications/consultancy/2005/LMLF
Doble RC, Simmons CT, Jolly ID, Walker GR. 2006. Spatial relation- Project Phase One report.pdf.
ships between vegetation cover and irrigation-induced groundwater
discharge on a semi-arid floodplain, Australia. Journal of Hydrology
329: 75–97.

Copyright  2009 John Wiley & Sons, Ltd. Hydrol. Process. 23, 3428– 3439 (2009)
DOI: 10.1002/hyp

You might also like