Introduction
Quantifying groundwater inflows to streams and rivers is
critical to understanding hydrogeological systems, protecting riverine
ecosystems, and managing water resources (e.g. Winter, 1999; Sophocleous,
2002; Brodie et al., 2007). Groundwater inflows may form the majority of
water in gaining rivers during periods of low streamflow, and riverine
ecosystems are commonly sustained by groundwater inflows at those times
(Kløve et al., 2011; Barron et al., 2012; Cartwright and Gilfedder, 2015).
Thus, understanding the distribution and magnitude of groundwater inflows is
important for managing and protecting these commonly vulnerable ecosystems.
Failure to understand groundwater contributions to rivers may also result in
the double allocation of water resources (i.e. surface water and groundwater
allocations might represent the same water). Documenting the distribution and
quantity of groundwater inflows to rivers is also required for flood
forecasting, understanding the impacts of contaminants on rivers, and
assessing the potential impacts of climate or land use changes on river
systems.
In many catchments globally there are insufficient groundwater bores to
understand the exchange between rivers and groundwater on anything other than
a regional scale. In these cases geochemical tracers provide an alternative
tool to understand groundwater–river interaction. Providing that groundwater
and surface water have significantly different geochemistry, changes in the
geochemistry of the river may be used to map and quantify groundwater inflows
(e.g. Cook, 2013). Tracers such as major ions, stable isotopes, radioactive
isotopes, and chlorofluorocarbons have been used to quantify groundwater
inflows to rivers (e.g. Ellins et al., 1990; Genereux and Hemond, 1992;
Négrel et al., 2001; Stellato et al., 2008; Cartwright et al., 2011,
2014; Cook, 2013; Bourke et al., 2014a, b). Geochemical tracers only quantify
groundwater inflows, and while they are commonly used to determine the
distribution of gaining and losing reaches, they do not quantify the
magnitude of any groundwater outflows.
River water also interacts with the sediments beneath and adjacent to the
streams in the hyporheic and parafluvial zones. The hyporheic zone comprises
the sediments of the streambed and sides through which the river water flows
due to irregularities in the streambed, and hyporheic flow generally occurs
on the centimetre to tens of centimetres scale (Boulton et al., 1998). In
rivers that have coarse-grained unconsolidated sediments on their floodplain,
metre to hundreds of metres scale parafluvial flow may also occur (Holmes et
al., 1994; Edwardson et al., 2003; Cartwright et al., 2014; Bourke et
al., 2014a; Briody et al., 2016). By contrast with hyporheic exchange that
occurs along the entire river, water enters the parafluvial zone in river reaches that are losing and then
re-enters the river where it is gaining, augmenting the groundwater inflows.
Both hyporheic exchange and parafluvial flow may impact the geochemistry of
the rivers (Boulton et al., 1998; Edwardson et al., 2003; Cook et al., 2006;
Cartwright et al., 2014; Bourke et al., 2014a; Briody et al., 2016) and must
be taken into account when using geochemical tracers to determine groundwater
inflows to rivers.
222Rn as a tracer of groundwater inflows
222Rn, which is an intermediate isotope in the 238U to 206Pb
decay series, is an important tracer for quantifying groundwater inflows to
rivers. 222Rn has a half-life of 3.8 days and the activity of 222Rn
reaches secular equilibrium with its parent isotope 226Ra over 3–4 weeks
(Cecil and Green, 2000). Because 226Ra activities in minerals in
the aquifer matrix are several orders of magnitude higher than those in
surface water, groundwater 222Rn activities are commonly 2 or 3
orders of magnitude higher than those of surface water (Cecil and Green,
2000). This makes 222Rn a viable tracer of groundwater inflows in
catchments where the groundwater has similar major ion concentrations and/or
stable isotope ratios to the river water. As 222Rn activities in rivers
decline downstream from regions of groundwater inflow due to radioactive
decay and degassing to the atmosphere (Ellins et al., 1990; Genereux and
Hemond, 1992), 222Rn is also useful in determining locations of
groundwater inflow, even if the inflows are not quantified.
The successful application of 222Rn to determine groundwater inflows,
however, requires careful consideration of several processes and
uncertainties. 222Rn activities in groundwater may be spatially or
temporally heterogeneous (Cook et al., 2006; Mullinger et al., 2007; Unland
et al., 2013; Yu et al., 2013; Cartwright et al., 2011; Atkinson et
al., 2015). Additionally, while it is well established that the rate of
222Rn degassing increases with increasing river turbulence and
decreasing river depth, it is difficult to reliably quantify the rate of
degassing (Genereux and Hemond, 1992; Mullinger et al., 2007; Cook, 2013;
Cartwright et al., 2014). Finally, in rivers that run through coarse alluvial
sediments, water from the hyporheic or parafluvial zones may provide a source
of additional 222Rn to groundwater inflow (Cook et al., 2006; Cartwright
et al., 2014; Bourke et al., 2014a). As has been outlined in several studies,
comparison of the calculated groundwater inflows from 222Rn with those
made from other geochemical tracers or with streamflow measurements is a
crucial test of the calculations (Cook et al., 2003, 2006; Mullinger et
al., 2007, 2009; Cartwright et al., 2011, 2014; McCallum et al., 2012; Unland
et al., 2013). Carrying out studies at baseflow conditions when most of the
water contributing to the streams is from groundwater inflows allows for a
comparison between the calculated groundwater inflows and the observed
increase in streamflows, which in turn provides for a test of the parameters
used in the 222Rn mass balance (Cartwright et al., 2014).
Objectives
This paper examines groundwater–river interaction in the Avon River,
southeast Australia, primarily using 222Rn as a tracer. The incised
nature of the Avon River and the fact that it rarely ceases to flow has led
to an assumption that it receives significant groundwater inflows (Gippsland
Water, 2012). There has been little attempt, however, to quantify groundwater
inflows or determine their distribution, and there are insufficient
groundwater monitoring bores in the catchment to understand the relationship
of groundwater to the river using hydraulic data. Understanding
groundwater–river interaction is required to protect and manage the Avon
River, especially in assessing the potential impacts of increased groundwater
or surface water use.
Summary geological and hydrogeological map of the Avon River
catchment (Hofmann and Cartwright, 2015; Department of Environment and
Primary Industries, 2015). Arrows show general direction of groundwater flow.
Main sampling sites (in order of distance downstream are) are BR: Browns,
WF: Wombat Flat, SM: Smyths Road, VA: Valencia, BP: Bushy Park, PL: Pearces
Lane, RL: Ridleys Lane, SC: Schools Lane, ST: Stewarts Lane, SA: Stratford,
KR: Knobs Reserve, RB: Redbank, CB: Chinns Bridge. Unnamed sampling sites are
the additional sites from February 2015 (Table A1).
The paper has two specific aims. Firstly, we use data from a 6-year period
to examine whether periodic major flooding events, which alter the geometry
of the Avon River floodplain, change the locations of groundwater inflows.
Understanding whether the locations of groundwater inflows change following
major flood events, and whether we can monitor those changes, is important
to understanding groundwater–river interactions. Secondly, we assess the
impacts of parafluvial exchange on the 222Rn budget. The Avon River
floodplain comprises coarse-grained unconsolidated alluvial sediments with
gravel banks, point bars, and pool and riffle sections that likely host
parafluvial flows. Rivers with similar coarse-grained sediments on their
floodplains are common at mountain fronts and parafluvial flow is likely to
be an important process in these settings. Despite parafluvial inflows being
a potential important contributor of 222Rn budget to rivers, few
studies have explicitly considered this process in the 222Rn mass
balance (e.g. Bourke et al., 2014a; Cartwright et al., 2014). Thus, the
results of this study will help improve the general utility of 222Rn as
a tracer of groundwater inflows into rivers.
The Avon catchment
The Avon River is an unregulated river in the Gippsland Basin of southeast
Australia (Fig. 1) that has a total catchment area of
∼ 1830 km2 (Cochrane et al., 1991; Department of Environment
and Primary Industries, 2015). It drains the southern slopes of the Victorian
Alps (maximum elevation in the catchment is 1634 m) and discharges
into Lake Wellington, which is a coastal saline lake connected to the
Southern Ocean. The highland areas represent ∼ 30 % of the Avon
catchment and are dominated by temperate native eucalyptus forest, whereas
the majority of the plains representing ∼ 70 % of the catchment
have been cleared for agriculture, which includes dairying, sheep grazing,
and vegetable production. The estimated population of the Avon catchment is
∼ 4000 with Stratford being the largest town (population ∼ 2000).
The highlands of the Victorian Alps comprise indurated Palaeozoic and
Mesozoic igneous rocks and metasediments that only host groundwater flow in
fractures or in near-surface weathered zones (Walker and Mollica, 1990;
Cochrane et al., 1991). These rocks form the basement to the Tertiary and
Quaternary sediments of the Gippsland Basin (Fig. 1). The shallowest regional
aquifer within the Avon catchment is the Pliocene to Pleistocene Haunted Hill
Formation which comprises up to 40 m of interbedded alluvial sands
and clays that have hydraulic conductivities between 10-7 and
10-5 ms-1 (Brumley et al., 1982; Walker and Mollica, 1990).
Quaternary sediments that consist of coarse-grained sand and gravels
interbedded with finer-grained silts occur mainly within the river valleys
and have hydraulic conductivities of 10-5 and
10-2 ms-1 (Brumley et al., 1982; Walker and Mollica, 1990).
Average rainfall within the Avon catchment ranges from
∼ 1.5 myr-1 in the highlands to
∼ 0.9 myr-1 on the plains with most precipitation
occurring in the austral winter (June–September) (Bureau of Meteorology,
2015). The Avon River displays strong seasonal flows with ∼ 80 % of
annual streamflow occurring during winter (Department of Environment and
Primary Industries, 2015). This study focusses on the reaches of the Avon
River located on the plains formed by the Gippsland Basin sediments that are
upstream of tidal influence. Streamflow is measured continuously at three
sites (the Channel, Stratford, and Chinns Bridge; Fig. 1). Total annual
streamflow at Stratford between 1977 and 2014 was between 1.3×107
and 9.0×108 m3yr-1 (median = 3.0×108 m3yr-1) and varied with total annual rainfall (Department
of Environment and Primary Industries, 2015). The Avon River only ceases to
flow during the summers of severe drought years (e.g. 1983) and experiences
periodic floods during high rainfall periods (Fig. 2). Streamflow generally
increases downstream at all times, except at very low flows when streamflow
decreases between Stratford and Chinns Bridge. Valencia Creek and Freestone
Creek are the main tributaries; both have streamflow measurements (Department
of Environment and Primary Industries, 2015) and enter the Avon in the upper
reaches of the studied section (Fig. 1).
The Avon River has incised through the Haunted Hill and Quaternary sediments
to create terraces that are up to 30 m high with a lower floodplain that is
up to 500 m wide. Where it crosses the sedimentary plains, the Avon River
comprises a sequence of slow-flowing pools that are typically 10–30 m wide,
up to 2 m deep at low flows, and up to 2 km long. These pools are connected
by shorter (typically ten to hundreds of metres long) and narrow (typically
< 5 m) faster-flowing riffle sections that commonly have steep
longitudinal gradients.
(a) Variation in streamflow at Stratford (Fig. 1) between
January 2009 and February 2015. The major floods (highlighted) caused
significant changes to the geometry of the floodplain. (b) Flow
frequency curve for Stratford for streamflows between January 2000 and March
2015 and the percentiles of discharge in the sampling campaigns (data from
Department of Environment and Primary Industries, 2015).
The floodplain of the Avon River between Browns (0.0 km) and Redbank
(41.3 km) (Fig. 1) comprises numerous gravel banks and point bars of
coarse-gained immature unconsolidated sediments with clasts of up to
50 cm in diameter. In regions where the river is incised, there are
seeps of water at the base of the slope and permanent patches of
water-tolerant vegetation. The alluvial sediments on the floodplain are
sparsely vegetated and the geometry of the floodplain changes markedly
following major flood events, such as those in 2011, 2012, and 2013 (Fig. 2).
These changes include the downstream migration of pools (often by several
tens of metres), scouring of the alluvial sediments, and changes to the
location of the sediment banks. Downstream of Redbank, the Avon River
occupies an incised channel with banks of finer-grained (clay- to sand-sized)
sediments. The banks and floodplain are more vegetated and do not change
markedly during the flood events.
Groundwater flows from the Victorian Alps to the coast (Hofmann and
Cartwright, 2013; Fig. 1). Use of water from the Avon River and its
tributaries for irrigation is up to 8×106 m3yr-1
(∼ 2.6 % of the median annual streamflow at Stratford); however, there is a prohibition on river
water use when the streamflow at Stratford is
< 104 m3day-1 (Gippsland Water, 2012).
Methods
Sampling
Sampling took place between February 2009 and February 2015 in six campaigns
at a variety of streamflows (Fig. 2a). These sampling campaigns were both
before and after four major flood events that occurred between 2011 and 2013
and which caused the redistribution of the position of pools and sediment
banks in the river. Each sampling campaign involved sampling the river sites
(Table A1, Fig. 1) over a 2- to 3-day period, with the February 2015
sampling campaign involving additional sites to the others. Distances are
measured relative to the first sampling site at Browns (0.0 km)
(Fig. 1). Streamflow is measured at three permanent gauging stations: the
Channel, which is close to the first sampling site at Browns; Stratford; and
Chinns Bridge (Department of Environment and Primary Industries, 2015;
Fig. 1). Streamflow was relatively constant during the sampling periods (the
variation in streamflow at Stratford over each sampling period was
< 5 %). River samples were collected from 0.5 to 1 m below the
river surface using a manual collector mounted on a pole. Groundwater was
sampled from bores installed on the river bank and floodplain at Stratford
and Pearces Lane (Fig. 1) that have 1 to 3 m long screens. Water was
extracted using an impeller pump set at the screened interval and at least
3-bore volumes of water were purged before sampling. Water was also extracted
from the alluvial gravels at a number of locations along the Avon River
during low flow periods either from open holes or from piezometers driven
1–2 m below the surface of the gravels.
Analytical techniques
Analytical techniques were similar to those in other studies (e.g. Unland et
al., 2013; Yu et al., 2013; Cartwright et al., 2014). Cations (Tables A1, A2)
were analysed on samples that had been filtered through 0.45 µm
cellulose nitrate filters and acidified to pH < 2 using a ThermoFinnigan
quadrupole ICP-MS at Monash University. Anions (Tables A1, A2) were analysed
on filtered unacidified samples using a Metrohm ion chromatograph at Monash
University. The precision of major ion concentrations based on replicate
analyses is 2–5 %. A suite of anions and cations were measured; however,
only Cl and Na are discussed in this study. 222Rn activities in
groundwater (Table A2) and surface water (Table A1) were determined using a
portable radon-in-air monitor (RAD-7, Durridge Co.) following methods
described by (Burnett and Dulaiova, 2006) and are expressed in becquerels per
m3 of water (Bqm-3). A sample of 0.5 L was
collected by bottom filling a glass flask and 222Rn was subsequently
degassed for 5 min into a closed air loop of known volume. Counting
times were 2 h for surface water and 20 min for groundwater.
Typical relative precision based on repeat sample measurements in this and
other studies (e.g. Cartwright et al., 2011, 2014) is < 3 % at
10 000 Bqm-3 and ∼ 10 % at 100 Bqm-3.
A total of 44 samples of riverbed sediments from sites along the Avon River
were collected in March 2014 and February 2015. 222Rn emanation rates
(γ) from these were determined by sealing a known dry weight of
sediment in airtight containers with water and allowing 222Rn to
accumulate (Lamontagne and Cook, 2007). Following incubation of 4–5 weeks, by which time the rate of
222Rn production and decay will have reached steady state,
20–40 mL of pore water was extracted and analysed for 222Rn
activities using the same method as above but with counting times of
6–12 h. γ (Table 2) was calculated from 222Rn produced
per unit mass of sediment Em, sediment density ρs,
and porosity φ by
γ=Em1-φρsλφ
(parameters summarized in Table 1).
Radon mass balance
Assuming that the atmosphere contains negligible radon, the change in
222Rn activities along a river is
Qdcrdx=Icgw-cr+wEcr+Fh+Fp-kdwcr-λdwcr
(modified from Mullinger et al., 2007; Cartwright et al., 2011; and Cook,
2013). In Eq. (2) Q is streamflow; cr and cgw are
the 222Rn activities in the river and groundwater, respectively; I is
the groundwater flux per unit length of river; E is the evaporation rate;
x is distance along the river; w is river width; d is river depth;
Fh and Fp are the inputs of 222Rn resulting from
exchange with the hyporheic zone and inflows of parafluvial waters,
respectively; k is the gas-transfer coefficient; and λ is the
decay constant (Table 1). A similar mass balance also applies to major ion
concentrations. Since the concentration of a conservative tracer such as Cl
is controlled only by groundwater inflows and evaporation, only the first two
terms on the right-hand side of Eq. (2) are relevant. If the river is gaining
throughout and solely fed by groundwater, the increase in streamflow
downstream is
dQdx=I-Ew.
The 222Rn activity in the hyporheic zone waters (ch) is
governed by the 222Rn activity of the water flowing into the hyporheic
zone (cin), the 222Rn emanation rate γ, and the
residence time th:
ch=γλ-cin1-e-λth+cin
(Hoehn et al., 1992; Hoehn and Cirpka, 2006) (Fig. 3a). An identical
expression relates the 222Rn activity in the parafluvial zone waters
(cp) to the residence time of that water in the parafluvial zone
(tp). ch increases with th until secular
equilibrium is approached at which point ch=γ/λ.
In a losing or neutral (i.e. neither gaining nor losing) river cin=cr. In a gaining river, water derived from the river will mix in
the alluvial sediments with upwelling regional groundwater that has high
222Rn activities. Cartwright et al. (2014) discussed using the
concentration of a conservative ion such as Cl to estimate the degree of
mixing within the alluvial sediments to estimate cin. Assuming that
all the water entering the hyporheic zone subsequently re-enters the river,
the 222Rn flux from the hyporheic zone (Fh) is
Fh=γAhφ1+λth-λAhφ1+λthcin,
where Ah is the cross-sectional area of the hyporheic zone
(Lamontagne and Cook, 2007). Equation (5) treats the hyporheic zone as a
homogeneous region adjacent to the river in which river water resides for a
certain period of time and then re-enters the river. While recognizing that
this is an oversimplification, it provides a means of calculating the changes
in 222Rn in the hyporheic zone from estimates of emanation rates and the
dimensions of the hyporheic zone.
Summary of parameters used in 222Rn mass balance.
Symbol
Parameter
Units
Comments
Q
Streamflow
m3day-1
E
Evaporation
mday-1
x
Distance downstream
m
w
Stream width
m
d
Stream depth
m
v
Stream velocity
mday-1
cgw, cr, ch, cp
222Rn activities in groundwater, river, hyporheic zone, parafluvial zone
Bqm-3
cin
222Rn activity of water entering the hyporheic or parafluvial zone
Bqm-3
k
Gas-transfer coefficient
day-1
λ
Decay constant
0.181 day-1
I
Groundwater inflows
m3m-1day-1
Eq. (2)
Fh
222Rn flux from hyporheic zone
Bqm-1day-1
Eq. (5)
Fp
222Rn flux from parafluvial zone
Bqm-1day-1
Eq. (6)
γ
222Rn emanation rate
Bqm-3day-1
Eq. (1)
Em
222Rn produced from sediments
Bqkg-1
ρs
Sediment density
kgm-3
Ip
Inflows from parafluvial zone
m3m-1day-1
th, tp
Residence time in hyporheic or parafluvial zone
day
φ
Porosity
Vp
Volume of sediments that parafluvial inflows interact with
m3m-1
Ah,Ap
Cross-sectional area of the hyporheic or parafluvial zone
m2
Ap=Vp
222Rn emanation rates from floodplain sediments.
Site*/Sample
Em
γ
γ/λ
(Bqkg-1)
(Bqm-3day-1)
(Bqm-3)
Chinns Bridge 1
2.01
1473
8138
Chinns Bridge 2
4.02
2949
16 293
Wombat Flat 1
4.04
2964
16 376
Wombat Flat 2
4.52
3311
18 295
Wombat Flat 3
4.19
3075
16 988
Wombat Flat 4
6.13
4492
24 819
Valencia 1
3.95
2899
16 016
Valencia 2
1.86
1362
7525
Pearces Lane 1
0.62
454
2506
Pearces Lane 2
3.25
2383
13 167
Pearces Lane 3
1.41
1034
5722
Pearces Lane 4
2.63
1925
10 636
Pearces Lane 5
6.76
4952
27 360
Pearces Lane 6
5.60
4107
22 689
Pearces Lane 7
4.12
3018
16 674
Pearces Lane 8
1.54
1127
6225
Stewarts Lane 1
3.41
2497
13 797
Stewarts Lane 2
5.78
4239
23 418
Stewarts Lane 3
3.08
2258
12 475
Stewarts Lane 4
2.88
2110
11 656
Stewarts Lane 5
4.63
3391
18 732
Stewarts Lane 6
3.64
2669
14 745
Stewarts Lane 7
4.52
3311
18 294
Stewarts Lane 8
4.58
3354
18 530
Stewarts Lane 9
1.96
1434
7925
Stewarts Lane 10
5.09
3733
20 622
Stewarts Lane 11
4.25
3119
17 230
Stewarts Lane 12
3.68
2699
14 910
Stewarts Lane 13
1.77
1294
7150
Stewarts Lane 14
2.89
2122
11 723
Stratford 1
2.13
1563
8634
Stratford 2
0.66
482
2663
Stratford 3
3.01
2206
12 190
Stratford 4
3.77
2762
15 259
Stratford 5
0.39
288
1591
Stratford 6
1.24
911
5032
Stratford 7
2.00
1469
8117
Stratford 8
2.71
1985
10 965
Stratford 9
0.91
668
3692
Stratford 10
1.01
738
4077
Stratford 11
4.55
3334
18 419
Stratford 12
3.13
2293
12 667
Stratford 13
0.81
491
3282
Mean
2308
12 751
σ
1197
6615
* sites in Fig. 2.
Equation (5) may also be used to calculate cp from tp
and γ (e.g. Cartwright et al., 2014). However, where parafluvial flow
involves long flow paths through alluvial sediments, an alternative
conceptualization is to consider the flux of 222Rn into the river at the
end of discrete flow paths through the parafluvial zone (Hoehn and Von
Gunten, 1989; Hoehn and Cirpka, 2006; Bourke et al., 2014a). In that case,
Fp is given by a similar expression to that which accounts for
the input of 222Rn due to groundwater inflows:
Fp=Ipcp-cr,
where Ip is the flux of water from the parafluvial zone per unit
length of the river. The minimum Ip required to produce a given
Fp is achieved when cp approaches steady state
(Fig. 3b), which requires th to be at least several days
(cp is ∼ 95 % of the steady state activity after
16 days; Fig. 3a). If th is less than the time required to
achieve steady state, cp is lower, and a higher Ip is
required to achieve the same Fp. The volume of sediments with
which the water has interacted during flow through the parafluvial zone
(Vp in m3 per meter of flow path length of river) is governed by Ip,
tp, and φ. If the flow paths through the parafluvial zone
are regular, Vp will be the cross-sectional area of the
parafluvial zone through which the water from the river flows
(Ap):
Vp=Ap=tpIpφ
(Bourke et al., 2014a). For the same input parameters, Eqs. (5) and (6) yield
closely similar estimates of Fp (Bourke et al., 2014a) and the
least well-known parameters in both cases are Ap and tp.
(a) Variation in the 222Rn activity in the parafluvial
or hyporheic zone (cp or ch) with residence time
(tp or th) and 222Rn emanation rate (γ)
(Eq. 3). (b) Variation in the water flux from the parafluvial zone
(Ip) with the flux of 222Rn from the parafluvial zone
(Fp) and tp (Eq. 5). In both cases cr=cin= 1000 Bqm-3.
There are several approaches that may be used to estimate the rate of
222Rn degassing from rivers. Firstly, as degassing involves diffusion of
222Rn through the boundary layer at the river surface, the stagnant film
model yields a gas transfer velocity as D/z (which is closely related to
k), where z is the thickness of the boundary layer at the water surface
(Ellins et al., 1990; Stellato et al., 2008). z and by extension D/z can
be calculated from differences in river 222Rn concentrations in losing
reaches. The gas transfer coefficient k may be estimated in a similar way
from the change in 222Rn activities in losing reaches (e.g. Cartwright
et al., 2011; Cook 2013) or even in gaining reaches if groundwater inflows
have been estimated using other tracers, numerical models, streamflow
measurements, and/or streambed temperature profiles (Cook et al., 2003;
Cartwright et al., 2014; Cartwright and Gilfedder, 2015). Determining k or
z by comparing calculated and measured 222Rn activities requires that
the 222Rn contributed from the hyporheic or parafluvial zones is
quantified, and that there are no inflows of water from tributaries that may
increase or decrease 222Rn activities. Since k values are typically
calculated from these methods for a few specific well-understood river
reaches, it is possible that they are not valid for all river reaches.
Downstream variations in 222Rn activities (a) and Cl
concentrations (b) for the six sampling campaigns (data from
Table A1, abbreviations are as for Fig. 2). Closed symbols for February 2015 are
from the main sites, open symbols are from the additional sites specific to
that sampling campaign (Table A1). Site abbreviations are as for Fig. 1.
It is also possible to measure k directly by using introduced gas tracers
such as SF6 (Cook et al., 2003, 2006; McCallum et al., 2012; Bourke et
al., 2014a), which has the advantage of estimating k for the river being
studied. However, such measurements are generally made along small reaches of
a river that may not be representative of the river as a whole. Additionally,
if the experiments were made at specific flow conditions, the gas transfer
coefficients may or may not be applicable to sampling campaigns made at
different flow conditions.
There are several empirical relationships that estimate k from river
velocities (v) and depths. The commonly used O'Connor and Dobbins (1958)
and Negulescu and Rojanski (1969) gas transfer equations as modified for
222Rn are
k=9.301×10-3v0.5d1.5,k=4.87×10-4vd0.85
(Mullinger et al., 2007). As discussed by Genereux and Hemond (1992),
however, there are numerous formulations that can yield very different
estimates of k for the same flow conditions and some independent assessment
of k (for example, by matching the predicted and observed decline in
222Rn activities in losing reaches) is needed.
Results
Streamflow
Between January 2000 and February 2015 streamflow at Stratford varied between
500 and 1.38×108 m3day-1 (Department of Environment
and Primary Industries, 2015). Despite this period including years with well
below average rainfall (for example, in 2006 when rainfall was
∼ 50 % of the long-term average; Bureau of Meteorology, 2015),
there were no periods of zero streamflow. Mean daily streamflows at Stratford
during the sampling rounds ranged from 10 670 to
88 800 m3day-1 (Table A1, Fig. 2a) which represent streamflow
percentiles of 39.5–89.9 (Fig. 2b). In February 2015, which is the sampling
round discussed in most detail below, the mean daily streamflow was
12 510 m3day-1 at the Channel, 23 090 m3day-1
at Stratford, and 25 780 m3day-1 at Chinns Bridge. Inflows
from Valencia Creek and Freestone Creek in February 2015 were 1410 and
600 m3day-1, respectively (Department of Environment and
Primary Industries, 2015).
River geochemistry
Figure 4a shows the 222Rn activities of the Avon River for the six
sampling campaigns. There are several distinct zones of elevated 222Rn
activities, notably at Wombat Flat (4.8 km) where 222Rn
activities are up to 2040 Bqm-3 and between Bushy Park and
Schools Lane (16.3–25.3 km) where 222Rn activities are up to
3690 Bqm-3. Zones of lower 222Rn activities in the upper
reaches occur at Smyths Road (8.1 km) and in the reaches between
Stewarts Lane and Stratford (30.1–35.1 km). The downstream river
reaches between Knobs Reserve and Chinns Bridge (37.8–49.7 km) also
have relatively low 222Rn activities that generally decline downstream.
The position of the highest 222Rn activities changed in the periods
prior to and after the 2011–2013 floods. In March 2014 and February 2015,
the highest 222Rn activities were at Bushy Park (16.3 km),
whereas this site had relatively low 222Rn activities in February 2009
and April 2010 when the highest 222Rn activities were at Pearces Lane
(20.0 km). The distribution of 222Rn activities in the detailed
sampling campaign in February 2015 is similar to that at other periods of low
to moderate streamflow (e.g. March 2014). The lowest overall 222Rn
activities were recorded during the periods of highest flow (September 2010
and July 2014).
EC values and Cl concentrations generally increase downstream from 54 to
131 µScm-1 and from 4 to 10 mgL-1 at Browns
(0.0 km) to as high as 934 µScm-1 and
98 mgL-1 at Chinns Bridge (49.7 km) (Table A1, Fig. 4b).
Cl concentrations at low streamflows in March 2014 were generally higher (up
to 98 mgL-1) than in the other sampling campaigns, while Cl
concentrations were < 20 mgL-1 during the highest streamflows
in September 2010. A marked increase in EC values and Cl concentrations
occurs downstream of Smyths Road (8.1 km) in the reaches where
222Rn activities are highest at low streamflows. The concentrations of
other major ions (e.g. Na) increase downstream in a similar manner
(Table A1).
Groundwater geochemistry
Groundwater from the near-river bores at Pearces Lane and Stratford has
222Rn activities that vary from 480 to 28 980 Bqm-3
(Table A2). There is some variation in 222Rn activities in individual
bores between the sampling rounds with relative standard deviations between 6
and 34 %. The mean value of all groundwater 222Rn activities (n=26) is 12 890 Bqm-3. Bore 5 at Pearces Lane is immediately
adjacent to the Avon River and possibly samples water from the parafluvial
zone rather than groundwater. Excluding data from that bore, the mean value
of 222Rn activities is 13 830 Bqm-3 (n=24) with a
standard error of 1273 Bqm-3 and a 95 % confidence interval
(calculated using the descriptive statistics tool in Excel 2010 which assumes
that the data follow a t distribution) of 2634 Bqm-3. EC
values of groundwater from the bores at Pearces Lane and Stratford are
between 100 and 680 µScm-1 and Cl concentrations range from
46 to 147 mgL-1 with a mean value of 79±34 mgL-1 (n=16) (Table A2). If Bore 5 at Pearces Lane is
again excluded, the mean Cl concentration is 87±28 mgL-1 (n=14) with a standard error of 8 mgL-1 and a 95 %
confidence interval of 16 mgL-1. These Cl concentrations are
typical of groundwater elsewhere in the Avon valley and neighbouring
catchments (Department of Environment and Primary Industries, 2015).
Geochemistry of water from the alluvial gravels
EC values of water within the gravels further than 1–2 m from the
edge of the river are between 120 and 550 µScm-1 (n=52)
(Fig. 5b). These EC values are higher than those of the adjacent river water
but similar to those of the groundwater. Only water extracted from within 1
to 2 m from the river had EC values similar to the river and in some
cases the EC of water from the gravels within a few centimetres of the river
edge was higher than the adjacent river. 222Rn activities of these
samples were between 7000 and 28 000 Bqm-3 (n=21)
(Fig. 5a), which are also significantly higher than the 222Rn activities
in the adjacent river. As discussed below, these data are interpreted as
indicating that the gravels contain a mixture of groundwater and parafluvial
water.
Variations in 222Rn activities (a) and EC
values (b) of water extracted from river bank gravels. Shaded boxes
show range of values in the groundwater (excluding Bore 5 at Pieces Lane) and
the Avon River (data from Tables A1 and A2).
222Rn emanation rates
222Rn emanation rates were determined via Eq. (1). The matrix density
was assigned as 2700 kgm-3, which is appropriate for sediments
rich in quartz (ρ= 2650 kgm-3), and a porosity of 0.4 was
used, which is appropriate for unconsolidated poorly sorted riverine
sediments (Freeze and Cherry, 1979). γ values range from 288 to
4950 Bqm-3 with a mean value of 2308±1197 Bqm-3 (n=44) and a standard error of
183 Bqm-3. The mean emanation rates for sediments from the
different sites vary between 1484 and 3461 Bqm-3; however, there
is no systematic variation with position in the catchment. The relative
variability in γ between the sediments is similar to that reported
elsewhere (e.g. Bourke et al., 2014a; Cartwright et al., 2014). 222Rn
activities of water in equilibrium with the sediments are given by γ/λ (Cecil and Green, 2000), and the mean γ/λ value is
12 751±6615 Bqm-3 with a standard error of
1009 Bqm-3. These γ/λ values are not significantly
different (p∼ 0.5) to the measured 222Rn activities of the
groundwater.
Discussion
The following observations imply that overall the Avon is a gaining river:
(1) even during periods of prolonged low rainfall the river continues to flow
and streamflow commonly increases between the Channel and Chinns Bridge
gauges; (2) 222Rn activities are higher than those that could be
maintained by hyporheic exchange alone (Cartwright et al., 2011; Cook, 2013);
(3) Cl concentrations increase downstream; and (4) there are seeps of water
(presumed to be groundwater) at the base of steep slopes at the edge of the
floodplain. In the following section, the 222Rn activities and Cl
concentrations are used to assess the location and magnitude of groundwater
inflows.
Distribution of groundwater inflows
The February 2009, April 2010, March 2014, and February 2015 sampling
campaigns represent lower streamflows. Because the majority of water in the
Avon River at these times is likely to be provided by groundwater, the
222Rn activities from these sampling campaigns are most useful in
understanding the distribution of groundwater inflows. The region between
Smyths Road and Ridleys Lane (8.1–23.0 km) where 222Rn
activities increase and remain high (Fig. 4a), especially at lower
streamflows, and where there is a marked increase in Cl concentrations
(Fig. 4b), is interpreted as receiving major groundwater inflows. This section
of the Avon River is incised up to 4 m below the floodplain which
likely produces steep hydraulic head gradients that result in groundwater
discharge on the floodplain and into the river. There are also groundwater
seeps and patches of perennial water-tolerant vegetation at the edge of the
floodplain in this area. The reaches between Browns and Wombat Flat
(0.0–4.8 km) and Stewarts Lane and Stratford (30.1–35.1 km)
are also characterized by high 222Rn activities and are again
interpreted as receiving groundwater inflows.
The reaches between Wombat Flat and Smyths Road (4.8–8.1 km),
Ridleys Lane and Stewarts Lane (23.0–30.1 km), and Knobs Reserve and Chinns
Bridge (37.8–49.7 km) where there is a gradual decline in 222Rn
activities and little change in Cl concentrations (Fig. 4) are interpreted as
either being losing or receiving minor groundwater inflows. The landscape is
flatter and the river is less incised in these areas which results in lower
hydraulic gradients and consequently less groundwater inflows to the river.
The difference in the location of the highest 222Rn activities between
the sampling campaigns that were conducted before and after the major floods
(i.e. pre-2011 vs. post-2013) indicates that the locations of groundwater
inflows changed. The major floods changed the location of pools and sediment
banks on the Avon River and caused scouring, which would change the
relationship of the river to the groundwater.
Quantifying groundwater inflows
This section concentrates on modelling the 222Rn activities for the
detailed February 2015 sampling campaign (Fig. 4a). It was considered that
groundwater inflows, hyporheic exchange, and parafluvial flow all contributed
222Rn to the river. The groundwater 222Rn activity was assumed to
be 13 000 Bqm-3, which is consistent both with the measured
222Rn activities of groundwater (Table A2) and the calculated 222Rn
activities of water in equilibrium with the alluvial sediments.
The flux of 222Rn from the hyporheic zone was estimated from Eq. (5)
using the mean γ value of 2300 Bqm-3day-1 (Table 2),
a porosity of 0.4 (which is appropriate for coarse-grained unconsolidated
sediments), and a value for cin that is the 222Rn activity of
the river in that reach. The residence time of water within the hyporheic
zone is likely to be short (Boulton et al., 1998; Tonina and Buffington,
2011; Zarnetske et al., 2011; Cartwright et al., 2014), and th= 0.1 days is assumed here; for th< 1 day, Fh is
relatively insensitive to the actual residence times in the hyporheic zone
(Lamontagne and Cook, 2007; Cartwright et al., 2014). The width of the
hyporheic zone has been assigned as the river width. The thickness of the
hyporheic zone is less well known; however, by analogy with rivers elsewhere,
it is likely to be a few centimetres thick (Boulton et al., 1998; Hester and
Doyle, 2008; Tonina and Buffington, 2011) and a value of 10 cm is
initially adopted.
Parafluvial flow is conceived to occur on the tens of metres to kilometre
scale and to represent water that is lost from the river into the floodplain
sediments that subsequently re-enters the river downstream. The Cl and
222Rn data from the water contained within the gravels (Fig. 5) are
interpreted as reflecting mixing of groundwater and parafluvial flows in the coarse-grained gravel. The generation of paraluvial flow requires
that the river is locally losing. As discussed above, on the kilometre scale
the Avon River may contain losing reaches. Additionally, the reaches that are
interpreted as being overall gaining may contain smaller sections that are
losing. In particular, the riffle sections commonly have steep longitudinal
gradients and may transition from losing at the upstream end to gaining at
the downstream end. Parafluvial flow is probably hosted mainly within the
coarser-grained alluvial sediments (although conceivably it could also
include water that flows through the upper levels of the aquifers underlying
the alluvial sediments). By contrast with hyporheic exchange which occurs
along all reaches (whether gaining or losing), inflows from the parafluvial
zone require upward head gradients and only occur where the river is gaining.
The parafluvial inflows will increase the 222Rn activities in the river
in a similar manner to inflowing groundwater. However, because it represents
water that originated from the river, the inflows from the parafluvial zone
do not increase the overall streamflow. If the parafluvial zone water is in
secular equilibrium with the sediments,
cp ∼ 12 700 Bqm-3 (Table 2).
Average evaporation rates in southeast Australia in February to April are
3×10-3 to 5×10-3 mday-1 (Bureau of
Meteorology, 2015) and a value of 4×10-3 mday-1 was
adopted. Average river width and depth is 10 and 0.5 m, respectively,
upstream of Wombat Flat (0.0–4.8 km) and 20 and 1 m,
respectively, for the rest of the river
The gas transfer coefficient was estimated from the decline in 222Rn
activities between Ridleys Lane and Schools Lane (23.0–25.3 km)
(Fig. 4a). This approach estimates the net kdwcr term and k was
estimated as 0.3 day-1 using the measured widths, depths, and 222Rn
concentrations. This requires that this is a losing stretch of the river, so
that there are no groundwater or parafluvial inflows. That Cl concentrations
do not increase over this stretch of river (Fig. 4b) are consistent with it
being losing. A k value of 0.3 day-1 is at the lower end of estimates
of Rn gas transfer coefficients (Genereux and Hemond, 1992; Cook et
al., 2003, 2006; Cartwright et al., 2011, 2014; Unland et al., 2013; Yu et
al., 2013; Atkinson et al., 2015). However, as the Avon River is dominated by
slow-flowing pools, degassing rates are expected to be low.
(a) Calculated and observed 222Rn activities for
February 2015 resulting from assigning 50 % of the calculated inflows as
parafluvial flow. (b) Variation in groundwater and parafluvial
inflows. (c) Calculated streamflow resulting from the groundwater
inflows (Eq. 2) vs. measured streamflow at Stratford and Chinns Bridge.
(d) Predicted vs. observed Cl concentrations. Shaded field is the
range resulting from varying groundwater Cl concentrations within the
95 % confidence interval.
(a) Calculated vs. observed 222Rn activities in the
Avon River for February 2015 assuming both uniform groundwater inflow within
each reach and the situation where groundwater inflow occurs immediately
upstream of the measurement point. Site abbreviations are as for Fig. 2.
(b) Calculated groundwater inflows (I) assuming uniform inflows
within each reach. (c) Calculated increase in streamflow from
groundwater inflows (Eq. 2). Both uniform groundwater inflow within each
reach and the situation where groundwater enters the river immediately
upstream of the measurement point overestimate the measured streamflow.
Shaded area is the range of streamflow resulting from varying cgw
within the 95 % confidence interval. (d) Predicted vs. observed
Cl concentrations. Shaded field is the range resulting from varying
groundwater Cl concentrations within the 95 % confidence interval.
Groundwater inflows were calculated from the 222Rn activities by solving
Eq. (2) using a finite difference approach in a spreadsheet with a distance
step of 10 m (the use of smaller or larger distance steps does not
significantly change the results). The streamflow at the Channel gauge was
used as the initial streamflow and Q was increased after each distance step
via Eq. (3). The calculations estimated the values of I and Ip
in each reach by matching the calculated and measured 222Rn activities
along the river with the additional constraint that the total groundwater
inflows cannot exceed the net increase in streamflow between the Channel
gauge and the gauges at Stratford and Chinns Bridge (Fig. 1). Since there are
few streamflow measurements, the calculations assumed that the ratio of I
to Ip was the same in all gaining reaches of the river.
Assuming that in the gaining reaches there are 50 % parafluvial inflows
and 50 % groundwater inflows allows both the 222Rn variations and
the increase in streamflow to be accounted for (Fig. 6a). Calculated
groundwater and parafluvial inflows are highest in the reaches between Smyths
Road and Pearces Lane (8.1–20.0 km) (Fig. 6b), which is the region
where Cl concentrations also increase markedly (Fig. 4b). Assuming that the
waters are in secular equilibrium with the sediments, the combined inflows of
groundwater and parafluvial water for this reach are up
2.5 m3m-1day-1 of which groundwater inflows are
∼ 1.26 m3m-1day-1.
There is no process in the parafluvial or hyporheic zones other than mixing
with groundwater that increases the Cl concentrations of the through-flowing
water. Thus, the Cl concentrations in the river reflect only the groundwater
inflows and in theory it would be possible to use Cl to quantify these (c.f.,
McCallum et al., 2012). However, the high variability of Cl concentrations in
the groundwater and the relatively small difference between groundwater and
river Cl concentrations results in large uncertainties. The change in Cl
concentrations (Fig. 6d) was calculated from the groundwater inflows assuming
that groundwater has a Cl concentration of 85 mgL-1. The
calculated Cl concentrations are slightly higher than those observed, but if
the Cl concentration of the groundwater is allowed to vary within the
95 % confidence interval (±16 mgL-1) the observed trend
can be reproduced.
If residence times in the parafluvial zone are shorter than those required to
attain secular equilibrium, cp will be lower and the inflows from
the parafluvial zone (Ip) required to produce a given flux of
222Rn (Fp) increases (Fig. 3). For example, if cr= 2300 Bqm-3, which is a typical value in many of the reaches
between Valencia and Bushy Park (10.9–16.3 km) and
cp= 12 700 Bqm-3, then (cp-cr)= 10 400 Bqm-3. If Ip= 1 m3m-1day-1, Fp= 10 400 Bqm-1day-1 (Eq. 6). If γ= 2300 Bqm-3day-1, cp is 2487, 4023, and
11 004 Bqm-3 where tp is 0.1, 1, and 10 days,
respectively. To produce a value of Fp of
10 400 Bqm-1day-1 requires
Ip ∼ 58 m3m-1day-1 for tp= 0.1 days, ∼ 6.0 m3m-1day-1 for tp= 1 day, and ∼ 1.2 m3m-1day-1 for tp= 10 days. For tp > 30 days the system is close to secular
equilibrium and cpand Ip are near constant (Fig. 3).
The cross-sectional area of the parafluvial zone Ap required to
accommodate these parafluvial flows with φ= 0.4 and tp
between 0.1 and 100 days is between 14 and 250 m2 (Eq. 7). The
floodplain of the Avon River is tens of metres wide with sediment thicknesses
of several metres and even the higher estimates of the cross-sectional area
are not unreasonable given the volume of gravels on the floodplain.
Calculated and observed 222Rn activities for February 2015
resulting from varying individual parameters in Eq. (1). In all cases the new
parameters result in significant overestimation of 222Rn activities in
many reaches and are unlikely to be realistic. Site abbreviations are as for
Fig. 1.
Uncertainties and sensitivity
The proposal that parafluvial flow is important in the Avon River is
consistent with the local hydrogeology and allows both the 222Rn and net
increase in streamflow to be reproduced. The conclusion that inflows of
parafluvial zone waters only occur in the gaining reaches is justifiable as
the conditions required for groundwater inflows (gaining river with steep
hydraulic gradients and high-hydraulic conductivity sediments) will likely
drive the return of parafluvial waters to the river. By contrast, losing
reaches are likely to be where the water enters the parafluvial sediments.
Given the multiple parameters in Eq. (2) and their inherent uncertainties,
however, consideration needs to be given to whether both the 222Rn
activities and the increases in streamflow can be accounted for without
parafluvial inflows being a significant source of 222Rn.
Matching the 222Rn profile along the Avon River using the parameters
discussed above but without input of 222Rn from parafluvial zone would
imply net groundwater inflows of
28 300 m3day-1. However, these inflows exceed the measured
increase in streamflow between the Channel and Chinns Bridge of
15 500 m3day-1 by 180 % (Fig. 7a). The February 2015
sampling round took place at the end of summer when the small ephemeral
tributaries were dry and there was no overland flow; however, there were
still flows from Valencia Creek and Freestone Creek of 1410 and
200 m3day-1, respectively. If these were included, the
discrepancy between the calculated and observed streamflow increases. The
calculated Cl concentrations are also higher than observed (Fig. 7d),
although given the uncertainty in groundwater Cl
concentrations the discrepancy is not large.
In common with most studies, the calculations assumed that the groundwater
inflows are uniform along a particular reach. However, because 222Rn is
lost from rivers by degassing and decay, lower groundwater inflows are
required to replicate the observed 222Rn activities if the groundwater
inflows occur immediately upstream of a sampling point (Cook, 2013). Even
assigning the groundwater inflows in each reach to the 10 m section upstream
of the measurement point still results in the calculated streamflow
overestimating the measured streamflow (Fig. 7c). The predicted 222Rn
activities in the river in this case are also not realistic (Fig. 7a).
The evaporation term in Eq. (2) is 1–2 orders of magnitude lower than
most of the other terms, and errors in the assumed evaporation rate have
little influence on the calculations. The main parameter impacting calculated
groundwater inflows is the 222Rn activity of groundwater (Cartwright et
al., 2011; Cook, 2013). Allowing cgw to vary within the 95 %
confidence interval of the groundwater 222Rn activities
(±2600 Bqm-3) makes little difference to the discrepancy
between the calculated and observed increase in streamflow (Fig. 7c).
Increasing cgw to 27 000 Bqm-3 allows both the
222Rn profile and the observed increase in streamflow between the
Channel and Chinns Bridge to be reproduced without the requirement for the
input of 222Rn from the parafluvial zone (Fig. 8). However, there is no
known groundwater in the Avon catchment with such high 222Rn activities
and these activities are far higher than would be in equilibrium with the
alluvial sediments that comprise the near-river aquifer lithologies. Hence,
it is considered not possible that groundwater 222Rn activities could be
this high.
Calculated streamflows resulting from groundwater inflows for the
sampling rounds excluding February 2015 estimated without parafluvial flow.
Aside from the high flow periods (September 2010 and July 2014) the
calculated increase in streamflow exceeds the observed streamflow at
Stratford and Chinns Bridge. Site abbreviations are as for Fig. 1.
There is uncertainty in the gas transfer coefficient. k was estimated
assuming that the Avon River contains losing reaches; if those reaches were
actually gaining then this methodology underestimates k. However,
increasing k from 0.3 day-1 increases the calculated groundwater
inflows, which increases the discrepancy between the observed and calculated
increases in streamflow. k estimated from Eqs. (8) and (9) ranges between
0.1 and 0.3 day-1. Using k= 0.1 day-1 produces net groundwater
inflows that more closely match the observed increase in streamflow. However,
adopting k= 0.1 day1 results in the calculated 222Rn
activities in a number of reaches being overestimated (Fig. 8). This is
because even assuming no groundwater inflows into those reaches, the loss of
222Rn by degassing is insufficient to explain the observed decrease in
222Rn. Such a poor correspondence between predicted and observed
222Rn activities implies problems with the adopted variables.
While there are uncertainties in ch, the main uncertainty in the
contribution of hyporheic exchange to the 222Rn budget is the dimensions
of the hyporheic zone. Increasing Fh also reduces the calculated
groundwater inflows. Using the same emanation rates, residence times, and
porosities but assigning a thickness of the hyporheic zone of 50 cm
increases Fh and produces groundwater inflows that broadly match
the increase in streamflow. However, the higher values of Fh
again result in a poor fit between predicted and observed 222Rn
activities (Fig. 8).
Because the error in λ is negligible and the evaporation term is much
smaller than the other terms, it is generally possible to produce identical
trends in 222Rn activities with different combinations of k and
Fh (Cartwright et al., 2014). If Fh is calculated
assuming a 50 cm thick hyporheic zone, adopting k= 0.6 day-1
reproduces the observed 222Rn activities. Similarly, if k= 0.1 day-1 a match between the observed and the predicted 222Rn
activities is achieved with no hyporheic exchange (Fh= 0).
However, these combinations of parameters again result in estimated net
groundwater inflows that exceed the measured increase in streamflow.
There is an unknown error in the streamflow measurements, but it is unlikely
to be sufficient to explain the gross overestimation of groundwater inflows.
Uncertainties in the assumed river widths and depths will also impact the
calculations. Specifically, reducing the width or depth decreases the
magnitude of the last two terms on the right-hand side of Eq. (2), which in
turn reduces I. If widths were reduced by 50 % (an unrealistic error),
net groundwater inflows broadly match the increase in streamflow. However,
this again results in 222Rn activities being overestimated in many
reaches (Fig. 8). Increasing k to 0.65 day-1 would allow the
222Rn activities to be predicted using these lower widths but again
results in the estimated net groundwater inflow exceeding the measured
increase in streamflow. Overall it is concluded that there are no
combinations
of parameters that can reproduce both the observed 222Rn activities and
streamflows without incorporating parafluvial flow.
It would be possible to explain the observed 222Rn activities and
streamflows if there were losing reaches in the Avon River through which
significant volumes of river water were lost to the underlying aquifers and,
unlike parafluvial flow, this water did not subsequently return to the river.
For this scenario to be valid, approximately 50 % of the groundwater
inflows would have to be lost from the river in these losing reaches in
February 2015. The reaches between 25 are 30 km are interpreted as
losing. However, these reaches do not dry up even during prolonged drought
(Gippsland Water, 2012), and all reaches of the river were flowing during the
2009 sampling campaign (which had the lowest streamflows). Also while
streamflows were not measured, such a major reduction in streamflow over such
a short distance would be apparent in the field. Likewise, significant
pumping of water from the river would also reduce streamflows. While the
surface water is licensed for use, streamflow during February 2009 and March
2014 was below the minimum levels where that is permitted and the streamflows
in April 2010 and February 2015 were such that use would be restricted;
hence, large-scale pumping of river water at those times is unlikely.
Other sampling campaigns
The predicted distribution of groundwater inflows in February 2009, April
2010, and March 2014 when streamflows were low to moderate are similar to
those in February 2015 (Fig. 4). Due to the lower number of sampling points,
it is difficult to calculate groundwater inflows with certainty. The net
groundwater inflows calculated using the same parameters as above but
ignoring parafluvial flows are between 15 900 and
21 700 m3day-1, respectively (Fig. 9), which are up to
490 % of the measured increases in streamflow between the Channel and
Chinns Bridge. Again propagating the likely uncertainties in the parameters
through Eq. (2) cannot resolve this discrepancy, implying that the inflows of
water from the parafluvial zone must be a significant part of the 222Rn
budget.
At the higher streamflows there will likely be significant inputs to the
river from overland flow or interflow; hence, it is not possible to use the
comparison between calculated groundwater inflows and the net increase in
streamflow to independently test for the input of 222Rn from the
parafluvial zone. For example, without incorporating parafluvial flow, the
net groundwater inflows using widths of 15 m upstream of Wombat Flat and
25 m elsewhere, depths of 1.25 upstream of Wombat Flat and
1.6 m elsewhere, k= 0.3 day-1, Fh adjusted for
the higher river widths is 32 100 m3day-1 (September 2010)
and 44 600 m3day-1 (July 2014). These net groundwater inflows
are lower than the measured increases in streamflow between the Channel and
Stratford or Chinns Bridge (Fig. 9). However, it is likely that significant
parafluvial flow occurs at those times and consequently that these values
also represent an overestimation of the actual groundwater inflows.
Conclusions
The variation in 222Rn activities and Cl concentrations
clearly define the reaches of the Avon River that are gaining. The
distribution of 222Rn activities also indicate that the location of
groundwater inflows changed after major floods that occurred between 2011 and
2013. This approach can be applied to other rivers where flood events change
the geometry of the floodplain sediments and where the groundwater monitoring
bore network is insufficient to define groundwater–river interaction.
The Avon River has coarse-grained unconsolidated gravels along its floodplain
and it was concluded that parafluvial flow was a significant process in
controlling the 222Rn activities of the river. However, this proposition
is difficult to definitively test or explore in more detail. The groundwater
and parafluvial inflows have been assumed to occur in similar proportions in
each reach, which may not necessarily be the case. Parafluvial flow is likely
to be important in rivers with coarse-grained alluvial sediments on their
floodplains, especially where there are locally alternating gaining and
losing reaches, and must be taken into account in 222Rn mass balance
calculations. Unlike hyporheic exchange, which occurs in all stretches,
parafluvial inflows are likely to dominantly occur in gaining reaches
augmenting the groundwater inflows.
Theoretically, a conservative tracer such as Cl that is unaffected by
parafluvial flow could be used to separate groundwater inflows from
parafluvial inflows. However, the relatively high variability of groundwater
Cl concentrations and the relative small difference between groundwater and
river Cl concentrations make this impractical in the Avon catchment.
Nevertheless, this may be possible in other river catchments and illustrates
the advantage of using multiple geochemical tracers.
More generally, this study illustrates the importance of carrying out
geochemical studies at low streamflows where the majority of inflows into the
river are likely to be from groundwater. While this might appear redundant in
terms of determining the water balance, it does provide for a test of
assumptions and parameterization. It would be possible to interpret the
changes to 222Rn activities during the periods of higher streamflow as
being solely due to groundwater inflows because the net groundwater inflows
are lower than the net increases in streamflow (Fig. 9). However, it is
likely that there is groundwater and parafluvial inflows at all times, in
which case calculated groundwater inflows will be overestimated.