CSIRO PUBLISHING
Marine and Freshwater Research, 2010, 61, 842–856
www.publish.csiro.au/journals/mfr
Multi-scale environmental factors explain fish losses and
refuge quality in drying waterholes of Cooper Creek,
an Australian arid-zone river
Angela H. ArthingtonA,D, Julian D. OldenB, Stephen R. BalcombeA
and Martin C. ThomsC
A
Australian Rivers Institute and eWater Cooperative Research Centre, Griffith University,
Nathan, Qld 4111, Australia.
B
School of Aquatic and Fishery Sciences, University of Washington, Seattle, WA 98195, USA.
C
University of New England, Armidale, NSW 2351, Australia.
D
Corresponding author. Email: [email protected]
Abstract. Many rivers experience intermittent flows naturally or as a consequence of water abstraction. Climate change
is likely to exacerbate flow variability such that dry spells may become more common. It is important to understand the
ecological consequences of flow intermittency and habitat fragmentation in rivers, and to identify and protect habitat
patches that provide refugia for aquatic biota. This paper explores environmental factors influencing dry season fish losses
from isolated waterbodies in Cooper Creek, an unregulated arid-zone river in the Lake Eyre Basin, Australia. Multivariate
ordination techniques and classification and regression trees (CART) were used to decompose species–environment
relationships into a hierarchically structured data set, and to determine factors explaining changes in fish assemblage
structure and species losses over a single dry season. Canonical correspondence analysis (CCA) explained 74% of fish
assemblage change in terms of waterhole morphology (wetted perimeter, depth), habitat structure (bench development,
off-take channels), waterhole quality (eroded banks, gross primary production), the size of surrounding floodplains and
the relative isolation of waterholes. Classification trees for endemic and restricted species reaffirmed the importance of
these waterhole and floodplain variables as drivers of fish losses. The CCA and CART models offer valuable tools for
identification of refugia in Cooper Creek and, possibly, other dryland rivers.
Additional keywords: drought, dryland rivers, floodplains, habitat structure, spatial scale.
Introduction
Many of the world’s rivers are increasingly threatened by
anthropogenic pressures that impinge on their hydrologic
regimes and disrupt ecological processes dependent on naturally
dynamic wetting and drying cycles (Poff et al. 1997; Puckridge
et al. 1998; Nilsson et al. 2005). Climate change is likely to
further disrupt hydrologic processes in many rivers, such that
riverine habitats may become increasingly fragmented and
stressful for aquatic life during extended dry periods (Wishart
2006). Current knowledge about the effects of flow intermittency is fragmentary and no clear framework or principles
exist to support the management of intermittent streams and
arid-zone rivers. It is important to understand ecological
responses to intermittent flows, habitat fragmentation and
habitat isolation in all rivers, and to be able to identify and
protect habitat patches that are of sufficient quality to act as
refugia for aquatic biota in fragmented riverine landscapes.
Arid-zone rivers such as Cooper Creek in the Lake Eyre Basin,
Australia, belong to a class of perched or semi-perched alluvial
rivers that experience flow intermittence as a consequence of
Ó CSIRO 2010
evaporative and transmission losses, and depletion of bank
storage and the floodplain aquifer (Larned et al. 2010). Rivers
draining to Lake Eyre, such as Cooper Creek, experience infrequent high-magnitude flow events (linked to ENSO phenomena)
that can inundate extensive areas of floodplain, often for several
months at a time (Knighton and Nanson 1994a, 2001). After flood
recession, Cooper Creek dries down into a complex system of
anastomosing channels (the ‘channel country’) and several hundred relatively deep channel segments known locally as ‘waterholes’ (Knighton and Nanson 1994b; Davis et al. 2002). These
waterbodies can be likened to the isolated floodplain lakes and
lagoons left in tropical floodplain river systems after floods recede
(Rodriguez and Lewis 1997). Isolated waterholes patchily distributed along Cooper Creek act as refugia for fish and other
obligate aquatic species (cf. Morton et al. 1995; Sheldon et al.
2002; Magoulick and Kobza 2003). These refuge habitats can
support up to 12 native fish species of ecological and conservation
significance, and species that are valued socially and economically for recreational fishing (Wager and Unmack 2000; Pusey
et al. 2004).
10.1071/MF09096
1323-1650/10/080842
Factors driving fish losses in floodplain refugia
Marine and Freshwater Research
843
Cooper Creek is often described as a ‘boom and bust’
ecological system (Walker et al. 1995, 1997; Bunn et al.
2006a, 2006b) where fish and other biota have evolved resilience traits (e.g. high vagility, use of floodplain resources for
feeding, growth and recruitment) that enable them to respond
opportunistically to flooding when it occurs. Bursts of primary
and secondary production on freshly inundated floodplains
translate into high levels of fish recruitment and production that
are often correlated with flood height and/or the area of flooded
terrain (Christensen 1993; Puckridge et al. 2000; James et al.
2008). All fish species (with the possible exception of the
endemic Cooper tandan Neosiluroides cooperensis) utilise the
rich food resources of the floodplains, and grow appreciably
over periods of floodplain inundation (Balcombe et al. 2007;
Balcombe and Arthington 2009). During the ‘bust’ periods that
inevitably follow floods (or flow pulses confined within channels), obligate aquatic species die as water levels decline, or
retreat to wetted habitats that continue to hold water even when
there is little or no baseflow (Walker et al. 1995; Hamilton et al.
2005). For desert fishes left in isolated waterholes after the
floods recede, dry periods are times of substantial stress and high
mortality resulting from deteriorating habitat conditions and the
biotic pressures of competition, predation and disease (Ruello
1976; Wager and Unmack 2000; Arthington et al. 2005).
Fish living in arid-zone rivers typically possess well-developed
resistance traits, such as tolerance of thermal, salinity,
turbidity and oxygen extremes, habitat and trophic generalism and predator avoidance strategies (Matthews and MarshMatthews 2003). The Cooper Creek fauna is no exception
(Puckridge et al. 2000; Wager and Unmack 2000; Balcombe
et al. 2005). Even so, fish abundances and the assemblage
composition in individual waterholes have been found to
change markedly over dry periods (Puckridge et al. 2000;
Arthington et al. 2005; Balcombe and Arthington 2009). This
paper presents another perspective on the fish assemblages of
drying waterholes by modelling how changes in environmental factors associated with water loss can influence
changes in fish assemblage structure, where change is measured as ‘turnover’ – the number of local colonisations and
extinctions related to the total number of species in the
community (MacArthur and Wilson 1967). We tested three
hypotheses: (1) that changes in waterhole fish assemblages
would be driven primarily by species losses because waterhole isolation should prevent colonisation events; (2) that
changes in fish assemblage structure would vary among
waterholes as a function of environmental factors operating
at three interacting spatial scales – the floodplain, the waterhole and habitat patches within waterholes (Arthington et al.
2005; Thorp et al. 2008); and (3) that individual species
would respond to different sets of environmental factors
depending on their habitat requirements, trophic ecology
and tolerance characteristics.
(‘Windorah’ and ‘Tanbar’) on Cooper Creek below the confluence of the Thomson and Barcoo Rivers, and one reach on
Kyabra Creek (‘Springfield’) (Fig. 1). A large waterhole was
selected in each reach as well as several other waterholes situated
in an adjacent or more distant channel, or upstream/downstream
of the main waterhole; additional waterholes were generally
within 5–10 km of the large waterhole. Cooper Creek waterholes
range in length from a few hundred metres to over 20 km. Some
have well-developed levees and are typically 2–5-fold wider and
2–3-fold deeper than those on the associated distributary channels
(Knighton and Nanson 1994a). Floodplain size, waterhole
exposure, basin morphometry, riparian development and physical
habitat structure varied greatly in our study area and influenced
the persistence times of water in these refuge habitats (Davis et al.
2002; Arthington et al. 2005; Hamilton et al. 2005). All of the
study reaches were unregulated by water infrastructure and
extraction, apart from minor pumping to homesteads.
The climate of the Cooper Creek catchment is classed as
semiarid to arid, with a mean annual rainfall varying from 400 to
500 mm in the headwaters too100 mm at its entry to Lake Eyre
(Puckridge 1999; Bunn et al. 2003). The hydrology of the
Cooper catchment has been described in detail by Puckridge
et al. (1998), Knighton and Nanson (2001) and McMahon and
Finlayson (2003). Most stream flow is generated by seasonal
monsoon rainfall in the headwaters and periodic local rainfall,
resulting in highly variable patterns of floodplain inundation
and within channel-flow pulses that can connect waterholes.
At Currareva near Windorah (Fig. 1), the mean annual discharge
is 3.05 km3 year1 (97 m3 s1), with wide interannual variation
and periods of zero flow lasting up to 21 months (based on
49 years of records). Episodic floods can inundate tens of
thousands of square kilometres of floodplain (equivalent to
35% of the catchment) and serve to reconnect channels, anabranches, distributaries and isolated waterholes (Knighton and
Nanson 1994a, 1994b; Bunn et al. 2003).
Mean daily maximum air temperatures at Windorah range
from 38.18C in January to 21.48C in July and mean daily minima
range from 248C in January to 78C in July. Mean annual
evaporation exceeds 3 m and, together with transpiration and
groundwater recharge, results in transmission losses below
Windorah accounting for more than two-thirds of the discharge
by the time it reaches the South Australian border (Knighton and
Nanson 1994b). Prior to this study, there was a large flood in
January–February 2000 connecting all waterholes at Noonbah,
Windorah and Tanbar, but not those along Kyabra Creek
(Springfield), where flows remained within the channel (B.
Morrish, pers. comm.). A small in-channel flow in December
2000–January 2001 linked all Springfield sites and also connected all of the study waterholes at Noonbah, Windorah and
Tanbar. Thereafter, there was no flow through the study reaches
and all waterholes decreased in volume between the early
(April) and late (September) dry season of 2001.
Materials and methods
Study area
Studies on waterhole fish assemblages were conducted at four
locations in the Cooper Creek catchment: one reach of the
Thomson River near Stonehenge (‘Noonbah’), two reaches
Fish collection
All 15 waterhole sites in the study area were sampled twice for
fish (April and September 2001) using four standardised sampling methods (fyke nets, beach seine, small purse-shaped drag
net and a zooplankton net). Details of the gear and sampling
844
Marine and Freshwater Research
140°E
A. H. Arthington et al.
142°E
144°E
148°E
Tho
ms
22°E
on
R
.
N
146°E
24°E
Barcoo R.
Windorah
26°E
Thomson River
Kyabra Creek
28°E
1(8)
THN
2(8)
WLN
0(8)
Cooper Creek
PHN
6(10)
GMW
MFW
MWW
Noonbah
Barcoo River
5(12)
SHW
0(9)
6(12)
Windorah
BHN
1(9)
Springfield
3(9)
3(10)
Kyabra Creek
Cooper Creek
3(10)
3(10)
1(9)
3(8)
YPT
WSF
HSF
OSF
TBT
6(10)
YLT
57% turnover (maximum)
Tanbar
YKT
11% turnover (minimum)
Fig. 1. Location of the study area and Cooper Creek catchment and diagrammatic representation of the
study reaches (boxes) and waterholes (circles) showing net change in species richness (no parentheses),
species richness in April 2001 (in parentheses) and relative fish assemblage turnover value (¼ area of each
circle) for each waterhole. Site codes: Noonbah reach: THN, Top; WLN, Waterloo; BHN, Bottom; PHN,
Pelican; Springfield reach: WSF, Warrannee; HSF, Homestead; OSF, One-Mile; Windorah reach: MWW,
Murken; SHW, Shed; GMW, Glenmurken; MFW, Mayfield; Tanbar reach: YKT, Yorakah; YLT,
Yalungah; YPT, Yappi; TBT, Tanbar.
techniques are provided in Arthington et al. (2005). Fyke net
samples provided catch per unit effort data for each species in
each waterhole in April and September 2001. Beach seining was
undertaken during daylight hours in the littoral area to collect
individuals and species that were too small to be collected by the
fyke nets. A small purse-shaped drag net used for the collection
of small fish and invertebrates for genetic analysis occasionally
captured additional small fish species (especially Hypseleotris
spp.) that were missed in the fyke nets and the beach seine.
Plankton tows also caught small fish on occasion. Additional
species collected by these methods were added to the presence/
absence data for each waterhole. Fish were identified to species
Factors driving fish losses in floodplain refugia
level using Allen et al. (2002) and species lists for the Lake Eyre
Basin (Puckridge 1999; Wager and Unmack 2000). Apart from
alien species (which were destroyed under the conditions of our
fishing permit) and small samples retained for genetic and food
web analyses, all fish were returned to the water alive.
Waterhole physical characteristics and water chemistry
Fish assemblages in rivers are influenced by abiotic and biotic
factors and ecological processes that operate at multiple spatial
scales. In Cooper Creek, we recognised three potentially interacting spatial scales – the floodplain, the waterhole and habitat
patches within each waterhole (after Davis et al. 2002). We
measured 25 physical and water quantity/quality variables at the
three spatial scales using remotely sensed imagery, aerial photographs or ground surveys (see methods described in Davis et al.
2002). Physical floodplain and waterhole information was collected in April 2001 in association with fish sampling, with
additional observations made and photographs taken during
September 2001 to document changes in physical waterhole
characteristics over time. Estimates of water loss in each waterhole between the two fish sampling occasions were provided by
Hamilton et al. (2005) based on changes in conservative ions
(Naþ and Cl) and from evaporative fractionation of oxygen and
hydrogen isotopes in the water. The temperature, pH and dissolved oxygen (DO) concentration were measured at each
waterhole on both sampling occasions. Gross primary productivity (GPP) levels in the shallow littoral zone of each
waterhole were determined by measuring dissolved oxygen and
temperature within open-bottom in situ chambers over 24 h. The
GPP was calculated as the sum of the DO produced during daylight hours plus the DO consumed by respiration during that
period of time based on the night-time respiration rate. Changes in
DO concentrations over time (mg O2 L1 h1) were multiplied by
chamber volume and divided by substrate surface area to obtain
values in units of mg O2 m2 h1. These rates were converted to
units of carbon, assuming that one mole of carbon is equivalent
to one mole of O2 for both respiration and photosynthesis (i.e.
1 mg O2 ¼ 0.375 mg C: Bender et al. 1987); the methods are
described in Bunn et al. (2003) and Fellows et al. (2009).
Statistical analysis
The total number of species present in each waterhole on each
sampling occasion was calculated from the pooled results of all
sampling methods. Assemblage-level turnover was estimated
from the presence–absence fish data for each waterhole, where
turnover is quantified according to the number of local colonisations and extinctions related to the number of species in the
community (MacArthur and Wilson 1967). A turnover index
(Tn) was calculated by dividing the sum total of species that
appeared and disappeared between April and September 2001
by the total number of species found during the two sampling
periods (cf. Russell et al. 1995). This turnover index ranges from
0 (no change in species composition) to 1 (complete change in
species composition).
Hierarchical canonical correspondence analysis (CCA) was
used to partition the total variance in species–environment
relationships into its independent and confounded components,
to facilitate tests of the relative importance of factors at different
Marine and Freshwater Research
845
organisational levels in driving changes in fish assemblages. This
analysis uses a series of partial canonical ordinations to decompose species–environment relationships into a hierarchically
structured data set (Anderson and Gribble 1998; Cushman and
McGarigal 2002). Following Cushman and McGarigal (2002),
environmental variables from the three spatial scales of measurement and five categories of factors (25 variables; Table 1) were
separately subjected to a principal components analysis (PCA) to
derive a reduced set of independent, composite variables that
summarise the major patterns of environmental variation in each
category. The statistical significance of the principal component
axes was evaluated using the broken-stick rule, where the
observed eigenvalues are compared with eigenvalues from random data (Peres-Neto et al. 2003). For all five categories of
environmental factors, the first two principal components were
found to be significant and were subsequently used as predictor
variables in the hierarchical CCA. In summary, the hierarchical
CCA was based on 10 principal components, four describing
within-waterhole variation (two habitat, two morphology), four
entire waterhole (two quality/quantity, two morphology) and two
catchment/floodplain variables. Using the results from the PCA,
we selected eight waterhole variables that explained the dominant
gradients of variation for inclusion in a CCA. Univariate normality of the variables was explored and present in all instances, thus
meeting the assumptions required for the PCA and CCA.
Classification and regression tree (CART) analysis was then
used to model community-level turnover and species change as
a function of the 25 variables describing waterhole physical
characteristics and water chemistry (Table 1). Three Lake Eyre
Basin endemics were selected for the species level analysis:
desert rainbowfish (Melanotaenia splendida tatei), the Lake
Eyre yellowbelly (Macquaria sp. B) and the Cooper Creek
tandan (Neosiluroides cooperensis). We also included northwest ambassis (Ambassis sp.) and the Barcoo grunter (Scortum
barcoo) because we were interested in why these particular
species were lost so often from drying waterholes.
The CART analysis is a machine-learning technique that can
operate on mixed variable types; these analyses are invariant
to monotonic transformations of data and are well suited for
modelling the non-linear and non-additive associations typical of
species–environment relationships (De’ath and Fabricus 2000;
Olden et al. 2008). CART uses a recursive partitioning algorithm
to split data into a nested series of mutually exclusive groups with
the goal of maximising homogeneity of the response variable. We
used the Gini impurity criterion to determine the optimal variable
splits (minimum parent node size: n ¼ 5; minimal terminal node
size: n ¼ 2), and determined the optimal size of the decision tree
by constructing a series of cross-validated trees and selecting the
smallest tree based on the one-standard-error rule (De’ath and
Fabricus 2000). Model performance for community turnover
(regression tree) was assessed using the Pearson’s productmoment correlation coefficient and for species change (classification tree) using correct classification rate; both were calculated
using n-fold cross-validation. Variable importance was determined by calculating for each variable at each node the change in
Gini impurity attributed to the best surrogate split on that variable.
The values of these deltas are summed over each node and
totalled, and scaled relative to the best-performing variable, that
is, they are expressed as relative importance on a scale of 0–100
846
Marine and Freshwater Research
A. H. Arthington et al.
Table 1. List of the 25 physical and water quantity/quality variables measured or estimated at the three spatial scales (floodplain, entire waterhole
and within waterhole) on one or both fish sampling occasions (April and September 2001)
The eight waterhole variables from the principal components analysis that explained the dominant gradients of variation for inclusion in the canonical
correspondence analysis are presented in bold face
Variable category
Description
Units
Floodplain
Effective floodplain width (EFW)
Floodplain setting (FS)
Bifurcation ratio (BR)
Channel distance to nearest waterhole (CD)
Straight-line distance to nearest waterhole (SLD)
m
Entire Waterhole
Morphology/Size
Water Quality/Quantity
Within Waterhole
Morphology/Complexity
Habitat
Cross-sectional area (CSA)
Bankfull depth (D)
Hydraulic radius (HR)
Wetted perimeter (WP)
Circularity index (CI)
Gross primary production (GPP)
Water loss (WL)
Water temperature (WT)
pH (PH)
Dissolved O2 (O2)
Bench 0 ] 1/3 (B1)
Bench 1/3 – 2/3 (B2)
Bench 2/3 – 3/3 (B3)
Sum of bars (BAR)
Off-take channels (OC)
Bed and bank complexity (BBC)
Eroding banks (EB)
Snags (SN)
Boulders (BOU)
Sum of vegetation (VEG)
km
km
m2
m
m
%
%
%
%
%
Range
748–7076
0.0126–1.9369
0.0004–0.0023
1.18–18.98
1.11–7.95
77.3–474.1
2.48–6.72
1.51–4.41
8.9–149.5
0.019–0.765
3.8–800
4.3–73.9
28.3–32.8
6–10.1
71.3–144.4
0.3–1.5
0.03–1.20
0–0.94
0.6–2.7
0.077–1.324
0.005–0.076
0–0.778
0–1.864
0–0.18
0.2–1.9
EFW, distance from obstructions such as sand hills to the edge of the floodplain; FS, ratio of the distance from a waterhole to the right and left floodplain edge;
BR, degree of floodplain dissection (number of channels present on a floodplain divided by floodplain width).
Bankfull cross-sectional area, width : depth ratio, wetted perimeter, bankfull depth, volume, hydraulic radius (ratio of cross-sectional area to the wetted
perimeter, effectively giving a measure of mean depth) and circularity index (CI ¼ 4pA/P2, where A is the surface area and P is the perimeter of the waterhole;
waterholes are more circular in shape when values approach one compared with more elongate waterhole shapes with CI values close to zero) were calculated
from cross-sections (5–20) surveyed at 500-m intervals along each waterhole.
Waterhole variables (morphology/complexity, habitat) were recorded as presence/absence at 100-m intervals along each waterhole and standardised for
waterhole size by dividing the number of positive readings by the number of observation intervals.
(Breiman et al. 1984). All analyses were performed in R
(R Development Core Team 2008).
Results
Fish assemblages of Cooper Creek waterholes
The fish fauna of Cooper Creek waterholes comprised 12 indigenous fish species in eight families and two alien species:
the goldfish Carassius auratus and the mosquitofish Gambusia
holbrooki (Table 2). Indigenous species richness per waterhole
based on all sampling methods was 7–11 species in April and
4–8 species in September 2001. Patterns of species occurrence
across the 15 waterholes, and species abundances in each
waterhole on the two sampling occasions, were highly variable
based on catch per unit effort data derived from standardised
fyke net sampling (Table 2). Five widespread indigenous species
contributed 96% of the total catch based on standardised fyke net
sampling of waterhole assemblages. These species were the
silver tandan Porochilus argenteus (45.5%), north-west
Ambassis (19.9%), spangled perch Leiopotherapon unicolor
(12.5%), bony herring Nematalosa erebi (9.7%) and Hyrtl’s
tandan Neosilurus hyrtlii (8.4%). The Barcoo grunter Scortum
barcoo comprised 2.6% of the total catch and was collected from
all 15 waterholes in April 2001 (Table 2). The three Lake Eyre
Basin endemics (the desert rainbowfish Melanotaenia splendida
tatei, the Lake Eyre yellowbelly Macquaria sp. B and Welch’s
grunter Bidyanus welchi) were found in variable numbers of
waterholes in April 2001, and were relatively abundant at some
sites, whereas the endemic Cooper Creek tandan (Neosiluroides
cooperensis) was rare in comparison (Table 2).
Assemblage level changes over time
Waterhole fish assemblages showed substantial changes in
species composition over time. For the entire fish fauna, the
predominant pattern was the persistence of widespread and
14
15
2
12
14
4
15
15
2
4
15
15
10
3
Bony herring
Carp gudgeons
Desert rainbowfish
Lake Eyre golden perch
Cooper Creek tandan
Hyrtl’s tandan
Silver tandan
Australian smelt
Welch’s grunter
Spangled perch
Barcoo grunter
Goldfish
Mosquitofish
No. waterholes where
present in April 2001
North-west ambassis
Common name
2
8
2
13
14
2
0
13
4
6
1
15.3 11.4 (0–171)
14.9 3.6 (0–41)
1.0 0.5 (0–6)
418 215 (1–2872)
2201 1411 (21–20 229)
0.13 0.13 (0–2)
0.87 0.47 (0–6)
664 406 (0–2)
146 120 (0–1825)
3.7 1.2 (0–13)
0.13 0.13 (0–2)
9
13
396 323 (11–4920)
0A
9
No. of waterholes where
present in September 2001
1071 1009 (0–15 179)
Mean s.e. (range) of
CPUE in April 2001
Hypseleotris spp. was not collected in fyke net samples in April 2001, but was collected by other methods so is included in the species list.
A
Indigenous species
Chandidae
Ambassis sp.
Clupeidae
Nematalosa erebi (Günther, 1868)
Gobiidae
Hypseleotris spp.
Melanotaeniidae
Melanotaenia splendida tatei (Zietz, 1896)
Percichthyidae
Macquaria sp. B (after Musyl and Keenan, 1992)
Plotosidae
Neosiluroides cooperensis Allen and Feinberg, 1998
Neosilurus hyrtlii Steindachner, 1867
Porochilus argenteus (Zietz, 1896)
Retropinnidae
Retropinna semoni (Weber, 1895)
Terapontidae
Bidyanus welchi (McCulloch and Waite, 1917)
Leiopotherapon unicolor (Günther, 1859)
Scortum barcoo (McCulloch and Waite, 1917)
Alien species
Cyprinidae
Carassius auratus Linnaeus, 1758
Poeciliidae
Gambusia holbrooki (Girard, 1859)
Family/species
0
1.2 0.6 (0–13)
0
25.8 7.1 (0–93)
1.20 0.84 (0–12)
0.07 0.07 (0–1)
0.50 0.41 (0–6)
19.8 6.2 (0–63)
162 49 (0–689)
4.7 2.0 (0–23)
0.13 0.13 (0–1)
0.13 0.09 (0–2)
70.9 23.7 (0–279)
32.6 27.1 (0–410)
Mean s.e. (range) of
CPUE in September 2001
Table 2. Fish species found in waterholes of the Thomson River, Cooper Creek and Kyabra Creek based on all four sampling methods, and the mean 6 standard error and (range) of catch per
unit effort (CPUE) for fyke net samples collected in April and September 2001
Factors driving fish losses in floodplain refugia
Marine and Freshwater Research
847
848
Marine and Freshwater Research
A. H. Arthington et al.
relatively abundant species, contrasting with losses of less
abundant species (Table 2). In total, there were 55 species losses
across the 15 waterholes. Gains of one or occasionally two
species were recorded in 11 of the 15 individual waterholes.
Seven of these ‘gains’ involved small gudgeons of the genus
Hypseleotris, which are usually too small to capture in fyke nets
(see zero to extremely low catches recorded in Table 2), but are
more readily caught by seine and plankton nets. Three other
uncommon species were recorded as ‘gains’ – the Cooper tandan
(N. cooperensis), the goldfish (C. auratus) and the mosquitofish
(G. holbrooki). The desert rainbowfish (M. s. tatei), a common,
but generally not abundant species, was recorded as ‘gained’ in
one waterhole and lost from 11.
Community turnover rates varied from 0.11 (one species
gained, one lost) at Glenmurken waterhole in the Windorah reach,
to 0.57 (seven species lost, one gained) at Yorakah waterhole in
the Tanbar reach (Fig. 2). Among the common species, the silver
tandan P. argenteus (45.5% of total catch) was lost from only one
waterhole (Yorakah in the Tanbar reach), the spangled perch
L. unicolor (12.5% of total catch) and the bony herring N. erebi
(9.7% of total catch) were each lost from two waterholes, while
the species most often lost (from 11 waterholes) were desert
rainbowfish and Barcoo grunter (Scortum barcoo).
Environmental variables explaining assemblage level
change over time
Results from the CCA based on the eight most important
environmental variables identified by the PCA show the similarities and differences in assemblage change among the 15
waterholes in relation to the dominant environmental gradients
(Fig. 2). The first and second CCA axes explained 20.5% and
15.0% of the total variation in turnover, respectively. In general,
CCA 1 represents a gradient of increasing change in fish
assemblages from right to left of this axis, with high eroding
banks and depth corresponding to high assemblage change.
Waterholes OSF (One-Mile, Springfield) and TBT (Tanbar)
showed similar patterns of assemblage change and were characterised by high GPP and locations towards the centre of their
respective floodplains, whereas YLT (Yalungah, Tanbar) and
YPT (Yappi, Tanbar) had similar patterns of assemblage change
associated with high floodplain width.
We used hierarchical CCA based on 10 principal components
from the PCA to determine the relative amount of variation in
assemblage change explained by different levels of the environmental hierarchy. For all five categories, the first two components of the PCA were significant, and together explained the
following percentages of variation: catchment/floodplain scale
(72%), waterhole morphology (79%), waterhole water quality/
quantity (67%), within-waterhole morphology (76%) and withinwaterhole habitat (70%). Results of the hierarchical CCA are
shown in a Venn diagram (Fig. 3). This analysis found that 74%
of variation in fish assemblage change could be explained by all
10 descriptors of floodplain and waterhole characteristics taking
effect independently and through interactions. Waterhole variables (depth, wetted perimeter, temperature and GPP) explained
most variation in assemblage change (33.7%), followed by
waterhole habitat variables (morphological complexity and
eroded banks), which explained 21.4%. Variables describing
floodplain size and waterhole position on the floodplain
explained 12.1% of the variation in fish assemblage change,
and interactions across the three spatial scales and their variable
sets explained 6.8%.
The CART analysis revealed the specific interactions of the
environmental variables and their effects on assemblage
change and species’ losses. Regression tree analysis formulated five relatively homogenous groups of sites and five
3
CCA 2 (15.0% variation explained)
OSF
2
GPP
1
YKT
TBT
FloodplainSet
Eroding banks
0
MFW
Depth
MWW
BHN
HSF
Perimeter
THN
Bench1
TemperatureWSF SHW
WLN
PHN
GMW
1
FloodplainWidth
YLT
2
2
1
0
1
YPT
2
CCA 1 (20.5% variation explained)
Fig. 2. Ordination plot resulting from the canonical correspondence analysis examining the
association between species loss and the physical and water quantity/quality variables of
the waterholes. The three letter codes refer to the waterholes (codes defined in Fig. 1) and the
vectors represent the strength and direction of the association between the predictor variables
and the ordination axes.
Factors driving fish losses in floodplain refugia
Marine and Freshwater Research
849
Habitat
only
Catchment
Morphology
only
Within Waterhole
Habitat
Waterhole
Morphology
Quality/Quantity
only
Quality/Quantity
Morphology
only
Morphology
Fig. 3. Results of first-tier and second-tier decompositions of the influence of catchment-level, waterhole-level and
within-waterhole-level factors on fish species turnover. The area of each rectangular cell is proportional to the variance
accounted for by that component. The numbers in the components list the percentage of the explained variance
accounted for by each component. The solid arrows are from first-tier circles to second-tier marginal decompositions,
and the dotted arrows are from first-tier independent effects to second-tier conditional decompositions.
Turnover 0.28
s
Ye
Eroding banks 0.64
Turnover 0.50
Turnover 0.25
Wetted
perimeter 66.4 m
Turnover 0.20
pH 0.6%
Turnover 0.28
Turnover 0.15
No
Turnover 0.30
Floodplain width 1194 m
Turnover Turnover 0.31
0.23
Bench 0
(B1) 0.52
Turnover 0.33
Turnover 0.29
Fig. 4. Regression tree predicting community-level species change as a
function of key waterhole characteristics. Note that the specific splitdefining values in the tree should not be interpreted as distinct thresholds,
but rather as a relative threshold compared with the entire data set. Variable
codes are defined in Table 1.
environmental variables were identified as predictors of fish
assemblage change (Fig. 4), with eroded banks (EB) selected
as the first variable splitting the data set and explaining the
largest amount of variation in assemblage change (0.50).
Although fish assemblage change was lower (0.25) in waterholes with less bank erosion, large waterholes (indicated by a
high bankfull wetted perimeter, WP 466.4 m) showed greater
change (0.30), especially those located on a wide floodplain
(indicated by effective floodplain width, EFW 41194 m) with
only shallow in-channel benches (B1 0.52). In this group of
waterholes, assemblage change was 0.33. In smaller waterholes (W 66.4 m), changes in fish assemblages were higher
(0.28) where pH increased by o0.6%.
Environmental variables explaining species level change
over time
Three Lake Eyre Basin endemics, the north-west Ambassis and
the Barcoo grunter (S. barcoo) were selected for modelling
environmental factors driving species losses. Classification
trees revealed the individual and interacting environmental
variables that most affected losses of these species.
The north-west Ambassis (80% correct classification) persisted in 10 waterholes, but was lost from five (Fig. 5a). Losses
in four waterholes were associated with high levels of bank
erosion (EB44.2) and low channel and bank irregularities
(hydraulic radius (HR) 3.0). In shallower waterholes (bankful
depth 6.3 m) with low levels of bank erosion, there were no
losses of this species over time. The desert rainbowfish M. s. tatei
850
Marine and Freshwater Research
A. H. Arthington et al.
(a) Ambassis sp.
10
5
No
(b) Melanotaenia splendida tatei
9
1
Depth 6.3 m
1
4
1 3
Hydraulic
radius 3.0
9
1
4
1
No change
Loss
Loss
No change
11
11
Bench 1 1.4
Ye
s
Ye
s
Eroding
banks 4.2
11
2
Temp. 33%
1
11
1
No change
6
Ye
s
9
Off-take
channels 0.68
No
8
1
1
No change Loss
1
Gain
11
Loss
2
5
(d ) Neosiluroides cooperensis
Floodplain
setting 0.56
2
1
Ye
Loss
No change
2
9
s
Ye
No
4
9
2
1
No
3
Loss
9
Depth 5.3 m
9
1
No change
Loss
2
Floodplain
width 9.534 m
Loss
1
11
Gain Loss
11
Loss
4
GPP 2.5%
(e) Scortum barcoo
BBC 0.04
9
Depth 5.4 m
s
5
4
No change
Water
loss 6.2%
(c) Macquaria sp. B
No
4
No change
KEY:
Gain
No change
Loss
Fig. 5. Classification trees discriminating among Cooper Creek waterholes that lost a species (black bar), gained a species
(grey bar) or showed no change (white bar) between April and September 2001 as a function of waterhole characteristics.
Species modelled included (a) Ambassis sp., (b) Melanataenia s. tatei, (c) Macquaria sp. B, (d) Neosiluroides cooperensis
and (e) Scortum barcoo. Bar width reflects the number of waterholes (values also presented as text).
(87% correct classification) showed three instances of no change
over time, one gain and 11 losses; these were associated with low
levels of bench development, especially in waterholes with
marked decreases in water temperature and46.2% loss of water
by volume over the dry season (Fig. 5b). The Lake Eyre yellowbelly Macquaria sp. B (87% correct classification) persisted in
eight waterholes and was lost from six others (Fig. 5c). Most of
these losses were from waterholes with higher numbers of off-take
channels, except in waterholes located close to the middle of the
left and right floodplain boundaries (indicated by floodplain
setting 40.56). Waterholes located closer to the edges of the
floodplain (i.e. more isolated from other waterholes and from
the influence of floods) showed greater losses. At this bifurcation
of the classification tree, water temperature was a surrogate
variable, where decreases in temperature over the dry season were
associated with more losses of Lake Eyre yellowbelly. The Cooper
Creek tandan (93% correct classification) showed nine instances
of no change over time, two gains and four losses (Fig. 5d). Deeper
waterholes (45.4 m) experienced most of the losses of this species,
whereas shallow waterholes with increased GPP over the dry
season were able to retain this species. For the Barcoo grunter
(93% correct classification), the most important variables
Factors driving fish losses in floodplain refugia
explaining 11 losses were low levels of bed and bank complexity
(BBC) and low floodplain width (Fig. 5e). There were no gains
recorded for this species and four waterholes remained unchanged.
Discussion
Fish assemblages of Cooper Creek waterholes
The fauna of the study area included four species confined
to central Australia – the endemic Cooper Creek tandan
(N. cooperensis), the Lake Eyre yellowbelly (Macquaria sp. B),
the desert rainbowfish (M. s. tatei) and Welch’s grunter
(B. welchi). Five species collected have a tropical and central
Australian distribution – P. argenteus, N. hyrtlii, L. unicolor,
north-west Ambassis and S. barcoo. The Australian smelt, R.
semoni, has a subtropical and temperate distribution and N. erebi
is widespread in central, eastern and northern Australia (Allen
et al. 2002; Pusey et al. 2004).
Assemblage level changes over time
Patterns of species assemblage change in Cooper Creek waterholes during the 2001 seasonal dry period were dominated by
the persistence of common and abundant species and by losses
of species of lower abundance, including four Lake Eyre Basin
endemics. There were 55 instances of species loss across 15
waterholes. However, five species not captured or sighted in
particular waterholes in April 2001 were caught in those
waterholes in September, after 6 months of waterhole drying and
presumed disconnection. These species ‘gains’ were recorded in
11 of the 15 individual waterholes; however, seven ‘gains’
involved small gudgeons of the genus Hypseleotris. This species
is usually too small to capture in fyke nets, but can be more
readily caught by seine and plankton nets. The occurrence of
gudgeons in September, but not in April probably reflects the
greater probability of capture of this species, by all methods, in
drying waterholes where fish are more concentrated in low
volumes of water.
Across 15 waterholes and 14 native and alien fish species,
the proportion of ‘gains’ relative to losses was low (20%), and
for M. s. tatei it was o10%. These figures may serve as rough
estimates of the number of instances where a species recorded as
‘lost’ in September was actually missed during sampling. We
suggest that there was an even lower likelihood of misses in
September when drying waterholes had decreased in volume,
fish were more concentrated, and all sampling methods were
potentially more efficient (Arthington et al. 2005). Certainly
there was no evidence of significant fish movements into
waterholes. We therefore affirm that changes in waterhole fish
assemblages were driven primarily by species losses because a
lack of hydrological connectivity and waterhole isolation prevented colonisation events (Hypothesis 1).
The prevalence of species losses means that none of the 15
waterholes was able to sustain all 12 indigenous fish species
over the drying phase of 2001, with up to six species lost from
individual waterholes (maximum turnover 0.57 at Yorakah
waterhole in the Tanbar reach). The species most often lost
(from 11 waterholes) were the desert rainbowfish (M. s. tatei)
and the Barcoo grunter (S. barcoo).
Marked assemblage change and the loss of aquatic species from
remnant aquatic habitats is a common event in drought-stressed
Marine and Freshwater Research
851
streams and rivers (Matthews and Marsh-Matthews 2003; Bond
et al. 2008). Such losses are often associated with the loss of
suitable stream habitat (Labbe and Fausch 2000; Matthews and
Marsh-Matthews 2003; Boys and Thoms 2006). In this study, we
have demonstrated that a complex spatial hierarchy of independent
and interacting environmental factors affected the habitat characteristics, availability and quality of isolated waterholes, and
drove the 55 instances of species loss over the dry season of 2001.
The importance of spatial scale was revealed first by CCA, where
74% of variation in fish assemblage change could be explained by
descriptors of waterhole morphology, habitat structure and floodplain setting. Most variation in assemblage change (33.7%) was
uniquely explained by waterhole morphology (wetted perimeter,
depth) and waterhole ‘quality’ (temperature and GPP), with an
additional 21.4% of variation explained by within waterhole
variables (morphological complexity, habitat structure and bank
stability). These findings are consistent with previous studies that
explain differences in fish assemblage structure in isolated floodplain waterbodies in terms of habitat structure and quality (Rodriguez and Lewis 1997; Tejerina-Garro et al. 1998; Welcomme et al.
2006).
Our study adds another dimension to models of fish assemblage change over dry periods in floodplain rivers – the
importance of floodplain size (i.e. width as a surrogate for
floodable area) and waterbody position in the surrounding
floodplain. In combination, these factors explained an additional
12.1% of variation in change at the assemblage level in Cooper
Creek waterholes. Previous studies in floodplain rivers have
given prominence to the influence of floodplain size on fish
assemblages, with little attention to the floodplain setting of
waterbodies. Many studies have found relationships between
fish catches in dry season habitats in one year and the size of the
flood pulse (or area of floodplain inundated) in the same or in
preceding years (Puckridge et al. 2000; Arthington et al. 2005;
Welcomme et al. 2006). These relationships arise because
inundated floodplains provide important nursery areas for larval
and juvenile fishes and also allow older fish to disperse, feed and
accumulate sufficient fat reserves to survive the dry season and
complete reproduction (Junk et al. 1989; Welcomme 2001; King
et al. 2003). In Cooper Creek, all indigenous species (with the
possible exception of the Cooper tandan N. cooperensis) use
inundated floodplains as juveniles and maturing individuals
and/or as spent adults (Balcombe et al. 2005, 2007; Balcombe
and Arthington 2009). Our evidence of the influence of antecedent flooding on dry season fish assemblages is also consistent
with the work of Puckridge et al. (2000) on the dynamics of
fish in the Coongie Lakes, situated on the lower Cooper Creek
system.
When floodwaters recede off large floodplains, many thousands of fish are available to colonise dry season lagoons, lakes
and waterholes (Welcomme 2001; Balcombe et al. 2007).
Waterholes that receive more fish, and fish in better condition,
are more likely to sustain their fish assemblages over the dry
season than those receiving lower numbers and/or fish in
relatively poor condition.
The importance of the floodplain setting of individual waterholes is probably linked to connectivity potential, with waterholes far out on the edges of a floodplain less likely to be
inundated than those situated closer to the centre. However,
852
Marine and Freshwater Research
biological connectivity of isolated waterholes may depend also
on specific types of pathway for fish movement across a floodplain. For example, channels draining the floodplain can provide
depth, velocity and directional cues that ‘funnel’ fish into
particular waterholes or floodplain lakes (Bishop et al. 1995;
Welcomme 2001). Thus, the characteristics and/or number of
floodplain channels connecting individual waterholes, as well as
the straight-line distance between waterholes, appear likely to
influence overall hydrological and biological connectivity in
floodplain systems (Sheldon et al. 2002; Arthington et al. 2005).
All of these factors and processes may affect colonisation rates
into formerly isolated waterholes and would also influence
biophysical conditions within waterholes, and hence fish survival after flood recession and natural drying processes.
Our analysis identified the particular environmental factors
within the three primary spatial factor sets that were responsible
for assemblage and species level changes. The regression tree
analysis produced a strong initial result – low levels of bank
erosion were conducive to fish persistence, high levels were not.
Several erosion processes are likely to be involved (e.g. gully
and sheet erosion, bank instability associated with fallen trees
and with undercutting of root masses stranded above the falling
waterline). Gully and sheet erosion would contribute to siltation
of waterholes and possibly to less differentiation of habitat
structure, and overall loss of habitat complexity within waterholes (Unmack 2001). Some of the waterholes of Cooper Creek
develop at points of flow convergence where the upper layer of
clay becomes incised, allowing the sand beneath to be excavated
into large pools. As floodwaters subside, the clay particles settle
and seal the depressions to create relatively permanent waterholes (Knighton and Nanson 2000). Habitat complexity is not
particularly high in Cooper Creek compared with other dryland
rivers (e.g. the Finke River in the Alice Springs area; Unmack
2001) and more mesic rivers where rock formations, coarse
substrates and large woody debris loads contribute to greater
habitat diversity. Bank instability and erosion may contribute to
hydrological and sedimentation processes that influence the
physical characteristics of waterholes, bringing about changes
that have an adverse influence on habitat heterogeneity. Fallen
trees, root masses and undercut banks exposed by falling water
levels would result in loss of in-stream habitat structure, cover
and shelter from piscivorous birds and other fish (Kennard 1995;
Pusey et al. 2004; Boys and Thoms 2006).
Although assemblage level change was lower in waterholes
showing less bank erosion, the regression analysis showed that
large waterholes with high wetted perimeter showed greater
change, especially those located on wide floodplains and with
shallow in-channel benches. Large waterholes situated on wide
floodplains are exposed to wind and turbulence factors (Davis
et al. 2002) that can influence rates of evaporative water loss, the
main process affecting drying rates in Cooper Creek waterholes.
Hamilton et al. (2005) found that evaporation rates are highly
variable among Cooper Creek waterholes, in part owing to their
different sizes, their effective width for wind action, the degree
of channel incision below levees and the height and width of
riparian vegetation. These features affect wind-induced turbulence at the water surface, exposure of the surface to solar
heating and convective air circulation above the water surface
(Brutsaert 1982). These factors contribute to the rate of water
A. H. Arthington et al.
loss in particular waterholes and associated changes in habitat
structure, availability and quality.
A surprising result was that water loss per se over the dry
period did not emerge as a major driving variable in our CART
analyses. Instead, the influence of water loss was manifest
indirectly by impacts on habitat availability for fishes. With a
decline in waterhole depth and volume, waterhole features such
as fallen trees, debris from riparian vegetation, root masses,
shallow benches and bars and deeper depressions can become
exposed and increasingly desiccated, depending on their elevation along the basin profile (Arthington et al. 2005). At low
water levels, exposure of such features would reduce the range
of sheltered places where fish can rest and forage, take cover
from predators, or launch ambush attacks on prey (Kennard
1995; Crook and Robertson 1999; Pusey et al. 2004). In isolated
waterholes of Cooper Creek, water loss translates to absolute
loss of aquatic habitat as well as changes in habitat heterogeneity, in accord with the first flow ecology principle of Bunn and
Arthington (2002). The influence of habitat structure (especially
structurally complex habitat such as submerged branches, leaf
litter and aquatic vegetation) on fish assemblage structure has
been demonstrated in other dry season floodplain habitats
(Kennard 1995; Welcomme 2001). Loss/change of habitat is
also important in intermittent streams subject to natural periods
of low flow or drought (Magoulick and Kobza 2003; Matthews
and Marsh-Matthews 2003; Larned et al. 2010).
Many studies have found that deterioration in water quality
(e.g. salinity, dissolved oxygen, turbidity, pH) is a major factor
in species losses from drying streams and isolated floodplain
habitats (Rodriguez and Lewis 1997; Magoulick and Kobza
2003; Matthews and Marsh-Matthews 2003). In our study,
changes in water chemistry, such as conductivity and nutrient
concentrations, did not emerge as important influences on the
survival of Cooper Creek fishes, apart from slightly higher levels
of assemblage change in small water holes with slight increases
in pH. These pH changes may be indicative of biogeochemical
processes that can affect fish physiology and behaviour. Other
studies have shown that several species in our study area are
affected by changes in pH. Balcombe and Arthington (2009)
found strong correlations between declining numbers of Barcoo
grunter (S. barcoo), spangled perch (L. unicolor) and silver
tandan (P. argenteus) and increasing pH associated with waterhole drying in the Windorah reach of Cooper Creek.
In summary, this study has demonstrated that a large amount
of variation in loss of species from isolated waterholes can be
explained by environmental factors operating at three interacting spatial scales (Hypothesis 2). Floodplain size, the position
of waterholes on the floodplain, waterhole morphology and
bank condition, and the heterogeneity of habitat patches within
waterholes are all important correlates of species persistence
and species losses. We suggest that these factors can be used to
identify waterholes most likely to sustain their fish assemblages
over dry periods.
Environmental variables explaining species level change
over time
We examined the drivers of change in the occurrence of three
Lake Eyre Basin endemics, north-west Ambassis and the Barcoo
Factors driving fish losses in floodplain refugia
grunter (S. barcoo). These geographically restricted and, in
some cases, relatively uncommon species may be more at risk
from the effects of waterhole drying and loss of high-quality
refugia than the more common species with wider geographic
distributions. Ensuring the persistence of endemic species in
the regional species pool is an important tenet of conservation
biology and species recovery plans in disturbed landscapes
(Bond and Lake 2003; Knight and Arthington 2008).
The desert rainbowfish (M. s. tatei) was lost from more
waterholes (11/15) than all other species except the Barcoo
grunter (S. barcoo), particularly from waterholes with low levels
of bench development, and where increased loss of water over
the dry season was associated with decreased water temperature.
Rainbowfish are surface feeders and depend on habitat structure
and food resources in the upper levels of the water column
(Pusey et al. 2004) where water temperatures would be most
likely to change over winter months. Surface water temperatures
closely mimic air temperature in exposed waterbodies and
evaporative loss can also reduce surface temperature (Matthews
and Marsh-Matthews 2003). Another small species, the northwest Ambassis, seemed very sensitive to high levels of bank
erosion in shallow waterholes (as measured by the hydraulic
radius, equivalent to mean depth). This response to seasonal
drying may reflect the preference of ambassids for mid to lower
water column areas over fine substrates and proximity to cover,
such as woody debris and leaf litter (Pusey et al. 2004). Bank
erosion may bring about loss or smothering of such habitat
elements in the more eroded waterholes, while declining water
levels would reduce opportunities for fish to find cover.
The Lake Eyre yellowbelly (Macquaria sp. B) is an important recreational fish in Australia’s inland rivers (Wager and
Unmack 2000; Pusey et al. 2004). It was lost from six waterholes, particularly those with higher numbers of off-take channels. As water levels fall, pools of water can remain in off-take
channels, but these patches of wetted habitat may become
isolated from the main body of a waterhole, possibly stranding
any fish left behind (Arthington et al. 2005). The CART analysis
revealed additional processes at work in that losses of the
yellowbelly were higher in waterholes located closer to the
edges of the floodplain. Waterholes at these locations are more
isolated from the influence of floods and from other waterholes
and therefore have lower potential hydrological connectivity
with source populations of fish. They may have started the 2001
dry season with less water, poor water quality and diminished
food resources compared with waterholes that were connected
during the floods of January–February 2000. We recorded
relatively low numbers of yellowbelly even at the start of the
dry season (mean CPUE 14.9 3.6, range 0–41). Low initial
abundances and adverse environmental conditions could have
been inimical to fish persistence in waterholes experiencing
further water loss over the 2001 dry season. In the decision tree
for Macquaria sp. B, decreasing water temperature over the dry
season also affected losses of this species. Exposure on the
floodplain may have influenced the thermal regime of isolated
waterholes, as described above in relation to the effect of
decreased water temperatures on the desert rainbowfish (M. s.
tatei).
Deeper waterholes exhibited losses of the endemic Cooper
Creek tandan (N. cooperensis), whereas shallow waterholes
Marine and Freshwater Research
853
with increases in GPP over the dry season were able to sustain
this species. The fishes of Cooper Creek are ultimately dependent on a shallow ‘bath-tub ring’ of benthic algae (Bunn et al.
2003), which would be likely to achieve greater biomass in
shallow sloping littoral zones than in steeper, deeper waterholes.
Our result and other studies of trophic ecology in Australian
floodplain rivers suggest that the food resources of drying
waterholes play an important part in sustaining their fish
assemblages (Balcombe et al. 2005; Medeiros and Arthington
2008). This is perhaps especially important for larger species
such as the Cooper Creek tandan (N. cooperensis) that feed on
large-bodied invertebrates, such as crustaceans and gastropods
(Allen et al. 2002), that are themselves sustained by high levels
of littoral primary production (Bunn et al. 2003). In this
instance, waterhole morphology appears to play a significant
role in supporting the aquatic food web and ultimately the
production of food resources needed to sustain fish during dry
periods.
The Barcoo grunter (S. barcoo) was lost from as many
waterholes as the rainbowfish (11), but different factors came
into play. Losses were associated with low levels of bed and
bank complexity and less cover to afford protection from
predators while foraging (Kennard 1995; Pusey et al. 2004).
This species of grunter was able to persist in waterholes offering
more complex habitat structure, but not in those situated on
narrow floodplains. This association may reflect the limited
capacity of small floodplains to support fish populations during
and after flood periods. Arthington et al. (2005) found a
correlation between floodplain width (a simple surrogate for
the area of floodplain inundated the year before sampling) and
numbers of S. barcoo in drying waterholes of Cooper Creek.
Smaller floodplains either produced lower numbers of this
species, or fewer individuals were returned to waterholes after
flood recession, or both processes were involved. An alternative
or additional factor affecting the persistence of this grunter in
drying waterholes may involve limited delivery of energy
sources and nutrients back into waterholes after flood recession
(i.e. low floodplain subsidies; Junk et al. 1989). In both cases,
fish that benefit from flooding can be expected to transform
floodplain energy sources into maintenance and growth, and
perhaps elevated survivorship, while they are confined to isolated waterholes (Puckridge et al. 2000).
This analysis of endemic and geographically restricted species has revealed the influence of suites of factors that explain
a large amount of variation in individual species’ losses from
isolated waterholes of Cooper Creek. Each species has been
shown to respond to different sets of environmental factors that,
in turn, underlie mechanisms of response to drying conditions.
The main biological mechanisms are associated with the particular habitat requirements, trophic ecology and physical/
chemical tolerance characteristics of each species (Hypothesis 3).
Role of biotic processes
The CCA analysis showed that 26% of variation in assemblage
level change was unexplained by the 25 factors included in
our spatially nested hierarchy of environmental variables. It is
always possible that additional variation might be explained by
the inclusion of other environmental factors or more refined
854
Marine and Freshwater Research
measurement of those included here. A further source of variation would be the influence of biotic processes, such as predation, competition for food resources or space, or exposure
to parasites and diseases. These biotic processes account for
significant fish mortality in drying waterbodies of mesic and
dryland rivers worldwide (Ruello 1976; Jackson et al. 2001;
Magoulick and Kobza 2003; Balcombe et al. 2005). Studies in
large floodplain rivers suggest that changes in habitat conditions
and consequences for predation and competition success
become increasingly important with progression of the dry
season (Kennard 1995; Rodriguez and Lewis 1997; TejerinaGarro et al. 1998). However, evidence of predation and competition impacts on Cooper Creek fish assemblages is very
limited. There appears to be relatively little piscivory in this fish
assemblage, with only Ambassis sp., L. unicolor and M. s. tatei
reported to consume fish (Balcombe et al. 2005). Moreover, it
appears that piscivory occurs mainly on inundated floodplains,
where all three species feed on larval or juvenile bony herring
(N. erebi) that dominate floodplain fish catches (Balcombe et al.
2007). When fish are confined to isolated waterholes after flood
recession, most species in the Cooper Creek assemblage feed on
a narrow range of benthic invertebrates and zooplankton; there
is no evidence of piscivory in waterhole dietary data. Whether
fish are affected by resource or interference competition when
habitat availability is declining and food resources are not
diverse is unclear. Balcombe et al. (2005) showed that this suite
of opportunistic carnivores is able to maintain stomach fullness
even when the diversity of food types in isolated waterholes is
low. The ability to feed and persist on a few types of food
(especially calanoid copepods) during dry periods is an important feeding strategy, given the intermittency of flooding and
uncertainty of access to rich floodplain food resources in this
arid-zone river (Balcombe et al. 2005).
Evidence of disease and parasitism as mortality factors for
Cooper Creek fishes is very limited. Fish of the Coongie Lakes
region of Cooper Creek are regularly affected by dermatitis
caused by Achlya sp. (Puckridge and Drewien 1988). Species
affected include the bony herring (N. erebi), Lake Eyre yellowbelly (Macquaria sp.), desert rainbowfish (M. s. tatei) and
goldfish (C. auratus). Increased rates of infection by the protozoan parasite Chilodonella hexasticha in central Australian
populations of N. erebi have also been associated with decreased
winter water temperatures (Pusey et al. 2004). Although N. erebi
is generally regarded as a sensitive species, vulnerable to low
temperatures and to physical ‘handling’, it is one of the most
widespread Australian freshwater fishes (Pusey et al. 2004).
This clupeid was able to persist in all but two waterholes over the
dry season studied here.
Implications for management
The survival of fish and other riverine biota during seasonal and
prolonged dry periods is a topical issue in Australia during the
present ‘millennium drought’ (Bond et al. 2008; Murphy and
Timbal 2008). The identification and protection of refugia for
obligate aquatic species is of particular interest, for it is these
special places and habitats that will sustain enough individuals
to repopulate a wider range of habitats when seasonal or prolonged drought conditions come to an end, and more favourable
A. H. Arthington et al.
conditions are restored (Bond et al. 2008). These are significant
issues in every river prone to intermittent flows and habitat
fragmentation (Rayner et al. 2009; Larned et al. 2009).
In Cooper Creek, a diversity of waterhole fish assemblages is
highly variable in space and time, with up to 6 of 12 native
species lost from individual waterholes over a single 6-month
dry season. However, overall g diversity was constant across the
15 waterholes. Even the rarest species were able to persist in
some waterholes. The crucial issue is: which waterholes should
be protected to conserve g diversity? Our CCA and CART
models offer valuable tools for understanding the mechanisms
underlying fish losses from isolated waterholes in Cooper
Creek, and should assist in the identification of quality refugia
in this river system and, possibly, other dryland rivers. However,
in such a dynamic environment, models such as these should
be tested over several dry seasons, and preferably over longer
periods of waterhole isolation, before we can be certain of their
consistency and predictive power and rely on them to nominate
particular waterholes as vital refugia for fish. Even then, it will
be unwise to base the management of such rivers on the
protection of a few select waterholes, given the potential for
failure of the wet season and the unpredictable frequency and
spatial patterns of flooding, channel flows, connectivity and
waterhole replenishment (Puckridge et al. 1998, 2000; Sheldon
et al. 2010). A better strategy will be to recognise the special
dynamic characteristics of arid-zone rivers and adopt a set of
scientific and management principles built around the maintenance of ecosystem resilience. Chief among these scientific
principles for arid-zone rivers are the hydrological, sedimentary
and biogeochemical processes, and changes to these wrought
by human activities, that can affect the characteristics and
quality of aquatic refugia and their very existence in the wider
landscape (Davies et al. 1994; Kingsford 2006; Wishart 2006).
We concur with Larned et al. (2010) and Sheldon et al. (2010)
that preservation or restoration of habitat mosaics, connectivity,
natural flow intermittence, and identification of flow requirements for highly valued species and processes, are fundamental
management principles for all intermittent rivers.
Acknowledgements
This study contributed to the Dryland River Refugia Project, a multidisciplinary and multi-institutional investigation of environmental factors
and processes sustaining waterhole biodiversity in dryland rivers of western
Queensland. It was funded by the former Cooperative Research Centre
(CRC) for Freshwater Ecology, Canberra. Completion of this paper was
supported by funding from the eWater CRC (Refugium Project). We thank
colleagues from the Australian Rivers Institute at Griffith University, the
former Queensland Department of Natural Resources and Mines, and the
Murray-Darling Basin Freshwater Research Centre (Northern Basin
Laboratory) for field assistance, data on river discharge and valuable
discussions. Two anonymous referees and Editors Andrew Boulton and
Fran Sheldon provided insightful reviews and comments that greatly
improved the manuscript. Thanks are also due to Ben Stuart-Koster for his
input to statistical discussions. We are indebted to landowners Bob
Morrish (Springfield), Angus Emmott (Noonbah), Sandy Kidd (Mayfield),
David Smith (Hammond Downs) and George Scott (Tanbar) for access to
waterholes on their properties and for their hospitality and encouragement.
Our research was conducted under Queensland Fisheries Permit
PRM00157K.
Factors driving fish losses in floodplain refugia
References
Allen, G. R., Midgley, S. H., and Allen, M. (2002). ‘Field Guide to the
Freshwater Fishes of Australia.’ (Western Australian Museum: Perth.)
Anderson, M. J., and Gribble, N. A. (1998). Partitioning the variation among
spatial, temporal and environmental components in a multivariate dataset.
Australian Journal of Ecology 23, 158–167. doi:10.1111/J.1442-9993.
1998.TB00713.X
Arthington, A. H., Balcombe, S. R., Wilson, G. A., Thoms, M. C., and
Marshall, J. (2005). Spatial and temporal variation in fish assemblage
structure in isolated waterholes during the 2001 dry season of an aridzone river, Cooper Creek, Australia. Marine and Freshwater Research
56, 25–35. doi:10.1071/MF04111
Balcombe, S. R., and Arthington, A. H. (2009). Temporal changes in fish
abundance in response to hydrological variability in a dryland floodplain
river. Marine and Freshwater Research 60, 146–159. doi:10.1071/
MF08118
Balcombe, S. R., Bunn, S. E., McKenzie-Smith, F. J., and Davies, P. E.
(2005). Variability of fish diets between dry and flood periods in an
arid zone floodplain river. Journal of Fish Biology 67, 1552–1567.
doi:10.1111/J.1095-8649.2005.00858.X
Balcombe, S. R., Bunn, S. E., Arthington, A. H., Fawcett, J. H., McKenzieSmith, F. J., et al. (2007). Fish larvae, growth and biomass relationships
in an Australian arid zone river: links between floodplains and waterholes. Freshwater Biology 52, 2385–2398. doi:10.1111/J.1365-2427.
2007.01855.X
Bender, M., Grande, K., Johnson, K., Sieburth, J., Pilson, M., et al. (1987).
A comparison of four methods for determining planktonic community
production. Limnology and Oceanography 32, 1085–1098. doi:10.4319/
LO.1987.32.5.1085
Bishop, K. A., Pidgeon, R. W., and Walden, D. J. (1995). Studies on fish
movement dynamics in a tropical floodplain river: prerequisites for a
procedure to monitor the impacts of mining. Australian Journal of
Ecology 20, 81–107. doi:10.1111/J.1442-9993.1995.TB00524.X
Bond, N. R., and Lake, P. S. (2003). Characterising fish-habitat assocations
in streams as the first step in ecological restoration. Austral Ecology 28,
611–621.
Bond, N. R., Lake, P. S., and Arthington, A. H. (2008). The impacts of
drought on freshwater ecosystems: an Australian perspective. Hydrobiologia 600, 3–16. doi:10.1007/S10750-008-9326-Z
Boys, C. A., and Thoms, M. C. (2006). A hierarchical scale approach to the
assessment of fish assemblages and their habitat associations in large
dryland rivers. Hydrobiologia 572, 11–31. doi:10.1007/S10750-0050004-0
Breiman, L., Friedman, J. H., Olshen, R. A., and Stone, C. J. (1984).
‘Classification and Regression Trees.’ (Chapman & Hall: Boca Raton, FL.)
Brutsaert, W. (1982). ‘Evaporation into the Atmosphere: Theory, History,
and Applications.’ (Reidel: Boston, MA.)
Bunn, S. E., and Arthington, A. H. (2002). Basic principles and consequences of altered hydrological regimes for aquatic biodiversity.
Environmental Management 30, 492–507. doi:10.1007/S00267-0022737-0
Bunn, S. E., Davies, P. M., and Winning, M. (2003). Sources of organic
carbon supporting the food web of an arid zone floodplain river. Freshwater Biology 48, 619–635. doi:10.1046/J.1365-2427.2003.01031.X
Bunn, S. E., Balcombe, S. R., Davies, P. M., Fellows, C. S., and McKenzieSmith, F. J. (2006a). Aquatic productivity and food webs of desert river
ecosystems. In ‘Ecology of Desert Rivers’. (Ed. R. T. Kingsford.)
pp. 76–99. (Cambridge University Press: Melbourne.)
Bunn, S. E., Thoms, M. C., Hamilton, S. K., and Capon, S. J. (2006b). Flow
variability in dryland rivers: boom, bust and the bits in between. River
Research and Applications 22, 179–186. doi:10.1002/RRA.904
Christensen, M. S. (1993). The artisanal fishery of the Mahakam River
floodplain in East Kalimantan, Indonesia. III. Actual and estimated
yields, their relationship to water levels and management options.
Marine and Freshwater Research
855
Journal of Applied Ichthyology 9, 202–209. doi:10.1111/J.1439-0426.
1993.TB00396.X
Crook, D. A., and Robertson, A. I. (1999). Relationships between riverine
fish and woody debris: implications for lowland rivers. Marine and
Freshwater Research 50, 941–953. doi:10.1071/MF99072
Cushman, S. A., and McGarigal, K. (2002). Hierarchical, multi-scale
decomposition of species–environment relationships. Landscape
Ecology 17, 637–646. doi:10.1023/A:1021571603605
Davies, B. R., Thoms, M. C., Walker, K. F., O’Keefe, J., and Gore, J. A.
(1994). Arid and semi-arid-land river ecosystems: perspectives on
ecological functioning and problems of management and conservation.
In ‘Rivers Handbook, Volume 2’. (Eds P. Calow and G. E. Petts.)
pp. 484–511. (Blackwell Scientific: Oxford.)
Davis, L., Thoms, M. C., Fellows, C. S., and Bunn, S. E. (2002). Physical
and ecological associations in dryland refugia: waterholes of the Cooper
Creek, Australia. International Association of Hydrological Sciences
276, 77–84.
De’ath, G., and Fabricus, K. E. (2000). Classification and regression trees: a
powerful yet simple technique for ecological data analysis. Ecology 81,
3178–3192. doi:10.1890/0012-9658(2000)081[3178:CARTAP]2.0.CO;2
Fellows, C. S., Bunn, S. E., Sheldon, F., and Beard, N. J. (2009). Benthic
metabolism in two turbid floodplain rivers. Freshwater Biology 54,
236–253. doi:10.1111/J.1365-2427.2008.02104.X
Hamilton, S. K., Bunn, S. E., Thoms, M. C., and Marshall, J. C. (2005).
Persistence of aquatic refugia between flow pulses in a dryland river
system (Cooper Creek, Australia). Limnology and Oceanography 50,
743–754.
Jackson, D. A., Peres-Neto, P. R., and Olden, J. D. (2001). What controls
who is where in freshwater fish communities – the roles of biotic, abiotic
and spatial factors. Canadian Journal of Fisheries and Aquatic Sciences
58, 157–170. doi:10.1139/CJFAS-58-1-157
James, C. S., Thoms, M. C., and Quinn, G. P. (2008). Zooplankton dynamics
from inundation to drying in a complex ephemeral floodplain-wetland.
Aquatic Sciences 70, 259–271. doi:10.1007/S00027-008-8034-0
Junk, W. J., Bayley, P. B., and Sparks, R. E. (1989). The Flood Pulse Concept
in river–floodplain systems. Proceedings of the International Large
Rivers Symposium (LARS) (Ed. D. P. Dodge). Canadian Journal of
Fisheries and Aquatic Sciences Special Publication 106, 110–127.
Kennard, M. J. (1995). Factors influencing freshwater fish assemblages in
floodplain lagoons of the Normanby River, Cape York Peninsula, a large
tropical Australian river. M Phil. Thesis, Griffith University, Brisbane.
King, A. J., Humphries, P., and Lake, P. S. (2003). Fish recruitment on
floodplains: the roles of patterns of flooding and life history characteristics. Canadian Journal of Fisheries and Aquatic Sciences 60, 773–786.
doi:10.1139/F03-057
Kingsford, R. T. (2006). Changing desert rivers. In ‘Ecology of Desert
Rivers’. (Ed. R. T. Kingsford.) pp. 336–345. (Cambridge University
Press: Melbourne.)
Knight, J. T., and Arthington, A. H. (2008). Distribution and habitat
associations of the endangered Oxleyan pygmy perch, Nannoperca
oxleyana Whitley, in eastern Australia. Aquatic Conservation: Marine
& Freshwater Ecosystems 18, 1240–1254. doi:10.1002/AQC.936
Knighton, A. D., and Nanson, G. C. (1994a). Waterholes and their
significance in the anastomosing channel system of Cooper Creek,
Australia. Geomorphology 9, 311–324. doi:10.1016/0169-555X(94)
90052-3
Knighton, A. D., and Nanson, G. C. (1994b). Flow transmission along an
arid zone anastomosing river, Cooper Creek, Australia. Hydrological
Processes 8, 137–154. doi:10.1002/HYP.3360080205
Knighton, A. D., and Nanson, G. C. (2000). Waterhole form and process
in the anastomosing channel system of Cooper Creek, Australia.
Geomorphology 35, 101–117. doi:10.1016/S0169-555X(00)00026-X
Knighton, A. D., and Nanson, G. C. (2001). An event based approach to
the hydrology of arid zone rivers in the Channel Country of Australia.
856
Marine and Freshwater Research
A. H. Arthington et al.
Journal of Hydrology (Amsterdam) 254, 102–123. doi:10.1016/S00221694(01)00498-X
Labbe, T. R., and Fausch, K. D. (2000). Dynamics of intermittent stream
habitat regulate persistence of a threatened fish at multiple scales.
Ecological Applications 10, 1774–1791. doi:10.1890/1051-0761(2000)
010[1774:DOISHR]2.0.CO;2
Larned, S. T., Datry, T., Arscott, D. B., and Tockner, T. (2010). Emerging
concepts in temporary-river ecology. Freshwater Biology 55, 717–738.
MacArthur, R. H., and Wilson, E. O. (1967). ‘The Theory of Island
Biogeography.’ (Princeton University Press: Princeton, NJ.)
Magoulick, D. D., and Kobza, R. M. (2003). The role of refugia for fishes
during drought: a review and synthesis. Freshwater Biology 48, 1186–
1198. doi:10.1046/J.1365-2427.2003.01089.X
Matthews, W. J., and Marsh-Matthews, E. (2003). Effects of drought on
fish across axes of space, time and ecological complexity. Freshwater
Biology 48, 1232–1253. doi:10.1046/J.1365-2427.2003.01087.X
McMahon, T. A., and Finlayson, B. L. (2003). Droughts and anti-droughts:
the low flow hydrology of Australian rivers. Freshwater Biology 48,
1147–1160. doi:10.1046/J.1365-2427.2003.01098.X
Medeiros, S. E. F., and Arthington, A. H. (2008). The importance of
zooplankton in the diets of three native fish species in floodplain
waterholes of a dryland river, the Macintyre River, Australia. Hydrobiologia 614, 19–31. doi:10.1007/S10750-008-9533-7
Morton, S. J., Short, J., and Baker, R. D. (1995). Refugia for biological
diversity in arid and semi-arid Australia. Biodiversity Series, Paper
No. 4. (Department of Environment, Sport and Territories: Canberra.)
Murphy, B. F., and Timbal, B. (2008). A review of recent climate variability
and climate change in southeastern Australia. International Journal of
Climatology 28, 859–879. doi:10.1002/JOC.1627
Nilsson, C., Reidy, C. A., Dynesius, M., and Revenga, C. (2005). Fragmentation and flow regulation of the world’s large river systems. Science
308, 405–408. doi:10.1126/SCIENCE.1107887
Olden, J. D., Lawler, J. J., and Poff, N. L. (2008). Machine learning methods
without tears: a primer for ecologists. The Quarterly Review of Biology
83, 171–193. doi:10.1086/587826
Peres-Neto, P. R., Jackson, D. A., and Somers, K. M. (2003). Giving
meaningful interpretation to ordination axes: assessing loading significance in principal component analysis. Ecology 84, 2347–2363.
doi:10.1890/00-0634
Poff, N. L., Allan, J. D., Bain, M. B., Karr, J. R., Prestegaard, K. L., et al.
(1997). The natural flow regime – a paradigm for river conservation and
restoration. Bioscience 47, 769–784. doi:10.2307/1313099
Puckridge, J. T. (1999). The role of hydrology in the ecology of Cooper
Creek, Central Australia: implications for the flood pulse concept. Ph.D.
Thesis, The University of Adelaide, Australia.
Puckridge, J. T., and Drewien, M. (1988). The aquatic fauna of the northwest branch of Cooper Creek. In ‘The Coongie Lakes Study’. (Eds J. R.
Reid and J. Gillen.) pp. 69–108. (South Australian Department of
Environment and Planning: Adelaide.)
Puckridge, J. T., Sheldon, F., Walker, K. F., and Boulton, A. J. (1998). Flow
variability and the ecology of large rivers. Marine and Freshwater
Research 49, 55–72. doi:10.1071/MF94161
Puckridge, J. T., Walker, K. F., and Costelloe, J. F. (2000). Hydrological
persistence and the ecology of dryland rivers. Regulated Rivers:
Research and Management 16, 385–402. doi:10.1002/1099-1646
(200009/10)16:5o385::AID-RRR59243.0.CO;2-W
Pusey, B. J., Kennard, M. J., and Arthington, A. H. (2004). ‘Freshwater
Fishes of North-eastern Australia.’ (CSIRO Publishing: Melbourne.)
R Development Core Team (2008). R: a language and environment for
statistical computing. R Foundation for Statistical Computing, Vienna.
Available at http://www.r-project.org/ [verified July 2010].
Rayner, T. S., Jenkins, K. M., and Kingsford, R. T. (2009). Small environmental flows, drought and the role of refugia for freshwater fish in the
Macquarie Marshes, arid Australia. Ecohydrology 2, 440–453.
Rodriguez, M. A., and Lewis, W. M., Jr (1997). Structure of fish assemblages along environmental gradients in floodplain lakes of the Orinoco
River. Ecological Monographs 67, 109–128.
Ruello, N. V. (1976). Observations on some fish kills in Lake Eyre.
Australian Journal of Marine and Freshwater Research 27, 667–672.
doi:10.1071/MF9760667
Russell, G. J., Diamond, J. M., Pimm, S. L., and Reid, T. M. (1995). A
century of turnover: community dynamics at three timescales. Journal of
Animal Ecology 64, 628–641. doi:10.2307/5805
Sheldon, F., Boulton, A. J., and Puckridge, J. T. (2002). Conservation value
of variable connectivity: aquatic invertebrate assemblages of channel
and floodplain habitats of a central Australian arid-zone river, Cooper
Creek. Biological Conservation 103, 13–31. doi:10.1016/S0006-3207
(01)00111-2
Sheldon, F., Bunn, S. E., Hughes, J. M., Arthington, A. H., Balcombe, S. R.,
et al. (2010). Ecological roles and threats to aquatic refugia in arid
landscapes: dryland river waterholes. Marine and Freshwater Research
61, 885–895. doi:10.1071/MF09239
Tejerina-Garro, F. L., Fortin, R., and Rodriguez, M. A. (1998). Fish
community structure in relation to environmental variation in floodplain
lakes of the Araguaia River, Amazon Basin. Environmental Biology of
Fishes 51, 399–410. doi:10.1023/A:1007401714671
Thorp, J. H., Thoms, M. C., and Delong, M. D. (2008). ‘The Riverine
Ecosystem Synthesis.’ (Elsevier: San Diego, CA.)
Unmack, P. J. (2001). Fish persistence and fluvial geomorphology in central
Australia. Journal of Arid Environments 49, 653–669. doi:10.1006/
JARE.2001.0813
Wager, R., and Unmack, P. J. (2000). ‘Fishes of the Lake Eyre Catchment
of Central Australia’. (Department of Primary Industries Queensland
Fisheries Service: Brisbane.)
Walker, K. F., Sheldon, F., and Puckridge, J. T. (1995). An ecological
perspective on dryland river ecosystems. Regulated Rivers: Research
and Management 11, 85–104. doi:10.1002/RRR.3450110108
Walker, K. F., Puckridge, J. T., and Blanch, S. J. (1997). Irrigation
development on Cooper Creek, central Australia – prospects for a
regulated economy in a boom-and-bust ecology. Aquatic Conservation:
Marine & Freshwater Ecosystems 7, 63–73. doi:10.1002/(SICI)10990755(199703)7:1o63::AID-AQC21843.0.CO;2-5
Welcomme, R. L. (2001). ‘Inland Fisheries Ecology and Management.’
(Fishing News Books, Blackwell Science: Oxford.)
Welcomme, R. L., Bene, C., Brown, C. A., Arthington, A., Dugan, P., et al.
(2006). Predicting the water requirements of river fisheries. In ‘Wetlands
and Natural Resource Management’. (Eds J. T. A. Verhoeven,
B. Beltman, R. Bobbink and D. F. Whigham.) Ecological Studies
Vol. 190, pp. 123–154. (Springer-Verlag: Berlin.)
Wishart, M. J. (2006). Water scarcity: politics, populations and the ecology
of desert rivers. In ‘Ecology of Desert Rivers’. (Ed. R. T. Kingsford.)
pp. 315–335. (Cambridge University Press: Melbourne.)
Manuscript received 30 April 2009, accepted 10 February 2010
http://www.publish.csiro.au/journals/mfr