Content uploaded by Mirela G Tulbure
Author content
All content in this area was uploaded by Mirela G Tulbure on Jul 16, 2018
Content may be subject to copyright.
Accepted Article
This article has been accepted for publication and undergone full peer review but has not
been through the copyediting, typesetting, pagination and proofreading process, which may
lead to differences between this version and the Version of Record. Please cite this article as
doi: 10.1002/eap.1507
This article is protected by copyright. All rights reserved.
MR. ROBBI BISHOP-TAYLOR (Orcid ID : 0000-0002-1533-2599)
Received Date : 05-Oct-2016
Revised Date : 22-Dec-2016
Accepted Date : 10-Jan-2017
Article type : Articles
Running head
Dynamic surface-water connectivity
Title
Surface-water dynamics and land use influence landscape connectivity across a major dryland region
List of authors
Robbi Bishop-Taylor (Corresponding author1)
School of Biological, Earth & Environmental Sciences,
University of New South Wales, Sydney, NSW 2052, Australia
Mirela G Tulbure
School of Biological, Earth & Environmental Sciences,
University of New South Wales, Sydney, NSW 2052, Australia
Mark Broich
School of Biological, Earth & Environmental Sciences,
University of New South Wales, Sydney, NSW 2052, Australia
1 r.bishop-taylor@unsw.edu.au
Accepted Article
This article is protected by copyright. All rights reserved.
Abstract
Landscape connectivity is important for the long-term persistence of species inhabiting dryland
freshwater ecosystems, with spatiotemporal surface-water dynamics (e.g., flooding) maintaining
connectivity by both creating temporary habitats and providing transient opportunities for dispersal.
Improving our understanding of how landscape connectivity varies with respect to surface-water
dynamics and land use is an important step to maintaining biodiversity in dynamic dryland
environments. Using a newly available validated Landsat TM and ETM+ surface-water time series,
we modelled landscape connectivity between dynamic surface-water habitats within Australia’s 1
million km2 semi-arid Murray Darling Basin across a 25-year period (1987 to 2011). We identified
key habitats that serve as well-connected ‘hubs’, or ‘stepping-stones’ that allow long-distance
movements through surface-water habitat networks. We compared distributions of these habitats
for short- and long-distance dispersal species during dry, average and wet seasons, and across land-
use types. The distribution of stepping-stones and hubs varied both spatially and temporally, with
temporal changes driven by drought and flooding dynamics. Conservation areas and natural
environments contained higher than expected proportions of both stepping-stones and hubs
throughout the time series; however, highly modified agricultural landscapes increased in
importance during wet seasons. Irrigated landscapes contained particularly high proportions of well-
connected hubs for long-distance dispersers, but remained relatively disconnected for less vagile
organisms. The habitats identified by our study may serve as ideal high-priority targets for land-use
specific management aimed at maintaining or improving dispersal between surface-water habitats,
potentially providing benefits to biodiversity beyond the immediate site scale. Our results also
highlight the importance of accounting for the influence of spatial and temporal surface-water
dynamics when studying landscape connectivity within highly variable dryland environments.
Keywords: landscape connectivity, surface-water dynamics, graph theory, land use, protected areas,
wetlands, flooding, Australia, Murray-Darling Basin, Landsat
Introduction
Dryland freshwater environments are among the most vulnerable ecosystems globally (Williams
1999), faced with the combined threats of land-use intensification and changes in water availability
driven by increased human consumption and a changing climate (Vörösmarty et al. 2010, Davis et al.
2015). This has resulted in extensive loss (up to 87% by area since 1700 AD) and degradation of
biodiverse wetlands due to agricultural and urban development and flow modification (Davidson
2014), isolation of remaining habitats separated by an increasingly modified landscape matrix
(Collins et al. 2014), and increasing severity of extreme drought and flooding events (Leblanc et al.
2012). The combined impacts of these changes to surrounding land use and spatiotemporal surface-
water dynamics (i.e., drought or flooding) are likely to disproportionately impact the population
persistence of dryland freshwater biota by reducing opportunities for dispersal through already
variable and fragmented surface-water habitat networks (Ward and Stanford 1995, Davis et al. 2015,
Saunders et al. 2015). Maintaining vital landscape-scale processes such as landscape connectivity in
ecosystems affected by increasingly intense human use and environmental change remains a key
Accepted Article
This article is protected by copyright. All rights reserved.
challenge facing conservation in the 21st century (Rudnick et al. 2012).
To mitigate these threats posed to freshwater ecosystems, biodiversity conservation approaches
have attempted to preserve ecologically significant surface-water habitats using protected area
strategies (Suski and Cooke 2006). Despite being effective at reducing site-scale threats such as
conversion of land for agriculture, protected area strategies can fail if they do not address
threatening processes operating at landscape, catchment or basin scales, such as changes to
hydrology or reductions in landscape connectivity (Pringle 2001, Berney and Hosking 2016). This may
be of particular concern for surface-water ecosystems, given the reliance of water-dependant biota
on water availability and variable flooding dynamics for recruitment and dispersal between habitats
(Crook et al. 2015). Identifying and managing habitats that preserve or enhance landscape
connectivity within habitat networks may provide a promising means to maintain population
persistence and biological diversity in threatened landscapes (Saunders et al. 2015). These areas
could serve as high-priority targets for the provision of limited and contested environmental water,
potentially facilitating dispersal and gene flow among isolated surface-water habitats and providing
biodiversity benefits that extend beyond protected area boundaries (Arthington 2012, Crook et al.
2015). Prioritising dynamic surface-water habitats for conservation at a network scale, however,
requires a detailed understanding of how spatial and temporal surface-water patterns influence the
potential for connectivity, particularly within heterogeneous landscapes affected by varying degrees
of modification (Watts et al. 2015).
Graph theory network analysis provides a powerful approach for studying spatiotemporal
connectivity patterns. Graph theory allows connectivity to be assessed across large spatial extents by
analysing entire landscapes in a network data structure (Minor and Urban 2007, 2008). For example,
individual surface-water habitats in a spatial surface-water habitat network can be depicted as a
series of ‘nodes’, with each node representing a potential habitat that can be connected to others by
an ‘edge’ (or link) if an ecological connection exists between them (Urban and Keitt 2001). In a
spatial habitat network, this may be the case when the distance and landscape conditions between
two neighbouring habitats allow resident water-dependent species to disperse successfully between
them. This spatially explicit analysis framework allows large habitat networks to be compared
consistently across both time and space. Graph theory analyses have provided insights into many
aspects of connectivity including how the loss or creation of habitats may impact long-distance
dispersal and gene flow (Uden et al. 2014), how migrating species may respond to a changing
climate (McIntyre et al. 2014, Dilts et al. 2016, McGuire et al. 2016) and how drought or flooding
may affect surface-water habitat network structure or metapopulation viability (Fortuna et al. 2006,
Wright 2010, Tulbure et al. 2014, Bishop-Taylor et al. 2015).
Algorithms provided by graph theory network analysis can assist in prioritising conservation
management actions by identifying individual habitats that are particularly important for facilitating
connectivity within habitat networks. Two of the most important roles a habitat can play are as
‘stepping-stones’ or ‘hubs’ (Ruiz et al. 2014, Saura et al. 2014). Stepping-stones are habitats that
allow dispersing organisms or populations to make long-distance movements through networks by
providing connections between larger groups of connected habitats. Stepping-stones can be critical
for range expansion, particularly as changing climate conditions and habitat loss forces organisms to
move long distances through increasingly modified and hostile landscapes (Kramer-Schadt et al.
2011, Saura et al. 2014). Hubs are habitats that are connected to a high number of neighbours
Accepted Article
This article is protected by copyright. All rights reserved.
relative to other habitats in the network. These habitats provide extensive opportunities for local
dispersal, and may serve as important ecological refuges into which organisms can retreat and
subsequently disperse from after periods of environmental disturbance (Davis et al. 2013, Bishop-
Taylor et al. 2015). Obtaining a better understanding of the importance of individual habitats within
habitat networks can therefore assist in maintaining landscape connectivity at scales far beyond
immediate sites.
While stepping-stone and hub habitats may represent ideal targets for habitat conservation, their
importance is unlikely to be consistent through time (Tulbure et al. 2014). In highly dynamic dryland
surface-water environments, only a small proportion of habitats may remain present from year to
year or season to season (Roshier et al. 2001). Other temporary or ephemeral habitats may play
important roles as connectivity providers only during specific environmental conditions, such as
during periods of extreme drought or flood (Zeigler and Fagan 2014). Despite the important
implications of spatiotemporal surface-water dynamics for connectivity, previous research has
focused on assessing connectivity between static or modelled sets of surface-water habitats (e.g.,
Fortuna et al. 2006, O’Farrill et al. 2014), or between waterbodies at discrete ‘snapshots’ in time
(e.g., Wright 2010, Uden et al. 2014). Little research has quantified how the distribution of important
stepping-stones and hubs varies across both space and time in the context of dynamic periods of
drought or flooding (Tulbure et al. 2014) or evaluated how these changing distributions interact with
current protected area listings or local land-use patterns (Ruiz et al. 2014, Bishop-Taylor et al. 2015).
In this study, we used graph theory network analysis to model spatially explicit patterns of important
stepping-stones and hubs using a newly available 25-year seasonally-continuous and statistically-
validated time series of potential surface-water habitats based on Landsat TM and ETM+ satellite
imagery (Tulbure et al. 2016). We focused on assessing connectivity within Australia’s Murray-
Darling Basin, a major inland river basin containing some of the world’s most ecologically significant
dryland floodplain and wetland environments (Rogers and Ralph 2010). By conducting to our
knowledge the largest spatially explicit ecological network analysis to date (a total of 5.4 million
nodes processed across 99 time-steps and two dispersal abilities), our study sought to answer the
following research questions relating to interactions among landscape connectivity, surface-water
dynamics and land-use and conservation management:
1. How does the distribution of important stepping-stones and hubs vary among dry,
average or wet conditions?
2. Are modified land-use types (e.g., grazing, dryland and irrigated agriculture) less likely to
contain stepping-stones and hubs than protected or natural areas?
3. How does dispersal ability affect relationships between land use and the distribution of
stepping-stones and hubs during dry, average or wet conditions?
Accepted Article
This article is protected by copyright. All rights reserved.
Methods
Study area
We focused on the Murray-Darling Basin (MDB), a 1 million km2 semi-arid basin in south-eastern
Australia (Figure 1). Despite being one of the driest catchments in the world by runoff (MDBA 2010),
the MDB supports some of Australia’s most biologically diverse and ecologically significant floodplain
and wetland habitats, including 16 sites listed as Wetlands of International Importance under the
Ramsar Convention (Rogers and Ralph 2010, Pittock and Finlayson 2011). Extreme variability in
water availability in the MDB drives a characteristic ‘boom and bust’ ecology, with populations of
many species fluctuating greatly in response to periods of severe drought and flooding (Ballinger and
Mac Nally 2006). Although these extremes are a natural feature of its regional climate, the MDB has
recently experienced both the most severe drought and wettest two-year period since instrumental
records began: the 1999–2010 Millennium Drought and the 2010–2011 La Niña floods (Mac Nally et
al. 2014). These extremes have combined with impacts of wetland loss, water diversions for
agriculture and flow modification to place increasing pressure on surface-water habitats (Finlayson
et al. 2011). Reduction in flows to wetlands is believed to be the major factor influencing the
ecological health of these ecosystems, with reduced landscape connectivity between habitats a key
threat to the persistence of water-dependent organisms (Finlayson et al. 2011).
Seasonal surface water
We studied the connectivity of surface-water habitats across the entire MDB between 1987–2011
using seasonally continuous observations of surface-water mapped from Landsat TM and ETM+
satellite imagery (Tulbure et al. 2016). To convert surface-water observations into potential habitat
layers, we identified pixels that contained water for greater than 50% of satellite observations within
each southern hemisphere season from 1987–2011 (i.e., autumn, winter, spring and summer
beginning on 1 March, 1 June, 1 September and 1 December, respectively). This majority threshold
ensured that resulting habitat pixels for each season were typically inundated for approximately 1.5
months, an ecologically relevant value for many aquatic invertebrates (e.g., 1–2 months from eggs to
independence for several MDB crustaceans; Jones 2010), amphibians (e.g., 2–4 month tadpole
lifespan for the majority of MDB frogs; Wassens 2010) and freshwater turtles (e.g., 60–70 day
incubation period near water for common MDB species; Thompson 1983, Kennett and Georges
1990).
We filled areas of ‘nodata’ on a per-pixel basis (Tulbure and Broich 2013) when a pixel was missing
data for an entire season (i.e. through consistent cloud cover). ‘Nodata’ pixels were assigned the
most common state (surface water or non-water) from the closest three years with data for the
respective season, commencing with the previous year (e.g., a ‘nodata’ pixel in autumn 2007 was
filled with values from autumn 2006, 2008 and 2005). To facilitate computationally intensive
connectivity analyses for the entire extensive spatial (> 1 million km2) and temporal (99 seasons)
data extent, the original 30m raster pixels were spatially aggregated to 120m pixels using a
maximum aggregation rule that considered aggregated cells as potential habitat if at least one of the
16 contributing pixels contained habitat (McRae 2008). All habitat layer manipulation was conducted
using the GDAL (GDAL Development Team 2015), NumPy and SciPy packages (Walt et al. 2011) for
Python 2.6 (Python Software Foundation 2015).
Accepted Article
This article is protected by copyright. All rights reserved.
Temporally consistent habitat IDs
To consistently compare connectivity across time and space, we assigned each flooded pixel with a
unique, temporally consistent habitat identifier (unique ID). Previous studies have defined discrete,
temporally consistent regions based on the maximum extent of flooding across an entire time series
to account for surface-water habitats growing, shrinking or fragmenting over time (Tulbure et al.
2014). In the MDB, however, rare but very large flooding events form extensive ‘mega-patches’ in
composite maximum extent datasets that can span hundreds of km2, potentially obscuring
ecologically significant seasonal surface-water dynamics. To resolve this ‘mega-patch problem’, we
used a simplified version of a method based on habitat asynchrony and graph theory community
detection previously used to identify temporally consistent subregions in remotely sensed giant kelp
forest time series (Cavanaugh et al. 2014). We initially generated an m x n matrix of n observations
(surface water or non-water for each of the 99 seasons) for each raster pixel m in the study area
(Figure 2a). We converted this matrix into a network graph by connecting surface-water pixels
(nodes) to their immediate 8 spatial neighbours, and weighted the resulting graph edges by the
pairwise Pearson correlation in temporal dynamics between each pixel (i.e., high correlation if the
two pixels consistently occurred together as either surface water or non-water throughout the time
series; Figure 2b).
We used the Igraph Python package (Csardi and Nepusz 2006) ‘community_multilevel’
implementation of the Louvain community detection algorithm to divide ‘mega-patches’ into
discrete asynchronous potential habitats (Cavanaugh et al. 2014). The Louvain algorithm efficiently
partitions large connected graphs into unique ‘communities’ by maximising the graph’s modularity, a
measure of the density of within-community links compared to between-community links (Newman
2006, Blondel et al. 2008). As the Louvain algorithm does not accept negative weights, we truncated
graph edge weights (correlations ranging from -1 to 1) to between 0 and 1 prior to analysis. The
community membership identifiers produced by the algorithm were finally converted back into a
raster containing 277,874 temporally consistent unique IDs informed by both the spatial structure
and temporal synchrony of habitats throughout the entire 1987–2011 time series (Figure 2c). These
unique IDs served as individual graph nodes in subsequent network analyses.
To test how successfully each unique ID habitat represented a spatially discrete area of surface
water across each season in the time series, we evaluated the proportion of seasons where surface
water was entirely restricted to within a habitat’s unique ID boundary (i.e., where no flooding
crossed between habitat boundaries). On average, the 277,874 habitats identified by the algorithm
had no water crossing habitat boundaries for over 90% of seasons, indicating that ‘mega-patches’
were successfully broken into unique regions which remained spatially discrete except during the
largest, most infrequent floods.
Circuit theory analysis
We calculated effective distances between unique habitats using circuit theory, a modelling
approach based on random walks that takes into account all possible movement pathways through a
heterogeneous landscape matrix separating two habitats (McRae et al. 2008). Landscape moisture
has been shown to strongly affect movement and connectivity for aquatic invertebrates (Morán-
Ordóñez et al. 2015), amphibians (Murphy et al. 2010, Watts et al. 2015), and turtles (Patrick et al.
Accepted Article
This article is protected by copyright. All rights reserved.
2012). To account for landscape resistance to movement for a broad range of water-dependent
organisms inhabiting the MDB, we used a generalised resistance layer based on landscape moisture
(Goldberg and Waits 2010). This resistance layer was generated by classifying Australian Dynamic
Land Cover Dataset v1.0 classes (Lymburner et al. 2010) and potentially significant movement
barriers or corridors such as roads, streams and urban areas from Geoscience Australia (2006)
topographic mapping into six exponentially increasing resistance classes based on assumed ground
level moisture content, and combining this with areas that received any inundation during each of
the 99 seasons (Table 1).
Pairwise circuit theory resistances (effective distances that account for landscape resistance to
movement) between neighbouring habitats were calculated for each season using a moving window
approach. To allow connectivity between windows and minimise circuit theory edge effects by
providing a large buffered area relative to the longest investigated inter-patch distance (Pelletier et
al. 2014), we divided the entire study area into 12 x 12 km ‘windows’ (100 x 100 raster pixels) that
overlapped by 50% (McIntyre and Strauss 2013). Pairwise resistance distances for each window were
calculated using the Circuitscape v4.04 software package (Shah and McRae 2008) with seasonal
habitat and resistance layers as inputs (connecting raster pixels to their eight immediate
neighbours). When pairs of habitats were analysed multiple times due to overlapping windows, the
lowest resistance distance was selected to eliminate artificially high resistance values associated
with circuit theory edge effects.
Graph theory network modelling
We generated potential surface-water habitat networks for each of the 99 seasonal time-steps in the
1987–2011 time series using graph theory network modelling, with the matrix of Circuitscape
resistance distances between all neighbouring habitat pairs as an input. We treated unique habitats
as graph nodes, and compared connectivity for two maximum circuit theory resistance distances
converted from Euclidean distance values. These resistance distances (0.53 and 1.61) were identified
by calculating the 95th percentile of circuit theory movement costs associated with maximum
Euclidean dispersal distances of 1000 m and 5000 m (Bishop-Taylor et al. 2015). Distances of 1000 m
and 5000 m were selected to encompass the dispersal abilities of a range of common water-
dependent organisms within the MDB, including aquatic invertebrates (e.g., 99% of movements
below ~1 km for a widespread MDB freshwater crustacean; O’Connor 1986), amphibians (e.g., mean
maximum dispersal distance of ~2 km globally; Smith and Green 2005) and freshwater turtles (e.g.,
dispersal up to ~5 km for a common south-eastern Australian freshwater turtle; Roe et al. 2009).
Many studies have highlighted the importance of habitats with high network centrality for
maintaining and enhancing landscape connectivity (Urban and Keitt 2001, Estrada and Bodin 2008).
We used local graph centrality metrics to evaluate the potential importance of each unique habitat
(node) across each of the 99 seasonal time-steps. Degree centrality (DC) measures the numbers of
edges (connections) between a node and its immediate neighbours. In a habitat network, habitats
with particularly high DC values (e.g. in the top 1% of values) may serve as ‘hub’ habitats whose
location and spatial structure provide abundant opportunities for local dispersal (Estrada and Bodin
2008). Betweenness centrality (BC) quantifies the number of times a node occurs on the shortest
path between any two nodes in a network, with high values (e.g. top 1%) indicating potential
Accepted Article
This article is protected by copyright. All rights reserved.
‘stepping-stone’ habitats vital for enabling long-distance movements through a habitat network. We
calculated DC and BC on networks restricted to each of the two dispersal distances using the Igraph
Python module (Csardi and Nepusz 2006), using resistance distances as edge weights to ensure
shortest paths for BC ran through areas of least resistance to movement.
To identify potential stepping-stones and hubs, we mapped the top 1% of values for each metric and
seasonal time-step by ranking habitats from 0 (low importance) to 1 (high importance). This top 1%
cutoff was selected to identify a potentially manageable subset of important habitats given the large
number of total habitats identified in our study area (i.e. 277,874), and to approximate cutoffs used
to identify stepping-stones and hubs in previous dynamic connectivity studies (e.g. the top 0.7% or
20 out of a maximum of 2955 wetlands by DC and BC used to identify hubs and stepping stones by
Ruiz et al. 2014). Habitats were mapped separately for the driest 25%, average 25–75%, and wettest
25% of seasons by surface water area across the entire MDB to compare how the distribution of top
1% habitats varied spatially and in context of drought and flood. To evaluate whether stepping-
stones and hubs were under- or over-represented within various land-use types, we extracted data
for seven land-use categories based on the 2010–2011 Catchment Scale Land Use of Australia
(CLUM) dataset’s ‘Primary’ categories (ABARES 2014; Table 2) and identified the majority land-use
category for each unique ID habitat using the Extract to Table tool in ArcMap 10.3 (ESRI 2016).
Assuming that 1% of habitats within each land-use type would be in the top 1% of habitats by BC and
DC basin-wide if these important habitats were randomly distributed with respect to land use, we
then calculated the proportion of top 1% basin-wide habitats compared to all habitats within each
land use (i.e., a proportion greater than 1% indicated a higher than expected proportion of stepping-
stones and hubs within a specific land use, whereas a proportion less than 1% indicated stepping-
stones or hubs were underrepresented). This proportion was calculated separately for each seasonal
time-step, and plotted by driest to wettest seasons to evaluate how the representation of stepping-
stones or hubs per land-use category varied along a spectrum from dry to wet conditions.
Results
Spatial distribution of stepping-stones and hubs
The distribution of important habitats for connectivity varied considerably both between dry,
average and wet seasons and by dispersal ability. During dry seasons, stepping-stones (habitats in
the top 1% of BC values) were largely restricted to permanent riverine corridors for both short and
long dispersal distances (i.e. ~1000 m and 5000 m, respectively; Figure 3). These regions included the
floodplains of the perennial River Murray and, to a lesser extent, the Murrumbidgee and Darling
where persistent waterbodies and wetlands supported long-distance connectivity even during dry
conditions. For both dispersal abilities, persistently important stepping-stones occurred along an
~650 km stretch of the Murray between the Ramsar-listed Coorong, Lower Lakes and Murray Mouth
lake and wetland complex (top 1% stepping-stones for up to 94 out of the total 99 seasonal time-
steps) to the confluence of the Murray-Murrumbidgee Rivers (up to 97 seasons), and within and
upstream of the Barmah-Millewa Ramsar site and wetlands of River Murray Reserve (up to 95
seasons). Although most dry-season stepping-stones for long-distance dispersers followed the main
channels of the Murray, Murrumbidgee and Darling Rivers, other important stepping-stones
Accepted Article
This article is protected by copyright. All rights reserved.
occurred within the upper reaches of the Gwydir, Namoi, Border Rivers and Condamine catchments
to the north-east of the MDB, and in highly ephemeral surface-water habitat within the Paroo River
catchment and the Greater Darling Anabranch (Figure 3).
During average and wet seasons, top 1% stepping-stone habitats expanded along the
Murrumbidgee, Darling and other smaller ephemeral river systems and agricultural regions in the
less modified northern MDB (Figure 3). Some of the largest increases in stepping-stone prevalence
for short-distance dispersing organisms occurred within the highly ephemeral and unregulated
Paroo basin. Although a small number of permanent habitats in the Paroo remained connected via
stepping-stones for long-distance dispersal organisms during averagely wet periods, these habitats
were highly disconnected for short-distance dispersers until large increases in connectivity during
wet conditions. Other ephemeral catchments including the Lachlan and Condamine-Balonne saw
large increases in connectivity during only the wettest seasons in the time series for both dispersal
distances but more pronounced for long-distance dispersal organisms. New stepping-stones during
wet periods also occurred away from main river channels within irrigated agricultural regions, such
as within the Riverina region where stepping-stones provided shortcuts for movement between the
floodplains of the upper Murray and Murrumbidgee Rivers (Figure 3). These regions typically served
as stepping-stones only for long-distance dispersers, with stepping-stones for less vagile organisms
being located in closer proximity to river channels.
Hub habitats were less consistent across time and displayed a more dispersed distribution compared
to stepping-stones (Figure 3). During dry seasons, habitats along the Murray and Darling Rivers
frequently served as both stepping-stones and hubs, particularly for short-dispersal organisms.
These habitats included the nationally significant Riverland Wetland Complex and Wallpolla Island
floodplain wetlands along the central Murray (top 1% hubs for 86 and 97 seasons, respectively) and
wetlands to the south of the Ramsar-listed Gunbower-Koondrook-Perricoota Forest group (75
seasons). During wet seasons, hubs for long-distance dispersers were more likely to occur across
extensive floodplain regions away from the immediate river channel, or within irrigated agricultural
regions (e.g., the Lachlan). The prevalence of hubs within irrigated agricultural areas also increased
during average or wet conditions, with some of the largest increases occurring in irrigated areas
within the Gwydir, Namoi, Border Rivers and Condamine catchments in the north-eastern MDB.
Connectivity in natural and modified landscapes
Conservation and natural environment land uses exhibited significantly higher than expected
proportions of stepping-stone habitats for both short- and long-distance dispersal organisms. During
dry seasons, conservation areas contained up to 9.5 times more top 1% BC habitats than expected
(i.e., 9.5% vs 1% expected), while natural environments contained up to 7.1 times (Figure 4).
Although these values decreased during increasingly wet seasons, conservation and natural
environments retained the highest proportion of stepping-stones of any land-use type across the
entire time series. Hub habitats were also consistently overrepresented within conservation areas
(up to 7.6 times more than expected), with no clear trend apparent with increasing inundated
habitat area aside from a gradual decrease for short-distance dispersal organisms across dry (~3.7%),
average (~1.7%) and wet conditions (~1.2%). Although natural environments also saw a steady
decrease in the proportion of stepping-stones during increasingly wet seasons, this proportion never
dropped significantly below the expected 1% of habitats.
Accepted Article
This article is protected by copyright. All rights reserved.
Artificial and modified land-use types typically contained fewer than expected stepping-stone or hub
habitats (Figure 4) despite hosting a far greater proportion of potential surface-water habitat (~87%
by area). Several land-use categories consistently showed significantly lower than expected
proportions of stepping-stones across the time series. This included dryland agriculture (including
cropping and grazing on modified pastures), artificial water (e.g., reservoirs and dams) and intensive
land uses (e.g., urban infrastructure and roads), which exhibited extremely small proportions of
stepping-stones for short-distance dispersers across dry, average and wet seasons (~0–0.5%
compared to 1% expected). Stepping-stone proportions improved marginally for long-distance
dispersers during increasingly wet seasons but remained significantly lower than expected (~0.3–
0.6%). Hub habitats were also underrepresented in these land-use types, although with greater
differences between dispersal distances (higher representation for long-distance dispersal) and with
higher and more variable proportions overall (i.e., typical values between 0 to 0.7% with occasional
seasons up to 2%).
The proportions of stepping-stones and hubs within production from natural environments (e.g.,
forestry and grazing on native vegetation) and irrigated agriculture were higher and showed
stronger relationships with inundated area compared to other modified land uses (Figure 4).
Although both land uses showed lower than expected proportions of stepping-stones and hubs
during dry seasons, increasingly wet conditions saw both stepping-stone and hub proportions
increase to above expected values. Although the increase in stepping-stone and hub proportions for
production from natural environments was modest (increases to 1.1–1.4% and 1.1–1.8% for
stepping-stones and hubs, respectively), the prevalence of this land use across much of the north
and western MDB was sufficient to account for the majority of decreases in proportions observed in
other land-use types during wet seasons (e.g., conservation and natural environments). Increases in
the importance of irrigated agricultural areas with flooding were far more extreme, with several
seasons distributed across the spectrum from average to wet showing extremely high proportions of
top hub habitats (up to 10.1% for long-distance and 3.8% for short-distance dispersal). Despite also
containing a disproportionately high number of stepping-stone habitats for long-distance dispersal
organisms during wet seasons (up to 3%), irrigated agricultural areas contained fewer than expected
stepping-stones and hubs for short-distance dispersers across the majority of the time series.
Discussion
Spatial distribution of stepping-stones and hubs
Landscape connectivity plays a key role in maintaining biodiversity within dynamic dryland
freshwater ecosystems (Davis et al. 2015, Murphy et al. 2015). Conservation and management
approaches aimed at maximising connectivity are likely to grow in importance as habitat loss and a
changing climate require water-dependent organisms to move long distances in search of suitable
habitat or ecological refuges (Nuñez et al. 2013, Davis et al. 2015). However, changing land-use
patterns, intensifying water resource demands and increasingly extreme surface-water dynamics
require approaches to maintaining or enhancing connectivity that account for and embrace rapid
environmental change and temporal variability (Bond et al. 2008). In this study, we used a
comprehensive and spatiotemporally consistent graph theory network analysis approach to model
Accepted Article
This article is protected by copyright. All rights reserved.
landscape connectivity between potential surface-water habitats within Australia’s MDB, a highly
modified semi-arid region containing some of Australia’s most biologically significant wetland
ecosystems (Rogers and Ralph 2010, Pittock and Finlayson 2011). We processed over 5.4 million
nodes across 99 seasonal time-steps and two dispersal abilities, making our study to our knowledge
the largest spatially explicit ecological network analyses yet conducted.
We identified a subset of potential habitats that facilitate high levels of connectivity both at local
(hubs) and at regional or network scales (stepping-stones). Prioritising surface-water habitats for
conservation based on their potential for improving connectivity through habitat networks may
provide an opportunity for maximising benefits from limited conservation funding and reducing
conflict in regions like the MDB faced with changing land-use patterns and increasing demands for
water resources. In the MDB, the importance of surface-water habitats as connectivity providers
depended on both their spatial position in a habitat network and their persistence and dynamics
over time. Large, persistent hub habitats such as those within the Coorong, Lower Lakes and Murray
Mouth lake and wetland complex may serve as important ecological refuges, allowing organisms to
survive periods of extreme environmental conditions, including drought. However, landscape
connectivity and processes of dispersal into and out of habitats during and after disturbances are
critical to refuge function (Chester and Robson 2013). Persistent habitats with highly suitable local-
scale habitat attributes (e.g., dense aquatic or riparian vegetation and a lack of predators) may be
unable to serve as refuges if they are functionally isolated from other habitats either by geographic
distance or by anthropogenic barriers (Sheldon et al. 2010, Davis et al. 2013). Conversely, hub
habitats may be less likely to serve as functional refuges if they are present in the landscape for a
short period, or if they are highly degraded or otherwise unsuitable for organisms to establish
populations or reproduce. Our results therefore do not reduce the importance of collecting site-
scale ecological data, but can complement field studies with insights on habitat importance informed
by a habitat’s context within its larger-scale habitat network.
Stepping-stone habitats that support long-distance dispersal though habitat networks will perform
different ecological roles depending on their permanence and relationship with flooding (Bishop-
Taylor et al. 2015). Many stepping-stone habitats in the MDB were located within riverine
floodplains, where a matrix of permanent and temporary habitats facilitated longitudinal
connectivity during even periods of severe drought. By providing connections between clusters of
habitats during dry seasons, these stepping-stones may enhance the ability of locally well-connected
hubs to serve as functional ecological refuges. This was particularly apparent along the lower Murray
River, where perennial stepping-stone habitats provided continuous opportunities for movement
between persistent hub habitats within the Lower Lakes and the floodplains of the central Murray.
However, other more ephemeral stepping-stones appeared during major flooding across extensive
natural floodplains of the northern MDB including within the Condamine-Balonne and Paroo
catchments. By creating both new temporary surface-water habitat and stepping-stones connecting
existing isolated habitats, transient flooding events in these regions are likely to play a key role in
maintaining regional-scale landscape connectivity.
Accepted Article
This article is protected by copyright. All rights reserved.
Connectivity in conservation and natural environments
Conservation areas and natural environments in the MDB contained significantly higher than
expected proportions of important stepping-stones and hubs throughout almost all seasonal time-
steps in our 25-year time series. Many of these regions contain surface-water habitats already
regarded as significant at a national (e.g., Directory of Important Wetlands in Australia) or global
(i.e., Ramsar Wetlands of International Importance) scale, of which some have served as previous
targets for the allocation of environmental water aimed at maintaining or improving ecological
health of habitats stressed by drought. Along the Murray River, several of the stepping-stones and
hubs identified by this study (e.g., within the Lower Lakes, Coorong and Murray Mouth, Gunbower-
Koondrook-Perricoota Forests, Lindsay-Wallpolla-Mulcra Islands and along the River Murray
Channel) have been allocated proportions of over 500 gigaliters of environmental water under The
Living Murray program, in part to improve connectivity between floodplain and waterways and
maintain ecological refuges during the worst of the 1999–2010 Millennium Drought (MDBC 2005).
The provision of strategic environmental flows to high-priority stepping-stone and hub habitats will
be important to maintain ecological connections between both local and distant clusters of habitats,
and may assist in greatly reducing the adverse impacts of severe drought on water-dependent
species (Arthington 2012).
The high importance of conservation and natural environments corroborates previous static
connectivity modelling in the Murray-Darling that found individual habitats located within protected
areas displayed higher BC and DC values than unprotected habitats, especially during large modelled
flooding scenarios (Bishop-Taylor et al. 2015). However, our dynamic study showed that while
individual habitats within conservation areas may be important for maintaining connectivity during
flooding, they represent an increasingly small proportion of total stepping-stones and hubs across
the entire MDB during wet seasons. This is likely to reflect two major underlying processes: a basin-
wide shift in important habitats towards the less-modified northern MDB during wet periods, and a
local shift away from immediate river channels and into more extensive floodplain environments.
Conservation areas cover a relatively small proportion of the MDB (~7.5% by area), with protected
rivers and wetlands exhibiting a strong bias towards the southern MDB (Bino et al. 2016). Although
this spatial distribution provides disproportionally high protection to persistent stepping-stones and
hubs located along the floodplains of the Murray River, conservation areas provide less protection
for more transiently important habitats that appear across the northern MDB during major
inundation events.
Increasing the regional- or continental-scale effectiveness of the Australian protected area network
for preserving freshwater biodiversity is likely to require an increase in the representation of
stepping-stones and hubs that are important for facilitating landscape connectivity across the
northern MDB. Many of these surface-water habitats are located in relatively natural environments
used predominantly for grazing and other forms of low intensity production, and so may still provide
local-scale habitat attributes required to support populations of water-dependent species, provided
they are protected from future land-use intensification (Chester and Robson 2013). Where protected
area listings are not compatible with increasingly intensive land uses, increased management may be
required to ensure that potential stepping-stones and hubs can support long-distance movement
through ecological networks, or serve as ecologically functional refuges. A range of management
techniques have shown promise in improving the suitability of artificial surface water to serve as
Accepted Article
This article is protected by copyright. All rights reserved.
ecological refuges or stepping-stones. These include promoting riparian and aquatic plant growth by
reducing the steepness of dam or canal edges (e.g., Robson and Clay 2005, Hamer et al. 2012),
fencing to minimise livestock erosion (e.g., Markwell and Fellows 2008, Canals et al. 2011) and
attempts to reduce predation by removing introduced fish species (e.g., Vredenburg 2004). The
stepping-stones and hubs identified by this study may provide priority targets for management,
allowing limited conservation funding to be allocated to habitats that serve as the most effective
connectivity providers.
Connectivity in modified environments
Surface water within irrigated agricultural areas in the MDB served as important stepping-stones and
hubs during average and wet seasons, but predominantly for organisms with long-distance dispersal
abilities. These landscapes were typically characterised by a shifting mosaic of uniformly distributed,
temporarily flooded fields and irrigation canals. Although this habitat structure did not prevent inter-
patch movement for long-distance dispersers, potential irrigated agricultural habitats remained
disconnected for short-distance dispersal organisms less able to move between the spatially
dispersed areas of surface water present at any moment in time. Other less transient artificial water
sources (i.e., farm dams and reservoirs) also displayed significantly lower than expected connectivity
across the majority of the time series, particularly for short dispersal distances (Figure 4). A similar
result was observed by Uden et al. (2014) in a graph theory study of amphibian habitat networks in
the Rainwater Basin (central USA), where smaller, shorter-distance dispersers displayed sharply
reduced connectivity in agricultural landscapes dominated by uniformly distributed irrigation water
storage pits.
These findings suggest that the regular, dispersed spatial structure of surface-water habitats in
highly modified agricultural landscapes disproportionately affect connectivity for short-distance
dispersers compared to the tightly clustered matrix of ephemeral and permanent floodplain and
wetland habitats they have replaced. Although organisms with long-distance dispersal abilities (i.e.,
dispersal distances of ~5000 m and above) may be able to successfully move through these
landscapes, less vagile organisms may suffer increased genetic isolation and heightened risks of local
extinction events due to reduced opportunities for dispersal (Johst et al. 2002). Hydrologic
management practices (i.e., maintaining a subset of fields flooded during fallow phases) aimed at
mimicking natural mosaics of temporary and permanent habitats have shown success in maximising
aquatic macroinvertebrate biodiversity (e.g., Stenert et al. 2009), and may be required to enhance
the persistence of short-distance dispersal species in these modified landscapes. In addition to
enhancing connectivity, managing artificial surface-water hydroperiods may also provide ecological
benefits at site scales, including ensuring that during even dry periods aquatic organisms such as
amphibians or invertebrates have continuous access to relatively persistent surface water in which
to complete larval or juvenile stages of their lifecycles (Wassens et al. 2008, Canals et al. 2011).
Limitations and future work
By allowing us to model spatiotemporal connectivity dynamics consistently across space and over a
~25-year period, our application of remotely sensed surface-water datasets advances previous graph
theory network analyses that relied on modelled flooding scenarios (e.g., Bishop-Taylor et al. 2015)
or historical and hypothetical future wetland maps (e.g., Uden et al. 2014). Nevertheless, some
Accepted Article
This article is protected by copyright. All rights reserved.
important caveats in our approach should be noted. As our work was based on remote sensing
datasets, our findings are conditional on both the accuracy and availability of input data. This is
particularly significant for graph theory network analyses that can be highly sensitive to variation in
the number and distribution of nodes (i.e., habitats) used to model connectivity (Butts 2009). The
surface-water time series we used to define potential habitats through time was subject to a
statistically rigorous accuracy assessment (Tulbure et al. 2016), with a high resulting accuracy for
surface water during both wet and dry years (producer's accuracy of 87% ± 3% SE). In addition,
temporally aggregating surface water by ensuring that a cell was flooded for at least 50% of each
seasonal time-step is likely to have significantly reduced errors of commission by eliminating noisy
pixels that may have only occurred once throughout the entire time series.
Where minimal satellite coverage or extended presence of clouds resulted in areas with no data that
persisted throughout an entire season, we used a conservative gap-filling method in which cells
were filled based on the nearest seasons with data. As reported by Tulbure et al. (2016), large areas
of no data in the MDB surface-water time series were concentrated in a minority of (mostly winter)
seasons towards both ends of the time series, minimising their influence on the analysis. The
resulting surface-water layers are expected to be far more accurate than previous static layers used
to model connectivity in the MDB, such as the Geoscience Australia (GA 2006) topographic
waterbody features used by Bishop-Taylor et al. (2015), which have coarse resolution (1:250,000), a
highly variable provenance (1965–2004 in a single static dataset) and inconsistent attribution (Bino
et al., 2016). Our use of freely available satellite data with global coverage (i.e., Landsat TM and
ETM+) also ensures our approach is generalizable to other regions, allowing our findings to be
compared and assessed consistently against other dynamic surface-water environments.
To focus on identifying habitats whose location within surface-water habitat networks was
important for facilitating landscape connectivity, we limited this study to using centrality metrics (DC
and BC) that quantified only network topology or structure. However, our flexible graph theory
framework supports future approaches that could make use of advanced metrics which additionally
incorporate habitat quantity, quality or the relative probabilities of movement between two
habitats. In particular, metrics based on the concept of habitat availability or reachability including
the Integral Index of Connectivity and Probability of Connectivity have been empirically shown to
provide information on the importance of habitats for connectivity that complements and improves
upon centrality methods (Saura and Pascual-Hortal 2007, Saura et al. 2011). Combining dynamic
surface-water data with these advanced methods would allow interactions between flooded area
and regional-scale connectivity to be compared holistically across time and space, potentially
providing valuable insights into the structure and resilience of natural and artificial surface-water
habitat networks during periods of extreme hydroclimatic variability.
Acknowledgements
This work was funded through an Australian Research Council Discovery Early Career Researcher
Award (DE140101608) to Tulbure. We thank B McRae for his timely assistance with the Circuitscape
software package.
Accepted Article
This article is protected by copyright. All rights reserved.
Literature cited
ABARES (Australian Bureau of Agricultural and Resource Economics and Sciences). 2014. Catchment
Scale Land Use of Australia - Update March 2014. ABARES, Canberra, ACT.
Arthington, A. H. 2012. Environmental Flows: Saving Rivers in the Third Millennium. University of
California Press, Berkeley, CA.
Ballinger, A., and R. C. Mac Nally. 2006. The landscape context of flooding in the Murray-Darling
Basin. Floods in an Arid Continent (Advances in Ecological Research) 39:85–105.
Berney, P., and T. Hosking. 2016. Opportunities and challenges for water-dependent protected area
management arising from water management reform in the Murray-Darling Basin: a case study from
the Macquarie Marshes in Australia. Aquatic Conservation: Marine and Freshwater Ecosystems
26:12–28.
Bino, G., R. T. Kingsford, and K. Brandis. 2016. Australia’s wetlands – learning from the past to
manage for the future 22:116–129.
Bino, G., R. T. Kingsford, and J. Porter. 2015. Prioritizing wetlands for waterbirds in a boom and bust
system: Waterbird refugia and breeding in the Murray-Darling Basin. PLoS ONE 10:e0132682
Bishop-Taylor, R., M. G. Tulbure, and M. Broich. 2015. Surface water network structure, landscape
resistance to movement and flooding vital for maintaining ecological connectivity across Australia’s
largest river basin. Landscape Ecology 30:2045–2065.
Blondel, V. D., J.-L. Guillaume, R. Lambiotte, and E. Lefebvre. 2008. Fast unfolding of communities in
large networks. Journal of Statistical Mechanics: Theory and Experiment 10:6.
Bond, N. R., P. S. Lake, and A. H. Arthington. 2008. The impacts of drought on freshwater
ecosystems: An Australian perspective. Hydrobiologia 600:3–16
Butts, C. T. 2009. Revisiting the foundations of network analysis. Science 325:414–6.
Canals, R. M., V. Ferrer, A. Iriarte, S. Cárcamo, L. S. Emeterio, and E. Villanueva. 2011. Emerging
conflicts for the environmental use of water in high-valuable rangelands. Can livestock water ponds
be managed as artificial wetlands for amphibians? Ecological Engineering 37:1443–1452.
Cavanaugh, K. C., D. A. Siegel, P. T. Raimondi, and F. Alberto. 2014. Patch definition in
metapopulation analysis: A graph theory approach to solve the mega-patch problem. Ecology
95:316–328.
Chester, E. T., and B. J. Robson. 2013. Anthropogenic refuges for freshwater biodiversity: Their
ecological characteristics and management. Biological Conservation 166:64–75.
Collins, S. D., L. J. Heintzman, S. M. Starr, C. K. Wright, G. M. Henebry, and N. E. McIntyre. 2014.
Hydrological dynamics of temporary wetlands in the southern Great Plains as a function of
surrounding land use. Journal of Arid Environments 109:6–14.
Crook, D. A., W. H. Lowe, F. W. Allendorf, T. Erős, D. S. Finn, B. M. Gillanders, W. L. Hadwen, C.
Accepted Article
This article is protected by copyright. All rights reserved.
Harrod, V. Hermoso, S. Jennings, R. W. Kilada, I. Nagelkerken, M. M. Hansen, T. J. Page, C. Riginos, B.
Fry, and J. M. Hughes. 2015. Human effects on ecological connectivity in aquatic ecosystems:
Integrating scientific approaches to support management and mitigation. Science of the Total
Environment 534:52–64.
Csardi, G., and T. Nepusz. 2006. The igraph software package for complex network research.
InterJournal Complex Systems:1695.
Davidson, N. C. 2014. How much wetland has the world lost? Long-term and recent trends in global
wetland area. Marine and Freshwater Research 65:934–941.
Davis, J., A. P. O’Grady, A. Dale, A. H. Arthington, P. A. Gell, P. D. Driver, N. Bond, M. Casanova, M.
Finlayson, R. J. Watts, S. J. Capon, I. Nagelkerken, R. Tingley, B. Fry, T. J. Page, and A. Specht. 2015.
When trends intersect: The challenge of protecting freshwater ecosystems under multiple land use
and hydrological intensification scenarios. Science of the Total Environment 534:65–78.
Davis, J., A. Pavlova, R. Thompson, and P. Sunnucks. 2013. Evolutionary refugia and ecological
refuges: key concepts for conserving Australian arid zone freshwater biodiversity under climate
change. Global Change Biology 19:1970–1984.
Dilts, T. E., P. J. Weisberg, P. Leitner, M. D. Matocq, R. D. Inman, K. E. Nussear, and T. C. Esque. 2016.
Multiscale connectivity and graph theory highlight critical areas for conservation under climate
change. Ecological Applications 26:1223–1237.
ESRI. 2016. ArcGIS Desktop: Release 10.3. Environmental Systems Research Institute, Redlands, CA.
Estrada, E., and Ö. Bodin. 2008. Using network centrality measures to manage landscape
connectivity. Ecological Applications 18:1810–1825.
Finlayson, C. M., J. A. Davis, P. A. Gell, R. T. Kingsford, and K. A. Parton. 2011. The status of wetlands
and the predicted effects of global climate change: the situation in Australia. Aquatic Sciences
75:73–93.
Fortuna, M. A., C. Gómez-Rodríguez, and J. Bascompte. 2006. Spatial network structure and
amphibian persistence in stochastic environments. Proceedings of the Royal Society B: Biological
Sciences 273:1429–1434.
GDAL Development Team. 2015. GDAL - Geospatial Data Abstraction Library, version 1.11.1. Open
Source Geospatial Foundation.
Geoscience Australia. 2006. GEODATA TOPO 250K Series 3 (Packaged - Shape file format).
Geoscience Australia, Canberra, ACT.
Goldberg, C. S., and L. P. Waits. 2010. Comparative landscape genetics of two pond-breeding
amphibian species in a highly modified agricultural landscape. Molecular Ecology 19:3650–63.
Hamer, A. J., P. J. Smith, and M. J. McDonnell. 2012. The importance of habitat design and aquatic
connectivity in amphibian use of urban stormwater retention ponds. Urban Ecosystems 15:451–471.
Johst, K., R. Brandl, and S. Eber. 2002. Metapopulation persistence in dynamic landscapes: the role
Accepted Article
This article is protected by copyright. All rights reserved.
of dispersal distance. Oikos 2:263–270.
Jones, H. A. 2010. Crustaceans and molluscs. Pages 275–310 in K. Rogers and T. J. Ralph, editors.
Floodplain wetland biota in the Murray-Darling Basin: Water and habitat requirements. CSIRO
Publishing, Canberra, ACT.
Kennett, R. M., and A. Georges. 1990. Habitat utilization and its relationship to growth and
reproduction of the Eastern Long-Necked Turtle, Chelodina longicollis (Testudinata: Chelidae), from
Australia. Herpetologica 46:22–33.
Kramer-Schadt, S., T. Kaiser, K. Frank, and T. Wiegand. 2011. Analyzing the effect of stepping stones
on target patch colonisation in structured landscapes for Eurasian lynx. Landscape Ecology 26:501–
513.
Leblanc, M., S. Tweed, A. Van Dijk, and B. Timbal. 2012. A review of historic and future hydrological
changes in the Murray-Darling Basin. Global and Planetary Change 80–81:226–246.
Lymburner, L., P. Tan, N. Mueller, R. Thackway, A. Lewis, M. Thankappan, and U. Senarath. 2010. The
National Dynamic Land Cover Dataset. National Earth Observation Group, Geoscience Australia,
Canberra.
Mac Nally, R., S. Nerenberg, J. R. Thomson, H. Lada, and R. H. Clarke. 2014. Do frogs bounce, and if
so, by how much? Responses to the “Big Wet” following the “Big Dry” in south-eastern Australia.
Global Ecology and Biogeography 23:223–234.
Markwell, K. A., and C. S. Fellows. 2008. Habitat and biodiversity of on-farm water storages: A case
study in Southeast Queensland, Australia. Environmental Management 41:234–249.
McGuire, J. L., J. J. Lawler, B. H. McRae, T. Nuñez, and D. M. Theobald. 2016. Achieving climate
connectivity in a fragmented landscape. Proceedings of the National Academy of Sciences of the
United States of America 113.
McIntyre, N. E., and R. E. Strauss. 2013. A new, multi-scaled graph visualization approach: an
example within the playa wetland network of the Great Plains. Landscape Ecology 28:769–782.
McIntyre, N. E., C. K. Wright, S. Swain, K. Hayhoe, G. Liu, F. W. Schwartz, and G. M. Henebry. 2014.
Climate forcing of wetland landscape connectivity in the Great Plains. Frontiers in Ecology and the
Environment 12:59–64.
McRae, B. H., B. G. Dickson, T. H. Keitt, and V. B. Shah. 2008. Using circuit theory to model
connectivity in ecology, evolution, and conservation. Ecology 89:2712–2724.
MDBA. 2010. Guide to the proposed Basin Plan: overview. Murray-Darling Basin Authority, Canberra,
ACT.
MDBC. 2005. Living Murray Business Plan. Murray-Darling Basin Commission, Canberra, ACT.
Minor, E. S., and D. L. Urban. 2007. Graph theory as a proxy for spatially explicit population models
in conservation planning. Ecological Applications 17:1771–1782.
Accepted Article
This article is protected by copyright. All rights reserved.
Minor, E. S., and D. L. Urban. 2008. A graph-theory framework for evaluating landscape connectivity
and conservation planning. Conservation Biology 22:297–307.
Morán-Ordóñez, A., A. Pavlova, A. M. Pinder, L. Sim, P. Sunnucks, R. M. Thompson, and J. Davis.
2015. Aquatic communities in arid landscapes: local conditions, dispersal traits and landscape
configuration determine local biodiversity. Diversity and Distributions 10:1230-1241
Murphy, A. L., A. Pavlova, R. Thompson, J. Davis, and P. Sunnucks. 2015. Swimming through sand:
Connectivity of aquatic fauna in deserts. Ecology and Evolution 5:5252–5264.
Murphy, M. A., J. S. Evans, and A. Storfer. 2010. Quantifying Bufo boreas connectivity in Yellowstone
National Park with landscape genetics. Ecology 91:252–261.
Newman, M. 2006. Modularity and community structure in networks. Proceedings of the National
Academy of Sciences 103:8577–82.
Nuñez, T. A., J. J. Lawler, B. H. McRae, D. J. Pierce, M. B. Krosby, D. M. Kavanagh, P. H. Singleton, and
J. J. Tewksbury. 2013. Connectivity planning to address climate change. Conservation Biology
27:407–416.
O’Connor, P. 1986. The biology of the Murray crayfish, Euastacus armatus (Decapoda: Parastacidae)
and recommendations for the future management of the fishery (unpublished NSW Department of
Agriculture data summary). NSW Fisheries, Narrandera, NSW.
O’Farrill, G., K. G. Schampaert, B. Rayfield, Ö. Bodin, S. Calme, R. Sengupta, and A. Gonzalez. 2014.
The potential connectivity of waterhole networks and the effectiveness of a protected area under
various drought scenarios. PloS one 9:e95049.
Patrick, D. A., J. P. Gibbs, V. D. Popescu, and D. A. Nelson. 2012. Multi-scale habitat-resistance
models for predicting road mortality “hotspots” for turtles and amphibians. Herpetological
Conservation and Biology 7:407–426.
Pelletier, D., M. Clark, M. G. Anderson, B. Rayfield, M. A. Wulder, and J. A. Cardille. 2014. Applying
circuit theory for corridor expansion and management at regional scales: tiling, pinch points, and
omnidirectional connectivity. PloS one 9:e84135.
Pittock, J., and C. M. Finlayson. 2011. Australia’s Murray - Darling Basin: freshwater ecosystem
conservation options in an era of climate change. Marine and Freshwater Research 62:232.
Pringle, C. M. 2001. Hydrologic connectivity and the management of biological reserves: a global
perspective. Ecological Applications 11:981–998.
Python Software Foundation. 2015. Python Language Reference, version 2.6.
Robson, B. J., and C. J. Clay. 2005. Local and regional macroinvertebrate diversity in the wetlands of a
cleared agricultural landscape in south-western Victoria, Australia. Aquatic Conservation: Marine
and Freshwater Ecosystems 15:403–414.
Roe, J. H., A. C. Brinton, and A. Georges. 2009. Temporal and spatial variation in landscape
connectivity for a freshwater turtle in a temporally dynamic wetland system. Ecological Applications
Accepted Article
This article is protected by copyright. All rights reserved.
19:1288–99.
Rogers, K., and T. J. Ralph. 2010. Floodplain wetlands of the Murray-Darling Basin and their
freshwater biota. Pages 1–16in K. Rogers and T. J. Ralph, editors.Floodplain Wetland Biota in the
Murray-Darling Basin: Water and Habitat Requirements. CSIRO Publishing, Collingwood, Victoria.
Roshier, D. A., P. H. Whetton, R. J. Allan, and A. I. Robertson. 2001. Distribution and persistence of
temporary wetland habitats in arid Australia in relation to climate. Austral Ecology 26:371–384.
Rudnick, D., S. Ryan, P. Beier, S. A. Cushman, F. Dieffenbach, C. Epps, L. Gerber, J. Hartter, J. Jenness,
J. Kintsch, A. Merenlender, R. Perkl, D. Preziosi, and S. Trombulak. 2012. The role of landscape
connectivity in planning and implementing conservation and restoration priorities. Issues in Ecology
16:1–20
Ruiz, L., N. Parikh, L. J. Heintzman, S. D. Collins, S. M. Starr, C. K. Wright, G. M. Henebry, N. van
Gestel, and N. E. McIntyre. 2014. Dynamic connectivity of temporary wetlands in the southern Great
Plains. Landscape Ecology 29:507–516.
Saunders, M. I., C. J. Brown, M. M. Foley, C. M. Febria, R. Albright, M. Mehling, M. T. Kavanaugh, and
D. D. Burfeind. 2015. Human impacts on connectivity in marine and freshwater ecosystems assessed
using graph theory: a review. Marine and Freshwater Research:1–2.
Saura, S., Ö. Bodin, and M.-J. Fortin. 2014. Stepping stones are crucial for species’ long-distance
dispersal and range expansion through habitat networks. Journal of Applied Ecology 51:171–182.
Saura, S., C. Estreguil, C. Mouton, and M. Rodríguez-Freire. 2011. Network analysis to assess
landscape connectivity trends: Application to European forests (1990–2000). Ecological Indicators
11:407–416.
Saura, S., and L. Pascual-Hortal. 2007. A new habitat availability index to integrate connectivity in
landscape conservation planning: Comparison with existing indices and application to a case study.
Landscape and Urban Planning 83:91–103.
Shah, V. B., and B. H. McRae. 2008. Circuitscape: a tool for landscape ecology. Pages 62–66 in G.
Varoquaux, T. Vaught, and J. Millman, editors. Proceedings of the 7th Python in Science Conference
(SciPy 2008). Pasadena, CA.
Sheldon, F., S. E. Bunn, J. M. Hughes, A. H. Arthington, S. R. Balcombe, and C. S. Fellows. 2010.
Ecological roles and threats to aquatic refugia in arid landscapes: Dryland river waterholes. Marine
and Freshwater Research 61:885–895.
Smith, M. A., and D. M. Green. 2005. Dispersal and the metapopulation paradigm in amphibian
ecology and conservation: Are all amphibian populations metapopulations? Ecography 28:110–128.
Stenert, C., R. C. Bacca, L. Maltchik, and O. Rocha. 2009. Can hydrologic management practices of
rice fields contribute to macroinvertebrate conservation in southern Brazil wetlands? Hydrobiologia
635:339–350.
Suski, C. D., and S. J. Cooke. 2006. Conservation of aquatic resources through the use of freshwater
Accepted Article
This article is protected by copyright. All rights reserved.
protected areas: opportunities and challenges. Biodiversity and Conservation 16:2015–2029.
Thompson, M. B. 1983. The physiology and ecology of the eggs of the pleurodiran tortoise Emydura
macquarii (Gray), 1831. PhD thesis. University of Adelaide.
Tulbure, M. G., and M. Broich. 2013. Spatiotemporal dynamic of surface water bodies using Landsat
time series data from 1999 to 2011. ISPRS Journal of Photogrammetry and Remote Sensing 79:44–
52.
Tulbure, M. G., M. Broich, S. V. Stehman, and A. Kommareddy. 2016. Surface water extent dynamics
from three decades of seasonally continuous Landsat time series at subcontinental scale in a semi-
arid region. Remote Sensing of Environment 178:142–157.
Tulbure, M. G., S. J. Kininmonth, and M. Broich. 2014. Spatiotemporal dynamics of surface water
networks across a global biodiversity hotspot—implications for conservation. Environmental
Research Letters 9:114012.
Uden, D., M. Hellman, D. G. Angeler, and C. R. Allen. 2014. The role of reserves and anthropogenic
habitats for functional connectivity and resilience of ephemeral wetlands. Ecological Applications
24:1569–1582.
Urban, D. L., and T. H. Keitt. 2001. Landscape connectivity: a graph-theoretic perspective. Ecology
82:1205–1218.
Vörösmarty, C. J., P. B. McIntyre, M. O. Gessner, D. Dudgeon, a Prusevich, P. Green, S. Glidden, S. E.
Bunn, C. a Sullivan, C. R. Liermann, and P. M. Davies. 2010. Global threats to human water security
and river biodiversity. Nature 467:555–561.
Vredenburg, V. T. 2004. Reversing introduced species effects: Experimental removal of introduced
fish leads to rapid recovery of a declining frog. Proceedings of the National Academy of Sciences of
the United States of America 101:7646–7650.
Walt, S. van der, S. C. Colbert, and G. Varoquaux. 2011. The NumPy array: a structure for efficient
numerical computation. Computing in Science & Engineering 13.
Ward, J. V., and J. A. Stanford. 1995. Ecological connectivity in alluvial river ecosystems and its
disruption by flow regulation. Regulated Rivers: Research & Managments 11:105–119.
Wassens, S. 2010. Frogs. Pages 253–274 in K. Rogers and T. J. Ralph, editors. Floodplain Wetland
Biota in the Murray-Darling Basin: Water and Habitat Requirements. CSIRO Publishing, Canberra,
ACT.
Wassens, S., R. J. Watts, A. Jansen, and D. Roshier. 2008. Movement patterns of southern bell frogs
(Litoria raniformis) in response to flooding. Wildlife Research 35:50–58.
Watts, A. G., P. Schlichting, S. Billerman, B. Jesmer, M. Fortin, C. Funk, P. Hapeman, E. L. Muths, and
A. Melanie. 2015. How spatio-temporal habitat connectivity affects amphibian genetic structure.
Frontiers in Genetics 6:275
Williams, W. D. 1999. Conservation of wetlands in drylands: A key global issue. Aquatic
Accepted Article
This article is protected by copyright. All rights reserved.
Conservation: Marine and Freshwater Ecosystems 9:517–522.
Wright, C. K. 2010. Spatiotemporal dynamics of prairie wetland networks: power-law scaling and
implications for conservation planning. Ecology 91:1924–1930.
Zeigler, S. L., and W. F. Fagan. 2014. Transient windows for connectivity in a changing world.
Movement Ecology 2:1.
Data Availability
Data associated with this paper have been deposited in the Dryad digital repository
http://dx.doi.org/10.5061/dryad.qf83q
Tables
Table 1 Resistance to movement values used to develop resistance surfaces (based on GA
topographic features and Australian Dynamic Land Cover Dataset v1.0 classes; GA 2006, Lymburner
et al. 2010)
Value Resistance Example DLCD land cover or GA topographic features
1 Very low
Surface water (including seasonal maximum flooding
extent), aquatic vegetation
2 Low Closed natural vegetation, irrigated cropping and pasture
4 Moderate Open natural vegetation, dryland pasture
8 Moderate-high Sparse natural vegetation, dryland cropping, minor roads
16 High Non-vegetated areas, major roads
32 Very high Urban infrastructure and highways, saline waterbodies
Accepted Article
This article is protected by copyright. All rights reserved.
Table 2 Land-use categories used to compare distributions of stepping-stones and hubs, based on
the 2010–2011 Catchment Scale Land Use Map (CLUM) classes (ABARES 2014)
Land-use category CLUM codes Example CLUM land uses
Conservation 1.1.1–1.1.7 Protected areas
Natural environments 1.2.0–1.3.4, 6.0.0–6.1.3,
6.3.0–6.3.3, 6.5.0–6.6.3
Non-protected but natural
vegetation or surface water
Production from natural
environments 2.1.0–2.2.2 Grazing on relatively natural
vegetation, forestry
Dryland agriculture 3.0.0–3.6.4 Dryland cropping, modified
pastures
Irrigated agriculture 4.0.0–4.6.3 Irrigated cropping and horticulture
Intensive uses 5.0.0–5.9.5 Urban areas, mining
Artificial water 6.2.0–6.2.3, 6.4.0–6.4.3 Reservoirs, farm dams
Accepted Article
This article is protected by copyright. All rights reserved.
Figures
Figure 1 Murray-Darling Basin study area, showing the distribution and frequency of potential
surface-water habitats across the entire 99 season time series (dark blue indicates persistent
habitats; Tulbure et al. 2016). Numbered annotations highlight significant regions of the MDB
discussed in this study: (1) Lower Lakes, Coorong and Murray Mouth; (2) the River Murray Channel
including Riverland Wetland Complex and Lindsay-Wallpolla-Mulcra Islands; (3) Gunbower-
Koondrook-Perricoota and Barmah-Millewa Forests; (4) Greater Darling Anabranch; (5) the Riverina
agricultural region; (6) irrigated agriculture and floodplains of the Condamine, Gwydir, Namoi and
Border Rivers catchments; and (7) floodplains of the Condamine-Balonne catchment
Figure 2 Example of process followed to break potential habitat areas into discrete ‘communities’ of
raster pixels (unique ID regions or graph nodes) that consistently occurred together across time. For
all neighbouring pairs of pixels within the study area, we assigned weights depending on their
asynchrony across the entire 1987–2011 time series (A). These weights (B; yellow to red indicating
pixel pairs that rarely occurred together in a single time-step) were then used to identify (C)
potential unique ID regions (graph nodes) that typically occurred as discrete areas of habitat except
during the largest flooding events.
Figure 3 Distribution of top 1% stepping-stones and hubs across the MDB. Important habitats are
shown separately for two dispersal abilities (short-distance, ~1000 m; long-distance, ~ 5000 m) and
the driest 25%, average (25–75%) and the wettest 25% of seasons by inundated habitat area.
Figure 4 Top 1% stepping-stones and hubs basin-wide as a proportion of all habitats within each
land-use category in the MDB. Proportions are plotted against inundated habitat area (percent rank)
for two dispersal abilities (short-distance in orange, ~1000 m; long-distance in blue, ~ 5000 m), and
compared against the expected proportion of 1% (values above 1% indicate more stepping-stones or
hubs than expected within a land-use category).
Accepted Article
This article is protected by copyright. All rights reserved.
Accepted Article
This article is protected by copyright. All rights reserved.
Accepted Article
This article is protected by copyri
g
ht. All ri
g
hts reserved.
Accepted Article
This article is protected by copyright. All rights reserved.
