Professional Documents
Culture Documents
Abstract:
The term connectivity is increasingly being applied in hydrological and geomorphological studies. Relevant research
encompasses aspects of landscape connectivity, hydrological connectivity and sedimentological connectivity. Unlike other
disciplines, notably ecology, published studies show no consensus on a standard definition. This paper provides an overview
of how existing research relates to the concept of connectivity in both ecology and hydrology by proposing and evaluating
a conceptual model of hydrological connectivity that includes five major components: climate; hillslope runoff potential;
landscape position; delivery pathway and lateral connectivity. We also evaluate a proposed measure of connectivity called the
volume to breakthrough to quantify changing connectivity between different environments and catchments. Copyright 2007
John Wiley & Sons, Ltd.
KEY WORDS
INTRODUCTION
The concept of connectivity is increasingly being
applied within a range of disciplines in the Earth and
Environmental Sciences as researchers recognize the need
to move beyond the traditional view that runoff is generated by either Hortonian infiltration excess or by the
variable source area (VSA) model. Recently, the VSA
model itself has been the focus of critique, with calls
for a new theory of runoff generation (e.g. McDonnell, 2003; Ambroise, 2004). Hydrological connectivity
is one possible concept of runoff generation and flood
production that could provide a way forward, but currently there is much confusion in the literature about
how the term is used and how it relates to existing
research.
In hydrology and geomorphology we can identify
three major types of connectivity, broadly classified
as: (1) landscape connectivity, which relates to the physical coupling of landforms (e.g. hillslope to channel)
within a drainage basin; (2) hydrological connectivity,
which refers to the passage of water from one part of
the landscape to another and is expected to generate some
catchment runoff response; and (3) sedimentological connectivity, which relates to the physical transfer of sediments and attached pollutants through the drainage basin
and may vary considerably with, amongst others, particle
size. Work on coupling of hillslopechannel systems (e.g.
Harvey, 1996; Michaelides and Wainwright, 2002) has
* Correspondence to: Louise J. Bracken, Department of Geography, University of Durham, Durham DH1 3LE, UK.
E-mail: l.j.bracken@durham.ac.uk
Ne
e Bull.
Copyright 2007 John Wiley & Sons, Ltd.
1750
(a)
Ru
no
ff
Deposited fine
sediment
Mean river level
(b)
Th
(belo roughflow
w wa
ter ta
ble)
Figure 1. Illustrations of connectivity: (a) hypothetical models to illustrate simple channel coupling (modified from Michaelides and Wainwright
(2003)); (b) a more complex floodplain environment that may limit connectivity
1751
CONNECTIVITY IN ECOLOGY
Ecologists have been using the concept of connectivity
for a number of years as a critical property in the persistence of spatially structured populations (Metzger and
Decamps, 1997). In this context, hydrological connectivity is defined by Pringle (2001) as the water-mediated
transfer of matter, energy, and/or organisms within or
between elements of the hydrologic cycle. This definition, although encompassing other material fluxes as
well as water, bears some similarity to the application
of the term hydrological connectivity. Using this more
rigorous definition, numerous ecological studies have
contributed to a clearer understanding of why hydrological connectivity is essential to the ecological integrity
of the landscape, where alterations that either reduce or
enhance connectivity are seen to have potentially significant environmental impacts (Pringle, 2003). Likewise,
biologists define landscape connectivity as the degree to
which a landscape facilitates or impedes the movement
of individuals (Taylor et al., 1993). Hence, the concept
is both a landscape- and species-specific characteristic
(Wiens et al., 1997; Brooks, 2003). Important landscape
characteristics include shape, size and location of feature in the landscape, whilst key species qualities are
behaviour and patterns of gene flow (Brooks, 2003). No
index allows cross-system comparisons (Davidson, 1998)
and no empirical studies consider the various scales that
are implicit in exploring connectivity (Brooks, 2003).
As with geomorphology, ecology faces conceptual and
empirical issues with heterogeneity between landscapes
and species (Wiens et al., 1997; Brooks, 2003). Interestingly, an ecologists definition of connectivity combines
a measure of spatial integration of the landscape with the
movement of species (Table I).
Overall, there appears widespread recognition that
the term connectivity in any of these geo-ecological
senses is useful in promoting the interconnection between
the morphological components of the landscape and
the material fluxes that move across, and through, the
drainage basin. The term has also allowed us to think
of dis-connectivity (or isolation as used in ecological
studies) between both landscape parts and sources of
1752
At-a-site
HYSS
Infiltration
Roughness
Soil / Lithology
Tillage
Vegetation
Profile form
Plan form
Drainage
Catena
HRU
Gradient
Connected area
Catchment response
Storm volume
Dist'n of HRUs
Effective intensity
0 metres 4
First response
Gutter system
Bare area (runoff source)
Second response
Drainage divide
Runoff
Potential
Delivery
Pathway
Lateral
Buffering
Landscape
Position
Climate
Figure 4. The components of catchment connectivity
Copyright 2007 John Wiley & Sons, Ltd.
Third response
Flow line
Vegetation
subsurface flow waiting to be reconnected during rainfall. Hence, wet area expansion occurs frequently and
connectivity in runoff can be relatively easily achieved.
The direction of connectivity within a catchment may be
working headwards from the channels as saturated areas
expand, but also downslope from saturated hollows to
stream channels (Figure 5). A review of issues related to
connectivity in temperate areas can be found in Burt and
Pinay (2005).
The VSA concept does not hold in semi-arid and arid
environments because most runoff is produced by HOF
(Bryan and Yair, 1982). There are no areas of saturation
to initiate quick runoff because storms are infrequent and
of short duration. Run-on infiltration and transmission
losses are also high, which limits the development of
continuous hydrological pathways. Key runoff-producing
areas tend to be found on steep slopes, either abandoned
after agriculture or with sparse vegetation, and are
composed of runoff-promoting soils such as marls (e.g.
Bull et al., 2000). These areas do not necessarily relate
to the channel network, and the mosaic pattern they form
is the key to producing floods in ephemeral channels.
Hence, in semi-arid and arid areas, the conceptual
model of connectivity is very different from humidtemperate areas. As a general rule, it is more difficult for
connectivity to be achieved in drylands and it occurs less
frequently than in humid-temperate areas. Antecedent
conditions are relatively unimportant in drylands and
there is no base flow or stores of through flow waiting to
be reactivated. Connectivity occurs by the interaction and
response of a mosaic of patches to rainfall, and depends
on the rainfall duration and intensity being great enough
to allow transmission of water over hillslopes and into
channels, and then to propagate down channels overcoming transmission losses to connect whole catchments
(Figure 5). For example, Lavee et al. (1998) found that
in semi-arid areas there was a mosaic of arid watercontributing areas and moist water-accepting patches.
The two scenarios outlined above represent the opposite ends of a continuous spectrum of runoff response for
a range of different environments. In most areas, both
SOF and HOF will be produced in different proportions
depending on rainfall intensity and surface characteristics. Hydrological connectivity depends on the interaction
Humid
Dryland
Channel network
1753
of saturated areas that follow the river channels and hillslope hollows, and the mosaic of areas that are sensitive
to intense rainfall. In such areas, therefore, the response
to rainfall and hydrological connectivity is dynamic and
will change depending on the nature of rainfall input,
antecedent conditions and catchment characteristics.
Storm intensity and duration. It has been proposed that
rainfall intensities are of prime importance for flood generation in all catchment sizes (Costa, 1987; Schick, 1988;
Pitlick, 1994), and it has been appreciated that high discharges are produced by an optimal combination of basin
morphometry and storm intensity (Costa, 1987). Storm
intensity and duration are thus crucial to producing connectivity, and are the inputs that combine with the spatial
pattern of integration to produce connected flow. Other
storm attributes that influence connectivity are the temporal variation in rainfall throughout the storm and the
variation of rainfall spatially within a catchment. For
example, the timing of a high-intensity pulse of rain during a storm event can be crucial for connectivity. Light
rainfall can have the effect of wetting up a catchment, so
that when a high-intensity pulse occurs the runoff generation is rapid and transmission losses low, thus encouraging connected flow. However, a high-intensity pulse on
a dry catchment may have little impact, since infiltration
rates are high and thresholds for runoff generation and
connected flow are not exceeded. Schick (1988) demonstrated that two-thirds of runoff events in the Negev and
Sinai are generated in small to medium catchments by
high-intensity, short-duration storm events, which have
extremely flashy hydrographs and have dramatic effects
on geomorphology. The remaining one-third are more
long-term floods, produced by low rainfall intensities
that may last a few days. In contrast, research undertaken on the Colorado Plateau in Utah by Dick et al.
(1997) showed that rapid runoff was produced during
short-duration storms, which resulted in complex hydrographs that reflected channel network geometry. The
largest flow event recorded in the area was produced by
a low-intensity storm of long duration, which occurred
when antecedent moisture was high (Dick et al., 1997).
Romkens et al. (2001) suggest that storms with initially
high intensities have the potential for the early onset of
concentrated flow, which has obvious implications for the
probability of connectivity.
The temporal pattern of high-intensity rainfall is also
important for determining flow distances of runoff which
connect source areas and result in connectivity at the
hillslope and catchment scale. Relationships are complex and difficult to investigate, so there is not a large
a body of literature to draw upon. Yair and Lavee (1985)
conducted a series of experiments on hillslope runoff
response using varying rainfall intensity. Uniform runoff
generation across slopes was only found for storms with
rainfall intensities greater than 9 mm h1 and for durations greater than 45 min. They found that the properties of storm events explained discontinuities in runoff
from semi-arid and arid areas. Wainwright and Parsons
Hydrol. Process. 21, 1749 1763 (2007)
DOI: 10.1002/hyp
1754
(2002) investigated the relationship between varying rainfall and spatial variation in infiltration capacity, slope
length and runoff coefficients. It was found that the temporal structure of the storm influenced the runoff coefficient because runoff produced during the start of the
storm could infiltrate as the storm continued. Temporal
variation in rainfall intensity was significant in controlling the scale dependency of the runoff coefficient. It was
suggested that using constant rainfall intensities in modelling underpredicts runoff generation. These two studies
are supported by Van de Giesen et al. (2000), who found
that the runoff coefficient varied between storms with
different spatial and temporal dynamics.
Figure 6 presents an example of the variability in
storm characteristics that produce floods in the semiarid environment in southeast Spain. The dominant
relationship is between the total rainfall in a spell and
the size of the flood peak. In this analysis, spells were
defined as periods of rainfall without breaks greater
than 12 h. Thus, a new spell begins once it rains and
has been dry for over 12 h. The seven largest spells
of rainfall all exceeded 50 mm in total, and produced
flood peaks in the main channel; however, smaller events
can also produce a hydrograph peak. The higher flood
peaks fit more closely with the trend identified compared
with the smaller flood peaks, which have more scatter
(Figure 6). There is also a slight relationship between
the duration of a spell of rain and the flood peak,
although this relationship is likely to be caused by the
correlation between spell duration and spell total rainfall.
More results such as these are necessary to understand
fully the impact of storm intensity and duration on
connectivity.
Catchment-scale hydrological connectivity and flooding, therefore, overall requires prolonged, high-intensity
storms, whereas hillslope hydrological connectivity can
be initiated by shorter duration, or lower intensity events.
Each catchment has a base spatial pattern in terms of connectivity, depending on key runoff-generating areas, and
a response curve as the catchments wets up.
Hillslope runoff potential
Related factors. It is now commonly accepted that
catchments have a non-uniform and non-linear hydrological response to rainfall (e.g. Cerd`a, 1995; Bergkamp
et al., 1996). This is caused by spatial variability in
hydrological properties due to complex geological, pedalogical and management histories (Fitzjohn et al., 1998)
combined with temporally and spatially variable rainfall
(Thornes, 2001). The hillslope is the major landscape
unit and is the scale at which most research on runoff
generation has taken place. The processes of runoff and
run-on lead to dynamic interactions between different
zones of spatially variable hillslopes, and these interactions affect the large-scale response of catchments to
produce floods (Fiedler et al., 2002). As both Lal (1990)
and Bryan (2000) highlight, there is no single, measurable soil property that can fully represent the integrated
response that constitutes soil surface. There are many
factors that influence hillslope runoff, including crusting
and surface roughness (Auzet et al., 1993; Helming et al.,
1998; Singer and Le Bissonnais, 1998), heterogeneity
within the soil (Fitzjohn et al., 1998), the impact of variable density and type of vegetation (Imeson et al., 1992;
Bergkamp et al., 1996), changing catchment morphometry, transmission losses in tributaries and main channels
(Reid and Frostick, 1997), and the impact of land use
(Bull et al., 2000; Lasanta et al., 2000). We do not discuss
each variable in detail; rather, we specifically highlight
the role of each factor for connectivity.
Infiltration is a key process in runoff generation and,
hence, connectivity. The infiltration rate is probably
the most important hillslope characteristic for runoff
generation at the plot scale. In contrast, the spatial
1000
Mean storm intensity
(mm/hr)
1000
100
10
1
0.01
0.1
100
10
1
0.01
10
10
10
Storm duration (days)
1000
100
10
1
0.01
0.1
0.1
10
0.1
0.01
0.01
0.1
1
Peak stage (m)
10
Figure 6. The relationship between floods and storm characteristics for the Rambla de Torrealvilla, southeast Spain
Copyright 2007 John Wiley & Sons, Ltd.
1755
of efficient pathways (Imeson and Verstraten, 1988; SoleBenet et al., 1997). Hydrological connectivity produced
by surface roughness is influenced by slope. Smooth surfaces are not greatly affected by gradient, but for rough
surfaces the surface area changes dramatically and can
affect depression storage, runoff and the development of
rills as effective coupling between areas of a hillslope.
Runoff is usually concentrated along channels determined
by topography or agricultural practices (Ludwig et al.,
1995), and some research has shown that, as roughness
increases, flow concentration occurs, thus increasing erosion (Abrahams and Parsons, 1990; Helming et al., 1998;
Romkens et al., 2001) and, in turn, increasing connectivity.
Vegetation has a complex relationship with all the
factors mentioned above and is a major influence on
hydrological connectivity at all scales. Indeed, Imeson
and Prinsen (2004) suggest that patterns of vegetationbare soil can be used as indicators of the extent, distribution and connectivity of runoff and sediment sources
and sinks. Vegetation has a positive effect on infiltration by increasing organic matter, lowering bulk density
(Thornes, 1976; Boix-Fayos et al., 1998; Wilcox et al.,
1998) and increasing hydraulic conductivity (Nicolau
et al., 1996). At the plot scale, research has shown that
infiltration rates beneath bushes can increase infiltration
up to three times that of the inter-shrub area (Lyford and
Qashu, 1969). Vegetation also reduces rainsplash and,
hence, crust formation, but increases surface roughness
(hence hydraulic resistance) and ponding. Many investigations have noted patterning of vegetation at the hillslope scale (Sanchez and Puidefabregas, 1994; Dunkerley,
1999; Valentin et al., 1999; Puigedefabregas, 2005; Boer
and Puigedefabregas, 2005), which can greatly increase
infiltration compared with the inter-vegetated areas. During storms, therefore, runoff is generated on the bare
areas, but runoff infiltrates when it reaches a vegetated
area, limiting connectivity.
However, the spatial configuration of bare and vegetated areas controls runoff, not just the actual areas of
each (Morgan, 1995; Fitzjohn et al., 1998; Cammeraat
and Imeson, 1999; Ludwig et al., 2005). Within ecology
there has been a lot of research based on spatial variations in vegetation and how this is related to hydrological
processes. Banding in vegetation and resulting patchiness has been explored and evaluated in terms of theory
(Wiens, 1995), causes (Belsky, 1995), genesis (Thiery
et al., 1995), dynamics (Mauchamp et al., 1994) and
function (Ludwig et al., 1999a,b, 2005). There have been
many different terms for vegetation patchiness, varying
from two-phase (Archer, 1990; Belsky, 1995), two-phase
mosaics (Montana, 1992), stripes (Cornet et al., 1991;
Thiery et al., 1995) and banding (Slatyner, 1961; Boyland, 1973; Mabbutt and Fanning, 1987). These different
patterns are related to soil depth and soil fertility (Tongway and Ludwig, 1990; Ludwig and Tongway, 1995).
Loss of patchiness has been shown to be the greatest
influence on the ability of hillslopes to capture rainfall
Hydrol. Process. 21, 1749 1763 (2007)
DOI: 10.1002/hyp
1756
(which is related to plant productivity), with the difference between different types of patchiness (stripes,
strands or patches) being small compared with the differences between vegetation patch types (Ludwig et al.,
1999a). Ludwig et al. (1999b) also found that the redistribution of runoff into run-on spaces between vegetation
patches is greater in drier environments relative to wetter ones. The ratio of bare : patch length decreased as
the rainfall increased in intermediate textured soils (Ludwig et al., 1999b). Ludwig et al. (2000) also propose
that scaling equations determined for the relationships
between bare areas and vegetation patches can be used
in computer simulation models or in hypothetical situations to collapse the complexity of scaling issues into
practical management frameworks.
This type of research in ecology benefits the discussion of hydrological connectivity in two ways. First, it
develops our understanding of factors influencing hillslope infiltration and runoff. Second, and possibly more
important, it also examines the spatial variability in hillslope characteristics. One obvious benefit may be the
development of spatial scaling equations for use in upscaling simulation models.
Land management affects runoff response and the spatial integration of source areas at all scales of the landscape and is a key factor in connectivity. Many studies
have shown that changing from natural vegetation to
cultivation can increase soil erosion rates by an order
of magnitude or more (Abernethy, 1990; Morgan, 1995;
Walling, 1999). Increasing soil erosion by rill development has a direct relationship with increased connectivity.
Land management includes crop planting, road building,
land-use change and reservoir construction. Ploughing is
one of the major controls of surface properties, removing
vegetation, destroying crusting, and moving rock fragments to the surface (Oostwoud Wijdenes et al., 1997),
but even abandonment of previously farmed land has a
complex relationship with runoff (Lasanta et al., 2000).
Ludwig et al. (1995) found that interactions between agricultural and topographic features resulted in a complex
temporary hydrological hillslope structure, which consisted of a runoff collector network composed of topographic or agricultural depressions and the distribution of
runoff contribution areas, that varied over time.
Temporal variability in the factors affecting connectivity is important, because over the duration of a storm,
as well as for longer time periods, many factors affecting
connectivity are subject to change. For instance, during an
individual storm the infiltration capacity decreases as the
soil becomes saturated and fines block pores. The microscale surface roughness also changes as soil aggregates
are broken down and crusts develop, also decreasing the
infiltration capacity. However, erosional processes, such
as rill development, operate during a storm and, once
formed, rills result in increased form roughness and act
as efficient pathways that channel runoff. Clay minerals
also swell during wet periods and shrink during drying, also affecting the infiltration capacity, and hence
runoff. Intense cracking can also result in marcopores. At
Copyright 2007 John Wiley & Sons, Ltd.
S
#
#
##
#
# # ##
#
# ## #
S
#
#
Gullied
Connected Dispersive
Forest Road
Natural Drainage
0
#S # #
# #
# ##
##
100
Scale -
S
#
#
S
#
S
S
##
S
#
##
S S
#
#
#
##
S
#
#
S
#
S
#
##
S
#
#
SS
#
#
#
#
##
S
#
1757
1758
channel. Likewise, the area labelled B is an area of tributaries disconnected by human-built drainage ditches to
channel flood water into wheat fields. However, with different input rainfall, such as a 1 in 10-year event, the areas
labelled A and B may produce runoff that connects with
the main channels to produce flood flow. Hence, there are
different responses with varying amounts of hydrological
runoff and resulting connectivity for different events.
In an attempt to characterize the complex distribution of spatial and temporal factors better, research
has focused on approaches to group or classify similar
hydrological surfaces. For example, for a storm to initiate catchment-scale runoff it must overcome the spatial
arrangement and threshold values of hydrologically similar surfaces (HYSS) at all scales (Kirkby et al., 2002).
Similarly, HRUs are defined as distributed, heterogeneously structured entities having a common climate,
land use and underlying pedologicaltopographical
geologic association controlling their hydrological
dynamics (Flugel, 1995). For hillslopes and channels to
be connected and produce floods, HRUs need to be interconnected and connect to main channels. Therefore, the
overland flow contribution from responsive areas depends
on their location within the catchment and their spatial
arrangement with respect to each other (Sharma et al.,
1987; Yair, 1992; Fitzjohn et al., 1998). Where there is no
connection between runoff-producing areas, only those
areas located adjacent to the channel contribute to catchment outflow, and runoff from spatially isolated areas
will be reinfiltrated (Yair, 1992; Fitzjohn et al., 1998).
Hence, hillslopes and catchments can be spatially isolated or spatially interactive (Sharma et al., 1987; Cerd`a,
1995; Fitzjohn et al., 1998); however, connectivity also
depends on the input of rainfall (duration and intensity)
to the system to produce connected flow.
By dynamic hydrological connectivity, we mean both
longer term landscape development, such as changes
following abandonment of agriculture, and short-term
variation in antecedent conditions and rainfall inputs
to systems that result in non-linearities in hillslope
and catchment response to rainfall. Dynamic aspects of
connectivity are more difficult to quantify, but they are
possibly more important than static elements because
it is the combination of static and dynamic elements
High potential runoff
A
B
1759
Vbt (l)
1000
100
10
U. Tyers
Hairsine et al.
Study area
Figure 9. Observed pattern of overland flow plumes following a highintensity storm on a severely burnt hillslope in southeastern Australia
Copyright 2007 John Wiley & Sons, Ltd.
1760
CONCLUSIONS
This paper has introduced a framework for connectivity
that identifies some of the key factors that affect the
physical linkage of water and, indirectly, sediment in
catchments. We have also proposed and illustrated a
working definition of hydrological connectivity that we
suggest can be measured and thus incorporated into
future studies of runoff generation and sediment delivery.
Connectivity is one way forward in the discussion of postVSA models of hillslope response to rainfall and provides
a basis for integrating processes of runoff generation with
landscape characteristics.
One of the key aspects of connectivity is the combination of both static and dynamic process responses.
After several decades of hydrological research, current
understanding is well developed in terms of the static elements, but we know less about the dynamic components
of hydrological connectivity. This is one significant reason for developing a definition that can be measured and
estimated in the field, but also incorporated into computer
simulation models. Validation needs to be more than just
a comparison with stream hydrographs, because whereas
hydrographs are representative of process interactions,
they do not enable us to discern the complex interplay of
factors that result in flow generation. For connectivity to
be accepted as a viable post-VSA and Hortonian framework, we should acknowledge the benefits experienced
by ecologists in using a standard terminology and use this
to modify how we conceptualize runoff relationships.
REFERENCES
Abdul AS, Gillham RW. 1989. Field studies of the effects of the
capillary-fringe on streamflow generation. Journal of Hydrology 112:
(1-2) 1-18.
Abernethy C. 1990. The use of river and reservoir sedimentation
data for the study of regional erosion rates and trends. In
International Symposium on Water Erosion Sedimentation and Resource
Conservation, India.
Abrahams AD, Parsons AJ. 1990. Determining the mean depth of
overland flow in field studies of flow hydraulics. Water Resources
Research 26: 501 503.
Ambroise B. 2004. Variable active versus contributing areas
or periods: a necessary distinction. Hydrological Processes 18:
1149 1155.
Andreassen HP, Halle S, Ims RA. 1996a. Optimal design of movement
corridors in root volesnot too wide and not too narrow. Journal of
Applied Ecology 33: 6370.
Andreassen HP, Ims RA, Steinset OK. 1996b. Discontinuous habitat
corridors: the effect on male root vole movements. Journal of Applied
Ecology 33: 555560.
Archer S. 1990. Development and stability of grasswoody mosaics
in a subtropical parkland Texas, USA. Journal of Biogeography 17:
453 462.
Aryal SK, Mein RG, OLoughlin EM. 2003. The concept of effective
length in hillslopes: assessing the influence of climate and topography
on the contributing area of catchments. Hydrological Processes 17:
131 151.
Auzet AV, Boiffin J, Papy F, Ludwig B, Maucorp J. 1993. Rill erosion
as a function of the characteristics of cultivated catchments in the north
of France. Catena 20: 41 62.
Barling RD. 1992. Saturation zones and ephemeral gullies on arable land
in south-eastern Australia. PhD thesis, The University of Melbourne,
Victoria, Australia.
Belsky AJ. 1995. Spatial and temporal landscape patterns in arid and
semi-arid Africa savannas. In Mosaic Landscapes and Ecological
Copyright 2007 John Wiley & Sons, Ltd.
1761
1762
1763