Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
Prof J. V. Ward, Department of Limnology, EAWAG/ETH, Ueberlandstrasse 133, CH-8600 Duebendorf, Switzerland. E-mail: firstname.lastname@example.org
1. This review is presented as a broad synthesis of riverine landscape diversity, beginning with an account of the variety of landscape elements contained within river corridors. Landscape dynamics within river corridors are then examined in the context of landscape evolution, ecological succession and turnover rates of landscape elements. This is followed by an overview of the role of connectivity and ends with a riverine landscape perspective of biodiversity.
2. River corridors in the natural state are characterised by a diverse array of landscape elements, including surface waters (a gradient of lotic and lentic waterbodies), the fluvial stygoscape (alluvial aquifers), riparian systems (alluvial forests, marshes, meadows) and geomorphic features (bars and islands, ridges and swales, levees and terraces, fans and deltas, fringing floodplains, wood debris deposits and channel networks).
3. Fluvial action (erosion, transport, deposition) is the predominant agent of landscape evolution and also constitutes the natural disturbance regime primarily responsible for sustaining a high level of landscape diversity in river corridors. Although individual landscape features may exhibit high turnover, largely as a function of the interactions between fluvial dynamics and successional phenomena, their relative abundance in the river corridor tends to remain constant over ecological time.
4. Hydrological connectivity, the exchange of matter, energy and biota via the aqueous medium, plays a major though poorly understood role in sustaining riverine landscape diversity. Rigorous investigations of connectivity in diverse river systems should provide considerable insight into landscape-level functional processes.
5. The species pool in riverine landscapes is derived from terrestrial and aquatic communities inhabiting diverse lotic, lentic, riparian and groundwater habitats arrayed across spatio-temporal gradients. Natural disturbance regimes are responsible for both expanding the resource gradient in riverine landscapes as well as for constraining competitive exclusion.
6. Riverine landscapes provide an ideal setting for investigating how complex interactions between disturbance and productivity structure species diversity patterns.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
River corridors consist of a dynamic mosaic of spatial elements and ecological processes arrayed hierarchically. As such, they fit comfortably within the framework of landscape ecology, defined by Turner (1998) as the study of interactions between spatial patterns and ecological processes in the context of spatial heterogeneity across a range of scales. This review of riverine landscape diversity focuses on pattern and process at habitat, floodplain, and corridor spatial scales across seasonal and successional time scales. River corridors are addressed as integrated ecological systems, as opposed to the traditional view of corridors as conduits between similar elements within the landscape.
Consideration is primarily restricted to natural conditions. It is our contention that the remarkable degree of spatio-temporal heterogeneity characterising riverine landscapes has been masked in much of the world by a long history of river engineering (see Petts, Moller & Roux, 1989; Benke, 1990; Dynesius & Nilsson, 1994). Floodplain reaches, which exhibit the highest heterogeneity in the natural state, have been the most severely altered (e.g. Ward & Stanford, 1995), resulting in a distorted perception of patterns and processes in riverine landscapes. An accurate comprehension of spatio-temporal heterogeneity in the unaltered state is crucial for a holistic understanding of the structure and function of river ecosystems and is essential for successful protection and restoration. As stated by Bornette et al. (1998a), `The sustainable conservation of diversity in riverine wetlands implies knowledge of the basic geomorphological and ecological processes that interplay at the landscape scale.'
In the natural state, riverine landscapes exemplify the `new paradigm in ecology' (sensuTalbot, 1996), in which ecological systems are widely recognised as non-deterministic, open systems in continual states of flux, rather than internally regulated, homeostatic systems exhibiting equilibrium conditions. Yet, despite their highly dynamic nature, riverine landscapes provide predictable ecological conditions. Although individual landscape features exhibit high turnover, largely as a function of interactions between fluvial dynamics and successional phenomena, their relative abundances in the river corridor tend to remain constant over ecological time.
This review begins with an account of the diverse array of landscape elements encompassing surface waters, alluvial aquifers, riparian systems and geomorphic features. Then landscape dynamics within river corridors are examined in the context of landscape evolution, ecological succession and turnover rates of landscape elements. This is followed by analyses of connectivity between landscape elements. Finally, a riverine landscape perspective of biodiversity is presented. Special attention is given to floodplains throughout the paper because of the high level of landscape diversity they exhibit. Riverine landscape diversity is a broad topic with a vast literature; in this review it is possible to include only a portion of even the most important work. Detailed treatments of the fauna and flora of riverine landscapes are presented elsewhere (Ward et al., 2000a; Adis & Junk, 2002; Amoros, 2002; Robinson, Tockner & Ward, 2002).
Landscape elements of river corridors
The river corridor
River corridors are linear features of the landscape, structured along ribbons of alluvium extending from the headwaters to the sea. Narrow canyon-constrained reaches typically alternate with alluvial floodplains, like `beads on a string' (Fig. 1; Stanford & Ward, 1993). In constrained reaches, the hillslopes descend abruptly to a single-thread channel bordered by a narrow band of riparian vegetation. Alluvial deposits consist of only a thin layer of sediment covering the bedrock and hydrological exchange is predominantly unidirectional (downstream).
The floodplain reaches (the `beads') of river corridors, in contrast, are expansive, with multiple channels and deep alluvial deposits. Alluvium is the matrix upon and within which the landscape elements are embedded. Hydrological exchange occurs along longitudinal, lateral and vertical dimensions. Terraces are former floodplain surfaces that were formed when the river was flowing at a higher level. Nanson & Croke (1992) developed a typology, in an attempt to relate the complex landforms of floodplains to two key variables: stream power (the ability of a river to entrain and transport sediment) and sediment cohesiveness (the erosional resistance of the alluvium). Three basic types of floodplains were identified: (1) Disequilibrium floodplains (high energy, non-cohesive) which erode in response to extreme episodic flow events. Such floodplains tend to be located in steep headwaters where channel migration is constrained by coarse substratum and narrow valleys. Floodplain construction is mainly by vertical accretion. Dominant landforms include boulder levees, sand and gravel splays, back channels and scour holes. (2) Equilibrium floodplains (medium energy, non-cohesive) are formed by regular flow events in broad valleys. Such floodplains are thought to be in dynamic equilibrium with the flow regime. Fluvial energy from extreme floods is dissipated as floodwaters overtop the channel banks and disperse across an expansive surface. Floodplain construction involves lateral point bar accretion or braid channel accretion. Channels may be braided, anastomosed or meandering. Characteristic landforms include abandoned channels, bars and islands, oxbows, meander scrolls and backswamps. (3) Low-gradient floodplains (low energy, cohesive) are also formed by regular flow events in broad valleys, but with channels laterally stabilised by erosion-resistant banks of fine cohesive alluvium. Floodplain construction involves mainly vertical accretion of fine sediment deposits and occasional channel avulsion. Channels may be braided, anastomosed or meandering. Landforms include levees, islands, splays and backswamps.
Ideally, these three classes of floodplains occur sequentially from the headwaters to the lower reaches of a given river corridor, corresponding in general to the sediment production, transfer and storage zones of Schumm's (1977) `fluvial system'. Many complicating factors, however, such as glacial activity, may disrupt the idealised sequence.
The major aquatic, semi-aquatic and terrestrial elements of riverine landscapes may be grouped within four interactive categories (Fig. 2), one of which is surface waterbodies. Here we address the diversity of surface waters in river corridors; subsequent sections deal with their permanence, connectivity and successional trajectories.
A diversity of lotic, semi-lotic and lentic waterbodies occur within river corridors. In dynamic river ecosystems these three classes of surface waters merely represent different positions along an inundation continuum and they are typically interconnected during floods. During the dry phase, however, each has distinctive attributes, as described below and illustrated in Fig. 3.
Lotic waters flow within the main channel and the side arms having both upstream and downstream connections to the main channel. In addition, springbrooks originate as upwelling ground water from the alluvial aquifer beneath the river corridor (alluvial springs) and from hillslope aquifers that emerge along the edge of the river corridor (hillslope springs). Surface drainage from the hillslope forms tributaries which may flow some distance within the corridor before forming a confluence or sinking into the alluvium.
Semi-lotic waters include abandoned braided channel segments and dead arms that retain a connection to the main channel only at their downstream ends. Both typically require only minor flooding to reconnect with the main channel, thereby frequently alternating between lentic and lotic conditions. Even when disconnected, inflowing ground water may sustain a slight current within these waterbodies.
Lentic waters within the river corridor are of both fluviatile and non-fluviatile origins. Lakes of non-fluvial origin that may be associated with riverine landscapes form by different mechanisms, including damming stream/river courses by lava flows, landslides, moraines or beavers (see Hutchinson, 1957; for a detailed account of the origin of lake basins). The fluviatile action of running water forms lakes (and other lentic waterbodies) in various ways. Lateral lakes form when entering tributaries are dammed by sediment deposited as levees along the main channel during floods. Tributaries may also be blocked by wood debris to form riparian lakes (Triska, 1984). Alluvial fans from steep mountain tributaries may dam the main channel to form lakes, although most are soon drained by erosion of the fan. Fluviatile action originating within the floodplain also forms lentic waterbodies. As the main channel migrates across the river corridor meander loops are abandoned to form oxbow lakes. Unlike abandoned braids, which are smaller, shallower and without surrounding levees, oxbow lakes are less frequently reconnected to the main channel, thereby developing a more truly lentic character. Ponds and marshes occur in the depressions between successive meander scrolls, where relict levees, abandoned channels and point bars form a ridge and swale topography on the floodplain (Nanson & Beach, 1977). Marshes and swamps often occur adjacent to the hillslope or terrace, where erosion scours the floodplain surface during floods.
A precise quantification of surface waters was conducted in six geomorphic reaches along the corridor of an Alpine river, the Fiume Tagliamento, Italy (Arscott, Tockner & Ward, 2000). Using digitised maps and aerial photographs in concert with field surveys, four major surface water habitats were distinguished: surface-connected channels (SC), alluvial channels (AC), tributary channels (TC) and isolated standing waters (ISO). The first three habitats included primary, secondary and tertiary branches plus backwaters (e.g. SC1, SC2, SC3, SCB), resulting in a total of 13 habitat types (Fig. 4). The total area of surface waters per river km exhibited a unimodal pattern, with the highest value in the island-braided lower reach and the lowest in the constrained headwaters. The Shannon diversity index and Simpson's index of dominance were employed to examine aquatic habitat heterogeneity along the river corridor (Fig. 4b), with the habitat types representing `species' and the relative abundance of habitats representing `species abundance' values. Shannon diversity exhibited a unimodal pattern with maximum diversity in the island-braided lower reach; dominance exhibited the inverse pattern as expected. The remarkable degree of habitat heterogeneity along the Fiume Tagiamento is attributed to its morphologically intact river corridor, forested catchment and natural flood regime (Ward et al., 1999).
The geomorphic features of floodplains reflect complex interactions between climate, catchment geology, topographic relief and fluvial dynamics, mediated by vegetation. In this section, we present geomorphic features largely as spatial landscape elements, reserving consideration of the dynamics for a later section.
Alluvial channel networks have been traditionally classified into four types (straight, meandering, braided and anastomosed), with recognition that these types are really part of a continuum. Straight channels have a sinuous thalweg and alternate bars that slowly move downstream. Meandering rivers have sinuous single-thread channels that erode the concave banks, causing the lateral and downstream migration of the channel network. Point bars that form on the convex bends of meanders build the floodplain by lateral accretion. Braided rivers consist of multiple shifting channels that are highly unstable if separated by sand/gravel bars devoid of vegetation (bar braided), but less so if the bars have been stabilised by vegetation (island braided). Anastomosed rivers may be regarded as an extreme form of island-braiding, the prominent landform consisting of large, relatively permanent vegetated islands.
Other geomorphic features of riverine landscapes include levees, crevasse splays, alluvial fans and deltas (Allen, 1965). Levees are ridges of sediment along channels formed when a river overtops its banks and deposits the coarsest part of its load as water velocity abruptly decreases. In large rivers levees may be more than 5 m above the floodplain surface. Crevasse splays are sediment deposits that form when sediment-laden water breaches the levee and spreads over the floodplain dropping its load. Alluvial fans are depositional structures that form where steep tributaries enter a river corridor. They commonly occur in arid climates where intermittent tributaries carry large sediment loads associated with violent rainstorms or in mountainous topography where tributaries carry snowmelt runoff. Deltas form where rivers empty into a lake or the sea, the channels forming a distributive network in which the streams divide and become increasingly smaller.
Deposits of large woody debris (LWD) are common landscape elements in natural rivers with forested catchments. In a survey of LWD retention in a river corridor, Gurnell et al. (2000) estimated wood storage values of c. 1 t ha–1 for open gravel reaches with single-thread channels, 6 t ha–1 for open gravel reaches with multiple-thread channels and 80 t ha–1 on islands. Wood debris plays major geomorphologic and ecological roles in riverine landscapes (Keller & Swanson, 1979; Maser & Sedell, 1994). Moreover, LWD may serve as nuclei for the development of vegetated islands (Abbe & Montgomery, 1996), the presence of which is believed to contribute directly and indirectly to landscape heterogeneity (Gurnell et al., 2001; Ward et al., 2000b).
The term fluvial stygoscape refers to the subterranean extension of the riverine landscape (Ward, 1997). Aquifer systems (water-saturated alluvium) associated with rivers, traditionally regarded as rather homogeneous, are now known to exhibit heterogeneity at multiple scales and may be highly interactive with surface waters (Gibert et al., 1994). A highly specialised fauna inhabits the water-filled interstices between sediment particles (Ward et al., 2000a). The fluvial stygoscape is a three-dimensional mosaic, the spatial structure of which is a function of sediment particle size, particle size heterogeneity, pore size, porosity, pore geometry and hydraulic conductivity, and the distribution of preferential flow paths, buried organic matter, and chemical and thermal gradients within the alluvial aquifer. Ground-penetrating radar may be used to identify subterranean structures such as fills of former channels, ancient bars and intercalations of buried organic debris (Fig. 5).
The hyporheic zone, the ecotone between surface waters and the alluvial ground waters of the phreatic zone, is a dynamic boundary zone characterised by nutrient transformations, steep gradients (physical/chemical/biological) and hydrologic exchange (Castany, 1985; Brunke & Gonser, 1997; Boulton et al., 1998). As the name implies, hyporheic zones occur beneath running water channels; however, the hyporheic zone may also extend some distance (up to kilometres) beyond the channel margin beneath the floodplain (Stanford & Ward, 1988) or be missing altogether under special conditions, such as where a headwater stream flows directly on bedrock.
The vegetation of river corridors forms a complex mosaic in response to gradients of climate, inundation/soil moisture, disturbance and nutrients (Gregory et al., 1991; Malanson, 1993; Décamps, 1996; Naiman & Décamps, 1997). Precipitation during the dry phase is a critical determinant of broad scale patterns of floodplain forest development. As stated by Junk, Bayley & Sparks (1989), `when local precipitation at low water is high, floodplains are forested, e.g. in the middle and upper Amazon, Zaire and Mississippi rivers. Conversely, when local precipitation is low, savannas with gallery forest develop, e.g. in the floodplains of the lower Nile, Zambezi, and Volta rivers.' Within a catchment, elevation exerts primary influence on controlling variables, and therefore on riparian vegetation composition and structure, at three scales: (1) as altitude decreases along the river's course from mountain headwaters to sea level, (2) as elevation increases along the lateral gradient from the main channel to the uplands, and (3) as a function of local changes in elevation reflecting topographic features on the floodplain surface, such as levees, islands, ridges and swales.
Under natural conditions, alders (Alnus) and willows (Salix) dominate the woody riparian vegetation in the upper reaches of Central European river corridors (Pinay et al., 1990; Statzner & Kohmann, 1995; Schnitzler, 1997). As rivers enter the foothills and develop a wider floodplain, alders begin to decline in relative abundance, willows increase, poplars (Populus) appear, and an alluvial hardwood forest dominated by oak (Quercus) and elm (Ulmus) occupies the less frequently inundated outer portions of the corridor. In the lowlands, willows and poplar predominate on frequently inundated sites, with an alluvial forest dominated by elm and oak occupying the remainder of the expansive floodplain. Different portions of the altitudinal gradient are characterised by different species of both Salix and Alnus. Riparian species differ markedly in the ability to sprout vegetatively from pieces of live wood deposited on sediments following flood recession (Gurnell et al., 2001). Because vegetative sprouting is one mechanism by which alluvial islands form, the altitudinal distribution of woody species influences landscape-level processes.
Along the lateral gradient from the channel to the uplands, river corridors with expansive floodplains may contain a series of riparian zones, primarily reflecting species-specific responses to soil moisture/oxygenation, sediment deposition, the frequency and duration of inundation, and the erosive action of flooding. An example of woody vegetation zones along a lateral floodplain gradient is illustrated in Fig. 6. Pautou (1984) developed a typology for alluvial forests of the Upper Rhone based on their position along three axes: (1) site elevation above water level, (2) sediment particle size composition/concentration of organic matter in the A1 horizon, and (3) the seasonal regime, including precipitation patterns, which determines the time of year a given floodplain unit is inundated.
The idealised lateral gradient at the floodplain scale is, however, complicated by local topographic features (Brinson, 1990). High levees, composed of coarse-grained well-drained soils, often support riparian plants unable to tolerate conditions at lower levels on the floodplain. The ridge and swale topography formed by relict point bars, levees and abandoned channels, produces an undulating surface across the floodplain. Depressions are colonised by species adapted to long hydroperiods, whereas the levees and ridges may contain species that also occur in mesic upland sites. Woody alluvial species collectively occupy a broad range of flood tolerance (Schnitzler, 1997). Salix alba L. tolerates up to 300 days of inundation; Quercus robur L. and Ulmus minor Mill. tolerate up to 151 days; Fraxinus excelsior L. 102 days; Acer campestre L. and Tilia cordata Mill. 13 days; whereas Acer pseudoplatanus L. and Fagus sylvatica L. cannot withstand even brief periods of inundation and are confined to high terraces.
Riparian vegetation moderates myriad ecological processes in river corridors by influencing temperature and light regimes; producing organic detritus (leaf litter, woody debris); by routing water and sediment; by structuring the physical habitat at several scales; by providing a substrate for biological activity and habitat/cover for aquatic, amphibious and terrestrial animals. Riparian plants serve as roughness elements during periods of inundation, thereby influencing the hydrogeomorphological processes that structure the riverine landscape (Johnson, 2000).
Landscape dynamics in river corridors
Contemporary riverine landscapes reflect processes occurring over a wide range of time scales (Table 1). The following brief discussion of landscape evolution is intended as a prelude to the treatment of dynamic patterns and processes occurring in ecological time (seasonal to centennial).
Table 1. Time scales of some major phenomena that structure patterns and processes in riverine landscapes
The action of running waters (erosion, transport, deposition) is the predominant agent of landscape evolution. Hydraulic action in concert with corrosion and corrasion erodes mineral matter from the banks and bed. These materials are transported by the current and deposited at downstream locations (within the channel, on the floodplain, in standing waterbodies).
On a newly formed landmass not previously subjected to fluvial denudation, the river corridor is in a youthful stage characterised by numerous on-channel lakes, falls and rapids (Fig. 7a). Eventually vertical erosion drains the lakes and reduces the falls. Moreover, the transport capacity of the high-gradient stream exceeds the available sediment load, resulting in additional bed degradation and the creation of deep gorges (Fig. 7b). As weathering and mass wasting act on the canyon walls, the corridor begins to widen (Fig. 7c). As tributary corridors develop in the same manner, the sediment supply to the main corridor increases. As the sediment load increases and the gradient decreases, a time is reached when the sediment supply matches the transport capacity, resulting in a graded channel. This ends the period of rapid downcutting and initiates the formation of the floodplain, as the channel undergoes lateral migration. In the advanced stage of landscape evolution (Fig. 7e), the river corridor consists of a complex low-gradient floodplain containing an expansive alluvial aquifer. The idealised landscape ontogenesis portrayed over geological time in Fig. 7 is analogous to the spatial landscape transformations occurring along river corridors arising in youthful mountains.
Ecological time scales
Rivers are interactive four-dimensional systems (Ward, 1989) characterised by high levels of natural disturbance. The disturbance regime of riverine landscapes encompasses debris flows, fire, disease, and animal activities among other factors. Here we focus on fluvial action as the primary driver of landscape dynamics. Fluvial action influences the riverine landscape directly, by structuring channel networks and other geomorphic features, and indirectly, by influencing successional phenomena. Although landscape dynamics within river corridors are more intense during overbank flood events (the `flood pulse' of Junk et al., 1989), flow pulses below bankfull discharge also have landscape-level implications (Tockner, Malard & Ward, 2000).
Channel migration and avulsion are two phenomena responsible for inducing channel change. Along meandering reaches lateral migration of channels tends to be unidirectional, as is apparent from the zonal array of riparian vegetation with increasing distance from the active channel (Kalliola et al., 1992). In contrast, the migration pattern in anastomosed reaches is patchy rather than unidirectional and the alluvial forest exhibits a mosaic pattern. Annual channel migration rates ranging from 25 to 400 m have been reported from large rivers in the Upper Amazon basin (Terborgh & Petren, 1991; Kalliola et al., 1992), with an extreme rate of 115 m day–1 reported by Puhakka et al. (1992). Based on the lateral erosion rate of 25 m year–1 determined by Terborgh & Petren (1991), the Manu River, Peru, would require 240 years to traverse the 6-km wide floodplain. Much lower lateral migration rates are reported for temperate rivers (Gilvear & Bravard, 1996), which undoubtedly reflects, at least in part, anthropogenic suppression of natural dynamics.
Avulsion refers to the process by which a channel segment is abandoned by an abrupt change in the water's course, such as when a meander loop is cut off to form an oxbow lake. In braided river networks avulsion is a major process of channel dynamics, involving infilling of active channels by sediment and subsequent diversion of the main flow path (Leddy, Ashworth & Best, 1993). Avulsion and lateral channel migration play major roles in the riverine landscape by structuring topographic features of the floodplain, influencing successional trajectories, and determining the turnover rates of landscape elements.
Despite high levels of disturbance and corresponding turnover rates, the relative abundance of landscape elements in a natural river corridor may remain relatively constant over ecological time (e.g. Kollmann et al., 1999), corresponding to the `shifting-mosaic steady state' model of Bormann & Likens (1979). For example, a landscape feature destroyed by localised shear forces is typically re-formed by aggradation processes elsewhere in the corridor. To be more specific, floods eliminate islands at some locations while initiating the formation of new islands at other locations. Therefore, a landscape element that appears to be increasing (or decreasing) at a fine scale of resolution, may in fact be in a steady state if viewed at a broader scale. Fig. 8 illustrates aquatic habitat turnover during a late summer–autumn period in the bar-braided geomorphic reach of an Alpine river with a natural flood regime. The high rate of turnover (53%) that occurred during an average flood season elucidates the high level of landscape dynamics that must have characterised Alpine rivers in the pristine state. Of course different geomorphic reaches responded differently to this flood season. An even higher turnover (62%) was documented in the island-braided headwater reach; the lowest turnover of a floodplain segment occurred in the meandering reach (22%). Turnover in the constrained headwater reach was less than 10%. Avulsion was the mechanism of channel change in the island-braided headwaters and lateral channel migration the dominant mechanism in the meandering reach, whereas both avulsion and lateral channel migration contributed to turnover in the other geomorphic reaches.
Employing GIS-based analysis of aerial photographs from 1984, 1986, and 1991, Kollmann et al. (1999) investigated turnover of landscape elements in the island-braided lower reach of the Tagliamento. Although erosive floods caused dramatic changes in landscape configuration, the relative proportion of river channels, gravel bars and vegetated islands did not change significantly. The turnover rate between 1984 and 1986 ranged from 15% for established islands to 83% for pioneer islands/large woody debris. We believe that high turnover is a general feature of unimpaired riverine landscapes.
The seasonal expansion and contraction of channel networks without overbank flooding, a common phenomenon in river corridors (Stanley, Fisher & Grimm, 1997; Tockner et al., 2000), has received little detailed study. An exception is the detailed analysis of the channel network in the Val Roseg, a glacial stream with a predictable expansion/contraction cycle (Tockner et al., 1997; Malard, Tockner & Ward, 1999; Ward et al., 1999). Total channel length within the 2.6 km long braided floodplain varied from 5.9 to 20.9 km, the highest values occurring during peak glacial melt in summer and the lowest values in winter. The entire floodplain shifts from dominance by glacial meltwater during the expansion phase in summer to a ground water-controlled system during the contraction phase in winter. The aquatic habitats of the glacial floodplain were by no means uniformly cold and turbid. Rather, a remarkable degree of spatio-temporal heterogeneity was documented. Seasonal changes in discharge, the spatial array of the channel network, and landscape patterns in water chemistry within the floodplain were linked to the shifting dominance of different hydrological reservoirs (englacial and subglacial water, hillslope and alluvial aquifers) and flow paths (surface and subsurface) within the catchment. Six lotic habitat types were identified within the floodplain, based on water source, flow paths and hydrochemical indicators: (1) main channel, (2) side channels, (3) intermittently connected channels, (4) mixed channels, (5) entering tributaries and (6) groundwater channels (three subtypes). Although individual groundwater channels were characterised by temporally stable environmental conditions, as a group groundwater channels exhibited higher spatial habitat heterogeneity than any other lotic habitat type, reflecting a diversity of water sources (hillslope aquifer, deep and shallow alluvial aquifers). The main channel and intermittently connected channels, in contrast, exhibited high temporal but low spatial heterogeneity. The other aquatic habitat types were characterised by intermediate levels of temporal and spatial heterogeneity. The high level of collective habitat heterogeneity ensured that clear-water refugia were available in the floodplain year round, even when highly turbid glacial meltwater dominated the system. Thermal heterogeneity between aquatic habitats was also higher than expected for a glacial system (U. Uehlinger, unpublished data).
The influence of the expansion/contraction cycle on riverine landscapes is expected to reflect geomorphic style (Ward & Tockner, 2001). Overall habitat diversity was highest near maximum expansion in the glacial floodplain of Val Roseg, where heterogeneity is largely a function of diversification of water sources and flow paths, rather than autogenic processes within aquatic habitats. In contrast, the Fiume Tagliamento, a flashy Mediterranean river characterised by extensive island-braided reaches, attains the highest landscape-level diversity at intermediate discharge. At higher levels inundation of the entire Tagliamento floodplain has a homogenising effect, whereas the glacial floodplain sustains a distinct network of running water channels even during major spates. Expansion/contraction cycles are intimately related to connectivity, a topic addressed in a later section of this paper.
Over longer ecological time scales, the erosive action of flooding is countered by ecological succession. In constrained reaches succession is maintained in early seral stages by the concentration of fluvial energy in a single-thread channel, whereas a diversity of seral stages is apparent in floodplain reaches. Therefore, at the river corridor scale, concomitant with the alternation of constrained and floodplain reaches (see Fig. 1), there is a downstream spatial pattern of alternating successional trajectories. The following discussion focuses on the different types of succession occurring in alluvial floodplains: as some floodplain waterbodies slowly fill and undergo terrestialisation, other waterbodies are newly formed or rejuvenated; as portions of floodplain forests are undercut by channel migration, seedlings are established on newly deposited alluvium; as mature islands are eroded by a flood, LWD deposited on gravel bars serves as nuclei for island formation. These three types of succession – hydrarch succession, floodplain forest succession and alluvial island succession – are briefly addressed below. First, however, we will examine the major successional trends for key components of the riparian flora, aquatic vegetation and aquatic fauna associated with the rapid terrestrialisation of a meander that in 1956–57 was cut off from the Ain River in France (Fig. 9). Willows (Salix eleagnos L.) occurring on the highest alluvial surfaces were the only woody riparian plants in 1965. Subsequently, as the floodplain surface was elevated relative to the water level, poplar (Populus nigra L.) and ash (Fraxinus) initially displaced S. eleagnos to lower elevation sites (1980). Terrestrialisation was predicted to eliminate S. eleagnos by 2000, at which time Populus and Fraxinus would co-dominate, hawthorn (Crataegus) would become abundant, and elm (Ulmus) and the willow S. purpurea would occupy sediment deposits in filled channel segments. The open water habitats with coarse substratum and some hydrophytes (not shown) in 1965 were partially filled by fine organic sediment and densely colonised by hydrophytes and helophytes by 1980 and predicted to become temporary waters by the year 2000. Salmonids (Thymallus thymallus L.), which occurred in the abandoned meander in 1965, were replaced by pike (Esox lucius L.) by 1980; the fish fauna would disappear completely by 2000 concomitant with the disappearance of permanent water. Gammarus, a predominant member of the zoobenthos in 1965 gave way to Asellus, anisopteran and coleopteran larvae as eutrophication proceeded. Benthic chydorids were also abundant in 1980.
Hydrarch (hydrosere) succession refers to the changes that occur over time (eutrophication/terrestrialisation) following the initial formation of lentic waterbodies. In river corridors successional trajectories are confounded by floods, which not only initiate hydrarch succession by forming new waterbodies by channel abandonment, but also reset successional sequences through rejuvenation of existing waterbodies. Waterbodies that are situated far from the active channel or those otherwise shielded from fluvial action by their position in the riverine landscape, undergo different sucessional trajectories than waterbodies subjected to frequent flooding. This was well illustrated by a study of six braided channel segments isolated from the main course of the Rhone at about the same time in the late 19th century (Bornette, Amoros & Chessel, 1994). Only two of the six channels had progressed to an advanced stage of hydrarch succession, characterised by eutrophic species and the accumulation of fine organic sediment. Those abandoned braids are protected from flood scouring by the forest vegetation growing on the alluvial plug. Conversely, the four other channels of the same age exhibited much retarded successional development, as indicated by oligo-mesotrophic species assemblages. Two of these had coarse alluvial plugs without woody vegetation and lacked protection from flood scouring. The other two were located far from the active channel and had forested alluvial plugs, but were fed by cold, nutrient-poor ground water, that slowed the rate of succession.
Floodplain forest succession is initiated on freshly deposited alluvium, a major landscape element of natural rivers. For example, recent alluvial deposits were estimated to occupy about 26% of the Amazonian lowlands of Peru (Salo et al., 1986). Terborgh & Petren (1991) provide a useful summary of alluvial forest succession along the Rio Manu, a meandering Peruvian river with a 6-km wide floodplain. During annual floods, as the forest is undercut along the concave banks of meander bends, primary succession is initiated on fresh alluvium deposited as point bars on convex bends. As water levels decline during the dry season, the woody composite Tessaria integrifolia Ruiz & Pav. germinates on the point bars and grows to heights up to 2 m before the next flood season. The next flood removes many Tessaria plants, but some stems are flattened and buried within the new layer of sediment. The buried stems sprout vegetatively and Tessaria attains heights of 3–4 m during the next dry season. Vertical accretion on the floodplain surface and a greater distance from the migrating river allows upright plants to withstand the next flood season and grow to full size (8–10 m) within 3–5 years. The cane Gynarium invades and eventually overtops Tessaria, causing its demise. Cane stands provide suitable conditions (adequate light penetration, reduced flood scour) for seedling establishment by many tree species. Subsequent invasion of the cane by Cecropia, a fast-growing, short-lived tree, forms a Gynaria–Cecropia association that lasts several years, but is eventually replaced by taller, slower-growing trees. A transitional mixed forest stand persists for several decades during which a dense herbaceous understory suppresses further tree recruitment. Two slow-growing trees (Ficus and Cedrella) that persist for about a century and attain heights of 35–40 m are co-dominant in the late successional stage of the floodplain forest. The mature floodplain forest has five vertically superimosed strata (the highest exceeding 50 m) with over 200 tree species per hectare. This mature stage requires 300–500 years to develop and is therefore limited to areas of the floodplain with lower than average rates of channel migration. The spatially variable rates of channel migration, in concert with the temporal sequence of seral stages, each characterised by distinctive three-dimensional structure, results in a diversity of forest patches across the dynamic riverine landscape.
Forest succession on temperate zone floodplains, although involving fewer species and lower structural diversity, exhibits generally similar trajectories to that described above for Amazonia. In the north temperate zone, the pioneer stage consists of patches of shrubs and tree saplings of Populus and Salix, the predominant woody plants on fresh alluvial deposits (Schnitzler, 1997). Like Tessaria of Amazonia, they are ruderal strategists (sensuGrime, 1977), enabling them to colonise frequently disturbed environments. These pioneer softwoods are tolerant of fluvial action, burial and submersion, but are vulnerable to shade and drought. They are weak competitors that are replaced by other species as landforms stabilise. The transitional or pre-equilibrium forest of European floodplains (defined by canopy closure of the Salix/Populus association) is characterised by two strata, a canopy c. 20 m high and the grass layer of 2 m height (Schnitzler, 1997). Canopy closure occurs within one or two decades. These softwoods are gradually replaced by hardwoods emerging from the understory. The `climax' (terminal successional stage) of the floodplain forest, characterised by the hardwoods Ulmus, Fraxinus and Quercus, consists of five strata with a canopy reaching 30–35 m in height. It takes two to three centuries to reach the terminal hardwood stage of alluvial forest succession. However, given that the rivers of Europe have been regulated for centuries, it is possible that the hardwood association may not represent the primeval floodplain forest of the active floodplain; prior to flow regulation natural fluvial dynamics may have prevented alluvial forests from developing beyond the softwood successional stage.
Island formation is typically initiated when LWD, often an uprooted tree that is being transported downstream, lodges on the crest of a gravel bar on the declining limb of a flood hydrograph. The root plate of the tree usually faces upstream with the trunk and branches trailing downstream. A plume of fine sediment accumulates immediately downstream from the LWD accumulation, partially or completely burying the trunk and branches. Shoots sprout from the buried wood of Salix and Populus during the first growing season. This complex of LWD, sediment, entrapped fine organic matter and regrowth from buried wood constitutes the incipient stage of island formation, termed phase I islands by Edwards et al. (1999). If suitable conditions prevail, phase I islands develop into pioneer islands (phase II) with young Salix and Populus trees several years old forming a canopy up to 4 m high. The woody vegetation of established islands (phase III) is dominated by patches of shrubs consisting of Salix spp. and a tree canopy up to 20 m high formed by Populus, Alnus and S. alba. Established islands have a distinctive community composition, including an assemblage of species less tolerant of inundation, but more tolerant of drought and low nutrient levels (Kollmann et al., 1999). Succession from gravel bars to phase III islands requires 10–20 years, during which 1–2 m of fine sediment is deposited on top of the underlying gravel bar. Although the probability of an island being washed away decreases as islands develop, islands of all phases may succumb to erosive forces, their accumulated materials contributing to island formation downstream. Nonetheless, an important effect of island succession is the creation of relatively stable habitats in the active floodplains of highly dynamic rivers. In this process of island succession, the plants serve as autogenic ecosystem engineers (sensuJones, Lawton & Shackak, 1994), trapping sediment and organic matter, thereby structuring the riverine landscape.
It is clear that phenomena associated with hydrarch, alluvial forest and island succession greatly contribute to riverine landscape diversity. The broad range of seral stages occurring simultaneously in natural river corridors results in landscape patches differing markedly in successional trajectories, geomorphic features, environmental conditions and biotic structure. However, if unchecked, succession eventually leads to reduced environmental heterogeneity: waterbodies undergo rapid terrestrialisation; islands merge with the floodplain, forming a single-thread channel; alluvial forests senesce. A high degree of natural disturbance is necessary to counter such successional trends. Connectivity, induced by the kinetic energy of flooding (fluvial dynamics), is the countervailing mechanism that sustains heterogeneity across the riverine landscape.
Connectivity between landscape elements
The concept of connectivity, initially applied to gene flow between subpopulations of a metapopulation (Merriam, 1984), here refers to hydrological connectivity, the exchange of matter, energy and biota between different elements of the riverine landscape via the aqueous medium (Amoros & Roux, 1988). Connectivity operates at different scales, as illustrated by the following examples of the distribution of animals in subsurface waters. At a fine scale of resolution (centimetres to meters), different faunal elements exhibit different degrees of connectivity between surface waters and ground waters (Gibert et al., 1990). The subterranean amphipod Niphargus freely traverses the ecotone (hyporheic zone) between ground water and surface water, whereas the hyporheic zone is a barrier for the strictly subterranean amphipod Salentinella. The surface-dwelling amphipod Gammarus and some aquatic insects freely colonise the hyporheic zone, whereas other epigean forms are confined to surface waters. Floods alter connectivity within the alluvium, with corresponding fine-scale shifts in faunal distribution patterns within the sediment interstices (Marmonier & Dole, 1986). In an expansive alluvial aquifer characterised by high hydraulic conductivity, specialised stoneflies migrate up to 2 km from the river beneath the floodplain during their 2-year larval stage, returning to surface waters of the river to emerge, mate and oviposit (Stanford & Ward, 1988). True groundwater crustaceans, in contrast, are confined to the alluvial aquifer. Hydrological connectivity varies greatly between different sedimentary units within the fluvial stygoscape. Preferential flow paths (buried former river channels) transmit a disproportionate amount of water through alluvial aquifers and provide especially favourable environmental conditions for groundwater fauna.
Fluvial dynamics, including the expansion/contraction of surface waters, is the primary driving force that sustains connectivity in alluvial rivers. Different types of floodplain waters may be arrayed along a gradient based on their degree of surface connectivity with the main channel. For example, abandoned braids lack surface connectivity during base flow, but are reconnected frequently by relatively small increases in water level. In contrast, abandoned meanders may require exceptional floods to re-establish connectivity. Reconnection serves as a reset mechanism, rejuvenating abandoned waters through scouring action that removes accumulations of fine sediments, organic detritus and organisms. Other materials, including propagules, are imported during reconnection.
Connectivity in alluvial rivers is, however, a more complex phenomenon than implied by simple gradients. The example presented earlier demonstrated how hydrological connectivity of abandoned braids may be reduced by vegetation growing on alluvial plugs or increased by upwelling of nutrient-poor ground waters, in both cases altering hydrarch successional trajectories. Some additional manifestations of connectivity, based on examples from European rivers, follow.
Fig. 10 compares shoreline ecotone length as a function of water level in floodplains of two rivers, both of which have dynamic flow regimes but differ in connectivity. Low connectivity in the Danube is engendered by artificial levees that limit surface connectivity between the river and the floodplain, whereas the Tagliamento is unconstrained. The Tagliamento had an ecotone length 2–3 times greater that that of the Danube during the study period. In addition, ecotone length remained constant over a wide range of discharge and exhibited high resilience to major floods. Despite dramatic expansion/contraction cycles, in the intact floodplain suitable shoreline habitats for edge species were available in similar quantities throughout the year (K. Tockner, unpublished data).
Landscape indices (Gustafson, 1998) were used to investigate seasonal dynamics in the spatial configuration of the braided channel network of Val Roseg, a glacial floodplain in the Swiss Alps (Malard et al., 2000). Connectivity was expressed as the relative proportion of the channel network length having an upstream surface connection with the main river channel. When surface connectivity was plotted against total channel length, two distinct phases were apparent. During the first expansion phase channel length more than doubled without a substantial increase in surface connectivity. During this period snow melt recharged hillslope and floodplain aquifers via subsurface pathways; increased surface flow was largely restricted to groundwater-fed tributaries and alluvial channels. Only when total channel length exceeded a threshold (15 km in this 3-km long floodplain) was there a corresponding increase in surface connectivity. These relationships have important implications, not only for groundwater–surface water interactions, but for colonisation pathways and refuge use by the aquatic fauna in this harsh alpine landscape.
Water temperature was recorded continuously for 1 year in 18 waterbodies along a connectivity gradient in the Alluvial Zone National Park, Austrian Danube (Fig. 11). Habitats ranged from the main river channel to an isolated pond at the edge of the floodplain. During the winter months there was a more or less homothermous spatial pattern. However, for much of the year, a remarkable degree of thermal heterogeneity is apparent across the riverine landscape, caused in part by strong upwelling of cooler ground waters at some locations. During the summer differences in daily mean temperatures between the warmest and the coolest habitats were as much as 16 °C, greater than that which occurs along the entire length of the main channel. Therefore, the habitats within this single floodplain collectively provide a wide range of temperature regimes, meeting the diverse thermal requirements of different aquatic species.
These few examples demonstrate that connectivity plays a major role in riverine landscapes (see also Amoros, 2002), although this phenomenon has not been afforded the attention it deserves. Detailed analysis of connectivity in diverse river systems should provide considerable insight into structural and functional attributes of riverine landscapes, including a greater understanding of the factors structuring biodiversity patterns.
Biodiversity in riverine landscapes
Biodiversity (sensuNoss, 1990; Ward & Tockner, 2001) includes structural diversity and functional diversity, thereby encompassing much of what has already been presented in this paper. In this final section, we explore the relationship between spatio-temporal heterogeneity and species diversity in a landscape context.
As stated by Levine (2000a), `The study of biodiversity is the study of how competitive exclusion is foiled – through the exploitation of heterogeneity and pattern in the environment….' The biota inhabiting riverine landscapes are indeed adapted to exploit the spatio-temporal heterogeneity engendered by natural disturbance regimes (Junk et al., 1989; Petts & Amoros, 1996; Naiman & Décamps, 1997; Schnitzler, 1997). Maintenance of community diversity is a composite function of the range of resources available and the degree to which species segregate those resources. In this context, landscape spatial heterogeneity expands the resource gradient (e.g. Fig. 4) and temporal heterogeneity (e.g. Fig. 11) increases the potential for niche overlap.
Natural disturbances are responsible both for structuring spatial heterogeneity and for creating conditions under which niche overlap can occur. For example, Schnitzler (1994) considers disturbance by fluvial processes largely responsible for the fact that the alluvial hardwood association (Querco–Ulmetum) is one of the most productive and speciose forest ecosystems of the temperate zone. In the Upper Rhine, the Querco–Ulmetum association, with 25 species of trees and 23 species of shrubs, consisted of five forest units (phytosociological sub-associations). Three of these were arrayed along a flood-risk gradient on the active floodplain. The fourth was associated with terraces and the fifth forest unit occurred in a part of the floodplain influenced by a major tributary carrying more acidic water and suspended clay. Schnitzler, Carbiener & Trémolières (1992) emphasise the role of fine-scale environmental gradients in the ecological segregation of closely related plant species in alluvial forests. For example, nine congeneric species of willows coexist in close proximity in the upper Rhine, segregated along gradients of soil acidity, sediment grain size, habitat stability and groundwater level, among other factors.
Dole (1983) examined interstitial faunal diversity within alluvial sediments (– 0.5 m depth) in three floodplain habitats representing a disturbance gradient on a Rhone River floodplain. A total of 112, 157 and 107 species were identified, respectively, from the active channel (high disturbance), side arm (moderate disturbance) and an isolated meander arm (low disturbance). The active channel contained a species assemblage resistant to disturbance, whereas an assemblage sensitive to disturbance occurred under the stable conditions characterising the abandoned meander. The proportion of hypogean forms was strongly correlated with the disturbance gradient, constituting 8, 20 and 47% of the total fauna from the active arm to the meander.
The intermediate disturbance hypothesis (IDH; Connell, 1978) predicts low community diversity in environments exposed to high disturbance levels where only highly tolerant or fugitive species can occur, as well as in low disturbance environments where superior competitors monopolise resources. According to the IDH, species diversity is highest in situations experiencing intermediate levels of disturbance, which constrains competitive exclusion, thereby allowing the coexistence of taxa with divergent species traits and adaptive strategies. Salo et al. (1986) proposed that river dynamics, by constraining competitive exclusion, is responsible in part for the high biodiversity of the upper Amazon basin. Pollock, Naiman & Hadley (1998) reported highest species richness in wetland plant communities subjected to intermediate flood frequencies. However, only limited attempts have been made to empirically test the IDH in complex riverine landscapes.
The only investigation known to us designed specifically to test the IDH in riverine floodplains examined the relationship between connectivity and aquatic plant diversity in 12 waterbodies that were inundated by the river from 0 to 37 days year–1 (Bornette et al., 1998b). The authors postulated that high connectivity would constrain diversity by excessive flood scouring; that low connectivity would constrain diversity through competitive exclusion; whereas intermediate connectivity would result in the highest species richness. No clear relationship between connectivity and species diversity was demonstrated; however, a complex of interacting factors structuring aquatic plant diversity in floodplain waterbodies was identified and incorporated into a predictive model (Amoros & Bornette, 1999). They found that the scouring effect is determined not only by frequency of inundation, but also by the configuration of the waterbody in the landscape; that groundwater connectivity may influence aquatic plant diversity; that nutrient levels, shading by phytoplankton or turbidity may override other factors; that submerged and floating macrophytes respond differently to connectivity; that silting and scouring are both important; that degree of connectivity determines whether sexual or vegetative reproduction predominate and whether antiherbivore defences are needed; and that retention structures determine whether or not entering propagules become established.
The relationship between species richness and connectivity is therefore determined by complex relationships among several interacting variables. In addition, species richness maxima for different faunal and floral elements occur at different positions along the connectivity gradient (Tockner, Schiemer & Ward, 1998). Fish diversity, for example, may peak in highly connected habitats, whereas amphibian diversity tends to be highest in habitats with low connectivity. Other groups attain maximum species richness between these two extremes. The resulting pattern is a series of overlapping species diversity peaks along the connectivity gradient, suggesting that habitats collectively traversing a broad range of connectivity will optimise community diversity in riverine landscapes.
The dynamic equilibrium model of Huston (1979, 1994) adds a resource dimension (productivity) along the disturbance axis of the IDH. This integrated model predicts that the level of disturbance at which species diversity is maximised varies as a function of the resource base, because greater disturbance is needed to prevent competitive exclusion as productivity increases. Alluvial floodplains provide an ideal setting for investigating how the balance between productivity and disturbance influences biodiversity patterns. Apart from Pollock et al. (1998), whose work on plant species richness in wetlands, including some floodplain sites, provides partial support for the dynamic equilibrium model, we are not aware of research that has been conducted in this context. However, Yachi & Loreau (1999) developed a general stochastic dynamic model to examine the relationship between biodiversity and productivity in fluctuating environments. The model output predicts increased ecosystem productivity and reduced variance in productivity in more speciose systems. Tilman (1999) predicts that the more complete utilisation of limiting resources by diverse communities reduces their susceptibility to invasion by exotic species. Results from experimental manipulations of riparian plant tussocks, however, suggest that the intrinsic invasion resistance of diverse local assemblages may not be realised at larger scales (Levine, 2000b).
There is a great need for rigorous empirical studies that examine the relationship between landscape diversity and species diversity in alluvial river corridors. The species pool in riverine landscapes is derived from terrestrial and aquatic communities inhabiting diverse lotic, lentic, riparian and groundwater habitats arrayed across spatio-temporal gradients (Fig. 12). Understanding the complex interactions between disturbance regimes, spatial heterogeneity and biodiversity in riverine landscapes provides a major challenge for the future.
Riverine landscapes in the natural state exhibit high levels of complexity across a range of scales. The expansive perspective provided by landscape ecology holds the potential for developing a truly holistic perspective of river corridors, one that rigorously integrates structure, dynamics, and function. In addition, landscape ecology offers an effective framework to integrate pattern and process in river corridors, to examine environmental dynamics and interactive pathways between landscape elements, to link research with management, and to develop viable strategies for river conservation. Remote sensing provides a variety of techniques for analysing landscape patterns and dynamics in river corridors, with promises of even greater utility in the future (Mertes, 2002). A more complete understanding of ecological processes from a landscape perspective, will not only advance the discipline of river ecology, but will also enhance the effectiveness of conservation and restoration initiatives. However, present understanding indicates the futility of managing for the optimal level of a process (connectivity for example). A more balanced approach is to assure that a given process varies sufficiently among individual habitat patches to sustain a broad range of that process at the landscape scale.
Our recent research has been supported by grants from the Swiss National Science Foundation (SNF grants 21-49243.96; 31-50440.97; 31-50444.97) and the ETH Forschungskommission (0-20572-98).