The science of connected ecosystems: What is the role of catchment-scale connectivity for healthy river ecology?

1 INTRODUCTION

Globally, rivers and streams are under considerable and increasing threat from a wide variety of anthropogenic stresses (Dudgeon, 2010; Dudgeon et al., 2006; Vorosmarty et al., 2010). Freshwater biodiversity is declining considerably more rapidly than other equivalent habitats from multiple interacting stressors (Leps, Tonkin, Dahm, Haase, & Sundermann, 2015; Matthaei, Piggott, & Townsend, 2010; Piggott, Lange, Townsend, & Matthaei, 2012; Wagenhoff, Townsend, Phillips, & Matthaei, 2011) including water abstraction for consumptive and agricultural needs (Dewson, James, & Death, 2007; McDowell, van der Weerden, & Campbell, 2011; Poff & Zimmerman, 2010), invasive species (Collier & Grainger, 2015; Olden et al., 2010), channelisation, sedimentation, eutrophication (Allan, 2004; Carpenter et al., 1998), and changing climate regimes (Death, Fuller, & Macklin, 2015; Palmer et al., 2008). Similarly, the geomorphic structure of river systems has been and continues to be extensively modified by changing flow regimes from damming, abstraction, and climate alteration (Death et al., 2015; Lobera et al., 2015; Magdaleno & Fernández, 2011). Furthermore, channelisation has isolated waterways from people and their properties, disconnecting river channels from adjacent floodplains (Fuller & Basher, 2013; Gilvear & Winterbottom, 1992; Surian, 1999; Surian & Rinaldi, 2003).

Considerable effort is now being directed towards halting and restoring some of the dramatic declines in river ecosystem health by reducing and removing the changes humans have imposed on the biological and/or geomorphological components of those rivers (Death, Bowie, & O'Donnell, 2016; Kail, Brabec, Poppe, & Januschke, 2015; Leps, Sundermann, Tonkin, Lorenz, & Haase, 2016; Palmer, Hondula, & Koch, 2014). Approaches to river restoration often involve re-establishing connections between the active channel and the floodplain, or between estuaries and lowland reaches and river headwaters at a catchment-scale. This can range from permitting simple channel widening to the removal of often large dams that re-establish flow regime, sediment regime, and geomorphology, and over time, much of the ecology (Hart et al., 2002; Tullos, Finn, & Walter, 2014). However, although local habitat condition may be restored, this does not always result in the desired restoration of healthy river systems (Bernhardt et al., 2005; Bernhardt & Palmer, 2011; Lake, Bond, & Reich, 2007; Palmer et al., 2014; Palmer, Menninger, & Bernhardt, 2010). The field of dreams hypothesis postulates that if habitat is restored then the appropriate organisms will recolonise, and correct ecosystem functioning will resume, however, that is often not the case (Hilderbrand, Watts, & Randle, 2005; Palmer et al., 2014; Palmer, Ambrose, & Poff, 1997; Sudduth, Hassett, Cada, & Bernhardt, 2011). Although this may have happened in spatially diverse pristine landscapes, humans have altered the species pool and biophysical flows to such an extent that this often does not occur. In a review of 644 river restoration projects where the physical habitat was restored, Palmer et al. (2014) found only 16% that showed a concomitant change in biodiversity, even many years postrestoration.

Increasing connectivity can take several forms depending on the nature of the biophysical flow. The focus of this paper is sediment connectivity, defined as the ‘connected transfer of sediment from a source to a sink’ (Bracken, Turnbull, Wainwright, & Bogaart, 2015), which can be construed as occurring at a catchment scale (see Brierley, Fryirs, & Jain, 2006). In a given catchment, multiple sediment sources exist. Connected sediment transfer takes place on hillslopes, between hillslopes and channels (slope–channel coupling), between floodplain and channel (bank erosion, floodplain deposition: lateral and vertical accretion), and within channels (longitudinal connectivity associated with sediment transport). Sediment conveyance within a catchment is maximised by a high degree of sediment connectivity (e.g., Fuller, Riedler, Bell, Marden, & Glade, 2016; Hooke, 2003; López-Tarazón, Batalla, Vericat, & Francke, 2009); and connectivity is critical to both the functioning of sediment cascades and sensitivity of the landscape to change (Fryirs, 2013; Fryirs, Brierley, Preston, & Kasai, 2007; Harvey, 2001; Hooke, 2003; Wohl et al., 2015). A high degree of sediment connectivity means that sediment generated by an erosion event is effectively transferred through the sediment cascade, and may register a geomorphic response (Harvey, 2001). In these highly connected systems, the structural resistance to change is reduced as there is a free transfer of energy and matter in the system (Brunsden, 2001). In this paper, we refer to these transfers at the catchment scale via both lateral and longitudinal connectivity. We define lateral connectivity as sediment transfers between hillslopes and river channels (potentially via floodplains), which delivers sediments from the hillslope domain via geomorphic processes. At the small scale, this can be construed as analogous to hillslope–channel coupling (sensu Harvey, 2001; Fuller & Marden, 2011). We define longitudinal connectivity as sediment transfers taking place within (along) the river channel itself, together with its interaction with the floodplain (sensu Hooke, 2003; Wohl et al., 2017).

In contrast, although enhanced sediment connectivity can have major effects on geomorphological condition and be potentially detrimental to most fauna and flora, the transmission of water and nutrients, on which these fauna and flora depend, relates to the hydrologic connectivity of a system (Jencso, McGlynn, Gooseff, Bencala, & Wondzell, 2010), because hydrological processes control most nutrient export from the land surface (Stieglitz et al., 2003). More broadly, and in an ecological sense, Pringle (2001, p. 981) defines hydrologic connectivity as ‘the water-mediated transfer of matter, energy, and /or organisms within or between elements of the hydrologic cycle,’ which is essential to ecological integrity (Pringle, 2003), and develops the concept proposed by Amoros and Roux (1988). Ecological condition is thus dependent on this hydrologic connectivity, which relates to transfers along surface and subsurface hydrological pathways laterally, vertically, and longitudinally, at a range of temporal scales (Covino, 2017). In terms of units within a catchment, hydrologic connectivity operates broadly between hillslopes and channels and more specifically via floodplains, hyporheic zones, downstream reaches, and upstream tributaries. Hydrologic connectivity essentially relates to connections of water and nutrients between these key riverine habitats. Fauna are usually highly mobile and seek refuge in discrete refugia within these zones during high or low flow disturbances and/or move between these habitats throughout their life history (Death et al., 2015). Thus, although high sediment connectivity may be detrimental to the fauna and flora of particularly sensitive rivers, the fragmentation and severing of hydrological connectivity and its impacts on habitat may be equally, if not more, detrimental to healthy ecosystems. Importantly, Wohl et al. (2015) recognise the need for both flow and sediment regimes to be understood for effective management of river systems, but all too often this integration is lacking.

2 THE DRIVERS OF SEDIMENT CONNECTIVITY AND IMPLICATIONS FOR RIVER HABITAT

Short-term channel dynamics and catchment connectivity in headwater catchments in the Southeastern Ruahine Ranges of New Zealand (Figure 1) have been examined by Schwendel, Fuller, and Death (2010) and Schwendel and Fuller (2011). Considerable variability in short-term reach-scale morphodynamics appeared to be driven by differences in sediment supply rate, which was modulated or amplified by the degree of sediment connectivity in the catchments (Schwendel et al., 2010; Schwendel & Fuller, 2011). This work found sediment supply to the study reaches was over time increasingly dependent on reworking of material stored in-channel and on valley floors from past intensive slope erosion from Cyclone Alison in 1975. In this case, longitudinal connectivity pathways downvalley were providing the conduits for sediment, not lateral connectivity with the adjacent slopes. Valley floors in this landscape had a clear memory of past events from the legacy of extreme climate events (cf. Brierley, 2010). It is important to recognise that geomorphic history of these sites drives sediment connectivity (Fuller et al., 2016). Changes in the landscape, in this case, deforestation of foot-slopes in the ranges, combined with storms, determined sediment connectivity pathways (e.g., Blakely, 1977; Grant, 1981; Grant, Hawkins, & Christie, 1978; Marden, 1984; Mosley, 1977; Mosley, 1978a; Schumm, 1977). Blakely (1977) suggested that this disturbance resulted in retrograding incision upstream, scour of old in-channel deposits, and slope failure induced by bank erosion. Storm-induced landsliding coincided with a small reduction in vegetated area by 2.8% between 1946 and 1974 (Grant, 1989), during which time a 91% increase in erosion was observed (Marden, 1984). An increase in erosion in this landscape is significant in terms of sediment connectivity because Fuller et al. (2016) suggested up to 78% of landslides in the southern Ruahines connect with headwater channels. Importantly, Fuller et al. (2016) also suggested that 100% of material is evacuated from landslide source zones (i.e., landslide scars). Their research indicated that the lateral connectivity in this environment must be considered high. However, although lateral connectivity between hillslopes and channels was strong, Fuller et al. (2016) found longitudinal connectivity was comparatively weaker, because the legacy sediment (cf. James, 2013; Wohl, 2015) from 1970s erosion remained in the valley floor for several decades, being progressively reworked, but not evacuated (Schwendel et al., 2010; Schwendel & Fuller, 2011).

The angularity and poor sorting of stream sediments in the Ruahine headwater streams (Fuller et al., 2016; and cf. Figure 2) indicate a local, landslide origin. Thus, ‘background’ frequent slope erosion may at first sight appear not to be readily transmitted from headwater streams to the trunk channels, as occurs elsewhere (Johnson et al., 2008). However, the volume of sediment in these valley floors should not be overlooked, and it is likely that the complete overwhelming of these valley floors by sediment from erosion in the 1970s is the reason for large quantities of this sediment continuing to contribute to the downstream sediment cascade. The reworking and delivery of this legacy sediment downstream (which has been observed elsewhere in Australia and New Zealand in response to postsettlement disturbance, e.g., Brierley et al., 1999; Fryirs & Brierley, 1999, 2001; Chappell & Brierley, 2014; Fuller, Macklin, & Richardson 2015), with its effects on channel dynamics and habitats at the range front (Schwendel et al., 2010), is an example of effective longitudinal connectivity. Benches abandoned by valley floor excavation indicate that the deposits are being actively reworked and contributing sediment downstream (Figure 2). However, the degree of valley confinement may be insufficient to promote as efficient a conveyance downstream when compared with systems described in Taiwan by Kuo and Brierley (2014). There, very short residence times were attributed by Kuo and Brierley (2014) to valley floor configuration and a high frequency of large magnitude events responsible for downstream sediment transfer. In the Ruahines, a high magnitude event, such as a cyclone, would also be likely to further enhance longitudinal connectivity, delivering large quantities of coarse sediment downstream (cf. Mosley, 1978a). The catchment configuration, which sets the limits on lateral (hillslope–channel) connectivity and down-valley longitudinal connectivity, together with event magnitude, frequency and timing, all condition the efficacy of sediment transfer in this type of landscape (Fuller et al., 2016). With an increasing frequency of storm events predicted with climate change, it will be important for river managers to consider the longer term dynamics and sources of sediment and the connectivity pathways. It will be pointless restoring river reaches if sediment from past legacy events upstream continues to migrate downstream and smother habitat and biological restoration, cf. Brierley and Fryirs (2009). There is thus a need to take into account present and future sediment transport if restoration is to be effective.

An individual coupling event facilitating lateral sediment connectivity (e.g., landslide to channel) may in the Ruahines have a return period of decades, but down-system longitudinal sediment connectivity appears to operate over decades to centuries, which is the likely timescale for sediment to be reworked from the low-order headwater channels to the trunk streams (Fuller et al., 2016). This timeframe is consistent with the timescales of coupling envisaged by Harvey (2002). The differential behaviour of channels observed by Schwendel et al. (2010) and Schwendel and Fuller (2011) in selected valleys along the foot-slope of the range front, where channel slope abruptly declines and valley floor width increases, could be attributed to differential recovery following major disturbance attributed to Cyclone Alison in 1975, which in turn may reflect different structural and transmission resistance in these catchments (cf. Brunsden, 2001), with resistance increasing southwards (Fuller et al., 2016). Harvey (2007) observed adjacent systems in the Howgill Fells recovered to differing degrees following a ‘100-year’ annual recurrence interval storm event in 1982. Catchments with less resistance do not recover as much, or as quickly, as those with a greater degree of resistance, because sufficiently frequent landslide activity in low-resistance catchments sustains sediment delivery to the stream network (cf. Fryirs & Brierley, 2016). In these systems, lateral connectivity is enhanced and longitudinal connectivity transfers this sediment downstream, resulting in highly unstable channels over decades to centuries (Fuller et al., 2016; Schwendel et al., 2010). Structurally, these catchments have changed, with connectivity enhanced, and sensitivity to disturbance from storm events increased, coincident with the degradation of forest canopy (Mosley, 1977).

The highly unstable nature of the channels and substrate, in turn, determine the diversity and composition of the invertebrate communities living in these streams (Minchin & Death, 2002; Schwendel, Death, Fuller, & Joy, 2011). As discharge increases during rainfall, movement of the substrate crushes invertebrates, or for many species results in them launching themselves, deliberately or accidently, into the water column (termed drift, this is the passive dispersal of invertebrates downstream in the water current; Death, 2008). The time before the next substrate-moving event then determines the recovery of species from refugia where substrate did not move. Thus, the more frequently and severely these substrate-moving events occur, the fewer species and animals that are able to recolonise; as a result, streams with such highly mobile sediments have fewer animals and species (Death, 2002; Death & Winterbourn, 1995; Death & Zimmermann, 2005). The speed of recolonisation is also affected by the recovery of their principal food resource, periphyton (unicellular algae growing on the substrate), which will also have been removed by scour and substrate movement (Death & Zimmermann, 2005). Not only will the diversity and abundance of invertebrates be limited by this unstable substrate movement (Figure 3), but also recolonisation will be limited to those types of taxa able to escape such flood events and/or recolonise quickly (Death, 1995; Death, 2004; Death, 2008). As a result, the few species able to survive in these Ruahine streams are the mayfly Deleatidium sp., blackfly Austrosimulium sp., and several species of Chironomidae (Minchin & Death, 2002; Schwendel, Death, et al., 2011).

3 BIGGER STORMS RESULT IN LARGER AND LONGER LASTING EFFECTS ON GEOMORPHOLOGY AND ECOLOGY

The East Coast of New Zealand's North Island delivers some of the highest sediment yields to the ocean on the planet, accounting for 33% of New Zealand's total sediment yield from 2.5% of the area (Page et al., 2007). In particular, the Waiapu catchment has an annual average suspended sediment yield of 35 Mt yr^-1-, which equates to 20,520 t km⁻² yr⁻¹ (Hicks, Gomez, & Trustrum, 2000), or 0.2% of the global sediment yield from 0.0002% of the terrestrial surface. These catchments are characterised by strong lateral connectivity between a range of erosion sources (notably gullies, landslides, and earthflows) and stream channels (Page et al., 2007). Such a high degree of sediment connectivity is attributed in part to a large magnitude event in March 1988 (Cyclone Bola), with a 100-year annual recurrence interval, which resulted in widespread erosion and off-slope conveyance of sediment to the stream network. This 100-year event changed the geomorphic structure in some catchments, opening up large erosion sources in previously pristine forest slopes (Parkner, Page, Marden, & Marutani, 2007). The consequent strong lateral connectivity resulted in large-scale and widespread valley-floor aggradation (Page et al., 2007; Tunnicliffe et al., 2018; Tunnicliffe, Leenman, & Reeve, 2014). Valley floor aggradation was further facilitated by strong longitudinal connectivity moving sediment downstream, especially fine grained sediment generated from erosion of weak, chemically weathered, and deformed Cretaceous shales. Cyclone Bola generated a wave of sediment working through the catchment, responsible for over 20 m of aggradation in some rivers (Tunnicliffe et al., 2018) and an almost completely denuded instream biological fauna (A. Death pers comm.).

Small headwater systems such as the Weraamaia, which initially also aggraded rapidly, have since degraded following vegetation regrowth and stabilisation of slopes, reducing lateral sediment connectivity to channels within 10 years of Cyclone Bola (Kasai, 2006; Kasai, Brierley, Page, Marutani, & Trustrum, 2005). This catchment response has improved stream health (cf. Parkyn et al., 2006). These landscape responses in the East Coast Region suggest rapid structural change (Brunsden, 2001) and sediment conveyance, although Jones and Preston (2012) also identified far lower rates of sediment transfer from slopes in lower magnitude events responsible for shallow landsliding in less confined soft-rock terrain in the Waipaoa due to effective buffering (cf. Fryirs, Brierley, Preston, & Kasai, 2007) at the toe of the slope, which reduced the efficiency of lateral connectivity. It is therefore important to understand landscape context and the differing operation of sediment cascades contingent upon the operation of both structural and transmission resistance (Brunsden, 2001; Fryirs & Brierley, 2009).

Fuller and Marden (2011) looked in detail at the delivery of sediment from a large gully complex to a headwater tributary of the Waipaoa River via its conveyance along a fan to understand the patterns and drivers of connectivity. Lateral erosion of the lower fan by the trunk stream during high flows produced a knickpoint that incised the lower portion of the fan, generating sediment to the stream system enhancing longitudinal connectivity. The behaviour of the lower fan was conditioned both by interaction with the Te Weraroa Stream (trimming the fan) and sediment supplied from up fan (Fuller & Marden, 2011). Sediment supply from the upper portion of the fan was in turn conditioned by cycles of repeated aggradation and incision in response to variable sediment supply rates from the contributing gully complex (lateral connectivity). During periods in which mass movement activity ceased or reduced, the upper fan, incised, which conveyed sediment down fan. The rapid nature of processes operating in this gully-fan system means that classic models of fanhead trenching and mid fan intersection points (e.g., Davies & Korup, 2007; Harvey, 1987) do not readily apply because Fuller and Marden (2011) observed trenching of the upper fan in a matter of months, and its reversal as quickly. Incision may be propagated through the entire length of the fan as well as being confined to either the fan head or distal portion of the fan (Taylor et al., 2018). The significance of this example for connectivity lies in its functioning as a microcosm of a catchment with strong lateral and longitudinal connectivity. The system itself has a very low structural resistance (cf. Brunsden, 2001) and is very sensitive to externally driven change. As such, it can be used to represent the behaviour of larger-scale catchments that share low resistance characteristics. Such catchments must be considered vulnerable to disturbance and very efficient conveyors of sediment, both laterally from slopes and longitudinally in response to propagation of waves of incision or aggradation longitudinally. In such highly connected systems, stream habitats are too unstable for invertebrates or periphyton to recolonise following spates or even survive, as much of the substrate is almost continually moving (cf. Parkyn et al., 2006).

4 THE IMPORTANCE OF SEDIMENT CONNECTIVITY TO ECOLOGICAL HEALTH

As we illustrated above, high sediment connectivity can result in unstable and rapidly changing riverine habitat that can limit the diversity, types, and abundance of plants and animals living in some streams and rivers, often long after a disturbance event. However, reduced or truncated sediment supply often occurs downstream of dams (Hart et al., 2002; Liermann, Nilsson, Robertson, & Ng, 2012; Stanley & Doyle, 2002; Tullos et al., 2016), and can similarly reduce habitat quality and consequently the organisms living in the downstream reaches.

Ecological communities are highly variable and dynamic biological systems intimately linked with a highly changeable habitat (although see Fryirs & Brierley, 2009), often stressed by both high and/or low flow events throughout the year, or in certain seasons depending on their geographic location (Death, 2008; Death, 2010; Ledger, Brown, Edwards, Milner, & Woodward, 2013; Vander Vorste, Corti, Sagouis, & Datry, 2016). Thus, although the stability of substrates will limit many taxa, river organisms in general have developed a variety of morphological, behavioural, and life history characteristics to survive during those high shear stress flows or hot low oxygen low flows (Death, 2008; Lake, 2000; Lake, 2011; Lancaster & Belyea, 1997; Lancaster & Downes, 2013). Despite these adaptive characteristics, floods and droughts can result in the loss of most, if not all, organisms from a river reach (Death, 1996; McEwan & Joy, 2013; Vander Vorste et al., 2016). However, in appropriate circumstances, recovery can occur within days or weeks, depending on the frequency and/or severity of the event (Datry, 2012; Datry, Moya, Zubieta, & Oberdorff, 2016; Death et al., 2015; Hauer & Habersack, 2009; Ilg et al., 2008; Lake, 2000; Vander Vorste et al., 2016).

Recovery from the defaunation after flood events in steepland rivers is rapid because many animals are able to escape the destructive forces of high shear stress and substrate movement where hydrologic connectivity is strong. The refugia for escape also often relies on spatially heterogeneous rivers, that in turn require some, but not too much, sediment supply. However, less geomorphologically dynamic systems may be more spatially homogeneous (cf. Fryirs & Brierley, 2009), with fauna adapted to lower energy system dynamics, which again emphasises the need to recognise the natural state or condition of connectivity in a given catchment to understand the extent of degradation, recovery, and/or rehabilitation. Many invertebrates are washed into or actively move into the floodplain or river margins during floods (e.g., Matthaei & Townsend, 2000; Rempel, Richardson, & Healey, 1999; Scrimgeour, Davidson, & Davidson, 1988). Beetle and hemipteran species in desert streams leave the stream for riparian areas in response to rain storms to avoid the coming flood (Lytle, 2008; Lytle, Bogan, & Finn, 2008; Lytle & White, 2007). Several species of caddisfly move from the tops of stones to lower shear stress positions behind or under boulders in response to increased turbidity and velocity (Cobb, Galloway, & Flannagan, 1992; Lake, 2000). A number of Galaxias species of fish in New Zealand move under the same large boulders for refuge during high flow events (McEwan & Joy, 2013). More stable areas of substrate and microform bed clusters, organised stacks of stone in the stream bed, also appear to accumulate species of invertebrate during floods (Biggs, Duncan, Francoeur, & Meyer, 1997; Hauer & Stanford, 1982; Matthaei, Arbuckle, & Townsend, 2000). Spatially heterogeneous river corridors will provide refugia (e.g., braided rivers with multiple secondary channels, or laterally mobile rivers with abandoned meander bends) from which recolonisation postflood can occur. It has also been postulated for some time that taxa seek refuge in the hyporheic zone during high flow events (Poff & Ward, 1990; Resh et al., 1988). However, there is limited evidence for the use of this escape route (Death, 2008) and its use is likely to depend on the geomorphic structure, substrate size, and permeability of sediments in a stream (Dole-Olivier, Marmonier, & Beffy, 1997; Marmonier & Chatelliers, 1991), that is, the strength of hydrologic connectivity (sensu Covino, 2017), modulated by sediment connectivity.

Connections to, and availability of, these refugia, which are products of hydrologic and sediment connectivity, will therefore determine the extent and subsequent recovery from flood and drought events (Death, 2008; Death et al., 2015; Lake, 2000; Vander Vorste et al., 2016). Rivers and streams with more heterogeneous stream beds composed of dead zones, debris dams, and microform bed clusters and/or those with good upstream, hyporheic, and floodplain connections will provide more refugia, and thus more colonists once high flows reside. Overwhelming the streambed with a large quantity of gravel, or blanketing it in fine sediment, where catchment erosion sources are strongly connected to the channel system, will compromise this heterogeneity, and thus stream health. Increasingly, connections between reaches and these refugia are also being severed by human modification of rivers. Floodplains have been drained or connections severed from the main channel by rock lining of channels. Smaller upstream reaches, which provide the bulk of recolonising invertebrates through drift (Brittain & Eikeland, 1988; Death, 2008; Muller, 1974), may be severed from downstream by drying, drought, channelisation, chemical discharges, piping, and culverts (Blakely, Harding, McIntosh, & Winterbourn, 2006; Tonkin, Stoll, Sundermann, & Haase, 2014). Blakely et al. (2006) found even short culverts underneath roads prevented the majority of adult caddisflies from flying upstream to colonise upstream reaches. Modification of geomorphology by river engineering may also remove backwaters, pools, and side channels that can also act as refugia. Finally, increasing armouring, embeddedness, and deposition of fine sediment removes access for many fish and invertebrates to the hyporheic zone, and refuges under logs and boulders (Ryan, 1991; Wood & Armitage, 1997). Fine sediment in particular, having a high degree of longitudinal connectivity moving quickly through the system (Hooke, 2003), is a major threat to stream health.

5 THE CHALLENGES OF HYDROLOGICAL CONNECTIVITY TO ECOLOGICAL HEALTH

As argued above, in some landscapes, increased sediment supply can degrade habitat for some riverine organisms, especially in higher energy gravelly rivers (bedload or mixed load dominated), or where bedrock reaches are overwhelmed with sediment (see e.g., Young, Olley, Prosser, & Warner, 2001). Where sediment supply is similar to that in prehuman times, the spatially heterogeneous riverscapes created and maintained will provide habitat for a diverse array of species both during base flows and/or as refugia during high flows. Furthermore, the branching hydrologically connected river network and its connected floodplains provide a riverscape over which many organisms play out sometimes complicated life cycles. The adults of most riverine invertebrates are terrestrial and require access to the appropriate riparian habitats to complete their life cycles (Lancaster & Downes, 2013). A considerable number of fish species migrate up and down river networks as part of often complex life cycles (Closs, Krkosek, & Olden, 2016; Helfman, 2007; McDowall, 1990). The New Zealand longfin eel (Anguilla dieffenbachii), for example, lives as an adult for 50–100 years in the upper reaches of rivers before swimming to sea somewhere near Tonga, spawning and dying. The larval eels migrate back to New Zealand and slowly move upstream transforming from glass eels, to elvers, to adults as they mature (McDowall, 1990). Most New Zealand galaxiid fish in contrast lay their eggs on the river banks where they live as adults, but hatching eggs are washed into estuaries by floods before young fish (whitebait) swim back upstream to their adult habitat (Charteris, Allibone, & Death, 2003; McDowall, 1990). Severing of hydrological connectivity by dams, perched culverts, malfunctioning fish passes and channelisation isolates upstream populations, removing recruitment and locally extirpating species (Joy & Death, 2001).

Hydrological connectivity for flood- and drought-refugia-seeking activity and diverse life-history patterns is critical for healthy functioning river ecosystems systems (Hermoso, Kennard, & Linke, 2012; Rodriguez-Iturbe, Muneepeerakul, Bertuzzo, Levin, & Rinaldo, 2009; Seymour & Altermatt, 2014). As a consequence, many restoration efforts are focused on returning upstream/downstream and channel/floodplain connectivity by widening channel form, installing fish passes, and even dam removal (Hart et al., 2002; Stanley & Doyle, 2002; Tullos et al., 2016). However, connectivity is being returned to river systems that are now considerably different in their species composition from what it was when that hydrological connectivity was severed because of the widespread introduction of invasive, often detrimental species. In New Zealand, introduced brown trout (Salmo trutta) have decimated populations of a number of native Galaxias species (Crowl, Townsend, & McIntosh, 1992; McIntosh et al., 2010; Townsend, 1996; Woodford, Cochrane, McHugh, & McIntosh, 2011). Healthy populations of several species now only occur where trout have been deliberately or accidentally excluded from the native fish habitat by severing of hydrological connectivity (Death et al., 2016). Thus, restoration of hydrological connectivity without consideration of the wider catchment or riverscape context may result in negative effects for river ecosystems from well-meaning restoration activities. Depending on the species and landscape position, good hydrological connectivity can be a bad or a good thing for river health.

We present the postulated response of invertebrates and fish to changes in hydrological and sediment connectivity in Figure 4. Invertebrate taxa richness is predominantly determined by sediment connectivity or substrate movement. However, taxa richness can recover from flood disturbance quickly if there is recovery by drift from an upstream source pool. In contrast, fish species richness (in faunas dominated by migratory species such as in New Zealand) will be primarily determined by the hydrological connectivity, particularly to downstream reaches where parts of their life history occur. They appear to be less affected by sediment connectivity as they are able to escape the effects of floods much more easily than the invertebrates. This outcome may differ in riverine fish faunas where migratory fish are not so common.

6 IMPLICATIONS FOR RIVER MANAGEMENT

The implications of highly connected catchments for ongoing stream management are complex. Many rivers are highly susceptible to efficient conveyance of coarse sediment from slopes to channels and downstream longitudinally, although the structural and transmission resistance of the catchments (cf. Brunsden, 2001) is such that longitudinal conveyance may be less efficient in higher frequency, low magnitude sediment transfer events. Limited longitudinal conveyance may build up a legacy of alluvium in low-order tributaries that will supply trunk streams for decades to come. Attempts to construct gravel reserves (willow-planted areas of active channel designed to trap coarse bedload), for example, at the range front of the Ruahines have been successful in mitigating flooding and aggradation downstream, which had adversely affected adjacent farmland (Mosley, 1978b). However, these reserves are approaching capacity and new mitigation approaches will be required, given the volumes of sediment being supplied by ongoing landsliding in these catchments and the efficient lateral and longitudinal sediment connectivity in this system (cf. Figure 2). The invertebrate communities in these streams are still, however, depauperate in diversity and abundance with only a few highly specialised species surviving (Schwendel, Joy, Death, & Fuller, 2011). Furthermore, the key to their survival depends on their continued connection with refuge habitats.

River management that functions at a catchment scale is most effective if it takes landscape and channel connectivity into account to interpret geomorphic and ecological recovery pathways (Brierley & Fryirs, 2009). Brierley, Reid, Fryirs, and Trahan (2010) go on to argue that geomorphic structure, function, and evolutionary trajectory of a river system are vital to understand if river management is to improve river condition. In some catchments, it is the disconnectivity within catchments that poses an issue for river rehabilitation and repair, because the sediments and animals required to repair degraded reaches are locked up within the catchment, for example, behind dams, bank protection, and levees. Such disconnectivity has been observed in Australia, for example, where limited sediment availability constrains the prospects for river recovery (Brooks & Brierley, 2004; Fryirs, Brierley, Preston, & Spencer, 2007; Fryirs & Brierley, 2000; Fryirs & Brierley, 2001; Fryirs & Brierley, 2016; Wethered et al., 2015) and in the USA, dams are being removed to restore connectivity for fish and invertebrates (Hart et al., 2002; Poff & Hart, 2002; Tullos et al., 2016). However, it must also be recognised that in some systems, the natural state is one of disconnectivity, and channel degradation has resulted where disconnected wetlands and ponds have been drained and artificially connected (Brierley et al., 1999; Fryirs & Brierley, 1999; Mould & Fryirs, 2017). In steepland New Zealand catchments, a super-abundance of sediment exists, which has been perceived as allowing for potentially rapid and effective geomorphological river recovery in some cases (Fryirs, Brierley, Preston, & Kasai, 2007). However, biological diversity has already been compromised as a result of strong lateral and longitudinal sediment connectivity in steep East Coast North Island catchments where deforestation has severely altered the natural disturbance regime. Here, clear-flowing boulder- and cobble-bedded rivers have been transformed into rapidly aggrading, high sediment yielding rivers with matrix-rich gravelly beds (Gomez & Livingston, 2012; Gomez, Rosser, Peacock, Hicks, & Palmer, 2001; Hamilton & Kelman, 1952; Page et al., 2007). The sediment conveyor is effectively in overdrive in these systems, and until the source of the problem, which is accelerated hillslope erosion in steep, soft-rock terrain, is properly addressed, the conveyor will continue to deliver sediment superabundantly via strong lateral and longitudinal sediment connectivity. The need here, paradoxically, is to shut down and limit sediment connectivity to allow these freshwater systems to recover, and recognise the available sediment stores in valley floors that will continue to be reworked for some considerable time to come given the system memory and legacy sediments of past events (Brierley, 2010; James, 2013; Wohl, 2015).

Interestingly, Jackson and Pringle (2010) also advocated for a reduction in hydrologic connectivity in human-modified landscapes. Given the quantum of sediment involved in the steepland New Zealand systems identified in this discussion, system recovery is likely to take multiple decades, if not centuries, and even millennia. Such timescales for recovery are actually very similar to those cited in low relief systems (Fryirs, Brierley, Preston, & Kasai, 2007). So what should we do? In the Waiapu catchment, a 100-year management plan was proposed in 2015, which will, if successful, begin to address the issues posed by erosion and strong connectivity in this system. In the Ruahines, in 2016, the New Zealand government announced plans for a predator-free New Zealand by 2050, if successful, this would limit the browsing and damage to forest canopy in these ranges, allowing forest to thrive again, but such forest takes some centuries to attain full maturity and provide full protection. Meanwhile, these catchments remain vulnerable to disturbance, as a result of their innate geomorphic sensitivity and their high connectivity characteristics. Nevertheless, recognising these connectivity characteristics and geomorphic structure of catchments remains essential to provide effective river management and begin to tackle river health, even if this is a (very) long-term objective. Furthermore, it is vital to understand the context of the natural process regime in a catchment for mitigation and management to be effective (Brierley & Fryirs, 2005).

Sediment and hydrological connectivity also remain essential components for management of river ecology, although it is often overlooked by river managers because of public pressure advocating water quality as the perceived most important management lever for healthy rivers (Death et al., 2015). But, as with geomorphological management in a human altered landscape, simply restoring historical connectivity pathways will not necessarily result in the desired outcomes. Too much or too little sediment supply is likely to be detrimental for river ecosystems. Similarly, restoring hydrological connectivity may be vital for restoring healthy ecological condition or it may result in the extirpation of isolated endangered species as a result of more widespread movement of invasive species. Not surprisingly, reengineering rivers to a healthy ecosystem condition is not a simple task; both the geomorphological and ecological components need to be considered at both the local and the catchment scale. An overly narrow focus on a single component is unlikely to meet with success (cf. Brierley et al., 2010).

7 CONCLUSIONS

The functioning of river ecosystems is intimately dependent on the geomorphic structure of a catchment, its sediment and hydrologic connectivity and linkages between habitats. The flows and transfers of sediment in a system must be assessed to effect rehabilitation to improve river health, however, in doing so, it must be remembered that hydrologic connectivity and transfer pathways between habitats are crucial for fish and invertebrates being able to survive high sediment flux and/or the natural sediment movement regime when it has been restored. Understanding lateral and longitudinal connectivity of sediment and ecological habitat plays an important role in this, but can act at different scales and directions for the ecological and geomorphological components. It is a complex and challenging task for ecologists and geomorphologists to understand the differing mechanisms and scales at which connectivity interacts. However, for truly effective restoration and management, it is important for ecologists and geomorphologists to understand each perspective. Attempts at ecological or geomorphological restoration in isolation of the other is doomed to failure. Healthy river ecology requires good habitat with the appropriate refuge connections, whereas simply considering the effects of sediment supply on a reach in isolation will ignore the dynamic nature of rivers and how the biology copes with such changes.

CONFLICT OF INTEREST

The authors declare no conflict of interest.