Hydrol Earth Syst Sci 23 2305ndash2319 2019httpsdoiorg105194hess-23-2305-2019copy Author(s) 2019 This work is distributed underthe Creative Commons Attribution 40 License

Integrating network topology metrics into studiesof catchment-level effects on river characteristicsEleanore L Heasley1 Nicholas J Clifford2 and James D A Millington1

1Department of Geography Kingrsquos College London London UK2School of Social Political and Geographical Sciences Loughborough University Leicestershire UK

Correspondence Eleanore L Heasley (eleanoreheasleykclacuk)

Received 22 February 2018 ndash Discussion started 19 March 2018Accepted 15 April 2019 ndash Published 14 May 2019

Abstract The spatial arrangement of the river network is afundamental characteristic of the catchment acting as a con-duit between catchment-level effects and reach morphologyand ecology Yet river network structure is often simplified toreflect an upstream-to-downstream gradient of river charac-teristics commonly represented by stream order The aim ofthis study is to quantify network topological structure usingtwo network density metrics ndash one that represents networkdensity over distance and the other over elevation ndash that caneasily be extracted from digital elevation models and so maybe applied to any catchment across the globe These met-rics should better account for the multi-dimensional nature ofthe catchment than stream order and be functionally applica-ble across geomorphological hydrological and ecological at-tributes of the catchment The functional utility of the metricsis assessed by appropriating monitoring data collected forregulatory compliance to explore patterns of river character-istics in relation to network topology This method is appliedto four comparatively low-energy anthropogenically modi-fied catchments in the UK using river characteristics derivedfrom Englandrsquos River Habitat Survey database The patternsin river characteristics explained by network density metricsare compared to stream order as a standard measure of topol-ogy The results indicate that the network density metrics of-fer a richer and functionally more relevant description of net-work topology than stream order highlighting differences inthe density and spatial arrangement of each catchmentrsquos in-ternal network structure Correlations between the networkdensity metrics and river characteristics show that habitatquality score consistently increases with network density inall catchments as hypothesized For other measures of rivercharacter ndash modification score flow-type speed and sediment

size ndash there are varying responses in different catchments tothe two network density metrics There are few significantcorrelations between stream order and the river characteris-tics highlighting the limitations of stream order in account-ing for network topology Overall the results suggest thatnetwork density metrics are more powerful measures whichconceptually and functionally provide an improved methodof accounting for the impacts of network topology on thefluvial system

1 Introduction

Rivers are integrators of many elements of their catchments(Dovers and Day 1988) Consequently integrated catchmentmanagement has long been seen as the gold standard forriver management and has been adopted in catchments acrossthe globe (Newson 2009) Research linking patterns of riverreach characteristics to catchment-level functioning is cur-rently focussed on characteristics of the terrestrial catchmentsuch as land cover geology and topography (eg Cohen etal 1998 Harvey et al 2008 Jusik et al 2015 Naura et al2016 Richards et al 1996 1997) Yet ldquohot-spotsrdquo of activ-ity within catchments are identified based on the hydrologicalconnectivity of the catchment (Newson 2010) a character-istic that is often neglected by catchment-level studies Thismissing component of the catchment is critical for true in-tegrated catchment management as the impacts of key man-agement features (eg water channel land ecology and hu-man activity) are transmitted throughout the river network(Downs et al 1991) By investigating the impacts of hydro-logical connectivity on river form and function our under-

Published by Copernicus Publications on behalf of the European Geosciences Union

2306 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

standing of catchment functioning can become more holisticand beneficial to catchment management

Effective catchment management rests not only on im-proving scientific understanding of river form and functionacross multiple scales but also on better integration betweenthe key disciplines of catchment studies geomorphology hy-drology and ecology This type of interdisciplinary approachis critical for understanding complex multi-casual relation-ships in river systems (Dollar et al 2007) However catch-ment connectivity is parameterized differently by differentdisciplines based on their interests The discipline of geo-morphology focusses on characterizing the morphometry ofthe catchment either using general variables which are con-tinuous across the landscape (eg elevation slope curvature)or specific variables which represent individual features suchas catchments (eg drainage density shape area) or streams(eg stream order stream length) (Evans and Minaacuter 2011)Hydrology focusses on how the catchment influences hydro-graph and flood peak timing and magnitude Methods such asthe Geomorphic Instantaneous Unit Hydrograph (Rodriguez-Iturbe and Valdes 1979) focus on predicting the travel timeof water reaching channels and travelling downstream basedthe morphology of the catchment drainage network andprecipitation Aquatic ecology takes a network-centric ap-proach utilizing dendritic ecological networks (Peterson etal 2013) This method aims to take a spatially continuousview of rivers (Fausch et al 2002) in order to appreciatethe influence of flow and location in the network on dis-crete sites chosen for ecological sampling Spatial statisticalstream network models based on the branching of the net-work (Ver Hoef and Peterson 2010) are shown to be moreaccurate than a standard Euclidean distance kriging modelyet only worthwhile if data sites are distributed across thenetwork and are spatially correlated (Peterson et al 2013)Alternate methods for exploring relationships between net-work structure and ecological functioning are also based onEuclidean distance along the network (Ver Hoef and Peter-son 2010)

Each discipline represents the elements of the catchmentcritical to their field focussing on describing catchmentform catchment flow responses and ecological responsesHowever the geomorphology hydrology and ecology of thecatchment are interconnected across spatial and temporaldimensions in the fluvial hydrosystem (Petts and Amoros1996) We argue that the overlap between disciplinary meth-ods can be utilized to create a metric to represent the catch-ment that is meaningful across all disciplines and offers in-creased potential for effective catchment management utiliz-ing a multi- or inter-disciplinary approach

This paper repurposes metrics that focus on the topologyof the river network for a novel application to assess thekey link between the catchment and reach-level function-ing The metrics represent network density variation withincatchments and have functional applications across the fieldsof geomorphology hydrology and ecology (Sect 11) The

impacts of internal network structure on patterns of rivercharacteristics within catchments are explored by utilizingdatasets that are collected for regulatory purposes with ar-eas of higher network density likely to support greater riverquality and diversity (Sect 12) The utility of the topologicalmetrics is compared against stream order a classic but over-simplified method of accounting for network topology Thetopology metrics are calculated for catchments with compar-atively low energy and that are influenced by anthropogenicmodification as much of the previous evidence for increasesin diversity in network-dense areas has been from highly ero-sive mountainous catchments

11 Quantifying the river network at different scalesand dimensions

River network structure or network topology is one wayto conceptualize the integrated transport of water sedimentand nutrients from the upstream catchment to downstreamreaches The spatial arrangement of links (river channels)and nodes (confluences) concentrates the catchment effectin some areas of the landscape making network topology auseful archetype of catchment functioning (Gupta and Mesa1988)

Drainage density (the total length of the network dividedby catchment area) is most commonly used to compare theamount of the catchment covered by river channels but thisfails to quantify spatial variation within catchments and sooffers only a partial means for functionally assessing catch-ment similarities and differences To represent within catch-ment network structure stream order (Strahler 1957) order-ing river links along an upstream-to-downstream gradientbased on their upstream connectivity (Fig 1a) is also com-monly used However stream order does not account forthe spatial arrangement of links only their relative positionin the distance dimension of the catchment Conceptualiz-ing the catchment in this one dimension leads also to over-simplification for example first-order streams are thoughtof as upland headwater streams furthest away from the rivermouth yet often first-order streams are tributaries to high-order lowland streams with different characteristics than up-land streams

This paper argues that the spatial arrangement of linkswithin catchments must be considered across the dis-tance and height of the catchment to obtain a full three-dimensional appreciation of catchment effect through net-work topology Two methods from the field of hydrologyndash network width function (NWF Kirkby 1976) and linkconcentration function (LCF Gupta et al 1986) ndash offer in-creased dimensionality by accounting for the width of thenetwork (ie the number of links) at successive distances forthe NWF or elevations for the LCF

These methods quantify network topology within catch-ments with functional significance NWF has hydrologicalapplication representing the travel time of water through the

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2307

Figure 1 Topological metrics explored in this paper and the dimensions of the network they represent (a) Strahler stream ordering repre-senting only the distance dimension of the network (b) Distance network density representing the width dimension of the network at eachdistance interval (inspired by the network width function Kirkby 1976) (c) Elevation network density representing the width dimension ofthe network at each elevation interval (inspired by the link concentration function Gupta et al 1986)

network to predict the timing and magnitude of unit hydro-graphs and flood peaks (Rodriguez-Iturbe and Valdes 1979)with a more functionally specific method than the traditionalstream-ordering approach (Gupta and Waymire 1983) Ex-tending applications beyond the field of hydrology the tim-ing and magnitude of the hydrograph have direct influenceon instream ecology controlling the formation maintenanceand disturbance of physical habitats (Bunn and Arthing-ton 2002) Longitudinal connectivity of water and sedimentthrough the network is also one of the multiple dimensions ofthe fluvial hydrosystems approach to catchment ecohydrol-ogy (Petts and Amoros 1996) influencing the capacity forlateral and vertical connectivity and the development of theriparian corridor over time LCF is less frequently appliedin hydrograph prediction than NWF However it may betterreflect catchment hydrology by incorporating the effect ofgradient on the travel time of water rather than the constanttravel time suggested by NWF (Gupta et al 1986) Thesemetrics also have morphometric significance reflecting theinternal shape of the network by segmenting catchments intointervals to represent how network density changes withincatchments (Stepinski and Stepinski 2005)

This paper repurposes these metrics to reflect network den-sity as a feature of the catchment rather than as a method forhydrograph prediction Distance network density (modelledon the NWF) (Fig 1b) and elevation network density (mod-elled on the LCF) (Fig 1c) allow for the comparison andquantification of network topological variation both withinand between catchment with improved interdisciplinary andfunctional applicability than the stream-ordering approach

12 Network topology effects on river reach functioning

The topological structure of the river network configures theriver ecosystem (Bravard and Gilvear 1996) by impactingfunctioning at the reach and sub-reach scales The distancedimension of the catchment often represented by streamorder (Fig 1a) reflects upstream-to-downstream gradual

changes exhibited by many in-channel features and speciesIt forms the basis of classic geomorphic models highlight-ing the zones of sediment supply in the headwaters sedi-ment transfer in the mid reaches and sediment storage nearthe outlet (Schumm 1977) It is also a key component inclassic ecological models such as the River Continuum Con-cept (Vannote et al 1980) which describes gradual changesin grain size channel width invertebrates fish species andenergy sources along the gradient Both models suggest thatdiversity in channel morphology and biota may be highest inthe mid reaches as channels transition from erosional to de-positional environments The River Continuum Concept is apopular model but is critiqued for being too simplistic andfor neglecting discontinuity introduced by changes at conflu-ences (Perry and Schaeffer 1987 Rice et al 2001) Conflu-ences as nodes in the network are associated with changesin hydrological geomorphological (Best 1987 Church andKellerhals 1978) and ecological (Kiffney et al 2006 Riceet al 2001) conditions and have therefore been termed bio-diversity ldquohotspotsrdquo (Benda et al 2004b)

Confluence impacts extend throughout the river networkwith increased channel heterogeneity in the tributary andmain channel upstream and downstream of the confluence(Rice 2017) This has led to several theories relating to theimpact of numerous confluences in the context of the widernetwork The Link Discontinuity Concept shows the impactof confluences throughout the length on the main channelcreating step changes in sediment size before fining contin-ues downstream towards the next confluence along a ldquosed-imentary linkrdquo (Rice et al 2001) The Network DynamicsHypothesis posits that catchments with higher drainage den-sity and thus more confluences will have greater channelheterogeneity (Benda et al 2004b) despite drainage densityfailing to be a useful catchment characteristic for predictinglocal habitat features (Davies et al 2000) The hypothesisalso suggests that catchment shape will influence the impactof confluences as more compact catchments will have moresimilarly sized tributaries (Benda et al 2004b) which have

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2308 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

the greatest impact on channel morphology (Benda et al2004a) the greatest flow diversity (Schindfessel et al 2015)and the greatest fish community diversity (Osborne and Wi-ley 1992) In contrast others have found that tributaries thatdiffer most in size have the greatest impact For exampleJones and Schmidt (2016) suggest that high densities of smalltributaries flowing into a large channel cause small cumula-tive changes and Milesi and Melorsquos (2013) study concludedthat small tributaries flowing into large channels in the pe-ripheral regions of the catchment have the greatest impact onmacroinvertebrate assemblages

Interestingly there is little evidence of anthropogenic im-pacts at confluences in the literature but as confluences areproposed concentration points of catchment effects it seemslikely that they may be focal points for anthropogenic im-pacts For example flood events may occur downstream oflarge confluences as flood peaks converge creating the needfor flood defence measures (Depettris et al 2000) and scourand erosion at confluence junctions (Best 1986) increase theneed for bed and bank protection Also sediment size at con-fluences is shown to increase in many studies (Church andKellerhals 1978 Knighton 1980) but in tributaries whosewatersheds are dominated by agricultural land uses fine sed-iments may become dominant at confluences potentially al-tering river functioning (Owens et al 2005)

Many previous studies citing the impact of the networkspecifically confluences on river characteristics were con-ducted in highly erosive relatively natural environments(Network Dynamics Hypothesis Benda et al 2004b LinkDiscontinuity Concept Rice et al 2001) Therefore it willbe interesting to assess the impact of network structure onriver characteristics in catchments in England a landscapethat has undergone modification impacting catchment func-tioning for centuries (Macklin and Lewin 2003)

2 Methods

21 Study sites

Four catchments are selected for testing the impact of net-work topology on river characteristics in England The catch-ments are from the Demonstration Test Catchment pro-gramme (Fig 2) which are representative of 80 of soil andrainfall combinations in the UK (McGonigle et al 2014)This demonstrates the potential use of topological metricsfor catchments with varying geologies and land uses

The Avon and Wensum catchments have similar charac-teristics both being dominated by chalk geology with loweraverage annual rainfall and a high percentage of arable farm-ing land cover In comparison the Eden and Tamar are dom-inated by less permeable bedrock with higher average annualrainfall and a high percentage of grassland land covers Interms of their morphometry the Avon and Wensum both havean elongated shape and low drainage density The Wensum

has the lowest relief with a maximum elevation of 95 m TheTamar has the smallest catchment area (928 km2) and is themost circular The Eden is the largest catchment (2295 km2)and has the highest maximum elevation (246 m)

22 Network topology metrics

Network topology metrics were calculated for each catch-ment using the 1 50000 river network map derived fromboth a digital terrain model (DTM) and Ordnance Sur-vey data (Moore et al 1994) Anabranches and incorrectlydigitized links in the network are identified using RivEX(Hornby 2010) and removed Removing anabranches wasnecessary as the topological metrics were designed for den-dritic networks so multi-thread channels either naturally oc-curring or artificial ditches would distort the calculationsThis resulted in a total of 448 2812 1516 and 532 links inthe Avon Eden Tamar and Wensum respectively

Elevation data were extracted from the Integrated Hydro-logical DTM (Morris and Flavin 1994) a 50times50 m griddedelevation raster with a 10 cm vertical resolution Average el-evation of each link and the distance from each link to thenetwork outlet were extracted using RivEX (Hornby 2010)

To extract a measure of network density that varies spa-tially within the catchment each network is divided into20 intervals each of which represents 5 of the total dis-tance or highest elevation in the network (Fig 2) The net-work is divided in this manner based on the methods of theNWF and LCF which have functional application to hydro-graph prediction Twenty intervals provide a relatively coarsesampling of the network compared to the 100 intervals de-scribed by Stepinski and Stepinski (2005) when they adapteda morphometric variable circularity ratio to represent inter-nal catchment elongation Here a total of 20 intervals is cho-sen so that most intervals contain links for the density cal-culation whilst ensuring the spatial distribution of networkdensity within the catchment is characterized

Distance network density was calculated followingEq (1)

Distance network density=

[n(d0) n(di) n(dN)

](dNtimes 005)

(1)

where the number of links (n( )) within each 5 distanceinterval (di) from the outlet (d0) to the maximum distance inthe network (dN) is normalized by the width of the interval(dNtimes 005)

Elevation network density was calculated followingEq (2)

Elevation network density=

[n(z0) n(zi) n(zN)

](zNtimes 005)

(2)

where the number of links (n( )) within each 5 elevationinterval (zi) from the outlet (z0) to the maximum height ofthe network (zN) is normalized by the width of the inter-val (zNtimes 005) Normalization allows network densities to

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2309

Figure 2 Distance and elevation intervals for each Demonstration Test Catchment Avon (A) Eden (E) Tamar (T) and Wensum (W)(a) Percentage distance intervals used to calculate distance network density (b) Percentage elevation intervals used to calculate elevationnetwork density Map of catchment locations in England in the bottom-right corner

be compared between catchments controlling for differencesin size and elevation as well as within catchments

To assess the utility of the multi-dimensional topologymetrics in accounting for the spatial structure of the networkthe metrics are compared to the one-dimensional Strahlerstream-order metric extracted from the river network datasetusing RivEX (Hornby 2010)

23 River characteristics

The impact of network topology on channel functioning isexplored using a broad-scale approach ie adapting data col-lected for regulatory compliance to answer scientific ques-tions Adapting such datasets to scientific enquiry allowsanalysis to be conducted in many catchments across a widespatial extent There are many habitat monitoring methodsacross the globe with 121 survey methods recorded in over

26 different countries (Belletti et al 2015) so this methodmay be adapted to other countries

This study utilizes the River Habitat Survey (RHS Ravenet al 1996) a regulatory dataset collected by Englandrsquos En-vironment Agency which is used to reflect the river reachcharacteristics in each catchment This dataset has been usedto identify catchment effects on river characteristics in broad-scale studies by previous research (eg Harvey et al 2008Naura et al 2016 Vaughan et al 2013) but none have in-cluded the effects of network topology

Since 1994 over 24 000 sites have been sampled incatchments across England and Wales including the Avon(n= 418) Eden (n= 398) Tamar (n= 189) and Wensum(n=315) Surveys were conducted at random sites withineach 10 km2 of England and Wales to ensure geographic cov-erage however this produces sampling bias as streams inhigh-density areas will be under-represented in the datasetwhich is acknowledged in this study and discussed below

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2310 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

Table 1 RHS variables calculated from RHS observations

RHS variable Calculation from RHS observations Units

Habitat Quality A score indicating the degree of naturalness and diversity of the HQA scaleAssessment riparian zone based on observations in the reach of flow types(HQA) substrate channel and bank features riparian vegetation etc

Habitat A score indicating the degree of artificial modification of the channel HMS scaleModification based on observations in the reach of reinforcements re-sectioningScore (HMS) embankments weed cutting realignment culverts dams weirs etc

Sediment size =(minus8middotBOminus7middotCOminus35middotGPminus15middotSA+15middotSI+9middotCL)

(BO+CO+GP+SA+SI+CL) ApproxBO (boulder) CO (cobble) GP (gravelndashpebble) SA (sand) SI (silt) φ unitsand CL (clay) represent the number of spot checks allocated to eachsediment size class

Flow-type speed =(0middotDR+1middotNP+2middotUP+3middotSM+4middotRP+5middotUW+6middotBW+7middotCF+8middotCH+9middotFF)

(DR+NP+UP+SM+RP+UW+BW+CF+CH+FF) Flow typeDR (dry) NP (no perceptible flow) UP (upwelling) SM (smooth) speed scaleRP (rippled) UW (unbroken wave) BW (broken wave) CF (chaoticflow) CH (chute) FF (free-fall) represent the number of spot checksallocated to each flow speed class

At each site over 100 features are recorded along a 500 mreach with 10 ldquospot-checkrdquo surveys conducted every 50 mand a ldquosweep-uprdquo survey conducted across the whole reach(see Raven et al 1996 for details) Variables of interest thatare hypothesized to be impacted by network structure can becalculated from the RHS observations (Table 1)

The Habitat Quality Assessment (HQA) and Habitat Mod-ification Score (HMS) variables are both amalgamations ofRHS observations with individual features given a score de-rived by expert opinion (see Raven et al 1998 for more de-tails) The scoring systems are subjective but HQA and HMSprovide overviews of the channel condition that are widelyapplied for regulatory compliance The scores are thereforeincluded in this study to reflect how they may be impactedby network topology

The remaining RHS variables are calculated directly fromRHS observations and so are more objective Sediment sizeis calculated as a reach average of spot-check observationsusing the same method as previous studies (Davenport et al2004 Emery et al 2004 Harvey et al 2008) Flow-typespeed was calculated in the same manner as sediment sizeusing values of flow which represent an approximate flowvelocity gradient defined in Davenport et al (2004) Thesevariables were chosen to reflect dominant geomorphic pro-cesses occurring in each reach and due to the prominence ofsediment size and flow type in defining physical habitats forinstream biota (Rowntree and Wadeson 1996) The variablesare likely to be impacted by the density of the river networkas they have been shown to be impacted by individual con-fluences For example channels are shown to become moregeomorphologically heterogeneous (Benda et al 2004a) andsubstrate size has been shown to coarsen at confluences (Riceet al 2001) Surface flow types are also likely to become

more diverse at confluences as the convergence of channelscreates a number of different flow environments (Best 1985)that result in different water-surface topographies (Biron etal 2002)

It must be noted that the RHS dataset was collected forregulatory compliance and was not directly intended for sci-entific enquiry Therefore there is a limitation in the amountof detail that can be extracted about physical processes as theobservations recorded are an average across a 500 m reachDespite this the wide spatial coverage of the dataset makesit a powerful tool allowing analysis to be conducted acrossmultiple catchments with differing characteristics

For each distance and elevation interval created by the net-work topology metrics descriptive statistics (mean median90th and 10th percentiles) of each RHS variable are calcu-lated Despite the RHS sampling strategy (Jeffers 1998b) bi-asing site selection towards less dense areas of the networkmost distance and elevation intervals contained RHS sites(with only some low-density intervals not containing RHSsites) This method is designed to account for natural varia-tion and modification at individual RHS sites in order to as-sess broad patterns of reach characteristics at the catchmentlevel

24 Statistical analysis

Analysis is conducted with all catchments combined into asingle population to identify overall trends across all catch-ments a method used in previous broad-scale studies Theanalysis is also split into individual catchments to identifyhow the relationship between network topology metrics andriver reach characteristics differed between catchments

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2311

Kendall correlation

Correlation tests are used to determine the strength and di-rection of the association between the descriptive statisticsof the RHS variables and distance network density eleva-tion network density and stream order to ascertain how reachcharacteristics respond to network topology Kendallrsquos cor-relation method was used as the variables have non-normaldistributions a small sample size and tied data values (Helseland Hirsch 2002) The effect size of Kendallrsquos τ is lowerthan other correlation methods with strong correlations oc-curring with τ values greater than 07 (Helsel and Hirsch2002)

As multiple correlations are conducted false discoveryrate (Benjamini and Hochberg 1995) corrections were ap-plied to the p-values produced from the Kendall correlationsto reduce the risk of type I error The false discovery ratemethod has been found to be more powerful than other pro-cedures for controlling for multiple tests (Glickman et al2014)

3 Results

31 Differences in network topology metrics betweencatchments

The topological metrics developed in this study show the in-ternal structure of the network for each catchment The sep-aration of the catchments into distance and elevation inter-vals emphasizes different features of the catchment The dis-tance intervals (Fig 2a) are arranged longitudinally withinthe catchment highlighting sub-basins within each catch-ment The elevation intervals (Fig 2b) have a radial arrange-ment centering around the incised main channel of eachcatchment

Distance network density is higher in the Eden (284plusmn103) and Tamar (441plusmn 219) compared to the Avon (47plusmn19) and Wensum (68plusmn 07) interesting as the Tamar is thesmallest catchment by area The shape of the distance net-work density function reflects the internal shape of the net-work (Fig 3a) For example the Tamar has a peaked den-sity distribution reflecting the circular shape of the catch-ment such that the majority of links are at 55 ndash65 dis-tance from the catchment outlet The Avon and Eden reflectsimilar internal network structures both exhibiting a bimodaldensity distribution despite the differences in the number oflinks in the catchments The density distribution of the Wen-sum has a more complex internal distribution of links withmultiple peaks in density

Elevation network density (Fig 3b) shows similar densitydistribution shapes to distance network density with a uni-modal distribution for the Tamar and multi-modal distribu-tions in the other catchments In contrast to distance networkdensity elevation network density shows the highest peaks in

density in the Tamar (103plusmn 50) and Wensum (101plusmn 43)despite the Wensum having the lowest network elevation andhas lower values in the Avon (34plusmn10) and Eden (58plusmn25)The peak densities in the Wensum occur at similar positionsin the elevation and distance intervals whereas the peaks inthe other catchments are negatively skewed showing the net-work density is highest at lowndashmid elevations

Nearly half of the links in each catchment are classified asfirst-order streams and the number of links declines exponen-tially towards the highest orders in all four catchments Thereare weak correlations (τ =minus003 to 017) between the threenetwork topology metrics distance network density eleva-tion network density and stream order This suggests that themetrics are independent and reflect different aspects of rivernetwork topology

32 River characteristic relationships with networktopology metrics

RHS sites in the Avon and Wensum have similar river char-acteristics Both have lower habitat quality high modifica-tion fine sediment and slower flow types than the Eden andTamar When all catchments are combined there are signif-icant (p lt 005 after p-value correction) correlations withmost descriptive statistics for each RHS variable and dis-tance network density (Fig 4) There are consistently pos-itive correlations with HQA and flow-type speed and neg-ative correlations with HMS and sediment size There arefewer and weaker significant correlations with elevation net-work density (Fig 4) The only significant correlations withstream order are with HMS which shows a negative correla-tion (Fig 4)

There are numerous significant correlations between thenetwork topology metrics and RHS variables for individualcatchments many of which were also shown to be signifi-cant after the correction to the p-value The results show thatcatchments have different responses to the network topol-ogy metrics of distance network density and elevation net-work density Distance network density only has significantcorrelations with the regulatory scoring variables (HQA andHMS) in the Eden and Tamar (Fig 4) Elevation networkdensity however has a wider array of significant correlationswith the scoring variables particularly HQA which showssubtle peaks and troughs reflecting the distribution of bothnetwork density metrics (Fig 3a and b) HMS shows mostlynegative correlations mainly with elevation network densityapart from the Eden which has significant positive correla-tion across all HMS descriptive statistics (Fig 4) Visually10th percentile HMS is most variable to network density withpeaks and troughs responding to the network density distri-butions (Fig 3a and b)

For individual RHS features the response to networktopology varies between catchments (Fig 4) The Avon hasnegative correlations between flow-type speed and distancenetwork density with an evident drop in 10th percentile

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2312 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

Figure 3 Network topology metrics (a) distance network density and (b) elevation network density Descriptive statistics of each RHSvariable over (a) distance and (b) elevation for each catchment with smooth loess lines to indicate trend Network topology metrics arenormalized between 0 and 1 and HMS score is logarithmically transformed for display purposes

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2313

Figure 4 Summary of correlations between distance network density elevation network density stream order and RHS variables for allcatchments combined and each individual catchment Significance of correlation is indicated by point size with the largest points significantpost p-value correction Effect size (Kendallrsquos τ ) is indicated by colour No correlation was possible between stream order and 10th percentilesediment size in the Avon due to no variation in the RHS variable

flow-type speed associated with peaks in network density(Fig 3a) The Eden and Tamar however have positive cor-relations with mean and 90th percentile flow-type speed fordistance network density but negative correlations with me-dian and 10th percentile elevation network density The Wen-sum shows positive correlations between flow-type speed andelevation network density Sediment size has a consistent re-sponse to distance network density with the Eden and Wen-sum showing negative correlations with the sediment size(Fig 4) For elevation network density the Avon shows neg-ative correlations with sediment size whereas the Tamar andWensum show positive correlations (Fig 4)

There were few significant correlations between stream or-der and the RHS variables in individual catchments (Fig 4)The only significant correlation after p-value correction iswith 90th percentile HMS in the Wensum which shows astrong negative relationship

4 Discussion

41 A new approach to utilizing network topology incatchment-level analysis

Network density metrics represent an alternative approach toaccount for network topology in catchment-level studies op-timizing the width dimension of the network (or the num-

ber of links in the network) as opposed to the commonplacestream-order metric which only reflects the longitudinal po-sition of links in a network (Fig 1) This study demonstratesthat two topology metrics can be calculated simply from aDTM with GIS and using a broad-scale analysis of river at-tributes can be used to investigate the functional processeswithin catchments

While the two network density metrics have similar forms(ie forms are consistently unimodal or multi-modal) thespatial configuration of the distance and elevation intervalsused in the calculation of network density varies and mayimpact the effectiveness of each topological metric Distancenetwork density separates the catchment into intervals basedon distance which spread upstream from the outlet (Fig 2a)reflecting natural sub-basins within the fractal structure of thecatchment (Lashermes and Foufoula-Georgiou 2007) Thisdiffers from elevation network density which separates thecatchment into intervals based on elevation which radiate outfrom the main channel of the network (Fig 2b) The config-uration means that distance intervals contain streams that arein closer proximity to one another rather than the more distalconfiguration created by the elevation intervals suggesting adegree of spatial dependency in river functioning This hasbeen highlighted in previous studies where spatial networkstructure has a stronger influence on some in-channel pro-cesses than predictor variables such as elevation (Steel et al

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2314 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

2016) However elevation intervals contain RHS sites thatalthough they may be distal may have similar properties aselevation has been strongly related to RHS variables includ-ing flow type substrate etc in a number of studies (Jeffers1998a Naura et al 2016 Vaughan et al 2013)

42 Impacts of network topology on rivercharacteristics

River characteristics are assessed using the RHS dataset Theobservations made by the RHS dataset (Table 1) cannot of-fer the level of detail regarding geomorphological processthat river classifications which consider multiple scales canoffer (eg Brierley and Fryirs 2000 Gurnell et al 2016)While process-based classifications are preferable broad-scale monitoring datasets such as the River Habitat Surveymay still be useful when combined with map-derived datato explore controls on river characteristics (Harvey et al2008 Naura et al 2016 Vaughan et al 2013) Howeverthere are biases in RHS data collection an inherent limitationwhen using existing datasets (Vaughan and Ormerod 2010)specifically the standardized survey length of 500 m reachthat will capture differing amounts of natural variability de-pending on the size of the river While this must be notedthere are few significant correlations between river charac-teristics identified with stream order (Fig 4) which suggeststhat channel size is not influencing the RHS variables to agreat degree in these catchments

In this study it is anticipated that sites in high networkdensity areas will have higher levels of habitat diversity asindicated by previous studies of confluences and networks(Benda et al 2004a Best 1985 Rice 2017) in turn in-creasing mean sediment size and flow-type speed comparedto sites in low-density areas The results of the correlationsbetween distance network density and river characteristicswhen all catchments are combined support this hypothesiswith greater HQA flow-type speed and coarser sedimentsizes observed in areas with high distance network density(Fig 4)

For individual catchments elevation network density in-duces a stronger positive HQA response across all catch-ments than distance network density (Fig 4) This supportsthe evidence that individual confluences (Rice et al 2006)and high densities of confluences increase physical hetero-geneity within the river network (Benda et al 2004b Rice2017) However flow-type speed and sediment size responddifferently to network density in individual catchments

Slower flow types are observed in high network densityareas of the Avon and Tamar whereas faster flow types areobserved in high-elevation network density areas of the Wen-sum Individual confluences are shown to create numeroushigh- and low-speed flow environments (Best 1987) thatmay be observed in surface water topography (Biron et al2002) It was expected that the introduction of the additionalflow types by a high density of confluences in relatively low-

energy rivers would increase mean reach flow-type speedhowever the correlation analysis (Fig 4) suggests that insome catchments mean flow-type speed is reduced

Sediment size response also shows variation betweencatchments Sediment size is coarser in network-dense ar-eas of the Avon and Eden as expected but is finer in boththe Tamar and Wensum (Fig 4) The evidence from high-energy rivers shows that sediment becomes coarser down-stream and finer upstream of certain confluences (Benda etal 2004a Rice 1998) and in this case high numbers of con-fluences were expected to increase mean sediment size of thereach However the impact on sediment size is dependent onthe sediment calibre of the incoming tributary being higherthan the main channel with enough energy to transport thecoarse sediment to the confluence for numerous tributariesin an area The Tamar and Wensum have different rangesof sediment sizes with the Tamar having coarser sedimentthan the Wensum on average (Fig 3) This implies that trib-utaries in the Tamar may be energy limited not transportingcoarse sediments to confluences and the Wensum may suf-fer from inputs of fine sediments from the high percentageof arable land that concentrates in network-dense areas Thishas before been observed in a low-energy modified catch-ment where anthropogenic modifications in tributaries re-duced coarse sediment and flow capacity causing either lim-ited confluence impact or localized sediment fining (Singer2008)

Others have related the capacity of confluences to alterreach features to the morphometry of catchments with largerand more circular catchments containing a higher percentageof confluences that have a significant impact (Rice 2017)Based on this theory the Eden and Tamar are likely to havethe greatest impact as they are the most circular and steepestof the four catchments (although the Tamar is the smallestby area) Yet there is no clear pattern indicating that thesecatchments respond differently than the Avon and Wensum(Fig 4) with catchments responding differently to differ-ent variables This perhaps suggests that rather than networkdensity having a directional impact on factors such as flow-type speed and sediment size it has an impact on overall het-erogeneity at the reach level as suggested by previous stud-ies and that specific directional change occurs at the sub-reach level

An increase in channel modification is also hypothesizeddue to the increase in flood peak downstream from conflu-ences (Depettris et al 2000) and the scour and erosion as-sociated with confluence junctions (Best 1986) potentiallyincreasing the need for bed and bank protection The corre-lations between distance network density and HMS when allcatchments are combined undermine this hypothesis show-ing less channel modification where distance network den-sity is higher (Fig 4) There were few significant correlationswith individual catchments but HMS in the Eden was consis-tently observed to be higher in network-dense areas (Fig 4)This may be due to the Edenrsquos high elevation and steep to-

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2315

pography inducing a higher energy environment where scourand erosion processes in areas with high numbers of conflu-ences would be more likely to be present

Differences in RHS variable responses also differed be-tween the descriptive statistics considered Often the ex-tremes 90th and 10th percentiles showed more significantand stronger correlations than the mean or median (Fig 4)This may reflect findings from previous studies which sug-gest that not all confluences cause reach-scale changes (Rice1998) that perhaps the changes to river character induced bycertain confluences only influence certain reaches whereasothers are left unaffected creating less pronounced responsesin the mean and median of variables External factors mayalso influence this trend for example 10th percentile HMSvisually responds dramatically to network density metrics(Fig 3a and b) compared to the other descriptive statisticsThis suggests that the most natural sites (ie with the low-est HMS score) are responding differently to network den-sity with the most natural sites having less modificationin network-dense areas whereas less natural sites becomemore modified in network-dense areas This reflects the HQAscore results which visually (Fig 3a and b) and statistically(Fig 4) vary with distance network density except for the10th percentile These sites with the lowest habitat qualityand naturalness are likely influenced by anthropogenic fac-tors that are independent of network density reducing habitatquality scores at impacted sites

The two network density metrics have differing impacts onriver characteristics While distance network density showsconsistently significant correlations when all catchments arecombined individual catchments respond more frequentlyand more strongly to elevation network density (Fig 4)This may be because there is a dramatic split in distancenetwork density values between the more upland drainage-dense catchments Eden and Tamar than the lowland chalklow-drainage catchments Avon and Wensum The combinedcorrelation will therefore in part reflect the difference be-tween the catchments which have different ranges of rivercharacteristics (Fig 3) This is not the case for elevation net-work density which has higher density values in the Tamarand Wensum than the Avon and Eden so therefore the char-acteristics of the catchments will have less bearing on thecombined correlation However there are patterns identifiedwith distance network density in individual catchments thatare not present with elevation network density increasedflow-type speed and sediment size in the Eden and reducedflow-type speed in the Avon with network density (Fig 4)which show its usefulness

The results suggest that the distance network density andelevation network density metrics quantify different dimen-sions of network topology which are shown to exhibit func-tionally meaningful patterns for river reach characteristicsbased on the correlation results Perhaps within catchmentselevation network density provides the more powerful met-ric for individual catchments but distance network density

better accounts for the drainage density of the catchmentsallowing it to be applied across multiple catchments

43 Comparison of stream order to network densitymetrics

The classic method of accounting for network topologystream order is critiqued for failing to represent disconti-nuities in the network and simplifying the network into agradient The number of links in different stream orders isconsistent across all catchments not reflecting the internalstructure of the network or the variety between catchmentsthat is achieved by the distance network density and eleva-tion network density metrics The analysis of the two net-work density metrics presented in this paper shows that dis-tances from source and elevation are not mutually exclusive(Fig 2a and b) contrary to the stream-order metric whichrepresents streams as upstream to downstream or upland tolowland

Stream order has few significant correlations with manyof the river characteristics considered in this study Negativecorrelations with HMS in the Wensum were statistically sig-nificant post p-value correction likely driving the significantrelationship with all catchments combined for this variable(Fig 4) This suggests that modification is greater upstreamin the Wensum contrary to ideas that downstream reachesmay show greater anthropogenic modification Intense agri-cultural land use in the upper reaches of the Wensum islikely to be the cause of the high HMS scores upstreamThe lack of significant correlations suggests that stream or-der and therefore an upstream-to-downstream gradient arenot the predominant pattern in river characteristics despitethe description of such a gradient by geomorphic (Schumm1977) and ecological frameworks (Vannote et al 1980)This is surprising as distance and elevation which both re-flect the upstream-to-downstream gradient have proven tobe important factors in previous studies explaining patternsof RHS features at a national level (Jeffers 1998a Vaughanet al 2013) This implies that upstream-to-downstream gra-dient may not sufficiently reflect patterns of river charac-teristics through the river network within individual catch-ments Others have also found that stream order has weakand inconsistent relationships with biodiversity patterns inriver systems arguing that the topological measure has no di-rect mechanistic control on biodiversity (Vander Vorste et al2017) Instead this study finds that the network density met-rics are a more powerful metric which conceptually providean improved method of accounting for the impacts of net-work topology on the fluvial system exhibiting relationshipswith river characteristics particularly habitat quality score(Fig 4)

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2316 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

44 Applicability of network topology metrics todifferent environments

Much of the seminal work on network and confluence im-pacts (eg the Network Dynamics Hypothesis Benda et al2004b and Link Discontinuity Concept Rice et al 2001) isconducted in natural highly erosive catchments with first-hand empirical measurements However in an age whenrivers and their catchments are increasingly altered by an-thropogenic modification (Meybeck 2003) contemporarystudies must not only aim to expand knowledge but also findmethods of transferring knowledge to many increasingly al-tered catchments (Clifford 2002)

The catchments selected by this study are more greatlymodified and although they reflect a range of fluvial envi-ronments in England are lower-energy catchments than theseminal studies Benda et al (2004a) suggest that confluenceeffects in less active landscapes would be subdued comparedto highly erosive landscapes but the evidence presented heredemonstrates the utility of evaluating network topologicalstructure in studies on catchment-level effects in any typeof fluvial environment including those with low-energy andwidespread anthropogenic changes The response of someriver characteristics varied between catchments observationsof flow-type speed sediment size and modifications showeddifferent responses to network density in different catch-ments This suggests that the functional effect of these topo-logical metrics is catchment dependent and likely is influ-enced by external catchment characteristics such as land usenot considered in this study although impact did not appearto vary with catchment topography or circularity as has beenshown in prior studies (Benda et al 2004a Rice 2017) Thisshould be explored further in future research to enable rec-ommendations to be made regarding where and how reachesmay respond to network density The response of habitatquality score was however consistent across catchments andbetween metrics showing that habitat quality is greater inareas with high network density (Fig 4) as hypothesizedby the Network Dynamics Hypothesis (Benda et al 2004b)and demonstrated by studies on individual catchments (Rice2017 Rice et al 2006)

The methods presented in this paper are designed to be im-plemented in any catchment with a dendritic network struc-ture The topology metrics can easily be calculated from anydendritic network with DTM data using GIS and comparedto any site-scale data This study uses regulatory monitoringdatasets so that analysis is targeted to assessment scores andphysical features of interest to river managers Also the highvolume and wide spatial extent of data available from regu-latory sources allow for between-catchment comparisons

5 Conclusions

Although appreciation of catchment-level effects is consid-ered the epitome of understanding river functioning a keycomponent of the catchment ndash the river network ndash is over-looked and oversimplified by catchment-level studies Thisstudy finds that river network density plays a role in struc-turing the distribution river characteristics throughout thecatchment offering more detailed explanation than the clas-sic stream-order metric Network-dense areas are generallyfound to have higher habitat quality and diversity but modi-fication flow-type speed and sediment size show different re-sponses in different catchments This study suggests two po-tential reasons for this (1) there is evidence that confluencesin the river network increase diversity as is observed in thisstudy so the direction of mean river characteristic responsemay not be consistent and (2) there may be external factorssuch as sediment availability land cover and anthropogenicmodification that alter the direction of mean river characteris-tic response This paper demonstrates the functional responseof river characteristics to network topology and suggests thatthe inclusion of network topology in catchment-level studieswould add a layer of function-based understanding to suchstudies linking reaches to their catchments

The broad-scale methodology adopted by this study allowsthe network density metrics which are easily extracted fromopen-source data using GIS software to be compared to anyregulatory dataset The use of regulatory datasets allows notonly for analysis over a wider spatial extent but also for moreapplicable results for regulatory bodies Therefore the inter-disciplinary approach to characterizing network topology canbe applied efficiently and effectively to capture catchment-level impacts on reach-level functioning in any catchmentacross the globe

Data availability River Habitat Survey data are freely availablefrom the Environment Agency England River network data andelevation data are available from the Centre for Ecology and Hy-drology UK

Author contributions EH and NC conceptualised and designed thestudy EH performed the data analysis with the support of JMEH wrote the paper with comments provided by NC and JM

Competing interests The authors declare that they have no conflictof interest

Acknowledgements The authors would like to thank the Environ-ment Agency and the Centre for Ecology and Hydrology for accessto the data used in this paper This research is funded by the NaturalEnvironmental Research Council (NERC)

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2317

Review statement This paper was edited by Nandita Basu and re-viewed by Simone Bizzi and one anonymous referee

References

Belletti B Rinaldi M Buijse A D Gurnell A Mand Mosselman E A review of assessment methods forriver hydromorphology Environ Earth Sci 73 2079ndash2100httpsdoiorg101007s12665-014-3558-1 2015

Benda L Andras K Miller D and Bigelow P Confluenceeffects in rivers Interactions of basin scale network geom-etry and disturbance regimes Water Resour Res 40 1ndash15httpsdoiorg1010292003WR002583 2004a

Benda L Poff N L Miller D Dunne T Reeves GPess G and Pollock M The network dynamics hy-pothesis how channel networks structure riverine habi-tats Bioscience 54 413ndash427 httpsdoiorg1016410006-3568(2004)054[0413TNDHHC]20CO2 2004b

Benjamini Y and Hochberg Y Controlling the False DiscoveryRate A Practical and Powerful Approach to Multiple TestingJ Roy Stat Soc B 57 289ndash300 1995

Best J L Flow dynamics and sediment transport at river channelconfluences Birbeck University of London London 1985

Best J L The morphology of river channel con-fluences Prog Phys Geogr 10 157ndash174httpsdoiorg101177030913338601000201 1986

Best J L Flow dynamics at river channel confluences implica-tions for sediment transport and bed morphology in Recent de-velopments in fluvial sedimentology edited by Ethridge F GFlores R M and Harvey M D Spec Publ Soc Econ Pale-ontol Miner Tulsa Okla 27ndash35 1987

Biron P M Richer A Kirkbride A D Roy A G andHan S Spatial patterns of water surface topography ata river confluence Earth Surf Proc Land 27 913ndash928httpsdoiorg101002esp359 2002

Bravard J P and Gilvear D J Hydrological and geomorpho-logical structure of hydrosystems in The Fluvial Hydrosystemedited by Petts G E and Amoros C Springer Netherlands98ndash116 1996

Brierley G and Fryirs K River Styles a Geomorphic Approachto Catchment Characterization Implications for River Rehabili-tation in Bega Catchment New South Wales Australia EnvironManage 25 661ndash679 2000

Bunn S E and Arthington A H Basic principles and ecologi-cal consequences of altered flow regimes for aquatic biodiversityEnviron Manage 30 492ndash507 httpsdoiorg101007s00267-002-2737-0 2002

Church M and Kellerhals R On the statistics of grain size varia-tion along a gravel river Can J Earth Sci 15 1151ndash1160 1978

Clifford N J Hydrology the changing paradigm Prog Phys Ge-ogr 26 290ndash301 2002

Cohen P Andriamahefa H and Wasson J-G Towardsa regionalizaton of aquatic habitat distribution of meso-habitats at the scale of a large basin Regul RiversRes Manage 14 391ndash404 httpsdoiorg101002(SICI)1099-1646(19980910)145lt391AID-RRR513gt30CO2-W 1998

Davenport A J Gurnell A M and Armitage P D Habitat sur-vey and classification of urban rivers River Res Appl 20 687ndash704 httpsdoiorg101002rra785 2004

Davies N M Norris R H and Thoms M C Prediction and as-sessment of local stream habitat features using large-scale catch-ment characteristics Freshwater Biol 45 343ndash369 2000

Depettris C Mendiondo E M Neiff J and Rohrmann H Flooddefence strategy at the confluence of the Paranaacute-Paraguay riversProc Int Symp Flood Def Kassel He C31 C40 2000

Dollar E James C Rogers K and Thoms M A framework forinterdisciplinary understanding of rivers as ecosystems Geomor-phology 89 147ndash162 2007

Dovers S R and Day D G Australian rivers and statute law En-viron Plan Law J 5 90ndash108 1988

Downs P W Gregory K J and Brookes A How integratedis river basin management Environ Manage 15 299ndash309httpsdoiorg101007BF02393876 1991

Emery J C Gurnell A M Clifford N J and Petts G E Char-acteristics and controls of gravel-bed riffles An analysis of datafrom the river-habitat survey Water Environ J 18 210ndash216httpsdoiorg101111j1747-65932004tb00535x 2004

Evans I S and Minaacuter J A classification of geomorphometric vari-ables in International Geom-orphometry 2011 Geomoprhome-tryorg Redlabds CA 105ndash108 2011

Fausch K D Torgersen C E Baxter C V Li H WView C The O F Is R To N How U Among ISet S For C and Fishes S Landscapes to riverscapesBridging the gap between research and conservation of streamfishes Bioscience 52 483ndash498 httpsdoiorg1016410006-3568(2002)052[0483LTRBTG]20CO2 2002

Glickman M E Rao S R and Schultz M R False discov-ery rate control is a recommended alternative to Bonferroni-typeadjustments in health studies J Clin Epidemiol 67 850ndash857httpsdoiorg101016jjclinepi201403012 2014

Gupta V K and Mesa O J Runoff generation and hydro-logic response via channel network geomorphology - Re-cent progress and open problems J Hydrol 102 3ndash28httpsdoiorg1010160022-1694(88)90089-3 1988

Gupta V K and Waymire E D On the formulation of an analyt-ical approach to hydrologic response and similarity at the basinscale J Hydrol 65 95ndash123 1983

Gupta V K Waymire E D and Rodriguez-Iturbe I On scalesgravity and network structure in basin runoff in Scale problemsin hydrology Springer Netherlands 159ndash184 1986

Gurnell A M Rinaldi M Belletti B Bizzi S Blamauer BBraca G Buijse A D Bussettini M Camenen B ComitiF Demarchi L Garciacutea de Jaloacuten D Gonzaacutelez del TaacutenagoM Grabowski R C Gunn I D M Habersack H Hen-driks D Henshaw A J Kloumlsch M Lastoria B LatapieA Marcinkowski P Martiacutenez-Fernaacutendez V Mosselman EMountford J O Nardi L Okruszko T OrsquoHare M T PalmaM Percopo C Surian N van de Bund W Weissteiner Cand Ziliani L A multi-scale hierarchical framework for devel-oping understanding of river behaviour to support river manage-ment Aquat Sci 78 1ndash16 httpsdoiorg101007s00027-015-0424-5 2016

Harvey G L Gurnell A M and Clifford N J Characterisationof river reaches The influence of rock type Catena 76 78ndash88httpsdoiorg101016jcatena200809010 2008

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2306 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

standing of catchment functioning can become more holisticand beneficial to catchment management

Effective catchment management rests not only on im-proving scientific understanding of river form and functionacross multiple scales but also on better integration betweenthe key disciplines of catchment studies geomorphology hy-drology and ecology This type of interdisciplinary approachis critical for understanding complex multi-casual relation-ships in river systems (Dollar et al 2007) However catch-ment connectivity is parameterized differently by differentdisciplines based on their interests The discipline of geo-morphology focusses on characterizing the morphometry ofthe catchment either using general variables which are con-tinuous across the landscape (eg elevation slope curvature)or specific variables which represent individual features suchas catchments (eg drainage density shape area) or streams(eg stream order stream length) (Evans and Minaacuter 2011)Hydrology focusses on how the catchment influences hydro-graph and flood peak timing and magnitude Methods such asthe Geomorphic Instantaneous Unit Hydrograph (Rodriguez-Iturbe and Valdes 1979) focus on predicting the travel timeof water reaching channels and travelling downstream basedthe morphology of the catchment drainage network andprecipitation Aquatic ecology takes a network-centric ap-proach utilizing dendritic ecological networks (Peterson etal 2013) This method aims to take a spatially continuousview of rivers (Fausch et al 2002) in order to appreciatethe influence of flow and location in the network on dis-crete sites chosen for ecological sampling Spatial statisticalstream network models based on the branching of the net-work (Ver Hoef and Peterson 2010) are shown to be moreaccurate than a standard Euclidean distance kriging modelyet only worthwhile if data sites are distributed across thenetwork and are spatially correlated (Peterson et al 2013)Alternate methods for exploring relationships between net-work structure and ecological functioning are also based onEuclidean distance along the network (Ver Hoef and Peter-son 2010)

Each discipline represents the elements of the catchmentcritical to their field focussing on describing catchmentform catchment flow responses and ecological responsesHowever the geomorphology hydrology and ecology of thecatchment are interconnected across spatial and temporaldimensions in the fluvial hydrosystem (Petts and Amoros1996) We argue that the overlap between disciplinary meth-ods can be utilized to create a metric to represent the catch-ment that is meaningful across all disciplines and offers in-creased potential for effective catchment management utiliz-ing a multi- or inter-disciplinary approach

This paper repurposes metrics that focus on the topologyof the river network for a novel application to assess thekey link between the catchment and reach-level function-ing The metrics represent network density variation withincatchments and have functional applications across the fieldsof geomorphology hydrology and ecology (Sect 11) The

impacts of internal network structure on patterns of rivercharacteristics within catchments are explored by utilizingdatasets that are collected for regulatory purposes with ar-eas of higher network density likely to support greater riverquality and diversity (Sect 12) The utility of the topologicalmetrics is compared against stream order a classic but over-simplified method of accounting for network topology Thetopology metrics are calculated for catchments with compar-atively low energy and that are influenced by anthropogenicmodification as much of the previous evidence for increasesin diversity in network-dense areas has been from highly ero-sive mountainous catchments

11 Quantifying the river network at different scalesand dimensions

River network structure or network topology is one wayto conceptualize the integrated transport of water sedimentand nutrients from the upstream catchment to downstreamreaches The spatial arrangement of links (river channels)and nodes (confluences) concentrates the catchment effectin some areas of the landscape making network topology auseful archetype of catchment functioning (Gupta and Mesa1988)

Drainage density (the total length of the network dividedby catchment area) is most commonly used to compare theamount of the catchment covered by river channels but thisfails to quantify spatial variation within catchments and sooffers only a partial means for functionally assessing catch-ment similarities and differences To represent within catch-ment network structure stream order (Strahler 1957) order-ing river links along an upstream-to-downstream gradientbased on their upstream connectivity (Fig 1a) is also com-monly used However stream order does not account forthe spatial arrangement of links only their relative positionin the distance dimension of the catchment Conceptualiz-ing the catchment in this one dimension leads also to over-simplification for example first-order streams are thoughtof as upland headwater streams furthest away from the rivermouth yet often first-order streams are tributaries to high-order lowland streams with different characteristics than up-land streams

This paper argues that the spatial arrangement of linkswithin catchments must be considered across the dis-tance and height of the catchment to obtain a full three-dimensional appreciation of catchment effect through net-work topology Two methods from the field of hydrologyndash network width function (NWF Kirkby 1976) and linkconcentration function (LCF Gupta et al 1986) ndash offer in-creased dimensionality by accounting for the width of thenetwork (ie the number of links) at successive distances forthe NWF or elevations for the LCF

These methods quantify network topology within catch-ments with functional significance NWF has hydrologicalapplication representing the travel time of water through the

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2307

Figure 1 Topological metrics explored in this paper and the dimensions of the network they represent (a) Strahler stream ordering repre-senting only the distance dimension of the network (b) Distance network density representing the width dimension of the network at eachdistance interval (inspired by the network width function Kirkby 1976) (c) Elevation network density representing the width dimension ofthe network at each elevation interval (inspired by the link concentration function Gupta et al 1986)

network to predict the timing and magnitude of unit hydro-graphs and flood peaks (Rodriguez-Iturbe and Valdes 1979)with a more functionally specific method than the traditionalstream-ordering approach (Gupta and Waymire 1983) Ex-tending applications beyond the field of hydrology the tim-ing and magnitude of the hydrograph have direct influenceon instream ecology controlling the formation maintenanceand disturbance of physical habitats (Bunn and Arthing-ton 2002) Longitudinal connectivity of water and sedimentthrough the network is also one of the multiple dimensions ofthe fluvial hydrosystems approach to catchment ecohydrol-ogy (Petts and Amoros 1996) influencing the capacity forlateral and vertical connectivity and the development of theriparian corridor over time LCF is less frequently appliedin hydrograph prediction than NWF However it may betterreflect catchment hydrology by incorporating the effect ofgradient on the travel time of water rather than the constanttravel time suggested by NWF (Gupta et al 1986) Thesemetrics also have morphometric significance reflecting theinternal shape of the network by segmenting catchments intointervals to represent how network density changes withincatchments (Stepinski and Stepinski 2005)

This paper repurposes these metrics to reflect network den-sity as a feature of the catchment rather than as a method forhydrograph prediction Distance network density (modelledon the NWF) (Fig 1b) and elevation network density (mod-elled on the LCF) (Fig 1c) allow for the comparison andquantification of network topological variation both withinand between catchment with improved interdisciplinary andfunctional applicability than the stream-ordering approach

12 Network topology effects on river reach functioning

The topological structure of the river network configures theriver ecosystem (Bravard and Gilvear 1996) by impactingfunctioning at the reach and sub-reach scales The distancedimension of the catchment often represented by streamorder (Fig 1a) reflects upstream-to-downstream gradual

changes exhibited by many in-channel features and speciesIt forms the basis of classic geomorphic models highlight-ing the zones of sediment supply in the headwaters sedi-ment transfer in the mid reaches and sediment storage nearthe outlet (Schumm 1977) It is also a key component inclassic ecological models such as the River Continuum Con-cept (Vannote et al 1980) which describes gradual changesin grain size channel width invertebrates fish species andenergy sources along the gradient Both models suggest thatdiversity in channel morphology and biota may be highest inthe mid reaches as channels transition from erosional to de-positional environments The River Continuum Concept is apopular model but is critiqued for being too simplistic andfor neglecting discontinuity introduced by changes at conflu-ences (Perry and Schaeffer 1987 Rice et al 2001) Conflu-ences as nodes in the network are associated with changesin hydrological geomorphological (Best 1987 Church andKellerhals 1978) and ecological (Kiffney et al 2006 Riceet al 2001) conditions and have therefore been termed bio-diversity ldquohotspotsrdquo (Benda et al 2004b)

Confluence impacts extend throughout the river networkwith increased channel heterogeneity in the tributary andmain channel upstream and downstream of the confluence(Rice 2017) This has led to several theories relating to theimpact of numerous confluences in the context of the widernetwork The Link Discontinuity Concept shows the impactof confluences throughout the length on the main channelcreating step changes in sediment size before fining contin-ues downstream towards the next confluence along a ldquosed-imentary linkrdquo (Rice et al 2001) The Network DynamicsHypothesis posits that catchments with higher drainage den-sity and thus more confluences will have greater channelheterogeneity (Benda et al 2004b) despite drainage densityfailing to be a useful catchment characteristic for predictinglocal habitat features (Davies et al 2000) The hypothesisalso suggests that catchment shape will influence the impactof confluences as more compact catchments will have moresimilarly sized tributaries (Benda et al 2004b) which have

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2308 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

the greatest impact on channel morphology (Benda et al2004a) the greatest flow diversity (Schindfessel et al 2015)and the greatest fish community diversity (Osborne and Wi-ley 1992) In contrast others have found that tributaries thatdiffer most in size have the greatest impact For exampleJones and Schmidt (2016) suggest that high densities of smalltributaries flowing into a large channel cause small cumula-tive changes and Milesi and Melorsquos (2013) study concludedthat small tributaries flowing into large channels in the pe-ripheral regions of the catchment have the greatest impact onmacroinvertebrate assemblages

Interestingly there is little evidence of anthropogenic im-pacts at confluences in the literature but as confluences areproposed concentration points of catchment effects it seemslikely that they may be focal points for anthropogenic im-pacts For example flood events may occur downstream oflarge confluences as flood peaks converge creating the needfor flood defence measures (Depettris et al 2000) and scourand erosion at confluence junctions (Best 1986) increase theneed for bed and bank protection Also sediment size at con-fluences is shown to increase in many studies (Church andKellerhals 1978 Knighton 1980) but in tributaries whosewatersheds are dominated by agricultural land uses fine sed-iments may become dominant at confluences potentially al-tering river functioning (Owens et al 2005)

Many previous studies citing the impact of the networkspecifically confluences on river characteristics were con-ducted in highly erosive relatively natural environments(Network Dynamics Hypothesis Benda et al 2004b LinkDiscontinuity Concept Rice et al 2001) Therefore it willbe interesting to assess the impact of network structure onriver characteristics in catchments in England a landscapethat has undergone modification impacting catchment func-tioning for centuries (Macklin and Lewin 2003)

2 Methods

21 Study sites

Four catchments are selected for testing the impact of net-work topology on river characteristics in England The catch-ments are from the Demonstration Test Catchment pro-gramme (Fig 2) which are representative of 80 of soil andrainfall combinations in the UK (McGonigle et al 2014)This demonstrates the potential use of topological metricsfor catchments with varying geologies and land uses

The Avon and Wensum catchments have similar charac-teristics both being dominated by chalk geology with loweraverage annual rainfall and a high percentage of arable farm-ing land cover In comparison the Eden and Tamar are dom-inated by less permeable bedrock with higher average annualrainfall and a high percentage of grassland land covers Interms of their morphometry the Avon and Wensum both havean elongated shape and low drainage density The Wensum

has the lowest relief with a maximum elevation of 95 m TheTamar has the smallest catchment area (928 km2) and is themost circular The Eden is the largest catchment (2295 km2)and has the highest maximum elevation (246 m)

22 Network topology metrics

Network topology metrics were calculated for each catch-ment using the 1 50000 river network map derived fromboth a digital terrain model (DTM) and Ordnance Sur-vey data (Moore et al 1994) Anabranches and incorrectlydigitized links in the network are identified using RivEX(Hornby 2010) and removed Removing anabranches wasnecessary as the topological metrics were designed for den-dritic networks so multi-thread channels either naturally oc-curring or artificial ditches would distort the calculationsThis resulted in a total of 448 2812 1516 and 532 links inthe Avon Eden Tamar and Wensum respectively

Elevation data were extracted from the Integrated Hydro-logical DTM (Morris and Flavin 1994) a 50times50 m griddedelevation raster with a 10 cm vertical resolution Average el-evation of each link and the distance from each link to thenetwork outlet were extracted using RivEX (Hornby 2010)

To extract a measure of network density that varies spa-tially within the catchment each network is divided into20 intervals each of which represents 5 of the total dis-tance or highest elevation in the network (Fig 2) The net-work is divided in this manner based on the methods of theNWF and LCF which have functional application to hydro-graph prediction Twenty intervals provide a relatively coarsesampling of the network compared to the 100 intervals de-scribed by Stepinski and Stepinski (2005) when they adapteda morphometric variable circularity ratio to represent inter-nal catchment elongation Here a total of 20 intervals is cho-sen so that most intervals contain links for the density cal-culation whilst ensuring the spatial distribution of networkdensity within the catchment is characterized

Distance network density was calculated followingEq (1)

Distance network density=

[n(d0) n(di) n(dN)

](dNtimes 005)

(1)

where the number of links (n( )) within each 5 distanceinterval (di) from the outlet (d0) to the maximum distance inthe network (dN) is normalized by the width of the interval(dNtimes 005)

Elevation network density was calculated followingEq (2)

Elevation network density=

[n(z0) n(zi) n(zN)

](zNtimes 005)

(2)

where the number of links (n( )) within each 5 elevationinterval (zi) from the outlet (z0) to the maximum height ofthe network (zN) is normalized by the width of the inter-val (zNtimes 005) Normalization allows network densities to

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2309

Figure 2 Distance and elevation intervals for each Demonstration Test Catchment Avon (A) Eden (E) Tamar (T) and Wensum (W)(a) Percentage distance intervals used to calculate distance network density (b) Percentage elevation intervals used to calculate elevationnetwork density Map of catchment locations in England in the bottom-right corner

be compared between catchments controlling for differencesin size and elevation as well as within catchments

To assess the utility of the multi-dimensional topologymetrics in accounting for the spatial structure of the networkthe metrics are compared to the one-dimensional Strahlerstream-order metric extracted from the river network datasetusing RivEX (Hornby 2010)

23 River characteristics

The impact of network topology on channel functioning isexplored using a broad-scale approach ie adapting data col-lected for regulatory compliance to answer scientific ques-tions Adapting such datasets to scientific enquiry allowsanalysis to be conducted in many catchments across a widespatial extent There are many habitat monitoring methodsacross the globe with 121 survey methods recorded in over

26 different countries (Belletti et al 2015) so this methodmay be adapted to other countries

This study utilizes the River Habitat Survey (RHS Ravenet al 1996) a regulatory dataset collected by Englandrsquos En-vironment Agency which is used to reflect the river reachcharacteristics in each catchment This dataset has been usedto identify catchment effects on river characteristics in broad-scale studies by previous research (eg Harvey et al 2008Naura et al 2016 Vaughan et al 2013) but none have in-cluded the effects of network topology

Since 1994 over 24 000 sites have been sampled incatchments across England and Wales including the Avon(n= 418) Eden (n= 398) Tamar (n= 189) and Wensum(n=315) Surveys were conducted at random sites withineach 10 km2 of England and Wales to ensure geographic cov-erage however this produces sampling bias as streams inhigh-density areas will be under-represented in the datasetwhich is acknowledged in this study and discussed below

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2310 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

Table 1 RHS variables calculated from RHS observations

RHS variable Calculation from RHS observations Units

Habitat Quality A score indicating the degree of naturalness and diversity of the HQA scaleAssessment riparian zone based on observations in the reach of flow types(HQA) substrate channel and bank features riparian vegetation etc

Habitat A score indicating the degree of artificial modification of the channel HMS scaleModification based on observations in the reach of reinforcements re-sectioningScore (HMS) embankments weed cutting realignment culverts dams weirs etc

Sediment size =(minus8middotBOminus7middotCOminus35middotGPminus15middotSA+15middotSI+9middotCL)

(BO+CO+GP+SA+SI+CL) ApproxBO (boulder) CO (cobble) GP (gravelndashpebble) SA (sand) SI (silt) φ unitsand CL (clay) represent the number of spot checks allocated to eachsediment size class

Flow-type speed =(0middotDR+1middotNP+2middotUP+3middotSM+4middotRP+5middotUW+6middotBW+7middotCF+8middotCH+9middotFF)

(DR+NP+UP+SM+RP+UW+BW+CF+CH+FF) Flow typeDR (dry) NP (no perceptible flow) UP (upwelling) SM (smooth) speed scaleRP (rippled) UW (unbroken wave) BW (broken wave) CF (chaoticflow) CH (chute) FF (free-fall) represent the number of spot checksallocated to each flow speed class

At each site over 100 features are recorded along a 500 mreach with 10 ldquospot-checkrdquo surveys conducted every 50 mand a ldquosweep-uprdquo survey conducted across the whole reach(see Raven et al 1996 for details) Variables of interest thatare hypothesized to be impacted by network structure can becalculated from the RHS observations (Table 1)

The Habitat Quality Assessment (HQA) and Habitat Mod-ification Score (HMS) variables are both amalgamations ofRHS observations with individual features given a score de-rived by expert opinion (see Raven et al 1998 for more de-tails) The scoring systems are subjective but HQA and HMSprovide overviews of the channel condition that are widelyapplied for regulatory compliance The scores are thereforeincluded in this study to reflect how they may be impactedby network topology

The remaining RHS variables are calculated directly fromRHS observations and so are more objective Sediment sizeis calculated as a reach average of spot-check observationsusing the same method as previous studies (Davenport et al2004 Emery et al 2004 Harvey et al 2008) Flow-typespeed was calculated in the same manner as sediment sizeusing values of flow which represent an approximate flowvelocity gradient defined in Davenport et al (2004) Thesevariables were chosen to reflect dominant geomorphic pro-cesses occurring in each reach and due to the prominence ofsediment size and flow type in defining physical habitats forinstream biota (Rowntree and Wadeson 1996) The variablesare likely to be impacted by the density of the river networkas they have been shown to be impacted by individual con-fluences For example channels are shown to become moregeomorphologically heterogeneous (Benda et al 2004a) andsubstrate size has been shown to coarsen at confluences (Riceet al 2001) Surface flow types are also likely to become

more diverse at confluences as the convergence of channelscreates a number of different flow environments (Best 1985)that result in different water-surface topographies (Biron etal 2002)

It must be noted that the RHS dataset was collected forregulatory compliance and was not directly intended for sci-entific enquiry Therefore there is a limitation in the amountof detail that can be extracted about physical processes as theobservations recorded are an average across a 500 m reachDespite this the wide spatial coverage of the dataset makesit a powerful tool allowing analysis to be conducted acrossmultiple catchments with differing characteristics

For each distance and elevation interval created by the net-work topology metrics descriptive statistics (mean median90th and 10th percentiles) of each RHS variable are calcu-lated Despite the RHS sampling strategy (Jeffers 1998b) bi-asing site selection towards less dense areas of the networkmost distance and elevation intervals contained RHS sites(with only some low-density intervals not containing RHSsites) This method is designed to account for natural varia-tion and modification at individual RHS sites in order to as-sess broad patterns of reach characteristics at the catchmentlevel

24 Statistical analysis

Analysis is conducted with all catchments combined into asingle population to identify overall trends across all catch-ments a method used in previous broad-scale studies Theanalysis is also split into individual catchments to identifyhow the relationship between network topology metrics andriver reach characteristics differed between catchments

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2311

Kendall correlation

Correlation tests are used to determine the strength and di-rection of the association between the descriptive statisticsof the RHS variables and distance network density eleva-tion network density and stream order to ascertain how reachcharacteristics respond to network topology Kendallrsquos cor-relation method was used as the variables have non-normaldistributions a small sample size and tied data values (Helseland Hirsch 2002) The effect size of Kendallrsquos τ is lowerthan other correlation methods with strong correlations oc-curring with τ values greater than 07 (Helsel and Hirsch2002)

As multiple correlations are conducted false discoveryrate (Benjamini and Hochberg 1995) corrections were ap-plied to the p-values produced from the Kendall correlationsto reduce the risk of type I error The false discovery ratemethod has been found to be more powerful than other pro-cedures for controlling for multiple tests (Glickman et al2014)

3 Results

31 Differences in network topology metrics betweencatchments

The topological metrics developed in this study show the in-ternal structure of the network for each catchment The sep-aration of the catchments into distance and elevation inter-vals emphasizes different features of the catchment The dis-tance intervals (Fig 2a) are arranged longitudinally withinthe catchment highlighting sub-basins within each catch-ment The elevation intervals (Fig 2b) have a radial arrange-ment centering around the incised main channel of eachcatchment

Distance network density is higher in the Eden (284plusmn103) and Tamar (441plusmn 219) compared to the Avon (47plusmn19) and Wensum (68plusmn 07) interesting as the Tamar is thesmallest catchment by area The shape of the distance net-work density function reflects the internal shape of the net-work (Fig 3a) For example the Tamar has a peaked den-sity distribution reflecting the circular shape of the catch-ment such that the majority of links are at 55 ndash65 dis-tance from the catchment outlet The Avon and Eden reflectsimilar internal network structures both exhibiting a bimodaldensity distribution despite the differences in the number oflinks in the catchments The density distribution of the Wen-sum has a more complex internal distribution of links withmultiple peaks in density

Elevation network density (Fig 3b) shows similar densitydistribution shapes to distance network density with a uni-modal distribution for the Tamar and multi-modal distribu-tions in the other catchments In contrast to distance networkdensity elevation network density shows the highest peaks in

density in the Tamar (103plusmn 50) and Wensum (101plusmn 43)despite the Wensum having the lowest network elevation andhas lower values in the Avon (34plusmn10) and Eden (58plusmn25)The peak densities in the Wensum occur at similar positionsin the elevation and distance intervals whereas the peaks inthe other catchments are negatively skewed showing the net-work density is highest at lowndashmid elevations

Nearly half of the links in each catchment are classified asfirst-order streams and the number of links declines exponen-tially towards the highest orders in all four catchments Thereare weak correlations (τ =minus003 to 017) between the threenetwork topology metrics distance network density eleva-tion network density and stream order This suggests that themetrics are independent and reflect different aspects of rivernetwork topology

32 River characteristic relationships with networktopology metrics

RHS sites in the Avon and Wensum have similar river char-acteristics Both have lower habitat quality high modifica-tion fine sediment and slower flow types than the Eden andTamar When all catchments are combined there are signif-icant (p lt 005 after p-value correction) correlations withmost descriptive statistics for each RHS variable and dis-tance network density (Fig 4) There are consistently pos-itive correlations with HQA and flow-type speed and neg-ative correlations with HMS and sediment size There arefewer and weaker significant correlations with elevation net-work density (Fig 4) The only significant correlations withstream order are with HMS which shows a negative correla-tion (Fig 4)

There are numerous significant correlations between thenetwork topology metrics and RHS variables for individualcatchments many of which were also shown to be signifi-cant after the correction to the p-value The results show thatcatchments have different responses to the network topol-ogy metrics of distance network density and elevation net-work density Distance network density only has significantcorrelations with the regulatory scoring variables (HQA andHMS) in the Eden and Tamar (Fig 4) Elevation networkdensity however has a wider array of significant correlationswith the scoring variables particularly HQA which showssubtle peaks and troughs reflecting the distribution of bothnetwork density metrics (Fig 3a and b) HMS shows mostlynegative correlations mainly with elevation network densityapart from the Eden which has significant positive correla-tion across all HMS descriptive statistics (Fig 4) Visually10th percentile HMS is most variable to network density withpeaks and troughs responding to the network density distri-butions (Fig 3a and b)

For individual RHS features the response to networktopology varies between catchments (Fig 4) The Avon hasnegative correlations between flow-type speed and distancenetwork density with an evident drop in 10th percentile

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2312 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

Figure 3 Network topology metrics (a) distance network density and (b) elevation network density Descriptive statistics of each RHSvariable over (a) distance and (b) elevation for each catchment with smooth loess lines to indicate trend Network topology metrics arenormalized between 0 and 1 and HMS score is logarithmically transformed for display purposes

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2313

Figure 4 Summary of correlations between distance network density elevation network density stream order and RHS variables for allcatchments combined and each individual catchment Significance of correlation is indicated by point size with the largest points significantpost p-value correction Effect size (Kendallrsquos τ ) is indicated by colour No correlation was possible between stream order and 10th percentilesediment size in the Avon due to no variation in the RHS variable

flow-type speed associated with peaks in network density(Fig 3a) The Eden and Tamar however have positive cor-relations with mean and 90th percentile flow-type speed fordistance network density but negative correlations with me-dian and 10th percentile elevation network density The Wen-sum shows positive correlations between flow-type speed andelevation network density Sediment size has a consistent re-sponse to distance network density with the Eden and Wen-sum showing negative correlations with the sediment size(Fig 4) For elevation network density the Avon shows neg-ative correlations with sediment size whereas the Tamar andWensum show positive correlations (Fig 4)

There were few significant correlations between stream or-der and the RHS variables in individual catchments (Fig 4)The only significant correlation after p-value correction iswith 90th percentile HMS in the Wensum which shows astrong negative relationship

4 Discussion

41 A new approach to utilizing network topology incatchment-level analysis

Network density metrics represent an alternative approach toaccount for network topology in catchment-level studies op-timizing the width dimension of the network (or the num-

ber of links in the network) as opposed to the commonplacestream-order metric which only reflects the longitudinal po-sition of links in a network (Fig 1) This study demonstratesthat two topology metrics can be calculated simply from aDTM with GIS and using a broad-scale analysis of river at-tributes can be used to investigate the functional processeswithin catchments

While the two network density metrics have similar forms(ie forms are consistently unimodal or multi-modal) thespatial configuration of the distance and elevation intervalsused in the calculation of network density varies and mayimpact the effectiveness of each topological metric Distancenetwork density separates the catchment into intervals basedon distance which spread upstream from the outlet (Fig 2a)reflecting natural sub-basins within the fractal structure of thecatchment (Lashermes and Foufoula-Georgiou 2007) Thisdiffers from elevation network density which separates thecatchment into intervals based on elevation which radiate outfrom the main channel of the network (Fig 2b) The config-uration means that distance intervals contain streams that arein closer proximity to one another rather than the more distalconfiguration created by the elevation intervals suggesting adegree of spatial dependency in river functioning This hasbeen highlighted in previous studies where spatial networkstructure has a stronger influence on some in-channel pro-cesses than predictor variables such as elevation (Steel et al

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2314 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

2016) However elevation intervals contain RHS sites thatalthough they may be distal may have similar properties aselevation has been strongly related to RHS variables includ-ing flow type substrate etc in a number of studies (Jeffers1998a Naura et al 2016 Vaughan et al 2013)

42 Impacts of network topology on rivercharacteristics

River characteristics are assessed using the RHS dataset Theobservations made by the RHS dataset (Table 1) cannot of-fer the level of detail regarding geomorphological processthat river classifications which consider multiple scales canoffer (eg Brierley and Fryirs 2000 Gurnell et al 2016)While process-based classifications are preferable broad-scale monitoring datasets such as the River Habitat Surveymay still be useful when combined with map-derived datato explore controls on river characteristics (Harvey et al2008 Naura et al 2016 Vaughan et al 2013) Howeverthere are biases in RHS data collection an inherent limitationwhen using existing datasets (Vaughan and Ormerod 2010)specifically the standardized survey length of 500 m reachthat will capture differing amounts of natural variability de-pending on the size of the river While this must be notedthere are few significant correlations between river charac-teristics identified with stream order (Fig 4) which suggeststhat channel size is not influencing the RHS variables to agreat degree in these catchments

In this study it is anticipated that sites in high networkdensity areas will have higher levels of habitat diversity asindicated by previous studies of confluences and networks(Benda et al 2004a Best 1985 Rice 2017) in turn in-creasing mean sediment size and flow-type speed comparedto sites in low-density areas The results of the correlationsbetween distance network density and river characteristicswhen all catchments are combined support this hypothesiswith greater HQA flow-type speed and coarser sedimentsizes observed in areas with high distance network density(Fig 4)

For individual catchments elevation network density in-duces a stronger positive HQA response across all catch-ments than distance network density (Fig 4) This supportsthe evidence that individual confluences (Rice et al 2006)and high densities of confluences increase physical hetero-geneity within the river network (Benda et al 2004b Rice2017) However flow-type speed and sediment size responddifferently to network density in individual catchments

Slower flow types are observed in high network densityareas of the Avon and Tamar whereas faster flow types areobserved in high-elevation network density areas of the Wen-sum Individual confluences are shown to create numeroushigh- and low-speed flow environments (Best 1987) thatmay be observed in surface water topography (Biron et al2002) It was expected that the introduction of the additionalflow types by a high density of confluences in relatively low-

energy rivers would increase mean reach flow-type speedhowever the correlation analysis (Fig 4) suggests that insome catchments mean flow-type speed is reduced

Sediment size response also shows variation betweencatchments Sediment size is coarser in network-dense ar-eas of the Avon and Eden as expected but is finer in boththe Tamar and Wensum (Fig 4) The evidence from high-energy rivers shows that sediment becomes coarser down-stream and finer upstream of certain confluences (Benda etal 2004a Rice 1998) and in this case high numbers of con-fluences were expected to increase mean sediment size of thereach However the impact on sediment size is dependent onthe sediment calibre of the incoming tributary being higherthan the main channel with enough energy to transport thecoarse sediment to the confluence for numerous tributariesin an area The Tamar and Wensum have different rangesof sediment sizes with the Tamar having coarser sedimentthan the Wensum on average (Fig 3) This implies that trib-utaries in the Tamar may be energy limited not transportingcoarse sediments to confluences and the Wensum may suf-fer from inputs of fine sediments from the high percentageof arable land that concentrates in network-dense areas Thishas before been observed in a low-energy modified catch-ment where anthropogenic modifications in tributaries re-duced coarse sediment and flow capacity causing either lim-ited confluence impact or localized sediment fining (Singer2008)

Others have related the capacity of confluences to alterreach features to the morphometry of catchments with largerand more circular catchments containing a higher percentageof confluences that have a significant impact (Rice 2017)Based on this theory the Eden and Tamar are likely to havethe greatest impact as they are the most circular and steepestof the four catchments (although the Tamar is the smallestby area) Yet there is no clear pattern indicating that thesecatchments respond differently than the Avon and Wensum(Fig 4) with catchments responding differently to differ-ent variables This perhaps suggests that rather than networkdensity having a directional impact on factors such as flow-type speed and sediment size it has an impact on overall het-erogeneity at the reach level as suggested by previous stud-ies and that specific directional change occurs at the sub-reach level

An increase in channel modification is also hypothesizeddue to the increase in flood peak downstream from conflu-ences (Depettris et al 2000) and the scour and erosion as-sociated with confluence junctions (Best 1986) potentiallyincreasing the need for bed and bank protection The corre-lations between distance network density and HMS when allcatchments are combined undermine this hypothesis show-ing less channel modification where distance network den-sity is higher (Fig 4) There were few significant correlationswith individual catchments but HMS in the Eden was consis-tently observed to be higher in network-dense areas (Fig 4)This may be due to the Edenrsquos high elevation and steep to-

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2315

pography inducing a higher energy environment where scourand erosion processes in areas with high numbers of conflu-ences would be more likely to be present

Differences in RHS variable responses also differed be-tween the descriptive statistics considered Often the ex-tremes 90th and 10th percentiles showed more significantand stronger correlations than the mean or median (Fig 4)This may reflect findings from previous studies which sug-gest that not all confluences cause reach-scale changes (Rice1998) that perhaps the changes to river character induced bycertain confluences only influence certain reaches whereasothers are left unaffected creating less pronounced responsesin the mean and median of variables External factors mayalso influence this trend for example 10th percentile HMSvisually responds dramatically to network density metrics(Fig 3a and b) compared to the other descriptive statisticsThis suggests that the most natural sites (ie with the low-est HMS score) are responding differently to network den-sity with the most natural sites having less modificationin network-dense areas whereas less natural sites becomemore modified in network-dense areas This reflects the HQAscore results which visually (Fig 3a and b) and statistically(Fig 4) vary with distance network density except for the10th percentile These sites with the lowest habitat qualityand naturalness are likely influenced by anthropogenic fac-tors that are independent of network density reducing habitatquality scores at impacted sites

The two network density metrics have differing impacts onriver characteristics While distance network density showsconsistently significant correlations when all catchments arecombined individual catchments respond more frequentlyand more strongly to elevation network density (Fig 4)This may be because there is a dramatic split in distancenetwork density values between the more upland drainage-dense catchments Eden and Tamar than the lowland chalklow-drainage catchments Avon and Wensum The combinedcorrelation will therefore in part reflect the difference be-tween the catchments which have different ranges of rivercharacteristics (Fig 3) This is not the case for elevation net-work density which has higher density values in the Tamarand Wensum than the Avon and Eden so therefore the char-acteristics of the catchments will have less bearing on thecombined correlation However there are patterns identifiedwith distance network density in individual catchments thatare not present with elevation network density increasedflow-type speed and sediment size in the Eden and reducedflow-type speed in the Avon with network density (Fig 4)which show its usefulness

The results suggest that the distance network density andelevation network density metrics quantify different dimen-sions of network topology which are shown to exhibit func-tionally meaningful patterns for river reach characteristicsbased on the correlation results Perhaps within catchmentselevation network density provides the more powerful met-ric for individual catchments but distance network density

better accounts for the drainage density of the catchmentsallowing it to be applied across multiple catchments

43 Comparison of stream order to network densitymetrics

The classic method of accounting for network topologystream order is critiqued for failing to represent disconti-nuities in the network and simplifying the network into agradient The number of links in different stream orders isconsistent across all catchments not reflecting the internalstructure of the network or the variety between catchmentsthat is achieved by the distance network density and eleva-tion network density metrics The analysis of the two net-work density metrics presented in this paper shows that dis-tances from source and elevation are not mutually exclusive(Fig 2a and b) contrary to the stream-order metric whichrepresents streams as upstream to downstream or upland tolowland

Stream order has few significant correlations with manyof the river characteristics considered in this study Negativecorrelations with HMS in the Wensum were statistically sig-nificant post p-value correction likely driving the significantrelationship with all catchments combined for this variable(Fig 4) This suggests that modification is greater upstreamin the Wensum contrary to ideas that downstream reachesmay show greater anthropogenic modification Intense agri-cultural land use in the upper reaches of the Wensum islikely to be the cause of the high HMS scores upstreamThe lack of significant correlations suggests that stream or-der and therefore an upstream-to-downstream gradient arenot the predominant pattern in river characteristics despitethe description of such a gradient by geomorphic (Schumm1977) and ecological frameworks (Vannote et al 1980)This is surprising as distance and elevation which both re-flect the upstream-to-downstream gradient have proven tobe important factors in previous studies explaining patternsof RHS features at a national level (Jeffers 1998a Vaughanet al 2013) This implies that upstream-to-downstream gra-dient may not sufficiently reflect patterns of river charac-teristics through the river network within individual catch-ments Others have also found that stream order has weakand inconsistent relationships with biodiversity patterns inriver systems arguing that the topological measure has no di-rect mechanistic control on biodiversity (Vander Vorste et al2017) Instead this study finds that the network density met-rics are a more powerful metric which conceptually providean improved method of accounting for the impacts of net-work topology on the fluvial system exhibiting relationshipswith river characteristics particularly habitat quality score(Fig 4)

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2316 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

44 Applicability of network topology metrics todifferent environments

Much of the seminal work on network and confluence im-pacts (eg the Network Dynamics Hypothesis Benda et al2004b and Link Discontinuity Concept Rice et al 2001) isconducted in natural highly erosive catchments with first-hand empirical measurements However in an age whenrivers and their catchments are increasingly altered by an-thropogenic modification (Meybeck 2003) contemporarystudies must not only aim to expand knowledge but also findmethods of transferring knowledge to many increasingly al-tered catchments (Clifford 2002)

The catchments selected by this study are more greatlymodified and although they reflect a range of fluvial envi-ronments in England are lower-energy catchments than theseminal studies Benda et al (2004a) suggest that confluenceeffects in less active landscapes would be subdued comparedto highly erosive landscapes but the evidence presented heredemonstrates the utility of evaluating network topologicalstructure in studies on catchment-level effects in any typeof fluvial environment including those with low-energy andwidespread anthropogenic changes The response of someriver characteristics varied between catchments observationsof flow-type speed sediment size and modifications showeddifferent responses to network density in different catch-ments This suggests that the functional effect of these topo-logical metrics is catchment dependent and likely is influ-enced by external catchment characteristics such as land usenot considered in this study although impact did not appearto vary with catchment topography or circularity as has beenshown in prior studies (Benda et al 2004a Rice 2017) Thisshould be explored further in future research to enable rec-ommendations to be made regarding where and how reachesmay respond to network density The response of habitatquality score was however consistent across catchments andbetween metrics showing that habitat quality is greater inareas with high network density (Fig 4) as hypothesizedby the Network Dynamics Hypothesis (Benda et al 2004b)and demonstrated by studies on individual catchments (Rice2017 Rice et al 2006)

The methods presented in this paper are designed to be im-plemented in any catchment with a dendritic network struc-ture The topology metrics can easily be calculated from anydendritic network with DTM data using GIS and comparedto any site-scale data This study uses regulatory monitoringdatasets so that analysis is targeted to assessment scores andphysical features of interest to river managers Also the highvolume and wide spatial extent of data available from regu-latory sources allow for between-catchment comparisons

5 Conclusions

Although appreciation of catchment-level effects is consid-ered the epitome of understanding river functioning a keycomponent of the catchment ndash the river network ndash is over-looked and oversimplified by catchment-level studies Thisstudy finds that river network density plays a role in struc-turing the distribution river characteristics throughout thecatchment offering more detailed explanation than the clas-sic stream-order metric Network-dense areas are generallyfound to have higher habitat quality and diversity but modi-fication flow-type speed and sediment size show different re-sponses in different catchments This study suggests two po-tential reasons for this (1) there is evidence that confluencesin the river network increase diversity as is observed in thisstudy so the direction of mean river characteristic responsemay not be consistent and (2) there may be external factorssuch as sediment availability land cover and anthropogenicmodification that alter the direction of mean river characteris-tic response This paper demonstrates the functional responseof river characteristics to network topology and suggests thatthe inclusion of network topology in catchment-level studieswould add a layer of function-based understanding to suchstudies linking reaches to their catchments

The broad-scale methodology adopted by this study allowsthe network density metrics which are easily extracted fromopen-source data using GIS software to be compared to anyregulatory dataset The use of regulatory datasets allows notonly for analysis over a wider spatial extent but also for moreapplicable results for regulatory bodies Therefore the inter-disciplinary approach to characterizing network topology canbe applied efficiently and effectively to capture catchment-level impacts on reach-level functioning in any catchmentacross the globe

Data availability River Habitat Survey data are freely availablefrom the Environment Agency England River network data andelevation data are available from the Centre for Ecology and Hy-drology UK

Author contributions EH and NC conceptualised and designed thestudy EH performed the data analysis with the support of JMEH wrote the paper with comments provided by NC and JM

Competing interests The authors declare that they have no conflictof interest

Acknowledgements The authors would like to thank the Environ-ment Agency and the Centre for Ecology and Hydrology for accessto the data used in this paper This research is funded by the NaturalEnvironmental Research Council (NERC)

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2317

Review statement This paper was edited by Nandita Basu and re-viewed by Simone Bizzi and one anonymous referee

References

Belletti B Rinaldi M Buijse A D Gurnell A Mand Mosselman E A review of assessment methods forriver hydromorphology Environ Earth Sci 73 2079ndash2100httpsdoiorg101007s12665-014-3558-1 2015

Benda L Andras K Miller D and Bigelow P Confluenceeffects in rivers Interactions of basin scale network geom-etry and disturbance regimes Water Resour Res 40 1ndash15httpsdoiorg1010292003WR002583 2004a

Benda L Poff N L Miller D Dunne T Reeves GPess G and Pollock M The network dynamics hy-pothesis how channel networks structure riverine habi-tats Bioscience 54 413ndash427 httpsdoiorg1016410006-3568(2004)054[0413TNDHHC]20CO2 2004b

Benjamini Y and Hochberg Y Controlling the False DiscoveryRate A Practical and Powerful Approach to Multiple TestingJ Roy Stat Soc B 57 289ndash300 1995

Best J L Flow dynamics and sediment transport at river channelconfluences Birbeck University of London London 1985

Best J L The morphology of river channel con-fluences Prog Phys Geogr 10 157ndash174httpsdoiorg101177030913338601000201 1986

Best J L Flow dynamics at river channel confluences implica-tions for sediment transport and bed morphology in Recent de-velopments in fluvial sedimentology edited by Ethridge F GFlores R M and Harvey M D Spec Publ Soc Econ Pale-ontol Miner Tulsa Okla 27ndash35 1987

Biron P M Richer A Kirkbride A D Roy A G andHan S Spatial patterns of water surface topography ata river confluence Earth Surf Proc Land 27 913ndash928httpsdoiorg101002esp359 2002

Bravard J P and Gilvear D J Hydrological and geomorpho-logical structure of hydrosystems in The Fluvial Hydrosystemedited by Petts G E and Amoros C Springer Netherlands98ndash116 1996

Brierley G and Fryirs K River Styles a Geomorphic Approachto Catchment Characterization Implications for River Rehabili-tation in Bega Catchment New South Wales Australia EnvironManage 25 661ndash679 2000

Bunn S E and Arthington A H Basic principles and ecologi-cal consequences of altered flow regimes for aquatic biodiversityEnviron Manage 30 492ndash507 httpsdoiorg101007s00267-002-2737-0 2002

Church M and Kellerhals R On the statistics of grain size varia-tion along a gravel river Can J Earth Sci 15 1151ndash1160 1978

Clifford N J Hydrology the changing paradigm Prog Phys Ge-ogr 26 290ndash301 2002

Cohen P Andriamahefa H and Wasson J-G Towardsa regionalizaton of aquatic habitat distribution of meso-habitats at the scale of a large basin Regul RiversRes Manage 14 391ndash404 httpsdoiorg101002(SICI)1099-1646(19980910)145lt391AID-RRR513gt30CO2-W 1998

Davenport A J Gurnell A M and Armitage P D Habitat sur-vey and classification of urban rivers River Res Appl 20 687ndash704 httpsdoiorg101002rra785 2004

Davies N M Norris R H and Thoms M C Prediction and as-sessment of local stream habitat features using large-scale catch-ment characteristics Freshwater Biol 45 343ndash369 2000

Depettris C Mendiondo E M Neiff J and Rohrmann H Flooddefence strategy at the confluence of the Paranaacute-Paraguay riversProc Int Symp Flood Def Kassel He C31 C40 2000

Dollar E James C Rogers K and Thoms M A framework forinterdisciplinary understanding of rivers as ecosystems Geomor-phology 89 147ndash162 2007

Dovers S R and Day D G Australian rivers and statute law En-viron Plan Law J 5 90ndash108 1988

Downs P W Gregory K J and Brookes A How integratedis river basin management Environ Manage 15 299ndash309httpsdoiorg101007BF02393876 1991

Emery J C Gurnell A M Clifford N J and Petts G E Char-acteristics and controls of gravel-bed riffles An analysis of datafrom the river-habitat survey Water Environ J 18 210ndash216httpsdoiorg101111j1747-65932004tb00535x 2004

Evans I S and Minaacuter J A classification of geomorphometric vari-ables in International Geom-orphometry 2011 Geomoprhome-tryorg Redlabds CA 105ndash108 2011

Fausch K D Torgersen C E Baxter C V Li H WView C The O F Is R To N How U Among ISet S For C and Fishes S Landscapes to riverscapesBridging the gap between research and conservation of streamfishes Bioscience 52 483ndash498 httpsdoiorg1016410006-3568(2002)052[0483LTRBTG]20CO2 2002

Glickman M E Rao S R and Schultz M R False discov-ery rate control is a recommended alternative to Bonferroni-typeadjustments in health studies J Clin Epidemiol 67 850ndash857httpsdoiorg101016jjclinepi201403012 2014

Gupta V K and Mesa O J Runoff generation and hydro-logic response via channel network geomorphology - Re-cent progress and open problems J Hydrol 102 3ndash28httpsdoiorg1010160022-1694(88)90089-3 1988

Gupta V K and Waymire E D On the formulation of an analyt-ical approach to hydrologic response and similarity at the basinscale J Hydrol 65 95ndash123 1983

Gupta V K Waymire E D and Rodriguez-Iturbe I On scalesgravity and network structure in basin runoff in Scale problemsin hydrology Springer Netherlands 159ndash184 1986

Gurnell A M Rinaldi M Belletti B Bizzi S Blamauer BBraca G Buijse A D Bussettini M Camenen B ComitiF Demarchi L Garciacutea de Jaloacuten D Gonzaacutelez del TaacutenagoM Grabowski R C Gunn I D M Habersack H Hen-driks D Henshaw A J Kloumlsch M Lastoria B LatapieA Marcinkowski P Martiacutenez-Fernaacutendez V Mosselman EMountford J O Nardi L Okruszko T OrsquoHare M T PalmaM Percopo C Surian N van de Bund W Weissteiner Cand Ziliani L A multi-scale hierarchical framework for devel-oping understanding of river behaviour to support river manage-ment Aquat Sci 78 1ndash16 httpsdoiorg101007s00027-015-0424-5 2016

Harvey G L Gurnell A M and Clifford N J Characterisationof river reaches The influence of rock type Catena 76 78ndash88httpsdoiorg101016jcatena200809010 2008

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2307

Figure 1 Topological metrics explored in this paper and the dimensions of the network they represent (a) Strahler stream ordering repre-senting only the distance dimension of the network (b) Distance network density representing the width dimension of the network at eachdistance interval (inspired by the network width function Kirkby 1976) (c) Elevation network density representing the width dimension ofthe network at each elevation interval (inspired by the link concentration function Gupta et al 1986)

network to predict the timing and magnitude of unit hydro-graphs and flood peaks (Rodriguez-Iturbe and Valdes 1979)with a more functionally specific method than the traditionalstream-ordering approach (Gupta and Waymire 1983) Ex-tending applications beyond the field of hydrology the tim-ing and magnitude of the hydrograph have direct influenceon instream ecology controlling the formation maintenanceand disturbance of physical habitats (Bunn and Arthing-ton 2002) Longitudinal connectivity of water and sedimentthrough the network is also one of the multiple dimensions ofthe fluvial hydrosystems approach to catchment ecohydrol-ogy (Petts and Amoros 1996) influencing the capacity forlateral and vertical connectivity and the development of theriparian corridor over time LCF is less frequently appliedin hydrograph prediction than NWF However it may betterreflect catchment hydrology by incorporating the effect ofgradient on the travel time of water rather than the constanttravel time suggested by NWF (Gupta et al 1986) Thesemetrics also have morphometric significance reflecting theinternal shape of the network by segmenting catchments intointervals to represent how network density changes withincatchments (Stepinski and Stepinski 2005)

This paper repurposes these metrics to reflect network den-sity as a feature of the catchment rather than as a method forhydrograph prediction Distance network density (modelledon the NWF) (Fig 1b) and elevation network density (mod-elled on the LCF) (Fig 1c) allow for the comparison andquantification of network topological variation both withinand between catchment with improved interdisciplinary andfunctional applicability than the stream-ordering approach

12 Network topology effects on river reach functioning

The topological structure of the river network configures theriver ecosystem (Bravard and Gilvear 1996) by impactingfunctioning at the reach and sub-reach scales The distancedimension of the catchment often represented by streamorder (Fig 1a) reflects upstream-to-downstream gradual

changes exhibited by many in-channel features and speciesIt forms the basis of classic geomorphic models highlight-ing the zones of sediment supply in the headwaters sedi-ment transfer in the mid reaches and sediment storage nearthe outlet (Schumm 1977) It is also a key component inclassic ecological models such as the River Continuum Con-cept (Vannote et al 1980) which describes gradual changesin grain size channel width invertebrates fish species andenergy sources along the gradient Both models suggest thatdiversity in channel morphology and biota may be highest inthe mid reaches as channels transition from erosional to de-positional environments The River Continuum Concept is apopular model but is critiqued for being too simplistic andfor neglecting discontinuity introduced by changes at conflu-ences (Perry and Schaeffer 1987 Rice et al 2001) Conflu-ences as nodes in the network are associated with changesin hydrological geomorphological (Best 1987 Church andKellerhals 1978) and ecological (Kiffney et al 2006 Riceet al 2001) conditions and have therefore been termed bio-diversity ldquohotspotsrdquo (Benda et al 2004b)

Confluence impacts extend throughout the river networkwith increased channel heterogeneity in the tributary andmain channel upstream and downstream of the confluence(Rice 2017) This has led to several theories relating to theimpact of numerous confluences in the context of the widernetwork The Link Discontinuity Concept shows the impactof confluences throughout the length on the main channelcreating step changes in sediment size before fining contin-ues downstream towards the next confluence along a ldquosed-imentary linkrdquo (Rice et al 2001) The Network DynamicsHypothesis posits that catchments with higher drainage den-sity and thus more confluences will have greater channelheterogeneity (Benda et al 2004b) despite drainage densityfailing to be a useful catchment characteristic for predictinglocal habitat features (Davies et al 2000) The hypothesisalso suggests that catchment shape will influence the impactof confluences as more compact catchments will have moresimilarly sized tributaries (Benda et al 2004b) which have

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2308 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

the greatest impact on channel morphology (Benda et al2004a) the greatest flow diversity (Schindfessel et al 2015)and the greatest fish community diversity (Osborne and Wi-ley 1992) In contrast others have found that tributaries thatdiffer most in size have the greatest impact For exampleJones and Schmidt (2016) suggest that high densities of smalltributaries flowing into a large channel cause small cumula-tive changes and Milesi and Melorsquos (2013) study concludedthat small tributaries flowing into large channels in the pe-ripheral regions of the catchment have the greatest impact onmacroinvertebrate assemblages

Interestingly there is little evidence of anthropogenic im-pacts at confluences in the literature but as confluences areproposed concentration points of catchment effects it seemslikely that they may be focal points for anthropogenic im-pacts For example flood events may occur downstream oflarge confluences as flood peaks converge creating the needfor flood defence measures (Depettris et al 2000) and scourand erosion at confluence junctions (Best 1986) increase theneed for bed and bank protection Also sediment size at con-fluences is shown to increase in many studies (Church andKellerhals 1978 Knighton 1980) but in tributaries whosewatersheds are dominated by agricultural land uses fine sed-iments may become dominant at confluences potentially al-tering river functioning (Owens et al 2005)

Many previous studies citing the impact of the networkspecifically confluences on river characteristics were con-ducted in highly erosive relatively natural environments(Network Dynamics Hypothesis Benda et al 2004b LinkDiscontinuity Concept Rice et al 2001) Therefore it willbe interesting to assess the impact of network structure onriver characteristics in catchments in England a landscapethat has undergone modification impacting catchment func-tioning for centuries (Macklin and Lewin 2003)

2 Methods

21 Study sites

Four catchments are selected for testing the impact of net-work topology on river characteristics in England The catch-ments are from the Demonstration Test Catchment pro-gramme (Fig 2) which are representative of 80 of soil andrainfall combinations in the UK (McGonigle et al 2014)This demonstrates the potential use of topological metricsfor catchments with varying geologies and land uses

The Avon and Wensum catchments have similar charac-teristics both being dominated by chalk geology with loweraverage annual rainfall and a high percentage of arable farm-ing land cover In comparison the Eden and Tamar are dom-inated by less permeable bedrock with higher average annualrainfall and a high percentage of grassland land covers Interms of their morphometry the Avon and Wensum both havean elongated shape and low drainage density The Wensum

has the lowest relief with a maximum elevation of 95 m TheTamar has the smallest catchment area (928 km2) and is themost circular The Eden is the largest catchment (2295 km2)and has the highest maximum elevation (246 m)

22 Network topology metrics

Network topology metrics were calculated for each catch-ment using the 1 50000 river network map derived fromboth a digital terrain model (DTM) and Ordnance Sur-vey data (Moore et al 1994) Anabranches and incorrectlydigitized links in the network are identified using RivEX(Hornby 2010) and removed Removing anabranches wasnecessary as the topological metrics were designed for den-dritic networks so multi-thread channels either naturally oc-curring or artificial ditches would distort the calculationsThis resulted in a total of 448 2812 1516 and 532 links inthe Avon Eden Tamar and Wensum respectively

Elevation data were extracted from the Integrated Hydro-logical DTM (Morris and Flavin 1994) a 50times50 m griddedelevation raster with a 10 cm vertical resolution Average el-evation of each link and the distance from each link to thenetwork outlet were extracted using RivEX (Hornby 2010)

To extract a measure of network density that varies spa-tially within the catchment each network is divided into20 intervals each of which represents 5 of the total dis-tance or highest elevation in the network (Fig 2) The net-work is divided in this manner based on the methods of theNWF and LCF which have functional application to hydro-graph prediction Twenty intervals provide a relatively coarsesampling of the network compared to the 100 intervals de-scribed by Stepinski and Stepinski (2005) when they adapteda morphometric variable circularity ratio to represent inter-nal catchment elongation Here a total of 20 intervals is cho-sen so that most intervals contain links for the density cal-culation whilst ensuring the spatial distribution of networkdensity within the catchment is characterized

Distance network density was calculated followingEq (1)

Distance network density=

[n(d0) n(di) n(dN)

](dNtimes 005)

(1)

where the number of links (n( )) within each 5 distanceinterval (di) from the outlet (d0) to the maximum distance inthe network (dN) is normalized by the width of the interval(dNtimes 005)

Elevation network density was calculated followingEq (2)

Elevation network density=

[n(z0) n(zi) n(zN)

](zNtimes 005)

(2)

where the number of links (n( )) within each 5 elevationinterval (zi) from the outlet (z0) to the maximum height ofthe network (zN) is normalized by the width of the inter-val (zNtimes 005) Normalization allows network densities to

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2309

Figure 2 Distance and elevation intervals for each Demonstration Test Catchment Avon (A) Eden (E) Tamar (T) and Wensum (W)(a) Percentage distance intervals used to calculate distance network density (b) Percentage elevation intervals used to calculate elevationnetwork density Map of catchment locations in England in the bottom-right corner

be compared between catchments controlling for differencesin size and elevation as well as within catchments

To assess the utility of the multi-dimensional topologymetrics in accounting for the spatial structure of the networkthe metrics are compared to the one-dimensional Strahlerstream-order metric extracted from the river network datasetusing RivEX (Hornby 2010)

23 River characteristics

The impact of network topology on channel functioning isexplored using a broad-scale approach ie adapting data col-lected for regulatory compliance to answer scientific ques-tions Adapting such datasets to scientific enquiry allowsanalysis to be conducted in many catchments across a widespatial extent There are many habitat monitoring methodsacross the globe with 121 survey methods recorded in over

26 different countries (Belletti et al 2015) so this methodmay be adapted to other countries

This study utilizes the River Habitat Survey (RHS Ravenet al 1996) a regulatory dataset collected by Englandrsquos En-vironment Agency which is used to reflect the river reachcharacteristics in each catchment This dataset has been usedto identify catchment effects on river characteristics in broad-scale studies by previous research (eg Harvey et al 2008Naura et al 2016 Vaughan et al 2013) but none have in-cluded the effects of network topology

Since 1994 over 24 000 sites have been sampled incatchments across England and Wales including the Avon(n= 418) Eden (n= 398) Tamar (n= 189) and Wensum(n=315) Surveys were conducted at random sites withineach 10 km2 of England and Wales to ensure geographic cov-erage however this produces sampling bias as streams inhigh-density areas will be under-represented in the datasetwhich is acknowledged in this study and discussed below

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2310 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

Table 1 RHS variables calculated from RHS observations

RHS variable Calculation from RHS observations Units

Habitat Quality A score indicating the degree of naturalness and diversity of the HQA scaleAssessment riparian zone based on observations in the reach of flow types(HQA) substrate channel and bank features riparian vegetation etc

Habitat A score indicating the degree of artificial modification of the channel HMS scaleModification based on observations in the reach of reinforcements re-sectioningScore (HMS) embankments weed cutting realignment culverts dams weirs etc

Sediment size =(minus8middotBOminus7middotCOminus35middotGPminus15middotSA+15middotSI+9middotCL)

(BO+CO+GP+SA+SI+CL) ApproxBO (boulder) CO (cobble) GP (gravelndashpebble) SA (sand) SI (silt) φ unitsand CL (clay) represent the number of spot checks allocated to eachsediment size class

Flow-type speed =(0middotDR+1middotNP+2middotUP+3middotSM+4middotRP+5middotUW+6middotBW+7middotCF+8middotCH+9middotFF)

(DR+NP+UP+SM+RP+UW+BW+CF+CH+FF) Flow typeDR (dry) NP (no perceptible flow) UP (upwelling) SM (smooth) speed scaleRP (rippled) UW (unbroken wave) BW (broken wave) CF (chaoticflow) CH (chute) FF (free-fall) represent the number of spot checksallocated to each flow speed class

At each site over 100 features are recorded along a 500 mreach with 10 ldquospot-checkrdquo surveys conducted every 50 mand a ldquosweep-uprdquo survey conducted across the whole reach(see Raven et al 1996 for details) Variables of interest thatare hypothesized to be impacted by network structure can becalculated from the RHS observations (Table 1)

The Habitat Quality Assessment (HQA) and Habitat Mod-ification Score (HMS) variables are both amalgamations ofRHS observations with individual features given a score de-rived by expert opinion (see Raven et al 1998 for more de-tails) The scoring systems are subjective but HQA and HMSprovide overviews of the channel condition that are widelyapplied for regulatory compliance The scores are thereforeincluded in this study to reflect how they may be impactedby network topology

The remaining RHS variables are calculated directly fromRHS observations and so are more objective Sediment sizeis calculated as a reach average of spot-check observationsusing the same method as previous studies (Davenport et al2004 Emery et al 2004 Harvey et al 2008) Flow-typespeed was calculated in the same manner as sediment sizeusing values of flow which represent an approximate flowvelocity gradient defined in Davenport et al (2004) Thesevariables were chosen to reflect dominant geomorphic pro-cesses occurring in each reach and due to the prominence ofsediment size and flow type in defining physical habitats forinstream biota (Rowntree and Wadeson 1996) The variablesare likely to be impacted by the density of the river networkas they have been shown to be impacted by individual con-fluences For example channels are shown to become moregeomorphologically heterogeneous (Benda et al 2004a) andsubstrate size has been shown to coarsen at confluences (Riceet al 2001) Surface flow types are also likely to become

more diverse at confluences as the convergence of channelscreates a number of different flow environments (Best 1985)that result in different water-surface topographies (Biron etal 2002)

It must be noted that the RHS dataset was collected forregulatory compliance and was not directly intended for sci-entific enquiry Therefore there is a limitation in the amountof detail that can be extracted about physical processes as theobservations recorded are an average across a 500 m reachDespite this the wide spatial coverage of the dataset makesit a powerful tool allowing analysis to be conducted acrossmultiple catchments with differing characteristics

For each distance and elevation interval created by the net-work topology metrics descriptive statistics (mean median90th and 10th percentiles) of each RHS variable are calcu-lated Despite the RHS sampling strategy (Jeffers 1998b) bi-asing site selection towards less dense areas of the networkmost distance and elevation intervals contained RHS sites(with only some low-density intervals not containing RHSsites) This method is designed to account for natural varia-tion and modification at individual RHS sites in order to as-sess broad patterns of reach characteristics at the catchmentlevel

24 Statistical analysis

Analysis is conducted with all catchments combined into asingle population to identify overall trends across all catch-ments a method used in previous broad-scale studies Theanalysis is also split into individual catchments to identifyhow the relationship between network topology metrics andriver reach characteristics differed between catchments

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2311

Kendall correlation

Correlation tests are used to determine the strength and di-rection of the association between the descriptive statisticsof the RHS variables and distance network density eleva-tion network density and stream order to ascertain how reachcharacteristics respond to network topology Kendallrsquos cor-relation method was used as the variables have non-normaldistributions a small sample size and tied data values (Helseland Hirsch 2002) The effect size of Kendallrsquos τ is lowerthan other correlation methods with strong correlations oc-curring with τ values greater than 07 (Helsel and Hirsch2002)

As multiple correlations are conducted false discoveryrate (Benjamini and Hochberg 1995) corrections were ap-plied to the p-values produced from the Kendall correlationsto reduce the risk of type I error The false discovery ratemethod has been found to be more powerful than other pro-cedures for controlling for multiple tests (Glickman et al2014)

3 Results

31 Differences in network topology metrics betweencatchments

The topological metrics developed in this study show the in-ternal structure of the network for each catchment The sep-aration of the catchments into distance and elevation inter-vals emphasizes different features of the catchment The dis-tance intervals (Fig 2a) are arranged longitudinally withinthe catchment highlighting sub-basins within each catch-ment The elevation intervals (Fig 2b) have a radial arrange-ment centering around the incised main channel of eachcatchment

Distance network density is higher in the Eden (284plusmn103) and Tamar (441plusmn 219) compared to the Avon (47plusmn19) and Wensum (68plusmn 07) interesting as the Tamar is thesmallest catchment by area The shape of the distance net-work density function reflects the internal shape of the net-work (Fig 3a) For example the Tamar has a peaked den-sity distribution reflecting the circular shape of the catch-ment such that the majority of links are at 55 ndash65 dis-tance from the catchment outlet The Avon and Eden reflectsimilar internal network structures both exhibiting a bimodaldensity distribution despite the differences in the number oflinks in the catchments The density distribution of the Wen-sum has a more complex internal distribution of links withmultiple peaks in density

Elevation network density (Fig 3b) shows similar densitydistribution shapes to distance network density with a uni-modal distribution for the Tamar and multi-modal distribu-tions in the other catchments In contrast to distance networkdensity elevation network density shows the highest peaks in

density in the Tamar (103plusmn 50) and Wensum (101plusmn 43)despite the Wensum having the lowest network elevation andhas lower values in the Avon (34plusmn10) and Eden (58plusmn25)The peak densities in the Wensum occur at similar positionsin the elevation and distance intervals whereas the peaks inthe other catchments are negatively skewed showing the net-work density is highest at lowndashmid elevations

Nearly half of the links in each catchment are classified asfirst-order streams and the number of links declines exponen-tially towards the highest orders in all four catchments Thereare weak correlations (τ =minus003 to 017) between the threenetwork topology metrics distance network density eleva-tion network density and stream order This suggests that themetrics are independent and reflect different aspects of rivernetwork topology

32 River characteristic relationships with networktopology metrics

RHS sites in the Avon and Wensum have similar river char-acteristics Both have lower habitat quality high modifica-tion fine sediment and slower flow types than the Eden andTamar When all catchments are combined there are signif-icant (p lt 005 after p-value correction) correlations withmost descriptive statistics for each RHS variable and dis-tance network density (Fig 4) There are consistently pos-itive correlations with HQA and flow-type speed and neg-ative correlations with HMS and sediment size There arefewer and weaker significant correlations with elevation net-work density (Fig 4) The only significant correlations withstream order are with HMS which shows a negative correla-tion (Fig 4)

There are numerous significant correlations between thenetwork topology metrics and RHS variables for individualcatchments many of which were also shown to be signifi-cant after the correction to the p-value The results show thatcatchments have different responses to the network topol-ogy metrics of distance network density and elevation net-work density Distance network density only has significantcorrelations with the regulatory scoring variables (HQA andHMS) in the Eden and Tamar (Fig 4) Elevation networkdensity however has a wider array of significant correlationswith the scoring variables particularly HQA which showssubtle peaks and troughs reflecting the distribution of bothnetwork density metrics (Fig 3a and b) HMS shows mostlynegative correlations mainly with elevation network densityapart from the Eden which has significant positive correla-tion across all HMS descriptive statistics (Fig 4) Visually10th percentile HMS is most variable to network density withpeaks and troughs responding to the network density distri-butions (Fig 3a and b)

For individual RHS features the response to networktopology varies between catchments (Fig 4) The Avon hasnegative correlations between flow-type speed and distancenetwork density with an evident drop in 10th percentile

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2312 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

Figure 3 Network topology metrics (a) distance network density and (b) elevation network density Descriptive statistics of each RHSvariable over (a) distance and (b) elevation for each catchment with smooth loess lines to indicate trend Network topology metrics arenormalized between 0 and 1 and HMS score is logarithmically transformed for display purposes

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2313

Figure 4 Summary of correlations between distance network density elevation network density stream order and RHS variables for allcatchments combined and each individual catchment Significance of correlation is indicated by point size with the largest points significantpost p-value correction Effect size (Kendallrsquos τ ) is indicated by colour No correlation was possible between stream order and 10th percentilesediment size in the Avon due to no variation in the RHS variable

flow-type speed associated with peaks in network density(Fig 3a) The Eden and Tamar however have positive cor-relations with mean and 90th percentile flow-type speed fordistance network density but negative correlations with me-dian and 10th percentile elevation network density The Wen-sum shows positive correlations between flow-type speed andelevation network density Sediment size has a consistent re-sponse to distance network density with the Eden and Wen-sum showing negative correlations with the sediment size(Fig 4) For elevation network density the Avon shows neg-ative correlations with sediment size whereas the Tamar andWensum show positive correlations (Fig 4)

There were few significant correlations between stream or-der and the RHS variables in individual catchments (Fig 4)The only significant correlation after p-value correction iswith 90th percentile HMS in the Wensum which shows astrong negative relationship

4 Discussion

41 A new approach to utilizing network topology incatchment-level analysis

Network density metrics represent an alternative approach toaccount for network topology in catchment-level studies op-timizing the width dimension of the network (or the num-

ber of links in the network) as opposed to the commonplacestream-order metric which only reflects the longitudinal po-sition of links in a network (Fig 1) This study demonstratesthat two topology metrics can be calculated simply from aDTM with GIS and using a broad-scale analysis of river at-tributes can be used to investigate the functional processeswithin catchments

While the two network density metrics have similar forms(ie forms are consistently unimodal or multi-modal) thespatial configuration of the distance and elevation intervalsused in the calculation of network density varies and mayimpact the effectiveness of each topological metric Distancenetwork density separates the catchment into intervals basedon distance which spread upstream from the outlet (Fig 2a)reflecting natural sub-basins within the fractal structure of thecatchment (Lashermes and Foufoula-Georgiou 2007) Thisdiffers from elevation network density which separates thecatchment into intervals based on elevation which radiate outfrom the main channel of the network (Fig 2b) The config-uration means that distance intervals contain streams that arein closer proximity to one another rather than the more distalconfiguration created by the elevation intervals suggesting adegree of spatial dependency in river functioning This hasbeen highlighted in previous studies where spatial networkstructure has a stronger influence on some in-channel pro-cesses than predictor variables such as elevation (Steel et al

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2314 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

2016) However elevation intervals contain RHS sites thatalthough they may be distal may have similar properties aselevation has been strongly related to RHS variables includ-ing flow type substrate etc in a number of studies (Jeffers1998a Naura et al 2016 Vaughan et al 2013)

42 Impacts of network topology on rivercharacteristics

River characteristics are assessed using the RHS dataset Theobservations made by the RHS dataset (Table 1) cannot of-fer the level of detail regarding geomorphological processthat river classifications which consider multiple scales canoffer (eg Brierley and Fryirs 2000 Gurnell et al 2016)While process-based classifications are preferable broad-scale monitoring datasets such as the River Habitat Surveymay still be useful when combined with map-derived datato explore controls on river characteristics (Harvey et al2008 Naura et al 2016 Vaughan et al 2013) Howeverthere are biases in RHS data collection an inherent limitationwhen using existing datasets (Vaughan and Ormerod 2010)specifically the standardized survey length of 500 m reachthat will capture differing amounts of natural variability de-pending on the size of the river While this must be notedthere are few significant correlations between river charac-teristics identified with stream order (Fig 4) which suggeststhat channel size is not influencing the RHS variables to agreat degree in these catchments

In this study it is anticipated that sites in high networkdensity areas will have higher levels of habitat diversity asindicated by previous studies of confluences and networks(Benda et al 2004a Best 1985 Rice 2017) in turn in-creasing mean sediment size and flow-type speed comparedto sites in low-density areas The results of the correlationsbetween distance network density and river characteristicswhen all catchments are combined support this hypothesiswith greater HQA flow-type speed and coarser sedimentsizes observed in areas with high distance network density(Fig 4)

For individual catchments elevation network density in-duces a stronger positive HQA response across all catch-ments than distance network density (Fig 4) This supportsthe evidence that individual confluences (Rice et al 2006)and high densities of confluences increase physical hetero-geneity within the river network (Benda et al 2004b Rice2017) However flow-type speed and sediment size responddifferently to network density in individual catchments

Slower flow types are observed in high network densityareas of the Avon and Tamar whereas faster flow types areobserved in high-elevation network density areas of the Wen-sum Individual confluences are shown to create numeroushigh- and low-speed flow environments (Best 1987) thatmay be observed in surface water topography (Biron et al2002) It was expected that the introduction of the additionalflow types by a high density of confluences in relatively low-

energy rivers would increase mean reach flow-type speedhowever the correlation analysis (Fig 4) suggests that insome catchments mean flow-type speed is reduced

Sediment size response also shows variation betweencatchments Sediment size is coarser in network-dense ar-eas of the Avon and Eden as expected but is finer in boththe Tamar and Wensum (Fig 4) The evidence from high-energy rivers shows that sediment becomes coarser down-stream and finer upstream of certain confluences (Benda etal 2004a Rice 1998) and in this case high numbers of con-fluences were expected to increase mean sediment size of thereach However the impact on sediment size is dependent onthe sediment calibre of the incoming tributary being higherthan the main channel with enough energy to transport thecoarse sediment to the confluence for numerous tributariesin an area The Tamar and Wensum have different rangesof sediment sizes with the Tamar having coarser sedimentthan the Wensum on average (Fig 3) This implies that trib-utaries in the Tamar may be energy limited not transportingcoarse sediments to confluences and the Wensum may suf-fer from inputs of fine sediments from the high percentageof arable land that concentrates in network-dense areas Thishas before been observed in a low-energy modified catch-ment where anthropogenic modifications in tributaries re-duced coarse sediment and flow capacity causing either lim-ited confluence impact or localized sediment fining (Singer2008)

Others have related the capacity of confluences to alterreach features to the morphometry of catchments with largerand more circular catchments containing a higher percentageof confluences that have a significant impact (Rice 2017)Based on this theory the Eden and Tamar are likely to havethe greatest impact as they are the most circular and steepestof the four catchments (although the Tamar is the smallestby area) Yet there is no clear pattern indicating that thesecatchments respond differently than the Avon and Wensum(Fig 4) with catchments responding differently to differ-ent variables This perhaps suggests that rather than networkdensity having a directional impact on factors such as flow-type speed and sediment size it has an impact on overall het-erogeneity at the reach level as suggested by previous stud-ies and that specific directional change occurs at the sub-reach level

An increase in channel modification is also hypothesizeddue to the increase in flood peak downstream from conflu-ences (Depettris et al 2000) and the scour and erosion as-sociated with confluence junctions (Best 1986) potentiallyincreasing the need for bed and bank protection The corre-lations between distance network density and HMS when allcatchments are combined undermine this hypothesis show-ing less channel modification where distance network den-sity is higher (Fig 4) There were few significant correlationswith individual catchments but HMS in the Eden was consis-tently observed to be higher in network-dense areas (Fig 4)This may be due to the Edenrsquos high elevation and steep to-

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2315

pography inducing a higher energy environment where scourand erosion processes in areas with high numbers of conflu-ences would be more likely to be present

Differences in RHS variable responses also differed be-tween the descriptive statistics considered Often the ex-tremes 90th and 10th percentiles showed more significantand stronger correlations than the mean or median (Fig 4)This may reflect findings from previous studies which sug-gest that not all confluences cause reach-scale changes (Rice1998) that perhaps the changes to river character induced bycertain confluences only influence certain reaches whereasothers are left unaffected creating less pronounced responsesin the mean and median of variables External factors mayalso influence this trend for example 10th percentile HMSvisually responds dramatically to network density metrics(Fig 3a and b) compared to the other descriptive statisticsThis suggests that the most natural sites (ie with the low-est HMS score) are responding differently to network den-sity with the most natural sites having less modificationin network-dense areas whereas less natural sites becomemore modified in network-dense areas This reflects the HQAscore results which visually (Fig 3a and b) and statistically(Fig 4) vary with distance network density except for the10th percentile These sites with the lowest habitat qualityand naturalness are likely influenced by anthropogenic fac-tors that are independent of network density reducing habitatquality scores at impacted sites

The two network density metrics have differing impacts onriver characteristics While distance network density showsconsistently significant correlations when all catchments arecombined individual catchments respond more frequentlyand more strongly to elevation network density (Fig 4)This may be because there is a dramatic split in distancenetwork density values between the more upland drainage-dense catchments Eden and Tamar than the lowland chalklow-drainage catchments Avon and Wensum The combinedcorrelation will therefore in part reflect the difference be-tween the catchments which have different ranges of rivercharacteristics (Fig 3) This is not the case for elevation net-work density which has higher density values in the Tamarand Wensum than the Avon and Eden so therefore the char-acteristics of the catchments will have less bearing on thecombined correlation However there are patterns identifiedwith distance network density in individual catchments thatare not present with elevation network density increasedflow-type speed and sediment size in the Eden and reducedflow-type speed in the Avon with network density (Fig 4)which show its usefulness

The results suggest that the distance network density andelevation network density metrics quantify different dimen-sions of network topology which are shown to exhibit func-tionally meaningful patterns for river reach characteristicsbased on the correlation results Perhaps within catchmentselevation network density provides the more powerful met-ric for individual catchments but distance network density

better accounts for the drainage density of the catchmentsallowing it to be applied across multiple catchments

43 Comparison of stream order to network densitymetrics

The classic method of accounting for network topologystream order is critiqued for failing to represent disconti-nuities in the network and simplifying the network into agradient The number of links in different stream orders isconsistent across all catchments not reflecting the internalstructure of the network or the variety between catchmentsthat is achieved by the distance network density and eleva-tion network density metrics The analysis of the two net-work density metrics presented in this paper shows that dis-tances from source and elevation are not mutually exclusive(Fig 2a and b) contrary to the stream-order metric whichrepresents streams as upstream to downstream or upland tolowland

Stream order has few significant correlations with manyof the river characteristics considered in this study Negativecorrelations with HMS in the Wensum were statistically sig-nificant post p-value correction likely driving the significantrelationship with all catchments combined for this variable(Fig 4) This suggests that modification is greater upstreamin the Wensum contrary to ideas that downstream reachesmay show greater anthropogenic modification Intense agri-cultural land use in the upper reaches of the Wensum islikely to be the cause of the high HMS scores upstreamThe lack of significant correlations suggests that stream or-der and therefore an upstream-to-downstream gradient arenot the predominant pattern in river characteristics despitethe description of such a gradient by geomorphic (Schumm1977) and ecological frameworks (Vannote et al 1980)This is surprising as distance and elevation which both re-flect the upstream-to-downstream gradient have proven tobe important factors in previous studies explaining patternsof RHS features at a national level (Jeffers 1998a Vaughanet al 2013) This implies that upstream-to-downstream gra-dient may not sufficiently reflect patterns of river charac-teristics through the river network within individual catch-ments Others have also found that stream order has weakand inconsistent relationships with biodiversity patterns inriver systems arguing that the topological measure has no di-rect mechanistic control on biodiversity (Vander Vorste et al2017) Instead this study finds that the network density met-rics are a more powerful metric which conceptually providean improved method of accounting for the impacts of net-work topology on the fluvial system exhibiting relationshipswith river characteristics particularly habitat quality score(Fig 4)

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2316 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

44 Applicability of network topology metrics todifferent environments

Much of the seminal work on network and confluence im-pacts (eg the Network Dynamics Hypothesis Benda et al2004b and Link Discontinuity Concept Rice et al 2001) isconducted in natural highly erosive catchments with first-hand empirical measurements However in an age whenrivers and their catchments are increasingly altered by an-thropogenic modification (Meybeck 2003) contemporarystudies must not only aim to expand knowledge but also findmethods of transferring knowledge to many increasingly al-tered catchments (Clifford 2002)

The catchments selected by this study are more greatlymodified and although they reflect a range of fluvial envi-ronments in England are lower-energy catchments than theseminal studies Benda et al (2004a) suggest that confluenceeffects in less active landscapes would be subdued comparedto highly erosive landscapes but the evidence presented heredemonstrates the utility of evaluating network topologicalstructure in studies on catchment-level effects in any typeof fluvial environment including those with low-energy andwidespread anthropogenic changes The response of someriver characteristics varied between catchments observationsof flow-type speed sediment size and modifications showeddifferent responses to network density in different catch-ments This suggests that the functional effect of these topo-logical metrics is catchment dependent and likely is influ-enced by external catchment characteristics such as land usenot considered in this study although impact did not appearto vary with catchment topography or circularity as has beenshown in prior studies (Benda et al 2004a Rice 2017) Thisshould be explored further in future research to enable rec-ommendations to be made regarding where and how reachesmay respond to network density The response of habitatquality score was however consistent across catchments andbetween metrics showing that habitat quality is greater inareas with high network density (Fig 4) as hypothesizedby the Network Dynamics Hypothesis (Benda et al 2004b)and demonstrated by studies on individual catchments (Rice2017 Rice et al 2006)

The methods presented in this paper are designed to be im-plemented in any catchment with a dendritic network struc-ture The topology metrics can easily be calculated from anydendritic network with DTM data using GIS and comparedto any site-scale data This study uses regulatory monitoringdatasets so that analysis is targeted to assessment scores andphysical features of interest to river managers Also the highvolume and wide spatial extent of data available from regu-latory sources allow for between-catchment comparisons

5 Conclusions

Although appreciation of catchment-level effects is consid-ered the epitome of understanding river functioning a keycomponent of the catchment ndash the river network ndash is over-looked and oversimplified by catchment-level studies Thisstudy finds that river network density plays a role in struc-turing the distribution river characteristics throughout thecatchment offering more detailed explanation than the clas-sic stream-order metric Network-dense areas are generallyfound to have higher habitat quality and diversity but modi-fication flow-type speed and sediment size show different re-sponses in different catchments This study suggests two po-tential reasons for this (1) there is evidence that confluencesin the river network increase diversity as is observed in thisstudy so the direction of mean river characteristic responsemay not be consistent and (2) there may be external factorssuch as sediment availability land cover and anthropogenicmodification that alter the direction of mean river characteris-tic response This paper demonstrates the functional responseof river characteristics to network topology and suggests thatthe inclusion of network topology in catchment-level studieswould add a layer of function-based understanding to suchstudies linking reaches to their catchments

The broad-scale methodology adopted by this study allowsthe network density metrics which are easily extracted fromopen-source data using GIS software to be compared to anyregulatory dataset The use of regulatory datasets allows notonly for analysis over a wider spatial extent but also for moreapplicable results for regulatory bodies Therefore the inter-disciplinary approach to characterizing network topology canbe applied efficiently and effectively to capture catchment-level impacts on reach-level functioning in any catchmentacross the globe

Data availability River Habitat Survey data are freely availablefrom the Environment Agency England River network data andelevation data are available from the Centre for Ecology and Hy-drology UK

Author contributions EH and NC conceptualised and designed thestudy EH performed the data analysis with the support of JMEH wrote the paper with comments provided by NC and JM

Competing interests The authors declare that they have no conflictof interest

Acknowledgements The authors would like to thank the Environ-ment Agency and the Centre for Ecology and Hydrology for accessto the data used in this paper This research is funded by the NaturalEnvironmental Research Council (NERC)

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2317

Review statement This paper was edited by Nandita Basu and re-viewed by Simone Bizzi and one anonymous referee

References

Belletti B Rinaldi M Buijse A D Gurnell A Mand Mosselman E A review of assessment methods forriver hydromorphology Environ Earth Sci 73 2079ndash2100httpsdoiorg101007s12665-014-3558-1 2015

Benda L Andras K Miller D and Bigelow P Confluenceeffects in rivers Interactions of basin scale network geom-etry and disturbance regimes Water Resour Res 40 1ndash15httpsdoiorg1010292003WR002583 2004a

Benda L Poff N L Miller D Dunne T Reeves GPess G and Pollock M The network dynamics hy-pothesis how channel networks structure riverine habi-tats Bioscience 54 413ndash427 httpsdoiorg1016410006-3568(2004)054[0413TNDHHC]20CO2 2004b

Benjamini Y and Hochberg Y Controlling the False DiscoveryRate A Practical and Powerful Approach to Multiple TestingJ Roy Stat Soc B 57 289ndash300 1995

Best J L Flow dynamics and sediment transport at river channelconfluences Birbeck University of London London 1985

Best J L The morphology of river channel con-fluences Prog Phys Geogr 10 157ndash174httpsdoiorg101177030913338601000201 1986

Best J L Flow dynamics at river channel confluences implica-tions for sediment transport and bed morphology in Recent de-velopments in fluvial sedimentology edited by Ethridge F GFlores R M and Harvey M D Spec Publ Soc Econ Pale-ontol Miner Tulsa Okla 27ndash35 1987

Biron P M Richer A Kirkbride A D Roy A G andHan S Spatial patterns of water surface topography ata river confluence Earth Surf Proc Land 27 913ndash928httpsdoiorg101002esp359 2002

Bravard J P and Gilvear D J Hydrological and geomorpho-logical structure of hydrosystems in The Fluvial Hydrosystemedited by Petts G E and Amoros C Springer Netherlands98ndash116 1996

Brierley G and Fryirs K River Styles a Geomorphic Approachto Catchment Characterization Implications for River Rehabili-tation in Bega Catchment New South Wales Australia EnvironManage 25 661ndash679 2000

Bunn S E and Arthington A H Basic principles and ecologi-cal consequences of altered flow regimes for aquatic biodiversityEnviron Manage 30 492ndash507 httpsdoiorg101007s00267-002-2737-0 2002

Church M and Kellerhals R On the statistics of grain size varia-tion along a gravel river Can J Earth Sci 15 1151ndash1160 1978

Clifford N J Hydrology the changing paradigm Prog Phys Ge-ogr 26 290ndash301 2002

Cohen P Andriamahefa H and Wasson J-G Towardsa regionalizaton of aquatic habitat distribution of meso-habitats at the scale of a large basin Regul RiversRes Manage 14 391ndash404 httpsdoiorg101002(SICI)1099-1646(19980910)145lt391AID-RRR513gt30CO2-W 1998

Davenport A J Gurnell A M and Armitage P D Habitat sur-vey and classification of urban rivers River Res Appl 20 687ndash704 httpsdoiorg101002rra785 2004

Davies N M Norris R H and Thoms M C Prediction and as-sessment of local stream habitat features using large-scale catch-ment characteristics Freshwater Biol 45 343ndash369 2000

Depettris C Mendiondo E M Neiff J and Rohrmann H Flooddefence strategy at the confluence of the Paranaacute-Paraguay riversProc Int Symp Flood Def Kassel He C31 C40 2000

Dollar E James C Rogers K and Thoms M A framework forinterdisciplinary understanding of rivers as ecosystems Geomor-phology 89 147ndash162 2007

Dovers S R and Day D G Australian rivers and statute law En-viron Plan Law J 5 90ndash108 1988

Downs P W Gregory K J and Brookes A How integratedis river basin management Environ Manage 15 299ndash309httpsdoiorg101007BF02393876 1991

Emery J C Gurnell A M Clifford N J and Petts G E Char-acteristics and controls of gravel-bed riffles An analysis of datafrom the river-habitat survey Water Environ J 18 210ndash216httpsdoiorg101111j1747-65932004tb00535x 2004

Evans I S and Minaacuter J A classification of geomorphometric vari-ables in International Geom-orphometry 2011 Geomoprhome-tryorg Redlabds CA 105ndash108 2011

Fausch K D Torgersen C E Baxter C V Li H WView C The O F Is R To N How U Among ISet S For C and Fishes S Landscapes to riverscapesBridging the gap between research and conservation of streamfishes Bioscience 52 483ndash498 httpsdoiorg1016410006-3568(2002)052[0483LTRBTG]20CO2 2002

Glickman M E Rao S R and Schultz M R False discov-ery rate control is a recommended alternative to Bonferroni-typeadjustments in health studies J Clin Epidemiol 67 850ndash857httpsdoiorg101016jjclinepi201403012 2014

Gupta V K and Mesa O J Runoff generation and hydro-logic response via channel network geomorphology - Re-cent progress and open problems J Hydrol 102 3ndash28httpsdoiorg1010160022-1694(88)90089-3 1988

Gupta V K and Waymire E D On the formulation of an analyt-ical approach to hydrologic response and similarity at the basinscale J Hydrol 65 95ndash123 1983

Gupta V K Waymire E D and Rodriguez-Iturbe I On scalesgravity and network structure in basin runoff in Scale problemsin hydrology Springer Netherlands 159ndash184 1986

Gurnell A M Rinaldi M Belletti B Bizzi S Blamauer BBraca G Buijse A D Bussettini M Camenen B ComitiF Demarchi L Garciacutea de Jaloacuten D Gonzaacutelez del TaacutenagoM Grabowski R C Gunn I D M Habersack H Hen-driks D Henshaw A J Kloumlsch M Lastoria B LatapieA Marcinkowski P Martiacutenez-Fernaacutendez V Mosselman EMountford J O Nardi L Okruszko T OrsquoHare M T PalmaM Percopo C Surian N van de Bund W Weissteiner Cand Ziliani L A multi-scale hierarchical framework for devel-oping understanding of river behaviour to support river manage-ment Aquat Sci 78 1ndash16 httpsdoiorg101007s00027-015-0424-5 2016

Harvey G L Gurnell A M and Clifford N J Characterisationof river reaches The influence of rock type Catena 76 78ndash88httpsdoiorg101016jcatena200809010 2008

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2308 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

the greatest impact on channel morphology (Benda et al2004a) the greatest flow diversity (Schindfessel et al 2015)and the greatest fish community diversity (Osborne and Wi-ley 1992) In contrast others have found that tributaries thatdiffer most in size have the greatest impact For exampleJones and Schmidt (2016) suggest that high densities of smalltributaries flowing into a large channel cause small cumula-tive changes and Milesi and Melorsquos (2013) study concludedthat small tributaries flowing into large channels in the pe-ripheral regions of the catchment have the greatest impact onmacroinvertebrate assemblages

Interestingly there is little evidence of anthropogenic im-pacts at confluences in the literature but as confluences areproposed concentration points of catchment effects it seemslikely that they may be focal points for anthropogenic im-pacts For example flood events may occur downstream oflarge confluences as flood peaks converge creating the needfor flood defence measures (Depettris et al 2000) and scourand erosion at confluence junctions (Best 1986) increase theneed for bed and bank protection Also sediment size at con-fluences is shown to increase in many studies (Church andKellerhals 1978 Knighton 1980) but in tributaries whosewatersheds are dominated by agricultural land uses fine sed-iments may become dominant at confluences potentially al-tering river functioning (Owens et al 2005)

Many previous studies citing the impact of the networkspecifically confluences on river characteristics were con-ducted in highly erosive relatively natural environments(Network Dynamics Hypothesis Benda et al 2004b LinkDiscontinuity Concept Rice et al 2001) Therefore it willbe interesting to assess the impact of network structure onriver characteristics in catchments in England a landscapethat has undergone modification impacting catchment func-tioning for centuries (Macklin and Lewin 2003)

2 Methods

21 Study sites

Four catchments are selected for testing the impact of net-work topology on river characteristics in England The catch-ments are from the Demonstration Test Catchment pro-gramme (Fig 2) which are representative of 80 of soil andrainfall combinations in the UK (McGonigle et al 2014)This demonstrates the potential use of topological metricsfor catchments with varying geologies and land uses

The Avon and Wensum catchments have similar charac-teristics both being dominated by chalk geology with loweraverage annual rainfall and a high percentage of arable farm-ing land cover In comparison the Eden and Tamar are dom-inated by less permeable bedrock with higher average annualrainfall and a high percentage of grassland land covers Interms of their morphometry the Avon and Wensum both havean elongated shape and low drainage density The Wensum

has the lowest relief with a maximum elevation of 95 m TheTamar has the smallest catchment area (928 km2) and is themost circular The Eden is the largest catchment (2295 km2)and has the highest maximum elevation (246 m)

22 Network topology metrics

Network topology metrics were calculated for each catch-ment using the 1 50000 river network map derived fromboth a digital terrain model (DTM) and Ordnance Sur-vey data (Moore et al 1994) Anabranches and incorrectlydigitized links in the network are identified using RivEX(Hornby 2010) and removed Removing anabranches wasnecessary as the topological metrics were designed for den-dritic networks so multi-thread channels either naturally oc-curring or artificial ditches would distort the calculationsThis resulted in a total of 448 2812 1516 and 532 links inthe Avon Eden Tamar and Wensum respectively

Elevation data were extracted from the Integrated Hydro-logical DTM (Morris and Flavin 1994) a 50times50 m griddedelevation raster with a 10 cm vertical resolution Average el-evation of each link and the distance from each link to thenetwork outlet were extracted using RivEX (Hornby 2010)

To extract a measure of network density that varies spa-tially within the catchment each network is divided into20 intervals each of which represents 5 of the total dis-tance or highest elevation in the network (Fig 2) The net-work is divided in this manner based on the methods of theNWF and LCF which have functional application to hydro-graph prediction Twenty intervals provide a relatively coarsesampling of the network compared to the 100 intervals de-scribed by Stepinski and Stepinski (2005) when they adapteda morphometric variable circularity ratio to represent inter-nal catchment elongation Here a total of 20 intervals is cho-sen so that most intervals contain links for the density cal-culation whilst ensuring the spatial distribution of networkdensity within the catchment is characterized

Distance network density was calculated followingEq (1)

Distance network density=

[n(d0) n(di) n(dN)

](dNtimes 005)

(1)

where the number of links (n( )) within each 5 distanceinterval (di) from the outlet (d0) to the maximum distance inthe network (dN) is normalized by the width of the interval(dNtimes 005)

Elevation network density was calculated followingEq (2)

Elevation network density=

[n(z0) n(zi) n(zN)

](zNtimes 005)

(2)

where the number of links (n( )) within each 5 elevationinterval (zi) from the outlet (z0) to the maximum height ofthe network (zN) is normalized by the width of the inter-val (zNtimes 005) Normalization allows network densities to

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2309

Figure 2 Distance and elevation intervals for each Demonstration Test Catchment Avon (A) Eden (E) Tamar (T) and Wensum (W)(a) Percentage distance intervals used to calculate distance network density (b) Percentage elevation intervals used to calculate elevationnetwork density Map of catchment locations in England in the bottom-right corner

be compared between catchments controlling for differencesin size and elevation as well as within catchments

To assess the utility of the multi-dimensional topologymetrics in accounting for the spatial structure of the networkthe metrics are compared to the one-dimensional Strahlerstream-order metric extracted from the river network datasetusing RivEX (Hornby 2010)

23 River characteristics

The impact of network topology on channel functioning isexplored using a broad-scale approach ie adapting data col-lected for regulatory compliance to answer scientific ques-tions Adapting such datasets to scientific enquiry allowsanalysis to be conducted in many catchments across a widespatial extent There are many habitat monitoring methodsacross the globe with 121 survey methods recorded in over

26 different countries (Belletti et al 2015) so this methodmay be adapted to other countries

This study utilizes the River Habitat Survey (RHS Ravenet al 1996) a regulatory dataset collected by Englandrsquos En-vironment Agency which is used to reflect the river reachcharacteristics in each catchment This dataset has been usedto identify catchment effects on river characteristics in broad-scale studies by previous research (eg Harvey et al 2008Naura et al 2016 Vaughan et al 2013) but none have in-cluded the effects of network topology

Since 1994 over 24 000 sites have been sampled incatchments across England and Wales including the Avon(n= 418) Eden (n= 398) Tamar (n= 189) and Wensum(n=315) Surveys were conducted at random sites withineach 10 km2 of England and Wales to ensure geographic cov-erage however this produces sampling bias as streams inhigh-density areas will be under-represented in the datasetwhich is acknowledged in this study and discussed below

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2310 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

Table 1 RHS variables calculated from RHS observations

RHS variable Calculation from RHS observations Units

Habitat Quality A score indicating the degree of naturalness and diversity of the HQA scaleAssessment riparian zone based on observations in the reach of flow types(HQA) substrate channel and bank features riparian vegetation etc

Habitat A score indicating the degree of artificial modification of the channel HMS scaleModification based on observations in the reach of reinforcements re-sectioningScore (HMS) embankments weed cutting realignment culverts dams weirs etc

Sediment size =(minus8middotBOminus7middotCOminus35middotGPminus15middotSA+15middotSI+9middotCL)

(BO+CO+GP+SA+SI+CL) ApproxBO (boulder) CO (cobble) GP (gravelndashpebble) SA (sand) SI (silt) φ unitsand CL (clay) represent the number of spot checks allocated to eachsediment size class

Flow-type speed =(0middotDR+1middotNP+2middotUP+3middotSM+4middotRP+5middotUW+6middotBW+7middotCF+8middotCH+9middotFF)

(DR+NP+UP+SM+RP+UW+BW+CF+CH+FF) Flow typeDR (dry) NP (no perceptible flow) UP (upwelling) SM (smooth) speed scaleRP (rippled) UW (unbroken wave) BW (broken wave) CF (chaoticflow) CH (chute) FF (free-fall) represent the number of spot checksallocated to each flow speed class

At each site over 100 features are recorded along a 500 mreach with 10 ldquospot-checkrdquo surveys conducted every 50 mand a ldquosweep-uprdquo survey conducted across the whole reach(see Raven et al 1996 for details) Variables of interest thatare hypothesized to be impacted by network structure can becalculated from the RHS observations (Table 1)

The Habitat Quality Assessment (HQA) and Habitat Mod-ification Score (HMS) variables are both amalgamations ofRHS observations with individual features given a score de-rived by expert opinion (see Raven et al 1998 for more de-tails) The scoring systems are subjective but HQA and HMSprovide overviews of the channel condition that are widelyapplied for regulatory compliance The scores are thereforeincluded in this study to reflect how they may be impactedby network topology

The remaining RHS variables are calculated directly fromRHS observations and so are more objective Sediment sizeis calculated as a reach average of spot-check observationsusing the same method as previous studies (Davenport et al2004 Emery et al 2004 Harvey et al 2008) Flow-typespeed was calculated in the same manner as sediment sizeusing values of flow which represent an approximate flowvelocity gradient defined in Davenport et al (2004) Thesevariables were chosen to reflect dominant geomorphic pro-cesses occurring in each reach and due to the prominence ofsediment size and flow type in defining physical habitats forinstream biota (Rowntree and Wadeson 1996) The variablesare likely to be impacted by the density of the river networkas they have been shown to be impacted by individual con-fluences For example channels are shown to become moregeomorphologically heterogeneous (Benda et al 2004a) andsubstrate size has been shown to coarsen at confluences (Riceet al 2001) Surface flow types are also likely to become

more diverse at confluences as the convergence of channelscreates a number of different flow environments (Best 1985)that result in different water-surface topographies (Biron etal 2002)

It must be noted that the RHS dataset was collected forregulatory compliance and was not directly intended for sci-entific enquiry Therefore there is a limitation in the amountof detail that can be extracted about physical processes as theobservations recorded are an average across a 500 m reachDespite this the wide spatial coverage of the dataset makesit a powerful tool allowing analysis to be conducted acrossmultiple catchments with differing characteristics

For each distance and elevation interval created by the net-work topology metrics descriptive statistics (mean median90th and 10th percentiles) of each RHS variable are calcu-lated Despite the RHS sampling strategy (Jeffers 1998b) bi-asing site selection towards less dense areas of the networkmost distance and elevation intervals contained RHS sites(with only some low-density intervals not containing RHSsites) This method is designed to account for natural varia-tion and modification at individual RHS sites in order to as-sess broad patterns of reach characteristics at the catchmentlevel

24 Statistical analysis

Analysis is conducted with all catchments combined into asingle population to identify overall trends across all catch-ments a method used in previous broad-scale studies Theanalysis is also split into individual catchments to identifyhow the relationship between network topology metrics andriver reach characteristics differed between catchments

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2311

Kendall correlation

Correlation tests are used to determine the strength and di-rection of the association between the descriptive statisticsof the RHS variables and distance network density eleva-tion network density and stream order to ascertain how reachcharacteristics respond to network topology Kendallrsquos cor-relation method was used as the variables have non-normaldistributions a small sample size and tied data values (Helseland Hirsch 2002) The effect size of Kendallrsquos τ is lowerthan other correlation methods with strong correlations oc-curring with τ values greater than 07 (Helsel and Hirsch2002)

As multiple correlations are conducted false discoveryrate (Benjamini and Hochberg 1995) corrections were ap-plied to the p-values produced from the Kendall correlationsto reduce the risk of type I error The false discovery ratemethod has been found to be more powerful than other pro-cedures for controlling for multiple tests (Glickman et al2014)

3 Results

31 Differences in network topology metrics betweencatchments

The topological metrics developed in this study show the in-ternal structure of the network for each catchment The sep-aration of the catchments into distance and elevation inter-vals emphasizes different features of the catchment The dis-tance intervals (Fig 2a) are arranged longitudinally withinthe catchment highlighting sub-basins within each catch-ment The elevation intervals (Fig 2b) have a radial arrange-ment centering around the incised main channel of eachcatchment

Distance network density is higher in the Eden (284plusmn103) and Tamar (441plusmn 219) compared to the Avon (47plusmn19) and Wensum (68plusmn 07) interesting as the Tamar is thesmallest catchment by area The shape of the distance net-work density function reflects the internal shape of the net-work (Fig 3a) For example the Tamar has a peaked den-sity distribution reflecting the circular shape of the catch-ment such that the majority of links are at 55 ndash65 dis-tance from the catchment outlet The Avon and Eden reflectsimilar internal network structures both exhibiting a bimodaldensity distribution despite the differences in the number oflinks in the catchments The density distribution of the Wen-sum has a more complex internal distribution of links withmultiple peaks in density

Elevation network density (Fig 3b) shows similar densitydistribution shapes to distance network density with a uni-modal distribution for the Tamar and multi-modal distribu-tions in the other catchments In contrast to distance networkdensity elevation network density shows the highest peaks in

density in the Tamar (103plusmn 50) and Wensum (101plusmn 43)despite the Wensum having the lowest network elevation andhas lower values in the Avon (34plusmn10) and Eden (58plusmn25)The peak densities in the Wensum occur at similar positionsin the elevation and distance intervals whereas the peaks inthe other catchments are negatively skewed showing the net-work density is highest at lowndashmid elevations

Nearly half of the links in each catchment are classified asfirst-order streams and the number of links declines exponen-tially towards the highest orders in all four catchments Thereare weak correlations (τ =minus003 to 017) between the threenetwork topology metrics distance network density eleva-tion network density and stream order This suggests that themetrics are independent and reflect different aspects of rivernetwork topology

32 River characteristic relationships with networktopology metrics

RHS sites in the Avon and Wensum have similar river char-acteristics Both have lower habitat quality high modifica-tion fine sediment and slower flow types than the Eden andTamar When all catchments are combined there are signif-icant (p lt 005 after p-value correction) correlations withmost descriptive statistics for each RHS variable and dis-tance network density (Fig 4) There are consistently pos-itive correlations with HQA and flow-type speed and neg-ative correlations with HMS and sediment size There arefewer and weaker significant correlations with elevation net-work density (Fig 4) The only significant correlations withstream order are with HMS which shows a negative correla-tion (Fig 4)

There are numerous significant correlations between thenetwork topology metrics and RHS variables for individualcatchments many of which were also shown to be signifi-cant after the correction to the p-value The results show thatcatchments have different responses to the network topol-ogy metrics of distance network density and elevation net-work density Distance network density only has significantcorrelations with the regulatory scoring variables (HQA andHMS) in the Eden and Tamar (Fig 4) Elevation networkdensity however has a wider array of significant correlationswith the scoring variables particularly HQA which showssubtle peaks and troughs reflecting the distribution of bothnetwork density metrics (Fig 3a and b) HMS shows mostlynegative correlations mainly with elevation network densityapart from the Eden which has significant positive correla-tion across all HMS descriptive statistics (Fig 4) Visually10th percentile HMS is most variable to network density withpeaks and troughs responding to the network density distri-butions (Fig 3a and b)

For individual RHS features the response to networktopology varies between catchments (Fig 4) The Avon hasnegative correlations between flow-type speed and distancenetwork density with an evident drop in 10th percentile

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2312 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

Figure 3 Network topology metrics (a) distance network density and (b) elevation network density Descriptive statistics of each RHSvariable over (a) distance and (b) elevation for each catchment with smooth loess lines to indicate trend Network topology metrics arenormalized between 0 and 1 and HMS score is logarithmically transformed for display purposes

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2313

Figure 4 Summary of correlations between distance network density elevation network density stream order and RHS variables for allcatchments combined and each individual catchment Significance of correlation is indicated by point size with the largest points significantpost p-value correction Effect size (Kendallrsquos τ ) is indicated by colour No correlation was possible between stream order and 10th percentilesediment size in the Avon due to no variation in the RHS variable

flow-type speed associated with peaks in network density(Fig 3a) The Eden and Tamar however have positive cor-relations with mean and 90th percentile flow-type speed fordistance network density but negative correlations with me-dian and 10th percentile elevation network density The Wen-sum shows positive correlations between flow-type speed andelevation network density Sediment size has a consistent re-sponse to distance network density with the Eden and Wen-sum showing negative correlations with the sediment size(Fig 4) For elevation network density the Avon shows neg-ative correlations with sediment size whereas the Tamar andWensum show positive correlations (Fig 4)

There were few significant correlations between stream or-der and the RHS variables in individual catchments (Fig 4)The only significant correlation after p-value correction iswith 90th percentile HMS in the Wensum which shows astrong negative relationship

4 Discussion

41 A new approach to utilizing network topology incatchment-level analysis

Network density metrics represent an alternative approach toaccount for network topology in catchment-level studies op-timizing the width dimension of the network (or the num-

ber of links in the network) as opposed to the commonplacestream-order metric which only reflects the longitudinal po-sition of links in a network (Fig 1) This study demonstratesthat two topology metrics can be calculated simply from aDTM with GIS and using a broad-scale analysis of river at-tributes can be used to investigate the functional processeswithin catchments

While the two network density metrics have similar forms(ie forms are consistently unimodal or multi-modal) thespatial configuration of the distance and elevation intervalsused in the calculation of network density varies and mayimpact the effectiveness of each topological metric Distancenetwork density separates the catchment into intervals basedon distance which spread upstream from the outlet (Fig 2a)reflecting natural sub-basins within the fractal structure of thecatchment (Lashermes and Foufoula-Georgiou 2007) Thisdiffers from elevation network density which separates thecatchment into intervals based on elevation which radiate outfrom the main channel of the network (Fig 2b) The config-uration means that distance intervals contain streams that arein closer proximity to one another rather than the more distalconfiguration created by the elevation intervals suggesting adegree of spatial dependency in river functioning This hasbeen highlighted in previous studies where spatial networkstructure has a stronger influence on some in-channel pro-cesses than predictor variables such as elevation (Steel et al

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2314 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

2016) However elevation intervals contain RHS sites thatalthough they may be distal may have similar properties aselevation has been strongly related to RHS variables includ-ing flow type substrate etc in a number of studies (Jeffers1998a Naura et al 2016 Vaughan et al 2013)

42 Impacts of network topology on rivercharacteristics

River characteristics are assessed using the RHS dataset Theobservations made by the RHS dataset (Table 1) cannot of-fer the level of detail regarding geomorphological processthat river classifications which consider multiple scales canoffer (eg Brierley and Fryirs 2000 Gurnell et al 2016)While process-based classifications are preferable broad-scale monitoring datasets such as the River Habitat Surveymay still be useful when combined with map-derived datato explore controls on river characteristics (Harvey et al2008 Naura et al 2016 Vaughan et al 2013) Howeverthere are biases in RHS data collection an inherent limitationwhen using existing datasets (Vaughan and Ormerod 2010)specifically the standardized survey length of 500 m reachthat will capture differing amounts of natural variability de-pending on the size of the river While this must be notedthere are few significant correlations between river charac-teristics identified with stream order (Fig 4) which suggeststhat channel size is not influencing the RHS variables to agreat degree in these catchments

In this study it is anticipated that sites in high networkdensity areas will have higher levels of habitat diversity asindicated by previous studies of confluences and networks(Benda et al 2004a Best 1985 Rice 2017) in turn in-creasing mean sediment size and flow-type speed comparedto sites in low-density areas The results of the correlationsbetween distance network density and river characteristicswhen all catchments are combined support this hypothesiswith greater HQA flow-type speed and coarser sedimentsizes observed in areas with high distance network density(Fig 4)

For individual catchments elevation network density in-duces a stronger positive HQA response across all catch-ments than distance network density (Fig 4) This supportsthe evidence that individual confluences (Rice et al 2006)and high densities of confluences increase physical hetero-geneity within the river network (Benda et al 2004b Rice2017) However flow-type speed and sediment size responddifferently to network density in individual catchments

Slower flow types are observed in high network densityareas of the Avon and Tamar whereas faster flow types areobserved in high-elevation network density areas of the Wen-sum Individual confluences are shown to create numeroushigh- and low-speed flow environments (Best 1987) thatmay be observed in surface water topography (Biron et al2002) It was expected that the introduction of the additionalflow types by a high density of confluences in relatively low-

energy rivers would increase mean reach flow-type speedhowever the correlation analysis (Fig 4) suggests that insome catchments mean flow-type speed is reduced

Sediment size response also shows variation betweencatchments Sediment size is coarser in network-dense ar-eas of the Avon and Eden as expected but is finer in boththe Tamar and Wensum (Fig 4) The evidence from high-energy rivers shows that sediment becomes coarser down-stream and finer upstream of certain confluences (Benda etal 2004a Rice 1998) and in this case high numbers of con-fluences were expected to increase mean sediment size of thereach However the impact on sediment size is dependent onthe sediment calibre of the incoming tributary being higherthan the main channel with enough energy to transport thecoarse sediment to the confluence for numerous tributariesin an area The Tamar and Wensum have different rangesof sediment sizes with the Tamar having coarser sedimentthan the Wensum on average (Fig 3) This implies that trib-utaries in the Tamar may be energy limited not transportingcoarse sediments to confluences and the Wensum may suf-fer from inputs of fine sediments from the high percentageof arable land that concentrates in network-dense areas Thishas before been observed in a low-energy modified catch-ment where anthropogenic modifications in tributaries re-duced coarse sediment and flow capacity causing either lim-ited confluence impact or localized sediment fining (Singer2008)

Others have related the capacity of confluences to alterreach features to the morphometry of catchments with largerand more circular catchments containing a higher percentageof confluences that have a significant impact (Rice 2017)Based on this theory the Eden and Tamar are likely to havethe greatest impact as they are the most circular and steepestof the four catchments (although the Tamar is the smallestby area) Yet there is no clear pattern indicating that thesecatchments respond differently than the Avon and Wensum(Fig 4) with catchments responding differently to differ-ent variables This perhaps suggests that rather than networkdensity having a directional impact on factors such as flow-type speed and sediment size it has an impact on overall het-erogeneity at the reach level as suggested by previous stud-ies and that specific directional change occurs at the sub-reach level

An increase in channel modification is also hypothesizeddue to the increase in flood peak downstream from conflu-ences (Depettris et al 2000) and the scour and erosion as-sociated with confluence junctions (Best 1986) potentiallyincreasing the need for bed and bank protection The corre-lations between distance network density and HMS when allcatchments are combined undermine this hypothesis show-ing less channel modification where distance network den-sity is higher (Fig 4) There were few significant correlationswith individual catchments but HMS in the Eden was consis-tently observed to be higher in network-dense areas (Fig 4)This may be due to the Edenrsquos high elevation and steep to-

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2315

pography inducing a higher energy environment where scourand erosion processes in areas with high numbers of conflu-ences would be more likely to be present

Differences in RHS variable responses also differed be-tween the descriptive statistics considered Often the ex-tremes 90th and 10th percentiles showed more significantand stronger correlations than the mean or median (Fig 4)This may reflect findings from previous studies which sug-gest that not all confluences cause reach-scale changes (Rice1998) that perhaps the changes to river character induced bycertain confluences only influence certain reaches whereasothers are left unaffected creating less pronounced responsesin the mean and median of variables External factors mayalso influence this trend for example 10th percentile HMSvisually responds dramatically to network density metrics(Fig 3a and b) compared to the other descriptive statisticsThis suggests that the most natural sites (ie with the low-est HMS score) are responding differently to network den-sity with the most natural sites having less modificationin network-dense areas whereas less natural sites becomemore modified in network-dense areas This reflects the HQAscore results which visually (Fig 3a and b) and statistically(Fig 4) vary with distance network density except for the10th percentile These sites with the lowest habitat qualityand naturalness are likely influenced by anthropogenic fac-tors that are independent of network density reducing habitatquality scores at impacted sites

The two network density metrics have differing impacts onriver characteristics While distance network density showsconsistently significant correlations when all catchments arecombined individual catchments respond more frequentlyand more strongly to elevation network density (Fig 4)This may be because there is a dramatic split in distancenetwork density values between the more upland drainage-dense catchments Eden and Tamar than the lowland chalklow-drainage catchments Avon and Wensum The combinedcorrelation will therefore in part reflect the difference be-tween the catchments which have different ranges of rivercharacteristics (Fig 3) This is not the case for elevation net-work density which has higher density values in the Tamarand Wensum than the Avon and Eden so therefore the char-acteristics of the catchments will have less bearing on thecombined correlation However there are patterns identifiedwith distance network density in individual catchments thatare not present with elevation network density increasedflow-type speed and sediment size in the Eden and reducedflow-type speed in the Avon with network density (Fig 4)which show its usefulness

The results suggest that the distance network density andelevation network density metrics quantify different dimen-sions of network topology which are shown to exhibit func-tionally meaningful patterns for river reach characteristicsbased on the correlation results Perhaps within catchmentselevation network density provides the more powerful met-ric for individual catchments but distance network density

better accounts for the drainage density of the catchmentsallowing it to be applied across multiple catchments

43 Comparison of stream order to network densitymetrics

The classic method of accounting for network topologystream order is critiqued for failing to represent disconti-nuities in the network and simplifying the network into agradient The number of links in different stream orders isconsistent across all catchments not reflecting the internalstructure of the network or the variety between catchmentsthat is achieved by the distance network density and eleva-tion network density metrics The analysis of the two net-work density metrics presented in this paper shows that dis-tances from source and elevation are not mutually exclusive(Fig 2a and b) contrary to the stream-order metric whichrepresents streams as upstream to downstream or upland tolowland

Stream order has few significant correlations with manyof the river characteristics considered in this study Negativecorrelations with HMS in the Wensum were statistically sig-nificant post p-value correction likely driving the significantrelationship with all catchments combined for this variable(Fig 4) This suggests that modification is greater upstreamin the Wensum contrary to ideas that downstream reachesmay show greater anthropogenic modification Intense agri-cultural land use in the upper reaches of the Wensum islikely to be the cause of the high HMS scores upstreamThe lack of significant correlations suggests that stream or-der and therefore an upstream-to-downstream gradient arenot the predominant pattern in river characteristics despitethe description of such a gradient by geomorphic (Schumm1977) and ecological frameworks (Vannote et al 1980)This is surprising as distance and elevation which both re-flect the upstream-to-downstream gradient have proven tobe important factors in previous studies explaining patternsof RHS features at a national level (Jeffers 1998a Vaughanet al 2013) This implies that upstream-to-downstream gra-dient may not sufficiently reflect patterns of river charac-teristics through the river network within individual catch-ments Others have also found that stream order has weakand inconsistent relationships with biodiversity patterns inriver systems arguing that the topological measure has no di-rect mechanistic control on biodiversity (Vander Vorste et al2017) Instead this study finds that the network density met-rics are a more powerful metric which conceptually providean improved method of accounting for the impacts of net-work topology on the fluvial system exhibiting relationshipswith river characteristics particularly habitat quality score(Fig 4)

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2316 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

44 Applicability of network topology metrics todifferent environments

Much of the seminal work on network and confluence im-pacts (eg the Network Dynamics Hypothesis Benda et al2004b and Link Discontinuity Concept Rice et al 2001) isconducted in natural highly erosive catchments with first-hand empirical measurements However in an age whenrivers and their catchments are increasingly altered by an-thropogenic modification (Meybeck 2003) contemporarystudies must not only aim to expand knowledge but also findmethods of transferring knowledge to many increasingly al-tered catchments (Clifford 2002)

The catchments selected by this study are more greatlymodified and although they reflect a range of fluvial envi-ronments in England are lower-energy catchments than theseminal studies Benda et al (2004a) suggest that confluenceeffects in less active landscapes would be subdued comparedto highly erosive landscapes but the evidence presented heredemonstrates the utility of evaluating network topologicalstructure in studies on catchment-level effects in any typeof fluvial environment including those with low-energy andwidespread anthropogenic changes The response of someriver characteristics varied between catchments observationsof flow-type speed sediment size and modifications showeddifferent responses to network density in different catch-ments This suggests that the functional effect of these topo-logical metrics is catchment dependent and likely is influ-enced by external catchment characteristics such as land usenot considered in this study although impact did not appearto vary with catchment topography or circularity as has beenshown in prior studies (Benda et al 2004a Rice 2017) Thisshould be explored further in future research to enable rec-ommendations to be made regarding where and how reachesmay respond to network density The response of habitatquality score was however consistent across catchments andbetween metrics showing that habitat quality is greater inareas with high network density (Fig 4) as hypothesizedby the Network Dynamics Hypothesis (Benda et al 2004b)and demonstrated by studies on individual catchments (Rice2017 Rice et al 2006)

The methods presented in this paper are designed to be im-plemented in any catchment with a dendritic network struc-ture The topology metrics can easily be calculated from anydendritic network with DTM data using GIS and comparedto any site-scale data This study uses regulatory monitoringdatasets so that analysis is targeted to assessment scores andphysical features of interest to river managers Also the highvolume and wide spatial extent of data available from regu-latory sources allow for between-catchment comparisons

5 Conclusions

Although appreciation of catchment-level effects is consid-ered the epitome of understanding river functioning a keycomponent of the catchment ndash the river network ndash is over-looked and oversimplified by catchment-level studies Thisstudy finds that river network density plays a role in struc-turing the distribution river characteristics throughout thecatchment offering more detailed explanation than the clas-sic stream-order metric Network-dense areas are generallyfound to have higher habitat quality and diversity but modi-fication flow-type speed and sediment size show different re-sponses in different catchments This study suggests two po-tential reasons for this (1) there is evidence that confluencesin the river network increase diversity as is observed in thisstudy so the direction of mean river characteristic responsemay not be consistent and (2) there may be external factorssuch as sediment availability land cover and anthropogenicmodification that alter the direction of mean river characteris-tic response This paper demonstrates the functional responseof river characteristics to network topology and suggests thatthe inclusion of network topology in catchment-level studieswould add a layer of function-based understanding to suchstudies linking reaches to their catchments

The broad-scale methodology adopted by this study allowsthe network density metrics which are easily extracted fromopen-source data using GIS software to be compared to anyregulatory dataset The use of regulatory datasets allows notonly for analysis over a wider spatial extent but also for moreapplicable results for regulatory bodies Therefore the inter-disciplinary approach to characterizing network topology canbe applied efficiently and effectively to capture catchment-level impacts on reach-level functioning in any catchmentacross the globe

Data availability River Habitat Survey data are freely availablefrom the Environment Agency England River network data andelevation data are available from the Centre for Ecology and Hy-drology UK

Author contributions EH and NC conceptualised and designed thestudy EH performed the data analysis with the support of JMEH wrote the paper with comments provided by NC and JM

Competing interests The authors declare that they have no conflictof interest

Acknowledgements The authors would like to thank the Environ-ment Agency and the Centre for Ecology and Hydrology for accessto the data used in this paper This research is funded by the NaturalEnvironmental Research Council (NERC)

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2317

Review statement This paper was edited by Nandita Basu and re-viewed by Simone Bizzi and one anonymous referee

References

Belletti B Rinaldi M Buijse A D Gurnell A Mand Mosselman E A review of assessment methods forriver hydromorphology Environ Earth Sci 73 2079ndash2100httpsdoiorg101007s12665-014-3558-1 2015

Benda L Andras K Miller D and Bigelow P Confluenceeffects in rivers Interactions of basin scale network geom-etry and disturbance regimes Water Resour Res 40 1ndash15httpsdoiorg1010292003WR002583 2004a

Benda L Poff N L Miller D Dunne T Reeves GPess G and Pollock M The network dynamics hy-pothesis how channel networks structure riverine habi-tats Bioscience 54 413ndash427 httpsdoiorg1016410006-3568(2004)054[0413TNDHHC]20CO2 2004b

Benjamini Y and Hochberg Y Controlling the False DiscoveryRate A Practical and Powerful Approach to Multiple TestingJ Roy Stat Soc B 57 289ndash300 1995

Best J L Flow dynamics and sediment transport at river channelconfluences Birbeck University of London London 1985

Best J L The morphology of river channel con-fluences Prog Phys Geogr 10 157ndash174httpsdoiorg101177030913338601000201 1986

Best J L Flow dynamics at river channel confluences implica-tions for sediment transport and bed morphology in Recent de-velopments in fluvial sedimentology edited by Ethridge F GFlores R M and Harvey M D Spec Publ Soc Econ Pale-ontol Miner Tulsa Okla 27ndash35 1987

Biron P M Richer A Kirkbride A D Roy A G andHan S Spatial patterns of water surface topography ata river confluence Earth Surf Proc Land 27 913ndash928httpsdoiorg101002esp359 2002

Bravard J P and Gilvear D J Hydrological and geomorpho-logical structure of hydrosystems in The Fluvial Hydrosystemedited by Petts G E and Amoros C Springer Netherlands98ndash116 1996

Brierley G and Fryirs K River Styles a Geomorphic Approachto Catchment Characterization Implications for River Rehabili-tation in Bega Catchment New South Wales Australia EnvironManage 25 661ndash679 2000

Bunn S E and Arthington A H Basic principles and ecologi-cal consequences of altered flow regimes for aquatic biodiversityEnviron Manage 30 492ndash507 httpsdoiorg101007s00267-002-2737-0 2002

Church M and Kellerhals R On the statistics of grain size varia-tion along a gravel river Can J Earth Sci 15 1151ndash1160 1978

Clifford N J Hydrology the changing paradigm Prog Phys Ge-ogr 26 290ndash301 2002

Cohen P Andriamahefa H and Wasson J-G Towardsa regionalizaton of aquatic habitat distribution of meso-habitats at the scale of a large basin Regul RiversRes Manage 14 391ndash404 httpsdoiorg101002(SICI)1099-1646(19980910)145lt391AID-RRR513gt30CO2-W 1998

Davenport A J Gurnell A M and Armitage P D Habitat sur-vey and classification of urban rivers River Res Appl 20 687ndash704 httpsdoiorg101002rra785 2004

Davies N M Norris R H and Thoms M C Prediction and as-sessment of local stream habitat features using large-scale catch-ment characteristics Freshwater Biol 45 343ndash369 2000

Depettris C Mendiondo E M Neiff J and Rohrmann H Flooddefence strategy at the confluence of the Paranaacute-Paraguay riversProc Int Symp Flood Def Kassel He C31 C40 2000

Dollar E James C Rogers K and Thoms M A framework forinterdisciplinary understanding of rivers as ecosystems Geomor-phology 89 147ndash162 2007

Dovers S R and Day D G Australian rivers and statute law En-viron Plan Law J 5 90ndash108 1988

Downs P W Gregory K J and Brookes A How integratedis river basin management Environ Manage 15 299ndash309httpsdoiorg101007BF02393876 1991

Emery J C Gurnell A M Clifford N J and Petts G E Char-acteristics and controls of gravel-bed riffles An analysis of datafrom the river-habitat survey Water Environ J 18 210ndash216httpsdoiorg101111j1747-65932004tb00535x 2004

Evans I S and Minaacuter J A classification of geomorphometric vari-ables in International Geom-orphometry 2011 Geomoprhome-tryorg Redlabds CA 105ndash108 2011

Fausch K D Torgersen C E Baxter C V Li H WView C The O F Is R To N How U Among ISet S For C and Fishes S Landscapes to riverscapesBridging the gap between research and conservation of streamfishes Bioscience 52 483ndash498 httpsdoiorg1016410006-3568(2002)052[0483LTRBTG]20CO2 2002

Glickman M E Rao S R and Schultz M R False discov-ery rate control is a recommended alternative to Bonferroni-typeadjustments in health studies J Clin Epidemiol 67 850ndash857httpsdoiorg101016jjclinepi201403012 2014

Gupta V K and Mesa O J Runoff generation and hydro-logic response via channel network geomorphology - Re-cent progress and open problems J Hydrol 102 3ndash28httpsdoiorg1010160022-1694(88)90089-3 1988

Gupta V K and Waymire E D On the formulation of an analyt-ical approach to hydrologic response and similarity at the basinscale J Hydrol 65 95ndash123 1983

Gupta V K Waymire E D and Rodriguez-Iturbe I On scalesgravity and network structure in basin runoff in Scale problemsin hydrology Springer Netherlands 159ndash184 1986

Gurnell A M Rinaldi M Belletti B Bizzi S Blamauer BBraca G Buijse A D Bussettini M Camenen B ComitiF Demarchi L Garciacutea de Jaloacuten D Gonzaacutelez del TaacutenagoM Grabowski R C Gunn I D M Habersack H Hen-driks D Henshaw A J Kloumlsch M Lastoria B LatapieA Marcinkowski P Martiacutenez-Fernaacutendez V Mosselman EMountford J O Nardi L Okruszko T OrsquoHare M T PalmaM Percopo C Surian N van de Bund W Weissteiner Cand Ziliani L A multi-scale hierarchical framework for devel-oping understanding of river behaviour to support river manage-ment Aquat Sci 78 1ndash16 httpsdoiorg101007s00027-015-0424-5 2016

Harvey G L Gurnell A M and Clifford N J Characterisationof river reaches The influence of rock type Catena 76 78ndash88httpsdoiorg101016jcatena200809010 2008

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2309

Figure 2 Distance and elevation intervals for each Demonstration Test Catchment Avon (A) Eden (E) Tamar (T) and Wensum (W)(a) Percentage distance intervals used to calculate distance network density (b) Percentage elevation intervals used to calculate elevationnetwork density Map of catchment locations in England in the bottom-right corner

be compared between catchments controlling for differencesin size and elevation as well as within catchments

To assess the utility of the multi-dimensional topologymetrics in accounting for the spatial structure of the networkthe metrics are compared to the one-dimensional Strahlerstream-order metric extracted from the river network datasetusing RivEX (Hornby 2010)

23 River characteristics

The impact of network topology on channel functioning isexplored using a broad-scale approach ie adapting data col-lected for regulatory compliance to answer scientific ques-tions Adapting such datasets to scientific enquiry allowsanalysis to be conducted in many catchments across a widespatial extent There are many habitat monitoring methodsacross the globe with 121 survey methods recorded in over

26 different countries (Belletti et al 2015) so this methodmay be adapted to other countries

This study utilizes the River Habitat Survey (RHS Ravenet al 1996) a regulatory dataset collected by Englandrsquos En-vironment Agency which is used to reflect the river reachcharacteristics in each catchment This dataset has been usedto identify catchment effects on river characteristics in broad-scale studies by previous research (eg Harvey et al 2008Naura et al 2016 Vaughan et al 2013) but none have in-cluded the effects of network topology

Since 1994 over 24 000 sites have been sampled incatchments across England and Wales including the Avon(n= 418) Eden (n= 398) Tamar (n= 189) and Wensum(n=315) Surveys were conducted at random sites withineach 10 km2 of England and Wales to ensure geographic cov-erage however this produces sampling bias as streams inhigh-density areas will be under-represented in the datasetwhich is acknowledged in this study and discussed below

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2310 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

Table 1 RHS variables calculated from RHS observations

RHS variable Calculation from RHS observations Units

Habitat Quality A score indicating the degree of naturalness and diversity of the HQA scaleAssessment riparian zone based on observations in the reach of flow types(HQA) substrate channel and bank features riparian vegetation etc

Habitat A score indicating the degree of artificial modification of the channel HMS scaleModification based on observations in the reach of reinforcements re-sectioningScore (HMS) embankments weed cutting realignment culverts dams weirs etc

Sediment size =(minus8middotBOminus7middotCOminus35middotGPminus15middotSA+15middotSI+9middotCL)

(BO+CO+GP+SA+SI+CL) ApproxBO (boulder) CO (cobble) GP (gravelndashpebble) SA (sand) SI (silt) φ unitsand CL (clay) represent the number of spot checks allocated to eachsediment size class

Flow-type speed =(0middotDR+1middotNP+2middotUP+3middotSM+4middotRP+5middotUW+6middotBW+7middotCF+8middotCH+9middotFF)

(DR+NP+UP+SM+RP+UW+BW+CF+CH+FF) Flow typeDR (dry) NP (no perceptible flow) UP (upwelling) SM (smooth) speed scaleRP (rippled) UW (unbroken wave) BW (broken wave) CF (chaoticflow) CH (chute) FF (free-fall) represent the number of spot checksallocated to each flow speed class

At each site over 100 features are recorded along a 500 mreach with 10 ldquospot-checkrdquo surveys conducted every 50 mand a ldquosweep-uprdquo survey conducted across the whole reach(see Raven et al 1996 for details) Variables of interest thatare hypothesized to be impacted by network structure can becalculated from the RHS observations (Table 1)

The Habitat Quality Assessment (HQA) and Habitat Mod-ification Score (HMS) variables are both amalgamations ofRHS observations with individual features given a score de-rived by expert opinion (see Raven et al 1998 for more de-tails) The scoring systems are subjective but HQA and HMSprovide overviews of the channel condition that are widelyapplied for regulatory compliance The scores are thereforeincluded in this study to reflect how they may be impactedby network topology

The remaining RHS variables are calculated directly fromRHS observations and so are more objective Sediment sizeis calculated as a reach average of spot-check observationsusing the same method as previous studies (Davenport et al2004 Emery et al 2004 Harvey et al 2008) Flow-typespeed was calculated in the same manner as sediment sizeusing values of flow which represent an approximate flowvelocity gradient defined in Davenport et al (2004) Thesevariables were chosen to reflect dominant geomorphic pro-cesses occurring in each reach and due to the prominence ofsediment size and flow type in defining physical habitats forinstream biota (Rowntree and Wadeson 1996) The variablesare likely to be impacted by the density of the river networkas they have been shown to be impacted by individual con-fluences For example channels are shown to become moregeomorphologically heterogeneous (Benda et al 2004a) andsubstrate size has been shown to coarsen at confluences (Riceet al 2001) Surface flow types are also likely to become

more diverse at confluences as the convergence of channelscreates a number of different flow environments (Best 1985)that result in different water-surface topographies (Biron etal 2002)

It must be noted that the RHS dataset was collected forregulatory compliance and was not directly intended for sci-entific enquiry Therefore there is a limitation in the amountof detail that can be extracted about physical processes as theobservations recorded are an average across a 500 m reachDespite this the wide spatial coverage of the dataset makesit a powerful tool allowing analysis to be conducted acrossmultiple catchments with differing characteristics

For each distance and elevation interval created by the net-work topology metrics descriptive statistics (mean median90th and 10th percentiles) of each RHS variable are calcu-lated Despite the RHS sampling strategy (Jeffers 1998b) bi-asing site selection towards less dense areas of the networkmost distance and elevation intervals contained RHS sites(with only some low-density intervals not containing RHSsites) This method is designed to account for natural varia-tion and modification at individual RHS sites in order to as-sess broad patterns of reach characteristics at the catchmentlevel

24 Statistical analysis

Analysis is conducted with all catchments combined into asingle population to identify overall trends across all catch-ments a method used in previous broad-scale studies Theanalysis is also split into individual catchments to identifyhow the relationship between network topology metrics andriver reach characteristics differed between catchments

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2311

Kendall correlation

Correlation tests are used to determine the strength and di-rection of the association between the descriptive statisticsof the RHS variables and distance network density eleva-tion network density and stream order to ascertain how reachcharacteristics respond to network topology Kendallrsquos cor-relation method was used as the variables have non-normaldistributions a small sample size and tied data values (Helseland Hirsch 2002) The effect size of Kendallrsquos τ is lowerthan other correlation methods with strong correlations oc-curring with τ values greater than 07 (Helsel and Hirsch2002)

As multiple correlations are conducted false discoveryrate (Benjamini and Hochberg 1995) corrections were ap-plied to the p-values produced from the Kendall correlationsto reduce the risk of type I error The false discovery ratemethod has been found to be more powerful than other pro-cedures for controlling for multiple tests (Glickman et al2014)

3 Results

31 Differences in network topology metrics betweencatchments

The topological metrics developed in this study show the in-ternal structure of the network for each catchment The sep-aration of the catchments into distance and elevation inter-vals emphasizes different features of the catchment The dis-tance intervals (Fig 2a) are arranged longitudinally withinthe catchment highlighting sub-basins within each catch-ment The elevation intervals (Fig 2b) have a radial arrange-ment centering around the incised main channel of eachcatchment

Distance network density is higher in the Eden (284plusmn103) and Tamar (441plusmn 219) compared to the Avon (47plusmn19) and Wensum (68plusmn 07) interesting as the Tamar is thesmallest catchment by area The shape of the distance net-work density function reflects the internal shape of the net-work (Fig 3a) For example the Tamar has a peaked den-sity distribution reflecting the circular shape of the catch-ment such that the majority of links are at 55 ndash65 dis-tance from the catchment outlet The Avon and Eden reflectsimilar internal network structures both exhibiting a bimodaldensity distribution despite the differences in the number oflinks in the catchments The density distribution of the Wen-sum has a more complex internal distribution of links withmultiple peaks in density

Elevation network density (Fig 3b) shows similar densitydistribution shapes to distance network density with a uni-modal distribution for the Tamar and multi-modal distribu-tions in the other catchments In contrast to distance networkdensity elevation network density shows the highest peaks in

density in the Tamar (103plusmn 50) and Wensum (101plusmn 43)despite the Wensum having the lowest network elevation andhas lower values in the Avon (34plusmn10) and Eden (58plusmn25)The peak densities in the Wensum occur at similar positionsin the elevation and distance intervals whereas the peaks inthe other catchments are negatively skewed showing the net-work density is highest at lowndashmid elevations

Nearly half of the links in each catchment are classified asfirst-order streams and the number of links declines exponen-tially towards the highest orders in all four catchments Thereare weak correlations (τ =minus003 to 017) between the threenetwork topology metrics distance network density eleva-tion network density and stream order This suggests that themetrics are independent and reflect different aspects of rivernetwork topology

32 River characteristic relationships with networktopology metrics

RHS sites in the Avon and Wensum have similar river char-acteristics Both have lower habitat quality high modifica-tion fine sediment and slower flow types than the Eden andTamar When all catchments are combined there are signif-icant (p lt 005 after p-value correction) correlations withmost descriptive statistics for each RHS variable and dis-tance network density (Fig 4) There are consistently pos-itive correlations with HQA and flow-type speed and neg-ative correlations with HMS and sediment size There arefewer and weaker significant correlations with elevation net-work density (Fig 4) The only significant correlations withstream order are with HMS which shows a negative correla-tion (Fig 4)

There are numerous significant correlations between thenetwork topology metrics and RHS variables for individualcatchments many of which were also shown to be signifi-cant after the correction to the p-value The results show thatcatchments have different responses to the network topol-ogy metrics of distance network density and elevation net-work density Distance network density only has significantcorrelations with the regulatory scoring variables (HQA andHMS) in the Eden and Tamar (Fig 4) Elevation networkdensity however has a wider array of significant correlationswith the scoring variables particularly HQA which showssubtle peaks and troughs reflecting the distribution of bothnetwork density metrics (Fig 3a and b) HMS shows mostlynegative correlations mainly with elevation network densityapart from the Eden which has significant positive correla-tion across all HMS descriptive statistics (Fig 4) Visually10th percentile HMS is most variable to network density withpeaks and troughs responding to the network density distri-butions (Fig 3a and b)

For individual RHS features the response to networktopology varies between catchments (Fig 4) The Avon hasnegative correlations between flow-type speed and distancenetwork density with an evident drop in 10th percentile

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2312 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

Figure 3 Network topology metrics (a) distance network density and (b) elevation network density Descriptive statistics of each RHSvariable over (a) distance and (b) elevation for each catchment with smooth loess lines to indicate trend Network topology metrics arenormalized between 0 and 1 and HMS score is logarithmically transformed for display purposes

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2313

Figure 4 Summary of correlations between distance network density elevation network density stream order and RHS variables for allcatchments combined and each individual catchment Significance of correlation is indicated by point size with the largest points significantpost p-value correction Effect size (Kendallrsquos τ ) is indicated by colour No correlation was possible between stream order and 10th percentilesediment size in the Avon due to no variation in the RHS variable

flow-type speed associated with peaks in network density(Fig 3a) The Eden and Tamar however have positive cor-relations with mean and 90th percentile flow-type speed fordistance network density but negative correlations with me-dian and 10th percentile elevation network density The Wen-sum shows positive correlations between flow-type speed andelevation network density Sediment size has a consistent re-sponse to distance network density with the Eden and Wen-sum showing negative correlations with the sediment size(Fig 4) For elevation network density the Avon shows neg-ative correlations with sediment size whereas the Tamar andWensum show positive correlations (Fig 4)

There were few significant correlations between stream or-der and the RHS variables in individual catchments (Fig 4)The only significant correlation after p-value correction iswith 90th percentile HMS in the Wensum which shows astrong negative relationship

4 Discussion

41 A new approach to utilizing network topology incatchment-level analysis

Network density metrics represent an alternative approach toaccount for network topology in catchment-level studies op-timizing the width dimension of the network (or the num-

ber of links in the network) as opposed to the commonplacestream-order metric which only reflects the longitudinal po-sition of links in a network (Fig 1) This study demonstratesthat two topology metrics can be calculated simply from aDTM with GIS and using a broad-scale analysis of river at-tributes can be used to investigate the functional processeswithin catchments

While the two network density metrics have similar forms(ie forms are consistently unimodal or multi-modal) thespatial configuration of the distance and elevation intervalsused in the calculation of network density varies and mayimpact the effectiveness of each topological metric Distancenetwork density separates the catchment into intervals basedon distance which spread upstream from the outlet (Fig 2a)reflecting natural sub-basins within the fractal structure of thecatchment (Lashermes and Foufoula-Georgiou 2007) Thisdiffers from elevation network density which separates thecatchment into intervals based on elevation which radiate outfrom the main channel of the network (Fig 2b) The config-uration means that distance intervals contain streams that arein closer proximity to one another rather than the more distalconfiguration created by the elevation intervals suggesting adegree of spatial dependency in river functioning This hasbeen highlighted in previous studies where spatial networkstructure has a stronger influence on some in-channel pro-cesses than predictor variables such as elevation (Steel et al

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2314 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

2016) However elevation intervals contain RHS sites thatalthough they may be distal may have similar properties aselevation has been strongly related to RHS variables includ-ing flow type substrate etc in a number of studies (Jeffers1998a Naura et al 2016 Vaughan et al 2013)

42 Impacts of network topology on rivercharacteristics

River characteristics are assessed using the RHS dataset Theobservations made by the RHS dataset (Table 1) cannot of-fer the level of detail regarding geomorphological processthat river classifications which consider multiple scales canoffer (eg Brierley and Fryirs 2000 Gurnell et al 2016)While process-based classifications are preferable broad-scale monitoring datasets such as the River Habitat Surveymay still be useful when combined with map-derived datato explore controls on river characteristics (Harvey et al2008 Naura et al 2016 Vaughan et al 2013) Howeverthere are biases in RHS data collection an inherent limitationwhen using existing datasets (Vaughan and Ormerod 2010)specifically the standardized survey length of 500 m reachthat will capture differing amounts of natural variability de-pending on the size of the river While this must be notedthere are few significant correlations between river charac-teristics identified with stream order (Fig 4) which suggeststhat channel size is not influencing the RHS variables to agreat degree in these catchments

In this study it is anticipated that sites in high networkdensity areas will have higher levels of habitat diversity asindicated by previous studies of confluences and networks(Benda et al 2004a Best 1985 Rice 2017) in turn in-creasing mean sediment size and flow-type speed comparedto sites in low-density areas The results of the correlationsbetween distance network density and river characteristicswhen all catchments are combined support this hypothesiswith greater HQA flow-type speed and coarser sedimentsizes observed in areas with high distance network density(Fig 4)

For individual catchments elevation network density in-duces a stronger positive HQA response across all catch-ments than distance network density (Fig 4) This supportsthe evidence that individual confluences (Rice et al 2006)and high densities of confluences increase physical hetero-geneity within the river network (Benda et al 2004b Rice2017) However flow-type speed and sediment size responddifferently to network density in individual catchments

Slower flow types are observed in high network densityareas of the Avon and Tamar whereas faster flow types areobserved in high-elevation network density areas of the Wen-sum Individual confluences are shown to create numeroushigh- and low-speed flow environments (Best 1987) thatmay be observed in surface water topography (Biron et al2002) It was expected that the introduction of the additionalflow types by a high density of confluences in relatively low-

energy rivers would increase mean reach flow-type speedhowever the correlation analysis (Fig 4) suggests that insome catchments mean flow-type speed is reduced

Sediment size response also shows variation betweencatchments Sediment size is coarser in network-dense ar-eas of the Avon and Eden as expected but is finer in boththe Tamar and Wensum (Fig 4) The evidence from high-energy rivers shows that sediment becomes coarser down-stream and finer upstream of certain confluences (Benda etal 2004a Rice 1998) and in this case high numbers of con-fluences were expected to increase mean sediment size of thereach However the impact on sediment size is dependent onthe sediment calibre of the incoming tributary being higherthan the main channel with enough energy to transport thecoarse sediment to the confluence for numerous tributariesin an area The Tamar and Wensum have different rangesof sediment sizes with the Tamar having coarser sedimentthan the Wensum on average (Fig 3) This implies that trib-utaries in the Tamar may be energy limited not transportingcoarse sediments to confluences and the Wensum may suf-fer from inputs of fine sediments from the high percentageof arable land that concentrates in network-dense areas Thishas before been observed in a low-energy modified catch-ment where anthropogenic modifications in tributaries re-duced coarse sediment and flow capacity causing either lim-ited confluence impact or localized sediment fining (Singer2008)

Others have related the capacity of confluences to alterreach features to the morphometry of catchments with largerand more circular catchments containing a higher percentageof confluences that have a significant impact (Rice 2017)Based on this theory the Eden and Tamar are likely to havethe greatest impact as they are the most circular and steepestof the four catchments (although the Tamar is the smallestby area) Yet there is no clear pattern indicating that thesecatchments respond differently than the Avon and Wensum(Fig 4) with catchments responding differently to differ-ent variables This perhaps suggests that rather than networkdensity having a directional impact on factors such as flow-type speed and sediment size it has an impact on overall het-erogeneity at the reach level as suggested by previous stud-ies and that specific directional change occurs at the sub-reach level

An increase in channel modification is also hypothesizeddue to the increase in flood peak downstream from conflu-ences (Depettris et al 2000) and the scour and erosion as-sociated with confluence junctions (Best 1986) potentiallyincreasing the need for bed and bank protection The corre-lations between distance network density and HMS when allcatchments are combined undermine this hypothesis show-ing less channel modification where distance network den-sity is higher (Fig 4) There were few significant correlationswith individual catchments but HMS in the Eden was consis-tently observed to be higher in network-dense areas (Fig 4)This may be due to the Edenrsquos high elevation and steep to-

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2315

pography inducing a higher energy environment where scourand erosion processes in areas with high numbers of conflu-ences would be more likely to be present

Differences in RHS variable responses also differed be-tween the descriptive statistics considered Often the ex-tremes 90th and 10th percentiles showed more significantand stronger correlations than the mean or median (Fig 4)This may reflect findings from previous studies which sug-gest that not all confluences cause reach-scale changes (Rice1998) that perhaps the changes to river character induced bycertain confluences only influence certain reaches whereasothers are left unaffected creating less pronounced responsesin the mean and median of variables External factors mayalso influence this trend for example 10th percentile HMSvisually responds dramatically to network density metrics(Fig 3a and b) compared to the other descriptive statisticsThis suggests that the most natural sites (ie with the low-est HMS score) are responding differently to network den-sity with the most natural sites having less modificationin network-dense areas whereas less natural sites becomemore modified in network-dense areas This reflects the HQAscore results which visually (Fig 3a and b) and statistically(Fig 4) vary with distance network density except for the10th percentile These sites with the lowest habitat qualityand naturalness are likely influenced by anthropogenic fac-tors that are independent of network density reducing habitatquality scores at impacted sites

The two network density metrics have differing impacts onriver characteristics While distance network density showsconsistently significant correlations when all catchments arecombined individual catchments respond more frequentlyand more strongly to elevation network density (Fig 4)This may be because there is a dramatic split in distancenetwork density values between the more upland drainage-dense catchments Eden and Tamar than the lowland chalklow-drainage catchments Avon and Wensum The combinedcorrelation will therefore in part reflect the difference be-tween the catchments which have different ranges of rivercharacteristics (Fig 3) This is not the case for elevation net-work density which has higher density values in the Tamarand Wensum than the Avon and Eden so therefore the char-acteristics of the catchments will have less bearing on thecombined correlation However there are patterns identifiedwith distance network density in individual catchments thatare not present with elevation network density increasedflow-type speed and sediment size in the Eden and reducedflow-type speed in the Avon with network density (Fig 4)which show its usefulness

The results suggest that the distance network density andelevation network density metrics quantify different dimen-sions of network topology which are shown to exhibit func-tionally meaningful patterns for river reach characteristicsbased on the correlation results Perhaps within catchmentselevation network density provides the more powerful met-ric for individual catchments but distance network density

better accounts for the drainage density of the catchmentsallowing it to be applied across multiple catchments

43 Comparison of stream order to network densitymetrics

The classic method of accounting for network topologystream order is critiqued for failing to represent disconti-nuities in the network and simplifying the network into agradient The number of links in different stream orders isconsistent across all catchments not reflecting the internalstructure of the network or the variety between catchmentsthat is achieved by the distance network density and eleva-tion network density metrics The analysis of the two net-work density metrics presented in this paper shows that dis-tances from source and elevation are not mutually exclusive(Fig 2a and b) contrary to the stream-order metric whichrepresents streams as upstream to downstream or upland tolowland

Stream order has few significant correlations with manyof the river characteristics considered in this study Negativecorrelations with HMS in the Wensum were statistically sig-nificant post p-value correction likely driving the significantrelationship with all catchments combined for this variable(Fig 4) This suggests that modification is greater upstreamin the Wensum contrary to ideas that downstream reachesmay show greater anthropogenic modification Intense agri-cultural land use in the upper reaches of the Wensum islikely to be the cause of the high HMS scores upstreamThe lack of significant correlations suggests that stream or-der and therefore an upstream-to-downstream gradient arenot the predominant pattern in river characteristics despitethe description of such a gradient by geomorphic (Schumm1977) and ecological frameworks (Vannote et al 1980)This is surprising as distance and elevation which both re-flect the upstream-to-downstream gradient have proven tobe important factors in previous studies explaining patternsof RHS features at a national level (Jeffers 1998a Vaughanet al 2013) This implies that upstream-to-downstream gra-dient may not sufficiently reflect patterns of river charac-teristics through the river network within individual catch-ments Others have also found that stream order has weakand inconsistent relationships with biodiversity patterns inriver systems arguing that the topological measure has no di-rect mechanistic control on biodiversity (Vander Vorste et al2017) Instead this study finds that the network density met-rics are a more powerful metric which conceptually providean improved method of accounting for the impacts of net-work topology on the fluvial system exhibiting relationshipswith river characteristics particularly habitat quality score(Fig 4)

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2316 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

44 Applicability of network topology metrics todifferent environments

Much of the seminal work on network and confluence im-pacts (eg the Network Dynamics Hypothesis Benda et al2004b and Link Discontinuity Concept Rice et al 2001) isconducted in natural highly erosive catchments with first-hand empirical measurements However in an age whenrivers and their catchments are increasingly altered by an-thropogenic modification (Meybeck 2003) contemporarystudies must not only aim to expand knowledge but also findmethods of transferring knowledge to many increasingly al-tered catchments (Clifford 2002)

The catchments selected by this study are more greatlymodified and although they reflect a range of fluvial envi-ronments in England are lower-energy catchments than theseminal studies Benda et al (2004a) suggest that confluenceeffects in less active landscapes would be subdued comparedto highly erosive landscapes but the evidence presented heredemonstrates the utility of evaluating network topologicalstructure in studies on catchment-level effects in any typeof fluvial environment including those with low-energy andwidespread anthropogenic changes The response of someriver characteristics varied between catchments observationsof flow-type speed sediment size and modifications showeddifferent responses to network density in different catch-ments This suggests that the functional effect of these topo-logical metrics is catchment dependent and likely is influ-enced by external catchment characteristics such as land usenot considered in this study although impact did not appearto vary with catchment topography or circularity as has beenshown in prior studies (Benda et al 2004a Rice 2017) Thisshould be explored further in future research to enable rec-ommendations to be made regarding where and how reachesmay respond to network density The response of habitatquality score was however consistent across catchments andbetween metrics showing that habitat quality is greater inareas with high network density (Fig 4) as hypothesizedby the Network Dynamics Hypothesis (Benda et al 2004b)and demonstrated by studies on individual catchments (Rice2017 Rice et al 2006)

The methods presented in this paper are designed to be im-plemented in any catchment with a dendritic network struc-ture The topology metrics can easily be calculated from anydendritic network with DTM data using GIS and comparedto any site-scale data This study uses regulatory monitoringdatasets so that analysis is targeted to assessment scores andphysical features of interest to river managers Also the highvolume and wide spatial extent of data available from regu-latory sources allow for between-catchment comparisons

5 Conclusions

Although appreciation of catchment-level effects is consid-ered the epitome of understanding river functioning a keycomponent of the catchment ndash the river network ndash is over-looked and oversimplified by catchment-level studies Thisstudy finds that river network density plays a role in struc-turing the distribution river characteristics throughout thecatchment offering more detailed explanation than the clas-sic stream-order metric Network-dense areas are generallyfound to have higher habitat quality and diversity but modi-fication flow-type speed and sediment size show different re-sponses in different catchments This study suggests two po-tential reasons for this (1) there is evidence that confluencesin the river network increase diversity as is observed in thisstudy so the direction of mean river characteristic responsemay not be consistent and (2) there may be external factorssuch as sediment availability land cover and anthropogenicmodification that alter the direction of mean river characteris-tic response This paper demonstrates the functional responseof river characteristics to network topology and suggests thatthe inclusion of network topology in catchment-level studieswould add a layer of function-based understanding to suchstudies linking reaches to their catchments

The broad-scale methodology adopted by this study allowsthe network density metrics which are easily extracted fromopen-source data using GIS software to be compared to anyregulatory dataset The use of regulatory datasets allows notonly for analysis over a wider spatial extent but also for moreapplicable results for regulatory bodies Therefore the inter-disciplinary approach to characterizing network topology canbe applied efficiently and effectively to capture catchment-level impacts on reach-level functioning in any catchmentacross the globe

Data availability River Habitat Survey data are freely availablefrom the Environment Agency England River network data andelevation data are available from the Centre for Ecology and Hy-drology UK

Author contributions EH and NC conceptualised and designed thestudy EH performed the data analysis with the support of JMEH wrote the paper with comments provided by NC and JM

Competing interests The authors declare that they have no conflictof interest

Acknowledgements The authors would like to thank the Environ-ment Agency and the Centre for Ecology and Hydrology for accessto the data used in this paper This research is funded by the NaturalEnvironmental Research Council (NERC)

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2317

Review statement This paper was edited by Nandita Basu and re-viewed by Simone Bizzi and one anonymous referee

References

Belletti B Rinaldi M Buijse A D Gurnell A Mand Mosselman E A review of assessment methods forriver hydromorphology Environ Earth Sci 73 2079ndash2100httpsdoiorg101007s12665-014-3558-1 2015

Benda L Andras K Miller D and Bigelow P Confluenceeffects in rivers Interactions of basin scale network geom-etry and disturbance regimes Water Resour Res 40 1ndash15httpsdoiorg1010292003WR002583 2004a

Benda L Poff N L Miller D Dunne T Reeves GPess G and Pollock M The network dynamics hy-pothesis how channel networks structure riverine habi-tats Bioscience 54 413ndash427 httpsdoiorg1016410006-3568(2004)054[0413TNDHHC]20CO2 2004b

Benjamini Y and Hochberg Y Controlling the False DiscoveryRate A Practical and Powerful Approach to Multiple TestingJ Roy Stat Soc B 57 289ndash300 1995

Best J L Flow dynamics and sediment transport at river channelconfluences Birbeck University of London London 1985

Best J L The morphology of river channel con-fluences Prog Phys Geogr 10 157ndash174httpsdoiorg101177030913338601000201 1986

Best J L Flow dynamics at river channel confluences implica-tions for sediment transport and bed morphology in Recent de-velopments in fluvial sedimentology edited by Ethridge F GFlores R M and Harvey M D Spec Publ Soc Econ Pale-ontol Miner Tulsa Okla 27ndash35 1987

Biron P M Richer A Kirkbride A D Roy A G andHan S Spatial patterns of water surface topography ata river confluence Earth Surf Proc Land 27 913ndash928httpsdoiorg101002esp359 2002

Bravard J P and Gilvear D J Hydrological and geomorpho-logical structure of hydrosystems in The Fluvial Hydrosystemedited by Petts G E and Amoros C Springer Netherlands98ndash116 1996

Brierley G and Fryirs K River Styles a Geomorphic Approachto Catchment Characterization Implications for River Rehabili-tation in Bega Catchment New South Wales Australia EnvironManage 25 661ndash679 2000

Bunn S E and Arthington A H Basic principles and ecologi-cal consequences of altered flow regimes for aquatic biodiversityEnviron Manage 30 492ndash507 httpsdoiorg101007s00267-002-2737-0 2002

Church M and Kellerhals R On the statistics of grain size varia-tion along a gravel river Can J Earth Sci 15 1151ndash1160 1978

Clifford N J Hydrology the changing paradigm Prog Phys Ge-ogr 26 290ndash301 2002

Cohen P Andriamahefa H and Wasson J-G Towardsa regionalizaton of aquatic habitat distribution of meso-habitats at the scale of a large basin Regul RiversRes Manage 14 391ndash404 httpsdoiorg101002(SICI)1099-1646(19980910)145lt391AID-RRR513gt30CO2-W 1998

Davenport A J Gurnell A M and Armitage P D Habitat sur-vey and classification of urban rivers River Res Appl 20 687ndash704 httpsdoiorg101002rra785 2004

Davies N M Norris R H and Thoms M C Prediction and as-sessment of local stream habitat features using large-scale catch-ment characteristics Freshwater Biol 45 343ndash369 2000

Depettris C Mendiondo E M Neiff J and Rohrmann H Flooddefence strategy at the confluence of the Paranaacute-Paraguay riversProc Int Symp Flood Def Kassel He C31 C40 2000

Dollar E James C Rogers K and Thoms M A framework forinterdisciplinary understanding of rivers as ecosystems Geomor-phology 89 147ndash162 2007

Dovers S R and Day D G Australian rivers and statute law En-viron Plan Law J 5 90ndash108 1988

Downs P W Gregory K J and Brookes A How integratedis river basin management Environ Manage 15 299ndash309httpsdoiorg101007BF02393876 1991

Emery J C Gurnell A M Clifford N J and Petts G E Char-acteristics and controls of gravel-bed riffles An analysis of datafrom the river-habitat survey Water Environ J 18 210ndash216httpsdoiorg101111j1747-65932004tb00535x 2004

Evans I S and Minaacuter J A classification of geomorphometric vari-ables in International Geom-orphometry 2011 Geomoprhome-tryorg Redlabds CA 105ndash108 2011

Fausch K D Torgersen C E Baxter C V Li H WView C The O F Is R To N How U Among ISet S For C and Fishes S Landscapes to riverscapesBridging the gap between research and conservation of streamfishes Bioscience 52 483ndash498 httpsdoiorg1016410006-3568(2002)052[0483LTRBTG]20CO2 2002

Glickman M E Rao S R and Schultz M R False discov-ery rate control is a recommended alternative to Bonferroni-typeadjustments in health studies J Clin Epidemiol 67 850ndash857httpsdoiorg101016jjclinepi201403012 2014

Gupta V K and Mesa O J Runoff generation and hydro-logic response via channel network geomorphology - Re-cent progress and open problems J Hydrol 102 3ndash28httpsdoiorg1010160022-1694(88)90089-3 1988

Gupta V K and Waymire E D On the formulation of an analyt-ical approach to hydrologic response and similarity at the basinscale J Hydrol 65 95ndash123 1983

Gupta V K Waymire E D and Rodriguez-Iturbe I On scalesgravity and network structure in basin runoff in Scale problemsin hydrology Springer Netherlands 159ndash184 1986

Gurnell A M Rinaldi M Belletti B Bizzi S Blamauer BBraca G Buijse A D Bussettini M Camenen B ComitiF Demarchi L Garciacutea de Jaloacuten D Gonzaacutelez del TaacutenagoM Grabowski R C Gunn I D M Habersack H Hen-driks D Henshaw A J Kloumlsch M Lastoria B LatapieA Marcinkowski P Martiacutenez-Fernaacutendez V Mosselman EMountford J O Nardi L Okruszko T OrsquoHare M T PalmaM Percopo C Surian N van de Bund W Weissteiner Cand Ziliani L A multi-scale hierarchical framework for devel-oping understanding of river behaviour to support river manage-ment Aquat Sci 78 1ndash16 httpsdoiorg101007s00027-015-0424-5 2016

Harvey G L Gurnell A M and Clifford N J Characterisationof river reaches The influence of rock type Catena 76 78ndash88httpsdoiorg101016jcatena200809010 2008

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2310 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

Table 1 RHS variables calculated from RHS observations

RHS variable Calculation from RHS observations Units

Habitat Quality A score indicating the degree of naturalness and diversity of the HQA scaleAssessment riparian zone based on observations in the reach of flow types(HQA) substrate channel and bank features riparian vegetation etc

Habitat A score indicating the degree of artificial modification of the channel HMS scaleModification based on observations in the reach of reinforcements re-sectioningScore (HMS) embankments weed cutting realignment culverts dams weirs etc

Sediment size =(minus8middotBOminus7middotCOminus35middotGPminus15middotSA+15middotSI+9middotCL)

(BO+CO+GP+SA+SI+CL) ApproxBO (boulder) CO (cobble) GP (gravelndashpebble) SA (sand) SI (silt) φ unitsand CL (clay) represent the number of spot checks allocated to eachsediment size class

Flow-type speed =(0middotDR+1middotNP+2middotUP+3middotSM+4middotRP+5middotUW+6middotBW+7middotCF+8middotCH+9middotFF)

(DR+NP+UP+SM+RP+UW+BW+CF+CH+FF) Flow typeDR (dry) NP (no perceptible flow) UP (upwelling) SM (smooth) speed scaleRP (rippled) UW (unbroken wave) BW (broken wave) CF (chaoticflow) CH (chute) FF (free-fall) represent the number of spot checksallocated to each flow speed class

At each site over 100 features are recorded along a 500 mreach with 10 ldquospot-checkrdquo surveys conducted every 50 mand a ldquosweep-uprdquo survey conducted across the whole reach(see Raven et al 1996 for details) Variables of interest thatare hypothesized to be impacted by network structure can becalculated from the RHS observations (Table 1)

The Habitat Quality Assessment (HQA) and Habitat Mod-ification Score (HMS) variables are both amalgamations ofRHS observations with individual features given a score de-rived by expert opinion (see Raven et al 1998 for more de-tails) The scoring systems are subjective but HQA and HMSprovide overviews of the channel condition that are widelyapplied for regulatory compliance The scores are thereforeincluded in this study to reflect how they may be impactedby network topology

The remaining RHS variables are calculated directly fromRHS observations and so are more objective Sediment sizeis calculated as a reach average of spot-check observationsusing the same method as previous studies (Davenport et al2004 Emery et al 2004 Harvey et al 2008) Flow-typespeed was calculated in the same manner as sediment sizeusing values of flow which represent an approximate flowvelocity gradient defined in Davenport et al (2004) Thesevariables were chosen to reflect dominant geomorphic pro-cesses occurring in each reach and due to the prominence ofsediment size and flow type in defining physical habitats forinstream biota (Rowntree and Wadeson 1996) The variablesare likely to be impacted by the density of the river networkas they have been shown to be impacted by individual con-fluences For example channels are shown to become moregeomorphologically heterogeneous (Benda et al 2004a) andsubstrate size has been shown to coarsen at confluences (Riceet al 2001) Surface flow types are also likely to become

more diverse at confluences as the convergence of channelscreates a number of different flow environments (Best 1985)that result in different water-surface topographies (Biron etal 2002)

It must be noted that the RHS dataset was collected forregulatory compliance and was not directly intended for sci-entific enquiry Therefore there is a limitation in the amountof detail that can be extracted about physical processes as theobservations recorded are an average across a 500 m reachDespite this the wide spatial coverage of the dataset makesit a powerful tool allowing analysis to be conducted acrossmultiple catchments with differing characteristics

For each distance and elevation interval created by the net-work topology metrics descriptive statistics (mean median90th and 10th percentiles) of each RHS variable are calcu-lated Despite the RHS sampling strategy (Jeffers 1998b) bi-asing site selection towards less dense areas of the networkmost distance and elevation intervals contained RHS sites(with only some low-density intervals not containing RHSsites) This method is designed to account for natural varia-tion and modification at individual RHS sites in order to as-sess broad patterns of reach characteristics at the catchmentlevel

24 Statistical analysis

Analysis is conducted with all catchments combined into asingle population to identify overall trends across all catch-ments a method used in previous broad-scale studies Theanalysis is also split into individual catchments to identifyhow the relationship between network topology metrics andriver reach characteristics differed between catchments

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2311

Kendall correlation

Correlation tests are used to determine the strength and di-rection of the association between the descriptive statisticsof the RHS variables and distance network density eleva-tion network density and stream order to ascertain how reachcharacteristics respond to network topology Kendallrsquos cor-relation method was used as the variables have non-normaldistributions a small sample size and tied data values (Helseland Hirsch 2002) The effect size of Kendallrsquos τ is lowerthan other correlation methods with strong correlations oc-curring with τ values greater than 07 (Helsel and Hirsch2002)

As multiple correlations are conducted false discoveryrate (Benjamini and Hochberg 1995) corrections were ap-plied to the p-values produced from the Kendall correlationsto reduce the risk of type I error The false discovery ratemethod has been found to be more powerful than other pro-cedures for controlling for multiple tests (Glickman et al2014)

3 Results

31 Differences in network topology metrics betweencatchments

The topological metrics developed in this study show the in-ternal structure of the network for each catchment The sep-aration of the catchments into distance and elevation inter-vals emphasizes different features of the catchment The dis-tance intervals (Fig 2a) are arranged longitudinally withinthe catchment highlighting sub-basins within each catch-ment The elevation intervals (Fig 2b) have a radial arrange-ment centering around the incised main channel of eachcatchment

Distance network density is higher in the Eden (284plusmn103) and Tamar (441plusmn 219) compared to the Avon (47plusmn19) and Wensum (68plusmn 07) interesting as the Tamar is thesmallest catchment by area The shape of the distance net-work density function reflects the internal shape of the net-work (Fig 3a) For example the Tamar has a peaked den-sity distribution reflecting the circular shape of the catch-ment such that the majority of links are at 55 ndash65 dis-tance from the catchment outlet The Avon and Eden reflectsimilar internal network structures both exhibiting a bimodaldensity distribution despite the differences in the number oflinks in the catchments The density distribution of the Wen-sum has a more complex internal distribution of links withmultiple peaks in density

Elevation network density (Fig 3b) shows similar densitydistribution shapes to distance network density with a uni-modal distribution for the Tamar and multi-modal distribu-tions in the other catchments In contrast to distance networkdensity elevation network density shows the highest peaks in

density in the Tamar (103plusmn 50) and Wensum (101plusmn 43)despite the Wensum having the lowest network elevation andhas lower values in the Avon (34plusmn10) and Eden (58plusmn25)The peak densities in the Wensum occur at similar positionsin the elevation and distance intervals whereas the peaks inthe other catchments are negatively skewed showing the net-work density is highest at lowndashmid elevations

Nearly half of the links in each catchment are classified asfirst-order streams and the number of links declines exponen-tially towards the highest orders in all four catchments Thereare weak correlations (τ =minus003 to 017) between the threenetwork topology metrics distance network density eleva-tion network density and stream order This suggests that themetrics are independent and reflect different aspects of rivernetwork topology

32 River characteristic relationships with networktopology metrics

RHS sites in the Avon and Wensum have similar river char-acteristics Both have lower habitat quality high modifica-tion fine sediment and slower flow types than the Eden andTamar When all catchments are combined there are signif-icant (p lt 005 after p-value correction) correlations withmost descriptive statistics for each RHS variable and dis-tance network density (Fig 4) There are consistently pos-itive correlations with HQA and flow-type speed and neg-ative correlations with HMS and sediment size There arefewer and weaker significant correlations with elevation net-work density (Fig 4) The only significant correlations withstream order are with HMS which shows a negative correla-tion (Fig 4)

There are numerous significant correlations between thenetwork topology metrics and RHS variables for individualcatchments many of which were also shown to be signifi-cant after the correction to the p-value The results show thatcatchments have different responses to the network topol-ogy metrics of distance network density and elevation net-work density Distance network density only has significantcorrelations with the regulatory scoring variables (HQA andHMS) in the Eden and Tamar (Fig 4) Elevation networkdensity however has a wider array of significant correlationswith the scoring variables particularly HQA which showssubtle peaks and troughs reflecting the distribution of bothnetwork density metrics (Fig 3a and b) HMS shows mostlynegative correlations mainly with elevation network densityapart from the Eden which has significant positive correla-tion across all HMS descriptive statistics (Fig 4) Visually10th percentile HMS is most variable to network density withpeaks and troughs responding to the network density distri-butions (Fig 3a and b)

For individual RHS features the response to networktopology varies between catchments (Fig 4) The Avon hasnegative correlations between flow-type speed and distancenetwork density with an evident drop in 10th percentile

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2312 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

Figure 3 Network topology metrics (a) distance network density and (b) elevation network density Descriptive statistics of each RHSvariable over (a) distance and (b) elevation for each catchment with smooth loess lines to indicate trend Network topology metrics arenormalized between 0 and 1 and HMS score is logarithmically transformed for display purposes

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2313

Figure 4 Summary of correlations between distance network density elevation network density stream order and RHS variables for allcatchments combined and each individual catchment Significance of correlation is indicated by point size with the largest points significantpost p-value correction Effect size (Kendallrsquos τ ) is indicated by colour No correlation was possible between stream order and 10th percentilesediment size in the Avon due to no variation in the RHS variable

flow-type speed associated with peaks in network density(Fig 3a) The Eden and Tamar however have positive cor-relations with mean and 90th percentile flow-type speed fordistance network density but negative correlations with me-dian and 10th percentile elevation network density The Wen-sum shows positive correlations between flow-type speed andelevation network density Sediment size has a consistent re-sponse to distance network density with the Eden and Wen-sum showing negative correlations with the sediment size(Fig 4) For elevation network density the Avon shows neg-ative correlations with sediment size whereas the Tamar andWensum show positive correlations (Fig 4)

There were few significant correlations between stream or-der and the RHS variables in individual catchments (Fig 4)The only significant correlation after p-value correction iswith 90th percentile HMS in the Wensum which shows astrong negative relationship

4 Discussion

41 A new approach to utilizing network topology incatchment-level analysis

Network density metrics represent an alternative approach toaccount for network topology in catchment-level studies op-timizing the width dimension of the network (or the num-

ber of links in the network) as opposed to the commonplacestream-order metric which only reflects the longitudinal po-sition of links in a network (Fig 1) This study demonstratesthat two topology metrics can be calculated simply from aDTM with GIS and using a broad-scale analysis of river at-tributes can be used to investigate the functional processeswithin catchments

While the two network density metrics have similar forms(ie forms are consistently unimodal or multi-modal) thespatial configuration of the distance and elevation intervalsused in the calculation of network density varies and mayimpact the effectiveness of each topological metric Distancenetwork density separates the catchment into intervals basedon distance which spread upstream from the outlet (Fig 2a)reflecting natural sub-basins within the fractal structure of thecatchment (Lashermes and Foufoula-Georgiou 2007) Thisdiffers from elevation network density which separates thecatchment into intervals based on elevation which radiate outfrom the main channel of the network (Fig 2b) The config-uration means that distance intervals contain streams that arein closer proximity to one another rather than the more distalconfiguration created by the elevation intervals suggesting adegree of spatial dependency in river functioning This hasbeen highlighted in previous studies where spatial networkstructure has a stronger influence on some in-channel pro-cesses than predictor variables such as elevation (Steel et al

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2314 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

2016) However elevation intervals contain RHS sites thatalthough they may be distal may have similar properties aselevation has been strongly related to RHS variables includ-ing flow type substrate etc in a number of studies (Jeffers1998a Naura et al 2016 Vaughan et al 2013)

42 Impacts of network topology on rivercharacteristics

River characteristics are assessed using the RHS dataset Theobservations made by the RHS dataset (Table 1) cannot of-fer the level of detail regarding geomorphological processthat river classifications which consider multiple scales canoffer (eg Brierley and Fryirs 2000 Gurnell et al 2016)While process-based classifications are preferable broad-scale monitoring datasets such as the River Habitat Surveymay still be useful when combined with map-derived datato explore controls on river characteristics (Harvey et al2008 Naura et al 2016 Vaughan et al 2013) Howeverthere are biases in RHS data collection an inherent limitationwhen using existing datasets (Vaughan and Ormerod 2010)specifically the standardized survey length of 500 m reachthat will capture differing amounts of natural variability de-pending on the size of the river While this must be notedthere are few significant correlations between river charac-teristics identified with stream order (Fig 4) which suggeststhat channel size is not influencing the RHS variables to agreat degree in these catchments

In this study it is anticipated that sites in high networkdensity areas will have higher levels of habitat diversity asindicated by previous studies of confluences and networks(Benda et al 2004a Best 1985 Rice 2017) in turn in-creasing mean sediment size and flow-type speed comparedto sites in low-density areas The results of the correlationsbetween distance network density and river characteristicswhen all catchments are combined support this hypothesiswith greater HQA flow-type speed and coarser sedimentsizes observed in areas with high distance network density(Fig 4)

For individual catchments elevation network density in-duces a stronger positive HQA response across all catch-ments than distance network density (Fig 4) This supportsthe evidence that individual confluences (Rice et al 2006)and high densities of confluences increase physical hetero-geneity within the river network (Benda et al 2004b Rice2017) However flow-type speed and sediment size responddifferently to network density in individual catchments

Slower flow types are observed in high network densityareas of the Avon and Tamar whereas faster flow types areobserved in high-elevation network density areas of the Wen-sum Individual confluences are shown to create numeroushigh- and low-speed flow environments (Best 1987) thatmay be observed in surface water topography (Biron et al2002) It was expected that the introduction of the additionalflow types by a high density of confluences in relatively low-

energy rivers would increase mean reach flow-type speedhowever the correlation analysis (Fig 4) suggests that insome catchments mean flow-type speed is reduced

Sediment size response also shows variation betweencatchments Sediment size is coarser in network-dense ar-eas of the Avon and Eden as expected but is finer in boththe Tamar and Wensum (Fig 4) The evidence from high-energy rivers shows that sediment becomes coarser down-stream and finer upstream of certain confluences (Benda etal 2004a Rice 1998) and in this case high numbers of con-fluences were expected to increase mean sediment size of thereach However the impact on sediment size is dependent onthe sediment calibre of the incoming tributary being higherthan the main channel with enough energy to transport thecoarse sediment to the confluence for numerous tributariesin an area The Tamar and Wensum have different rangesof sediment sizes with the Tamar having coarser sedimentthan the Wensum on average (Fig 3) This implies that trib-utaries in the Tamar may be energy limited not transportingcoarse sediments to confluences and the Wensum may suf-fer from inputs of fine sediments from the high percentageof arable land that concentrates in network-dense areas Thishas before been observed in a low-energy modified catch-ment where anthropogenic modifications in tributaries re-duced coarse sediment and flow capacity causing either lim-ited confluence impact or localized sediment fining (Singer2008)

Others have related the capacity of confluences to alterreach features to the morphometry of catchments with largerand more circular catchments containing a higher percentageof confluences that have a significant impact (Rice 2017)Based on this theory the Eden and Tamar are likely to havethe greatest impact as they are the most circular and steepestof the four catchments (although the Tamar is the smallestby area) Yet there is no clear pattern indicating that thesecatchments respond differently than the Avon and Wensum(Fig 4) with catchments responding differently to differ-ent variables This perhaps suggests that rather than networkdensity having a directional impact on factors such as flow-type speed and sediment size it has an impact on overall het-erogeneity at the reach level as suggested by previous stud-ies and that specific directional change occurs at the sub-reach level

An increase in channel modification is also hypothesizeddue to the increase in flood peak downstream from conflu-ences (Depettris et al 2000) and the scour and erosion as-sociated with confluence junctions (Best 1986) potentiallyincreasing the need for bed and bank protection The corre-lations between distance network density and HMS when allcatchments are combined undermine this hypothesis show-ing less channel modification where distance network den-sity is higher (Fig 4) There were few significant correlationswith individual catchments but HMS in the Eden was consis-tently observed to be higher in network-dense areas (Fig 4)This may be due to the Edenrsquos high elevation and steep to-

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2315

pography inducing a higher energy environment where scourand erosion processes in areas with high numbers of conflu-ences would be more likely to be present

Differences in RHS variable responses also differed be-tween the descriptive statistics considered Often the ex-tremes 90th and 10th percentiles showed more significantand stronger correlations than the mean or median (Fig 4)This may reflect findings from previous studies which sug-gest that not all confluences cause reach-scale changes (Rice1998) that perhaps the changes to river character induced bycertain confluences only influence certain reaches whereasothers are left unaffected creating less pronounced responsesin the mean and median of variables External factors mayalso influence this trend for example 10th percentile HMSvisually responds dramatically to network density metrics(Fig 3a and b) compared to the other descriptive statisticsThis suggests that the most natural sites (ie with the low-est HMS score) are responding differently to network den-sity with the most natural sites having less modificationin network-dense areas whereas less natural sites becomemore modified in network-dense areas This reflects the HQAscore results which visually (Fig 3a and b) and statistically(Fig 4) vary with distance network density except for the10th percentile These sites with the lowest habitat qualityand naturalness are likely influenced by anthropogenic fac-tors that are independent of network density reducing habitatquality scores at impacted sites

The two network density metrics have differing impacts onriver characteristics While distance network density showsconsistently significant correlations when all catchments arecombined individual catchments respond more frequentlyand more strongly to elevation network density (Fig 4)This may be because there is a dramatic split in distancenetwork density values between the more upland drainage-dense catchments Eden and Tamar than the lowland chalklow-drainage catchments Avon and Wensum The combinedcorrelation will therefore in part reflect the difference be-tween the catchments which have different ranges of rivercharacteristics (Fig 3) This is not the case for elevation net-work density which has higher density values in the Tamarand Wensum than the Avon and Eden so therefore the char-acteristics of the catchments will have less bearing on thecombined correlation However there are patterns identifiedwith distance network density in individual catchments thatare not present with elevation network density increasedflow-type speed and sediment size in the Eden and reducedflow-type speed in the Avon with network density (Fig 4)which show its usefulness

The results suggest that the distance network density andelevation network density metrics quantify different dimen-sions of network topology which are shown to exhibit func-tionally meaningful patterns for river reach characteristicsbased on the correlation results Perhaps within catchmentselevation network density provides the more powerful met-ric for individual catchments but distance network density

better accounts for the drainage density of the catchmentsallowing it to be applied across multiple catchments

43 Comparison of stream order to network densitymetrics

The classic method of accounting for network topologystream order is critiqued for failing to represent disconti-nuities in the network and simplifying the network into agradient The number of links in different stream orders isconsistent across all catchments not reflecting the internalstructure of the network or the variety between catchmentsthat is achieved by the distance network density and eleva-tion network density metrics The analysis of the two net-work density metrics presented in this paper shows that dis-tances from source and elevation are not mutually exclusive(Fig 2a and b) contrary to the stream-order metric whichrepresents streams as upstream to downstream or upland tolowland

Stream order has few significant correlations with manyof the river characteristics considered in this study Negativecorrelations with HMS in the Wensum were statistically sig-nificant post p-value correction likely driving the significantrelationship with all catchments combined for this variable(Fig 4) This suggests that modification is greater upstreamin the Wensum contrary to ideas that downstream reachesmay show greater anthropogenic modification Intense agri-cultural land use in the upper reaches of the Wensum islikely to be the cause of the high HMS scores upstreamThe lack of significant correlations suggests that stream or-der and therefore an upstream-to-downstream gradient arenot the predominant pattern in river characteristics despitethe description of such a gradient by geomorphic (Schumm1977) and ecological frameworks (Vannote et al 1980)This is surprising as distance and elevation which both re-flect the upstream-to-downstream gradient have proven tobe important factors in previous studies explaining patternsof RHS features at a national level (Jeffers 1998a Vaughanet al 2013) This implies that upstream-to-downstream gra-dient may not sufficiently reflect patterns of river charac-teristics through the river network within individual catch-ments Others have also found that stream order has weakand inconsistent relationships with biodiversity patterns inriver systems arguing that the topological measure has no di-rect mechanistic control on biodiversity (Vander Vorste et al2017) Instead this study finds that the network density met-rics are a more powerful metric which conceptually providean improved method of accounting for the impacts of net-work topology on the fluvial system exhibiting relationshipswith river characteristics particularly habitat quality score(Fig 4)

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2316 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

44 Applicability of network topology metrics todifferent environments

Much of the seminal work on network and confluence im-pacts (eg the Network Dynamics Hypothesis Benda et al2004b and Link Discontinuity Concept Rice et al 2001) isconducted in natural highly erosive catchments with first-hand empirical measurements However in an age whenrivers and their catchments are increasingly altered by an-thropogenic modification (Meybeck 2003) contemporarystudies must not only aim to expand knowledge but also findmethods of transferring knowledge to many increasingly al-tered catchments (Clifford 2002)

The catchments selected by this study are more greatlymodified and although they reflect a range of fluvial envi-ronments in England are lower-energy catchments than theseminal studies Benda et al (2004a) suggest that confluenceeffects in less active landscapes would be subdued comparedto highly erosive landscapes but the evidence presented heredemonstrates the utility of evaluating network topologicalstructure in studies on catchment-level effects in any typeof fluvial environment including those with low-energy andwidespread anthropogenic changes The response of someriver characteristics varied between catchments observationsof flow-type speed sediment size and modifications showeddifferent responses to network density in different catch-ments This suggests that the functional effect of these topo-logical metrics is catchment dependent and likely is influ-enced by external catchment characteristics such as land usenot considered in this study although impact did not appearto vary with catchment topography or circularity as has beenshown in prior studies (Benda et al 2004a Rice 2017) Thisshould be explored further in future research to enable rec-ommendations to be made regarding where and how reachesmay respond to network density The response of habitatquality score was however consistent across catchments andbetween metrics showing that habitat quality is greater inareas with high network density (Fig 4) as hypothesizedby the Network Dynamics Hypothesis (Benda et al 2004b)and demonstrated by studies on individual catchments (Rice2017 Rice et al 2006)

The methods presented in this paper are designed to be im-plemented in any catchment with a dendritic network struc-ture The topology metrics can easily be calculated from anydendritic network with DTM data using GIS and comparedto any site-scale data This study uses regulatory monitoringdatasets so that analysis is targeted to assessment scores andphysical features of interest to river managers Also the highvolume and wide spatial extent of data available from regu-latory sources allow for between-catchment comparisons

5 Conclusions

Although appreciation of catchment-level effects is consid-ered the epitome of understanding river functioning a keycomponent of the catchment ndash the river network ndash is over-looked and oversimplified by catchment-level studies Thisstudy finds that river network density plays a role in struc-turing the distribution river characteristics throughout thecatchment offering more detailed explanation than the clas-sic stream-order metric Network-dense areas are generallyfound to have higher habitat quality and diversity but modi-fication flow-type speed and sediment size show different re-sponses in different catchments This study suggests two po-tential reasons for this (1) there is evidence that confluencesin the river network increase diversity as is observed in thisstudy so the direction of mean river characteristic responsemay not be consistent and (2) there may be external factorssuch as sediment availability land cover and anthropogenicmodification that alter the direction of mean river characteris-tic response This paper demonstrates the functional responseof river characteristics to network topology and suggests thatthe inclusion of network topology in catchment-level studieswould add a layer of function-based understanding to suchstudies linking reaches to their catchments

The broad-scale methodology adopted by this study allowsthe network density metrics which are easily extracted fromopen-source data using GIS software to be compared to anyregulatory dataset The use of regulatory datasets allows notonly for analysis over a wider spatial extent but also for moreapplicable results for regulatory bodies Therefore the inter-disciplinary approach to characterizing network topology canbe applied efficiently and effectively to capture catchment-level impacts on reach-level functioning in any catchmentacross the globe

Data availability River Habitat Survey data are freely availablefrom the Environment Agency England River network data andelevation data are available from the Centre for Ecology and Hy-drology UK

Author contributions EH and NC conceptualised and designed thestudy EH performed the data analysis with the support of JMEH wrote the paper with comments provided by NC and JM

Competing interests The authors declare that they have no conflictof interest

Acknowledgements The authors would like to thank the Environ-ment Agency and the Centre for Ecology and Hydrology for accessto the data used in this paper This research is funded by the NaturalEnvironmental Research Council (NERC)

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2317

Review statement This paper was edited by Nandita Basu and re-viewed by Simone Bizzi and one anonymous referee

References

Belletti B Rinaldi M Buijse A D Gurnell A Mand Mosselman E A review of assessment methods forriver hydromorphology Environ Earth Sci 73 2079ndash2100httpsdoiorg101007s12665-014-3558-1 2015

Benda L Andras K Miller D and Bigelow P Confluenceeffects in rivers Interactions of basin scale network geom-etry and disturbance regimes Water Resour Res 40 1ndash15httpsdoiorg1010292003WR002583 2004a

Benda L Poff N L Miller D Dunne T Reeves GPess G and Pollock M The network dynamics hy-pothesis how channel networks structure riverine habi-tats Bioscience 54 413ndash427 httpsdoiorg1016410006-3568(2004)054[0413TNDHHC]20CO2 2004b

Benjamini Y and Hochberg Y Controlling the False DiscoveryRate A Practical and Powerful Approach to Multiple TestingJ Roy Stat Soc B 57 289ndash300 1995

Best J L Flow dynamics and sediment transport at river channelconfluences Birbeck University of London London 1985

Best J L The morphology of river channel con-fluences Prog Phys Geogr 10 157ndash174httpsdoiorg101177030913338601000201 1986

Best J L Flow dynamics at river channel confluences implica-tions for sediment transport and bed morphology in Recent de-velopments in fluvial sedimentology edited by Ethridge F GFlores R M and Harvey M D Spec Publ Soc Econ Pale-ontol Miner Tulsa Okla 27ndash35 1987

Biron P M Richer A Kirkbride A D Roy A G andHan S Spatial patterns of water surface topography ata river confluence Earth Surf Proc Land 27 913ndash928httpsdoiorg101002esp359 2002

Bravard J P and Gilvear D J Hydrological and geomorpho-logical structure of hydrosystems in The Fluvial Hydrosystemedited by Petts G E and Amoros C Springer Netherlands98ndash116 1996

Brierley G and Fryirs K River Styles a Geomorphic Approachto Catchment Characterization Implications for River Rehabili-tation in Bega Catchment New South Wales Australia EnvironManage 25 661ndash679 2000

Bunn S E and Arthington A H Basic principles and ecologi-cal consequences of altered flow regimes for aquatic biodiversityEnviron Manage 30 492ndash507 httpsdoiorg101007s00267-002-2737-0 2002

Church M and Kellerhals R On the statistics of grain size varia-tion along a gravel river Can J Earth Sci 15 1151ndash1160 1978

Clifford N J Hydrology the changing paradigm Prog Phys Ge-ogr 26 290ndash301 2002

Cohen P Andriamahefa H and Wasson J-G Towardsa regionalizaton of aquatic habitat distribution of meso-habitats at the scale of a large basin Regul RiversRes Manage 14 391ndash404 httpsdoiorg101002(SICI)1099-1646(19980910)145lt391AID-RRR513gt30CO2-W 1998

Davenport A J Gurnell A M and Armitage P D Habitat sur-vey and classification of urban rivers River Res Appl 20 687ndash704 httpsdoiorg101002rra785 2004

Davies N M Norris R H and Thoms M C Prediction and as-sessment of local stream habitat features using large-scale catch-ment characteristics Freshwater Biol 45 343ndash369 2000

Depettris C Mendiondo E M Neiff J and Rohrmann H Flooddefence strategy at the confluence of the Paranaacute-Paraguay riversProc Int Symp Flood Def Kassel He C31 C40 2000

Dollar E James C Rogers K and Thoms M A framework forinterdisciplinary understanding of rivers as ecosystems Geomor-phology 89 147ndash162 2007

Dovers S R and Day D G Australian rivers and statute law En-viron Plan Law J 5 90ndash108 1988

Downs P W Gregory K J and Brookes A How integratedis river basin management Environ Manage 15 299ndash309httpsdoiorg101007BF02393876 1991

Emery J C Gurnell A M Clifford N J and Petts G E Char-acteristics and controls of gravel-bed riffles An analysis of datafrom the river-habitat survey Water Environ J 18 210ndash216httpsdoiorg101111j1747-65932004tb00535x 2004

Evans I S and Minaacuter J A classification of geomorphometric vari-ables in International Geom-orphometry 2011 Geomoprhome-tryorg Redlabds CA 105ndash108 2011

Fausch K D Torgersen C E Baxter C V Li H WView C The O F Is R To N How U Among ISet S For C and Fishes S Landscapes to riverscapesBridging the gap between research and conservation of streamfishes Bioscience 52 483ndash498 httpsdoiorg1016410006-3568(2002)052[0483LTRBTG]20CO2 2002

Glickman M E Rao S R and Schultz M R False discov-ery rate control is a recommended alternative to Bonferroni-typeadjustments in health studies J Clin Epidemiol 67 850ndash857httpsdoiorg101016jjclinepi201403012 2014

Gupta V K and Mesa O J Runoff generation and hydro-logic response via channel network geomorphology - Re-cent progress and open problems J Hydrol 102 3ndash28httpsdoiorg1010160022-1694(88)90089-3 1988

Gupta V K and Waymire E D On the formulation of an analyt-ical approach to hydrologic response and similarity at the basinscale J Hydrol 65 95ndash123 1983

Gupta V K Waymire E D and Rodriguez-Iturbe I On scalesgravity and network structure in basin runoff in Scale problemsin hydrology Springer Netherlands 159ndash184 1986

Gurnell A M Rinaldi M Belletti B Bizzi S Blamauer BBraca G Buijse A D Bussettini M Camenen B ComitiF Demarchi L Garciacutea de Jaloacuten D Gonzaacutelez del TaacutenagoM Grabowski R C Gunn I D M Habersack H Hen-driks D Henshaw A J Kloumlsch M Lastoria B LatapieA Marcinkowski P Martiacutenez-Fernaacutendez V Mosselman EMountford J O Nardi L Okruszko T OrsquoHare M T PalmaM Percopo C Surian N van de Bund W Weissteiner Cand Ziliani L A multi-scale hierarchical framework for devel-oping understanding of river behaviour to support river manage-ment Aquat Sci 78 1ndash16 httpsdoiorg101007s00027-015-0424-5 2016

Harvey G L Gurnell A M and Clifford N J Characterisationof river reaches The influence of rock type Catena 76 78ndash88httpsdoiorg101016jcatena200809010 2008

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2311

Kendall correlation

Correlation tests are used to determine the strength and di-rection of the association between the descriptive statisticsof the RHS variables and distance network density eleva-tion network density and stream order to ascertain how reachcharacteristics respond to network topology Kendallrsquos cor-relation method was used as the variables have non-normaldistributions a small sample size and tied data values (Helseland Hirsch 2002) The effect size of Kendallrsquos τ is lowerthan other correlation methods with strong correlations oc-curring with τ values greater than 07 (Helsel and Hirsch2002)

As multiple correlations are conducted false discoveryrate (Benjamini and Hochberg 1995) corrections were ap-plied to the p-values produced from the Kendall correlationsto reduce the risk of type I error The false discovery ratemethod has been found to be more powerful than other pro-cedures for controlling for multiple tests (Glickman et al2014)

3 Results

31 Differences in network topology metrics betweencatchments

The topological metrics developed in this study show the in-ternal structure of the network for each catchment The sep-aration of the catchments into distance and elevation inter-vals emphasizes different features of the catchment The dis-tance intervals (Fig 2a) are arranged longitudinally withinthe catchment highlighting sub-basins within each catch-ment The elevation intervals (Fig 2b) have a radial arrange-ment centering around the incised main channel of eachcatchment

Distance network density is higher in the Eden (284plusmn103) and Tamar (441plusmn 219) compared to the Avon (47plusmn19) and Wensum (68plusmn 07) interesting as the Tamar is thesmallest catchment by area The shape of the distance net-work density function reflects the internal shape of the net-work (Fig 3a) For example the Tamar has a peaked den-sity distribution reflecting the circular shape of the catch-ment such that the majority of links are at 55 ndash65 dis-tance from the catchment outlet The Avon and Eden reflectsimilar internal network structures both exhibiting a bimodaldensity distribution despite the differences in the number oflinks in the catchments The density distribution of the Wen-sum has a more complex internal distribution of links withmultiple peaks in density

Elevation network density (Fig 3b) shows similar densitydistribution shapes to distance network density with a uni-modal distribution for the Tamar and multi-modal distribu-tions in the other catchments In contrast to distance networkdensity elevation network density shows the highest peaks in

density in the Tamar (103plusmn 50) and Wensum (101plusmn 43)despite the Wensum having the lowest network elevation andhas lower values in the Avon (34plusmn10) and Eden (58plusmn25)The peak densities in the Wensum occur at similar positionsin the elevation and distance intervals whereas the peaks inthe other catchments are negatively skewed showing the net-work density is highest at lowndashmid elevations

Nearly half of the links in each catchment are classified asfirst-order streams and the number of links declines exponen-tially towards the highest orders in all four catchments Thereare weak correlations (τ =minus003 to 017) between the threenetwork topology metrics distance network density eleva-tion network density and stream order This suggests that themetrics are independent and reflect different aspects of rivernetwork topology

32 River characteristic relationships with networktopology metrics

RHS sites in the Avon and Wensum have similar river char-acteristics Both have lower habitat quality high modifica-tion fine sediment and slower flow types than the Eden andTamar When all catchments are combined there are signif-icant (p lt 005 after p-value correction) correlations withmost descriptive statistics for each RHS variable and dis-tance network density (Fig 4) There are consistently pos-itive correlations with HQA and flow-type speed and neg-ative correlations with HMS and sediment size There arefewer and weaker significant correlations with elevation net-work density (Fig 4) The only significant correlations withstream order are with HMS which shows a negative correla-tion (Fig 4)

There are numerous significant correlations between thenetwork topology metrics and RHS variables for individualcatchments many of which were also shown to be signifi-cant after the correction to the p-value The results show thatcatchments have different responses to the network topol-ogy metrics of distance network density and elevation net-work density Distance network density only has significantcorrelations with the regulatory scoring variables (HQA andHMS) in the Eden and Tamar (Fig 4) Elevation networkdensity however has a wider array of significant correlationswith the scoring variables particularly HQA which showssubtle peaks and troughs reflecting the distribution of bothnetwork density metrics (Fig 3a and b) HMS shows mostlynegative correlations mainly with elevation network densityapart from the Eden which has significant positive correla-tion across all HMS descriptive statistics (Fig 4) Visually10th percentile HMS is most variable to network density withpeaks and troughs responding to the network density distri-butions (Fig 3a and b)

For individual RHS features the response to networktopology varies between catchments (Fig 4) The Avon hasnegative correlations between flow-type speed and distancenetwork density with an evident drop in 10th percentile

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2312 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

Figure 3 Network topology metrics (a) distance network density and (b) elevation network density Descriptive statistics of each RHSvariable over (a) distance and (b) elevation for each catchment with smooth loess lines to indicate trend Network topology metrics arenormalized between 0 and 1 and HMS score is logarithmically transformed for display purposes

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2313

Figure 4 Summary of correlations between distance network density elevation network density stream order and RHS variables for allcatchments combined and each individual catchment Significance of correlation is indicated by point size with the largest points significantpost p-value correction Effect size (Kendallrsquos τ ) is indicated by colour No correlation was possible between stream order and 10th percentilesediment size in the Avon due to no variation in the RHS variable

flow-type speed associated with peaks in network density(Fig 3a) The Eden and Tamar however have positive cor-relations with mean and 90th percentile flow-type speed fordistance network density but negative correlations with me-dian and 10th percentile elevation network density The Wen-sum shows positive correlations between flow-type speed andelevation network density Sediment size has a consistent re-sponse to distance network density with the Eden and Wen-sum showing negative correlations with the sediment size(Fig 4) For elevation network density the Avon shows neg-ative correlations with sediment size whereas the Tamar andWensum show positive correlations (Fig 4)

There were few significant correlations between stream or-der and the RHS variables in individual catchments (Fig 4)The only significant correlation after p-value correction iswith 90th percentile HMS in the Wensum which shows astrong negative relationship

4 Discussion

41 A new approach to utilizing network topology incatchment-level analysis

Network density metrics represent an alternative approach toaccount for network topology in catchment-level studies op-timizing the width dimension of the network (or the num-

ber of links in the network) as opposed to the commonplacestream-order metric which only reflects the longitudinal po-sition of links in a network (Fig 1) This study demonstratesthat two topology metrics can be calculated simply from aDTM with GIS and using a broad-scale analysis of river at-tributes can be used to investigate the functional processeswithin catchments

While the two network density metrics have similar forms(ie forms are consistently unimodal or multi-modal) thespatial configuration of the distance and elevation intervalsused in the calculation of network density varies and mayimpact the effectiveness of each topological metric Distancenetwork density separates the catchment into intervals basedon distance which spread upstream from the outlet (Fig 2a)reflecting natural sub-basins within the fractal structure of thecatchment (Lashermes and Foufoula-Georgiou 2007) Thisdiffers from elevation network density which separates thecatchment into intervals based on elevation which radiate outfrom the main channel of the network (Fig 2b) The config-uration means that distance intervals contain streams that arein closer proximity to one another rather than the more distalconfiguration created by the elevation intervals suggesting adegree of spatial dependency in river functioning This hasbeen highlighted in previous studies where spatial networkstructure has a stronger influence on some in-channel pro-cesses than predictor variables such as elevation (Steel et al

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2314 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

2016) However elevation intervals contain RHS sites thatalthough they may be distal may have similar properties aselevation has been strongly related to RHS variables includ-ing flow type substrate etc in a number of studies (Jeffers1998a Naura et al 2016 Vaughan et al 2013)

42 Impacts of network topology on rivercharacteristics

River characteristics are assessed using the RHS dataset Theobservations made by the RHS dataset (Table 1) cannot of-fer the level of detail regarding geomorphological processthat river classifications which consider multiple scales canoffer (eg Brierley and Fryirs 2000 Gurnell et al 2016)While process-based classifications are preferable broad-scale monitoring datasets such as the River Habitat Surveymay still be useful when combined with map-derived datato explore controls on river characteristics (Harvey et al2008 Naura et al 2016 Vaughan et al 2013) Howeverthere are biases in RHS data collection an inherent limitationwhen using existing datasets (Vaughan and Ormerod 2010)specifically the standardized survey length of 500 m reachthat will capture differing amounts of natural variability de-pending on the size of the river While this must be notedthere are few significant correlations between river charac-teristics identified with stream order (Fig 4) which suggeststhat channel size is not influencing the RHS variables to agreat degree in these catchments

In this study it is anticipated that sites in high networkdensity areas will have higher levels of habitat diversity asindicated by previous studies of confluences and networks(Benda et al 2004a Best 1985 Rice 2017) in turn in-creasing mean sediment size and flow-type speed comparedto sites in low-density areas The results of the correlationsbetween distance network density and river characteristicswhen all catchments are combined support this hypothesiswith greater HQA flow-type speed and coarser sedimentsizes observed in areas with high distance network density(Fig 4)

For individual catchments elevation network density in-duces a stronger positive HQA response across all catch-ments than distance network density (Fig 4) This supportsthe evidence that individual confluences (Rice et al 2006)and high densities of confluences increase physical hetero-geneity within the river network (Benda et al 2004b Rice2017) However flow-type speed and sediment size responddifferently to network density in individual catchments

Slower flow types are observed in high network densityareas of the Avon and Tamar whereas faster flow types areobserved in high-elevation network density areas of the Wen-sum Individual confluences are shown to create numeroushigh- and low-speed flow environments (Best 1987) thatmay be observed in surface water topography (Biron et al2002) It was expected that the introduction of the additionalflow types by a high density of confluences in relatively low-

energy rivers would increase mean reach flow-type speedhowever the correlation analysis (Fig 4) suggests that insome catchments mean flow-type speed is reduced

Sediment size response also shows variation betweencatchments Sediment size is coarser in network-dense ar-eas of the Avon and Eden as expected but is finer in boththe Tamar and Wensum (Fig 4) The evidence from high-energy rivers shows that sediment becomes coarser down-stream and finer upstream of certain confluences (Benda etal 2004a Rice 1998) and in this case high numbers of con-fluences were expected to increase mean sediment size of thereach However the impact on sediment size is dependent onthe sediment calibre of the incoming tributary being higherthan the main channel with enough energy to transport thecoarse sediment to the confluence for numerous tributariesin an area The Tamar and Wensum have different rangesof sediment sizes with the Tamar having coarser sedimentthan the Wensum on average (Fig 3) This implies that trib-utaries in the Tamar may be energy limited not transportingcoarse sediments to confluences and the Wensum may suf-fer from inputs of fine sediments from the high percentageof arable land that concentrates in network-dense areas Thishas before been observed in a low-energy modified catch-ment where anthropogenic modifications in tributaries re-duced coarse sediment and flow capacity causing either lim-ited confluence impact or localized sediment fining (Singer2008)

Others have related the capacity of confluences to alterreach features to the morphometry of catchments with largerand more circular catchments containing a higher percentageof confluences that have a significant impact (Rice 2017)Based on this theory the Eden and Tamar are likely to havethe greatest impact as they are the most circular and steepestof the four catchments (although the Tamar is the smallestby area) Yet there is no clear pattern indicating that thesecatchments respond differently than the Avon and Wensum(Fig 4) with catchments responding differently to differ-ent variables This perhaps suggests that rather than networkdensity having a directional impact on factors such as flow-type speed and sediment size it has an impact on overall het-erogeneity at the reach level as suggested by previous stud-ies and that specific directional change occurs at the sub-reach level

An increase in channel modification is also hypothesizeddue to the increase in flood peak downstream from conflu-ences (Depettris et al 2000) and the scour and erosion as-sociated with confluence junctions (Best 1986) potentiallyincreasing the need for bed and bank protection The corre-lations between distance network density and HMS when allcatchments are combined undermine this hypothesis show-ing less channel modification where distance network den-sity is higher (Fig 4) There were few significant correlationswith individual catchments but HMS in the Eden was consis-tently observed to be higher in network-dense areas (Fig 4)This may be due to the Edenrsquos high elevation and steep to-

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2315

pography inducing a higher energy environment where scourand erosion processes in areas with high numbers of conflu-ences would be more likely to be present

Differences in RHS variable responses also differed be-tween the descriptive statistics considered Often the ex-tremes 90th and 10th percentiles showed more significantand stronger correlations than the mean or median (Fig 4)This may reflect findings from previous studies which sug-gest that not all confluences cause reach-scale changes (Rice1998) that perhaps the changes to river character induced bycertain confluences only influence certain reaches whereasothers are left unaffected creating less pronounced responsesin the mean and median of variables External factors mayalso influence this trend for example 10th percentile HMSvisually responds dramatically to network density metrics(Fig 3a and b) compared to the other descriptive statisticsThis suggests that the most natural sites (ie with the low-est HMS score) are responding differently to network den-sity with the most natural sites having less modificationin network-dense areas whereas less natural sites becomemore modified in network-dense areas This reflects the HQAscore results which visually (Fig 3a and b) and statistically(Fig 4) vary with distance network density except for the10th percentile These sites with the lowest habitat qualityand naturalness are likely influenced by anthropogenic fac-tors that are independent of network density reducing habitatquality scores at impacted sites

The two network density metrics have differing impacts onriver characteristics While distance network density showsconsistently significant correlations when all catchments arecombined individual catchments respond more frequentlyand more strongly to elevation network density (Fig 4)This may be because there is a dramatic split in distancenetwork density values between the more upland drainage-dense catchments Eden and Tamar than the lowland chalklow-drainage catchments Avon and Wensum The combinedcorrelation will therefore in part reflect the difference be-tween the catchments which have different ranges of rivercharacteristics (Fig 3) This is not the case for elevation net-work density which has higher density values in the Tamarand Wensum than the Avon and Eden so therefore the char-acteristics of the catchments will have less bearing on thecombined correlation However there are patterns identifiedwith distance network density in individual catchments thatare not present with elevation network density increasedflow-type speed and sediment size in the Eden and reducedflow-type speed in the Avon with network density (Fig 4)which show its usefulness

The results suggest that the distance network density andelevation network density metrics quantify different dimen-sions of network topology which are shown to exhibit func-tionally meaningful patterns for river reach characteristicsbased on the correlation results Perhaps within catchmentselevation network density provides the more powerful met-ric for individual catchments but distance network density

better accounts for the drainage density of the catchmentsallowing it to be applied across multiple catchments

43 Comparison of stream order to network densitymetrics

The classic method of accounting for network topologystream order is critiqued for failing to represent disconti-nuities in the network and simplifying the network into agradient The number of links in different stream orders isconsistent across all catchments not reflecting the internalstructure of the network or the variety between catchmentsthat is achieved by the distance network density and eleva-tion network density metrics The analysis of the two net-work density metrics presented in this paper shows that dis-tances from source and elevation are not mutually exclusive(Fig 2a and b) contrary to the stream-order metric whichrepresents streams as upstream to downstream or upland tolowland

Stream order has few significant correlations with manyof the river characteristics considered in this study Negativecorrelations with HMS in the Wensum were statistically sig-nificant post p-value correction likely driving the significantrelationship with all catchments combined for this variable(Fig 4) This suggests that modification is greater upstreamin the Wensum contrary to ideas that downstream reachesmay show greater anthropogenic modification Intense agri-cultural land use in the upper reaches of the Wensum islikely to be the cause of the high HMS scores upstreamThe lack of significant correlations suggests that stream or-der and therefore an upstream-to-downstream gradient arenot the predominant pattern in river characteristics despitethe description of such a gradient by geomorphic (Schumm1977) and ecological frameworks (Vannote et al 1980)This is surprising as distance and elevation which both re-flect the upstream-to-downstream gradient have proven tobe important factors in previous studies explaining patternsof RHS features at a national level (Jeffers 1998a Vaughanet al 2013) This implies that upstream-to-downstream gra-dient may not sufficiently reflect patterns of river charac-teristics through the river network within individual catch-ments Others have also found that stream order has weakand inconsistent relationships with biodiversity patterns inriver systems arguing that the topological measure has no di-rect mechanistic control on biodiversity (Vander Vorste et al2017) Instead this study finds that the network density met-rics are a more powerful metric which conceptually providean improved method of accounting for the impacts of net-work topology on the fluvial system exhibiting relationshipswith river characteristics particularly habitat quality score(Fig 4)

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2316 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

44 Applicability of network topology metrics todifferent environments

Much of the seminal work on network and confluence im-pacts (eg the Network Dynamics Hypothesis Benda et al2004b and Link Discontinuity Concept Rice et al 2001) isconducted in natural highly erosive catchments with first-hand empirical measurements However in an age whenrivers and their catchments are increasingly altered by an-thropogenic modification (Meybeck 2003) contemporarystudies must not only aim to expand knowledge but also findmethods of transferring knowledge to many increasingly al-tered catchments (Clifford 2002)

The catchments selected by this study are more greatlymodified and although they reflect a range of fluvial envi-ronments in England are lower-energy catchments than theseminal studies Benda et al (2004a) suggest that confluenceeffects in less active landscapes would be subdued comparedto highly erosive landscapes but the evidence presented heredemonstrates the utility of evaluating network topologicalstructure in studies on catchment-level effects in any typeof fluvial environment including those with low-energy andwidespread anthropogenic changes The response of someriver characteristics varied between catchments observationsof flow-type speed sediment size and modifications showeddifferent responses to network density in different catch-ments This suggests that the functional effect of these topo-logical metrics is catchment dependent and likely is influ-enced by external catchment characteristics such as land usenot considered in this study although impact did not appearto vary with catchment topography or circularity as has beenshown in prior studies (Benda et al 2004a Rice 2017) Thisshould be explored further in future research to enable rec-ommendations to be made regarding where and how reachesmay respond to network density The response of habitatquality score was however consistent across catchments andbetween metrics showing that habitat quality is greater inareas with high network density (Fig 4) as hypothesizedby the Network Dynamics Hypothesis (Benda et al 2004b)and demonstrated by studies on individual catchments (Rice2017 Rice et al 2006)

The methods presented in this paper are designed to be im-plemented in any catchment with a dendritic network struc-ture The topology metrics can easily be calculated from anydendritic network with DTM data using GIS and comparedto any site-scale data This study uses regulatory monitoringdatasets so that analysis is targeted to assessment scores andphysical features of interest to river managers Also the highvolume and wide spatial extent of data available from regu-latory sources allow for between-catchment comparisons

5 Conclusions

Although appreciation of catchment-level effects is consid-ered the epitome of understanding river functioning a keycomponent of the catchment ndash the river network ndash is over-looked and oversimplified by catchment-level studies Thisstudy finds that river network density plays a role in struc-turing the distribution river characteristics throughout thecatchment offering more detailed explanation than the clas-sic stream-order metric Network-dense areas are generallyfound to have higher habitat quality and diversity but modi-fication flow-type speed and sediment size show different re-sponses in different catchments This study suggests two po-tential reasons for this (1) there is evidence that confluencesin the river network increase diversity as is observed in thisstudy so the direction of mean river characteristic responsemay not be consistent and (2) there may be external factorssuch as sediment availability land cover and anthropogenicmodification that alter the direction of mean river characteris-tic response This paper demonstrates the functional responseof river characteristics to network topology and suggests thatthe inclusion of network topology in catchment-level studieswould add a layer of function-based understanding to suchstudies linking reaches to their catchments

The broad-scale methodology adopted by this study allowsthe network density metrics which are easily extracted fromopen-source data using GIS software to be compared to anyregulatory dataset The use of regulatory datasets allows notonly for analysis over a wider spatial extent but also for moreapplicable results for regulatory bodies Therefore the inter-disciplinary approach to characterizing network topology canbe applied efficiently and effectively to capture catchment-level impacts on reach-level functioning in any catchmentacross the globe

Data availability River Habitat Survey data are freely availablefrom the Environment Agency England River network data andelevation data are available from the Centre for Ecology and Hy-drology UK

Author contributions EH and NC conceptualised and designed thestudy EH performed the data analysis with the support of JMEH wrote the paper with comments provided by NC and JM

Competing interests The authors declare that they have no conflictof interest

Acknowledgements The authors would like to thank the Environ-ment Agency and the Centre for Ecology and Hydrology for accessto the data used in this paper This research is funded by the NaturalEnvironmental Research Council (NERC)

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2317

Review statement This paper was edited by Nandita Basu and re-viewed by Simone Bizzi and one anonymous referee

References

Belletti B Rinaldi M Buijse A D Gurnell A Mand Mosselman E A review of assessment methods forriver hydromorphology Environ Earth Sci 73 2079ndash2100httpsdoiorg101007s12665-014-3558-1 2015

Benda L Andras K Miller D and Bigelow P Confluenceeffects in rivers Interactions of basin scale network geom-etry and disturbance regimes Water Resour Res 40 1ndash15httpsdoiorg1010292003WR002583 2004a

Benda L Poff N L Miller D Dunne T Reeves GPess G and Pollock M The network dynamics hy-pothesis how channel networks structure riverine habi-tats Bioscience 54 413ndash427 httpsdoiorg1016410006-3568(2004)054[0413TNDHHC]20CO2 2004b

Benjamini Y and Hochberg Y Controlling the False DiscoveryRate A Practical and Powerful Approach to Multiple TestingJ Roy Stat Soc B 57 289ndash300 1995

Best J L Flow dynamics and sediment transport at river channelconfluences Birbeck University of London London 1985

Best J L The morphology of river channel con-fluences Prog Phys Geogr 10 157ndash174httpsdoiorg101177030913338601000201 1986

Best J L Flow dynamics at river channel confluences implica-tions for sediment transport and bed morphology in Recent de-velopments in fluvial sedimentology edited by Ethridge F GFlores R M and Harvey M D Spec Publ Soc Econ Pale-ontol Miner Tulsa Okla 27ndash35 1987

Biron P M Richer A Kirkbride A D Roy A G andHan S Spatial patterns of water surface topography ata river confluence Earth Surf Proc Land 27 913ndash928httpsdoiorg101002esp359 2002

Bravard J P and Gilvear D J Hydrological and geomorpho-logical structure of hydrosystems in The Fluvial Hydrosystemedited by Petts G E and Amoros C Springer Netherlands98ndash116 1996

Brierley G and Fryirs K River Styles a Geomorphic Approachto Catchment Characterization Implications for River Rehabili-tation in Bega Catchment New South Wales Australia EnvironManage 25 661ndash679 2000

Bunn S E and Arthington A H Basic principles and ecologi-cal consequences of altered flow regimes for aquatic biodiversityEnviron Manage 30 492ndash507 httpsdoiorg101007s00267-002-2737-0 2002

Church M and Kellerhals R On the statistics of grain size varia-tion along a gravel river Can J Earth Sci 15 1151ndash1160 1978

Clifford N J Hydrology the changing paradigm Prog Phys Ge-ogr 26 290ndash301 2002

Cohen P Andriamahefa H and Wasson J-G Towardsa regionalizaton of aquatic habitat distribution of meso-habitats at the scale of a large basin Regul RiversRes Manage 14 391ndash404 httpsdoiorg101002(SICI)1099-1646(19980910)145lt391AID-RRR513gt30CO2-W 1998

Davenport A J Gurnell A M and Armitage P D Habitat sur-vey and classification of urban rivers River Res Appl 20 687ndash704 httpsdoiorg101002rra785 2004

Davies N M Norris R H and Thoms M C Prediction and as-sessment of local stream habitat features using large-scale catch-ment characteristics Freshwater Biol 45 343ndash369 2000

Depettris C Mendiondo E M Neiff J and Rohrmann H Flooddefence strategy at the confluence of the Paranaacute-Paraguay riversProc Int Symp Flood Def Kassel He C31 C40 2000

Dollar E James C Rogers K and Thoms M A framework forinterdisciplinary understanding of rivers as ecosystems Geomor-phology 89 147ndash162 2007

Dovers S R and Day D G Australian rivers and statute law En-viron Plan Law J 5 90ndash108 1988

Downs P W Gregory K J and Brookes A How integratedis river basin management Environ Manage 15 299ndash309httpsdoiorg101007BF02393876 1991

Emery J C Gurnell A M Clifford N J and Petts G E Char-acteristics and controls of gravel-bed riffles An analysis of datafrom the river-habitat survey Water Environ J 18 210ndash216httpsdoiorg101111j1747-65932004tb00535x 2004

Evans I S and Minaacuter J A classification of geomorphometric vari-ables in International Geom-orphometry 2011 Geomoprhome-tryorg Redlabds CA 105ndash108 2011

Fausch K D Torgersen C E Baxter C V Li H WView C The O F Is R To N How U Among ISet S For C and Fishes S Landscapes to riverscapesBridging the gap between research and conservation of streamfishes Bioscience 52 483ndash498 httpsdoiorg1016410006-3568(2002)052[0483LTRBTG]20CO2 2002

Glickman M E Rao S R and Schultz M R False discov-ery rate control is a recommended alternative to Bonferroni-typeadjustments in health studies J Clin Epidemiol 67 850ndash857httpsdoiorg101016jjclinepi201403012 2014

Gupta V K and Mesa O J Runoff generation and hydro-logic response via channel network geomorphology - Re-cent progress and open problems J Hydrol 102 3ndash28httpsdoiorg1010160022-1694(88)90089-3 1988

Gupta V K and Waymire E D On the formulation of an analyt-ical approach to hydrologic response and similarity at the basinscale J Hydrol 65 95ndash123 1983

Gupta V K Waymire E D and Rodriguez-Iturbe I On scalesgravity and network structure in basin runoff in Scale problemsin hydrology Springer Netherlands 159ndash184 1986

Gurnell A M Rinaldi M Belletti B Bizzi S Blamauer BBraca G Buijse A D Bussettini M Camenen B ComitiF Demarchi L Garciacutea de Jaloacuten D Gonzaacutelez del TaacutenagoM Grabowski R C Gunn I D M Habersack H Hen-driks D Henshaw A J Kloumlsch M Lastoria B LatapieA Marcinkowski P Martiacutenez-Fernaacutendez V Mosselman EMountford J O Nardi L Okruszko T OrsquoHare M T PalmaM Percopo C Surian N van de Bund W Weissteiner Cand Ziliani L A multi-scale hierarchical framework for devel-oping understanding of river behaviour to support river manage-ment Aquat Sci 78 1ndash16 httpsdoiorg101007s00027-015-0424-5 2016

Harvey G L Gurnell A M and Clifford N J Characterisationof river reaches The influence of rock type Catena 76 78ndash88httpsdoiorg101016jcatena200809010 2008

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2312 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

Figure 3 Network topology metrics (a) distance network density and (b) elevation network density Descriptive statistics of each RHSvariable over (a) distance and (b) elevation for each catchment with smooth loess lines to indicate trend Network topology metrics arenormalized between 0 and 1 and HMS score is logarithmically transformed for display purposes

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2313

Figure 4 Summary of correlations between distance network density elevation network density stream order and RHS variables for allcatchments combined and each individual catchment Significance of correlation is indicated by point size with the largest points significantpost p-value correction Effect size (Kendallrsquos τ ) is indicated by colour No correlation was possible between stream order and 10th percentilesediment size in the Avon due to no variation in the RHS variable

flow-type speed associated with peaks in network density(Fig 3a) The Eden and Tamar however have positive cor-relations with mean and 90th percentile flow-type speed fordistance network density but negative correlations with me-dian and 10th percentile elevation network density The Wen-sum shows positive correlations between flow-type speed andelevation network density Sediment size has a consistent re-sponse to distance network density with the Eden and Wen-sum showing negative correlations with the sediment size(Fig 4) For elevation network density the Avon shows neg-ative correlations with sediment size whereas the Tamar andWensum show positive correlations (Fig 4)

There were few significant correlations between stream or-der and the RHS variables in individual catchments (Fig 4)The only significant correlation after p-value correction iswith 90th percentile HMS in the Wensum which shows astrong negative relationship

4 Discussion

41 A new approach to utilizing network topology incatchment-level analysis

Network density metrics represent an alternative approach toaccount for network topology in catchment-level studies op-timizing the width dimension of the network (or the num-

ber of links in the network) as opposed to the commonplacestream-order metric which only reflects the longitudinal po-sition of links in a network (Fig 1) This study demonstratesthat two topology metrics can be calculated simply from aDTM with GIS and using a broad-scale analysis of river at-tributes can be used to investigate the functional processeswithin catchments

While the two network density metrics have similar forms(ie forms are consistently unimodal or multi-modal) thespatial configuration of the distance and elevation intervalsused in the calculation of network density varies and mayimpact the effectiveness of each topological metric Distancenetwork density separates the catchment into intervals basedon distance which spread upstream from the outlet (Fig 2a)reflecting natural sub-basins within the fractal structure of thecatchment (Lashermes and Foufoula-Georgiou 2007) Thisdiffers from elevation network density which separates thecatchment into intervals based on elevation which radiate outfrom the main channel of the network (Fig 2b) The config-uration means that distance intervals contain streams that arein closer proximity to one another rather than the more distalconfiguration created by the elevation intervals suggesting adegree of spatial dependency in river functioning This hasbeen highlighted in previous studies where spatial networkstructure has a stronger influence on some in-channel pro-cesses than predictor variables such as elevation (Steel et al

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2314 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

2016) However elevation intervals contain RHS sites thatalthough they may be distal may have similar properties aselevation has been strongly related to RHS variables includ-ing flow type substrate etc in a number of studies (Jeffers1998a Naura et al 2016 Vaughan et al 2013)

42 Impacts of network topology on rivercharacteristics

River characteristics are assessed using the RHS dataset Theobservations made by the RHS dataset (Table 1) cannot of-fer the level of detail regarding geomorphological processthat river classifications which consider multiple scales canoffer (eg Brierley and Fryirs 2000 Gurnell et al 2016)While process-based classifications are preferable broad-scale monitoring datasets such as the River Habitat Surveymay still be useful when combined with map-derived datato explore controls on river characteristics (Harvey et al2008 Naura et al 2016 Vaughan et al 2013) Howeverthere are biases in RHS data collection an inherent limitationwhen using existing datasets (Vaughan and Ormerod 2010)specifically the standardized survey length of 500 m reachthat will capture differing amounts of natural variability de-pending on the size of the river While this must be notedthere are few significant correlations between river charac-teristics identified with stream order (Fig 4) which suggeststhat channel size is not influencing the RHS variables to agreat degree in these catchments

In this study it is anticipated that sites in high networkdensity areas will have higher levels of habitat diversity asindicated by previous studies of confluences and networks(Benda et al 2004a Best 1985 Rice 2017) in turn in-creasing mean sediment size and flow-type speed comparedto sites in low-density areas The results of the correlationsbetween distance network density and river characteristicswhen all catchments are combined support this hypothesiswith greater HQA flow-type speed and coarser sedimentsizes observed in areas with high distance network density(Fig 4)

For individual catchments elevation network density in-duces a stronger positive HQA response across all catch-ments than distance network density (Fig 4) This supportsthe evidence that individual confluences (Rice et al 2006)and high densities of confluences increase physical hetero-geneity within the river network (Benda et al 2004b Rice2017) However flow-type speed and sediment size responddifferently to network density in individual catchments

Slower flow types are observed in high network densityareas of the Avon and Tamar whereas faster flow types areobserved in high-elevation network density areas of the Wen-sum Individual confluences are shown to create numeroushigh- and low-speed flow environments (Best 1987) thatmay be observed in surface water topography (Biron et al2002) It was expected that the introduction of the additionalflow types by a high density of confluences in relatively low-

energy rivers would increase mean reach flow-type speedhowever the correlation analysis (Fig 4) suggests that insome catchments mean flow-type speed is reduced

Sediment size response also shows variation betweencatchments Sediment size is coarser in network-dense ar-eas of the Avon and Eden as expected but is finer in boththe Tamar and Wensum (Fig 4) The evidence from high-energy rivers shows that sediment becomes coarser down-stream and finer upstream of certain confluences (Benda etal 2004a Rice 1998) and in this case high numbers of con-fluences were expected to increase mean sediment size of thereach However the impact on sediment size is dependent onthe sediment calibre of the incoming tributary being higherthan the main channel with enough energy to transport thecoarse sediment to the confluence for numerous tributariesin an area The Tamar and Wensum have different rangesof sediment sizes with the Tamar having coarser sedimentthan the Wensum on average (Fig 3) This implies that trib-utaries in the Tamar may be energy limited not transportingcoarse sediments to confluences and the Wensum may suf-fer from inputs of fine sediments from the high percentageof arable land that concentrates in network-dense areas Thishas before been observed in a low-energy modified catch-ment where anthropogenic modifications in tributaries re-duced coarse sediment and flow capacity causing either lim-ited confluence impact or localized sediment fining (Singer2008)

Others have related the capacity of confluences to alterreach features to the morphometry of catchments with largerand more circular catchments containing a higher percentageof confluences that have a significant impact (Rice 2017)Based on this theory the Eden and Tamar are likely to havethe greatest impact as they are the most circular and steepestof the four catchments (although the Tamar is the smallestby area) Yet there is no clear pattern indicating that thesecatchments respond differently than the Avon and Wensum(Fig 4) with catchments responding differently to differ-ent variables This perhaps suggests that rather than networkdensity having a directional impact on factors such as flow-type speed and sediment size it has an impact on overall het-erogeneity at the reach level as suggested by previous stud-ies and that specific directional change occurs at the sub-reach level

An increase in channel modification is also hypothesizeddue to the increase in flood peak downstream from conflu-ences (Depettris et al 2000) and the scour and erosion as-sociated with confluence junctions (Best 1986) potentiallyincreasing the need for bed and bank protection The corre-lations between distance network density and HMS when allcatchments are combined undermine this hypothesis show-ing less channel modification where distance network den-sity is higher (Fig 4) There were few significant correlationswith individual catchments but HMS in the Eden was consis-tently observed to be higher in network-dense areas (Fig 4)This may be due to the Edenrsquos high elevation and steep to-

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2315

pography inducing a higher energy environment where scourand erosion processes in areas with high numbers of conflu-ences would be more likely to be present

Differences in RHS variable responses also differed be-tween the descriptive statistics considered Often the ex-tremes 90th and 10th percentiles showed more significantand stronger correlations than the mean or median (Fig 4)This may reflect findings from previous studies which sug-gest that not all confluences cause reach-scale changes (Rice1998) that perhaps the changes to river character induced bycertain confluences only influence certain reaches whereasothers are left unaffected creating less pronounced responsesin the mean and median of variables External factors mayalso influence this trend for example 10th percentile HMSvisually responds dramatically to network density metrics(Fig 3a and b) compared to the other descriptive statisticsThis suggests that the most natural sites (ie with the low-est HMS score) are responding differently to network den-sity with the most natural sites having less modificationin network-dense areas whereas less natural sites becomemore modified in network-dense areas This reflects the HQAscore results which visually (Fig 3a and b) and statistically(Fig 4) vary with distance network density except for the10th percentile These sites with the lowest habitat qualityand naturalness are likely influenced by anthropogenic fac-tors that are independent of network density reducing habitatquality scores at impacted sites

The two network density metrics have differing impacts onriver characteristics While distance network density showsconsistently significant correlations when all catchments arecombined individual catchments respond more frequentlyand more strongly to elevation network density (Fig 4)This may be because there is a dramatic split in distancenetwork density values between the more upland drainage-dense catchments Eden and Tamar than the lowland chalklow-drainage catchments Avon and Wensum The combinedcorrelation will therefore in part reflect the difference be-tween the catchments which have different ranges of rivercharacteristics (Fig 3) This is not the case for elevation net-work density which has higher density values in the Tamarand Wensum than the Avon and Eden so therefore the char-acteristics of the catchments will have less bearing on thecombined correlation However there are patterns identifiedwith distance network density in individual catchments thatare not present with elevation network density increasedflow-type speed and sediment size in the Eden and reducedflow-type speed in the Avon with network density (Fig 4)which show its usefulness

The results suggest that the distance network density andelevation network density metrics quantify different dimen-sions of network topology which are shown to exhibit func-tionally meaningful patterns for river reach characteristicsbased on the correlation results Perhaps within catchmentselevation network density provides the more powerful met-ric for individual catchments but distance network density

better accounts for the drainage density of the catchmentsallowing it to be applied across multiple catchments

43 Comparison of stream order to network densitymetrics

The classic method of accounting for network topologystream order is critiqued for failing to represent disconti-nuities in the network and simplifying the network into agradient The number of links in different stream orders isconsistent across all catchments not reflecting the internalstructure of the network or the variety between catchmentsthat is achieved by the distance network density and eleva-tion network density metrics The analysis of the two net-work density metrics presented in this paper shows that dis-tances from source and elevation are not mutually exclusive(Fig 2a and b) contrary to the stream-order metric whichrepresents streams as upstream to downstream or upland tolowland

Stream order has few significant correlations with manyof the river characteristics considered in this study Negativecorrelations with HMS in the Wensum were statistically sig-nificant post p-value correction likely driving the significantrelationship with all catchments combined for this variable(Fig 4) This suggests that modification is greater upstreamin the Wensum contrary to ideas that downstream reachesmay show greater anthropogenic modification Intense agri-cultural land use in the upper reaches of the Wensum islikely to be the cause of the high HMS scores upstreamThe lack of significant correlations suggests that stream or-der and therefore an upstream-to-downstream gradient arenot the predominant pattern in river characteristics despitethe description of such a gradient by geomorphic (Schumm1977) and ecological frameworks (Vannote et al 1980)This is surprising as distance and elevation which both re-flect the upstream-to-downstream gradient have proven tobe important factors in previous studies explaining patternsof RHS features at a national level (Jeffers 1998a Vaughanet al 2013) This implies that upstream-to-downstream gra-dient may not sufficiently reflect patterns of river charac-teristics through the river network within individual catch-ments Others have also found that stream order has weakand inconsistent relationships with biodiversity patterns inriver systems arguing that the topological measure has no di-rect mechanistic control on biodiversity (Vander Vorste et al2017) Instead this study finds that the network density met-rics are a more powerful metric which conceptually providean improved method of accounting for the impacts of net-work topology on the fluvial system exhibiting relationshipswith river characteristics particularly habitat quality score(Fig 4)

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2316 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

44 Applicability of network topology metrics todifferent environments

Much of the seminal work on network and confluence im-pacts (eg the Network Dynamics Hypothesis Benda et al2004b and Link Discontinuity Concept Rice et al 2001) isconducted in natural highly erosive catchments with first-hand empirical measurements However in an age whenrivers and their catchments are increasingly altered by an-thropogenic modification (Meybeck 2003) contemporarystudies must not only aim to expand knowledge but also findmethods of transferring knowledge to many increasingly al-tered catchments (Clifford 2002)

The catchments selected by this study are more greatlymodified and although they reflect a range of fluvial envi-ronments in England are lower-energy catchments than theseminal studies Benda et al (2004a) suggest that confluenceeffects in less active landscapes would be subdued comparedto highly erosive landscapes but the evidence presented heredemonstrates the utility of evaluating network topologicalstructure in studies on catchment-level effects in any typeof fluvial environment including those with low-energy andwidespread anthropogenic changes The response of someriver characteristics varied between catchments observationsof flow-type speed sediment size and modifications showeddifferent responses to network density in different catch-ments This suggests that the functional effect of these topo-logical metrics is catchment dependent and likely is influ-enced by external catchment characteristics such as land usenot considered in this study although impact did not appearto vary with catchment topography or circularity as has beenshown in prior studies (Benda et al 2004a Rice 2017) Thisshould be explored further in future research to enable rec-ommendations to be made regarding where and how reachesmay respond to network density The response of habitatquality score was however consistent across catchments andbetween metrics showing that habitat quality is greater inareas with high network density (Fig 4) as hypothesizedby the Network Dynamics Hypothesis (Benda et al 2004b)and demonstrated by studies on individual catchments (Rice2017 Rice et al 2006)

The methods presented in this paper are designed to be im-plemented in any catchment with a dendritic network struc-ture The topology metrics can easily be calculated from anydendritic network with DTM data using GIS and comparedto any site-scale data This study uses regulatory monitoringdatasets so that analysis is targeted to assessment scores andphysical features of interest to river managers Also the highvolume and wide spatial extent of data available from regu-latory sources allow for between-catchment comparisons

5 Conclusions

Although appreciation of catchment-level effects is consid-ered the epitome of understanding river functioning a keycomponent of the catchment ndash the river network ndash is over-looked and oversimplified by catchment-level studies Thisstudy finds that river network density plays a role in struc-turing the distribution river characteristics throughout thecatchment offering more detailed explanation than the clas-sic stream-order metric Network-dense areas are generallyfound to have higher habitat quality and diversity but modi-fication flow-type speed and sediment size show different re-sponses in different catchments This study suggests two po-tential reasons for this (1) there is evidence that confluencesin the river network increase diversity as is observed in thisstudy so the direction of mean river characteristic responsemay not be consistent and (2) there may be external factorssuch as sediment availability land cover and anthropogenicmodification that alter the direction of mean river characteris-tic response This paper demonstrates the functional responseof river characteristics to network topology and suggests thatthe inclusion of network topology in catchment-level studieswould add a layer of function-based understanding to suchstudies linking reaches to their catchments

The broad-scale methodology adopted by this study allowsthe network density metrics which are easily extracted fromopen-source data using GIS software to be compared to anyregulatory dataset The use of regulatory datasets allows notonly for analysis over a wider spatial extent but also for moreapplicable results for regulatory bodies Therefore the inter-disciplinary approach to characterizing network topology canbe applied efficiently and effectively to capture catchment-level impacts on reach-level functioning in any catchmentacross the globe

Data availability River Habitat Survey data are freely availablefrom the Environment Agency England River network data andelevation data are available from the Centre for Ecology and Hy-drology UK

Author contributions EH and NC conceptualised and designed thestudy EH performed the data analysis with the support of JMEH wrote the paper with comments provided by NC and JM

Competing interests The authors declare that they have no conflictof interest

Acknowledgements The authors would like to thank the Environ-ment Agency and the Centre for Ecology and Hydrology for accessto the data used in this paper This research is funded by the NaturalEnvironmental Research Council (NERC)

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2317

Review statement This paper was edited by Nandita Basu and re-viewed by Simone Bizzi and one anonymous referee

References

Belletti B Rinaldi M Buijse A D Gurnell A Mand Mosselman E A review of assessment methods forriver hydromorphology Environ Earth Sci 73 2079ndash2100httpsdoiorg101007s12665-014-3558-1 2015

Benda L Andras K Miller D and Bigelow P Confluenceeffects in rivers Interactions of basin scale network geom-etry and disturbance regimes Water Resour Res 40 1ndash15httpsdoiorg1010292003WR002583 2004a

Benda L Poff N L Miller D Dunne T Reeves GPess G and Pollock M The network dynamics hy-pothesis how channel networks structure riverine habi-tats Bioscience 54 413ndash427 httpsdoiorg1016410006-3568(2004)054[0413TNDHHC]20CO2 2004b

Benjamini Y and Hochberg Y Controlling the False DiscoveryRate A Practical and Powerful Approach to Multiple TestingJ Roy Stat Soc B 57 289ndash300 1995

Best J L Flow dynamics and sediment transport at river channelconfluences Birbeck University of London London 1985

Best J L The morphology of river channel con-fluences Prog Phys Geogr 10 157ndash174httpsdoiorg101177030913338601000201 1986

Best J L Flow dynamics at river channel confluences implica-tions for sediment transport and bed morphology in Recent de-velopments in fluvial sedimentology edited by Ethridge F GFlores R M and Harvey M D Spec Publ Soc Econ Pale-ontol Miner Tulsa Okla 27ndash35 1987

Biron P M Richer A Kirkbride A D Roy A G andHan S Spatial patterns of water surface topography ata river confluence Earth Surf Proc Land 27 913ndash928httpsdoiorg101002esp359 2002

Bravard J P and Gilvear D J Hydrological and geomorpho-logical structure of hydrosystems in The Fluvial Hydrosystemedited by Petts G E and Amoros C Springer Netherlands98ndash116 1996

Brierley G and Fryirs K River Styles a Geomorphic Approachto Catchment Characterization Implications for River Rehabili-tation in Bega Catchment New South Wales Australia EnvironManage 25 661ndash679 2000

Bunn S E and Arthington A H Basic principles and ecologi-cal consequences of altered flow regimes for aquatic biodiversityEnviron Manage 30 492ndash507 httpsdoiorg101007s00267-002-2737-0 2002

Church M and Kellerhals R On the statistics of grain size varia-tion along a gravel river Can J Earth Sci 15 1151ndash1160 1978

Clifford N J Hydrology the changing paradigm Prog Phys Ge-ogr 26 290ndash301 2002

Cohen P Andriamahefa H and Wasson J-G Towardsa regionalizaton of aquatic habitat distribution of meso-habitats at the scale of a large basin Regul RiversRes Manage 14 391ndash404 httpsdoiorg101002(SICI)1099-1646(19980910)145lt391AID-RRR513gt30CO2-W 1998

Davenport A J Gurnell A M and Armitage P D Habitat sur-vey and classification of urban rivers River Res Appl 20 687ndash704 httpsdoiorg101002rra785 2004

Davies N M Norris R H and Thoms M C Prediction and as-sessment of local stream habitat features using large-scale catch-ment characteristics Freshwater Biol 45 343ndash369 2000

Depettris C Mendiondo E M Neiff J and Rohrmann H Flooddefence strategy at the confluence of the Paranaacute-Paraguay riversProc Int Symp Flood Def Kassel He C31 C40 2000

Dollar E James C Rogers K and Thoms M A framework forinterdisciplinary understanding of rivers as ecosystems Geomor-phology 89 147ndash162 2007

Dovers S R and Day D G Australian rivers and statute law En-viron Plan Law J 5 90ndash108 1988

Downs P W Gregory K J and Brookes A How integratedis river basin management Environ Manage 15 299ndash309httpsdoiorg101007BF02393876 1991

Emery J C Gurnell A M Clifford N J and Petts G E Char-acteristics and controls of gravel-bed riffles An analysis of datafrom the river-habitat survey Water Environ J 18 210ndash216httpsdoiorg101111j1747-65932004tb00535x 2004

Evans I S and Minaacuter J A classification of geomorphometric vari-ables in International Geom-orphometry 2011 Geomoprhome-tryorg Redlabds CA 105ndash108 2011

Fausch K D Torgersen C E Baxter C V Li H WView C The O F Is R To N How U Among ISet S For C and Fishes S Landscapes to riverscapesBridging the gap between research and conservation of streamfishes Bioscience 52 483ndash498 httpsdoiorg1016410006-3568(2002)052[0483LTRBTG]20CO2 2002

Glickman M E Rao S R and Schultz M R False discov-ery rate control is a recommended alternative to Bonferroni-typeadjustments in health studies J Clin Epidemiol 67 850ndash857httpsdoiorg101016jjclinepi201403012 2014

Gupta V K and Mesa O J Runoff generation and hydro-logic response via channel network geomorphology - Re-cent progress and open problems J Hydrol 102 3ndash28httpsdoiorg1010160022-1694(88)90089-3 1988

Gupta V K and Waymire E D On the formulation of an analyt-ical approach to hydrologic response and similarity at the basinscale J Hydrol 65 95ndash123 1983

Gupta V K Waymire E D and Rodriguez-Iturbe I On scalesgravity and network structure in basin runoff in Scale problemsin hydrology Springer Netherlands 159ndash184 1986

Gurnell A M Rinaldi M Belletti B Bizzi S Blamauer BBraca G Buijse A D Bussettini M Camenen B ComitiF Demarchi L Garciacutea de Jaloacuten D Gonzaacutelez del TaacutenagoM Grabowski R C Gunn I D M Habersack H Hen-driks D Henshaw A J Kloumlsch M Lastoria B LatapieA Marcinkowski P Martiacutenez-Fernaacutendez V Mosselman EMountford J O Nardi L Okruszko T OrsquoHare M T PalmaM Percopo C Surian N van de Bund W Weissteiner Cand Ziliani L A multi-scale hierarchical framework for devel-oping understanding of river behaviour to support river manage-ment Aquat Sci 78 1ndash16 httpsdoiorg101007s00027-015-0424-5 2016

Harvey G L Gurnell A M and Clifford N J Characterisationof river reaches The influence of rock type Catena 76 78ndash88httpsdoiorg101016jcatena200809010 2008

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2313

Figure 4 Summary of correlations between distance network density elevation network density stream order and RHS variables for allcatchments combined and each individual catchment Significance of correlation is indicated by point size with the largest points significantpost p-value correction Effect size (Kendallrsquos τ ) is indicated by colour No correlation was possible between stream order and 10th percentilesediment size in the Avon due to no variation in the RHS variable

flow-type speed associated with peaks in network density(Fig 3a) The Eden and Tamar however have positive cor-relations with mean and 90th percentile flow-type speed fordistance network density but negative correlations with me-dian and 10th percentile elevation network density The Wen-sum shows positive correlations between flow-type speed andelevation network density Sediment size has a consistent re-sponse to distance network density with the Eden and Wen-sum showing negative correlations with the sediment size(Fig 4) For elevation network density the Avon shows neg-ative correlations with sediment size whereas the Tamar andWensum show positive correlations (Fig 4)

There were few significant correlations between stream or-der and the RHS variables in individual catchments (Fig 4)The only significant correlation after p-value correction iswith 90th percentile HMS in the Wensum which shows astrong negative relationship

4 Discussion

41 A new approach to utilizing network topology incatchment-level analysis

Network density metrics represent an alternative approach toaccount for network topology in catchment-level studies op-timizing the width dimension of the network (or the num-

ber of links in the network) as opposed to the commonplacestream-order metric which only reflects the longitudinal po-sition of links in a network (Fig 1) This study demonstratesthat two topology metrics can be calculated simply from aDTM with GIS and using a broad-scale analysis of river at-tributes can be used to investigate the functional processeswithin catchments

While the two network density metrics have similar forms(ie forms are consistently unimodal or multi-modal) thespatial configuration of the distance and elevation intervalsused in the calculation of network density varies and mayimpact the effectiveness of each topological metric Distancenetwork density separates the catchment into intervals basedon distance which spread upstream from the outlet (Fig 2a)reflecting natural sub-basins within the fractal structure of thecatchment (Lashermes and Foufoula-Georgiou 2007) Thisdiffers from elevation network density which separates thecatchment into intervals based on elevation which radiate outfrom the main channel of the network (Fig 2b) The config-uration means that distance intervals contain streams that arein closer proximity to one another rather than the more distalconfiguration created by the elevation intervals suggesting adegree of spatial dependency in river functioning This hasbeen highlighted in previous studies where spatial networkstructure has a stronger influence on some in-channel pro-cesses than predictor variables such as elevation (Steel et al

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2314 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

2016) However elevation intervals contain RHS sites thatalthough they may be distal may have similar properties aselevation has been strongly related to RHS variables includ-ing flow type substrate etc in a number of studies (Jeffers1998a Naura et al 2016 Vaughan et al 2013)

42 Impacts of network topology on rivercharacteristics

River characteristics are assessed using the RHS dataset Theobservations made by the RHS dataset (Table 1) cannot of-fer the level of detail regarding geomorphological processthat river classifications which consider multiple scales canoffer (eg Brierley and Fryirs 2000 Gurnell et al 2016)While process-based classifications are preferable broad-scale monitoring datasets such as the River Habitat Surveymay still be useful when combined with map-derived datato explore controls on river characteristics (Harvey et al2008 Naura et al 2016 Vaughan et al 2013) Howeverthere are biases in RHS data collection an inherent limitationwhen using existing datasets (Vaughan and Ormerod 2010)specifically the standardized survey length of 500 m reachthat will capture differing amounts of natural variability de-pending on the size of the river While this must be notedthere are few significant correlations between river charac-teristics identified with stream order (Fig 4) which suggeststhat channel size is not influencing the RHS variables to agreat degree in these catchments

In this study it is anticipated that sites in high networkdensity areas will have higher levels of habitat diversity asindicated by previous studies of confluences and networks(Benda et al 2004a Best 1985 Rice 2017) in turn in-creasing mean sediment size and flow-type speed comparedto sites in low-density areas The results of the correlationsbetween distance network density and river characteristicswhen all catchments are combined support this hypothesiswith greater HQA flow-type speed and coarser sedimentsizes observed in areas with high distance network density(Fig 4)

For individual catchments elevation network density in-duces a stronger positive HQA response across all catch-ments than distance network density (Fig 4) This supportsthe evidence that individual confluences (Rice et al 2006)and high densities of confluences increase physical hetero-geneity within the river network (Benda et al 2004b Rice2017) However flow-type speed and sediment size responddifferently to network density in individual catchments

Slower flow types are observed in high network densityareas of the Avon and Tamar whereas faster flow types areobserved in high-elevation network density areas of the Wen-sum Individual confluences are shown to create numeroushigh- and low-speed flow environments (Best 1987) thatmay be observed in surface water topography (Biron et al2002) It was expected that the introduction of the additionalflow types by a high density of confluences in relatively low-

energy rivers would increase mean reach flow-type speedhowever the correlation analysis (Fig 4) suggests that insome catchments mean flow-type speed is reduced

Sediment size response also shows variation betweencatchments Sediment size is coarser in network-dense ar-eas of the Avon and Eden as expected but is finer in boththe Tamar and Wensum (Fig 4) The evidence from high-energy rivers shows that sediment becomes coarser down-stream and finer upstream of certain confluences (Benda etal 2004a Rice 1998) and in this case high numbers of con-fluences were expected to increase mean sediment size of thereach However the impact on sediment size is dependent onthe sediment calibre of the incoming tributary being higherthan the main channel with enough energy to transport thecoarse sediment to the confluence for numerous tributariesin an area The Tamar and Wensum have different rangesof sediment sizes with the Tamar having coarser sedimentthan the Wensum on average (Fig 3) This implies that trib-utaries in the Tamar may be energy limited not transportingcoarse sediments to confluences and the Wensum may suf-fer from inputs of fine sediments from the high percentageof arable land that concentrates in network-dense areas Thishas before been observed in a low-energy modified catch-ment where anthropogenic modifications in tributaries re-duced coarse sediment and flow capacity causing either lim-ited confluence impact or localized sediment fining (Singer2008)

Others have related the capacity of confluences to alterreach features to the morphometry of catchments with largerand more circular catchments containing a higher percentageof confluences that have a significant impact (Rice 2017)Based on this theory the Eden and Tamar are likely to havethe greatest impact as they are the most circular and steepestof the four catchments (although the Tamar is the smallestby area) Yet there is no clear pattern indicating that thesecatchments respond differently than the Avon and Wensum(Fig 4) with catchments responding differently to differ-ent variables This perhaps suggests that rather than networkdensity having a directional impact on factors such as flow-type speed and sediment size it has an impact on overall het-erogeneity at the reach level as suggested by previous stud-ies and that specific directional change occurs at the sub-reach level

An increase in channel modification is also hypothesizeddue to the increase in flood peak downstream from conflu-ences (Depettris et al 2000) and the scour and erosion as-sociated with confluence junctions (Best 1986) potentiallyincreasing the need for bed and bank protection The corre-lations between distance network density and HMS when allcatchments are combined undermine this hypothesis show-ing less channel modification where distance network den-sity is higher (Fig 4) There were few significant correlationswith individual catchments but HMS in the Eden was consis-tently observed to be higher in network-dense areas (Fig 4)This may be due to the Edenrsquos high elevation and steep to-

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2315

pography inducing a higher energy environment where scourand erosion processes in areas with high numbers of conflu-ences would be more likely to be present

Differences in RHS variable responses also differed be-tween the descriptive statistics considered Often the ex-tremes 90th and 10th percentiles showed more significantand stronger correlations than the mean or median (Fig 4)This may reflect findings from previous studies which sug-gest that not all confluences cause reach-scale changes (Rice1998) that perhaps the changes to river character induced bycertain confluences only influence certain reaches whereasothers are left unaffected creating less pronounced responsesin the mean and median of variables External factors mayalso influence this trend for example 10th percentile HMSvisually responds dramatically to network density metrics(Fig 3a and b) compared to the other descriptive statisticsThis suggests that the most natural sites (ie with the low-est HMS score) are responding differently to network den-sity with the most natural sites having less modificationin network-dense areas whereas less natural sites becomemore modified in network-dense areas This reflects the HQAscore results which visually (Fig 3a and b) and statistically(Fig 4) vary with distance network density except for the10th percentile These sites with the lowest habitat qualityand naturalness are likely influenced by anthropogenic fac-tors that are independent of network density reducing habitatquality scores at impacted sites

The two network density metrics have differing impacts onriver characteristics While distance network density showsconsistently significant correlations when all catchments arecombined individual catchments respond more frequentlyand more strongly to elevation network density (Fig 4)This may be because there is a dramatic split in distancenetwork density values between the more upland drainage-dense catchments Eden and Tamar than the lowland chalklow-drainage catchments Avon and Wensum The combinedcorrelation will therefore in part reflect the difference be-tween the catchments which have different ranges of rivercharacteristics (Fig 3) This is not the case for elevation net-work density which has higher density values in the Tamarand Wensum than the Avon and Eden so therefore the char-acteristics of the catchments will have less bearing on thecombined correlation However there are patterns identifiedwith distance network density in individual catchments thatare not present with elevation network density increasedflow-type speed and sediment size in the Eden and reducedflow-type speed in the Avon with network density (Fig 4)which show its usefulness

The results suggest that the distance network density andelevation network density metrics quantify different dimen-sions of network topology which are shown to exhibit func-tionally meaningful patterns for river reach characteristicsbased on the correlation results Perhaps within catchmentselevation network density provides the more powerful met-ric for individual catchments but distance network density

better accounts for the drainage density of the catchmentsallowing it to be applied across multiple catchments

43 Comparison of stream order to network densitymetrics

The classic method of accounting for network topologystream order is critiqued for failing to represent disconti-nuities in the network and simplifying the network into agradient The number of links in different stream orders isconsistent across all catchments not reflecting the internalstructure of the network or the variety between catchmentsthat is achieved by the distance network density and eleva-tion network density metrics The analysis of the two net-work density metrics presented in this paper shows that dis-tances from source and elevation are not mutually exclusive(Fig 2a and b) contrary to the stream-order metric whichrepresents streams as upstream to downstream or upland tolowland

Stream order has few significant correlations with manyof the river characteristics considered in this study Negativecorrelations with HMS in the Wensum were statistically sig-nificant post p-value correction likely driving the significantrelationship with all catchments combined for this variable(Fig 4) This suggests that modification is greater upstreamin the Wensum contrary to ideas that downstream reachesmay show greater anthropogenic modification Intense agri-cultural land use in the upper reaches of the Wensum islikely to be the cause of the high HMS scores upstreamThe lack of significant correlations suggests that stream or-der and therefore an upstream-to-downstream gradient arenot the predominant pattern in river characteristics despitethe description of such a gradient by geomorphic (Schumm1977) and ecological frameworks (Vannote et al 1980)This is surprising as distance and elevation which both re-flect the upstream-to-downstream gradient have proven tobe important factors in previous studies explaining patternsof RHS features at a national level (Jeffers 1998a Vaughanet al 2013) This implies that upstream-to-downstream gra-dient may not sufficiently reflect patterns of river charac-teristics through the river network within individual catch-ments Others have also found that stream order has weakand inconsistent relationships with biodiversity patterns inriver systems arguing that the topological measure has no di-rect mechanistic control on biodiversity (Vander Vorste et al2017) Instead this study finds that the network density met-rics are a more powerful metric which conceptually providean improved method of accounting for the impacts of net-work topology on the fluvial system exhibiting relationshipswith river characteristics particularly habitat quality score(Fig 4)

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2316 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

44 Applicability of network topology metrics todifferent environments

Much of the seminal work on network and confluence im-pacts (eg the Network Dynamics Hypothesis Benda et al2004b and Link Discontinuity Concept Rice et al 2001) isconducted in natural highly erosive catchments with first-hand empirical measurements However in an age whenrivers and their catchments are increasingly altered by an-thropogenic modification (Meybeck 2003) contemporarystudies must not only aim to expand knowledge but also findmethods of transferring knowledge to many increasingly al-tered catchments (Clifford 2002)

The catchments selected by this study are more greatlymodified and although they reflect a range of fluvial envi-ronments in England are lower-energy catchments than theseminal studies Benda et al (2004a) suggest that confluenceeffects in less active landscapes would be subdued comparedto highly erosive landscapes but the evidence presented heredemonstrates the utility of evaluating network topologicalstructure in studies on catchment-level effects in any typeof fluvial environment including those with low-energy andwidespread anthropogenic changes The response of someriver characteristics varied between catchments observationsof flow-type speed sediment size and modifications showeddifferent responses to network density in different catch-ments This suggests that the functional effect of these topo-logical metrics is catchment dependent and likely is influ-enced by external catchment characteristics such as land usenot considered in this study although impact did not appearto vary with catchment topography or circularity as has beenshown in prior studies (Benda et al 2004a Rice 2017) Thisshould be explored further in future research to enable rec-ommendations to be made regarding where and how reachesmay respond to network density The response of habitatquality score was however consistent across catchments andbetween metrics showing that habitat quality is greater inareas with high network density (Fig 4) as hypothesizedby the Network Dynamics Hypothesis (Benda et al 2004b)and demonstrated by studies on individual catchments (Rice2017 Rice et al 2006)

The methods presented in this paper are designed to be im-plemented in any catchment with a dendritic network struc-ture The topology metrics can easily be calculated from anydendritic network with DTM data using GIS and comparedto any site-scale data This study uses regulatory monitoringdatasets so that analysis is targeted to assessment scores andphysical features of interest to river managers Also the highvolume and wide spatial extent of data available from regu-latory sources allow for between-catchment comparisons

5 Conclusions

Although appreciation of catchment-level effects is consid-ered the epitome of understanding river functioning a keycomponent of the catchment ndash the river network ndash is over-looked and oversimplified by catchment-level studies Thisstudy finds that river network density plays a role in struc-turing the distribution river characteristics throughout thecatchment offering more detailed explanation than the clas-sic stream-order metric Network-dense areas are generallyfound to have higher habitat quality and diversity but modi-fication flow-type speed and sediment size show different re-sponses in different catchments This study suggests two po-tential reasons for this (1) there is evidence that confluencesin the river network increase diversity as is observed in thisstudy so the direction of mean river characteristic responsemay not be consistent and (2) there may be external factorssuch as sediment availability land cover and anthropogenicmodification that alter the direction of mean river characteris-tic response This paper demonstrates the functional responseof river characteristics to network topology and suggests thatthe inclusion of network topology in catchment-level studieswould add a layer of function-based understanding to suchstudies linking reaches to their catchments

The broad-scale methodology adopted by this study allowsthe network density metrics which are easily extracted fromopen-source data using GIS software to be compared to anyregulatory dataset The use of regulatory datasets allows notonly for analysis over a wider spatial extent but also for moreapplicable results for regulatory bodies Therefore the inter-disciplinary approach to characterizing network topology canbe applied efficiently and effectively to capture catchment-level impacts on reach-level functioning in any catchmentacross the globe

Data availability River Habitat Survey data are freely availablefrom the Environment Agency England River network data andelevation data are available from the Centre for Ecology and Hy-drology UK

Author contributions EH and NC conceptualised and designed thestudy EH performed the data analysis with the support of JMEH wrote the paper with comments provided by NC and JM

Competing interests The authors declare that they have no conflictof interest

Acknowledgements The authors would like to thank the Environ-ment Agency and the Centre for Ecology and Hydrology for accessto the data used in this paper This research is funded by the NaturalEnvironmental Research Council (NERC)

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2317

Review statement This paper was edited by Nandita Basu and re-viewed by Simone Bizzi and one anonymous referee

References

Belletti B Rinaldi M Buijse A D Gurnell A Mand Mosselman E A review of assessment methods forriver hydromorphology Environ Earth Sci 73 2079ndash2100httpsdoiorg101007s12665-014-3558-1 2015

Benda L Andras K Miller D and Bigelow P Confluenceeffects in rivers Interactions of basin scale network geom-etry and disturbance regimes Water Resour Res 40 1ndash15httpsdoiorg1010292003WR002583 2004a

Benda L Poff N L Miller D Dunne T Reeves GPess G and Pollock M The network dynamics hy-pothesis how channel networks structure riverine habi-tats Bioscience 54 413ndash427 httpsdoiorg1016410006-3568(2004)054[0413TNDHHC]20CO2 2004b

Benjamini Y and Hochberg Y Controlling the False DiscoveryRate A Practical and Powerful Approach to Multiple TestingJ Roy Stat Soc B 57 289ndash300 1995

Best J L Flow dynamics and sediment transport at river channelconfluences Birbeck University of London London 1985

Best J L The morphology of river channel con-fluences Prog Phys Geogr 10 157ndash174httpsdoiorg101177030913338601000201 1986

Best J L Flow dynamics at river channel confluences implica-tions for sediment transport and bed morphology in Recent de-velopments in fluvial sedimentology edited by Ethridge F GFlores R M and Harvey M D Spec Publ Soc Econ Pale-ontol Miner Tulsa Okla 27ndash35 1987

Biron P M Richer A Kirkbride A D Roy A G andHan S Spatial patterns of water surface topography ata river confluence Earth Surf Proc Land 27 913ndash928httpsdoiorg101002esp359 2002

Bravard J P and Gilvear D J Hydrological and geomorpho-logical structure of hydrosystems in The Fluvial Hydrosystemedited by Petts G E and Amoros C Springer Netherlands98ndash116 1996

Brierley G and Fryirs K River Styles a Geomorphic Approachto Catchment Characterization Implications for River Rehabili-tation in Bega Catchment New South Wales Australia EnvironManage 25 661ndash679 2000

Bunn S E and Arthington A H Basic principles and ecologi-cal consequences of altered flow regimes for aquatic biodiversityEnviron Manage 30 492ndash507 httpsdoiorg101007s00267-002-2737-0 2002

Church M and Kellerhals R On the statistics of grain size varia-tion along a gravel river Can J Earth Sci 15 1151ndash1160 1978

Clifford N J Hydrology the changing paradigm Prog Phys Ge-ogr 26 290ndash301 2002

Cohen P Andriamahefa H and Wasson J-G Towardsa regionalizaton of aquatic habitat distribution of meso-habitats at the scale of a large basin Regul RiversRes Manage 14 391ndash404 httpsdoiorg101002(SICI)1099-1646(19980910)145lt391AID-RRR513gt30CO2-W 1998

Davenport A J Gurnell A M and Armitage P D Habitat sur-vey and classification of urban rivers River Res Appl 20 687ndash704 httpsdoiorg101002rra785 2004

Davies N M Norris R H and Thoms M C Prediction and as-sessment of local stream habitat features using large-scale catch-ment characteristics Freshwater Biol 45 343ndash369 2000

Depettris C Mendiondo E M Neiff J and Rohrmann H Flooddefence strategy at the confluence of the Paranaacute-Paraguay riversProc Int Symp Flood Def Kassel He C31 C40 2000

Dollar E James C Rogers K and Thoms M A framework forinterdisciplinary understanding of rivers as ecosystems Geomor-phology 89 147ndash162 2007

Dovers S R and Day D G Australian rivers and statute law En-viron Plan Law J 5 90ndash108 1988

Downs P W Gregory K J and Brookes A How integratedis river basin management Environ Manage 15 299ndash309httpsdoiorg101007BF02393876 1991

Emery J C Gurnell A M Clifford N J and Petts G E Char-acteristics and controls of gravel-bed riffles An analysis of datafrom the river-habitat survey Water Environ J 18 210ndash216httpsdoiorg101111j1747-65932004tb00535x 2004

Evans I S and Minaacuter J A classification of geomorphometric vari-ables in International Geom-orphometry 2011 Geomoprhome-tryorg Redlabds CA 105ndash108 2011

Fausch K D Torgersen C E Baxter C V Li H WView C The O F Is R To N How U Among ISet S For C and Fishes S Landscapes to riverscapesBridging the gap between research and conservation of streamfishes Bioscience 52 483ndash498 httpsdoiorg1016410006-3568(2002)052[0483LTRBTG]20CO2 2002

Glickman M E Rao S R and Schultz M R False discov-ery rate control is a recommended alternative to Bonferroni-typeadjustments in health studies J Clin Epidemiol 67 850ndash857httpsdoiorg101016jjclinepi201403012 2014

Gupta V K and Mesa O J Runoff generation and hydro-logic response via channel network geomorphology - Re-cent progress and open problems J Hydrol 102 3ndash28httpsdoiorg1010160022-1694(88)90089-3 1988

Gupta V K and Waymire E D On the formulation of an analyt-ical approach to hydrologic response and similarity at the basinscale J Hydrol 65 95ndash123 1983

Gupta V K Waymire E D and Rodriguez-Iturbe I On scalesgravity and network structure in basin runoff in Scale problemsin hydrology Springer Netherlands 159ndash184 1986

Gurnell A M Rinaldi M Belletti B Bizzi S Blamauer BBraca G Buijse A D Bussettini M Camenen B ComitiF Demarchi L Garciacutea de Jaloacuten D Gonzaacutelez del TaacutenagoM Grabowski R C Gunn I D M Habersack H Hen-driks D Henshaw A J Kloumlsch M Lastoria B LatapieA Marcinkowski P Martiacutenez-Fernaacutendez V Mosselman EMountford J O Nardi L Okruszko T OrsquoHare M T PalmaM Percopo C Surian N van de Bund W Weissteiner Cand Ziliani L A multi-scale hierarchical framework for devel-oping understanding of river behaviour to support river manage-ment Aquat Sci 78 1ndash16 httpsdoiorg101007s00027-015-0424-5 2016

Harvey G L Gurnell A M and Clifford N J Characterisationof river reaches The influence of rock type Catena 76 78ndash88httpsdoiorg101016jcatena200809010 2008

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2314 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

2016) However elevation intervals contain RHS sites thatalthough they may be distal may have similar properties aselevation has been strongly related to RHS variables includ-ing flow type substrate etc in a number of studies (Jeffers1998a Naura et al 2016 Vaughan et al 2013)

42 Impacts of network topology on rivercharacteristics

River characteristics are assessed using the RHS dataset Theobservations made by the RHS dataset (Table 1) cannot of-fer the level of detail regarding geomorphological processthat river classifications which consider multiple scales canoffer (eg Brierley and Fryirs 2000 Gurnell et al 2016)While process-based classifications are preferable broad-scale monitoring datasets such as the River Habitat Surveymay still be useful when combined with map-derived datato explore controls on river characteristics (Harvey et al2008 Naura et al 2016 Vaughan et al 2013) Howeverthere are biases in RHS data collection an inherent limitationwhen using existing datasets (Vaughan and Ormerod 2010)specifically the standardized survey length of 500 m reachthat will capture differing amounts of natural variability de-pending on the size of the river While this must be notedthere are few significant correlations between river charac-teristics identified with stream order (Fig 4) which suggeststhat channel size is not influencing the RHS variables to agreat degree in these catchments

In this study it is anticipated that sites in high networkdensity areas will have higher levels of habitat diversity asindicated by previous studies of confluences and networks(Benda et al 2004a Best 1985 Rice 2017) in turn in-creasing mean sediment size and flow-type speed comparedto sites in low-density areas The results of the correlationsbetween distance network density and river characteristicswhen all catchments are combined support this hypothesiswith greater HQA flow-type speed and coarser sedimentsizes observed in areas with high distance network density(Fig 4)

For individual catchments elevation network density in-duces a stronger positive HQA response across all catch-ments than distance network density (Fig 4) This supportsthe evidence that individual confluences (Rice et al 2006)and high densities of confluences increase physical hetero-geneity within the river network (Benda et al 2004b Rice2017) However flow-type speed and sediment size responddifferently to network density in individual catchments

Slower flow types are observed in high network densityareas of the Avon and Tamar whereas faster flow types areobserved in high-elevation network density areas of the Wen-sum Individual confluences are shown to create numeroushigh- and low-speed flow environments (Best 1987) thatmay be observed in surface water topography (Biron et al2002) It was expected that the introduction of the additionalflow types by a high density of confluences in relatively low-

energy rivers would increase mean reach flow-type speedhowever the correlation analysis (Fig 4) suggests that insome catchments mean flow-type speed is reduced

Sediment size response also shows variation betweencatchments Sediment size is coarser in network-dense ar-eas of the Avon and Eden as expected but is finer in boththe Tamar and Wensum (Fig 4) The evidence from high-energy rivers shows that sediment becomes coarser down-stream and finer upstream of certain confluences (Benda etal 2004a Rice 1998) and in this case high numbers of con-fluences were expected to increase mean sediment size of thereach However the impact on sediment size is dependent onthe sediment calibre of the incoming tributary being higherthan the main channel with enough energy to transport thecoarse sediment to the confluence for numerous tributariesin an area The Tamar and Wensum have different rangesof sediment sizes with the Tamar having coarser sedimentthan the Wensum on average (Fig 3) This implies that trib-utaries in the Tamar may be energy limited not transportingcoarse sediments to confluences and the Wensum may suf-fer from inputs of fine sediments from the high percentageof arable land that concentrates in network-dense areas Thishas before been observed in a low-energy modified catch-ment where anthropogenic modifications in tributaries re-duced coarse sediment and flow capacity causing either lim-ited confluence impact or localized sediment fining (Singer2008)

Others have related the capacity of confluences to alterreach features to the morphometry of catchments with largerand more circular catchments containing a higher percentageof confluences that have a significant impact (Rice 2017)Based on this theory the Eden and Tamar are likely to havethe greatest impact as they are the most circular and steepestof the four catchments (although the Tamar is the smallestby area) Yet there is no clear pattern indicating that thesecatchments respond differently than the Avon and Wensum(Fig 4) with catchments responding differently to differ-ent variables This perhaps suggests that rather than networkdensity having a directional impact on factors such as flow-type speed and sediment size it has an impact on overall het-erogeneity at the reach level as suggested by previous stud-ies and that specific directional change occurs at the sub-reach level

An increase in channel modification is also hypothesizeddue to the increase in flood peak downstream from conflu-ences (Depettris et al 2000) and the scour and erosion as-sociated with confluence junctions (Best 1986) potentiallyincreasing the need for bed and bank protection The corre-lations between distance network density and HMS when allcatchments are combined undermine this hypothesis show-ing less channel modification where distance network den-sity is higher (Fig 4) There were few significant correlationswith individual catchments but HMS in the Eden was consis-tently observed to be higher in network-dense areas (Fig 4)This may be due to the Edenrsquos high elevation and steep to-

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2315

pography inducing a higher energy environment where scourand erosion processes in areas with high numbers of conflu-ences would be more likely to be present

Differences in RHS variable responses also differed be-tween the descriptive statistics considered Often the ex-tremes 90th and 10th percentiles showed more significantand stronger correlations than the mean or median (Fig 4)This may reflect findings from previous studies which sug-gest that not all confluences cause reach-scale changes (Rice1998) that perhaps the changes to river character induced bycertain confluences only influence certain reaches whereasothers are left unaffected creating less pronounced responsesin the mean and median of variables External factors mayalso influence this trend for example 10th percentile HMSvisually responds dramatically to network density metrics(Fig 3a and b) compared to the other descriptive statisticsThis suggests that the most natural sites (ie with the low-est HMS score) are responding differently to network den-sity with the most natural sites having less modificationin network-dense areas whereas less natural sites becomemore modified in network-dense areas This reflects the HQAscore results which visually (Fig 3a and b) and statistically(Fig 4) vary with distance network density except for the10th percentile These sites with the lowest habitat qualityand naturalness are likely influenced by anthropogenic fac-tors that are independent of network density reducing habitatquality scores at impacted sites

The two network density metrics have differing impacts onriver characteristics While distance network density showsconsistently significant correlations when all catchments arecombined individual catchments respond more frequentlyand more strongly to elevation network density (Fig 4)This may be because there is a dramatic split in distancenetwork density values between the more upland drainage-dense catchments Eden and Tamar than the lowland chalklow-drainage catchments Avon and Wensum The combinedcorrelation will therefore in part reflect the difference be-tween the catchments which have different ranges of rivercharacteristics (Fig 3) This is not the case for elevation net-work density which has higher density values in the Tamarand Wensum than the Avon and Eden so therefore the char-acteristics of the catchments will have less bearing on thecombined correlation However there are patterns identifiedwith distance network density in individual catchments thatare not present with elevation network density increasedflow-type speed and sediment size in the Eden and reducedflow-type speed in the Avon with network density (Fig 4)which show its usefulness

The results suggest that the distance network density andelevation network density metrics quantify different dimen-sions of network topology which are shown to exhibit func-tionally meaningful patterns for river reach characteristicsbased on the correlation results Perhaps within catchmentselevation network density provides the more powerful met-ric for individual catchments but distance network density

better accounts for the drainage density of the catchmentsallowing it to be applied across multiple catchments

43 Comparison of stream order to network densitymetrics

The classic method of accounting for network topologystream order is critiqued for failing to represent disconti-nuities in the network and simplifying the network into agradient The number of links in different stream orders isconsistent across all catchments not reflecting the internalstructure of the network or the variety between catchmentsthat is achieved by the distance network density and eleva-tion network density metrics The analysis of the two net-work density metrics presented in this paper shows that dis-tances from source and elevation are not mutually exclusive(Fig 2a and b) contrary to the stream-order metric whichrepresents streams as upstream to downstream or upland tolowland

Stream order has few significant correlations with manyof the river characteristics considered in this study Negativecorrelations with HMS in the Wensum were statistically sig-nificant post p-value correction likely driving the significantrelationship with all catchments combined for this variable(Fig 4) This suggests that modification is greater upstreamin the Wensum contrary to ideas that downstream reachesmay show greater anthropogenic modification Intense agri-cultural land use in the upper reaches of the Wensum islikely to be the cause of the high HMS scores upstreamThe lack of significant correlations suggests that stream or-der and therefore an upstream-to-downstream gradient arenot the predominant pattern in river characteristics despitethe description of such a gradient by geomorphic (Schumm1977) and ecological frameworks (Vannote et al 1980)This is surprising as distance and elevation which both re-flect the upstream-to-downstream gradient have proven tobe important factors in previous studies explaining patternsof RHS features at a national level (Jeffers 1998a Vaughanet al 2013) This implies that upstream-to-downstream gra-dient may not sufficiently reflect patterns of river charac-teristics through the river network within individual catch-ments Others have also found that stream order has weakand inconsistent relationships with biodiversity patterns inriver systems arguing that the topological measure has no di-rect mechanistic control on biodiversity (Vander Vorste et al2017) Instead this study finds that the network density met-rics are a more powerful metric which conceptually providean improved method of accounting for the impacts of net-work topology on the fluvial system exhibiting relationshipswith river characteristics particularly habitat quality score(Fig 4)

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2316 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

44 Applicability of network topology metrics todifferent environments

Much of the seminal work on network and confluence im-pacts (eg the Network Dynamics Hypothesis Benda et al2004b and Link Discontinuity Concept Rice et al 2001) isconducted in natural highly erosive catchments with first-hand empirical measurements However in an age whenrivers and their catchments are increasingly altered by an-thropogenic modification (Meybeck 2003) contemporarystudies must not only aim to expand knowledge but also findmethods of transferring knowledge to many increasingly al-tered catchments (Clifford 2002)

The catchments selected by this study are more greatlymodified and although they reflect a range of fluvial envi-ronments in England are lower-energy catchments than theseminal studies Benda et al (2004a) suggest that confluenceeffects in less active landscapes would be subdued comparedto highly erosive landscapes but the evidence presented heredemonstrates the utility of evaluating network topologicalstructure in studies on catchment-level effects in any typeof fluvial environment including those with low-energy andwidespread anthropogenic changes The response of someriver characteristics varied between catchments observationsof flow-type speed sediment size and modifications showeddifferent responses to network density in different catch-ments This suggests that the functional effect of these topo-logical metrics is catchment dependent and likely is influ-enced by external catchment characteristics such as land usenot considered in this study although impact did not appearto vary with catchment topography or circularity as has beenshown in prior studies (Benda et al 2004a Rice 2017) Thisshould be explored further in future research to enable rec-ommendations to be made regarding where and how reachesmay respond to network density The response of habitatquality score was however consistent across catchments andbetween metrics showing that habitat quality is greater inareas with high network density (Fig 4) as hypothesizedby the Network Dynamics Hypothesis (Benda et al 2004b)and demonstrated by studies on individual catchments (Rice2017 Rice et al 2006)

The methods presented in this paper are designed to be im-plemented in any catchment with a dendritic network struc-ture The topology metrics can easily be calculated from anydendritic network with DTM data using GIS and comparedto any site-scale data This study uses regulatory monitoringdatasets so that analysis is targeted to assessment scores andphysical features of interest to river managers Also the highvolume and wide spatial extent of data available from regu-latory sources allow for between-catchment comparisons

5 Conclusions

Although appreciation of catchment-level effects is consid-ered the epitome of understanding river functioning a keycomponent of the catchment ndash the river network ndash is over-looked and oversimplified by catchment-level studies Thisstudy finds that river network density plays a role in struc-turing the distribution river characteristics throughout thecatchment offering more detailed explanation than the clas-sic stream-order metric Network-dense areas are generallyfound to have higher habitat quality and diversity but modi-fication flow-type speed and sediment size show different re-sponses in different catchments This study suggests two po-tential reasons for this (1) there is evidence that confluencesin the river network increase diversity as is observed in thisstudy so the direction of mean river characteristic responsemay not be consistent and (2) there may be external factorssuch as sediment availability land cover and anthropogenicmodification that alter the direction of mean river characteris-tic response This paper demonstrates the functional responseof river characteristics to network topology and suggests thatthe inclusion of network topology in catchment-level studieswould add a layer of function-based understanding to suchstudies linking reaches to their catchments

The broad-scale methodology adopted by this study allowsthe network density metrics which are easily extracted fromopen-source data using GIS software to be compared to anyregulatory dataset The use of regulatory datasets allows notonly for analysis over a wider spatial extent but also for moreapplicable results for regulatory bodies Therefore the inter-disciplinary approach to characterizing network topology canbe applied efficiently and effectively to capture catchment-level impacts on reach-level functioning in any catchmentacross the globe

Data availability River Habitat Survey data are freely availablefrom the Environment Agency England River network data andelevation data are available from the Centre for Ecology and Hy-drology UK

Author contributions EH and NC conceptualised and designed thestudy EH performed the data analysis with the support of JMEH wrote the paper with comments provided by NC and JM

Competing interests The authors declare that they have no conflictof interest

Acknowledgements The authors would like to thank the Environ-ment Agency and the Centre for Ecology and Hydrology for accessto the data used in this paper This research is funded by the NaturalEnvironmental Research Council (NERC)

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2317

Review statement This paper was edited by Nandita Basu and re-viewed by Simone Bizzi and one anonymous referee

References

Belletti B Rinaldi M Buijse A D Gurnell A Mand Mosselman E A review of assessment methods forriver hydromorphology Environ Earth Sci 73 2079ndash2100httpsdoiorg101007s12665-014-3558-1 2015

Benda L Andras K Miller D and Bigelow P Confluenceeffects in rivers Interactions of basin scale network geom-etry and disturbance regimes Water Resour Res 40 1ndash15httpsdoiorg1010292003WR002583 2004a

Benda L Poff N L Miller D Dunne T Reeves GPess G and Pollock M The network dynamics hy-pothesis how channel networks structure riverine habi-tats Bioscience 54 413ndash427 httpsdoiorg1016410006-3568(2004)054[0413TNDHHC]20CO2 2004b

Benjamini Y and Hochberg Y Controlling the False DiscoveryRate A Practical and Powerful Approach to Multiple TestingJ Roy Stat Soc B 57 289ndash300 1995

Best J L Flow dynamics and sediment transport at river channelconfluences Birbeck University of London London 1985

Best J L The morphology of river channel con-fluences Prog Phys Geogr 10 157ndash174httpsdoiorg101177030913338601000201 1986

Best J L Flow dynamics at river channel confluences implica-tions for sediment transport and bed morphology in Recent de-velopments in fluvial sedimentology edited by Ethridge F GFlores R M and Harvey M D Spec Publ Soc Econ Pale-ontol Miner Tulsa Okla 27ndash35 1987

Biron P M Richer A Kirkbride A D Roy A G andHan S Spatial patterns of water surface topography ata river confluence Earth Surf Proc Land 27 913ndash928httpsdoiorg101002esp359 2002

Bravard J P and Gilvear D J Hydrological and geomorpho-logical structure of hydrosystems in The Fluvial Hydrosystemedited by Petts G E and Amoros C Springer Netherlands98ndash116 1996

Brierley G and Fryirs K River Styles a Geomorphic Approachto Catchment Characterization Implications for River Rehabili-tation in Bega Catchment New South Wales Australia EnvironManage 25 661ndash679 2000

Bunn S E and Arthington A H Basic principles and ecologi-cal consequences of altered flow regimes for aquatic biodiversityEnviron Manage 30 492ndash507 httpsdoiorg101007s00267-002-2737-0 2002

Church M and Kellerhals R On the statistics of grain size varia-tion along a gravel river Can J Earth Sci 15 1151ndash1160 1978

Clifford N J Hydrology the changing paradigm Prog Phys Ge-ogr 26 290ndash301 2002

Cohen P Andriamahefa H and Wasson J-G Towardsa regionalizaton of aquatic habitat distribution of meso-habitats at the scale of a large basin Regul RiversRes Manage 14 391ndash404 httpsdoiorg101002(SICI)1099-1646(19980910)145lt391AID-RRR513gt30CO2-W 1998

Davenport A J Gurnell A M and Armitage P D Habitat sur-vey and classification of urban rivers River Res Appl 20 687ndash704 httpsdoiorg101002rra785 2004

Davies N M Norris R H and Thoms M C Prediction and as-sessment of local stream habitat features using large-scale catch-ment characteristics Freshwater Biol 45 343ndash369 2000

Depettris C Mendiondo E M Neiff J and Rohrmann H Flooddefence strategy at the confluence of the Paranaacute-Paraguay riversProc Int Symp Flood Def Kassel He C31 C40 2000

Dollar E James C Rogers K and Thoms M A framework forinterdisciplinary understanding of rivers as ecosystems Geomor-phology 89 147ndash162 2007

Dovers S R and Day D G Australian rivers and statute law En-viron Plan Law J 5 90ndash108 1988

Downs P W Gregory K J and Brookes A How integratedis river basin management Environ Manage 15 299ndash309httpsdoiorg101007BF02393876 1991

Emery J C Gurnell A M Clifford N J and Petts G E Char-acteristics and controls of gravel-bed riffles An analysis of datafrom the river-habitat survey Water Environ J 18 210ndash216httpsdoiorg101111j1747-65932004tb00535x 2004

Evans I S and Minaacuter J A classification of geomorphometric vari-ables in International Geom-orphometry 2011 Geomoprhome-tryorg Redlabds CA 105ndash108 2011

Fausch K D Torgersen C E Baxter C V Li H WView C The O F Is R To N How U Among ISet S For C and Fishes S Landscapes to riverscapesBridging the gap between research and conservation of streamfishes Bioscience 52 483ndash498 httpsdoiorg1016410006-3568(2002)052[0483LTRBTG]20CO2 2002

Glickman M E Rao S R and Schultz M R False discov-ery rate control is a recommended alternative to Bonferroni-typeadjustments in health studies J Clin Epidemiol 67 850ndash857httpsdoiorg101016jjclinepi201403012 2014

Gupta V K and Mesa O J Runoff generation and hydro-logic response via channel network geomorphology - Re-cent progress and open problems J Hydrol 102 3ndash28httpsdoiorg1010160022-1694(88)90089-3 1988

Gupta V K and Waymire E D On the formulation of an analyt-ical approach to hydrologic response and similarity at the basinscale J Hydrol 65 95ndash123 1983

Gupta V K Waymire E D and Rodriguez-Iturbe I On scalesgravity and network structure in basin runoff in Scale problemsin hydrology Springer Netherlands 159ndash184 1986

Gurnell A M Rinaldi M Belletti B Bizzi S Blamauer BBraca G Buijse A D Bussettini M Camenen B ComitiF Demarchi L Garciacutea de Jaloacuten D Gonzaacutelez del TaacutenagoM Grabowski R C Gunn I D M Habersack H Hen-driks D Henshaw A J Kloumlsch M Lastoria B LatapieA Marcinkowski P Martiacutenez-Fernaacutendez V Mosselman EMountford J O Nardi L Okruszko T OrsquoHare M T PalmaM Percopo C Surian N van de Bund W Weissteiner Cand Ziliani L A multi-scale hierarchical framework for devel-oping understanding of river behaviour to support river manage-ment Aquat Sci 78 1ndash16 httpsdoiorg101007s00027-015-0424-5 2016

Harvey G L Gurnell A M and Clifford N J Characterisationof river reaches The influence of rock type Catena 76 78ndash88httpsdoiorg101016jcatena200809010 2008

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2315

pography inducing a higher energy environment where scourand erosion processes in areas with high numbers of conflu-ences would be more likely to be present

Differences in RHS variable responses also differed be-tween the descriptive statistics considered Often the ex-tremes 90th and 10th percentiles showed more significantand stronger correlations than the mean or median (Fig 4)This may reflect findings from previous studies which sug-gest that not all confluences cause reach-scale changes (Rice1998) that perhaps the changes to river character induced bycertain confluences only influence certain reaches whereasothers are left unaffected creating less pronounced responsesin the mean and median of variables External factors mayalso influence this trend for example 10th percentile HMSvisually responds dramatically to network density metrics(Fig 3a and b) compared to the other descriptive statisticsThis suggests that the most natural sites (ie with the low-est HMS score) are responding differently to network den-sity with the most natural sites having less modificationin network-dense areas whereas less natural sites becomemore modified in network-dense areas This reflects the HQAscore results which visually (Fig 3a and b) and statistically(Fig 4) vary with distance network density except for the10th percentile These sites with the lowest habitat qualityand naturalness are likely influenced by anthropogenic fac-tors that are independent of network density reducing habitatquality scores at impacted sites

The two network density metrics have differing impacts onriver characteristics While distance network density showsconsistently significant correlations when all catchments arecombined individual catchments respond more frequentlyand more strongly to elevation network density (Fig 4)This may be because there is a dramatic split in distancenetwork density values between the more upland drainage-dense catchments Eden and Tamar than the lowland chalklow-drainage catchments Avon and Wensum The combinedcorrelation will therefore in part reflect the difference be-tween the catchments which have different ranges of rivercharacteristics (Fig 3) This is not the case for elevation net-work density which has higher density values in the Tamarand Wensum than the Avon and Eden so therefore the char-acteristics of the catchments will have less bearing on thecombined correlation However there are patterns identifiedwith distance network density in individual catchments thatare not present with elevation network density increasedflow-type speed and sediment size in the Eden and reducedflow-type speed in the Avon with network density (Fig 4)which show its usefulness

The results suggest that the distance network density andelevation network density metrics quantify different dimen-sions of network topology which are shown to exhibit func-tionally meaningful patterns for river reach characteristicsbased on the correlation results Perhaps within catchmentselevation network density provides the more powerful met-ric for individual catchments but distance network density

better accounts for the drainage density of the catchmentsallowing it to be applied across multiple catchments

43 Comparison of stream order to network densitymetrics

The classic method of accounting for network topologystream order is critiqued for failing to represent disconti-nuities in the network and simplifying the network into agradient The number of links in different stream orders isconsistent across all catchments not reflecting the internalstructure of the network or the variety between catchmentsthat is achieved by the distance network density and eleva-tion network density metrics The analysis of the two net-work density metrics presented in this paper shows that dis-tances from source and elevation are not mutually exclusive(Fig 2a and b) contrary to the stream-order metric whichrepresents streams as upstream to downstream or upland tolowland

Stream order has few significant correlations with manyof the river characteristics considered in this study Negativecorrelations with HMS in the Wensum were statistically sig-nificant post p-value correction likely driving the significantrelationship with all catchments combined for this variable(Fig 4) This suggests that modification is greater upstreamin the Wensum contrary to ideas that downstream reachesmay show greater anthropogenic modification Intense agri-cultural land use in the upper reaches of the Wensum islikely to be the cause of the high HMS scores upstreamThe lack of significant correlations suggests that stream or-der and therefore an upstream-to-downstream gradient arenot the predominant pattern in river characteristics despitethe description of such a gradient by geomorphic (Schumm1977) and ecological frameworks (Vannote et al 1980)This is surprising as distance and elevation which both re-flect the upstream-to-downstream gradient have proven tobe important factors in previous studies explaining patternsof RHS features at a national level (Jeffers 1998a Vaughanet al 2013) This implies that upstream-to-downstream gra-dient may not sufficiently reflect patterns of river charac-teristics through the river network within individual catch-ments Others have also found that stream order has weakand inconsistent relationships with biodiversity patterns inriver systems arguing that the topological measure has no di-rect mechanistic control on biodiversity (Vander Vorste et al2017) Instead this study finds that the network density met-rics are a more powerful metric which conceptually providean improved method of accounting for the impacts of net-work topology on the fluvial system exhibiting relationshipswith river characteristics particularly habitat quality score(Fig 4)

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2316 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

44 Applicability of network topology metrics todifferent environments

Much of the seminal work on network and confluence im-pacts (eg the Network Dynamics Hypothesis Benda et al2004b and Link Discontinuity Concept Rice et al 2001) isconducted in natural highly erosive catchments with first-hand empirical measurements However in an age whenrivers and their catchments are increasingly altered by an-thropogenic modification (Meybeck 2003) contemporarystudies must not only aim to expand knowledge but also findmethods of transferring knowledge to many increasingly al-tered catchments (Clifford 2002)

The catchments selected by this study are more greatlymodified and although they reflect a range of fluvial envi-ronments in England are lower-energy catchments than theseminal studies Benda et al (2004a) suggest that confluenceeffects in less active landscapes would be subdued comparedto highly erosive landscapes but the evidence presented heredemonstrates the utility of evaluating network topologicalstructure in studies on catchment-level effects in any typeof fluvial environment including those with low-energy andwidespread anthropogenic changes The response of someriver characteristics varied between catchments observationsof flow-type speed sediment size and modifications showeddifferent responses to network density in different catch-ments This suggests that the functional effect of these topo-logical metrics is catchment dependent and likely is influ-enced by external catchment characteristics such as land usenot considered in this study although impact did not appearto vary with catchment topography or circularity as has beenshown in prior studies (Benda et al 2004a Rice 2017) Thisshould be explored further in future research to enable rec-ommendations to be made regarding where and how reachesmay respond to network density The response of habitatquality score was however consistent across catchments andbetween metrics showing that habitat quality is greater inareas with high network density (Fig 4) as hypothesizedby the Network Dynamics Hypothesis (Benda et al 2004b)and demonstrated by studies on individual catchments (Rice2017 Rice et al 2006)

The methods presented in this paper are designed to be im-plemented in any catchment with a dendritic network struc-ture The topology metrics can easily be calculated from anydendritic network with DTM data using GIS and comparedto any site-scale data This study uses regulatory monitoringdatasets so that analysis is targeted to assessment scores andphysical features of interest to river managers Also the highvolume and wide spatial extent of data available from regu-latory sources allow for between-catchment comparisons

5 Conclusions

Although appreciation of catchment-level effects is consid-ered the epitome of understanding river functioning a keycomponent of the catchment ndash the river network ndash is over-looked and oversimplified by catchment-level studies Thisstudy finds that river network density plays a role in struc-turing the distribution river characteristics throughout thecatchment offering more detailed explanation than the clas-sic stream-order metric Network-dense areas are generallyfound to have higher habitat quality and diversity but modi-fication flow-type speed and sediment size show different re-sponses in different catchments This study suggests two po-tential reasons for this (1) there is evidence that confluencesin the river network increase diversity as is observed in thisstudy so the direction of mean river characteristic responsemay not be consistent and (2) there may be external factorssuch as sediment availability land cover and anthropogenicmodification that alter the direction of mean river characteris-tic response This paper demonstrates the functional responseof river characteristics to network topology and suggests thatthe inclusion of network topology in catchment-level studieswould add a layer of function-based understanding to suchstudies linking reaches to their catchments

The broad-scale methodology adopted by this study allowsthe network density metrics which are easily extracted fromopen-source data using GIS software to be compared to anyregulatory dataset The use of regulatory datasets allows notonly for analysis over a wider spatial extent but also for moreapplicable results for regulatory bodies Therefore the inter-disciplinary approach to characterizing network topology canbe applied efficiently and effectively to capture catchment-level impacts on reach-level functioning in any catchmentacross the globe

Data availability River Habitat Survey data are freely availablefrom the Environment Agency England River network data andelevation data are available from the Centre for Ecology and Hy-drology UK

Author contributions EH and NC conceptualised and designed thestudy EH performed the data analysis with the support of JMEH wrote the paper with comments provided by NC and JM

Competing interests The authors declare that they have no conflictof interest

Acknowledgements The authors would like to thank the Environ-ment Agency and the Centre for Ecology and Hydrology for accessto the data used in this paper This research is funded by the NaturalEnvironmental Research Council (NERC)

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2317

Review statement This paper was edited by Nandita Basu and re-viewed by Simone Bizzi and one anonymous referee

References

Belletti B Rinaldi M Buijse A D Gurnell A Mand Mosselman E A review of assessment methods forriver hydromorphology Environ Earth Sci 73 2079ndash2100httpsdoiorg101007s12665-014-3558-1 2015

Benda L Andras K Miller D and Bigelow P Confluenceeffects in rivers Interactions of basin scale network geom-etry and disturbance regimes Water Resour Res 40 1ndash15httpsdoiorg1010292003WR002583 2004a

Benda L Poff N L Miller D Dunne T Reeves GPess G and Pollock M The network dynamics hy-pothesis how channel networks structure riverine habi-tats Bioscience 54 413ndash427 httpsdoiorg1016410006-3568(2004)054[0413TNDHHC]20CO2 2004b

Benjamini Y and Hochberg Y Controlling the False DiscoveryRate A Practical and Powerful Approach to Multiple TestingJ Roy Stat Soc B 57 289ndash300 1995

Best J L Flow dynamics and sediment transport at river channelconfluences Birbeck University of London London 1985

Best J L The morphology of river channel con-fluences Prog Phys Geogr 10 157ndash174httpsdoiorg101177030913338601000201 1986

Best J L Flow dynamics at river channel confluences implica-tions for sediment transport and bed morphology in Recent de-velopments in fluvial sedimentology edited by Ethridge F GFlores R M and Harvey M D Spec Publ Soc Econ Pale-ontol Miner Tulsa Okla 27ndash35 1987

Biron P M Richer A Kirkbride A D Roy A G andHan S Spatial patterns of water surface topography ata river confluence Earth Surf Proc Land 27 913ndash928httpsdoiorg101002esp359 2002

Bravard J P and Gilvear D J Hydrological and geomorpho-logical structure of hydrosystems in The Fluvial Hydrosystemedited by Petts G E and Amoros C Springer Netherlands98ndash116 1996

Brierley G and Fryirs K River Styles a Geomorphic Approachto Catchment Characterization Implications for River Rehabili-tation in Bega Catchment New South Wales Australia EnvironManage 25 661ndash679 2000

Bunn S E and Arthington A H Basic principles and ecologi-cal consequences of altered flow regimes for aquatic biodiversityEnviron Manage 30 492ndash507 httpsdoiorg101007s00267-002-2737-0 2002

Church M and Kellerhals R On the statistics of grain size varia-tion along a gravel river Can J Earth Sci 15 1151ndash1160 1978

Clifford N J Hydrology the changing paradigm Prog Phys Ge-ogr 26 290ndash301 2002

Cohen P Andriamahefa H and Wasson J-G Towardsa regionalizaton of aquatic habitat distribution of meso-habitats at the scale of a large basin Regul RiversRes Manage 14 391ndash404 httpsdoiorg101002(SICI)1099-1646(19980910)145lt391AID-RRR513gt30CO2-W 1998

Davenport A J Gurnell A M and Armitage P D Habitat sur-vey and classification of urban rivers River Res Appl 20 687ndash704 httpsdoiorg101002rra785 2004

Davies N M Norris R H and Thoms M C Prediction and as-sessment of local stream habitat features using large-scale catch-ment characteristics Freshwater Biol 45 343ndash369 2000

Depettris C Mendiondo E M Neiff J and Rohrmann H Flooddefence strategy at the confluence of the Paranaacute-Paraguay riversProc Int Symp Flood Def Kassel He C31 C40 2000

Dollar E James C Rogers K and Thoms M A framework forinterdisciplinary understanding of rivers as ecosystems Geomor-phology 89 147ndash162 2007

Dovers S R and Day D G Australian rivers and statute law En-viron Plan Law J 5 90ndash108 1988

Downs P W Gregory K J and Brookes A How integratedis river basin management Environ Manage 15 299ndash309httpsdoiorg101007BF02393876 1991

Emery J C Gurnell A M Clifford N J and Petts G E Char-acteristics and controls of gravel-bed riffles An analysis of datafrom the river-habitat survey Water Environ J 18 210ndash216httpsdoiorg101111j1747-65932004tb00535x 2004

Evans I S and Minaacuter J A classification of geomorphometric vari-ables in International Geom-orphometry 2011 Geomoprhome-tryorg Redlabds CA 105ndash108 2011

Fausch K D Torgersen C E Baxter C V Li H WView C The O F Is R To N How U Among ISet S For C and Fishes S Landscapes to riverscapesBridging the gap between research and conservation of streamfishes Bioscience 52 483ndash498 httpsdoiorg1016410006-3568(2002)052[0483LTRBTG]20CO2 2002

Glickman M E Rao S R and Schultz M R False discov-ery rate control is a recommended alternative to Bonferroni-typeadjustments in health studies J Clin Epidemiol 67 850ndash857httpsdoiorg101016jjclinepi201403012 2014

Gupta V K and Mesa O J Runoff generation and hydro-logic response via channel network geomorphology - Re-cent progress and open problems J Hydrol 102 3ndash28httpsdoiorg1010160022-1694(88)90089-3 1988

Gupta V K and Waymire E D On the formulation of an analyt-ical approach to hydrologic response and similarity at the basinscale J Hydrol 65 95ndash123 1983

Gupta V K Waymire E D and Rodriguez-Iturbe I On scalesgravity and network structure in basin runoff in Scale problemsin hydrology Springer Netherlands 159ndash184 1986

Gurnell A M Rinaldi M Belletti B Bizzi S Blamauer BBraca G Buijse A D Bussettini M Camenen B ComitiF Demarchi L Garciacutea de Jaloacuten D Gonzaacutelez del TaacutenagoM Grabowski R C Gunn I D M Habersack H Hen-driks D Henshaw A J Kloumlsch M Lastoria B LatapieA Marcinkowski P Martiacutenez-Fernaacutendez V Mosselman EMountford J O Nardi L Okruszko T OrsquoHare M T PalmaM Percopo C Surian N van de Bund W Weissteiner Cand Ziliani L A multi-scale hierarchical framework for devel-oping understanding of river behaviour to support river manage-ment Aquat Sci 78 1ndash16 httpsdoiorg101007s00027-015-0424-5 2016

Harvey G L Gurnell A M and Clifford N J Characterisationof river reaches The influence of rock type Catena 76 78ndash88httpsdoiorg101016jcatena200809010 2008

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2316 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

44 Applicability of network topology metrics todifferent environments

Much of the seminal work on network and confluence im-pacts (eg the Network Dynamics Hypothesis Benda et al2004b and Link Discontinuity Concept Rice et al 2001) isconducted in natural highly erosive catchments with first-hand empirical measurements However in an age whenrivers and their catchments are increasingly altered by an-thropogenic modification (Meybeck 2003) contemporarystudies must not only aim to expand knowledge but also findmethods of transferring knowledge to many increasingly al-tered catchments (Clifford 2002)

The catchments selected by this study are more greatlymodified and although they reflect a range of fluvial envi-ronments in England are lower-energy catchments than theseminal studies Benda et al (2004a) suggest that confluenceeffects in less active landscapes would be subdued comparedto highly erosive landscapes but the evidence presented heredemonstrates the utility of evaluating network topologicalstructure in studies on catchment-level effects in any typeof fluvial environment including those with low-energy andwidespread anthropogenic changes The response of someriver characteristics varied between catchments observationsof flow-type speed sediment size and modifications showeddifferent responses to network density in different catch-ments This suggests that the functional effect of these topo-logical metrics is catchment dependent and likely is influ-enced by external catchment characteristics such as land usenot considered in this study although impact did not appearto vary with catchment topography or circularity as has beenshown in prior studies (Benda et al 2004a Rice 2017) Thisshould be explored further in future research to enable rec-ommendations to be made regarding where and how reachesmay respond to network density The response of habitatquality score was however consistent across catchments andbetween metrics showing that habitat quality is greater inareas with high network density (Fig 4) as hypothesizedby the Network Dynamics Hypothesis (Benda et al 2004b)and demonstrated by studies on individual catchments (Rice2017 Rice et al 2006)

The methods presented in this paper are designed to be im-plemented in any catchment with a dendritic network struc-ture The topology metrics can easily be calculated from anydendritic network with DTM data using GIS and comparedto any site-scale data This study uses regulatory monitoringdatasets so that analysis is targeted to assessment scores andphysical features of interest to river managers Also the highvolume and wide spatial extent of data available from regu-latory sources allow for between-catchment comparisons

5 Conclusions

Although appreciation of catchment-level effects is consid-ered the epitome of understanding river functioning a keycomponent of the catchment ndash the river network ndash is over-looked and oversimplified by catchment-level studies Thisstudy finds that river network density plays a role in struc-turing the distribution river characteristics throughout thecatchment offering more detailed explanation than the clas-sic stream-order metric Network-dense areas are generallyfound to have higher habitat quality and diversity but modi-fication flow-type speed and sediment size show different re-sponses in different catchments This study suggests two po-tential reasons for this (1) there is evidence that confluencesin the river network increase diversity as is observed in thisstudy so the direction of mean river characteristic responsemay not be consistent and (2) there may be external factorssuch as sediment availability land cover and anthropogenicmodification that alter the direction of mean river characteris-tic response This paper demonstrates the functional responseof river characteristics to network topology and suggests thatthe inclusion of network topology in catchment-level studieswould add a layer of function-based understanding to suchstudies linking reaches to their catchments

The broad-scale methodology adopted by this study allowsthe network density metrics which are easily extracted fromopen-source data using GIS software to be compared to anyregulatory dataset The use of regulatory datasets allows notonly for analysis over a wider spatial extent but also for moreapplicable results for regulatory bodies Therefore the inter-disciplinary approach to characterizing network topology canbe applied efficiently and effectively to capture catchment-level impacts on reach-level functioning in any catchmentacross the globe

Data availability River Habitat Survey data are freely availablefrom the Environment Agency England River network data andelevation data are available from the Centre for Ecology and Hy-drology UK

Author contributions EH and NC conceptualised and designed thestudy EH performed the data analysis with the support of JMEH wrote the paper with comments provided by NC and JM

Competing interests The authors declare that they have no conflictof interest

Acknowledgements The authors would like to thank the Environ-ment Agency and the Centre for Ecology and Hydrology for accessto the data used in this paper This research is funded by the NaturalEnvironmental Research Council (NERC)

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2317

Review statement This paper was edited by Nandita Basu and re-viewed by Simone Bizzi and one anonymous referee

References

Belletti B Rinaldi M Buijse A D Gurnell A Mand Mosselman E A review of assessment methods forriver hydromorphology Environ Earth Sci 73 2079ndash2100httpsdoiorg101007s12665-014-3558-1 2015

Benda L Andras K Miller D and Bigelow P Confluenceeffects in rivers Interactions of basin scale network geom-etry and disturbance regimes Water Resour Res 40 1ndash15httpsdoiorg1010292003WR002583 2004a

Benda L Poff N L Miller D Dunne T Reeves GPess G and Pollock M The network dynamics hy-pothesis how channel networks structure riverine habi-tats Bioscience 54 413ndash427 httpsdoiorg1016410006-3568(2004)054[0413TNDHHC]20CO2 2004b

Benjamini Y and Hochberg Y Controlling the False DiscoveryRate A Practical and Powerful Approach to Multiple TestingJ Roy Stat Soc B 57 289ndash300 1995

Best J L Flow dynamics and sediment transport at river channelconfluences Birbeck University of London London 1985

Best J L The morphology of river channel con-fluences Prog Phys Geogr 10 157ndash174httpsdoiorg101177030913338601000201 1986

Best J L Flow dynamics at river channel confluences implica-tions for sediment transport and bed morphology in Recent de-velopments in fluvial sedimentology edited by Ethridge F GFlores R M and Harvey M D Spec Publ Soc Econ Pale-ontol Miner Tulsa Okla 27ndash35 1987

Biron P M Richer A Kirkbride A D Roy A G andHan S Spatial patterns of water surface topography ata river confluence Earth Surf Proc Land 27 913ndash928httpsdoiorg101002esp359 2002

Bravard J P and Gilvear D J Hydrological and geomorpho-logical structure of hydrosystems in The Fluvial Hydrosystemedited by Petts G E and Amoros C Springer Netherlands98ndash116 1996

Brierley G and Fryirs K River Styles a Geomorphic Approachto Catchment Characterization Implications for River Rehabili-tation in Bega Catchment New South Wales Australia EnvironManage 25 661ndash679 2000

Bunn S E and Arthington A H Basic principles and ecologi-cal consequences of altered flow regimes for aquatic biodiversityEnviron Manage 30 492ndash507 httpsdoiorg101007s00267-002-2737-0 2002

Church M and Kellerhals R On the statistics of grain size varia-tion along a gravel river Can J Earth Sci 15 1151ndash1160 1978

Clifford N J Hydrology the changing paradigm Prog Phys Ge-ogr 26 290ndash301 2002

Cohen P Andriamahefa H and Wasson J-G Towardsa regionalizaton of aquatic habitat distribution of meso-habitats at the scale of a large basin Regul RiversRes Manage 14 391ndash404 httpsdoiorg101002(SICI)1099-1646(19980910)145lt391AID-RRR513gt30CO2-W 1998

Davenport A J Gurnell A M and Armitage P D Habitat sur-vey and classification of urban rivers River Res Appl 20 687ndash704 httpsdoiorg101002rra785 2004

Davies N M Norris R H and Thoms M C Prediction and as-sessment of local stream habitat features using large-scale catch-ment characteristics Freshwater Biol 45 343ndash369 2000

Depettris C Mendiondo E M Neiff J and Rohrmann H Flooddefence strategy at the confluence of the Paranaacute-Paraguay riversProc Int Symp Flood Def Kassel He C31 C40 2000

Dollar E James C Rogers K and Thoms M A framework forinterdisciplinary understanding of rivers as ecosystems Geomor-phology 89 147ndash162 2007

Dovers S R and Day D G Australian rivers and statute law En-viron Plan Law J 5 90ndash108 1988

Downs P W Gregory K J and Brookes A How integratedis river basin management Environ Manage 15 299ndash309httpsdoiorg101007BF02393876 1991

Emery J C Gurnell A M Clifford N J and Petts G E Char-acteristics and controls of gravel-bed riffles An analysis of datafrom the river-habitat survey Water Environ J 18 210ndash216httpsdoiorg101111j1747-65932004tb00535x 2004

Evans I S and Minaacuter J A classification of geomorphometric vari-ables in International Geom-orphometry 2011 Geomoprhome-tryorg Redlabds CA 105ndash108 2011

Fausch K D Torgersen C E Baxter C V Li H WView C The O F Is R To N How U Among ISet S For C and Fishes S Landscapes to riverscapesBridging the gap between research and conservation of streamfishes Bioscience 52 483ndash498 httpsdoiorg1016410006-3568(2002)052[0483LTRBTG]20CO2 2002

Glickman M E Rao S R and Schultz M R False discov-ery rate control is a recommended alternative to Bonferroni-typeadjustments in health studies J Clin Epidemiol 67 850ndash857httpsdoiorg101016jjclinepi201403012 2014

Gupta V K and Mesa O J Runoff generation and hydro-logic response via channel network geomorphology - Re-cent progress and open problems J Hydrol 102 3ndash28httpsdoiorg1010160022-1694(88)90089-3 1988

Gupta V K and Waymire E D On the formulation of an analyt-ical approach to hydrologic response and similarity at the basinscale J Hydrol 65 95ndash123 1983

Gupta V K Waymire E D and Rodriguez-Iturbe I On scalesgravity and network structure in basin runoff in Scale problemsin hydrology Springer Netherlands 159ndash184 1986

Gurnell A M Rinaldi M Belletti B Bizzi S Blamauer BBraca G Buijse A D Bussettini M Camenen B ComitiF Demarchi L Garciacutea de Jaloacuten D Gonzaacutelez del TaacutenagoM Grabowski R C Gunn I D M Habersack H Hen-driks D Henshaw A J Kloumlsch M Lastoria B LatapieA Marcinkowski P Martiacutenez-Fernaacutendez V Mosselman EMountford J O Nardi L Okruszko T OrsquoHare M T PalmaM Percopo C Surian N van de Bund W Weissteiner Cand Ziliani L A multi-scale hierarchical framework for devel-oping understanding of river behaviour to support river manage-ment Aquat Sci 78 1ndash16 httpsdoiorg101007s00027-015-0424-5 2016

Harvey G L Gurnell A M and Clifford N J Characterisationof river reaches The influence of rock type Catena 76 78ndash88httpsdoiorg101016jcatena200809010 2008

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2317

Review statement This paper was edited by Nandita Basu and re-viewed by Simone Bizzi and one anonymous referee

References

Belletti B Rinaldi M Buijse A D Gurnell A Mand Mosselman E A review of assessment methods forriver hydromorphology Environ Earth Sci 73 2079ndash2100httpsdoiorg101007s12665-014-3558-1 2015

Benda L Andras K Miller D and Bigelow P Confluenceeffects in rivers Interactions of basin scale network geom-etry and disturbance regimes Water Resour Res 40 1ndash15httpsdoiorg1010292003WR002583 2004a

Benda L Poff N L Miller D Dunne T Reeves GPess G and Pollock M The network dynamics hy-pothesis how channel networks structure riverine habi-tats Bioscience 54 413ndash427 httpsdoiorg1016410006-3568(2004)054[0413TNDHHC]20CO2 2004b

Benjamini Y and Hochberg Y Controlling the False DiscoveryRate A Practical and Powerful Approach to Multiple TestingJ Roy Stat Soc B 57 289ndash300 1995

Best J L Flow dynamics and sediment transport at river channelconfluences Birbeck University of London London 1985

Best J L The morphology of river channel con-fluences Prog Phys Geogr 10 157ndash174httpsdoiorg101177030913338601000201 1986

Best J L Flow dynamics at river channel confluences implica-tions for sediment transport and bed morphology in Recent de-velopments in fluvial sedimentology edited by Ethridge F GFlores R M and Harvey M D Spec Publ Soc Econ Pale-ontol Miner Tulsa Okla 27ndash35 1987

Biron P M Richer A Kirkbride A D Roy A G andHan S Spatial patterns of water surface topography ata river confluence Earth Surf Proc Land 27 913ndash928httpsdoiorg101002esp359 2002

Bravard J P and Gilvear D J Hydrological and geomorpho-logical structure of hydrosystems in The Fluvial Hydrosystemedited by Petts G E and Amoros C Springer Netherlands98ndash116 1996

Brierley G and Fryirs K River Styles a Geomorphic Approachto Catchment Characterization Implications for River Rehabili-tation in Bega Catchment New South Wales Australia EnvironManage 25 661ndash679 2000

Bunn S E and Arthington A H Basic principles and ecologi-cal consequences of altered flow regimes for aquatic biodiversityEnviron Manage 30 492ndash507 httpsdoiorg101007s00267-002-2737-0 2002

Church M and Kellerhals R On the statistics of grain size varia-tion along a gravel river Can J Earth Sci 15 1151ndash1160 1978

Clifford N J Hydrology the changing paradigm Prog Phys Ge-ogr 26 290ndash301 2002

Cohen P Andriamahefa H and Wasson J-G Towardsa regionalizaton of aquatic habitat distribution of meso-habitats at the scale of a large basin Regul RiversRes Manage 14 391ndash404 httpsdoiorg101002(SICI)1099-1646(19980910)145lt391AID-RRR513gt30CO2-W 1998

Davenport A J Gurnell A M and Armitage P D Habitat sur-vey and classification of urban rivers River Res Appl 20 687ndash704 httpsdoiorg101002rra785 2004

Davies N M Norris R H and Thoms M C Prediction and as-sessment of local stream habitat features using large-scale catch-ment characteristics Freshwater Biol 45 343ndash369 2000

Depettris C Mendiondo E M Neiff J and Rohrmann H Flooddefence strategy at the confluence of the Paranaacute-Paraguay riversProc Int Symp Flood Def Kassel He C31 C40 2000

Dollar E James C Rogers K and Thoms M A framework forinterdisciplinary understanding of rivers as ecosystems Geomor-phology 89 147ndash162 2007

Dovers S R and Day D G Australian rivers and statute law En-viron Plan Law J 5 90ndash108 1988

Downs P W Gregory K J and Brookes A How integratedis river basin management Environ Manage 15 299ndash309httpsdoiorg101007BF02393876 1991

Emery J C Gurnell A M Clifford N J and Petts G E Char-acteristics and controls of gravel-bed riffles An analysis of datafrom the river-habitat survey Water Environ J 18 210ndash216httpsdoiorg101111j1747-65932004tb00535x 2004

Evans I S and Minaacuter J A classification of geomorphometric vari-ables in International Geom-orphometry 2011 Geomoprhome-tryorg Redlabds CA 105ndash108 2011

Fausch K D Torgersen C E Baxter C V Li H WView C The O F Is R To N How U Among ISet S For C and Fishes S Landscapes to riverscapesBridging the gap between research and conservation of streamfishes Bioscience 52 483ndash498 httpsdoiorg1016410006-3568(2002)052[0483LTRBTG]20CO2 2002

Glickman M E Rao S R and Schultz M R False discov-ery rate control is a recommended alternative to Bonferroni-typeadjustments in health studies J Clin Epidemiol 67 850ndash857httpsdoiorg101016jjclinepi201403012 2014

Gupta V K and Mesa O J Runoff generation and hydro-logic response via channel network geomorphology - Re-cent progress and open problems J Hydrol 102 3ndash28httpsdoiorg1010160022-1694(88)90089-3 1988

Gupta V K and Waymire E D On the formulation of an analyt-ical approach to hydrologic response and similarity at the basinscale J Hydrol 65 95ndash123 1983

Gupta V K Waymire E D and Rodriguez-Iturbe I On scalesgravity and network structure in basin runoff in Scale problemsin hydrology Springer Netherlands 159ndash184 1986

Gurnell A M Rinaldi M Belletti B Bizzi S Blamauer BBraca G Buijse A D Bussettini M Camenen B ComitiF Demarchi L Garciacutea de Jaloacuten D Gonzaacutelez del TaacutenagoM Grabowski R C Gunn I D M Habersack H Hen-driks D Henshaw A J Kloumlsch M Lastoria B LatapieA Marcinkowski P Martiacutenez-Fernaacutendez V Mosselman EMountford J O Nardi L Okruszko T OrsquoHare M T PalmaM Percopo C Surian N van de Bund W Weissteiner Cand Ziliani L A multi-scale hierarchical framework for devel-oping understanding of river behaviour to support river manage-ment Aquat Sci 78 1ndash16 httpsdoiorg101007s00027-015-0424-5 2016

Harvey G L Gurnell A M and Clifford N J Characterisationof river reaches The influence of rock type Catena 76 78ndash88httpsdoiorg101016jcatena200809010 2008

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019

2318 E L Heasley et al Integrating network topology metrics into studies of catchment-level effects

Helsel B D R and Hirsch R M Chapter A3 Statistical Meth-ods in Water Resources B 4 Hydrol Anal Interpret TechWater-Resources Investig United States Geol Surv RestonVA 2002

Hornby D D RivEX 67 ed available at httpwwwrivexcouk(last access 23 January 2018) 2010

Jeffers J N R Characterization of river habitats and predictionof habitat features using ordination techniques Aquat ConservMar Freshw Ecosyst 8 529ndash540 1998a

Jeffers J N R The statistical basis of samplingstrategies for rivers An example using River Habi-tat Survey Aquat Conserv Mar Freshw Ecosyst8 447ndash454 httpsdoiorg101002(SICI)1099-0755(19980708)84lt447AID-AQC288gt30CO2-R 1998b

Jones N E and Schmidt B J Tributary effects in rivers interac-tions of spatial scale network structure and landscape character-istics Can J Fish Aquat Sci 74 503ndash510 2016

Jusik S Szoszkiewicz K Kupiec J M Lewin I and Samecka-Cymerman A Development of comprehensive river typologybased on macrophytes in the mountain-lowland gradient of dif-ferent Central European ecoregions Hydrobiologia 745 241ndash262 httpsdoiorg101007s10750-014-2111-2 2015

Kiffney P M Greene C M Hall J E and Davies J RTributary streams create spatial discontinuities in habitat bio-logical productivity and diversity in mainstem rivers Can JFish Aquat Sci 63 2518ndash2530 httpsdoiorg101139f06-138 2006

Kirkby M Tests of the random network model and its ap-plication to basin hydrology Earth Surf Proc 1 197ndash212httpsdoiorg101002esp3290010302 1976

Knighton A D Longitudinal changes in size and sort-ing of stream-bed material in four English rivers GeolSoc Am Bull 91 55ndash62 httpsdoiorg1011300016-7606(1980)91lt55LCISASgt20CO2 1980

Lashermes B and Foufoula-Georgiou E Area andwidth functions of river networks New results onmultifractal properties Water Resour Res 43 1ndash19httpsdoiorg1010292006WR005329 2007

Macklin M G and Lewin J River sediments great floods andcentennial-scale Holocene climate change J Quaternay Sci 18101ndash105 httpsdoiorg101002jqs751 2003

McGonigle D F Burke S P Collins A L Gartner R HaftM R Harris R C Haygarth P M Hedges M C HiscockK M and Lovett A A Developing Demonstration Test Catch-ments as a platform for transdisciplinary land management re-search in England and Wales Environ Sci Process Imp 161618ndash1628 httpsdoiorg101039c3em00658a 2014

Meybeck M Global analysis of river systems fromEarth system controls to Anthropocene syndromesPhilos T Roy Soc Lond B 358 1935ndash1955httpsdoiorg101098rstb20031379 2003

Milesi S V and Melo A S Conditional effects of aquatic in-sects of small tributaries on mainstream assemblages positionwithin drainage network matters Can J Fish Aquat Sci 711ndash9 2013

Moore R V Morris D G and Flavin R W Sub-set of UK digi-tal 1 50000 scale river centreline network NERC Inst HydrolWallingford 1994

Morris D G and Flavin R W Sub-set of UK 50 m by 50 m hydro-logical digital terrain model grids NERC Inst Hydrol Walling-ford 1994

Naura M Clark M J Sear D A Atkinson P M Hornby D DKemp P England J Peirson G Bromley C and Carter MG Mapping habitat indices across river networks using spatialstatistical modelling of River Habitat Survey data Ecol Indic66 20ndash29 httpsdoiorg101016jecolind201601019 2016

Newson M D Land water and development sustainable andadaptive management of rivers 3rd Edn Taylor and FrancisAbingdon 2009

Newson M D Understanding ldquohot-spotrdquo problems in catch-ments The need for scale-sensitive measures and mechanismsto secure effective solutions for river management and con-servation Aquat Conserv Mar Freshw Ecosyst 20 62ndash72httpsdoiorg101002aqc1091 2010

Osborne L L and Wiley M J Influence of tributary spatial po-sition on the structure of warmwater fish communities Can JFish Aquat Sci 49 671ndash681 httpsdoiorg101139f92-0761992

Owens P N Batalla R J Collins A J Gomez B Hicks DM Horowitz A J Kondolf G M Marden M Page M JPeacock D H Petticrew E L Salomons W and Trustrum NA Fine-grained sediment in river systems environmental signif-icance and management issues River Res Appl 21 693ndash717httpsdoiorg101002rra878 2005

Perry J A and Schaeffer D J The longitudinal distribution ofriverine benthos a river discontinuum Hydrobiologia 148257ndash268 1987

Peterson E E Ver Hoef J M Isaak D J Falke J A FortinM J Jordan C E McNyset K Monestiez P Ruesch A SSengupta A Som N Steel E A Theobald D M TorgersenC E and Wenger S J Modelling dendritic ecological networksin space An integrated network perspective Ecol Lett 16 707ndash719 httpsdoiorg101111ele12084 2013

Petts G E and Amoros C The Fluvial Hydrosystems SpringerDordrecht 1996

Raven P J Fox P Everard M Holmes N T H and Dawson FH River Habitat Survey a new system for classifying rivers ac-cording to their habitat quality in Freshwater Quality Definingthe Indefinable edited by Boon P J and Howell D L TheStationery Office Edinburgh 215ndash234 1996

Raven P J Holmes N T H Dawson F H andEverard M Quality assessment using River Habi-tat Survey data Aquat Conserv Mar FreshwEcosyst 8 477ndash499 httpsdoiorg101002(SICI)1099-0755(19980708)84lt477AID-AQC299gt30CO2-K 1998

Rice S P Which tributaries disrupt downstream finingalong gravel-bed rivers Geomorphology 22 39ndash56httpsdoiorg101016S0169-555X(97)00052-4 1998

Rice S P Tributary connectivity confluence aggradationand network biodiversity Geomorphology 277 6ndash16httpsdoiorg101016jgeomorph201603027 2017

Rice S P Greenwood M T and Joyce C B Tributaries sed-iment sources and the longitudinal organisation of macroinver-tebrate fauna along river systems Can J Fish Aquat Sci 58824ndash840 httpsdoiorg101139cjfas-58-4-824 2001

Rice S P Ferguson R I and Hoey T B Tributary con-trol of physical heterogeneity and biological diversity at

Hydrol Earth Syst Sci 23 2305ndash2319 2019 wwwhydrol-earth-syst-scinet2323052019

E L Heasley et al Integrating network topology metrics into studies of catchment-level effects 2319

river confluences Can J Fish Aquat Sci 63 2553ndash2566httpsdoiorg101139f06-145 2006

Richards C Johnson L B and Host G E Landscape-scale in-fluences on stream habitats and biota Can J Fish Aquat Sci53 295ndash311 httpsdoiorg101139f96-006 1996

Richards C Haro R Johnson L and Host G Catch-ment and reach-scale properties as indicators of macroin-vertebrate species traits Freshwater Biol 37 219ndash230httpsdoiorg101046j1365-24271997d01-540x 1997

Rodriguez-Iturbe I and Valdes J B The Geomorphologic Struc-ture of Hydrologic Response Water Resour Res 15 1409ndash1420 1979

Rowntree K M and Wadeson R A Translating channel morphol-ogy into hydraulic habitat application of the hydraulic biotopeconcept to an assessment of discharge related habitat changes inProceedings of the 2nd International Association for HydraulicResearch International Symposium on Hydraulics and HabitatsA281ndashA292 1996

Schindfessel L Creeumllle S and De Mulder T Flow Patterns in anOpen Channel Confluence with Increasingly Dominant TributaryInflow Water 7 4724ndash4751 httpsdoiorg103390w70947242015

Schumm S A The Fluvial System John Wiley amp Sons New York1977

Singer M B Downstream patterns of bed materialgrain size in a large lowland alluvial river subjectto low sediment supply Water Resour Res 44 1ndash7httpsdoiorg1010292008WR007183 2008

Steel E A Sowder C and Peterson E E Spatial and Tempo-ral Variation of Water Temperature Regimes on the SnoqualmieRiver Network J Am Water Resour Assoc 52 769ndash787httpsdoiorg1011111752-168812423 2016

Stepinski T F and Stepinski A P Morphology of drainagebasins as an indicator of climate on early Mars J GeophysRes-Planets 110 1ndash10 httpsdoiorg1010292005JE0024482005

Strahler A Quantitative analysis of watershed geomorphologyTrans Am Geophys Union 38 913ndash920 1957

Vannote R Minshall G Cummins K Sedell J and CushingC The River Continuum Concept Can J Fish Aquat Sci 37130ndash137 1980

Vaughan I P and Ormerod S J Linking ecological andhydromorphological data Approaches challenges and futureprospects for riverine science Aquat Conserv Mar FreshwEcosyst 20 125ndash130 2010

Vaughan I P Merrix-Jones F L and Constantine J A Success-ful predictions of river characteristics across England and Walesbased on ordination Geomorphology 194 121ndash131 2013

Vander Vorste R McElmurray P Bell S Eliason K M andBrown B L Does stream size really explain biodiversity pat-terns in lotic systems A call for mechanistic explanations Di-versity 9 1ndash21 httpsdoiorg103390d9030026 2017

Ver Hoef J M and Peterson E E A Moving Average Approachfor Spatial Statistical Models of Stream Networks J Am StatAssoc 105 6ndash18 httpsdoiorg101198jasa2009ap082482010

wwwhydrol-earth-syst-scinet2323052019 Hydrol Earth Syst Sci 23 2305ndash2319 2019