UC Davis UC Davis Previously Published Works

Title The role of topographic variability in river channel classification

Permalink https://escholarship.org/uc/item/9pg1x6j9

Journal Progress in Physical Geography, 41(5)

ISSN 0309-1333

Authors Lane, BA Pasternack, GB Dahlke, HE et al.

Publication Date 2017-10-01

DOI 10.1177/0309133317718133

Peer reviewed

eScholarship.org Powered by the California Digital Library University of California Article

Progress in Physical Geography 1–31 ª The Author(s) 2017 The role of topographic Reprints and permission: sagepub.co.uk/journalsPermissions.nav variability in river channel DOI: 10.1177/0309133317718133 classification journals.sagepub.com/home/ppg

Belize A. Lane Department of Land, Air and Water Resources, University of California at Davis, Davis, CA, USA Department of Civil and Environmental Engineering, Utah State University, Logan, UT, USA Gregory B. Pasternack Department of Land, Air and Water Resources, University of California at Davis, Davis, CA, USA Helen E. Dahlke Department of Land, Air and Water Resources, University of California at Davis, Davis, CA, USA Samuel Sandoval-Solis Department of Land, Air and Water Resources, University of California at Davis, Davis, CA, USA

Abstract To date, subreach-scale variations in flow width and bed elevation have rarely been included in channel classifications. Variability in topographic features of rivers, however, in conjunction with sediment supply and discharge produces a mosaic of channel forms that provides unique habitats for sensitive aquatic species. In this study we investigated the utility of topographic variability attributes (TVAs) in distinguishing channel types and dominant channel formation and maintenance processes in montane and lowland streams of the Sacramento River basin, California, USA. A stratified random survey of 161 stream sites was performed to ensure balanced sampling across groups of stream reaches with expected similar geomorphic settings. For each site surveyed, width and depth variability were measured at baseflow and bankfull stages, and then incorporated in a channel classification framework alongside traditional reach-averaged geomorphic attributes (e.g., channel slope, width-to-depth, confinement, and dominant substrate) to evaluate the significance of TVAs in differentiating channel types. In contrast to more traditional attributes such as slope and contributing area, which are often touted as the key indicators of hydrogeomorphic processes, bankfull width variance emerged as a first-order attribute for distinguishing channel types. A total of nine channel types were distinguished for the Sacramento Basin consisting of both previously identified and new channel types. The results indicate that incorporating TVAs in channel classification provides a quantitative basis for interpreting nonuniform as well as uniform geomorphic processes, which can improve our ability to distinguish linked channel forms and processes of geomorphic and ecological significance.

Keywords Channel classification, river topography, nonuniform process, channel form

I Introduction Corresponding author: Belize Lane, Department of Civil and Environmental Building on the classic premise of Davis (1909), Engineering, Utah State University, Logan, UT, USA. Thornbury (1954) stated that geomorphic Email: [email protected] 2 Progress in Physical Geography XX(X) processes create a characteristic assemblage of Considerable recent efforts have been invested landforms. Through judicious use of inverse in developing geomorphic attributes for river reasoning, investigation of landforms can pro- characterization, particularly in Europe through vide an understanding of linked geomorphic the implementation of the Water Framework processes. Over the past century, studies have Directive (e.g., Orr et al., 2008; Polvi et al., shown that ecological structure and function of 2014; Raven et al., 1998; Sear et al., 2009). rivers are strongly influenced by channel type Common attributes considered include uniform (e.g., Hack and Goodlett, 1960; Smith et al., metrics such as reach-averaged channel slope, 1995; Vannote et al., 1980). As a result of these width-to-depth ratio, entrenchment ratio, valley strong foundations, channel classification has confinement, sinuosity, stream power, and come to the forefront of river science and man- dominant channel substrate (Brierley and agement as a central feature of methods for Fryirs, 2005; Church, 1992; Kasprak et al., understanding, protecting, and restoring rivers 2016; Knighton, 1999; Montgomery and in North America (Buffington and Montgomery, Buffington, 1997; Rosgen, 1994). 2013; Kondolf, 1995; Rosgen, 1994), Europe However, nonuniform mechanisms not well (e.g., Gonza´lez del Tańago and Garcıá de Jaloń, characterized or indicated by reach-averaged 2004; Orr et al., 2008), Australia (Brierley and uniform metrics have been identified as primary Fryirs, 2005), and South Africa (Rowntree and drivers of channel formation and maintenance Wadeson, 1998). Channel classification is of crit- in many channel settings (Dietrich and Smith, ical importance today for river management, 1983; Lane and Carlson, 1953; Makaske, 2001; because anthropogenic changes to flow regimes Paustian et al., 1992; Powell et al. 2005; (Magilligan and Nislow, 2005; Molles et al., Thompson, 1986; White et al., 2010; Wilcox 1998), sediment regimes (Graf, 1980; Pitlick and and Wohl, 2006; Wohl and Thompson, 2000). Van Steeter, 1998; Wohl et al., 2015), and the For example, subreach-scale flow convergence physical structure of rivers (Price et al., 2012) routing has been shown to control riffle-pool have led to widespread degradation of river eco- formation and maintenance and the locations systems worldwide (Arthington, 2012; Dynesius of sediment deposition and bar instability and Nilsson, 1994). (MacWilliams et al., 2006). In meandering and Reach-scale geomorphic settings (e.g., pool- alternate bar morphologies, nonuniformity is riffle, step-pool (Montgomery and Buffington, maintained primarily by the alternating conver- 1997)) distinguished by attributes related to ging and diverging secondary transverse flow channel form and sediment transport and supply cells in and between bends, respectively, which have been shown to influence ecosystem help to maintain sediment routing through the dynamics and biological diversity (Biggs inside of meander bends (Thompson, 1986). et al., 2005; Meitzen et al., 2013; Milner et al., Topographic variability attributes (TVAs), 2015; Montgomery and Bolton, 2003), high- defined here as any measure of subreach-scale lighting channel reach classification as a critical variability (i.e., departures from average condi- step in river ecosystem management. Geo- tions in channel bed elevation, bankfull width, morphic attributes used in channel classification curvature, and floodplain width), are closely are often chosen to describe relevant, persistent tied to nonuniform channel processes and likely reach-scale characteristics that influence offer more appropriate metrics for characteriz- hydraulics and sediment dynamics and in turn ing and comparing dominant channel processes aquatic and riparian ecosystem functioning and habitat dynamics than their far more (Birkeland, 1996; Hupp and Osterkamp, 1996; common uniform counterparts used in many Kasprak et al., 2016; Merritt and Wohl, 2003). channel morphologies. For example, measures Lane et al. 3 of subreach-scale channel width and depth driven scour, and meander bend cut-off. One variance are expected to capture the frequency might conjecture that variability is indicated by and magnitude distribution of flow expansions reach-scale homogenous metrics like specific and contractions associated with flow conver- stream power, and thus not needed to define gence routing under a dynamic flow regime channel classes, but if the processes that control (MacWilliams et al., 2006). Furthermore, high channel form are governed by variability, then within-reach topographic variability is often the reverse should be taken as the dominant con- associated with heterogeneous habitat units jecture: reach-scale homogenous metrics are the available across a wider range of discharges that outcome of the interplay between channel varia- can support a variety of native biota and ecolo- bility and flow, not the controls on it. gical functions (Murray et al., 2006; Scown In spite of the established geomorphic et al., 2016), promoting high biodiversity (Brown et al., 2014, 2015; Gostner et al. (Fausch et al., 2002; Poff and Ward, 1990; 2013a, 2013b; MacWilliams et al., 2006; Townsend and Hildrew, 1994) and ecological Thompson, 1986; White et al., 2010) and eco- resilience (Elmqvist et al., 2003; McCluney logical (Elmqvist et al., 2003; McCluney et al., et al., 2014). 2014; Murray et al., 2006; Scown et al., 2016) Channel topographic variability exists natu- significance of subreach-scale topographic rally and is part of a dynamic equilibrium with variability, very few existing channel classifica- other channel variables. At the valley scale, tions consider TVAs. While the Rosgen (1994) there are nested layers of topographic variabil- and Montgomery and Buffington (1997) classi- ity, including variations in the width of hillsides, fications both consider the spacing of individual terraces and floodplains along a corridor (e.g., channel-unit types along a reach (e.g., non- Gangodagamage et al., 2007; White et al., dimensional pool spacing measured in channel 2010). When a flow of a set magnitude moves widths) in their suite of geomorphic attributes, through a layered topographic boundary, it no direct measure of channel width or depth engages one or more of these controls and a variability is included. The limited consider- specific scale of topographic steering is initi- ation of TVAs in past channel classifications ated. That specific type of steering then drives may be due to the preference by practitioners subreach variability in the hydraulic flow field to conduct rapid field surveys (sometimes at that focuses erosion and deposition locally only one cross-section per reach) in order to (Strom et al., 2016). For a dynamic flow regime, maximize the number of channel reaches sur- topographic steering changes with flow and this veyed in lieu of performing more in-depth sur- results in a diversity of stage-dependent hydrau- veys across fewer reaches (Buffington and lic patch behaviors (Scown et al., 2016; Strom Montgomery, 2013) given resource limitations. et al., 2016), each with a different capability to With the emergence of meter-scale remote sen- promote erosion or deposition (Brown and Pas- sing of rivers, datasets that support computing ternack, 2014; Grams et al., 2013). and analyzing TVAs will become more avail- As a result of these factors, rivers exhibit able, accurate, and useful (Gleason and Wang, complex patterns of topographic change pro- 2015; Gonzalez and Pasternack, 2015). There cesses that promote strong longitudinal varia- has already been significant progress on the use tion in width and depth (Wyrick and of high resolution aerial imagery from drones to Pasternack, 2015). Variability itself is expected map river characteristics (e.g., Lejot et al., 2007; to differ between reaches, because many geo- Rivas Casado et al., 2015, 2016). morphic processes control aspects of variability, A few exceptions include the studies by such as flow convergence, avulsion, turbulence- Trainor and Church (2003) and Jaeger (2015). 4 Progress in Physical Geography XX(X)

Trainor and Church (2003) included channel used channel classification system in North depth and width variability as key geomorphic America and globally (Kasprak et al., 2016), attributes in a channel comparison study, but the was adopted and expanded on in this study to focus on quantifying dissimilarity between facilitate ease of application of the proposed channel reach pairs precluded an evaluation of methods in future channel classifications. The the relative significance of individual attributes Rosgen channel classification is a stream-reach for distinguishing channel types. Jaeger (2015) taxonomy that classifies channel types using considered the standard deviation of channel field-collected geomorphic attributes (e.g., bed elevation (a measure of depth variability) slope, entrenchment ratio, width-to-depth ratio, in their classification of headwater streams. sinuosity, and median grain size). In an effort to However, the set-up of the study as an analysis support the incorporation of TVAs into field- of the geomorphic significance of mountaintop based mapping for channel classification given mining again precluded any evaluation of attri- the common constraint of resource limitations, bute significance. This major gap in the channel the Rosgen channel classification procedure classification literature indicates a need to test was extended in three ways: (1) the channel the value of incorporating TVAs into the suite of network was binned into hydrogeomorphically potentially significant geomorphic attributes similar groups prior to field data collection distinguishing ecologically relevant channel using a stratified analysis of hydrologic and types. This must be done before we can even topographic data in a Geographic Information begin to evaluate the geomorphic or ecological System (GIS); (2) four TVAs consisting of significance of these emerging attributes com- within-reach low flow and bankfull width and pared to the more traditional reach-averaged depth variance were measured in the field in attributes described above. addition to the traditional geomorphic attributes The purpose of this study was to investigate considered by Rosgen (1994); and (3) a heuris- how TVAs can be incorporated in a channel tic refinement procedure was used to distinguish classification framework to improve the utility the most parsimonious set of physically inter- of morphological analysis to distinguish domi- pretable channel types instead of associating the nant channel processes and habitat dynamics field-observed channel types with known along channel networks in varied landscapes. Rosgen classes. The specific study objectives were to test the use of TVAs in (i) distinguishing channel types across a landscape and (ii) characterizing domi- 1. Study area nant channel processes of interest. The utility The study was conducted in the Sacramento and ecological implications of incorporating Basin of California, USA, encompassing the TVAs in a channel classification of montane largest river in the State of California by dis- and lowland streams of a Mediterranean basin charge (producing * 30% of California’s sur- are then discussed and evaluated in the context face water runoff) and the second largest U.S. of the existing body of channel classification river draining into the Pacific Ocean (after the literature and current understanding of land- Columbia River) (Carter and Resh, 2005). This scape form – process linkages. 70,000-km2 basin lies between the Sierra Nevada and Cascade Range to the east and the Coast Range and Klamath Mountains to the II Methodology west. From its headwaters in the volcanic pla- The Rosgen channel classification (Level II, teau of northern California (Upper Sacramento, Rosgen, 1994), arguably the most commonly McCloud, and Pit Rivers), the Sacramento Lane et al. 5

River flows south for 715 km before reaching the Sacramento–San Joaquin River Delta and San Francisco Bay. The river has many small to moderate-sized tributaries (e.g., Clear, Cottonwood, Cow, Battle, Antelope, Mill, Deer, Stony, Big Chico, and Butte Creek) and two large tributaries, the Feather River and the American River. The basin primarily exhibits a Mediterranean climate with cold, wet winters (Oct–Apr) and warm, dry summers (May–Sep) (Leung et al., 2003). The basin’s diverse physiographic settings range from the glacially-carved Sierra Nevada Mountains to lowland marshes and agricultural lands, with a total relief of about 4300 m (US Geological Survey, 2011). The Sacramento Figure 1. Sacramento Basin physiographic prov- Basin is split into three overlying physiographic inces used to refine contributing area (Ac) based sediment composition thresholds for channel net- provinces: the Pacific Border, the Cascade- work stratification. Sierra Mountains, and the Basin and Range provinces (Fenneman and Johnson, 1946) California’s legacy of intensive and wide- (Figure 1). These provinces exhibit distinct spread hydrologic and geomorphic alteration landscape units (sensu Brierley and Fryirs, for water supply, flood control, land use change, 2005) based on differential tectonic uplift, hydropower, and mining has left the Sacra- lithology, and climate (California Geological mento Basin’s river ecosystems severely Survey, 2002) and are therefore expected to degraded (Hanak et al., 2011; Healey et al., account for major differences in geomorphic 2008). The basin simultaneously supports 2.8 processes and resulting channel morphologies million people and numerous federally endan- (Schmitt et al., 2007; Trainor and Church, gered and threatened aquatic species (e.g., 2003). For instance, the Basin and Range prov- winter-run Chinook salmon (Oncorhynchus ince consists primarily of a thick accumulation tschawytscha), Sacramento splittail (Pogo- of lava flows and tuff beds, supporting low slope nichthys macrolepidotus)) (Lindley et al., meandering streams and large marshlands with 2007; Moyle et al., 2011). Most of the Sacra- low sediment transport capacity. The Cascade- mento Basin valley is intensively cultivated, Sierra Mountains province consists of a massive with over 8100 km2 of irrigated agriculture. tilted fault block; the western slope descends in Major reservoirs in the basin include Lake a series of undulating low-relief upland surfaces Shasta (5.6 km3, upper Sacramento, McCloud punctuated by deeply incised river canyons, and Pit Rivers), Lake Oroville (4.4 km3, Feather driving high sediment transport rates (Stock River), Lake Folsom (1.2 km3, American et al., 2005). The Pacific Border province River), and New Bullards Bar Reservoir (1.2 delineates an alluvial basin that acts as a deposi- km3, Yuba River). In light of systemic anthro- tional trough (California Geological Survey, pogenic alteration promoting channel homoge- 2002). Relationships between contributing area nization and simplification (Arnold et al., 1982; and channel bed composition are expected to Booth and Jackson, 1997; Walsh et al., 2005), vary significantly between these provinces one might expect that topographic variability based on major differences in sediment regimes. would be suppressed. Therefore, if TVAs prove 6 Progress in Physical Geography XX(X) important here in the characterization of in- 2000) and hydrology (Thomas and Lewis, 1995; channel habitat dynamics, then they are likely Yang and Woo, 1999). Three landscape charac- even more important in undisturbed settings in teristics accounting for geologic structure, sedi- which topographic variability is expected to be ment availability, and sediment transport greater and thus influence habitat dynamics capacity were obtained from GIS data and anal- across a larger range of TVAs. yses as described below and used to stratify the This study was constrained to one hydrologic Sacramento Basin channel network into 15 sub- regime found within the Sacramento Basin to help groups or strata of potentially distinct reach- isolate factors that cause diverse hydrological and scale geomorphic characteristics. geomorphic effects. An existing regional hydrolo- Geologic structure (i.e., tectonic uplift and gic classification of California (Lane et al., 2017) lithology), derived from the overlying physio- was used to identify stream reaches exhibiting the graphic provinces (California Geological Sur- low-volume snowmelt and rain (LSR) regime. The vey, 2002; Fenneman and Johnson, 1946) LSR hydrologic regime was chosen as it captures (Figure 1), was used in conjunction with sedi- the transition from the montane snowmelt-driven ment availability and transport capacity to dis- to lowland rain-driven flow regime and has the tinguish 15 geomorphic strata. Sediment supply largest spatial footprint of hydrologic regimes in and transport capacity were represented using the Sacramento Basin (47%); stream reaches in contributing area to a reach (Ac) and the channel this hydrologic regime are expected to exhibit high bed slope of a reach (S). These were obtained geomorphic variability. through analysis of the National Hydrography Dataset (HUC 1802) (US Geological Survey, 2013) in conjunction with a 10-m digital 2. Channel network stratification elevation model (DEM) of the study area (US Given the large study domain with about Geological Survey, 2009). Ac isacommon 100,000 reaches and limited resources, the pro- topographically-derived surrogate for channel- cess of observing representative sites requires forming discharge (e.g., Hack, 1957; Rosgen, selecting a relatively small number of samples 1994; Schumm et al., 1984) and S is consistently compared to the scope of the system. If sites used in classifications to characterize local flow were selected at random, then the odds are energy dissipation (e.g., Gartner et al., 2015; that different geomorphic settings would be Montgomery and Buffington, 1997; Rosgen, observed in proportion to their frequency of 1994). The combination of the two variables is occurrence, and that would bias the assessment also prominent in hydrogeomorphic classifica- of classification, especially if too few sites of tion, as it is often conjectured that channel bed rare yet important classes were sampled. There- morphology arises as a function of reach-scale fore, instead of random sampling, a stratified shear stress and/or specific stream power, which random approach was used to obtain an equal are determined by both unit discharge and channel effort strategy mindful of process-based con- slope (Flores et al., 2006). Indices combining trols on river organization. Stratified random Ac and S as a measure of stream power (Lane, sampling and related variants using equal effort 1957; Leopold and Wolman, 1957; Sklar and Die- in each stratum have not been widely applied in trich, 1998) have been used to distinguish braided channel classification studies to date to capture from meandering rivers (Carson, 1984), to iden- reach-scale geomorphic heterogeneity, but are tify thresholds for channel incision (Schumm well known in field ecology (Columbia Habitat et al., 1984) and sediment transport capacity Monitoring Program, 2016; Johnson, 1980; (Bledsoe et al., 2002), and in reach-scale channel Manly and Alberto, 2014; Miller and Ambrose, classification (e.g., Schmitt et al., 2007). Lane et al. 7

The channel network was derived from the Table 1. Contributing area (Ac) thresholds for channel 10-m DEM and dissected into equidistant seg- composition across Sacramento Basin physiographic provinces (see Figure 1 for map of physiographic ments of 250 m length; S and Ac were subse- quently derived from the DEM for each provinces). segment. Within each physiographic province, Contributing area threshold channel segments were binned according to (km2) GIS-derived S and Ac thresholds to aid with sampling – the results of the study are not sen- Physiographic Bedrock/boulder cobble/gravel province to cobble/gravel to sand/silt sitive to the exact number of bins or thresholds between bins, as long as the procedure aids with Pacific Border 50 5,000 sampling the diversity in the system with equal Cascade-Sierra 300 9,000 effort. Five S bins were considered based on Mountains Rosgen’s (1994) channel classification thresh- Basin and Range 300 10,000 olds for ease of comparison: <0.1%, 0.1–2%,2– 4%,4–10%,and>10%.ThreeAc bins were Of the 15 geomorphic strata distinguished established based on estimated A threshold c across the Sacramento Basin by Ac and S com- transitions for prevalent sediment sizes: (1) bed- binations, 13 strata were exhibited by LSR rock/boulder, (2) cobble/gravel, and (3) sand/ reaches, indicating that LSR-dominated hydro- silt. The Ac thresholds assigned to distinguish logic regimes were 87% representative of the channel bed composition classes were unique full range of geomorphic variability in the for each of the three physiographic provinces Sacramento Basin as expressed by the geo- within the Sacramento Basin. This decision was morphic strata. The two geomorphic strata not based on the expected differences in Ac required found within LSR reaches consisted of the com- to transition from boulder- to cobble- and from binations of the highest Ac bin and 4–10% or gravel- to sand- dominated channels arising >10% slope bins. Based on reach accessibility from large-scale differences in geology, topo- and expected variability of geomorphic attri- graphy, and climate driving distinct sediment butes, 10 to 12 field surveys were performed regimes. The physiographic provinces provide within each of the 13 geomorphic strata exhib- bounds on what channels are potentially com- ited by LSR reaches for a total of 161 field parable in terms of relations between drainage survey reaches representing a large range of area and discharge, sediment supply, and sub- Ac–S combinations (Figure 3). Note that strate size (Montgomery and Buffington, 1993). DEM-derived S was not used further in this Within each province, Ac bin thresholds were study, as it is not highly accurate at representing estimated based on identified channel composi- reach-scale channel slope. tion transition locations reported in available literature combined with expert knowledge 3. Data-driven geomorphic channel relating Ac and sediment composition in the region (e.g., Gasparini et al., 2004; Montgomery classification and Buffington, 1993) (Table 1). Fifteen geo- Field surveys. Geomorphic field surveys were morphic strata were then distinguished as all pos- performed for each study reach identified sible combinations of topographically-derived through the stratified random sampling scheme Ac and S bins (Figure 2, top-left), and each described above. Surveys of 64 reaches were stream segment in the channel network was conducted by the authors’ crew and data from assigned to a stratum based on its particular another 97 reaches were obtained from the GIS-based Ac and S values (Figure 2a). Surface Water Ambient Monitoring Program 8 Progress in Physical Geography XX(X)

Figure 2. Map of geomorphic strata (a) across the Sacramento Basin and (b) across the low-volume snowmelt and rain (LSR) reaches of the Sacramento Basin. Yellow dots indicate the randomly chosen field survey locations across the 15 strata. The geomorphic strata are defined in the top-left table based on the combination of contributing area (Ac) and slope (S) bins, which are derived based on thresholds stated in the bottom-left table and Table 1.

Depending on whether the average wetted channel width was less than or greater than 10 m, a stream reach was surveyed over a length of 150 or 250 m, respectively (Ode, 2007), corresponding to 10 – 100 bankfull widths. Eleven evenly spaced cross-sectional transects were surveyed along each stream reach to quantify variability in 22 geomorphic attributes listed in Table 2 (Ode, 2007). These decisions were intended to balance geomorphic (Grant et al., 1990; Mon- tgomery and Buffington, 1997) and ecological Figure 3. The stratified random field survey loca- (Frissell et al., 1986) relevance with the practi- tions (n ¼ 161) represent a large range of GIS-based cal time and resource limitations of field sur- reach slopes (S) and contributing areas (Ac). Colors veying. The choice of reach length and and shading indicate the distinct S and Ac bins that transect spacing also enabled incorporation of correspond to the geomorphic strata listed in Figure the existing SWAMP geomorphic dataset for 2 based on the Cascade-Sierra Mountains physio- the study region that uses the same values. graphic province Ac thresholds in Table 1. Channel morphology and reach characteristics for the 161 surveyed reaches were measured (SWAMP) of the California State Water using a surveying level and stadia rod (Topcon Resources Control Board. Both field campaigns AT-B, 0.01m). Longitudinal streambed profiles used the same sampling protocols, outlined in were surveyed at consecutive transects along Ode (2007) and briefly summarized below. the thalweg for the entire length of the reach. Lane et al. 9

Table 2. Reach-scale geomorphic and topographic variability attributes considered in channel classification. Geomorphic Attribute Code Description Units wetted depth d average across 11 transects; 0 if dry channel m wetted width w average across 11 transects; 0 if dry channel m wetted width-to-depth w:d ratio of channel width to depth – wetted depth-to-D50 d:D50 low water roughness; channel depth standardized by median – grain size bankfull depth dBF average across 11 transects m bankfull width wBF average across 11 transects m bankfull width-to-depth w:dBF ratio of bankfull width to depth – bankfull depth-to-D50 dBF:D50 roughness; bankfull depth standardized by median grain size – entrenchment ratio e:ratio floodprone width / average bankfull width; floodprone width – manually estimated from high resolution aerial imagery (<1m) (Rosgen, 1994) shear stress shear depth–slope product approximation Pa shields stress shields non-dimensionalization of shear stress (Shields, 1936) – contributing area Ac drainage area to downstream end of reach km2 slope slope average water surface slope over 11 transects % sinuosity sin straightline distance/actual channel distance along *2000 m – of channel sediment distribution CVsed variance of transect sediment distribution (n¼10) across 11 – variance transects D50 D50 median grain size across reach (n¼110) mm D84 D84 84th percentile grain size across reach (n¼110) mm Dmax Dmax maximum grain size across reach (n ¼ 110) mm y wetted depth variance CVd std/mean across 11 transects; 0 if no water in channel – y wetted width variance CVw std/mean across 11 transects; 0 if no water in channel – y bankfull depth CVd.BF std/mean across 11 transects – variance y bankfull width CVw.BF std/mean across 11 transects – variance y Topographic variability attributes (TVAs)

Wolman pebble counts (Wolman, 1954) of geomorphic characteristics that influence 110 pebbles were performed at each reach hydraulics and sediment dynamics and in turn such that 10 pebbles were randomly selected aquatic and riparian ecosystem functioning from each of 11 transects to balance sam- (Birkeland, 1996; Hupp and Osterkamp, 1996; pling precision and effort across a range of Merritt and Wohl, 2003). The field-measured sediment material variability assuming nor- and computed attributes included traditional mally distributed sediment size (Bunte and Abt, reach-averaged diagnostic variables (e.g., 2001; Edwards and Glysson, 1999). slope (slope), contributing area (Ac), sinuosity (sin), entrenchment (e:ratioÞ, shear stress Reach-scale geomorphic attributes. Twenty-two (shearÞ, relative roughness (d:D50), sediment geomorphic attributes (Table 2) were chosen composition (i.e., D50, D84, and Dmax) and base to describe relevant, persistent reach-scale flow and bankfull depth (d), width (w), and 10 Progress in Physical Geography XX(X)

width-to-depth ratio (w:dBF)) as well as four 0 to 1 and examined for correlation to identify TVAs capturing within-reach variability in base and remove highly correlated attributes (Pearson’s flow and bankfull channel width (CVw) and bed correlation coefficient > 0.8) to meet the assump- elevation (CVd)(Table2). tion of lack of multicollinearity. Five of the origi- Reach-scale estimates of geomorphic attri- nal 22 attributes were highly correlated (d, w, butes were computed from field surveys by d:D50, D50, CVsed), reducing the dataset to 17 geo- averaging values across the eleven surveyed morphic attributes (Table 2). cross-sections within each reach. Entrench- A hierarchical clustering analysis using ment was calculated as flood-prone width Ward’s algorithm (Murtagh and Legendre, divided by bankfull width (Rosgen, 1994), 2013; Ward, 1963) was used to examine the where flood-prone width was measured manu- clustering structure of the uncorrelated, stan- ally from sub-meter resolution aerial imagery. dardized geomorphic attributes describing the Sinuosity was calculated as the linear valley 161 study reaches. The dataset also was ana- distance divided by the actual channel distance lyzed by k-means cluster analysis stipulating 2 along 2 km of channel straddling the field site to 15 (k) clusters that maximize the between- (Elliott et al., 2009). The coefficient of varia- group variation (Hartigan and Wong, 1979; tion (CV) of base flow and bankfull width and Kaufman and Rousseeuw, 1990). Slope breaks depth was calculated among the eleven cross- in the k-means scree plot of the within-group sections of each survey reach as a measure sum of squares for each clustering solution of within-reach variability. CV is a non- were interpreted as numbers of clusters at dimensional measure of standard deviation that which information content of the clustering provides a useful but not exclusive metric of process changed. Scree plot slope breaks and variability (Schneider, 1994) that is commonly the Davies–Bouldin internal clustering index used in spatial analysis of ecological patterns (DBI ¼ 0.91) indicated that 12 clusters created (Gostner et al., 2013a; Gubala et al., 1996; distinct groups of study reaches, similar to the Palmer et al., 1997; Rossi et al., 1992; Simon- hierarchical clustering results. sonetal.,1994;Thoms,2006).Alistofgeo- A combination of univariate and multivariate morphic attributes considered and their statistical methods was then applied to (i) exam- methods of measurement or calculation is ine the strength of variables for distinguishing provided in Table 2. When possible, these attri- identified channel types, (ii) test the hypothesis butes were made non-dimensional for applica- that channel types exhibit significantly different tion in a range of physiographic and climatic values of geomorphic attributes, (iii) examine settings (Parker, 1979; Parker et al., 2003). the potential range of values for variables of Given the dual aims of adapting the Rosgen interest between channel types, and (iv) validate classification to incorporate TVAs and com- the basis of the channel classification by predict- parability with existing field data for the study ing the channel type using geomorphic attributes. region, the present study omitted several These statistical methods included nonmetric potentially significant metrics (e.g., channel multidimensional scaling (NMDS) (Clarke, vegetation, bank material, dominant flow 1993), one-way analysis of variance (ANOVA) types (Raven et al., 1998), and stream power with Tukey’s honestly significant differences (Knighton, 1999; Orr et al., 2008)) that could (HSD) test, nonparametric permutational multi- be considered in future studies. variate analysis of variance (PerMANOVA) (Anderson, 2001), and classification and regres- Statistical analyses. The geomorphic attributes sion trees (CART) (Breiman et al., 1984; De’ath (Table 2) were initially re-scaled to range from and Fabricius, 2000). Lane et al. 11

An exploratory NMDS analysis (Clarke, physical understanding of the region. First, 1993; Oksanen, 2011) of the surveyed reaches potential splitting solutions were identified based on the uncorrelated geomorphic attributes based on the structure of the hierarchical clus- was performed to visually represent the struc- tering and the shape of the scree-plots from the ture of the multivariate dataset and evaluate the non-hierarchical k-means clustering. Each relative significance and correlation of attri- potential splitting solution was assessed itera- butes. NMDS is common in ecological studies, tively from largest to smallest splitting distance including those identifying differences in (based on Ward’s hierarchical clustering). biological communities based on geomorphic Heuristic (dis)aggregation of clusters was sub- variables (e.g., Virtanen et al., 2010; Walters sequently performed based on the physical dis- et al., 2003;) and is increasingly included in tinction and interpretability of the resulting dedicated geomorphic studies (e.g., Jaeger, clusters with the objective of minimizing the 2015; Merriam et al., 2011; Sutfin et al., 2014; final number of physically interpretable channel Varanka et al., 2014;). Histograms of each geo- types. For instance, if a particular splitting solu- morphic attribute were also used to evaluate the tion distinguished only some empirical clusters density distributions of attribute values across to a level of reasonable physical interpretability, the survey reaches and lend insight into the mul- the remaining clusters would be iteratively dis- tivariate clustering structure. aggregated based on the next potential splitting Individual one-way ANOVAs were con- solutions until the minimal number of physi- ducted to compare geomorphic attribute means cally meaningful clusters was identified. between channel types. A post-hoc Tukey’s HSD test at the 95% confidence level indicated the III Results best attributes for distinguishing between channel types. A PerMANOVA analysis (Anderson, 1. Relative significance of geomorphic 2001) (Euclidean distance, 9999 permutations attributes (Oksanen, 2011)) was performed to test the The two-dimensional NMDS ordination illu- hypothesis that the channel types distinguished strated the significance of TVAs and the relative through clustering analysis exhibit significant roles of geomorphic attributes in structuring the differences (p < 0.01) in geomorphic attributes. multivariate dataset. The NMDS minimized Toward the primary goal of the study, CART mean stress at 0.08 for 161 study reaches (Breiman et al., 1984) was then used to identify (Figure 4); stress values of <0.1 are considered the most explanatory geomorphic attributes dis- to be a good ordination with little risk of draw- tinguishing channel types and their threshold ing false inferences (McCune and Grace, values. CART yields a binary decision tree 2002). NMDS indicated that the first axis where the response variable (study reach) is par- (NMDS1) is dominated by CVd.BF, CVw.BF, titioned into groups (channel types) with mini- slope,andAc, while the second axis (NMDS2) mized within-group variance (based on ten-fold is dominated by cross-sectional geomorphic cross-validation, Therneau et al., 2010) and attributes (e.g., D84, D50, dBF:D50, w:d BF)as increasing purity (based on the Gini index, well as CVw.BF. As these axes represent gradi- De’ath and Fabricus, 2000). ents of maximum variation, dominant attributes on each axis control the structure of the Heuristic refinement of inductive clustering solution. multivariate dataset. The final number of clusters distinguished was Histograms of rescaled geomorphic attri- determined heuristically based on a combina- butes lend insight into how the density distribution of statistical analysis interpretation and tions of geomorphic attribute values control the 12 Progress in Physical Geography XX(X)

multivariate data structure (Figure 5). If an attribute is normally distributed with a predomi- nance of its values within a narrow band of its full range for most study reaches, then that attribute will likely yield a single grouping, so it cannot explain differences between those reaches; it may instead distinguish the few statistical outlier reaches. In contrast, an attribute with a more uniform distribution will tend to produce more, equally weighted groupings and thus be a dominant factor explaining differences among many reaches. Upon visual assessment Figure 4. Nonmetric dimensional scaling (NMDS) of the geomorphic attribute distributions, most for the first two axes with channel types of individual attributes exhibited highly skewed distributions study reaches indicated. Vectors of attributes are towards lower values (e.g., sin, e:ratio,and plotted based on the strength of their correlation to w ). In contrast, the TVAs (CV and the axis (e.g. longer vectors are more strongly cor- BF d.BF CV )andslope exhibited more uniform related to an axis). w.BF

Figure 5. Histograms of geomorphic attributes (re-scaled from 0 to 1) across the 161 study reaches illustrate the distribution of each attribute. In contrast to the highly skewed distributions exhibited by most attributes about a small range of values, the TVAs (CVd.BF and CVw.BF) and slope exhibit more uniform distributions. Lane et al. 13

Figure 6. Hierarchical clustering of study reaches using Ward’s method showing 12 distinct groups (boxed in red) representing nine physically distinct channel types following heuristic refinement. distributions, helping to explain their dominant Table 3. ANOVA results show that mean geo- roles in structuring the multivariate dataset. morphic attribute values differ between the nine channel types. Statistically significant attributes (p < 0.05) are indicated in bold. 2. Distinguishing channel types Geomorphic Agglomerative hierarchical clustering with attribute Mean square F p-value Ward’s linkage (Murtagh and Legendre, 2013; Ac 334.59 106.28 0.00 Ward, 1963) illustrated the clustering structure of dBF:D50 121.09 26.96 0.00 the 161 study reaches across the re-scaled uncor- CVw.BF 0.25 19.90 0.00 related geomorphic attributes (Figure 6). The first slope 37.06 18.63 0.00 split occurs at a distance of 20, distinguishing w:dBF 76.26 15.98 0.00 reaches of high (*0.2–1.7) and low (*0–0.2) CVd.BF 0.24 15.90 0.00 bankfull width variance. Splitting groups at a dis- dBF 59.50 12.20 0.00 tance of eight distinguished 12 groups that were e:ratio 20.43 10.27 0.00 then reduced to nine physically meaningful groups wBF 42.36 8.50 0.00 sin 28.36 5.59 0.02 by applying the heuristic clustering refinement D84 9.86 4.96 0.03 procedures explained in Section II.3. The nine shear 9.28 4.66 0.03 resulting groups represented physically distinct Dmax 17.66 3.43 0.07 channel types containing between 4 and 57 study shields 0.74 0.14 0.71 reaches each (average of 18 reaches). Individual one-way ANOVA results indicated that group means of 12 of 17 geomorphic using Tukey’s HSD post-hoc test at the attributes varied significantly between the nine 95% confidence level indicated particularly channel types (p < 0.05) (all attributes except w, significant channel types for specific attri- d, D50, Dmax, and shields) (Table 3). Multiple butes (Figure 7). For example, w:d BF is sig- comparisons of group means of each attribute nificantly higher for type 2 reaches than all 14 Progress in Physical Geography XX(X)

Figure 7. Box-and-whisker plots and Tukey’s honestly significant differences (HSD) test indicate differences in geomorphic and topographic variability attributes across the nine identified channel types: 1. confined headwater small boulder cascade, 2. partly-confined expansion pool – wide bar, 3. unconfined upland plateau large uniform, 4. confined cascade/step-pool, 5. partly-confined pool-riffle, 6. partly-confined large uniform, 7. unconfined anastomosing plateau small pool-riffle, 8. unconfined large uniform boulder, and 9. unconfined large meandering sand. other channel types. Conversely, CVw.BF differs Multivariate analyses revealed that the data- significantly between channel types 4 and 7 driven channel types identified exhibit signifi- and channel types 6, 8, and 9 while there is cantly different geomorphic settings and identi- no significant difference in the attribute within fied the geomorphic attribute ranges across each those groups. Box-and-whisker plots illustrate channel type in the study basin. PerMANOVA relative differences in geomorphic attributes results indicated that multivariate mean geo- within and across the nine identified channel morphic setting is not equal for all nine channel types (Figure 7). Finally, a map of the spatial types (p ¼ 0.0001; F-statistic ¼ 13), allowing distribution of classified channel types across for the rejection of the null hypothesis that chan- LSR-dominated reaches in the Sacramento nel types were identical. The CART analysis Basin is provided in Figure 8. identified the most explanatory geomorphic Lane et al. 15

Figure 9. CART classification trees considering (a) Figure 8. Map of the spatial distribution of field sites non-dimensional and (b) all geomorphic attributes, in the hydrological regime investigated and their indicating primary attributes and their threshold classified channel types across LSR-dominated values distinguishing channel types. Geomorphic and reaches (light blue lines) of the Sacramento Basin. topographic variability attributes are defined in Table 2 and circled numbers refer to channel types as attributes distinguishing channel types and their defined in Table 4. threshold values, providing potential ranges of attribute values expected for each channel type three primary attributes distinguishing channel (Figure 9). The classification tree model deter- types, emphasizing their persistent significance mined the relative strength of non-dimensional independent of individual field sites. Further- variables to be as follows: CVw.BF, sin, slope, more, CVw.BF emerged as a dominant attribute e:ratio, CVd.BF, w:d BF. This indicates that two above traditional Rosgen (1994) geomorphic of the six explanatory attributes identified by attributes in both models. the model were TVAs (i.e., CVw.BF,CVd.BF), while slope played a lesser role. The non- dimensional classification tree correctly 3. Physical interpretation of channel types classified 85% of survey reaches based on their Physical interpretation of the above statistical reach-averaged geomorphic attribute values analyses (summarized in Table 4) was used in (Figure 9(a)). Alternatively, 93% of reaches combination with expert evaluation and existing could be correctly classified by the classifica- channel classification literature to name the tion tree considering all attributes (Figure 9(b)). nine channel types based on their valley setting When both dimensional and non-dimensional and distinguishing channel attributes (this attributes were considered (n ¼ 17, Table 2), nomenclature is used for the remainder of this D84,Ac, and wBF emerged as additional signif- study): 1. confined headwater small boulder icant attributes for distinguishing channel types. cascade,2.partly-confined expansion pool – Separate classification tree models using only wide bar,3.unconfined upland plateau large the author’s field sites (n ¼ 64) and using both uniform,4.confined cascade/step-pool,5. the author’s and SWAMP field sites (n ¼ 161) partly-confined pool-riffle,6.partly-confined both identified CVw.BF, sin,andslope as the large uniform,7.unconfined anastomosing 16

Table 4. Descriptive name, literature analogs, key channel form characteristics, and physical process interpretation of identified channel types are provided.

Descriptive name Morphological characterization

Channel Literature Channel slope Cross-sectional number Valley setting Channel type analog and planform attributes Bed material TVAs Physical process interpretation

1 Confined Small boulder- Type Ay; Bedrock Very steep, Low w-to-d, highly Poorly sorted Moderate w and High stream power combined with variable headwater cascadez or Cascadez; straight entrenched boulder- d variance topography drive high sediment Steep dominated transport and high subreach-scale headwater* variability in scour and fill (Powell et al., 2005) 2 Partly-confined Pool-wide bar Moderate gradient Variable slope, Wide and shallow, Poorly sorted High w variance, Lateral flow divergence drives rapid expansion alluvial fan high sinuosity entrenched pebble- to moderate d deposition of unsorted alluvial sediment channel cobble-sized variance; þ (Paustian et al., 1992) (Paustian et al., covariance 1992) 3 Unconfined Large uniform Low slope, Large channel Homogenous Low variability Low energy depositional valley; uniform upland plateau straight dimensions, pebble- to topography drives sediment transport low entrenchment cobble-sized as uniform sheet (Miller and Burnett, 2008) 4 Confined Cascade / Cascade or Steep, Low w-to-d, Boulder- Extremely high High topographic variability drives complex Step-pool Step-poolz; moderate entrenched dominated variability;- subreach-scale flow resistance dynamics Type G1-2 or Ay; sinuosity covariance and generates turbulence (Wilcox and Gorge* Wohl 2006; Wohl and Thompson, 2000) 5 Partly-confined Pool-riffle Pool-Rifflez; Mid- to high- Moderate w-to-d, Gravel- to High d variance, Channel constraint by valley and floodplain Type Cy slope, moderate cobble-sized moderate w topographic controls drives localized moderate entrenchment variance; þ vertical and lateral flow convergence; sinuosity covariance convective accelerations reinforce non uniform flow convergence (Dietrich and Smith, 1983) 6 Partly-confined Large uniform Plane-bedz; Mid slope, Moderate w-to-d, Cobble- to Low variability No topographic steering controls on Type By straight moderate boulder-sized deposition or erosion; uniform entrenchment topography drives sediment transport as uniform sheet (Lane and Carlson, 1953; Miller and Burnett, 2008) (continued) Table 4. (continued)

Descriptive name Morphological characterization

Channel Literature Channel slope Cross-sectional number Valley setting Channel type analog and planform attributes Bed material TVAs Physical process interpretation

7 Unconfined Small pool- Type DAy; Type Ey Low slope, Small channel Poorly sorted High variability; Anastomosing channels formed by avulsion anastomosing riffle moderately dimensions, pebble- to þ/- covariance driven by rapid channel aggradation plateau sinuous low entrenchment cobble-sized (Makaske, 2001). High channel depth variance and poorly sorted sediment may be indicative of rapid, heterogeneous channel deposition triggering avulsion 8 Unconfined Large uniform Low slope, Large channel Cobble- to Low variability Underlying bedrock geology constrains boulder moderate dimensions, boulder-sized formation of meander bends and sinuosity entrenched determines sediment supply and composition 9 Unconfined Large Labile channel Low slope, high Very large channel Sand Low variability Meanders maintained by secondary meandering (Church 2006); sinuosity dimensions, highly transverse flow cells that drive sediment sand bed Meandering entrenched routing through inside of bends sand (Lane (Thompson, 1986); “live bed” sediment 1957) transport (Henderson, 1963)

yRosgen (1994). zMontgomery and Buffington (1997). *Brierley and Fryirs (2005). 17 18 Progress in Physical Geography XX(X)

Figure 10. Example images of channel types distinguished by classification from field and Google Earth@ imagery. plateau small pool-riffle,8.unconfined large low sinuosity, and a boulder-dominated bed. uniform boulder, and 9. unconfined large mean- High stream power combined with variable dering sand (Figure 10, Table 4). topography drive high sediment transport and The order of the identified channel types rep- high subreach-scale variability in scour and fill resents an idealized upstream to downstream (Powell et al., 2005) indicated by high CVd.BF. progression in the landscape from montane to The confined cascade/step-pool channel type lowland streams; however, some channel types (4) is distinguished from the boulder – cascade are less predictable along such a progression by slightly lower slopes and larger Ac, as well as (e.g., partly-confined expansion pool – wide slightly increased channel dimensions and a bar, unconfined upland plateau large uniform). reduction in w:d BF and dominant sediment size. Four of the identified channel types (i.e., 2, 3, 6, These changes are indicative of a downstream and 8) were not commonly identified by previ- progression from hillslope- to channel- ous classifications. The geomorphic character- dominated processes. Cascade/step-pool chan- istics of each channel type are described below, nels are also characterized by the highest CVd.BF organized and interpreted with respect to pre- and CVw.BF of any channel type and generally sumed dominant channel processes and related negatively co-varying bed and width undula- to TVAs where applicable. tions, indicating complex subreach-scale flow The confined headwater small boulder-cas- resistance dynamics. Flow resistance in these cade channel type (1) (sensu Hassan et al., 2005; channels is hypothesized to be generated by Montgomery and Buffington, 1997; Sullivan, the form drag of constricting step-forming 1986) is characterized by the highest slopes and roughness features and by tumbling flow lowest Ac of any channel type. These channels regimes in which critical or supercritical flow exhibit high entrenchment, low width-to-depth, over narrow step crests plunges into wider pools, Lane et al. 19 abruptly decreasing velocity and generating energy dissipation dominated by grain and bank substantial turbulence (Montgomery and roughness (Montgomery and Buffington, 1997). Buffington, 1997; Peterson and Mohanty, 1960; Thewell-armoredbedindicatedbythelarge Wilcox and Wohl, 2006; Wohl and Thompson, D50 and D84 suggest relative channel stability 2000; Wyrick and Pasternack, 2008). and a supply limited sediment transport regime The partly-confined pool-riffle channel type (Dietrich et al., 1989). (5) exhibits the next highest slopes and shear Partly-confined expansion pool – wide bar stress and slightly larger Ac than the cascade/ channels (2) generally occur at abrupt valley step-pool channel. Pool-riffle channels are con- widenings and exhibit very high w:d BF and het- strained by valley and floodplain topographic erogeneous sediment composition (CVsed). controls and characterized by positively Alluvial fans develop by the accumulation of co-varying bed and width undulations that gen- sediment where a channel exits an upland drai- erate subreach-scale width and depth constric- nage area (Drew, 1873). These lower-gradient tions and expansions (indicated by high CVw.BF type 2 channels running through alluvial fan and CVd.BF) which drive localized flow conver- style valley expansions likely have limited gence. Topographically-driven convective transport capacity due to reduced stream power accelerations have been shown to reinforce and lateral flow divergence, driving rapid these nonuniform convergent and divergent deposition of unsorted alluvial sediment flow patterns, and thus pool-riffle morphogen- (Paustian et al., 1992). These channels are dis- esis (Dietrich and Smith, 1983; Dietrich and tinguished by pool- wide bar morphology in Whiting, 1989; Nelson and Smith, 1989). The which positively co-varying bed and width pool-riffle channel type is morphologically sim- variability combine with mobile sediment ilar in many regards to the partly-confined large and limited lateral confinement to generate uniform channel type (6) except for signifi- extremely wide, entrenched bars between con- cantly higher topographic variability and stricted troughs. smaller sediment composition. This is inter- The unconfined upland plateau large uni- preted as a difference in sediment transport form channel type (3) exhibits very low mechanisms. In pool-riffle channels, topo- entrenchment due to moderate-sized channels graphic variability has been shown to control bordered by vast floodplains. The laterally sediment transport through mechanisms such unconfined upland plateau valleys through as topographic steering (MacWilliams et al., which these channels run are low-energy (low 2006; Whiting and Dietrich, 1991), flow con- slope and Ac) depositional environments in vergence (MacWilliams et al., 2006; Sawyer which sediment supply is presumed to exceed et al., 2010), and recirculating eddies (Lisle, transport capacity (Nagel et al., 2014). The uni- 1986; Rathburn and Wohl, 2003; Thompson and form topography, low sinuosity, and homoge- Wohl, 2009; Woodsmith and Hassan, 2005). nous sediment composition are indicative of Alternatively, in large uniform channels largely uniform geomorphic processes (e.g., sediment devoid of any organized or rhythmic bedforms, transport as a uniform sheet (Miller and Burnett, at the time of transport the whole bed is 2008)). The unconfined anastomosing plateau expected to move as a conveyor belt (Lane and small pool-riffle channel type (7), also charac- Carlson, 1953; Montgomery and Buffington, terized by low entrenchment and a laterally 1997). As there are no topographic steering con- unconfined valley setting, is distinguished from trols on where deposition or erosion takes place the large uniform channel type by much smaller in large uniform channels, the presumed result channel dimensions and higher topographic is maintenance of uniform width and depth with variability and sinuosity. Similar to partly- 20 Progress in Physical Geography XX(X) confined pool-riffle channels, these channels IV Discussion are expected to maintain nonuniform mor- 1. Lessons learned from channel phology through nonuniform mechanisms such as topographic steering, flow conver- classification modifications gence, and eddy recirculation. At the valley Channel network stratification. The initial GIS- scale, these channels appear to connect to based stratification of the channel network create multi-thread channels that diverge and based on catchment DEM-derived S and Ac converge around vegetated, rarely inundated proved effective at distinguishing underrepre- islands cut from the floodplain (Knighton and sented geomorphic settings in the landscape that Nanson, 1993). The high channel depth varia- would likely otherwise have been overlooked. bility that distinguishes this channel type While some channel types (e.g., pool-riffle, from the upland valley uniform channel may plane-bed, cascade/step-pool) spanned many be indicative of past avulsion triggered by S–Ac bins, indicating their limited dependence rapid, heterogeneous channel deposition on S or Ac, others were almost exclusively (Makaske, 2001). found in one bin (e.g., pool – wide bar, large Finally, unconfined large uniform boulder uniform boulder, large meandering sand). Bins (8) and large meandering sand bed channels with the largest representation across the land- (9) are characterized by very large Ac,large scape unsurprisingly captured the largest num- channel dimensions, low slopes, high sinuosity, ber of channel types. Bins 2, 3, and 4 (Figure 2) and very low width and depth variability. Large represented 28, 16, and 20% of the channel net- uniform boulder bed channels are distinguished work in the study domain and contained 7, 6, by boulder-dominated beds and lower bankfull and 5 channel types, respectively, compared depths, while the large meandering sand bed with 3 channel types per bin on average. channels are sand-dominated and exhibit Geomorphic bins 1–5 with the smallest Ac extremely high sinuosity and entrenchment accounted for 78% of LSR-dominated reaches typical of meandering morphologies (Hickin, in the Sacramento Basin while bins 11–13 with 1974). These differences likely indicate a dif- the largest Ac accounted for less than one per- ference in underlying geology and sediment cent of the study domain combined. However, supply constraining the formation of meanders field sites classified as large uniform boulder by lateral migration and influencing channel and large meandering sand channels fell almost bed composition. The large meandering sand exclusively in bins 11–13, emphasizing the channel type distinguished in this study value of stratified sampling for revealing natu- appears similar to the meandering sand bed rally underrepresented channel types. Slope channel described by Lane (1957) and the bins were more evenly distributed, but very low labile channel distinguished by Church (<0.1%) and very high (>10%) slopes each (2006). Meanders are hypothesized to be main- accounted for less than 10% of the study tained primarily by the alternating converging domain. The identification of low slope domi- and diverging secondary transverse flow cells nated channel types by the classification (e.g., in and between bends, respectively, which help anastomosing, large uniform boulder, and large to maintain sediment routing through the meandering sand) highlights the value of strati- inside of meander bends (Thompson, 1986). fied sampling as these channel types would Mobile bedforms provide the primary hydrau- likely not have been sampled sufficiently to dis- lic resistance in these channels (Kennedy, tinguish distinct classes in a uniform random 1975), driving “live bed” sediment transport sampling scheme given their limited represen- (Henderson, 1963). tation in the basin. Lane et al. 21

The stratified sampling scheme enabled a capturing the unique context of the landscape large proportion of the full range of geomorphic under study (inductive component). variability present in the study domain to be Identified channel types with strong analogs captured by the field sites. For example, bank- in the classification literature highlight the full channel width across all surveyed sites ran- physical basis of the classification results ged from 1.1 to 98.8 m. The smallest and largest achieved after heuristic classification refine- channels evident in the system from visual ment. For example, cascade channels as defined inspection are 0.8 and 100 m, respectively, indi- by Montgomery and Buffington (1997) gener- cating that the sampling scheme captured 98% ally occur on steep slopes, are narrowly con- of the total range of bankfull widths. Similarly, fined by valley walls, and are characterized by the sampling scheme captured 78% of the total longitudinally and laterally disorganized bed range of Ac and 65% of the total range of S. The material typically consisting of cobbles and 2 maximum Ac for a surveyed site was 7760 km boulders. This channel type corresponds while the maximum Ac of any reach in the LSR strongly to our identified confined cascade/ channel network was closer to 10,000 km2. The step-pool channel, characterized by valley- maximum surveyed S of 14.3% was substan- confined channels with steep slopes, low tially less than the estimated 22% maximum width-to-depth, high bankfull width and depth reach S. Overall, these results indicate that, variance, and cobble/boulder dominated sedi- while not entirely representative, stratifying ment. Montgomery and Buffington’s (1993) field data collection by GIS-based landscape plane-bed channel type refers to mid-slope pla- characteristics accounting for geologic struc- nar gravel- and cobble- bed channels generally ture, sediment availability, and sediment trans- lacking discrete bars or in-channel features. port capacity enabled the resulting field sites to This channel type is similar to our partly- capture a large range of geomorphic variabil- confined large uniform channel, characterized ity. Splitting the channel network into further by a moderate slope, cobble-dominated bed, and bins with more refined Ac and S requirements very low bankfull width and depth variance could increase the proportion of the total range (indicating absence of bars and planar longitu- of geomorphic variability captured by field dinal morphology). surveys. Alternatively, stratifying the network Some identified channel types have no ana- across other GIS-based characteristics such as log in the Montgomery and Buffington classifi- bankfull width or adjusting the Ac and S thresh- cation designed for the mountains of the Pacific olds for bin membership could potentially Northwest of the US, particularly those channel improve results. types associated with non-mountain environments. In these cases (e.g., unconfined anasto- Heuristic refinement of classification results. The mosing plateau small pool-riffle), the more nine channel types identified in this study cap- descriptive Rosgen (1994) channel types may ture a diverse range of reach-scale geomorphic provide a better analog (Table 4). Alternatively, settings including channel types previously the large meandering sand bed (9) channel type, identified by existing channel typologies and while not present in the Montgomery and Buf- new, thus far unidentified, channel types. fington (1993) or Rosgen (1994) channel clas- These findings emphasize the value of the a sifications, has been distinguished in numerous posteriori heuristic refinement of inductive other channel classification frameworks (e.g., classification results by suggesting that the Church, 2006; Lane, 1957; Schumm, 1963). resulting channel types retain a physical basis The partly-confined expansion pool – wide bar (deductive component) but are capable of channel type seems to only have an analog in the 22 Progress in Physical Geography XX(X) moderate gradient alluvial fan channel as they were significant attributes in CART. The described by Paustian et al. (1992). This simi- hierarchical clustering structure (Figure 6) and larity of our results with the process-based chan- classification tree (Figure 9) both identified nel types distinguished by Paustian et al. (1992) CVw.BF as the primary splitting variable distin- indicates that the classification framework as guishing channel types for LSR streams of the applied in this study is similarly capable of Sacramento Basin. revealing distinct associations between channel Unlike most geomorphic attributes, which morphology and processes. had overlapping value ranges across all but one Channel types with no clear analog in the channel type (e.g., w:d BF, e:ratio, sin, shear), literature were also identified (e.g., unconfined CVw.BF and CVd.BF exhibited more uniform den- upland plateau large uniform, unconfined large sity distributions (Figure 5) and expressed a uniform boulder), suggesting that the addition continuum of value ranges across all nine chan- of TVAs to the classification framework com- nel types (Figure 7). Thus, TVAs were found to bined with channel network stratification and be very important because they show that some heuristic refinement enabled the resulting chan- rivers have substantial channel bed and width nel classification to reveal the unique context of variability and some do not – it is the variability the landscape under study. For instance, upland in the variability that makes them powerful plateau large uniform channels were distin- classifiers compared to Ac and many other guished from anastomosing plateau small reach-average metrics. For example, the chan- pool-riffle channels primarily on the basis of nel classification distinguished four channel topographic variability. Distinct geomorphic types with very low, one with moderate, and channel formation and maintenance processes four with high topographic variability. Of the and associated ecosystem functions were thus highly variable channel types, two exhibited revealed from otherwise similar channel types primarily positive width and depth covariance, and valley settings based on differences in one exhibited primarily negative covariance, subreach-scale topographic variability. and one exhibited a mixture of both. It may be possible that the significance of TVAs in this study is influenced by the specific 2. Value of topographic variability attributes positioning or frequency of cross-sections along Distinguishing channel types. With respect to the each study reach. Topographic variability is first study objective, TVAs were found to play a often structured with quasi-periodic undula- major role in distinguishing channel types tions, so how sample locations align with those across the landscape. Numerous univariate and structures is very important and probably should multivariate statistical analyses all identified not be left to chance when designing observa- bankfull width and depth variability as first- tion protocols. Future studies with more cross- order predictors of geomorphic channel type. sections per reach or using near-census channel Even though S and Ac – frequently identified width measurements based on high-resolution as dominant variables controlling channel form remote sensing data would reduce the likelihood and geomorphic processes (Church 2002; Die- that the variability being measured is a function trich et al., 1992; Dunne and Leopold, 1978; of the cross-section locations. However, the sta- Leopold and Maddock, 1953; Montgomery and tistically distinct clustering solution and physi- Buffington, 1997) – were used to stratify the cal interpretability of results indicate that the channel network prior to random sampling, they significance of TVAs in the channel classifica- were not identified as the primary attributes dis- tion is fundamentally based on differences in tinguishing geomorphic channel types, though subreach-scale channel forms and processes. Lane et al. 23

Furthermore, study results indicate that the compare dominant channel processes in many history of land use and anthropogenic altera- channel types. For example, differences in tions in the Sacramento Basin are not artifi- TVAs and their covariance as distinguished by cially inflating the importance of TVAs in the the channel classification appeared to be indi- landscape. If any reaches with small degrees of cative of different sediment transport mechan- variability stood out given the simplified isms in partly-confined pool-riffle and large nature (e.g., dredged and straightened) of many uniform channels. Similarly, the high channel parts of the basin, one would expect to see a depth variance distinguishing unconfined highly skewed distribution of TVA values plateau small pool-riffle channels from large towards low variability. However, the uniform uniform channels supported the interpretation distributions exhibited by CVw.BF and CVd.BF of the dominant channel forming process as (Figure 5) negate this hypothesis, indicating avulsion and the dominant channel mainte- instead a large, relatively evenly distributed nance processes as topographic steering, flow range of width and depth variability across the convergence, and eddy recirculation in spite of landscape. very similar valley settings and traditional geomorphic attributes (e.g., slope, w:dBF, e:ratio, Characterizing dominant channel processes. With D84). Alternatively, unconfined large uniform respect to the second study objective, TVAs boulder and meandering sand bed channel were found to be extremely useful for charac- types were differentiated on the basis of under- terizing dominant channel processes that have lying geology rather than TVAs. been reported extensively in the literature but which have been neglected from quantitative Ecological implications. The spatial variability or classification studies prior to this. Most studies lack thereof of channel morphology and associ- only consider processes in terms of reach- ated geomorphic processes as distinguished by average erosive potential, sometimes relative TVAs has important ecological implications. to sediment supply. They have no basis for For example, differences in spatial patterns of describing channel types in terms of the actual hyporheic exchange (Kasahara and Wondzell, specific processes that occur in reaches, such as 2003; Tonina and Buffington, 2009) drive dif- knickpoint migration, bank erosion, and island ferences in local biogeochemistry (Poole et al., formation. By incorporating TVAs in a channel 2008) and habitat dynamics (Geist, 2000). classification framework, we were able to char- Channels with high subreach topographic varia- acterize and distinguish the type and magnitude bility and associated heterogeneous sediment of topographic variability within reaches. In scour and deposition (e.g., our pool-riffle and doing so, this study provided a quantitative cascade/step-pool channels) may exhibit highly basis for interpreting the resultant classes in localized hyporheic exchange (Kasahara and terms of a diversity of mechanisms for fluvial Wondzell, 2003; Poole et al., 2006, 2008), cre- landform formation and maintenance that rely ating local nutrient hotspots associated with on both nonuniform and uniform channel mor- algae or macrophyte growth (Fisher et al., phology (Dietrich and Smith, 1983; Lane and 1998) and preferential spawning habitat (Geist Carlson, 1953; Makaske, 2001; Paustian et al., 2000). In contrast, the uniform flow and sedi- 1992; Powell et al., 2005; Thompson, 1986; ment transport processes exhibited by very low White et al., 2010; Wilcox and Wohl, 2006; topographic variability (e.g., upland valley uni- Wohl and Thompson, 2000). As hypothesized, form channels) are associated with long hypor- TVAs – closely tied to nonuniform processes – heic flow paths that modify the reach’s mean improved the ability to characterize and daily temperature (Poole et al., 2008) and 24 Progress in Physical Geography XX(X) biogeochemistry (Findlay, 1995) from average 3. Future research channel conditions, in turn affecting habitat With the aim of characterizing dominant process quality (Poole et al., 2008; Tonina and Buffing- regimes of distinct channel types as differen- ton, 2009) and salmonid population structure tiated by TVAs, we speculated as to the physical (e.g., Burnett et al., 2003) throughout the reach. processes associated with each identified channel Unconfined uniform channels with the propen- type. We suggest direct measurement of these sity for these long hyporheic flow paths have hypothesized dominant subreach-scale processes also been shown to provide low-velocity refugia and their co-occurrence with distinct TVA for biota during periods of high flow (e.g., Wen- settings as an important direction for future ger et al., 2011) and support wider riparian work. For instance, measurement of hydraulic zones (Polvi et al. 2011). flow fields, hyporheic exchange, or sediment Incorporating TVAs in channel classifica- transport rates across channel types would tion is also expected to inform river restoration bolster physical understanding of the differ- efforts. For example, riparian species richness ences in processes regimes between distinct has been shown to increase with subreach- TVA settings. scale bed elevation variability (Pollock et al., With the emergence of meter-scale remote 1998), suggesting that characterizing TVAs in sensing of rivers, datasets that support comput- addition to more traditional geomorphic attri- ing and analyzing TVAs will become more butes may help predict the impact of distur- available, accurate, and useful (Gleason and bances on the biotic community across the Wang, 2015; Gonzalez and Pasternack, 2015). channel network. Targeting high variability In the meantime, by considering TVAs in addi- channel types (e.g., cascade/step-pool, pool- tion to more traditional channel classification riffle) for riparian restoration efforts may attributes, we hope to encourage future research increase the likelihood of success by increas- into how a stream reach is influenced by its ing the range of hydrogeomorphic and thus surrounding landscape at various scales based ecological responses to disturbance. Alterna- on hierarchical topographic variability relation- tively, channel change associated with channel ships. This could enable the application of unit to reach scale (e.g., 10 – 100 channel increasingly available larger-scale topographic widths) changes in TVAs may indicate datasets to distinguishing differences in multi- changes in flow regimes, sediment regimes, scale process controls on channel morphology or land use (Montgomery and Bolton, 2003), and predicting reach-scale geomorphic settings. indicating critical locations for larger-scale Further understanding of relationships between restoration efforts. For example, the conver- TVAs and multi-scale geomorphic processes is sion of fully forested riparian zones to grass- critical to developing insight into sediment lands has been associated with a significant transport and formative processes in these reduction in within-reach width variability diverse channel types. (Jackson et al., 2015). By identifying channels with rapidly changing CVBF.W, practitioners may more easily define management objec- V Conclusion tives and prioritize restoration activities. This study found that measures of subreach-scale Characteristic TVA values of ecologically topographic variability provided improved functional reaches could provide practitioners information on river geomorphic landforms and with a baseline level of channel and floodplain processes in channel networks of varied land- variability to incorporate into restoration scapes. When incorporated in a channel efforts for degraded reaches. classification framework among a suite of more Lane et al. 25 traditional geomorphic attributes, TVAs Arthington AH (2012) Environmental Flows: Saving improved the ability to distinguish dominant Rivers in the Third Millennium. Berkeley, CA: Uni- channel types and associated geomorphic versity of California Press. processes in LSR-dominated streams of a Med- Biggs BJ, Nikora VI and Snelder TH (2005) Linking scales iterranean region. Bankfull width variance was of flow variability to lotic ecosystem structure and function. River Research and Applications 21(2–3): identified as the primary attribute distinguishing 283–298. channel types over common attributes such as Birkeland GH (1996) Riparian vegetation and sandbar channel slope, width-to-depth ratio, confine- morphology along the lower Little Colorado River, ment, sinuosity, and dominant substrate. The Arizona. Physical Geography 17(6): 534–553. nine channel types distinguished for the Sacra- Bledsoe BP, Watson CC and Biedenharn DS (2002) mento Basin included both channel types with Quantification of incised channel evolution and equi- strong analogs in existing geomorphic literature librium Journal of the American Water Resources and novel channel types. By reenvisioning Association 38(3): 861–870. channel classification through the incorporation Booth DB and Jackson CR (1997) Urbanization of aquatic of TVAs, distinct channel landforms and pro- systems: degradation thresholds stormwater detection cesses were revealed from otherwise similar and the limits of mitigation. Journal of the American geomorphic settings with limited additional Water Resources Association 33(5): 1077–1090. Breiman L, Friedman J, Stone CJ, et al. (1984) Classifi- resource requirements. Results indicate that cation and Regression Trees. Boca Raton, FL: CRC incorporating TVAs in channel classification Press. may improve river restoration efforts by reveal- Brierley GJ and Fryirs KA (2005) Geomorphology and ing ecologically-significant differences in chan- River Management: Applications of the River Styles nel form and function. Framework. Oxford: Blackwell Publications. Brown RA and Pasternack GB (2014) Hydrologic and Declaration of Conflicting Interests topographic variability modulate channel change in The author(s) declared no potential conflicts of inter- mountain rivers Journal of Hydrology 510: 551–564. est with respect to the research, authorship, and/or Brown RA, Pasternack GB and Lin T (2015) The topo- publication of this article. graphic design of river channels for form-process linkages. Environmental Management 57(4): 929–942. Funding Brown RA, Pasternack GB and Wallender WW (2014) Synthetic river valleys: creating prescribed topography The author(s) disclosed receipt of the following for form–process inquiry and river rehabilitation financial support for the research, authorship, and/ design. Geomorphology 214: 40–55. or publication of this article: This project was sup- Buffington JM and Montgomery DR (2013) Geomorphic ported by the USDA National Institute of Food and classification of rivers. In: Shroder J and Wohl E (eds) Agriculture, Hatch project numbers #CA-D-LAW- Treatise on Geomorphology Fluvial Geomorphology. 7034-H and CA-D-LAW-2243-H. This research was Volume 9. San Diego, CA: Academic Press, 730–767. also supported by the California State Water Bunte K and Abt SR (2001) Sampling surface and Resources Control Board through the sub-agreement subsurface particle-size distributions in wadable 15MLSK0192. gravel- and cobble-bed streams for analyses in sediment transport, hydraulics, and streambed monitoring. References General technical report RMRS-GTR-74. Fort Collins, Anderson MJ (2001) A new method for non-parametric CO: US Department of Agriculture, Forest Service, multivariate analysis of variance. Austral Ecology Rocky Mountain Research Station. 26(1): 32–46 Burnett K, Reeves G, Miller D, et al. (2003) A first step Arnold CL, Boison PJ and Patton PC (1982) Sawmill toward broad-scale identification of freshwater pro- brook: an example of rapid geomorphic change related tected areas for Pacific salmon and trout in Oregon to urbanization. Journal of Geology 90: 155–166 26 Progress in Physical Geography XX(X)

USA. In: Aquatic protected areas: what works best and Dunne T and Leopold LB (1978) Water in Environmental how do we know? (ed Beumer JP, Grant A and Smith Planning. New York: WH Freeman and Company. DC), Cairns, Australia, ?? August 2002, 144–154. Dynesius M and Nilsson C (1994) Fragmentation and flow California Geological Survey (2002) California Geo- regulation of river systems in the northern third of the morphic Provinces Note 36. California Department of world. Science 266(5186): 753–762. Conservation, California Geological Survey. Edwards TK and Glysson GD (1999) Field Methods for Carson MA (1984) The meandering-braided river threshold: Measurement of Fluvial Sediment. Techniques of a reappraisal. Journal of Hydrology 73(3): 315–334. Water-Resources Investigations 03-C2. US Geological Carter J and Resh V (2005) Pacific Coast rivers of the Survey. coterminous United States. In: Benke A and Colbert C Elliott CM, Huhmann BL and Jacobson RB (2009) Geo- (eds) Rivers of North America. San Diego, CA: Else- morphic classification of the Lower Platte River, vier, 547–552. Nebraska. Report no. 2009-5198. Reston, VA: US Columbia Habitat Monitoring Program (2016) Scientific Geological Survey. protocol for salmonid habitat surveys within the Elmqvist T, Folke C, Nystro¨m M, et al. (2003) Response Columbia Habitat Monitoring Program. Report for the diversity ecosystem change and resilience. Frontiers in Bonneville Power Administration. Ecology and the Environment 1(9): 488–494. Church M (1992) Channel morphology and typology. In: Fausch KD, Torgersen CE, Baxter CV, et al. (2002) Calow P and Petts GE (eds) The Rivers Handbook. Landscapes to riverscapes: bridging the gap between Oxford: Blackwell Publications, 126–143. research and conservation of stream fishes a continuous Church M (2002) Geomorphic thresholds in riverine view of the river is needed to understand how processes landscapes. Freshwater Biology 47(4): 541–557. interacting among scales set the context for stream Church M (2006) Bed material transport and the mor- fishes and their habitat. BioScience 52(6): 483–498. phology of alluvial rivers. Annual Review of Earth and Fenneman NM and Johnson DW (1946) Physiographic Planetary Sciences 34: 325–354. divisions of the conterminous United States. Reston, Clarke KR (1993) Non-parametric multivariate analyses of VA: US Geological Survey. changes in community structure. Australian Journal of Findlay S (1995). Importance of surface-subsurface Ecology 18: 117–143. exchange in stream ecosystems: the hyporheic zone. Davis WM (1909) The geographical cycle. In: Geo- Limnology and Oceanography 40(1): 159–164. graphical Essays. New York: Ginn and Company. Fisher SG, Grimm NB, Martı´ E, et al. (1998) Hierarchy, De’ath G and Fabricius KE (2000) Classification and spatial configuration, and nutrient cycling in a desert regression trees: a powerful yet simple technique for stream. Australian Journal of Ecology 23(1): 41–52. ecological data analysis. Ecology 81(11): 3178–3192. Flores AN, Bledsoe BP, Cuhaciyan CO, et al. (2006) Dietrich WE and Smith JD (1983) Influence of the point Channel-reach morphology dependence on energy bar on flow through curved channels. Water Resources scale and hydroclimatic processes with implications for Research 19(5): 1173–1192. prediction using geospatial data. Water Resources Dietrich WE and Whiting PJ (1989) Boundary shear stress Research 42(6): ??-??. and sediment transport in river meanders of sand and Frissell CA, Liss WJ, Warren CE, et al. (1986) A hier- gravel. In: Ikeda S and Parker G (eds) River Meandering. archical framework for stream habitat classification: Washington, DC: American Geophysical Union, 1–50. viewing streams in a watershed context. Environmental Dietrich WE, Kirchner JW, Ikeda H, et al. (1989) Sediment Management 10(2): 199–214. supply and the development of the coarse surface layer Gangodagamage C, Barnes E and Foufoulageorgiou E in gravel-bedded rivers. Nature 340(6230): 215–217. (2007) Scaling in river corridor widths depicts orga- Dietrich WE, Wilson CJ, Montgomery DR, et al. (1992) nization in valley morphology. Geomorphology 91(3): Erosion thresholds and land surface morphology. 198–215. Geology 20(8): 675–679. Gartner JD, Dade WB, Renshaw CE, et al. (2015) Drew F (1873) Alluvial and lacustrine deposits and glacial Gradients in stream power influence lateral and records of the Upper-Indus Basin. Quarterly Journal of downstream sediment flux in floods. Geology 43(11): the Geological Society 29(1–2): 441–471. 983–6. Lane et al. 27

Gasparini NM, Tucker GE and Bras RL (2004) Network- aquatic habitat. In: Proceedings of the 2nd International scale dynamics of grain-size sorting: implications for Symposium on Habitats and Hydraulics (ed Leclerc M, downstream fining stream-profile concavity and drai- Capra H, Valentin S, et al.), Quebec, Canada, Date? nage basin morphology. Earth Surface Processes and 1996, 299–306. Landforms 29: 401–421. Hack JT (1957) Studies of Longitudinal Stream Profiles in Geist DR (2000) Hyporheic discharge of river water into Virginia and Maryland. US Geological Survey Pro- fall chinook salmon (Oncorhynchus tshawytscha) fessional Paper 294-B. US Geological Survey. spawning areas in the Hanford Reach, Columbia River. Hack JT and Goodlett JC (1960) Geomorphology and Canadian Journal of Fisheries and Aquatic Sciences Forest Ecology of a Mountain Region in the Central 57: 1647–1656. Appalachians. US Geological Survey Professional Gleason CJ and Wang J (2015) Theoretical basis for at-many- Paper 347. US Geological Survey. stations hydraulic geometry. Geophysical Research Hanak EE, Lund J, Dinar A, et al. (2011) Managing Cali- Letters 42(17): 7107–7114. fornia’s Water: From Conflict to Reconciliation.San Gonzalez RL and Pasternack GB (2015) Reenvisioning Francisco, CA: Public Policy Institute of California. cross-sectional at-a-station hydraulic geometry as Hartigan JA and Wong MA (1979) Algorithm AS 136: a spatially explicit hydraulic topography. Geomorphol- k-means clustering algorithm. Journal of the Royal ogy 246: 394–406. Statistical Society Series C: Applied Statistics 28(1): Gonza´lez del Tańago M and Garcıá de Jaloń D (2004) 100–108. Hierarchical classification of rivers: a proposal for eco- Hassan MA, Church M, Lisle TE, et al. (2005) Sediment geomorphic characterization of Spanish rivers within transport and channel morphology of small forested the European Water Frame Directive. In: Fifth inter- streams. Journal of the American Water Resources national symposium on ecohydraulics. Aquatic habi- Association 41(4): 853–876. tats: analysis and restoration (ed Garcıá de Jaloń D and Healey MC, Dettinger MD and Norgaard RB (eds) (2008) Vizcaıńo P), Madrid, 12–17 September 2004, pp. 205–211. The State of the Bay-Delta Science 2008. Sacramento, Madrid: IAHR Congress Proceedings. CA: CALFED Science Program. Gostner W, Alp M, Schleiss AJ, et al. (2013a) The hydro- Henderson FM (1963) Stability of alluvial channels. morphological index of diversity: a tool for describing Transactions of the American Society of Civil Engi- habitat heterogeneity in river engineering projects. neers 128: 657–686. Hydrobiologia 712(1): 43–60. Hickin EJ (1974) The development of meanders in natural Gostner W, Parasiewicz P and Schleiss AJ (2013b) A case river channels. American Journal of Science 274: study on spatial and temporal hydraulic variability in 414–442. an alpine gravel-bed stream based on the hydro- Hupp CR and Osterkamp WR (1996) Riparian vegetation morphological index of diversity. Ecohydrology 6(4): and fluvial geomorphic processes. Geomorphology 652–667. 14(4): 277–295. Graf WL (1980) The effect of dam closure on downstream Jackson CR, Leigh DS, Scarbrough SL, et al. (2015) Her- rapids. Water Resources Research 16(1): 129. baceous versus forested riparian vegetation: narrow and Grams PE, Topping DJ, Schmidt JC, et al. (2013) Linking simple versus wide woody and diverse stream habitat. morphodynamic response with sediment mass balance River Research and Applications 31(7): 847–857. on the Colorado River in Marble Canyon: issues of Jaeger KL (2015) Reach-scale geomorphic differences scale, geomorphic setting, and sampling design. Jour- between headwater streams draining mountaintop nal of Geophysical Research: Earth Surface 118(2): mined and unmined catchments. Geomorphology 236: 361–381. 25–33. Grant GE, Swanson FJ and Wolman MG (1990) Pattern Johnson DH (1980) The comparison of use and availability and origin of stepped-bed morphology in high-gradient measurements for evaluating resource preference. Ecol- streams Western Cascades Oregon. Geological Society ogy 71: 61–65. of America Bulletin 102: 340–352. Kasahara and Wondzell (2003) Geomorphic controls on Gubala CP, Eilers JM and Bemert JA (1996) The relation- hyporheic exchange flow in mountain streams. Water ships between river channel morphology complexity and Resources Research 39(1); ??-??. 28 Progress in Physical Geography XX(X)

Kasprak A, Hough-Snee N, Beechie T, et al. (2016) The Lindley ST, Schick RS, Mora E, et al. (2007) Framework blurred line between form and process: a comparison of for assessing viability of threatened and endangered stream channel classification frameworks. PLoS One Chinook salmon and steelhead in the Sacramento–San 11(3): e0150293. Joaquin Basin. San Francisco Estuary and Watershed Kaufman LR and Rousseeuw PJ (1990) Finding Groups in Science 5(1). Data: An Introduction to Cluster Analysis. Hoboken, Lisle TE (1986) Stabilization of a gravel channel by large NJ: John Wiley and Sons. streamside obstructions and bedrock bends Jacoby Kennedy JF (1975) Hydraulic relations for alluvial Creek northwest California. Geological Society of streams. In: Vanoni V (ed) Sedimentation Engineering America Bulletin 97: 999–1011. Manual 54. New York: American Society of Civil MacWilliams ML, Wheaton JM, Pasternack GB, et al. Engineers, 114–154. (2006) Flow convergence routing hypothesis for pool- Knighton AD (1999) Downstream variation in stream riffle maintenance in alluvial rivers. Water Resources power. Geomorphology 29(3): 293–306. Research 42(10): W10427. Knighton AD and Nanson GC (1993) Anastomosis and the Magilligan FJ and Nislow KH (2005) Changes in hydro- continuum of channel pattern. Earth Surface Processes logic regime by dams. Geomorphology 71(1): 61–78. and Landforms 18: 613–625. Makaske B (2001) Anastomosing rivers: a review of their Kondolf GM (1995) Geomorphological stream channel classification origin and sedimentary products. Earth- classification in aquatic habitat restoration: uses and Science Reviews 53(3): 149–196. limitations. Aquatic Conservation: Marine and Fresh- Manly BFJ and Alberto JA (2014) Introduction to Ecolo- water Ecosystems 5(2): 127–141. gical Sampling. Boca Raton, FL: Taylor and Francis Lane BA, Dahlke HE, Pasternack GB, et al. (2017) Group LLC. Revealing the diversity of natural hydrologic regimes McCluney KE, Poff NL, Palmer MA, et al. (2014) Riverine in California relevant to future environmental flows macrosystems ecology: sensitivity resistance and resi- applications. Journal of American Water Resources lience of whole river basins with human alterations. Association 53(2): 411–430. Frontiers in Ecology and the Environment 12(1): 48–58. Lane EW (1957) A study of the shape of channels formed McCune Bruce, James B. Grace and Dean L. Urban (2012) by natural streams flowing in erodible material. Analysis of Ecological Communities. Vol. 28. Gleneden Omaha, NE: United States Army Engineer Division, Beach, OR: MjM software design. Missouri River. Meitzen KM, Doyle MW, Thoms MC, et al. (2013) Geo- Lane EW and Carlson EJ (1953) Some factors affecting the morphology within the interdisciplinary science of stability of canals constructed in coarse granular environmental flows. Geomorphology 200: 143–154. material. US Bureau of Reclamation. Merriam ER, Petty JT, Merovich GT Jr, et al. (2011) Lejot J, Delacourt C, Pie´gay H, et al. (2007) Very high Additive effects of mining and residential development spatial resolution imagery for channel bathymetry and on stream conditions in a central Appalachian water- topography from an unmanned mapping controlled shed. Journal of the North American Benthological platform. Earth Surface Processes and Landforms Society 30(2): 399–418. 32(11): 1705–1725. Merritt DM and Wohl EE (2003) Downstream hydraulic Leopold LB and Maddock T Jr (1953) The Hydraulic geometry and channel adjustment during a flood along Geometry of Stream Channels and Some Physiographic an ephemeral arid-region drainage. Geomorphology Implications. US Geological Survey Professional paper 52(3): 165–180. 252. US Geological Survey. Miller AW and Ambrose RF (2000) Sampling patchy Leopold LB and Wolman MG (1957) River Channel Pat- distributions: comparison of sampling designs in rocky terns: Braided Meandering and Straight. US Geological intertidal habitat. Marine Ecology Progress Series 196: Survey Professional Paper 282-B. US Geological Survey. 1–14. Leung LR, Qian Y and Bian X (2003) Hydroclimate of the Miller DJ and Burnett KM (2008) A probabilistic model of western United States based on observations and debris-flow delivery to stream channels demonstrated regional climate simulation of 1981–2000. Part I. Sea- for the Coast Range of Oregon USA. Geomorphology sonal statistics. Journal of Climate 16: 1892–1911. 94(1): 184–205. Lane et al. 29

Milner VS, Willby NJ, Gilvear DJ, et al. (2015) Linkages Oksanen J (2011) Multivariate analysis of ecological between reach scale physical habitat and invertebrate communities. R: Vegan Tutorial R Package Version 1. assemblages in upland streams. Marine and Fresh- Orr HG, Large AR, Newson MD, et al. (2008) A predictive water Research 66: 438–448. typology for characterising hydromorphology Geo- Molles MC, Crawford CS, Ellis LM, et al. (1998) Managed morphology 100(1): 32–40. flooding for riparian ecosystem restoration. BioScience Palmer MA, Hakenkamp CC and Nelson-Baker K (1997) 48(9): 749–756. Ecological heterogeneity in streams: why variance Montgomery DR and Bolton SM (2003) Hydrogeomorphic matters. Journal of the North American Benthological variability and river restoration. In: Wissmar RC and Society 16(1):189–202. Bisson PA (eds) Strategies for Restoring River Ecosys- Parker G (1979) Hydraulic geometry of active gravel rivers. tems: Sources of Variability and Uncertainty in Natural Journal of the Hydraulics Division 105(9): 1185–1201. and Managed Systems.Bethesda,MD:AmericanFish- Parker G, Toro-Escobar CM, Ramey M, et al. (2003) eries Society, 39–80. Effect of floodwater extraction on mountain stream Montgomery DR and Buffington JM (1993) Channel morphology. Journal of Hydraulic Engineering classification prediction of channel response and 129(11): 885–895. assessment of channel condition. Report no. TFW- Paustian SJ, Trull SJ, Foster ND, et al. (1992) A channel SH10-93-002. Washington State Department of Natu- type user guide for the Tongass National Forest, ral Resources. southeast Alaska. R10 Technical Paper 26. US Forest Montgomery DR and Buffington JM (1997) Channel- Service, Alaska Region. reach morphology in mountain drainage basins. Geo- Peterson DF and Mohanty PK (1960) Flume studies of logical Society of America Bulletin 109: 596–611. flow in steep rough channels. Journal of the Hydraulics Moyle PB, Katz JV and Quin˜ones RM (2011) Rapid Division 86(9): 55–76. decline of California’s native inland fishes: a Pitlick J and Van Steeter MM (1998) Geomorphology and status assessment. Biological Conservation 144(10): endangered fish habitats of the upper Colorado River. 2414–2423. 2. Linking sediment transport to habitat maintenance. Murray O, Thoms M and Rayburg S (2006) The diversity Water Resources Research 34(2): 303–316. of inundated areas in semiarid flood plain ecosystems. Poff NL and Ward JV (1990) Physical habitat template of In: Rowan JS, Duck RW and Werritty A (eds) Sediment lotic systems: recovery in the context of historical Dynamics and the Hydromorphology of Fluvial Sys- pattern of spatiotemporal heterogeneity. Environmen- tems. Wallingford, UK: IAHS Press, 277–286. tal Management 14(5): 629–645. Murtagh F and Legendre P (2013) Ward’s hierarchical Pollock MM, Naiman RJ and Hanley TA (1998) Plant agglomerative clustering method: which algorithms species richness in riparian wetlands: a test of biodi- implement Ward’s criterion? Journal of Classification versity theory. Ecology 79(1): 94–105. 31: 274. Polvi LE, Nilsson C and Hasselquist EM (2014) Potential Nagel DE, Buffington JM, Parkes SL, et al. (2014) A and actual geomorphic complexity of restored head- landscape scale valley confinement algorithm: deli- water streams in northern Sweden. Geomorphology neating unconfined valley bottoms for geomorphic 210: 98–118. aquatic and riparian applications. General technical Polvi LE, Wohl EE and Merritt DM (2011) Geomorphic report RMRS-GTR. USDA Forest Service. and process domain controls on riparian zones in Nelson JM and Smith JD (1989) Mechanics of flow over the Colorado Front Range. Geomorphology 125(4): ripples and dunes. Journal of Geophysical Research: 504–516. Oceans 94(C6): 8146–8162. Poole GC, O’Daniel SJ, Jones KL, et al. (2008) Hydrologic Ode PR (2007) Standard operating procedures for collect- spiraling: the role of multiple interactive flow paths in ing benthic macroinvertebrate samples and associated stream ecosystems. River Research and Applications physical and chemical data for ambient bioassessments 24(7): 1018–1031. in California. Bioassessment SOP 1. California State Poole GC, Stanford JA, Running SW, et al. (2006) Mul- Water Resources Control Board Surface Water Ambient tiscale geomorphic drivers of groundwater flow paths: Monitoring Program. subsurface hydrologic dynamics and hyporheic habitat 30 Progress in Physical Geography XX(X)

diversity. Journal of the North American Benthological Schumm SA, Harvey MD and Watson CC (1984) Incised Society 25(2): 288–303. Channels: Morphology Dynamics and Control. Lit- Powell DM, Brazier R, Wainwright J, et al. (2005) tleton, CO: Water Resources Publications. Streambed scour and fill in low-order dryland channels. Scown MW, Thoms MC and De Jager NR (2016) An index Water Resources Research 41(5): W05019. of floodplain surface complexity Hydrology and Earth Price AE, Humphries P, Gawne B, et al. (2012) Effects of System Sciences Discussions 12: 4507–4540. discharge regulation on slackwater characteristics at Sear D, Newson M, Hill C, et al. (2009) A method multiple scales in a lowland river. Canadian Journal of for applying fluvial geomorphology in support of Fisheries and Aquatic Sciences 70(2): 253–262. catchment-scale river restoration planning. Aquatic Rathburn SL and Wohl EE (2003) Sedimentation hazards Conservation: Marine and Freshwater Ecosystems downstream from reservoirs. In: Graff WL (ed) Dam 19(5): 506–19. Removal Research: Status and Prospects. Washington, Simonson TD, Lyons J and Kanehl PD (1994) Quantifying DC: The Heinz Center, 105–118. fish habitat in streams: transect spacing sample size and Raven PJ, Holmes NT, Dawson FH, et al. (1998) Quality a proposed framework North American Journal of assessment using River Habitat Survey data. Aquatic Fisheries Management 14: 607–615. Conservation: Marine and Freshwater Ecosystems 8: Sklar L and Dietrich WE (1998) River longitudinal pro- 477–499. files and bedrock incision models: Stream power and Rivas Casado MR, Ballesteros Gonzalez RB, Wright R, the influence of sediment supply. In: Tinkler KJ and et al. (2016) Quantifying the effect of aerial imagery Wohl EE (eds) Rivers Over Rock: Fluvial Processes in resolution in automated hydromorphological river Bedrock Channels. Washington, DC: American Geo- characterisation. Remote Sensing 8(8): 650. physical Union, 237–260. Rivas Casado MR, Gonzalez RB, Kriechbaumer T, et al. Smith RD, Ammann A, Bartoldus C, et al. (1995) (2015) Automated identification of river hydro- An approach for assessing wetland functions using morphological features using UAV high resolution hydrogeomorphic classification reference wetlands aerial imagery. Sensors 15(11): 27969–27989. and functional indices. Vicksburg, MS: Army Engineer Rosgen DL (1994) A classification of natural rivers. Waterways Experiment Station. Catena 22(3): 169–199. Stock GM, Anderson RS and Finkel RC (2005) Rates of Rossi RE, Mulla DJ, Journel AG, et al. (1992) Geostatistical erosion and topographic evolution of the Sierra Nevada tools for modeling and interpreting ecological spatial California inferred from cosmogenic 26Al and 10Be dependence. Ecological Monographs 62(2): 277–314. concentrations. Earth Surface Processes and Land- Rowntree K and Wadeson R (1998) A geomorphological forms 30(8): 985–1006. framework for the assessment of instream flow Strom MA, Pasternack GB and Wyrick JR (2016) Reenvi- requirements. Aquatic Ecosystem Health & Manage- sioning velocity reversal as a diversity of hydraulic patch ment 1(2): 125–141. behaviors. Hydrologic Processes 30(13): 2348–2365. Sawyer AM, Pasternack GB, Moir HJ, et al. (2010) Riffle- Sullivan K (1986) Hydraulics and fish habitat in relation pool maintenance and flow convergence routing to channel morphology. PhD Thesis, Johns Hopkins observed on a large gravel-bed river. Geomorphology University, USA. 114(3): 143–160. Sutfin NA, Shaw JR, Wohl EE, et al. (2014) A geomorphic Schmitt L, Marie G, Nobelis P, et al. (2007) Quantitative classification of ephemeral channels in a mountainous morphodynamic typology of rivers: a methodological arid region southwestern Arizona USA. Geomorphol- study based on the French Upper Rhine basin. Earth ogy 221: 164–175. Surface Processes and Landforms 32: 1726–1746. Therneau TM, Atkinson B and Ripley MB (2010) Schneider R (1994) The role of hydrologic regime in main- The rpart package. Available at: http://mayoresearch. taining rare plant communities of New York’s coastal mayo.edu/mayo/research/biostat/splusfunctions.cfm plain pondshores. Biological Conservation 68: 253–260. (accessed 10 June 2015). Schumm SA (1963) A tentative classification of alluvial Thomas RB and Lewis J (1995) An evaluation of flow- river channels. United States Geological Survey Cir- stratified sampling for estimating suspended sediment cular 477. loads. Journal of Hydrology 170: 27–45. Lane et al. 31

Thompson A (1986) Secondary flows and the pool-riffle unit: Walters DM, Leigh DS, Freeman MC, et al. (2003) Geo- a case study of the processes of meander development. morphology and fish assemblages in a Piedmont river Earth Surface Processes and Landforms 11(6): 631–641. basin USA. Freshwater Biology 48: 1950–1970. Thompson DM and Wohl E (2009) The linkage between Ward JH Jr (1963) Hierarchical grouping to optimize an velocity patterns and sediment entrainment in a forced- objective function. Journal of the American Statistical pool and riffle unit. Earth Surface Processes and Association 58(301): 236–244. Landforms 34: 177–192. Wenger SJ, Isaak DJ, Dunham JB, et al. (2011) Role of Thoms MC (2006) Variability in riverine ecosystems. climate and invasive species in structuring trout dis- River Research and Applications 22: 115–151. tributions in the interior Columbia River Basin USA. Thornbury WB (1954) Principles of Geomorphology. New Canadian Journal of Fisheries and Aquatic Sciences York: Wiley. 68(6): 988–1008. Tonina D and Buffington JM (2009) Hyporheic exchange White JQ, Pasternack GB and Moir HJ (2010) Valley in mountain rivers. I. Mechanics and environmental width variation influences riffle-pool location and effects. Geography Compass 3(3): 1063–1086. persistence on a rapidly incising gravel-bed river. Townsend CR and Hildrew AG (1994) Species traits in Geomorphology 121: 206–221. relation to a habitat templet for river systems. Fresh- Whiting PJ and Dietrich WE (1991) Convective accel- water Biology 31(3): 265–275. erations and boundary shear stress over a channel bar. Trainor K and Church M (2003) Quantifying variability in Water Resources Research 27(5): 783–796. stream channel morphology. Water Resources Research Wilcox AC and Wohl EE (2006) Flow resistance dynamics 39(9): ESG5–1. in step-pool streams channels. 1. Large woody debris US Geological Survey (2009) National elevation dataset and controls on total resistance. Water Resources (NED). Available at: http://nhd.Usgs.gov/ (accessed 28 Research 42(5): W05418. May 2016). Wohl EE and Thompson DM (2000) Velocity character- US Geological Survey (2011) Study unit description for istics along a small step-pool channel. Earth Surface Sacramento River Basin NAWQA program. Available Processes and Landforms 25: 353–367. at: http://ca.water.usgs.gov/sac_nawqa/study_descrip Wohl EE, Bledsoe BP, Jacobson RB, et al. (2015) The tion.html (accessed 28 May 2016). natural sediment regime: broadening the foundation for US Geological Survey (2013) National hydrography geo- ecosystem management. BioScience 65: 358–371. database: the national map viewer. Available at: http:// Wolman MG (1954) A method of sampling coarse river-bed viewer.nationalmap.gov/viewer/nhd.html?p¼nhd material. American Geophysical Union 35(6): 951–956. (accessed 1 June 2015). Woodsmith RD and Hassan MA (2005) Maintenance of Vannote RL, Minshall GW, Cummins KW, et al. (1980) an obstruction-forced pool in a gravel-bed channel: The river continuum concept. Canadian Journal of streamflow channel morphology and sediment trans- Fisheries and Aquatic Sciences 37(1): 130–137. port. Developments in Earth Surface Processes 7: Varanka S, Hjort J and Luoto M (2014) Geomorphological 169–196. factors predict water quality in boreal rivers. Earth Wyrick JR and Pasternack GB (2008) Modeling energy Surface Processes and Landforms 40(15): 1989–1999. dissipation and hydraulic jump regime responses to Virtanen R, Luoto M, Ra¨ma¨ T, et al. (2010) Recent channel nonuniformity at river steps. Journal of Geo- vegetation changes at the high latitude tree line ecotone physical Research 113(F3): ??-??. are controlled by geomorphological disturbance pro- Wyrick JR and Pasternack GB (2015) Revealing the natural ductivity and diversity. Global Ecology and Biogeo- complexity of topographic change processes through graphy 19(6): 810–821. repeat surveys and decision-tree classification. Earth Walsh CJ, Roy AH, Feminella JW, et al. (2005) The urban Surface Processes and Landforms 41(6): 723–737. stream syndrome: current knowledge and the search for Yang D and Woo M-K (1999) Representativeness of local a cure. Journal of the North American Benthological snow data for large scale hydrologic investigations. Society 24(3): 706–723. Hydrological Processes 13:1977–1988.