Spatially Continuous Analysis of Fish–Habitat Relationships

Landscape Influences on Longitudinal Patterns of River Fishes: Spatially Continuous Analysis of Fish–Habitat Relationships

Christian E. Torgersen*,1 and Colden V. Baxter2 Oregon Cooperative Fish and Wildlife Research Unit, Department of Fisheries and Wildlife Oregon State University, Corvallis, Oregon 97331, USA

Hiram W. Li Oregon Cooperative Fish and Wildlife Research Unit (U.S. Geological Survey) Department of Fisheries and Wildlife, 104 Nash Hall Oregon State University, Corvallis, Oregon 97331, USA

Bruce A. McIntosh Oregon Department of Fish and Wildlife, Corvallis Research Laboratory 28655 Highway 34, Corvallis, Oregon 97333, USA

Abstract.—Longitudinal analysis of the distribution and abundance of river fishes provides a context-specific characterization of species responses to riverscape heterogeneity. We examined spatially continuous longitudinal profiles (35–70 km) of fish distribution and aquatic habitat (channel gradient, depth, temperature, and water velocity) for three northeastern Ore- gon rivers. We evaluated spatial patterns of river fishes and habitat using multivariate analysis to compare gradients in fish assemblage structure among rivers and at multiple spatial scales. Spatial structuring of fish assemblages exhibited a generalized pattern of cold- and coolwater fish assemblage zones but was variable within thermal zones, particularly in the warmest river. Landscape context (geographic setting and thermal condition) influenced the observed relationship between species distribution and channel gradient. To evaluate the effect of spatial extent and geographical context on observed assemblage patterns and fish–habitat relationships, we performed multiple ordinations on subsets of our data from varying lengths of each river and compared gradients in assemblage structure within and among rivers. The relative associations of water temperature increased and channel morphology decreased as the spatial scale of analysis increased. The crossover point where both variables explained equal amounts of variation was useful for identifying transitions between cool- and coldwater fish assemblages. Spatially continuous analysis of river fishes and their habitats revealed unexpected ecological patterns and provided a unique perspective on fish distribution that emphasized the importance of habitat heterogeneity and spatial variability in fish–habitat relationships.

INTRODUCTION *Corresponding author: [email protected] 1 Present address: USGS-FRESC Cascadia Field Station, Studies of river fish assemblages often focus on 344 Bloedel Hall, University of Washington, Seattle, describing and understanding patterns in species Washington 98195-2100, USA. composition that occur along the length of river 2 Present address: Department of Biological Sciences, systems. In general, fish assemblage structure is Idaho State University, Pocatello, Idaho 83209, USA. thought to change predictably from headwaters

473

22torgersen.p65 473 7/28/2006, 10:00 AM 474 Torgersen et al.

to downstream reaches, with biotic zones (e.g., of spatial variability in river fish–habitat relation- cold- and warmwater assemblages) occurring in ships remain unanswered: How finely tuned are which species are added or replaced in response longitudinal patterns in fish assemblages to key to continuous gradients in temperature, chan- habitat factors such as thermal heterogeneity, nel morphology, and water velocity (Huet 1959; channel morphology, and velocity? Can the ef- Sheldon 1968; Horwitz 1978; Hughes and fects of temperature on fish assemblages be iso- Gammon 1987; Li et al. 1987; Rahel and Hubert lated from the effects of other factors? How are 1991; Paller 1994; Belliard et al. 1997). Biotic assemblage patterns at one scale mediated by zonation and species addition are dominant, context at larger spatial scales? Does perception coarse-scale patterns that have been described of habitat relationships change with the spatial by numerous studies during the last century extent of a study? We propose that these ques- (Matthews 1998). Beyond these patterns, how- tions can be addressed only by adapting and ever, there is little understanding of spatial het- changing the manner in which fish assemblage erogeneity in fish distribution and habitat and habitat data are collected and analyzed. relationships within biotic zones or the effects Here we illustrate a new approach to collect- of spatial scale and context on observed fish as- ing and analyzing fish assemblage and habitat semblage patterns in rivers (Collares-Pereira et data that provides a more spatially continuous al. 1995; Duncan and Kubecka 1996; Poizat and view of fishes and the riverine landscapes, or Pont 1996; Bult et al. 1998; Fausch et al. 2002). “riverscapes,” they inhabit (Fausch et al. 2002). Consequently, there are likely many more pat- Our objectives were to (1) collect spatially con- terns and spatial relationships that have yet to tinuous data on fish assemblage structure and be described, and these may be essential to un- habitat along the length of three rivers with con- derstanding river fish assemblages. trasting physical environments, (2) characterize Discovery of new patterns and spatial rela- and compare longitudinal patterns and habitat tionships may be constrained by repeated use of relationships (water depth, velocity, channel gra- traditional study approaches. Relatively short dient, and water temperature) among and within sampling reaches (<500 m) spaced at wide in- these riverscapes, and (3) evaluate the effect of tervals (>10 km) along the longitudinal profile spatial extent and geographical context of sur- or throughout a channel network may provide vey data on observed fish–habitat relationships. the information necessary to detect coarse gradients in fish assemblage structure associated METHODS with factors such as temperature and stream order (Vannote et al. 1980). However, such site- Study Area based studies lack the spatial resolution necessary for detecting patterns in fish–habitat relation- We studied fish assemblages in three small riv- ships across a range of spatial scales. Conse- ers in the Blue Mountains of northeastern Ore- quently, the perception that river fish gon: the Middle Fork John Day (MFJD; upper assemblages change gradually with respect to 49 km), the North Fork John Day (NFJD; upper longitudinal habitat gradients may be driven 70 km), and the Wenaha River (WEN; lower 35 largely by the resolution and extent of data col- km; Figure 1). Study section elevations ranged lection and analysis (Naiman et al. 1988; Wiens from 500 m in the lower WEN to 1,700 m in the 1989; Poole 2002). As a consequence of the dis- upper NFJD and shared a similar geology of continuous and spatially limited manner in Columbia River basalt at lower elevations and which river fishes are traditionally sampled, fun- folded metamorphosed rocks partially overlain damental questions about the nature and extent by volcanic tuff in headwater reaches (Orr et al.

22torgersen.p65 474 7/28/2006, 10:00 AM Landscape Influences on Longitudinal Patterns of River Fishes 475

Figure 1. Study area and river sections surveyed for fish assemblages in northeastern Oregon. Study rivers included (A) the Middle Fork John Day (MFJD), (B) the North Fork John Day (NFJD), and (C) the Wenaha River (WEN). Black dots indicate the spatial extent and continuity of underwater visual surveys.

22torgersen.p65 475 7/28/2006, 10:00 AM 476 Torgersen et al.

1992). Although the NFJD study section had the ranches. Land-use impacts are minimal in the largest drainage area and the highest elevations, relatively pristine WEN compared to the NFJD the WEN received more annual precipitation and and the MFJD, which have experienced exten- had higher summer base flow (Table 1). Longi- sive mining, grazing, and logging during the last tudinal gradients in elevation and annual pre- century. cipitation were steepest in the WEN, followed by the NFJD and the MFJD. Maximum summer Fish Assemblages water temperature patterns reflected differences in streamflow among basins and represented a Native fish species common in the study rivers range of cool and cold thermal environments included four salmonids, three catostomids, four (Table 1). cyprinids, and two cottids. Two nonnative fishes Seasonal weather patterns throughout the (brook trout Salvelinus fontinalis and small- study area are typical of high desert climates with mouth bass Micropterus dolomieu) were ex- hot, dry summers and cold, relatively wet win- tremely rare and therefore not included in our ters (–15–38°C; Loy et al. 2001). The Blue Moun- analysis. We selected a subset of species for as- tains ecoregion is characterized by contrasts in semblage analysis based on their relative abun- temperature, precipitation, and vegetation cor- dance and ease of identification underwater responding with steep elevation gradients (Figure 2). We noted sculpins Cottus spp., (Clarke and Bryce 1997). Canyons and alluvial longnose dace Rhinichthys cataractae, and moun- valleys in the Wenaha and John Day River ba- tain sucker Catostomus platyrhynchus during sins are vegetated with mixed conifer forest (pon- surveys but did not include them in analysis be- derosa pine Pinus ponderosa, grand fir Abies cause they were difficult to detect and identify grandis, Douglas-fir Pseudotsuga menziesii, west- underwater, as determined by comparisons of ern larch Larix occidentalis, and lodgepole pine snorkeling and electrofishing in selected sections Pinus contorta) on the upslopes and broadleaf of the MFJD (H. W. Li, unpublished data). assemblages of black cottonwood Populus Fish assemblage zones overlapped in each of trichocarpa, willow Salix spp., and red alder Alnus the study rivers and provided an excellent op- rubra in the valley bottoms. The upper NFJD and portunity to evaluate patterns in assemblage the WEN are designated wild and scenic rivers structure in relation to water temperature and situated within public wilderness areas, whereas channel morphometry. Cold- and coolwater the MFJD flows mainly through private cattle temperature classifications were based on species

Table 1. Physical characteristics of study sections in the Middle Fork John Day (MFJD), the North Fork John Day (NFJD), and the Wenaha (WEN) rivers.

River Drainage Summer Water kilometer area Elevation Stream Precipitation base flow temperature River (rkm)a (km2)b (m) orderc (cm/year) (m3/s)d (°C)e MFJD 62–117 1,000 1,000–1,300 4th–5th 35–60 1.4 21.1–25.2 NFJD 95–165 1,600 800–1,700 4th–5th 50–90 2.8 19.1–25.0 WEN 0–35 750 500–1,100 4th–5th 50–150 5.7 15.1–21.3 a Distance upstream from mouth. b Drainage area at lower boundary of study section. c Range in stream order between upper and lower boundaries of study section (determined from 1:100,000-scale U.S. Geological Survey topographic maps). d Streamflow estimates are approximations of summer low-flow conditions based on field measurements in late August and September 1997– 1999. e Range in mean maximum water temperature on 1–7 August 1998 at the upstream and downstream boundaries of the study sections.

22torgersen.p65 476 7/28/2006, 10:00 AM Landscape Influences on Longitudinal Patterns of River Fishes 477

1 2

3 4

Salmonidae

5 8

6 7 9

Cyprinidae Catostomidae

IDSpecies Common name

1. Oncorhynchus mykiss rainbow trout (R T) 2. Salvelinus confluentus bull trout (BT) 3. Oncorhynchus tshawytscha juvenile Chinook salmon (JCS) 4. Prosopium williamsoni mountain whitefish (MW) 5. Ptychocheilus oregonensis northern pikeminnow (NP) 6. Richardsonius balteatus redside shiner (RS) 7. Rhinichthys osculus speckled dace (SD) 8. Catostomus macrocheilus largescale sucker (LS) 9. Catostomus columbianus bridgelip sucker (BS)

Figure 2. River fish assemblage surveyed in northeastern Oregon. Benthic fish species, including longnose dace Rhinichthys cataractae, mountain sucker Catostomus platyrhynchus, torrent sculpin Cottus rhotheus, and Paiute sculpin C. beldingii, were noted during surveys but not included in analyses. The species code used in subsequent figures is listed after the common name.

ranges, spawning seasons, spawning tempera- August 1996 (MFJD), 1997 (NFJD), and 1998 tures, and physiological optima (Zaroban et al. (WEN). Underwater snorkel surveys provide 1999). Coldwater species included only the accurate assessments of fish abundance in flow- salmonids, whereas coolwater species comprised ing waters and offer an alternative to electro- both catostomids and cyprinids. fishing when it is restricted by management agencies or when rivers are too large to sample Longitudinal Surveys of Fish Distribution effectively with a backpack electrofisher and too and Aquatic Habitat small to sample by boat (Cunjak et al. 1988; Zubik and Fraley 1988; Thurow and Schill 1996; We conducted extensive snorkel surveys to quan- Mullner et al. 1998; Joyce and Hubert 2003). We tify longitudinal patterns in river fish assemblages evaluated the distribution and abundance of during summer low-flow conditions in July– river fishes using a modified version of point

22torgersen.p65 477 7/28/2006, 10:00 AM 478 Torgersen et al.

abundance sampling (Persat and Copp 1990). data on fish assemblages, field crews collected The objective of modified point abundance sam- information on channel morphology (e.g., side pling was to collect large numbers of closely channel/main channel, depth, width, velocity) spaced samples (<100 m separation), providing and water temperature and recorded geographic a relatively continuous assessment of fish distri- coordinates (±100 m) of individual sample units bution (Figure 1). Although the size of the study with a handheld global positioning system rivers prevented us from estimating the efficiency (GPS). Field crews placed slow- and fast-water of our sampling procedure (e.g., via compari- habitats in four categories corresponding to wa- son with estimates from multiple-pass ter velocity: (1) pools, (2) slow-moving glides, electrofishing), other work conducted in rivers (3) fast-moving glides, and (4) riffles (Bisson et of similar size has shown that visual estimates al. 1982). Categorical estimates of current veloc- provide an accurate (though perhaps imprecise) ity explained 68% of the variation in current assessment of fish distribution (H. W. Li and P. velocity measured with a flowmeter (n = 33, P < B. Bayley, Oregon State University, unpublished 0.001, y = 0.39 + 0.26x + 0.08x2). data). Three short gaps (2–3 km) in the extensive surveys of the MFJD and the NFJD occurred Geographical Analysis where access was denied to private lands or where and Remote Sensing steep canyons and rapids made sampling too dangerous. We divided survey sections into A geographical information system (GIS) was reaches of equal length and sampled fishes and essential for mapping, displaying, and analyzing habitat with two-person crews consisting of a the large number of sample points required to diver and a data recorder walking along the shore. assess spatial patterns in aquatic habitat and fish Divers counted fish in two or more passes near distribution (Figure 1). We mapped sampled shore and mid-channel in an upstream or down- channel units as individual points linked to a stream direction depending upon water depth database containing information on fish abun- and velocity. Using this approach, a diver– dance and habitat characteristics. Longitudinal recorder crew was capable of surveying an aver- analysis was accomplished using route and dy- age of 2–4 km per day. namic segmentation procedures in ARC/INFO Divers recorded fish abundances in categories GIS (ESRI 1996; Radko 1997). We derived digi- indicating whether a species was dominant tal hydrography layers from 1:5,000-scale aerial (>50%), common (10–50%), or rare (<10%) in photographs (MFJD) and 1:100,000-scale topo- relation to the total number of fish observed in graphic maps (NFJD and WEN). Route-measure a sample unit. Relative abundance provided in- coordinates, defined as the distance upstream formation on the composition but not the abso- from the mouth (i.e., river kilometer, rkm), served lute abundance of fish species in a given channel as a common axis with which to compare longi- unit. Relative abundances were representative of tudinal profiles of fish distribution and aquatic the proportion of all fish of all species estimated habitat. We generated a channel gradient profile to be in a sampled channel unit. This measure from a 10-m digital elevation model (DEM) by of abundance is particularly useful for determin- sampling elevation every 100 m along the river ing the ecological relationships among fish spe- channel and then calculating gradient using a cies (Rahel 1990; Rahel and Hubert 1991; 500-m moving window. Overlays of evenly spaced Reynolds et al. 2003). In all cases, the divers were sample points on the spatially continuous chan- highly experienced in fish identification and nel gradient profile provided a coarse estimate evaluated their estimates of fish abundance regu- of gradient that was consistent among channel larly through repeat dives of the same channel units. Those units varied in length and generally unit by different divers. In addition to collecting decreased in size in an upstream direction.

22torgersen.p65 478 7/28/2006, 10:00 AM Landscape Influences on Longitudinal Patterns of River Fishes 479

We assessed spatially and temporally continu- gradients in assemblage structure (Hughes and ous patterns in water temperature with airborne Gammon 1987; Rahel and Hubert 1991; Paller thermal infrared (TIR) remote sensing and au- 1994; Taylor et al. 1996). However, these methods tomated instream thermographs (Torgersen et are not appropriate for analyzing nonnormally al. 2001). Aerial surveys occurred on cloudless distributed data sets, such as the spatially con- days 4–9 August 1998 at 1300–1400 hours. Ther- tinuous fish assemblage and habitat data col- mographs served as ground-truth points for TIR lected in this study (McCune 1997). Therefore, remote sensing and provided temporal data nec- we computed multivariate ordinations with essary for comparing relative difference in mean nonmetric multidimensional scaling (NMS) in and maximum water temperatures within and PC-ORD, a software package specifically de- among basins. signed for multivariate analysis of ecological data (McCune and Mefford 1999). Nonmetric mul- Data Analysis tidimensional scaling is a nonparametric procedure that calculates axis scores based on ranked To evaluate spatial patterns and associations in distances and therefore alleviates the problems fish distribution, channel morphology, water of zero truncation caused by heterogeneous eco- velocity, and temperature, we compared peaks logical data sets (Clarke 1993; Tabachnick and and troughs in fish abundance to longitudinal Fidell 2001). profiles of habitat. We scaled the relative abun- We calculated two-dimensional solutions in dance estimates for fish (dominant, common, NMS using the Sørensen distance measure and and rare) and categorical estimates of water ve- 15 runs of real data with up to 200 iterations to locity to 1.0, and then plotted fish and habitat evaluate stability. Because of the extremely large variables versus distance upstream from the river sample size, only 30 Monte Carlo runs were suf- mouth. To identify spatial trends in longitudinal ficient to evaluate the probability (␣ = 0.05) that profiles, we used locally weighted scatterplot ordination axes explained more variation than smoothing (LOWESS), a robust, nonparamet- would be expected by chance. To identify envi- ric regression technique used to identify trends ronmental gradients associated with ordination in heterogeneous ecological data (Trexler and axes, we constructed joint plots and biplots Travis 1993). Locally weighted regression calcu- (Jongman et al. 1995) of samples and species in lations used a second-degree polynomial ordination space and examined Pearson corre- smoothing function in SigmaPlot statistical soft- lations between variables in a habitat matrix ware (SPSS 2001). The objective of graphical (mean depth, maximum depth, water velocity, analysis with LOWESS was to explore spatially channel gradient, and water temperature) and continuous patterns and evaluate the extent to ordination axis scores. The statistical significance which longitudinal changes in species distribu- of Pearson correlations provided a relative means tion and habitat were gradual or abrupt in con- to compare correlation strength among habitat trasting riverine environments. variables and ordination axes rather than to test Multivariate analysis was necessary to distin- specific hypotheses (McCune and Grace 2002). guish patterns in fish assemblage structure both To facilitate interpretation of the ordinations, we within and among rivers. Standard parametric rotated the point cluster around the centroid to multivariate methods (e.g., principal compo- align habitat variable vectors in the joint plots nents analysis, detrended correspondence analy- with the primary and secondary ordination axes. sis, and canonical correspondence analysis) are We produced ordinations with fish species and commonly applied in studies of river fishes be- samples plotted in ordination space by calculat- cause they reduce complex species matrices into ing species scores with weighted averaging. We two or more dimensions, or axes, representing then labeled the primary and secondary gradients

22torgersen.p65 479 7/28/2006, 10:00 AM 480 Torgersen et al.

in fish assemblage structure (ordination axes 1 and trout (rkm 83–88). Catostomids (bridgelip 2, respectively) according to the two habitat vari- sucker Catostomus columbianus and largescale ables with which they were most highly correlated. sucker C. macrocheilus) had different patterns of To evaluate the effect of spatial extent and relative abundance depending on the species. geographical context on observed assemblage Peaks in the relative abundance of bridgelip patterns and fish–habitat relationships, we per- sucker occurred downstream of rkm 65, at rkm formed multiple ordinations on subsets of our 85, and upstream of rkm 110, and largescale data from varying lengths of each river and com- sucker were relatively abundant at rkm 72 and pared gradients in assemblage structure within rkm 97. Speckled dace Rhinichthys osculus and and among rivers. We divided each river section redside shiner Richardsonius balteatus were com- into 10 reaches of varying lengths (e.g., rkm 0– mon throughout the MFJD but exhibited local 70, rkm 0–65, rkm 0–60, etc.) and performed a peaks in relative abundance at rkm 83 (both spe- separate ordination for each reach. This process cies) and peaks and troughs, respectively, at rkm is essentially a scaling analysis that quantifies the 101. Northern pikeminnow Ptychocheilus effects of spatial extent and geographic context oregonensis were relatively rare in the MFJD ex- on assemblage composition. Specifically, by com- cept in reaches downstream of rkm 65 and at paring Pearson correlations of environmental rkm 107–112. variables with ordination axis scores along the Fish distribution in the NFJD exhibited dis- longitudinal profile, we were able to examine the tinct peaks and troughs in the relative abundance combined effects of spatial extent (i.e., reach of juvenile Chinook salmon and mountain length) and geographic context on the observed whitefish but was more gradual for rainbow relative influences of habitat on fish assemblage trout, bull trout Salvelinus confluentus, largescale structure. sucker, bridgelip sucker, speckled dace, redside shiner, and northern pikeminnow (Figure 3B). RESULTS Rainbow trout increased in abundance gradually in an upstream direction but were rare in Longitudinal Patterns of the uppermost reaches of the NFJD. Bull trout Individual Fish Species increased in abundance gradually in an upstream direction from the lowermost occurrence at rkm Longitudinal patterns in fish distribution were 150. Bull trout and coolwater species (catosto- gradual for some species and abrupt for others, mids and cyprinids) did not overlap spatially. and differed markedly among rivers. In the Salmonids dominated the fish assemblage in MFJD, patterns were driven by differences in the the WEN (Figure 3C). Juvenile Chinook salmon distribution of juvenile Chinook salmon were relatively abundant throughout the study Oncorhynchus tshawytscha and rainbow trout O. section but were most abundant in the middle mykiss versus mountain whitefish Prosopium reaches of the WEN (rkm 13–25). Rainbow trout williamsoni, catostomids, and cyprinids (Figure were common throughout the study section but 3A). Juvenile Chinook salmon and rainbow trout increased in relative abundance in downstream were both relatively abundant in the middle sec- reaches (rkm 0–8) and in the uppermost reach of tion of the river (rkm 85–100). Rainbow trout the study section (rkm 33–35). Relative abun- increased in relative abundance upstream of rkm dances of bull trout and juvenile Chinook salmon 105, whereas juvenile Chinook salmon were most increased gradually in an upstream direction and common in a single reach downstream of rkm reached a peak at rkm 20–23. Mountain white- 100. Mountain whitefish were relatively abun- fish were relatively abundant throughout the lower dant downstream of the reaches with high rela- 23 km of the WEN but decreased dramatically in tive abundances of Chinook salmon and rainbow relative abundance upstream of rkm 23. Largescale

22torgersen.p65 480 7/28/2006, 10:00 AM Landscape Influences on Longitudinal Patterns of River Fishes 481

A B

Figure 3. Longitudinal patterns of fish distribution in (A) the Middle Fork John Day (MFJD), (B) the North Fork John Day (NFJD), and (C) the Wenaha River (WEN). Trend lines are smoothed values from locally weighted scatterplot smoothing (LOWESS) of near-continuous fish survey data. Dashed horizontal bars below each trend line depict the spatial continuity of fish surveys and provide a relative indicator of the number of data points used to calculate LOWESS regressions. Relative abundance represents a continuum of rare to dominant on a scale of 0 to 1; panels are separated and scaled differently to clarify individual species-abundance patterns. See Figure 2 for definitions of species codes.

sucker and northern pikeminnow occurred Associations between Fish Species and throughout the lower 23 km of the WEN but rep- Longitudinal Patterns of Aquatic Habitat resented a relatively small part of the fish assemblage, except in the lower reaches (rkm 0–3) where Longitudinal patterns of fish distribution corre- largescale sucker were nearly as common as sponded with patterns in aquatic habitat, but mountain whitefish and rainbow trout. these associations were nonlinear and complex

22torgersen.p65 481 7/28/2006, 10:00 AM 482 Torgersen et al.

(Figures 3 and 4). In the MFJD, peaks in the rela- 4A). Peaks in the relative abundance of moun- tive abundance of juvenile Chinook salmon were tain whitefish, bridgelip sucker, speckled dace, associated with peaks in channel gradient and and redside shiner corresponded with the high- troughs in water temperature (Figures 3A and est peak in maximum depth at rkm 83 (Figures

A B

Figure 4. Longitudinal patterns of aquatic habitat in (A) the Middle Fork John Day (MFJD), (B) the North Fork John Day (NFJD), and (C) the Wenaha River (WEN). Trend lines are smoothed values from locally weighted scatterplot smoothing (LOWESS) of spatially continuous (channel gradient and temperature) and near-continuous survey data (depth and water velocity). Dashed horizontal bars below each trend line depict the spatial continuity of habitat surveys and provide a relative indicator of the number of data points used to calculate LOWESS regressions. Water velocity is scaled to 1.0 and represents a continuum of slow- to fast-water aquatic habitats. Mean daily water temperatures were recorded on the day of synoptic surveys with thermal infrared remote sensing.

22torgersen.p65 482 7/28/2006, 10:00 AM Landscape Influences on Longitudinal Patterns of River Fishes 483

3A and 4A). In the NFJD, spatial associations High A gradient between fish distribution and channel morphol- Cool MFJD ogy and water temperature were not as pro-

nounced as they were in the MFJD (Figures 3A, ) 19% 3B, 4A, and 4B). The two highest peaks in the (

relative abundance of juvenile Chinook salmon xis 2 corresponded with the highest peak (rkm 120) A and the lowest trough in water velocity (rkm Low 150). Peaks in the relative abundance of moun- gradient Warm tain whitefish corresponded with peaks in maxi- Axis 1 (73%) Shallow Deep mum depth. In the WEN, peaks in the relative Fast Slow abundance of rainbow trout were associated with high-velocity downstream reaches (rkm 0–8) B Deep and high-gradient reaches upstream (rkm 33– Slow NFJD 35) (Figures 3C and 4C). Peaks in the relative abundance of bull trout and juvenile Chinook ) salmon coincided with a peak in maximum wa- 26% ter depth and a trough in water velocity. ( xis 2 A Multivariate Gradients in Fish Assemblage Structure and Aquatic Habitat Shallow Fast Fishes exhibited distinct differences in assem- Axis 1 (57%) Warm Cold blage structure with respect to habitat variables Low gradient High gradient in the three rivers. Variation in fish assemblage composition in the MFJD corresponded with C habitat gradients in depth, water velocity, chan- Deep Slow WEN nel gradient, and, to a lesser degree, water temperature (Figure 5A and Table 2). The primary ordination axis (depth and water velocity) ex- ) 43% plained 73% of the variation in fish assemblage ( xis 2

structure, and the secondary axis (channel gra- A dient and water temperature) explained 19% of Shallow the variation (P < 0.05). Fishes were strongly seg- Fast regated among shallow riffles (rainbow trout, Axis 1 (46%) juvenile Chinook salmon, mountain whitefish, Cool Cold Low gradient High gradient and speckled dace) and deep pools (redside shiner, bridgelip sucker, northern pikeminnow, Figure 5. Ordination of nonmetric multidimensional and largescale sucker). Fish species most strongly scaling (NMS) analysis of fish assemblage structure in correlated with the primary axis (depth and wa- (A) the Middle Fork John Day (MFJD), (B) the North Fork John Day (NFJD), and (C) the Wenaha River ter velocity) included bridgelip sucker, northern (WEN). Fish species are plotted in ordination space, pikeminnow, largescale sucker, redside shiner, in which each fish outline indicates the position of the and rainbow trout. Species strongly associated species’ centroid with respect to the ordination axes. with the second axis (channel gradient and tem- Solid triangles are sample units in species space. The amount of variation explained by each ordination axis perature) included juvenile Chinook salmon, is shown in parentheses. See Figure 3 for a key to the rainbow trout, and northern pikeminnow (Table fishes.

22torgersen.p65 483 7/28/2006, 10:00 AM 484 Torgersen et al.

Table 2. Pearson correlation coefficients of species and habitat variables versus axis scores from ordinations of fish assemblage structure in entire survey reaches. The surveyed lengths in the Middle Fork John Day, the North Fork John Day, and the Wenaha rivers are 49, 70, and 35 km, respectively. Ordinations were calculated from relative abundance data using nonmetric multidimensional scaling (NMS). The statistical significance of correlations between axis scores and habitat variables, indicated with one or two asterisk symbols (P < 0.05 or P < 0.001, respectively), provides a relative means to compare correlation strength among variables and ordination axes.

Middle Fork North Fork John Day River John Day River Wenaha River (n = 261) (n = 244) (n = 179) Variable Axis 1 Axis 2 Axis 1 Axis 2 Axis 1 Axis 2 Species Bull trout – – 0.30 –0.11 0.22 0.31 Juvenile Chinook salmon –0.12 0.87 –0.24 0.29 0.50 0.76 Rainbow trout –0.69 0.79 0.79 –0.18 0.24 –0.46 Mountain whitefish –0.15 0.11 –0.29 0.71 –0.86 0.40 Northern pikeminnow 0.92 –0.50 –0.42 0.06 –0.29 0.31 Largescale sucker 0.76 0.05 –0.44 0.09 –0.50 0.19 Bridgelip sucker 0.95 –0.39 –0.40 –0.30 – – Redside shiner 0.74 –0.37 –0.54 –0.23 – – Speckled dace –0.51 –0.40 –0.78 –0.42 – – Habitat Temperature 0.28** –0.13* –0.76** 0.00 –0.65** –0.04 Channel gradient –0.16* 0.32** 0.57** –0.02 0.35** –0.07 Maximum depth 0.40** –0.12* –0.22** 0.30** –0.18* 0.35** Mean depth 0.32** -0.05 –0.18** 0.28** –0.23** 0.29** Water velocitya –0.38** 0.04 0.14 –0.22** 0.11 –0.28** a Water velocity is a categorical variable that represents a continuum of slow- to fast-water aquatic habitats.

2). Mountain whitefish and speckled dace occu- cluded rainbow trout, speckled dace, and redside pied intermediate positions. shiner (Table 2). The secondary axis explained Fish assemblages in the NFJD were structured 26% of the variation in fish assemblage structure along gradients of water temperature, channel (P < 0.05) and was associated primarily with mean gradient, and depth (Figure 5B). Temperature and and maximum water depth (Table 2). With the channel gradient were strongly correlated with the exception of bull trout and rainbow trout, fishes primary axis, which explained 57% of the varia- in the NFJD were generally grouped into deep- tion in fish assemblage structure (P < 0.05) (Fig- water (mountain whitefish, largescale sucker, ure 5B and Table 2). The distribution of fish juvenile Chinook salmon, and northern pike- species with respect to the primary ordination minnow) and shallow-water (bridgelip sucker, axis (temperature and channel gradient) indi- speckled dace, and redside shiner) assemblages. cated a separation between coolwater fishes Fishes in the WEN responded to gradients in (redside shiner, largescale sucker, bridgelip temperature, channel gradient, depth, and wa- sucker, northern pikeminnow, and speckled ter velocity (Figure 5C, Table 2). Coldwater fishes dace) and rainbow trout and bull trout. Juvenile (juvenile Chinook salmon, bull trout, and rain- Chinook salmon and mountain whitefish were bow trout) were most abundant in colder, up- positioned at an intermediate location with re- stream reaches, while coolwater fishes (northern spect to the primary ordination axis (tempera- pikeminnow and largescale sucker) were most ture and channel gradient). Fish species most common downstream. Mountain whitefish, strongly correlated with the primary axis in- largescale sucker, and juvenile Chinook salmon

22torgersen.p65 484 7/28/2006, 10:00 AM Landscape Influences on Longitudinal Patterns of River Fishes 485

were strongly correlated with ordination scores on the primary axis, which explained 46% of the variation in the ordination (P < 0.05) (Figure 5C and Table 2). On the secondary ordination axis, depth and water velocity explained 43% of the variation in fish assemblage structure (P < 0.05). Fish species most strongly associated with the secondary axis included juvenile Chinook salmon, rainbow trout, and mountain whitefish (Table 2). Of the three coldwater fishes in the WEN, juvenile Chinook salmon exhibited the strongest positive association with the secondary ordination axis (water depth and slow-water habitats) (Table 2).

Effects of Spatial Extent and Geographical Context on Observed Fish–Habitat Relationships

The observed relationships between fish assemblage structure and aquatic habitat changed depending on the spatial extent and geographical context of analysis (Figure 6). In the MFJD and the NFJD, the relative influences of temperature and channel morphology increased and decreased, respectively, when the spatial extent of the data set was increased. Crossover points in the trends of correlations indicated where temperature and channel morphology explained approximately equal amounts of variation in fish Figure 6. Scale-dependent effects of temperature and assemblage structure (Figure 6). In the MFJD, channel morphology on river fish assemblage struc- the warmest of the rivers, the crossover point ture in the Middle Fork John Day (MFJD), the North occurred in the upper portion of the study sec- Fork John Day (NFJD), and the Wenaha River (WEN). Pearson correlation coefficients (r) indicate the relation (rkm 48), whereas in the NFJD, this transi- tive influence of temperature and channel morphol- tion occurred in the lower 20 km of the study ogy on river fish assemblage structure over a range of section (rkm 115). In the WEN, the coldest and spatial extents. Crossover points indicate the spatial also the shortest river, the relative influence of extent and location where temperature and channel channel morphology on fish assemblage struc- morphology explained approximately equal amounts of variation in river fish assemblage structure. ture increased as the spatial extent of the data set was increased. In the WEN, there was no consistent spatial trend in the relative influence of DISCUSSION temperature on fish assemblage structure, and there was no crossover point at which water Longitudinal patterns of river fish in the MFJD, temperature and channel morphology explained NFJD, and WEN ranged from gradual to abrupt equal amounts of variation in fish assemblage and differed substantially among species and structure. among rivers. At the scale of entire river sections

22torgersen.p65 485 7/28/2006, 10:00 AM 486 Torgersen et al.

(30–70 km), longitudinal patterns of water tem- warmwater fishes in North America (Rahel and perature and channel gradient corresponded to Hubert 1991), but this relationship is difficult to zonation from a coldwater assemblage (Salmon- interpret because elevation, channel gradient, idae) to a coolwater minnow–sucker assemblage and coldwater temperatures are often closely (Cyprinidae–Catostomidae) as predicted by the correlated (Isaak and Hubert 2000; Isaak and river continuum concept and current models of Hubert 2001). However, we were able to isolate river fish distribution (Vannote et al. 1980; Li et the influence of channel gradient on species dis- al. 1987; Rahel and Hubert 1991). However, em- tribution from that of stream temperature. Our bedded within the broad-scale template of cool- observations of fish distribution in the MFJD and and coldwater fish assemblage zones, the distri- NFJD provided a unique opportunity to evalu- bution of fishes was highly variable and reflected ate the response of a coldwater fish (juvenile reach-scale variation in channel morphology Chinook salmon) to relatively high-gradient (i.e., depth and water velocity) and water tem- reaches over a range of water temperatures. For perature. Fish assemblage structure was particu- example, we found that the relative abundance larly variable along the length of the MFJD where patterns of juvenile Chinook salmon in the the longitudinal thermal gradient was not so pro- warmer MFJD and lower NFJD were positively nounced as in the NFJD and the WEN. Although associated with relatively high-gradient reaches. other studies have suggested a high degree of In contrast, in the upper, colder section of the spatial heterogeneity in river fish distribution at NFJD, a peak in the relative abundance of juve- spatial extents of 30–70 km (Stewart et al. 1992; nile Chinook salmon corresponded with a local Roper and Scarnecchia 1994), few studies have trough in channel gradient. Thus, landscape con- described such patterns with spatially continu- text (i.e., geographic and thermal conditions) ous data (Baxter 2002; Torgersen 2002). This reversed the observed relationship between this study provides an example of how spatially con- species and channel gradient. This effect of tinuous data can be collected and analyzed to landscape context was also observed in patterns evaluate landscape influences on longitudinal of fish assemblage structure. Comparisons be- patterns of cool- and coldwater fish assemblages tween assemblage structure in the MFJD and in three Pacific Northwest rivers. the WEN indicated that juvenile Chinook salmon were associated with the shallow, fast- Effects of Landscape Context on water fish assemblage at warm temperatures Fish–Habitat Relationships (MFJD) but were more associated with the deep, slow-water fish assemblage at cold tem- Longitudinal patterns of species and assem- peratures (WEN). blages.—Detailed studies of the spatial distribu- Potential ecological mechanisms.—The poten- tion of fishes within entire river sections (10–100 tial ecological mechanisms underlying this km) are useful for evaluating how fish–habitat changing habitat relationship require further relationships change across scales and in differ- investigation. The apparent reversal in habitat ent spatial contexts (Fausch et al. 1994). River selection by juvenile Chinook salmon may be fish responses to channel gradient provide a case related to species interactions and bioenerget- in point. Within a given river basin, rainbow ics. For example, coolwater fish species, such as trout/steelhead (anadromous rainbow trout) are redside shiner and northern pikeminnow, have generally associated with relatively steeper, a physiological advantage over coldwater fishes swifter habitats than juvenile Chinook salmon in the relatively warmer sections of the MFJD (McMichael and Pearsons 1998; Montgomery et and lower NFJD. Coolwater fishes were the most al. 1999). Similarly, salmonids are usually asso- abundant species in deep, slow-water habitats in ciated with higher channel gradient than the MFJD and the lower NFJD and may exclude

22torgersen.p65 486 7/28/2006, 10:00 AM Landscape Influences on Longitudinal Patterns of River Fishes 487

juvenile Chinook salmon through competition these two variables has not been previously de- with redside shiner (Reeves et al. 1987) or scribed. The quantitative approach that we em- through predation by northern pikeminnow ployed in this study may be useful in river fish (Isaak and Bjornn 1996). Increased riffle use by ecology and management both for understand- salmonids has also been shown to occur in re- ing fish–habitat relationships and for identifying sponse to higher metabolic demands at warmer transitions in fish assemblage structure. water temperatures (Smith and Li 1983) because Transition zones between cool- and coldwater faster current velocities provide higher inverte- fish assemblages were difficult to identify in this brate drift rates and may actually balance out the study because the assemblages overlapped con- increased metabolic costs of maintaining a po- siderably in all three study rivers. However, cross- sition in faster current. over points in the relative influences of The heterogeneous distribution of river fishes temperature and channel morphology on fish we observed in these watersheds may also be in- assemblage structure were useful for identifying fluenced by historical constraints on distribution potential habitat-specific transitions between (geomorphology and biogeographic history), cool- and coldwater fish assemblages. Differences land-use history, and the spawning distribution in the structure of deepwater assemblages in the of adults (Harding et al. 1998; Williams et al. 2003). MFJD, NFJD, and WEN indicated that deep pool All three factors likely play roles in the assemblage environments may be occupied by either cold- patterns of the study rivers, particularly in the or coolwater fishes, depending on water tempera- MFJD and NFJD, which have complex histories ture. Others have observed similar transitions at of land use and have undergone considerable deep pools when continuously electrofishing channel restructuring (Torgersen et al. 1999). large Pacific Northwest rivers (R. M. Hughes, Oregon State University, Corvallis, unpublished Spatial Scale of Observation and data). Without data of such high spatial resolu- Gradients in Fish Assemblage Structure tion and extent, we would not have detected such crossover points in the habitat relationships of The spatial scale of analysis influenced the ob- fish assemblages. served gradients in fish assemblage structure. A The crossover point in the relative influence of number of studies have evaluated changing habi- channel morphology (i.e., depth and velocity) tat relationships of individual fish species across versus water temperature on fish assemblage multiple spatial scales (Fausch et al. 1994; Poizat structure may provide a useful index for assess- and Pont 1996; Torgersen et al. 1999; Baxter and ing and monitoring biological potential for cool- Hauer 2000; Thompson et al. 2001; Torgersen and and coldwater fishes in rivers. In the MFJD and Close 2004). However, investigations of the effects the NFJD, crossover points in assemblage struc- of scale on observed patterns of species diversity ture occurred at 20–22°C (mean daily tempera- and assemblage structure are less common (Wil- ture during the hottest week of the year). This son et al. 1999). This is largely due to the expense temperature range corresponds with the highest of collecting data that are of sufficient resolution mean weekly temperatures recommended for and extent to conduct sequential analyses while coldwater fish species cited by Armour (1991) and varying the spatial dimensions of the data set. The the thermal transition zone recorded by Taniguchi relative roles of temperature and channel mor- et al. (1998) for trout and nontrout assemblages phology in structuring river fish assemblages are in the Rocky Mountains. Because this is the first known to change as the spatial extent and loca- description of such crossover points in fish assem- tion in the drainage are altered (Matthews 1998). blage structure, more examples are needed from However, the specific quantitative relationship a range of rivers over a broader geographic area between spatial scale and the observed effects of to test the application and further develop the

22torgersen.p65 487 7/28/2006, 10:00 AM 488 Torgersen et al.

utility of this approach for examining transitions Schill 1996; Mullner et al. 1998; Bateman et al. in fish assemblage structure. 2005). New sampling methods and statistical approaches, such as hydroacoustics, point abun- Spatially Continuous Analysis dance sampling, and replicate sampling (Barker of Fish–Habitat Relationships and Sauer 1995; Duncan and Kubecka 1996; Cao et al. 2001), are all applicable to surveys of river The response of riverine fishes to habitat het- fishes and represent an area needing more re- erogeneity at intermediate scales is poorly un- search in order to better describe and understand derstood (Fausch et al. 1994, 2002). In part, this the spatial distribution of river fishes. is due to the relative ease of either assessing broad- scale patterns in fish distribution with respect to ACKNOWLEDGMENTS geographic variation in elevation and air temperature (Rahel and Nibbelink 1999) or observ- We are indebted to the generous field support ing fine-scale patterns in fish behavior in provided by numerous technicians and volun- individual pool–riffle sequences and in the labo- teers, including L. Weaver-Baxter, C. Lacey, C. ratory (Reeves et al. 1987; Taniguchi et al. 1998). Krueger, R. Scarlett, S. Robertson, P. Howell, K. In both site-based and laboratory approaches, the Dwire, K. Wright, J. Li, B. Hasebe, P. Jacobs, B. number of samples used to evaluate statistical Landau, K. Hychka, E. Phillips, and S. Hastie. relationships is relatively small. To collect the Additional technical support was provided by the large number of samples necessary for evaluat- Umatilla and Malheur National Forests. We also ing spatially continuous patterns in fish distri- thank the land owners in the Middle Fork John bution, we used snorkeling and relative Day River basin for allowing us to conduct fish- abundance estimates. In many instances, relative eries research on their properties. Thermal re- abundance categories (abundant, common, rare) mote sensing flights were coordinated by R. Faux and presence–absence data are sufficient for with Watershed Sciences, Inc. and Snowy Butte identifying important trends in river fish assem- Helicopters. GIS facilities and consulting were blages (Rahel 1990); however, estimates of rela- provided by K. Christiansen and the Aquatic– tive abundance and presence–absence may be Land Interactions research group at the Pacific unreliable if they are uncorrected for sampling Northwest Research Laboratory, Forestry Sci- efficiency (Bayley and Dowling 1993). Neverthe- ences Laboratory in Corvallis, Oregon. Research less, many mountain rivers are too large to sample funding was provided by the U.S. Environmen- with a backpack electrofishing unit and too small tal Protection Agency/National Science Founda- to sample with boat electrofishing gear (Hughes tion Joint Watershed Research Program et al. 2002; Mebane et al. 2003). These logistical (R82–4774–010 for ecological research) and the challenges make it difficult to validate visual es- Bonneville Power Administration (project No. timates of fish abundance. To compensate for the 88–108 for salmon research). Insightful com- lack of precision in snorkeling surveys, we used ments and constructive criticism from R. Hughes a modified version of point abundance sampling and three anonymous reviewers improved the (Persat and Copp 1990) and found that large focus and quality of the manuscript. numbers of visual estimates of fish abundance were quite effective for quantifying spatial pat- REFERENCES terns in river fish assemblages. Other studies have successfully employed snorkeling and less rigor- Armour, C. L. 1991. Guidance for evaluating and rec- ous electrofishing methods (single-pass) to ommending temperature regimes to protect fish. evaluate patterns of fish distribution in small U.S. Fish and Wildlife Service Biological Report streams (Hankin and Reeves 1988; Thurow and 90(22), Washington, D.C.

22torgersen.p65 488 7/28/2006, 10:00 AM Landscape Influences on Longitudinal Patterns of River Fishes 489

Barker, R. J., and J. R. Sauer. 1995. Statistical aspects Clarke, S. E., and S. A. Bryce. 1997. Hierarchical sub- of point count sampling. Pages 125–130 in C. J. divisions of the Columbia Plateau and Blue Moun- Ralph, J. R. Sauer, and S. Droege, editors. Moni- tains ecoregions, Oregon and Washington. U.S. toring bird populations by point counts. U.S. Forest Forest Service, Portland, Oregon. Service, General Technical Report PSW-GTR-149, Collares-Pereira, M. J., M. F. Magalhaes, A. M. Albany, California. Geraldes, and M. M. Coelho. 1995. Riparian eco- Bateman, D. S., R. E. Gresswell, and C. E. Torgersen. tones and spatial variation of fish assemblages in 2005. Evaluating single-pass catch as a tool for iden- Portuguese lowland streams. Hydrobiologia tifying spatial pattern in fish distribution. Journal 303:93–101. of Freshwater Ecology 20(2):335–345. Cunjak, R. A., R. G. Randall, and E. M. P. Chadwick. Baxter, C. V. 2002. Fish movement and assemblage 1988. Snorkeling versus electrofishing: a compari- dynamics in a Pacific Northwest riverscape. Doc- son of census techniques in Atlantic salmon rivers. toral dissertation. Oregon State University, Corvallis. Naturaliste Canadien 115:89–93. Baxter, C. V., and F. R. Hauer. 2000. Geomorphology, Duncan, A., and J. Kubecka. 1996. Patchiness of lon- hyporheic exchange, and selection of spawning gitudinal fish distributions in a river as revealed habitat by bull trout (Salvelinus confluentus). Ca- by a continuous hydroacoustic survey. ICES Jour- nadian Journal of Fisheries and Aquatic Sciences nal of Marine Science 53:161–165. 57:1470–1481. ESRI (Environmental Systems Research Institute). Bayley, P. B., and D. C. Dowling. 1993. The effect of 1996. ARC/INFO GIS, version 7.0.4. ESRI, habitat in biasing fish abundance and species rich- Redlands, California. ness estimates when using various sampling meth- Fausch, K. D., S. Nakano, and K. Ishigaki. 1994. Dis- ods in streams. Polskie Archiwum Hydrobiologii tribution of two congeneric charrs in streams of 40:5–14. Hokkaido Island, Japan: Considering multiple fac- Belliard, J., P. Boet, and E. Tales. 1997. Regional and tors across scales. Oecologia 100:1–12. longitudinal patterns of fish community structure Fausch, K. D., C. E. Torgersen, C. V. Baxter, and H. W. in the Seine River basin, France. Environmental Li. 2002. Landscapes to riverscapes: bridging the Biology of Fishes 50:133–147. gap between research and conservation of stream Bisson, P. A., J. L. Nielsen, R. A. Palmason, and L. E. fishes. BioScience 52:483–498. Grove. 1982. A system of naming habitat types in Hankin, D. G., and G. H. Reeves. 1988. Estimating to- small streams, with examples of habitat utilization tal fish abundance and total habitat area in small by salmonids during low streamflow. Pages 62–73 streams based on visual estimation methods. Ca- in N. B. Armantrout, editor. Acquisition and utili- nadian Journal of Fisheries and Aquatic Sciences zation of aquatic habitat inventory information. 45:834–844. American Fisheries Society, Bethesda, Maryland. Harding, J. S., E. F. Benfield, P. V. Bolstad, G. S. Helfman, Bult, T. P., R. L. Haedrich, and D. C. Schneider. 1998. and E. D. B. Jones. 1998. Stream biodiversity: the New technique describing spatial scaling and habi- ghost of land use past. Proceedings of the National tat selection in riverine habitats. Regulated Rivers: Academy of Sciences 95:14843–14847. Research & Management 14:107–118. Horwitz, R. J. 1978. Temporal variability patterns and Cao, Y., D. P. Larsen, and R. M. Hughes. 2001. Evalu- the distributional patterns of stream fishes. Eco- ating sampling sufficiency in fish assemblage sur- logical Monographs 48:307–321. veys: a similarity-based approach. Canadian Huet, M. 1959. Profiles and biology of western Journal of Fisheries and Aquatic Sciences 58: European streams as related to fish management. 1782–1793. Transactions of the American Fisheries Society Clarke, K. R. 1993. Non-parametric analyses of 88:155–163. changes in community structure. Australian Jour- Hughes, R. M., and J. R. Gammon. 1987. Longitudi- nal of Ecology 18:117–143. nal changes in fish assemblages and water quality

22torgersen.p65 489 7/28/2006, 10:00 AM 490 Torgersen et al.

in the Willamette River, Oregon. Transactions of McCune, B., and J. B. Grace. 2002. Analysis of eco- the American Fisheries Society 116:196–209. logical communities. MjM Software Design, Hughes, R. M., P. R. Kaufmann, A. T. Herlihy, S. S. Gleneden Beach, Oregon. Intelmann, S. C. Corbett, M. C. Arbogast, and R. McCune, B., and M. J. Mefford. 1999. Multivariate C. Hjort. 2002. Electrofishing distance needed to analysis of ecological data, version 4.27. MjM Soft- estimate fish species richness in raftable Oregon ware, Gleneden Beach, Oregon. rivers. North American Journal of Fisheries Man- McMichael, G. A., and T. N. Pearsons. 1998. Effects of agement 22:1229–1240. wild juvenile spring chinook salmon on growth Isaak, D. J., and T. C. Bjornn. 1996. Movement of and abundance of wild rainbow trout. Transactions northern squawfish in the tailrace of a lower Snake of the American Fisheries Society 127:261–274. River dam relative to the migration of juvenile Mebane, C. A., T. R. Maret, and R. M. Hughes. 2003. anadromous salmonids. Transactions of the Ameri- An index of biological integrity (IBI) for Pacific can Fisheries Society 125:780–793. Northwest rivers. Transactions of the American Isaak, D. J., and W. A. Hubert. 2000. Are trout popula- Fisheries Society 132:239–262. tions affected by reach-scale stream slope? Cana- Montgomery, D. R., E. M. Beamer, G. R. Pess, and T. P. dian Journal of Fisheries and Aquatic Sciences Quinn. 1999. Channel type and salmonid spawn- 57:468–477. ing distribution and abundance. Canadian Jour- Isaak, D. J., and W. A. Hubert. 2001. A hypothesis about nal of Fisheries and Aquatic Sciences 56:377–387. factors that affect maximum summer stream tem- Mullner, S. A., W. A. Hubert, and T. A. Wesche. 1998. peratures across montane landscapes. Journal of Snorkeling as an alternative to depletion electro- the American Water Resources Association 37: fishing for estimating abundance and length-class 351–366. frequencies of trout in small streams. North Ameri- Jongman, R. H. G., C. J. F. ter Braak, and O. F. R. van can Journal of Fisheries Management 18:947–953. Tongeren. 1995. Data analysis in community and Naiman, R. J., H. Decamps, J. Pastor, and C. A. landscape ecology. Cambridge University Press, Johnston. 1988. The potential importance of Cambridge, UK. boundaries to fluvial ecosystems. Journal of the Joyce, M. P., and W. A. Hubert. 2003. Snorkeling as an North American Benthological Society 7:289–306. alternative to depletion electrofishing for assess- Orr, E. L., W. N. Orr, and E. M. Baldwin. 1992. Geol- ing cutthroat trout and brown trout in stream ogy of Oregon. Kendall/Hunt Publishing, pools. Journal of Freshwater Ecology 18:215–222. Dubuque, Iowa. Li, H. W., C. B. Schreck, C. E. Bond, and E. R. Rextad. Paller, M. H. 1994. Relationships between fish assem- 1987. Factors influencing changes in fish assemblage structure and stream order in South Caro- blages of Pacific Northwest streams. Pages 192–202 lina Coastal Plain streams. Transactions of the in W. J. Matthews, and D. C. Heins, editors. Com- American Fisheries Society 123:150–161. munity and evolutionary ecology of North Ameri- Persat, H., and G. H. Copp. 1990. Electric fishing and can stream fishes. University of Oklahoma Press, point abundance sampling for the ichthyology of Norman. large rivers. Pages 197–209 in I. G. Cowx, editor. Loy, W. G., S. Allan, A. R. Buckley, and J. E. Meacham. Developments in electric fishing. Cambridge Uni- 2001. Atlas of Oregon. University of Oregon Press, versity Press, Cambridge, UK. Eugene. Poizat, G., and D. Pont. 1996. Multi-scale approach to Matthews, W. J. 1998. Patterns in freshwater fish ecol- species-habitat relationships: juvenile fish in a large ogy. Chapman and Hall, New York. river section. Freshwater Biology 36:611–622. McCune, B. 1997. Influence of noisy environmental Poole, G. C. 2002. Fluvial landscape ecology: address- data on canonical correspondence analysis. Ecol- ing uniqueness within the river discontinuum. ogy 78:2617–2623. Freshwater Biology 47:641–660.

22torgersen.p65 490 7/28/2006, 10:00 AM Landscape Influences on Longitudinal Patterns of River Fishes 491

Radko, M. A. 1997. Spatially linking basin-wide stream Arbuckle Mountain stream. The American Natu- inventories in a geographic information system. ralist 37:138–147. U.S. Forest Service, General Technical Report INT- Tabachnick, B. G., and L. S. Fidell. 2001. Using multi- GTR-345, Ogden, Utah. variate statistics. Allyn and Bacon, Boston. Rahel, F. J. 1990. The hierarchical nature of commu- Taniguchi, Y., F. J. Rahel, D. C. Novinger, and K. G. nity persistence: a problem of scale. The American Gerow. 1998. Temperature mediation of competi- Naturalist 136:328–344. tive interactions among three fish species that re- Rahel, F. J., and W. A. Hubert. 1991. Fish assemblages place each other along longitudinal stream and habitat gradients in a Rocky Mountain-Great gradients. Canadian Journal of Fisheries and Plains stream: biotic zonation and additive patterns Aquatic Sciences 55:1894–1901. of community change. Transactions of the Ameri- Taylor, C. M., M. R. Winston, and W. J. Matthews. 1996. can Fisheries Society 120:319–332. Temporal variation in tributary and mainstem fish Rahel, F. J., and N. P. Nibbelink. 1999. Spatial patterns assemblages in a Great Plains stream system. in relations among brown trout (Salmo trutta) dis- Copeia 2:280–289. tribution, summer air temperature, and stream size Thompson, A. R., J. T. Petty, and G. D. Grossman. 2001. in Rocky Mountain streams. Canadian Journal of Multi-scale effects of resource patchiness on for- Fisheries and Aquatic Sciences 56:43–51. aging behaviour and habitat use by longnose dace, Reeves, G. H., F. H. Everest, and J. D. Hall. 1987. Inter- Rhinichthys cataractae. Freshwater Biology 46: actions between the redside shiner (Richardsonius 145–160. balteatus) and the steelhead trout (Salmo gairdneri) Thurow, R. F., and D. J. Schill. 1996. Comparison of in western Oregon: the influence of water tempera- day snorkeling, night snorkeling, and electro- ture. Canadian Journal of Fisheries and Aquatic fishing to estimate bull trout abundance and size Sciences 44:1603–1613. structure in a second-order Idaho stream. North Reynolds, L., A. T. Herlihy, S. V. Gregory, and R. M. American Journal of Fisheries Management Hughes. 2003. Electrofishing effort requirements 16:314–323. for assessing species richness and biotic integrity Torgersen, C. E. 2002. A geographical framework for in western Oregon streams. North American Jour- assessing longitudinal patterns in stream habitat nal of Fisheries Management 23:450–461. and fish distribution. Doctoral dissertation. Or- Roper, B. B., and D. L. Scarnecchia. 1994. Summer dis- egon State University, Corvallis. tribution of and habitat use by chinook salmon Torgersen, C. E., and D. A. Close. 2004. Influence of and steelhead within a major basin of the South habitat heterogeneity on the distribution of larval Umpqua River, Oregon. Transactions of the Ameri- Pacific lamprey (Lampetra tridentata) at two spa- can Fisheries Society 123:298–308. tial scales. Freshwater Biology 49:614–630. Sheldon, A. L. 1968. Species diversity and longitudinal Torgersen, C. E., R. N. Faux, B. A. McIntosh, N. J. Poage, succession in stream fishes. Ecology 49:193–198. and D. J. Norton. 2001. Airborne thermal remote Smith, J. J., and H. W. Li. 1983. Energetic factors in- sensing for water temperature assessment in rivers fluencing foraging tactics of juvenile steelhead and streams. Remote Sensing of Environment trout, Salmo gairdneri. Pages 173–180 in D. L. G. 76:386–398. Noakes, editor. Predators and prey in fishes. Dr. Torgersen, C. E., D. M. Price, H. W. Li, and B. A. McIn- W. Junk Publishers, The Hague, The Netherlands. tosh. 1999. Multiscale thermal refugia and stream SPSS. 2001. SigmaPlot, version 8.0. SPSS Science, habitat associations of chinook salmon in north- Chicago. eastern Oregon. Ecological Applications 9:301–319. Stewart, B. G., J. G. Knight, and R. C. Cashner. 1992. Trexler, J. C., and J. Travis. 1993. Nontraditional re- Longitudinal distribution and assemblages of fishes gression analysis. Ecology 74:1629–1637. of Byrd’s Mill Creek, a southern Oklahoma Vannote, R. L., G. W. Minshall, K. W. Cummins, J. R.

22torgersen.p65 491 7/28/2006, 10:00 AM 492 Torgersen et al.

Sedell, and C. E. Cushing. 1980. The river con- Wilson, J. B., J. Steel, W. King, and H. Gitay. 1999. The tinuum concept. Canadian Journal of Fisheries and effect of spatial scale on evenness. Journal of Veg- Aquatic Sciences 37:130–137. etation Science 10:463–468. Wiens, J. A. 1989. Spatial scaling in ecology. Functional Zaroban, D., M. Mulvey, T. Maret, R. Hughes, and G. Ecology 3:385–397. Merritt. 1999. Classification of species attributes Williams, L. R., C. M. Taylor, M. L. Warren, and J. A. for Pacific Northwest freshwater fishes. Northwest Clingenpeel. 2003. Environmental variability, his- Science 73:81–93. torical contingency, and the structure of regional Zubik, R. J., and J. J. Fraley. 1988. Comparison of snor- fish and macroinvertebrate faunas in Ouachita kel and mark-recapture estimates for trout popu- Mountain stream systems. Environmental Biology lations in large streams. North American Journal of Fishes 67:203–216. of Fisheries Management 8:58–62.

22torgersen.p65 492 7/28/2006, 10:00 AM