Fisheries Management and Ecology

Fisheries Management and Ecology, 2009, 16, 202218

Prey abundance, channel structure and the allometry of growth rate potential for juvenile trout
J. S. ROSENFELD
B.C. Ministry of the Environment, Vancouver, BC, Canada

J. TAYLOR
Jacques Whitford Environmental, Burnaby, BC, Canada

Abstract The application of a drift-foraging bioenergetic model to evaluate the relative inuence of prey abundance (invertebrate drift) and habitat (e.g. pool frequency) on habitat quality for young-of-the-year (YOY) and yearling juvenile cutthroat trout, Oncorhynchus clarki (Richardson) is described. Experiments and modelling indicated simultaneous limitation of sh growth by prey abundance and habitat, where depth and current velocity limit the volume of water and prey owing through a shs reactive eld as well as swimming costs and prey capture success. Predicted energy intake and growth increase along a depth gradient, with slower deeper pool habitat generating higher predicted growth for both YOY and yearling trout. Bioenergetic modelling indicated that sh are constrained to use progressively deeper habitats to meet increasing energy requirements as they grow. Sensitivity of growth to prey abundance identied the need to better understand how variation in invertebrate drift and terrestrial drop aects habitat quality and capacity for drift-feeding shes.
KEYWORDS:

bioenergetic modelling, habitat limitation, habitat quality, habitat quantity, invertebrate drift.

Introduction Physical habitat structure and food availability (prey abundance) are key factors affecting habitat use and production of stream shes. Habitat structure inuences metabolic (e.g. swimming) costs, the ability of sh to encounter and capture prey items and vulnerability to predation; prey abundance (encounter rate) directly inuences energy intake that a sh experiences in any given habitat (Hill & Grossman 1993; Grossman, Rincon, Farr & Ratajczak 2002; Hughes, Hayes, Shearer & Young 2003). Habitat structure and prey abundance therefore jointly determine both the quantity of habitat that generates positive growth and survival for an organism across a landscape (i.e. the area of useable habitat) as well as the quality of that habitat (realised growth and survival rates in dierent habitat types). The inuence of physical habitat structure on habitat use by stream shes has received enormous attention and been the focus of innumerable studies. By contrast, few studies have considered the

role of prey abundance in habitat selection (e.g. Hughes 1992; Nislow, Folt & Parrish 2000; Guensch, Hardy & Addley 2001), and fewer still how habitat and food interact to determine the distribution of useable habitats in streams (e.g. Po & Huryn 1998; Thompson, Petty & Grossman 2001). Consequently, there is a need to develop both theory and modelling tools to predict the outcome of environmental impacts that alter habitat structure and prey abundance in streams. Fish production is the outcome of complex interactions between habitat and prey abundance (Grant, Steingrimsson, Keeley & Cunjak 1998; Po & Huryn 1998), and separating their independent eects on habitat quality is dicult. Stream-rearing salmonids are exceptional in the simplicity of their foraging mode feeding on drifting invertebrates while swimming at a xed focal point in the water column. Consequently, bioenergetic and foraging models developed for salmonids can be used to convert the eects of habitat and prey abundance to estimates of energy intake, expenditure and, ultimately, growth. This allows direct

Correspondence: Jordan Rosenfeld, B.C. Ministry of the Environment, 2202 Main Mall, Vancouver, BC, Canada V6T 1Z4 (e-mail: jordan.rosenfeld@gov.bc.ca)

doi: 10.1111/j.1365-2400.2009.00656.x

2009 Crown in the right of Canada.

CHANNEL STRUCTURE AND ALLOMETRY OF TROUT GROWTH

203

quantication of the independent eects of prey abundance (invertebrate drift) and physical habitat (water depth, velocity, channel cross-section conguration) on growth rate potential (where growth rate potential (Brandt, Mason & Patrick 1992) represents densityindependent growth, i.e. the maximum growth rate realised in the absence of competition and predation). However, the suitability of drift-feeding salmonids as a model system to understand how habitat and food jointly control habitat quality depends on how well bioenergetic foraging models predict actual growth (Ney 1993). Fortunately, the ability to measure the growth of individuals experimentally conned to different habitats in natural streams (e.g. Rosenfeld & Boss 2001; Harvey, White & Nakomoto 2005) allows testing of bioenergetic models for drift feeding sh in ways that are extremely dicult for many other taxa (e.g. pelagic shes or birds). The size of the organism further modies the inuence of food and habitat on individual growth through the complex allometry of metabolism and habitat relationships (Peters 1983). Most of these relationships scale non-linearly with organism size, making it dicult to anticipate precisely how habitat quality will change with sh size and prey abundance. However, because the absolute energy requirements of sh (i.e. prey intake required for basal metabolism and growth) increase with size, the subset of habitats that can deliver adequate drift to support growth of larger sh should be limited in small streams and can be expected to decrease as sh grow larger (Rincon & Lobon-Cervia 2002). Similarly, the allometry of growth should result in a proportionally greater growth response of smaller sh to prey enrichment. This study describes and calibrates a bioenergetic model to predict the growth rate of drift-feeding sh [juvenile cutthroat trout, Oncorhynchus clarki (Richardson) and coho salmon, Oncorhynchus kisutch (Walbaum)] by comparing observed and modelled growth of individual sh from experiments in a natural stream and articial channels. Modelling scenarios were then used to assess the effect of independently varying both prey abundance and habitat structure on growth rate potential for juvenile trout in a small stream. Specic objectives were: (1) to determine how quality differs between habitats that represent the fundamental constituents of stream channels (e.g. pools vs rifes); (2) to assess how the extent and quality of useable habitat changes with sh size and whether there are sizerelated thresholds in the ability of sh to exploit habitats; and (3) to determine how physical habitat and prey abundance jointly limit habitat quality in small streams by modelling the independent effects of
2009 Crown in the right of Canada.

increasing pool frequency vs increasing prey abundance on growth rate potential for juvenile trout. Predictions were: (1) that the allometry of increasing energy requirements as sh grow should reduce both the extent and the quality of habitat available to larger drift-feeding sh in small streams (i.e. small trout should be capable of exploiting habitats with lower energy intake than larger trout); (2) that both physical habitat and prey abundance simultaneously limit habitat quality (growth rate potential) in streams; and (3) that increasing prey abundance should permit exploitation of habitats that are bioenergetically unavailable at lower productivities. Methods The structure of the drift-foraging and bioenergetic model is presented below, followed by a description of the experimental growth data used to calibrate it and the modelling scenarios used to assess food and habitat effects on sh growth.
Overview of the bioenergetic model

The bioenergetic model used was a modication of an earlier model for drift-feeding sh (Hughes & Dill 1990; Hughes et al. 2003) that calculates energy intake based on the volume of water owing past the focal point of a sh within its reactive distance to three sizeclasses of invertebrate prey. Energy expenditures were based on the swimming costs experienced by an individual sh at its focal point and the additional costs of leaving the focal point to intercept prey. The net energy left over for growth is simply energy intake less energy losses and expenditures. The model diers from many earlier models by including: (1) a relationship that adjusts prey capture success for water velocity, sh size and distance of prey from the focal point of the sh (but see Van Winkle, Jager, Railsback, Holcomb, Studley & Baldrige 1998; Nislow, Folt & Parrish 1999 and Railsback, Stauer & Harvey 2003 for similar implementations); (2) a scalar that adjusts energy expenditures for increased swimming costs associated with turbulence at higher velocities; and (3) a correction factor to adjust bioenergetic growth estimates for any positive bias associated with increased prey consumption (Bajer, Whitledge, Hayward & Zweifel 2003). These additions improve the accuracy of model predictions, as described in more detail below. Inputs to the bioenergetic model (see Hughes & Dill 1990; Rosenfeld & Boss 2001) are invertebrate drift concentration (mg m)3 dry weight) for three size

204

J. S. ROSENFELD & J. TAYLOR

classes of invertebrate prey (<2.5, 2.55.0, and >5.0 mm), sh mass (g wet weight, for metabolic calculations), water velocity (cm s)1) and depth (cm) at the focal point of the sh (to model swimming costs), and depth and velocity at 60% of total depth at 20-cm intervals along a transect through the focal point of the sh (perpendicular to ow) and along an additional transect 20 cm upstream of the focal point (to model energy intake). Velocity and depth transects were used to estimate discharge owing past the focal point of a sh [water volume per unit time, calculated as the product of cross-sectional area (CA) and velocity (V)] within a shs reactive distance to three size-classes of drifting prey (Hughes & Dill 1990; see Table 1). Three sizeclasses of prey were used because maximum capture distance (and therefore the volume of water scanned by a drift-feeding sh) increases with prey size. A more detailed description of the method for calculating the size of the drift foraging window is presented in Hughes & Dill (1990). Total energy available to a sh at a given focal point was calculated by multiplying total water volume by drift concentration (CONCb, mg m)3) for each size-class of invertebrate. Invertebrate biomass was converted to energy content using a factor of 5200 Cal (21 790 J) g)1 dry weight (Cummins & Wuycheck 1971). Within the reactive distance of a sh, capture success (CS) of prey decreases with increasing current velocity and lateral distance of prey from the focal point of the sh, and increases with sh size (Flore, Keckeis & Schiemer 2001; Grossman et al. 2002). Data from Hill & Grossman (1993) were used to develop piecewise and logistic regression models of capture success using these variables (Table 1; Fig. 1). The piecewise regression was used because it explained a slightly larger proportion of the variance in the original data (R2 = 0.92 for the piecewise regression, R2 = 0.88 for the logistic regression). Energy intake was adjusted for capture success by multiplying the volume of water passing within the reactive distance of a sh by the associated capture success probability (CS, range of 01) for each 20-cm interval on either side of a shs focal point. Energy intake was then expressed as EI R RCAi V CSi CONCb 21790 J g1
i1 3

as described in Hewett & Johnson (1992) with parameter values for coho salmon (Stewart & Ibarra 1991; Table 1). A modied Holling Disc function (after Hughes et al. 2003) was used to model the probability of trout missing prey because of handling eects (i.e. undetected prey drifting past the focal point when trout were intercepting another prey item). To correct for this eect at higher prey densities, energy intake (equation 1 above) for each size class of prey was multiplied by the scalar HD. HD 1=1 ER HTi 3

where ER (encounter rate) is the number of prey per second passing through the reactive window of a sh and HTi is handling time for a sh striking a prey item for each of the three (i) size classes of prey, calculated after Hayes, Stark & Shearer (2000) as the sum of the time required to intercept the prey and the time required to return to the focal point (Table 1). Swimming costs (inclusive of basal metabolism) were calculated as an exponential function of sh weight (g), fork length (cm) and focal velocity according to Hughes & Dill (1990) SC 10CMV 19mass 103 TS 4

0:6 3600 106

Maximum daily consumption (g g)1 d)1) of sh was calculated according to MDC 0:303mass0:275 KA KB 2

based on data for sockeye salmon at 10 C (from Brett & Glass 1973), where C = 2.070.37 log(length) and M = 0.0410.0196 log(length) (Hughes & Dill 1990). This swimming cost function likely underestimates the true cost of swimming in turbulent ow under natural conditions because it is based on forced swimming of sh in laminar ow (Enders, Boisclair & Roy 2003). A turbulence scalar (TS; Table 1, equation 4; Rosenfeld, Leiter, Lindner & Rothman 2005) was included to make swimming cost estimates more realistic by tting a positive exponential function to empirical data from Enders et al. (2003), assuming that variation in velocity at a focal point was equivalent to 33% of the mean focal velocity (the average turbulence used by Enders et al. (2003)). Although sh will make use of velocity refuges to minimise swimming costs and maximise water volume searched (Hayes & Jowett 1994; Railsback & Harvey 2002), the species modelled (juvenile cutthroat trout and coho salmon) tend to occupy focal locations in the water column in relatively slow velocity habitat, rather than holding in velocity refuges closer to the substrate like species adapted to higher velocity habitats [e.g. juvenile Atlantic salmon, Salmo salar L., and steelhead Oncorhynchus mykiss (Walbaum)]. Including the capture success function and a scalar
2009 Crown in the right of Canada.

CHANNEL STRUCTURE AND ALLOMETRY OF TROUT GROWTH

205

Table 1. Relationships and parameter values used in the drift-foraging and bioenergetic model Parameter Energy intake CS Piecewise regression if V < 15: CS = 1.02)(0.00634 V)) (0.00135 T)+(FL 0.00074 V))(V 0.0004 d) if V > 15: CS = 1.04)(0.0131 V))(0.038 d)+ (0.00567 T)+(FL 0.00119 V)) (V 0.00144 d)+(d 0.00478 FL) if V 15 and V 20 then average of functions above Logistic regression eu CS 1eu , where u = 1.28)0.0588 V+0.383 FL)0.0918 (d/RD))0.210 V (d/RD) 12 Prey length (1)e()0.2 FL)) (RD2)(V RD/Vmax)2)0.5 ER = CAi V CONCn HTi = (0.5 MCD/Vmax)+(0.5 MCD/(Vopt+Vfocal)) Estimated from graphical data in Hill & Grossman (1993) Units Value Equation Reference

RD MCD ER HTi CONCn Foraging costs Vmax Vopt Vfocal TS Metabolic costs KA KB L1 L1 G1 G1 CTL CTM CK4 CTO CQ CK1 F U SDA p FA FB FG UA UB UG

cm cm prey s)1 s prey m)3 cm s)1 cm s)1 cm s)1

Hughes & Dill 1990 Hughes & Dill 1990 Hughes 1998 Modied from Hayes et al. 2000

Vmax = 36.23 FL 0.19 Vopt = 17.6 weight 0.05 TS = 10

[(0.06 V))0.98]

Hughes & Dill 1990 Stewart et al. 1983 Rosenfeld et al. 2005 Hewett Hewett Hewett Hewett Hewett Hewett Hewett Hewett Hewett Hewett Hewett Hewett Hewett Hewett Hewett Hewett Hewett Hewett Hewett Hewett Hewett Hewett & & & & & & & & & & & & & & & & & & & & & & Johnson Johnson Johnson Johnson Johnson Johnson Johnson Johnson Johnson Johnson Johnson Johnson Johnson Johnson Johnson Johnson Johnson Johnson Johnson Johnson Johnson Johnson (1992) (1992) (1992) (1992) (1992) (1992) (1992) (1992) (1992) (1992) (1992) (1992) (1992) (1992) (1992) (1992) (1992) (1992) (1992) (1992) (1992) (1992)

+0.90

KA = (CK1 L1)/(1 + CK1 (L1)1)) KB = (CK4 L2)/(1 + CK4 (L2)1)) L1 = e(G1 (T)CQ)) L2 = e(G2 (CTL)T)) G1 = (1/(CTO)CQ)) ln((0.98 (1)CK1))/(0.02 CK1)) G2 = (1/(CTL)CTM)) ln((0.98 (1)CK4))/(0.02 CK4)) 24 18 0.01 15 5 36 F = FA TEMPFB e (FG p) U = UA TEMPUB e (UG p) 0.172 EI/MDC 0.212 )0.222 0.631 0.0314 0.58 )0.299 ED = (386.7PDM))3632 If FL < 6: PDM = 0.18 If FL 6 and FL 10 then PDM = (0.068 + 0.018 FL) If FL > 10 and FL 16 then PDM = (0.171 + 0.007 FL) If FL > 16 then PDM = 0.283

Estimation of growth rate ED J g)1 wet weight PDM %

Hartman & Brandt (1995) Post & Parkinson (2001)

2009 Crown in the right of Canada.

206

J. S. ROSENFELD & J. TAYLOR

100 80 60 Capture success (%) 40 20 100 80 60

7 cm FL

0 10

egestion, excretion, specic dynamic action (Jobling 1994) and swimming costs: NEI GEI1 F 1 U SDA SC Net Energy Intake was converted to growth increment using a generalised energy (J) to biomass (dry weight) conversion relationship for the family Salmonidae (Hartman & Brandt 1995), where energy density (ED) of tissue is an increasing function of percent dry mass (Table 1). As percent dry mass (PDM) of juvenile salmon increases with sh size (smaller sh have a higher water content), percent dry mass of body tissue was estimated using a piecewise regression (Table 1) t to data from Post & Parkinson (2001). Bajer et al. (2003) showed that bioenergetic models may systematically overestimate growth rates as a linear function of consumption. As consumption increases, assimilation eciency should decline (Ursin 1967; Jobling 1994) because of decreased residence time of food in the gut and there should be associated changes in the metabolic costs of specic dynamic action, egestion and excretion (Bajer et al. 2003). Deviations between actual and modelled growth at high consumption arise from errors in estimates of metabolic parameters [possibly because of the limited range of conditions when originally measured in laboratory experiments; (Bajer, Hayward, Whitledge & Zweifel 2004a; Bajer, Whitledge & Hayward 2004b)]. Bajer et al. (2003) recommended regressing growth rate error (EGR = modelled growth)observed growth) on consumption to generate a relationship to correct for this structural error in bioenergetic models when a positive correlation between model error and consumption is known to exist. This method is a form of model tting and should be viewed as a pragmatic approach to correcting model error until more accurate bioenergetic parameter estimates are derived (Bajer et al. 2004b). To determine whether data exhibited a positive relationship between growth rate error and consumption, the error in modelled growth rate (EGR) was regressed against consumption (estimated with the foraging model) for both cutthroat trout and coho. A signicant relationship was found for both species, and therefore used to correct for bias in modelled growth with increasing consumption as suggested by Bajer et al. (2003), i.e. EGR was subtracted from modelled growth to generate predicted growth.
Model calibration

20 30 40 0 10 20

30

40 20 0 0 5 10 15 20 Lateral distance from focal point (cm)

Figure 1. Capture success of rainbow trout feeding on invertebrate drift as a function of lateral distance of prey from the focal point of a sh, for 7 cm and 11 cm FL rainbow trout at current velocities ranging from 040 cm s)1 (based on data from Hill & Grossman 1993).
11 cm FL 40

for swimming costs in turbulent ow greatly reduced bioenergetic estimates of growth, particularly in higher velocity microhabitats; the model produces gross overestimates of growth rate if these functions are absent. Foraging costs were modelled following the method described in Hayes et al. (2000), where costs of swimming at maximum speed (Vmax; Table 1) were experienced for the duration of time required for the sh to intercept the prey item, and the return speed to the focal point was assumed to be the optimal swimming velocity for the sh (Vopt). Maximum speed during prey attack was multiplied by 0.59 since Hughes et al. (2003) observed brown trout striking prey at this average fraction of Vmax. Swimming costs for the duration of a strike were calculated using bioenergetic equations from Elliott (1976) for brown trout, Salmo trutta L. Swimming costs for the attack portion of the strike were multiplied by a factor of ve to account for the underestimation of active swimming costs by forced swimming models (Boisclair & Tang 1993; Hughes & Kelly 1996), as calculated using equation 4. Costs of egestion (F) and excretion (U) were calculated according to equations from Hewett & Johnson (1992, after Elliott 1976) including a correction for the eects of ration size on F and U with parameter values for coho (Table 1). Net Energy Intake (NEI), the energy available for growth, was calculated as gross energy intake less the costs of

The bioenergetic model was calibrated by comparing observed and modelled growth of individual sh from
2009 Crown in the right of Canada.

CHANNEL STRUCTURE AND ALLOMETRY OF TROUT GROWTH

207

two separate experimental data sets, one for juvenile cutthroat trout and the other for juvenile coho salmon. Both species are anadromous, with adults maturing in the ocean while juveniles rear in streams. The trout calibration data set was based on observed growth rates of individual young-of-the-year (YOY, 5.0 cm average FL; n = 8) and yearling and older sh (13.7 cm average FL; n = 14) conned to discrete pool or rife habitats for 2530 days in Husdon Creek on the Sunshine Coast of British Columbia (Rosenfeld & Boss 2001). Husdon Creek is a small (3 m bankfull channel width) coastal stream typical of the habitat where juvenile anadromous cutthroat trout and coho rear (combined juvenile salmonid density of 1.1 sh m)2; Rosenfeld, Porter & Parkinson 2000). Daytime invertebrate drift (referred to as ambient drift in subsequent modelling) was measured in 15 replicate pool and rie habitats over 2 days in the middle of the experiment (Rosenfeld & Boss 2001). Invertebrate drift as well as microhabitat observations on sh focal point depth, velocity and cross-sectional area through the focal point were used as inputs for modelling growth rate potential of individual sh in pool and rie enclosures (see Rosenfeld & Boss (2001) for more detail). This permitted tting of the model to observed growth rates of two size-classes of juvenile trout occupying dierent habitat types (pools and ries) in a natural stream. The juvenile coho salmon calibration data set was based on observed growth rates of dominant sh reared in experimental stream channels at different densities (2, 6 and 12 sh m)2; Rosenfeld et al. 2005). In this experiment, young-of-the-year coho (5.1 cm average FL; n = 12) were conned to experimental stream channels and growth was measured along an experimentally created gradient of natural invertebrate drift abundance (0.0470.99 mg m)3). Only the growth rates of dominant sh were modelled in each channel (n = 12) because of the difculty in estimating drift abundance for subdominants (due to upstream prey consumption). To increase the range of consumption and growth levels over which the model could be validated, growth data from four sub-dominant sh in the highest density treatments that achieved the lowest growth in these experiments were included (average growth of )1.0 0.2% per day; similar to a maximum daily weight loss for juvenile trout of 1% used in an earlier bioenergetics model; Clark & Rose 1997). These sub-dominant sh, which were located downstream of at least six to nine conspecics over a distance of 50 cm, were assumed to have an eective energy intake close to zero. It was also assumed that juvenile cutthroat trout would have similar negative
2009 Crown in the right of Canada.

growth rates at zero energy consumption and these four data points were included when tting the cutthroat trout growth rate errorconsumption relationship. Fish were assumed to forage only during the day, based on observations of sh occupying quiescent microhabitats with low water velocity at night. Nocturnal metabolic costs were therefore calculated assuming a focal velocity of 0 cm s)1 (i.e. basal metabolism only). Water temperature was xed at 10 C for all modelling, close to the average temperature in both experiments. Fit of both models was assessed in terms of the proportion of variance in observed growth rate explained by modelled growth (R2 = SStot)SSerror/SStot). To highlight the potential limitations of correcting for modelled growth rate error (EGR) based on estimated (as opposed to known) consumption, the relationship between EGR and consumption was calculated using: (1) dominant coho from all 12 channels; and (2) a subset of channels (n = 8) that excluded coho in the lowest density treatment. Behavioural observations of solitary juvenile coho in the lowest density treatment (n = 4 channels) indicated that solitary coho foraged more timidly alone than in the presence of conspecics, i.e. generally exhibited greater refuging behaviour and did not strike prey to the limit of their reactive distance. Modelled prey consumption should therefore exceed actual consumption for solitary sh, biasing the model-error consumption regression and causing an underestimate of corrected growth. Model performance was also assessed by comparing predicted growth of YOY at satiation to the expected range of maximum YOY growth rate for similar-sized salmonids ($56% per day, e.g. Post et al. 1999) to evaluate whether the EGR correction with or without solitary sh systematically underestimated predicted growth.
Quality of different habitat types as a function of sh size and prey abundance

Inuence of habitat on growth rate potential. Streams are characterised by discrete habitat types with different depth and velocity distributions (e.g. pools and rifes; Petersen & Rabeni 2001). Changes in growth rate potential along a gradient of increasing habitat depth (e.g. from ries to pools) were evaluated by using habitat-specic velocities and depths as input parameters to the bioenergetic model. As part of an earlier study on hydraulic conditions in Husdon Creek, depth and water velocity (at 60% of total depth) were

208

J. S. ROSENFELD & J. TAYLOR

measured at 20-cm intervals on multiple transects spaced 20 cm apart in ve replicate rie, run, glide and pool channel units in Husdon Creek (equivalent to measuring velocity and depth at the nodes of a 20-cm square grid superimposed on each habitat unit). Velocity and depths were measured at summer low ow; channel unit types were dierentiated as described in Rosenfeld et al. (2000). Channel unit lengths ranged from 1.1 to 9.2 m, with 67435 paired velocity and depth measurements in each channel unit (Table 3). These habitat data and summer low-ow invertebrate drift concentrations from Husdon Creek were used as inputs to the bioenergetic model described above to calculate growth rate potential at each grid point in each replicate channel unit. Growth was modelled separately for YOY (5 cm FL) and yearling (13.7 cm FL) trout. The proportion of channel unit surface area predicted to generate positive growth (hereafter referred to as useable foraging habitat) was calculated, and a mean calculated for each of the four habitat types (n = 5 replicate channel units for each habitat type). Average and maximum growth rate potential were also estimated for each channel unit (total n = 20) within the subset of area predicted to generate positive growth (not for the entire habitat unit, because absolute predictions of energy loss rates in highly unsuitable habitats (e.g. at V = 60 cm s)1) could not be validated). Average and maximum growth rates were then calculated for each habitat type (n = 5 replicate channel units per type). The volume of water that ows past the focal point of a drift-feeding sh is inuenced by both water depth and velocity, so that it is difcult to separate their independent effects on habitat quality. To understand how stream depth affects energy intake independent of velocity, and to test the prediction that smaller sh can obtain sufcient energy for growth in shallower water than larger sh, prey consumption for YOY and yearling trout was modelled along a simulated depth gradient from 10 to 100 cm. Modelling assumptions included a constant current velocity of 10 cm s)1, summer low-ow drift concentration from Husdon Creek, and an innitely wide stream channel (i.e. a foraging window width equal to the full reactive distance of the sh). Inuence of prey abundance on growth rate potential. The effect of prey abundance (a correlate of ecosystem productivity) on habitat quality in different habitat types was assessed by modelling growth rate potential for 5 cm and 13.7 cm juvenile trout over a gradient of increasing drift concentrations

set at 0.25, 0.5, 1, 2, 3, 4 and 5 times the ambient baseline drift measured in Husdon Creek at summer low ow during our growth rate experiment, as described above. Joint effects of percent pool and prey abundance on reach-scale habitat quality. To determine the relative effects on habitat quality of simultaneously changing habitat and prey abundance at a scale larger than a single channel unit, virtual stream reaches 20 channel units long were assembled by randomly selecting pools, rifes, runs and glides from the ve replicates of each that were measured in Husdon Creek. Habitat units were sampled in frequencies that generated pool habitat ranging from 10, 25, 40, 55 and 70% of channel area at drift concentrations of 0.25, 0.5, 1, 2, 3, 4 and 5 times the ambient measured in Husdon Creek at summer low ow. The relative proportions of habitats other than pools were xed at the natural frequencies observed in Husdon Creek (glide:run:rife ratio of 3:1:8). One thousand randomly assembled reaches (average length 77 m, average wetted width 1.9 m) were generated for each percent pool and drift combination for both 5 and 13.7 cm FL juvenile trout. To assess trends in habitat quality along the drift (productivity) gradient, reach average growth rate potential in useable foraging habitat and percent of habitat area with positive growth were plotted against drift concentration for each pool frequency. The independence of habitat and prey abundance effects on habitat quality was determined by testing for an interaction between percent pool and drift concentration. Results
Correction of growth for consumption

Consistent with the observation of Bajer et al. (2003), there was a signicant positive relationship between modelled growth rate error and consumption for both juvenile cutthroat trout (P < 0.002, F1,6 = 25.4) and coho salmon (P < 0.0001, F1,11 = 237 for all 12 channels, P < 0.002, F1,7 = 29.2 with low density channels removed; Fig. 2, Table 2). Slopes of the growth rate errorconsumption relationships (Fig. 2, Table 2) were similar to those described for perch by Bajer et al. (2003), as was the observation that the bioenergetic model systematically overestimated growth at high consumption and underestimated growth at low consumption. However, growth rate error appeared to be negligible at very low prey consumption, i.e. the model accurately predicted the
2009 Crown in the right of Canada.

CHANNEL STRUCTURE AND ALLOMETRY OF TROUT GROWTH

209

14 Growth rate error (%day1) 12 10 8 6 4 2 0 2 0 5 10 15 20 25 Mean daily consumption (%day1)

Figure 2. Relationship between growth rate error (percent of body weight per day) and mean daily consumption for uncorrected bioenergetic model predictions for juvenile cutthroat trout (black line; black circles YOY; grey circles yearling sh) and coho salmon (open diamonds. Solid grey line uses data for dominant sh from all stream channels (n = 12), broken line is with solitary sh excluded (n = 8); see Methods for details).

negative growth of sub-dominant coho with extremely low energy intake. Consequently, the growth rate error data exhibited a curvilinear relationship close to zero consumption; i.e. linear ts to the data would result in the unrealistic prediction of positive growth at zero consumption. A polynomial regression was therefore tted to growth rate error at low consumptions (Fig. 2; Table 3). Model error was signicantly larger (more negative) for YOY trout than for yearling and older juveniles (P < 0.0001, F1,19 = 33.7). There was also no signicant relationship between growth rate error and consumption for yearling and older juvenile trout (P < 0.13, F1,12 = 2.7; Fig. 2). The minor dierence in mean predicted and observed growth (0.196%, n = 14) was therefore used to model error correction for yearling trout. Including solitary sh from the low density channel treatment increased the slope of the growth rate error regression for coho (Fig. 2), indicating that the foraging model overestimated consumption for solitary sh and that predicted growth rates using an EGR correction are sensitive to error in estimated consumption rates.

Table 2. Growth rate errorconsumption relationships observed in this study and Bajer et al. (2003) Weight range (g) 0.62.6 Growth rate errorconsumption relationship (%%) If MDC 2% EGR = 0.538 MDC)1.17a,b EGR = 0.460 MDC)0.86c If MDC < 2% EGR = )0.70 MDC + 0.372 MDC2)0.027 MDC3)0.025 If MDC 2.5% EGR = 0.487 MDC)2.55 If MDC < 2.5% EGR = )1.15 MDC + 29.46 MDC2)183.9 MDC3 + 0.00023 EGR = )0.00196 EGR = 1.05 MDC)1.22 EGR = 0.61 MDC)0.88 R2

Species Coho salmon

Study This study (and Rosenfeld et al. 2005)

0.96 0.83

Cutthroat trout

1.22.1

0.71

This study (and Rosenfeld & Boss 2001)

Perch (Wisconsin model) Perch (KarasThoresson model)

a

1641 1830 1830

0.84 0.72

Bajer et al. 2003 Bajer et al. 2003

All dominant sh (n = 12). EGR and MDC in %. c Solitary sh excluded (n = 8).

b

Table 3. Average habitat characteristics (n = 5 replicates) of the channel unit types measured in Husdon Creek and used to assemble reaches of dierent pool frequency Mean velocity (cm s)1) 5.7 12.7 15.2 19.5 Maximum velocity (cm s)1) 26.0 34.0 42.4 66.2 Grid point count 238 156 86 211

Habitat Pool Glide Run Rie

Mean area (m2) 8.7 5.7 2.97 7.7

Mean depth (cm) 19.1 12.7 10.7 5.7

Maximum depth (cm) 43.6 24.8 18.8 14.6

Maximum width (m) 3.1 2.0 1.5 2.3

Length (m) 3.6 3.7 2.7 4.3

2009 Crown in the right of Canada.

210

J. S. ROSENFELD & J. TAYLOR

4.5
(a)

10
(a)

3.5 2.5 1.5 Daily growth (%) 0.5 0.5 1.0 3.5 2.5 1.5
(b)

8 6 4

Growth (% day1)

2 0 4
(b)

3 2 1

0.5 0 0.5 1.5 0 5 10 Fish length (cm) 15 1 0 0.4 0.8 Total drift concentration (mgm3) 1.2

Figure 3. Observed growth (open circles) and model estimates of daily growth (lled circles) for YOY (average 5 cm FL) and yearling (average 13.7 cm FL) coastal cutthroat trout from Husdon Creek (data from Rosenfeld & Boss 2001). Panel (a) shows unadjusted bioenergetic model predictions; panel (b) shows bioenergetic model predictions adjusted for systematic error associated with consumption.

Figure 4. Observed growth (open circles) and bioenergetic model estimates of growth (lled circles) for juvenile coho salmon trout feeding over a gradient of natural drift abundance in experimental stream channels (data from Rosenfeld et al. 2005). Panel (a) shows unadjusted bioenergetic model predictions; panel (b) shows bioenergetic model predictions adjusted for systematic error associated with consumption.

Evaluation of model performance

The bioenergetic model without any form of model tting explained 63% of the variance in growth of juvenile cutthroat trout observed in Husdon Creek enclosures (for YOY and yearling trout combined, P < 0.0001, F1,20 = 64; Fig. 3a). When growth rate was corrected for error associated with consumption (using parameters and equations listed in Table 2), the model t improved considerably (R2 = 0.90, P < 0.0001, F1,20 = 193; Fig. 3b). The slope of the regression of observed vs modelled growth was not signicantly dierent from one (0.954 0.069 SE) and the intercept was not signicantly dierent from zero (0.00028 0.00095 SE), indicating minimal model bias. Coho growth modelled over a larger range of prey abundance did not t the observed data as well as for trout (Fig. 4a), although correcting growth rate for consumption improved the model t (Fig. 4b; R2 = 0.42, P = 0.017, F1,10 = 8.2; R2 = 0.53, P = 0.009, F1,9 = 10.9 with the single negative growth outlier point in Fig. 4b removed). The slope and intercept of the regression of observed on modelled growth also approximated one and zero respec-

tively (0.80 0.24 SE for slope; 0.36 0.33 SE for intercept). Another criteria for effective model performance, and in particular application of the EGR correction, was whether maximum predicted growth rates were realistic at prey availabilities greatly in excess of the calibration data (e.g. ambient drift in Husdon Creek). Maximum bioenergetic model predictions for YOY growth at satiation were 5.8% for cutthroat trout at a fork length of 5 cm in model simulations where invertebrate drift was increased to very high concentrations (i.e. satiation). This is in the expected range for maximum growth of YOY salmonids observed in the wild (e.g. 56%; Post et al. 1999). Maximum bioenergetic model predictions of YOY growth at satiation for 5-cm coho were 3.1% with the EGR correction including solitary sh and 4.7% using the EGR correction excluded solitary sh; this is consistent with the expectation of lower predicted growth rates with an EGR correction based on overestimated consumption.
Inuence of habitat on growth rate potential

For smaller sh, simulated energy intake along a gradient of increasing depth asymptoted at approximately 60 cm (Fig. 5), as the increase in water volume
2009 Crown in the right of Canada.

CHANNEL STRUCTURE AND ALLOMETRY OF TROUT GROWTH

211

300 Gross energy Intake (Jh1)

Growth rate potential (%day1)

6
(a) (b)

250 200 150 100 50

5 4 3 2 1 0

0 0 10 20 30 40 50 60 70 80 90 100 Depth (cm)

Proportion of habitat with positive growth rate potential

(c)

(d)

0.8 0.6 0.4 0.2 0

0 1 2 3 4 5 1 2 3 4 5

Figure 5. Modelled estimates of gross energy intake (J h)1) for YOY (lled circles) and yearling (open circles) cutthroat trout across a gradient of increasing channel depth at a xed current velocity of 10 cm s)1.

scanned for drift becomes limited by reactive distance to prey rather than stream depth. By contrast, energy intake for larger trout continued to increase with depth because of their greater reactive distance. Yearling trout were predicted to have a threshold depth of 17 cm below which energy intake was insufcient for positive growth at ambient drift concentrations. The distribution of growth rate potential in Husdon Creek showed a similar pattern, with the highest average and maximum predicted growth rate potential in pool habitats and the lowest in rifes for both sizes of trout (Fig. 6a, b); the proportion of habitat with positive growth rate potential was also highest in pools (Fig. 6c, d). Similarly, the proportion of a channel unit with positive growth and average growth rate potential both were increasing functions of average channel unit depth for both YOY and older sh (Fig. 7). Young-ofthe-year were predicted to be able to use approximately 90% of a channel unit area at average depths in excess of 1520 cm (Fig. 7). For yearling trout, there appears to be a threshold average channel unit depth of 7.5 cm below which there are negligible microhabitats suitable for growth, with the proportion of suitable habitat increasing linearly with depth beyond this threshold (Fig. 7).
Inuence of increasing prey abundance on growth rate potential

Drift concentration (as multiples of ambient)

Figure 6. Average and maximum modelled growth rate potential for YOY (open circles) and yearling (lled circles) cutthroat trout in (a) pool and (b) rie channel units along a gradient of drift concentrations (expressed as multiples of the ambient concentration measured in Husdon Creek at summer low-ow). Each point represents modelled mean values from 5 replicate channel units. The lower two panels (c and d) illustrate increases in the average proportion of dierent habitat types (pools black circles; glides white circles; runs grey circles; ries open diamonds) that generate positive growth rate potential for (c) YOY or (d) yearling trout along a productivity gradient of increasing drift.

The extent of habitat predicted to generate positive growth was sensitive to prey abundance. As prey abundance increased, absolute increases in the propor 2009 Crown in the right of Canada.

tion of habitat suitable for growth tended to be higher for yearling than YOY trout because more habitat was suitable for YOY at all productivities (Fig. 6). The rate of change in suitable habitat was steepest at drift concentrations at or below the ambient observed in Husdon Creek (i.e. the baseline level observed at summer low ow), indicating that the extent of suitable habitat is most sensitive to incremental changes in drift at low productivities. The proportion of rie habitat suitable for growth of yearling trout was negligible at or below ambient drift concentrations (Fig. 6d), indicating that productivity-related thresholds are rst reached in shallower habitats and progressively deeper habitats become energetic sinks as prey abundance declines. Consequently, most of the useable habitat for yearling trout is in pools at lower drift levels (Fig. 6d), consistent with empirical observations (e.g. Rosenfeld & Boss 2001).

212

J. S. ROSENFELD & J. TAYLOR

Average positive growth rate potential (%day1)

1
Proportion of habitat with positive growth rate potential
(a)

3.0

YOY trout (5 cm)

0.8 0.6 0.4 0.2 0

(b)

2.0

1.0

0.0 1+ trout (13.7 cm) 0.5 0.0 0 1 2 3 4 5

1.0

0.0

Growth rate potential (%day1)

2 1.5 1 0.5 0 0 5 10 15 20 25 Mean habitat unit depth (cm) 30

Drift concentration (as multiples of ambient)

Figure 8. Reach-average growth rate potential for virtual reaches of randomly selected channel units varying in proportion of pool habitat over a gradient of drift abundance (expressed as multiples of summer low-ow drift concentrations in Husdon Creek) for YOY and yearling trout. Solid circles are 10% pool, open circles are 25% pool, lled squares and solid line are 40% pool (the average in Husdon Creek), diamonds are 55% pool and squares are 70% pool habitat. Error bars are omitted for clarity. The horizontal dotted line indicates the equivalent eects on average growth rate potential (in useable habitat) of tripling prey abundance in a reach with 10% pool vs increasing percent pool from 10 to 65% in a stream with ambient drift levels.

Figure 7. Proportion of habitat predicted to have positive growth rate potential (a) and average channel unit growth rate potential (b) as a function of mean channel unit depth for YOY (gray lines) and yearling trout (black lines). Habitats in order of increasing depth are ries (open circles), runs (grey squares), glides (open triangles) and pools (black circles).

Modelled growth rate potential increased steeply with elevated drift in both pools and rifes. Growth rates of 13.7 cm FL trout were a linear function of drift concentration (Fig. 6) because yearling trout were below satiation at all drift levels. Maximum growth rates of YOY plateaued at 5.8% per day at a drift concentration of three to four times ambient in pool habitat, although mean growth continued to increase along the productivity gradient in both pools and ries. Growth rate also appeared to be most sensitive to incremental increases in prey abundance at very low productivities. The steepest change in modelled growth for YOY and yearling trout was at low drift concentrations, particularly in rie habitats, which were predicted to be unusable for yearling trout and marginal for YOY growth at drift concentrations below ambient.
Joint effects of pool frequency and prey abundance on reach-scale habitat quality

Simulations of reach-scale habitat quality showed that increasing either the frequency of pools or drift

concentrations lead to systematic increases in average growth rate potential (Fig. 8) and useable foraging habitat (Fig. 9). Dierent combinations of pool frequency and prey abundance are predicted to generate similar reach-average habitat qualities. For example, if a channel with only 10% pool habitat is interpreted as degraded (Montgomery, Bungton, Smith, Schmidt & Pess 1995; Johnston & Slaney 1996), then increasing either pool frequency to 65% of wetted area or tripling prey abundance will achieve similar increases in reach-average growth rate potential within useable habitat (Fig. 8, dotted line). However, eects are not entirely substitutable, as increases in pool area generally result in a greater incremental increase in useable habitat than increases in prey abundance. Incremental effects of prey enrichment on reachscale habitat quality are also greatest at low productivities (Figs 8 & 9). Increases in relative growth rate are generally higher for YOY than yearling trout, although the proportion of useable habitat plateaued with increasing prey abundance for YOY but not older sh (Fig. 9). A signicant positive interaction between percent pool and drift concentration on growth rate potential (F1,34 = 17.8, P < 0.0002 for YOY, F1,26 = 63.1, P < 0.0001 for 15 cm trout) indicated that that the effects of habitat are synergistic with food
2009 Crown in the right of Canada.

CHANNEL STRUCTURE AND ALLOMETRY OF TROUT GROWTH

213

0.8 0.6

YOY trout (5 cm)

0.4 0.2 0 0.8 0.6 0.4 0.2 0 0 1 2 3 4 5

1+ trout (13.7 cm)

Drift concentration (as multiples of ambient)

Figure 9. Proportion of reach area with positive growth rate potential for YOY and yearling trout over a gradient of drift concentration for randomly assembled reaches varying in the proportion of pool habitat. Symbol values as indicated in Figure 8 caption.

resources, i.e. effects of increasing prey are greatest when pools are abundant. Discussion
Allometry and habitat quality for drift-feeding sh

Experiments and modelling demonstrated that physical habitat and prey abundance jointly limit growth of drift-feeding sh through their concurrent effects on energy intake. Drift concentration directly inuences prey encounter rate, while the velocity and depth prole through the focal point of a sh determines capture success and the volume of water searched for prey. Changes in either habitat structure or drift concentration can therefore be expected to simultaneously inuence capacity of individual habitats or entire streams. Contrasting stream habitat types can be characterised in terms of both habitat quality (realised growth rate potential in useable habitat) and habitat quantity (area of habitat with positive growth rate potential, i.e. useable foraging habitat for drift-feeding). Consistent with expectations, average modelled growth rate
2009 Crown in the right of Canada.

potential in the constituent habitat types of Husdon Creek (pools, glides, runs and rifes) at low ow was an increasing function of channel unit depth, as was the proportion of each habitat type with positive growth. This is consistent with numerous observations of habitat use by drift-feeding shes (e.g. Heggenes, Northcote & Peter 1991), and indicates that higher predicted energy intake in deeper habitats is sucient to account for dierential use of pools, irrespective of other demonstrated functions of pool habitat, such as refuge from predation (Power 1984; Lonzarich & Quinn 1995) or scouring ows (Quinn & Peterson 1996; Solazzi, Nickleson, Johnson & Rodgers 2000). Modelling indicates that allometric changes in absolute energy requirement also impose size-specic constraints on habitat use. Smaller energy requirements allow YOY sh to achieve positive growth in shallow habitats that provide inadequate ration to support larger trout at summer low ow, while growth of yearling cutthroat under summer low ow conditions was possible only in deeper pool habitat. Although the results apply most directly to juvenile salmonids, the general limitation of energy intake by channel dimensions provides a clear mechanism not only for preference of pool habitat but also for downstream migration to deeper habitat as sh grow. Based on optimal foraging theory, Bachman (1982) (building on work by Kerr 1971) proposed that drift abundance would limit sh growth and maximum size. While bioenergetic modelling demonstrated this to be true, energy intake and sh growth was also shown to be limited by physical habitat dimensions when channel depth is less than the reactive distance of a sh a common circumstance in small streams where maximum reactive distances range from 4580 cm, depending on sh and prey size (Hughes & Dill 1990). Maintaining growth once channel dimensions constrain the drift-foraging window therefore requires migration to deeper habitat, habitat with a higher drift concentration, or habitat that supports a shift to piscivory (Keeley & Grant 2001). The allometry of energy requirement and swimming performance should also affect longitudinal trends in habitat use along the River Continuum that result in peak abundance of juvenile salmon in lower-order streams (Mundie 1974; Rosenfeld et al. 2000). Small physically complex streams in low-gradient landscapes have lower velocities, so that a relatively high proportion of habitat is within the performance capacity of small sh while delivering sucient energy for growth. As streams increase in average depth and velocity along a downstream continuum (Leopold & Maddock 1953; Rosenfeld, Post, Robins & Hateld 2007), a

Proportion of habitat with positive GRP

214

J. S. ROSENFELD & J. TAYLOR

relatively higher proportion of habitat will be outside of the performance capacity of small sh (e.g. YOY), but a higher proportion should deliver sucient prey for growth of larger trout. As channel size and velocity increase further, the relative proportion of habitat within the performance range of both YOY and yearling trout should decline, leading to the expectation that channel-average densities of juvenile salmonids should peak in smaller streams (Murphy, Heifetz, Johnson, Koski & Thedinga 1986; Rosenfeld et al. 2000; Rosenfeld et al. 2007).
Effects of prey enrichment on habitat quality

Habitat use by drift-feeding sh is driven by the allometry of swimming abilities as much as energy requirement. Area of suitable drift-foraging habitat for YOY did not appreciably increase beyond a doubling of ambient drift concentration (Fig. 9), indicating that as productivity increases, the distribution of suitable habitat for YOY sh becomes rapidly limited by swimming performance (inability to use fast water habitat), rather than energy intake. Further response to enrichment is through increased growth rate potential in suitable habitat with minimal increase in area occupied. By contrast, both growth rate potential and area of usable habitat for larger trout continue to increase along a drift concentration gradient, indicating that distribution of older juvenile trout in Husdon Creek was primarily limited by energy intake rather than swimming performance. However, increases in mass-specic growth rates were much higher for YOY suggesting that smaller sh are more likely to benet from prey enrichment in small streams (unless there is signicant piscivory by yearling trout). Modelling also indicated that incremental changes in prey abundance should have the largest impact in oligotrophic streams. Similarly, as deep habitats are the only ones that generate positive growth at low productivity, incremental loss or restoration of deeper pool habitat should also have the largest relative effect in oligotrophic streams. Implications for wild populations are that growth of juvenile salmonids will be highly sensitive to natural seasonal and spatial variation in prey abundance as well as loss of deeper pool habitat, particularly in low productivity systems (cf. Grant et al. 1998). Simulations suggested that similar target levels of habitat quality can be achieved by increasing either pool frequency or invertebrate drift. Implications for habitat restoration are that increases in pool frequency or prey abundance may be substitutable in terms of

their effects on habitat quality. While increased prey alone permits exploitation of habitats that are bioenergetically unavailable at lower productivities, simultaneously increasing drift and pool frequency is predicted to be most effective at improving habitat quality, i.e. effects are predicted to be synergistic rather than additive. However, the cost of instream rehabilitation measures (e.g. log or boulder placement to increase pool frequency) may be much higher than nutrient or salmon carcass addition, so enrichment may be an appealing approach if funds are limited. This inference must be tempered by consideration of the additional functions performed by pools (e.g. cover from predators, hydraulic refuges) that are not substitutable through enrichment of shallow habitats. Degraded channel structure (e.g. absence of large woody debris and associated pools) will continue to depress habitat quality under enrichment and recovery of natural riparian and watershed processes to increase LWD inputs and natural pool-forming processes is essential for restoring pre-impact capacity (Beechie & Bolton 1999).
Model evaluation

Inclusion of key ecological processes absent in many models resulted in more accurate growth estimates. The most important of these is the capture success function (derived from Hill & Grossman 1993), which accounted for variation in prey capture as a function of current velocity, sh size and distance of prey from the focal point of the sh. This function greatly reduced modelled energy intake, which is otherwise overestimated by a factor of two or more if sh are assumed to consume all prey within their reactive distance (Hughes et al. 2003; Piccolo, Hughes & Bryant 2008a,b). Grossman et al. (2002) made a strong case that the primary ecological factor limiting exploitation of fast-water habitats was decreasing prey capture success at higher water velocities, rather than increased swimming costs. However, swimming costs may rise more quickly with velocity than previously thought because forced-swimming respiration models that assume laminar ow underestimate the true costs of swimming in turbulent water (Enders et al. 2003; but see Nikora, Aberle, Biggs, Jowett & Sykes 2003 for contrasting evidence). Regardless, inclusion of both capture success functions (e.g. Van Winkle et al. 1998; Nislow et al. 2000; Railsback & Harvey 2002) and realistic swimming cost relationships are essential to accurately model dierential changes in energy intake and expenditure along velocity gradients (Piccolo et al. 2008a).
2009 Crown in the right of Canada.

CHANNEL STRUCTURE AND ALLOMETRY OF TROUT GROWTH

215

The second key feature that contributed to a better t was correction for model error associated with prey consumption as recommended by Bajer et al. (2003). Systematic errors in estimated consumption from the foraging submodel (from inaccuracies in the foraging model or error in estimated drift biomass) may also have inuenced either the slope or intercept of the EGRconsumption relationship. However, the general similarity of relationships from this study to those observed by Bajer et al. (2003) suggested that the EGR consumption relationship is primarily driven by errors in metabolic parameter estimates rather than errors in estimated consumption, and that the foraging model does a reasonably good job of representing energy intake and how it changes with habitat. Nevertheless, inclusion of the model errorconsumption correction factor represents a form of model tting and the EGR correction applied in this study should not be viewed as directly transferable to other species or populations in different streams. The application of bioenergetics to evaluate habitat quality as presented in this study needs to be tempered by several additional considerations. First, the effects of predation risk were ignored. Predation risk alters habitat use because sh trade off predation risk with the need to grow sufciently to survive and reproduce (e.g. Fraser & Cerri 1982; Harvey 1991). Although predation risk aects habitat quality, it remains useful rst to understand the eect of habitat on growth rate independent of predation risk. Predation risk can then be treated as an additional habitat attribute that further modies habitat quality based on survival, which can be combined with growth rate potential to generate an index of habitat quality that better reects tness [e.g. the ratio of mortality risk to growth (Gilliam & Fraser 1987; Bradford & Higgins 2001) or the probability of a sh surviving to reproduce (Railsback & Harvey 2002)]. Second, habitat quality was modelled as densityindependent growth rate potential, i.e. the growth that a sh would realise in the absence of competition. Because drift-feeding sh can locally deplete drifting prey (Leung, Rosenfeld & Bernhardt 2009), this likely overestimates growth in the presence of conspecics (Hughes 1992). Although accounting for local prey depletion would reduce estimates of growth rate potential, it would be unlikely to alter the general conclusions or the relative ranking of habitats in terms of quality. However, habitat capacity models based on drift-foraging bioenergetics need to account for depletion and diusion dynamics of invertebrate drift (e.g. Hughes 1992) to realistically predict sh abundance (e.g. Railsback & Harvey 2002; Hayes, Hughes & Kelly 2007).
2009 Crown in the right of Canada.

Finally, it is implicitly assumed that drift concentration is independent of channel structure (e.g. the proportion of pool or rife habitat in a reach). Changing the relative proportion of pool and rife habitat in a stream channel (as a modelling exercise, or through habitat restoration) will affect reach-scale production of invertebrate drift as well as the distribution of available habitat for sh. How the length, frequency and juxtaposition of different channel units affect the production and depletion of invertebrate prey by drift-feeding sh is an area that demands further consideration to assess their relevance to instream production and restoration goals (Mundie 1974; Po & Huryn 1998; Leung et al. 2009). Despite these caveats, bioenergetic modelling is a viable approach for assessing habitat quality for driftfeeding sh (Nislow et al. 2000; Railsback & Harvey 2002; Hayes et al. 2007). In combination with driftforaging models, bioenergetics provide an approach for predicting the independent eects of prey abundance and habitat structure on sh growth and highlights the need to better understand how drift abundance varies through space and time to inuence productive capacity. Bioenergetics also provides a framework for understanding how the allometry of energy requirements and metabolic costs, superimposed on the physical habitat template in streams, jointly determine the distribution of useable habitat for drift-feeding sh. Acknowledgments We would like to thank Shelly Boss, Taja Lee, Thomas Leiter, Gerhard Lindner, Bob Little, Brent Matsuda, Marc Porter and Michelle Roberge for assistance with data collection and Dave Bates for help with site selection and logistics. We also thank Cliff Kraft, Eric Parkinson and Tom Johnston for reviewing an earlier version of this manuscript. Funding for this research was partially provided by Forest Renewal B.C., the Habitat Conservation Trust Fund and Bugs Unlimited. A spreadsheet version of the bioenergetic model is available on request from JSR. References
Bachman R.A. (1982) A growth model for drift-feeding salmonids: a selective pressure for migration. In: E.L. Brannon & E.O. Salo (eds) Proceedings of the Salmon and Trout Migratory Behaviour Symposium. Seattle, WA: University of Washington Press, pp. 128135. Bajer P.G., Whitledge G.W., Hayward R.S. & Zweifel R.D. (2003) Laboratory evaluation of two bioenergetics models

216

J. S. ROSENFELD & J. TAYLOR

applied to yellow perch: identication of a major source of systematic error. Journal of Fish Biology 62, 436454. Bajer P.G., Hayward R.S., Whitledge G.W. & Zweifel R.D. (2004a) Simultaneous identication and correction of systematic error in bioenergetics models: demonstration with a white crappie (Pomoxis annularis) model. Canadian Journal of Fisheries and Aquatic Sciences 61, 21682182. Bajer P.G., Whitledge G.W. & Hayward R.S. (2004b) Widespread consumption-dependent systematic error in sh bioenergetics models and its implications. Canadian Journal of Fisheries and Aquatic Sciences 61, 2158 2167. Beechie T.J. & Bolton S. (1999) An approach to restoring salmonid habitat-forming processes in Pacic Northwest watersheds. Fisheries 24(4), 615. Boisclair D. & Tang M. (1993) Empirical analysis of the inuence of swimming pattern on the net energetic cost of swimming in shes. Journal of Fish Biology 42, 169183. Bradford M.J. & Higgins P.S. (2001) Habitat-, season-, and size-specic variation in diel activity patterns of juvenile chinook salmon (Oncorhynchus tshawytscha) and steelhead trout (Oncorhynchus mykiss). Canadian Journal of Fisheries and Aquatic Sciences 58, 365374. Brandt S.B., Mason D.M. & Patrick E.V. (1992) Spatiallyexplicit models of sh growth rate. Fisheries 17(2), 2335. Brett J.R. & Glass N.R. (1973) Metabolic rates and critical swimming speeds of sockeye salmon (Oncorhynchus nerka) in relation to size and temperature. Journal of the Fisheries Research board of Canada 30, 379387. Clark M. & Rose K.A. (1997) Individual-based model of stream-resident brook trout and brook char: model description, corroboration, and eects of sympatry and spawning season duration. Ecological Modelling 94, 157175. Cummins K.W. & Wuycheck J.C. (1971) Caloric equivalents for investigations in ecological energetics. Mitteilungen Internationale Vereinigung fur Theoretische und angewandte Limnologie 18, 1158. Elliott J.M. (1976) The energetics of feeding, metabolism and growth of brown trout (Salmo trutta L.) in relation to body weight, water temperature and ration size. Journal of Animal Ecology 45, 923948. Enders E.C., Boisclair D. & Roy A.G. (2003) The eect of turbulence on the cost of swimming for juvenile Atlantic salmon (Salmo salar). Canadian Journal of Fisheries and Aquatic Sciences 60, 11491160. Flore L., Keckeis H. & Schiemer F. (2001) Feeding, energetic benet and swimming capabilities of 0+ nase (Chondrostoma nasus L.) in owing water: and integrative laboratory approach. Archive fur Hydrobiologie Supplement 135, 409424.

Fraser D.F. & Cerri R.D. (1982) Experimental evaluation of predator-prey relationships in a patchy environment: consequences for habitat use patterns in minnows. Ecology 63, 307313. Gilliam J.F. & Fraser D.F. (1987) Habitat selection under predation hazard: test of a model with foraging minnows. Ecology 68, 18561862. Grant J.W.A., Steingrimsson S.O., Keeley E.R. & Cunjak R.A. (1998) Implications of territory size for the measurement and prediction of salmonid abundance in streams. Canadian Journal of Fisheries and Aquatic Sciences 55(Suppl. 1), 181190. Grossman G., Rincon P.A., Farr M.D. & Ratajczak R.E. (2002) A new optimal foraging model predicts habitat use by drift-feeding stream minnows. Ecology of Freshwater Fish 11, 210. Guensch G.R., Hardy T.B. & Addley R.C. (2001) Examining feeding strategies and position choice of drift-feeding salmonids using and individual-based, mechanistic foraging model. Canadian Journal of Fisheries and Aquatic Sciences 58, 446457. Hartman K.J. & Brandt S.B. (1995) Estimating energy density of sh. Transactions of the American Fisheries Society 124, 347355. Harvey B.C. (1991) Interactions among stream shes: predator-induced habitat shifts and larval survival. Oecologia 87, 2936. Harvey B.C., White J.L. & Nakomoto R.J. (2005) Habitatspecic biomass, survival, and growth of rainbow trout (Oncorhynchus mykiss) during summer in a small coastal stream. Canadian Journal of Fisheries and Aquatic Sciences 62, 650658. Hayes J.W. & Jowett I.G. (1994) Microhabitat use by large brown trout in three New Zealand rivers. North American Journal of Fisheries Management 14, 710725. Hayes J.W., Stark J.D. & Shearer K.A. (2000) Development and test of a whole-lifetime foraging and bioenergetics growth model for drift-feeding brown trout. Transactions of the American Fisheries Society 129, 315 332. Hayes J.W., Hughes N.F. & Kelly L.H. (2007) Process-based modelling of invertebrate drift transport, net energy intake and reach carrying capacity for drift-feeding salmonids. Ecological Modelling 207, 171178. Heggenes J., Northcote T.G. & Peter A. (1991) Seasonal habitat selection and preferences by cutthroat trout (Oncorhynchus clarki) in a small, coastal stream. Canadian Journal of Fisheries and Aquatic Sciences 48, 1364 1370. Hewett S.W. & Johnson B.L. (1992) A Generalized Bioenergetics Model of Fish Growth for Microcomputer. Technical ReportWIS-SG-92-250. Madison, WI: University of Wisconsin Sea Grant Program, 79 pp.

2009 Crown in the right of Canada.

CHANNEL STRUCTURE AND ALLOMETRY OF TROUT GROWTH

217

Hill J. & Grossman G. (1993) An energetic model of microhabitat use for rainbow trout and rosyside dace. Ecology 74, 685698. Hughes N.F. (1992) Selection of positions by drift-feeding salmonids in dominance hierarchies: model and tests for arctic grayling (Thymallus arcticus) in subarctic mountain streams, interior Alaska. Canadian Journal of Fisheries and Aquatic Sciences 49, 19992008. Hughes N.F. & Dill L.M. (1990) Position choice by driftfeeding salmonids: model and test for arctic grayling (Thymallus arcticus) in subarctic mountain streams, interior Alaska. Canadian Journal of Fisheries and Aquatic Sciences 47, 20392048. Hughes N.F. & Kelly L.H. (1996) A hydrodynamic model for estimating the energetic cost of swimming maneuvers from a description of their geometry and dynamics. Canadian Journal of Fisheries and Aquatic Sciences 53, 24842493. Hughes N.F. (1998) A model of habitat selection by driftfeeding stream salmonids at dierent scales. Ecology 89, 281294. Hughes N.F., Hayes J.W., Shearer K.A. & Young R.G. (2003) Testing a model of drift-feeding using threedimensional videography of wild brown trout, Salmo trutta, in a New Zealand river. Canadian Journal of Fisheries and Aquatic Sciences 60, 14621476. Jobling M. (1994) Fish Bioenergetics. London: Chapman and Hall, 309 pp. Johnston N.T. & Slaney P.A. (1996) Fish Habitat Assessment Procedure. Watershed Restoration Technical Circular No. 8. Victoria, Canada: Queens Printer, 95 pp. Keeley E.R. & Grant J.W.A. (2001) Prey size of salmnoid shes in streams, lakes, and oceans. Canadian Journal of Fisheries and Aquatic Sciences 58, 11221132. Kerr S.R. (1971) Prediction of growth eciency in nature. Journal of the Fisheries Research Board of Canada 28, 809 814. Leopold L.B. & Maddock T. (1953) The Hydraulic Geometry of Stream Channels and some Physiographic Implications. U.S.G.S. Prof. Paper No. 252. Washington, DC: U.S.G.S., 56 pp. Leung E., Rosenfeld J. & Bernhardt J. (2009) Habitat eects on invertebrate drift in a small trout stream: implications for prey availability to drift-feeding sh. Hydrobiologia 623, 113125. Lonzarich D.G. & Quinn T.P. (1995) Experimental evidence for the eect of depth and structure on the distribution, growth, and survival of shes. Canadian Journal of Zoology 73, 22232230. Montgomery D.R., Bungton J.M., Smith R.D., Schmidt K.M. & Pess G. (1995) Pool spacing in forest channels. Water Resources Research 31, 10971105.

Mundie J.H. (1974) Optimization of the salmonid nursery stream. Journal of the Fisheries Research Board of Canada 31, 18271837. Murphy M.L., Heifetz J., Johnson S.W., Koski K.V. & Thedinga J.F. (1986) Eects of clear-cut logging with and without buer strips on juvenile salmonids in Alaskan stream. Canadian Journal of Fisheries and Aquatic Sciences 43, 15211533. Ney J.J. (1993) Bioenergetics modelling today: growing pain on the cutting edge. Transactions of the American Fisheries Society 122, 736748. Nikora V.I., Aberle J., Biggs B.J.F., Jowett I.G. & Sykes J.R.E. (2003) Eects of sh size, time-to-fatigue, and turbulence on swimming performance: a case study of Galaxias maculatus. Journal of Fish Biology 63, 13651382. Nislow K.H., Folt C.L. & Parrish D.L. (1999) Favorable foraging locations for young Atlantic salmon: application to habitat and population restoration. Ecological Applications 9, 10851099. Nislow K.H., Folt C.L. & Parrish D.L. (2000) Spatially explicit bioenergetic analysis of habitat quality for age-0 Atlantic salmon. Transactions of the American Fisheries Society 129, 10671081. Peters R.H. (1983) The Ecological Implications of Body Size. Cambridge: Cambridge University Press, 345 pp. Petersen J.T. & Rabeni C.F. (2001) Evaluating the physical characteristics of channel units in an Ozark stream. Transactions of the American Fisheries Society 130, 898910. Piccolo J.J., Hughes N.F. & Bryant M.D. (2008a) Development of net energy intake models for drift-feeding juvenile coho salmon and steelhead. Environmental Biology of Fish 83, 259267. Piccolo J.J., Hughes N.F. & Bryant M.D. (2008b) Water velocity inuences prey detection and capture by driftfeeding juvenile coho salmon (Oncorhynchus kisutch) and steelhead (Oncorhynchus mykiss irideus). Canadian Journal of Fisheries and Aquatic Sciences 65, 266275. Po N.L. & Huryn A.D. (1998) Multi-scale determinants of secondary production in Atlantic salmon (Salmo salar) streams. Canadian Journal of Fisheries and Aquatic Sciences 55(Suppl. 1), 201217. Post J.R. & Parkinson E.A. (2001) Energy allocation strategy in young sh: allometry and survival. Ecology 82, 10401051. Post J.R., Parkinson E.A. & Johnston N.T. (1999) Densitydependent processes in structured sh populations: interaction strengths in whole-lake experiments. Ecological Monographs 69, 155175. Power M.E. (1984) Depth distributions of armoured catsh: predator induced resource avoidance? Ecology 65, 523528.

2009 Crown in the right of Canada.

218

J. S. ROSENFELD & J. TAYLOR

Quinn T.P. & Peterson N.P. (1996) The inuence of habitat complexity and sh size on overwinter survival and growth of individually marked juvenile coho salmon (Oncorhynchus kisutch) in Big Beef Creek, Washington. Canadian Journal of Fisheries and Aquatic Sciences 53, 15551564. Railsback S.F. & Harvey B.C. (2002) Analysis of habitatselection rules using an individual-based model. Ecology 83, 18171830. Railsback S.F., Stauer H.B. & Harvey B.C. (2003) What can habitat preference models tell us? Tests using a virtual trout population. Ecological Applications 13, 15801594. Rincon P.A. & Lobon-Cervia J. (2002) Nonlinear self-thinning in a stream-resident population of brown trout (Salmo trutta). Ecology 83, 18081816. Rosenfeld J.S. (2003) Assessing the habitat requirements of stream shes: an overview and evaluation of dierent approaches. Transactions of the American Fisheries Society 132, 953968. Rosenfeld J.S. & Boss S. (2001) Fitness consequences of habitat use for juvenile cutthroat trout: energetic costs and benets in pools and ries. Canadian Journal of Fisheries and Aquatic Sciences 58, 585593. Rosenfeld J.S., Porter M. & Parkinson E. (2000) Habitat factors aecting the abundance and distribution of juvenile cutthroat trout (Oncorhynchus clarki) and coho salmon (Oncorhynchus kisutch). Canadian Journal of Fisheries and Aquatic Sciences 57, 766774. Rosenfeld J.S., Leiter T., Lindner G. & Rothman L. (2005) Food abundance and sh density alters habitat selection, growth, and habitat suitability curves for juvenile coho salmon (Oncorhynchus kisutch). Canadian Journal of Fisheries and Aquatic Sciences 62, 16911701.

Rosenfeld J.S., Post J.R., Robins G. & Hateld T. (2007) Hydraulic geometry as a physical template for the River Continuum: applications to optimal ows and longitudinal trends in sh habitat. Canadian Journal of Fisheries and Aquatic Sciences 64, 755. Solazzi M.F., Nickleson T.E., Johnson S.L. & Rodgers J.D. (2000) Eects of increasing winter rearing habitat on abundance of salmonids in two coastal Oregon streams. Canadian Journal of Fisheries and Aquatic Sciences 57, 906914. Stewart D.J. & Ibarra M. (1991) Predation and production by salmonine shes in Lake Michigan, 197888. Canadian Journal of Fisheries and Aquatic Sciences 48, 909922. Steward D.J., Weininger D., Rottiers D.V. & Edsall T.A. (1983) An energetics model for lake trout, Salvelinus namaycush: application to the Lake Michigan population. Canadian Journal of Fisheries and Aquatic Sciences 40, 681698. Thompson A.R., Petty J.T. & Grossman G.D. (2001) Multiscale eects of resource patchiness on foraging behaviour and habitat use by longnose dace, Rhinichthys cataracte. Freshwater Biology 46, 145160. Ursin E. (1967) A mathematical model of some aspects of sh growth, respiration, and mortality. Journal of the Fisheries Research Board of Canada 24, 23552453. Van Winkle W., Jager H.I., Railsback S.F., Holcomb B.D., Studley T.K. & Baldrige J.E. (1998) Individualbased model of sympatric populations of brown and rainbow trout for instream ow assessment: model description and calibration. Ecological Modelling 110, 175207.

2009 Crown in the right of Canada.