Influence of velocity and chlorophyll standingstock on periphyton δ13C and δ15N in theSte. Marguerite River system, Quebec

Joseph B. Rasmussen and Veronique Trudeau

Abstract: In a previous laboratory experiment, water velocity had a strong negative effect on periphyton δ13C andδ15N, probably reflecting the effect of flow on boundary-layer exchange. To determine whether similar patterns couldbe detected in situ at different spatial scales, we examined several reaches, each with a wide range in water velocityand chlorophyll standing stock. Periphyton samples were uniform in composition, composed predominantly of livingand dead diatoms of very few species. δ13C signatures were more variable (–17.7‰ to –31.5‰) than δ15N signatures(–0.4‰ to 5.7‰). Velocity and chlorophyll were significant predictors of periphyton δ13C within reaches, mainly atmedium scales (10–500 m) where persistent alternations between fast and slow current (riffles–runs–pools) producedinverse fluctuations of periphyton δ13C. However, none of the variability in δ15N could be explained, possibly a resultof severe boundary-layer N depletion. Similarly, the velocity and chlorophyll predictors could not explain the among-reach signature differences for either isotope. δ13C signatures tended to be 13C-depleted in the tributary streams and13C-enriched in the downstream reaches.

Résumé : Dans une expérience antérieure en laboratoire, la vitesse du courant avait un fort effet négatif sur le δ13C etle δ15N du périphyton, ce qui reflète probablement l’effet du courant sur les échanges dans la couche limite. Afin dedéterminer si de tels patrons s’observent in situ à diverses échelles spatiales, nous avons examiné plusieurs sectionsprésentant chacune une gamme étendue de courants et de biomasses de chlorophylle. Les échantillons de périphytonont une composition uniforme et comprennent principalement un très petit nombre d’espèces de diatomées vivantes etmortes. Les signatures de δ13C sont plus variables (–17,7 ‰ à –31,5 ‰) que celles de δ15N (–0,4 ‰ à 5,7 ‰). La vi-tesse du courant et la chlorophylle sont des variables explicatives significatives du δ13C du périphyton dans les sections,particulièrement aux échelles intermédiaires (10–500 m) où des alternances persistantes entre les zones de courants ra-pides et lents (rapides–coulées–fosses) causent des fluctuations inverses du δ13C du périphyton. Toutefois, rien de la va-riabilité de δ15N ne peut être expliqué, probablement à cause d’un important épuisement de N dans la couche limite.De même, les variables explicatives, la vitesse du courant et la chlorophylle, ne réussissent pas à élucider les différen-ces dans les signatures de ni l’un ni l’autre des deux isotopes. Les signatures de δ13C ont tendance à être réduites en13C dans les tributaires et enrichies en 13C dans les sections d’aval.

[Traduit par la Rédaction] Rasmussen and Trudeau 1381

Introduction

In recent years, stable isotopic signatures have becomean essential ecological tool due to both their ability to re-flect trophic dynamics and their importance as source indi-cators to provide spatial information about physical andbiological processes. Interpreting stable isotopic signaturesin food web studies requires understanding of the factorsthat contribute to spatial variability in primary producersignatures, including both the variability in fractionationrelative to primary nutrient sources and the biogeochemicalfactors that influence nutrient signatures. In river systems,there is a complex interplay of these two factors operatingat different spatial scales that makes consumer signatures

even more challenging to interpret (Finlay et al. 1999,2002; Finlay 2004). In a laboratory experiment, Trudeauand Rasmussen (2003) have shown that both 13C and 15Nfractionation by benthic algae increase with water velocity,although the effect on δ13C was much stronger. Finlay et al.(2002) found very similar results in a field study wherethey compared δ13C and δ15N signatures of benthic algaetaken from riffles vs. pools. Reduced fractionation underlow velocity likely results from nutrient depletion in a stag-nant boundary layer, and high velocities likely prevent thisfrom occurring. Moreover, thin diatom films had more neg-ative δ13C signatures than more dense communities domi-nated by filamentous green algae under the same velocityconditions, indicating that the thickness of the algal mat

Can. J. Fish. Aquat. Sci. 64: 1370–1381 (2007) doi:10.1139/F07-109 © 2007 NRC Canada

1370

Received 31 July 2006. Accepted 4 June 2007. Published on the NRC Research Press Web site at cjfas.nrc.ca on 7 October 2007.J19454

J.B. Rasmussen.1 Department of Biological Sciences, University of Lethbridge, 4401 University Drive W., Lethbridge,AB T1K 3M4, Canada.V. Trudeau. Department of Biology, McGill University, 1201 Dr. Penfield Avenue, Montreal, QC H3A 1B1, Canada.

1Corresponding author (e-mail: joseph.rasmussen@uleth.ca).

(chlorophyll standing stock) might interact with water ve-locity to influence signatures, both through its physicalpresence in the boundary layer and by enhancing the up-take rate of nutrients by the biofilm. Recent studies haveprovided strong support for both the velocity and the chlo-rophyll effect. Singer et al. (2005) showed that underhighly fluctuating velocities, periphyton δ13C signaturesstrongly reflected flow histories over several weeks, andHill and Middleton (2006) have shown that the buildup ofchlorophyll standing stock over time on artificial substratesenriched periphyton δ13C signatures. Signature effects re-sulting from boundary-layer effects on fractionation canalso be traced up the food web. Finlay et al. (1999, 2002)were able to discriminate δ13C signatures of fish and inver-tebrates from pools and riffles and used these to trace foodweb interactions. Morinville and Rasmussen (2003, 2006a)found that premigratory juvenile brook trout, which occupyfaster water habitats than residents, also had more depletedcarbon signatures than residents.

Although our controlled laboratory study generated pat-terns on a small spatial scale, where velocity was regulatedand light and water source were kept constant, the presentfield study was conducted to examine in situ patterns of δ13Cand δ15N in the Ste. Marguerite River (SMR), where manyfactors are varying simultaneously. Although flow conditionschange gradually along the SMR from reach to reach, eachsection of the river has sections where flow alternates be-tween fast (riffles) and slow (pools) velocity. By samplingperiphyton communities along these gradients of velocity, aswell as across the river, we hoped to generate a data set ofperiphyton δ13C and δ15N signatures that would allow us tomodel the relationships of signatures to velocity and chloro-phyll standing stock across a range of spatial scales anddetermine how much of the spatial variation in isotopic sig-

natures along the course of the river could be explained bythese variables.

We expected that periphyton from river sections with per-sistent fast currents would have significantly more 13C- and15N-depleted signatures than sections with persistent slowcurrents. In addition, we hypothesized that high chlorophyllstanding stock should result in enriched signatures due toboth enhanced nutrient flux and reduced boundary-layer ex-change resulting from the thicker biofilm. Because the riverexhibits velocity gradients through several different mecha-nisms that operate on different scales, we tested the hypothe-sis on (i) a local scale (<10 m), where cross-sectionalgradients in velocity predominate, (ii) a medium scale (10–500 m), where alternating pool–run–riffle patterns occur, andfinally, (iii) a larger scale (>500 m), where some reaches aresteeper and have faster average current speeds than others.

Materials and methods

Study areaThe study area was the Ste. Marguerite River system in the

Saguenay region, Quebec, Canada (Fig. 1). This river, with itstwo major branches, drains a 1800 km2 area (runoff coefficient600 mm; Fisheries and Environment Canada 1978), receives1140 mm of precipitation annually, 30% of which falls assnow, and delivers an average discharge of 30.3 m3·s–1 (Goodet al. 2001). Peak discharges are generally associated with thespring melt and exceed 100 m3·s–1 during late April throughMay; minimum flows occur during July and August and areusually <20 m3·s–1 (Good et al. 2001). Water chemistry wasmonitored weekly for 11 weeks at 4–6 stations in four studyreaches (i.e., Cascade, Onesime, Big Pool, Grande Rapide –Bardsville), during summer 2003 by J. Luce and M. Lapointe(Department of Geography, McGill University, Montréal, Qué-

© 2007 NRC Canada

Rasmussen and Trudeau 1371

Fig. 1. The Ste. Marguerite River, Quebec, with the tributaries and reaches sampled during this study: EPT, Epinette tributary; BPT,Big Pool tributary; CA, Cascade; ON, Onesime; BP, Big Pool; GR, Grande Rapide; BA, Bardsville; NE, northeast reach.

bec; Table 1). Like most rivers and streams on the CanadianShield, especially those of the upper north shore of the Gulf ofSt. Lawrence, the Ste. Marguerite has very low conductivity,alkalinity, dissolved inorganic carbon (DIC) (Lachance et al.1988), and nutrient concentrations, although its values for totalN and P are above the median values for Shield streams givenby Grenier et al. (2005).

Many streams feed the river and there are no polluting in-dustrial sources within the drainage system or recent impactsof clear-cut logging operations. Because we had limited roadaccess to much of the river, we performed detailed samplingin a total of eight reaches: five in the main branch of the river,one in the northeast branch, and one in each of two easily ac-cessible tributaries. The term reach, as we apply it, has no de-finitive meaning in terms of hydrology or river geometry. Byreach, we simply refer to a river or tributary section com-posed of a series of nearby sites encompassing a series ofpool–run–riffle sequences. The names applied to the reacheswe studied are traditional names used by salmon guides. Theriver reaches were up to a few kilometres in length, whereastributary reaches were only a few 100 m in length. The fivereaches in the main branch are listed in order from down-stream to upstream in Fig. 1. Bardsville (BA) is a wide sec-tion (>40 m) with no canopy cover. Just upstream, GrandeRapide (GR) is similarly wide with no canopy cover but issteeper with many boulders. Further upstream, Big Pool (BP)and Onesime (ON) are 10 and 20 km, respectively, upstreamof GR. These reaches are narrower (10–20 m), with smalltrees and shrubs along both banks. The upstream-most sectionsampled, Cascade (CA), was the narrowest (≈10 m), shallow-est, and fastest reach, with well-treed banks and more canopycover than any other reach on the main river. The streamreaches Epinette tributary (EPT) and Big Pool tributary (BPT)are very narrow (on average 4 and 2 m wide, respectively)and shallow with a continuously high percent canopy cover.Although a canopy cover is present, it should be noted thatthis is a boreal forest region with a history of logging andcontains no tall coniferous or deciduous trees. Consequently,none of the streams is shaded to anywhere near the same ex-tent as first- and second-order streams in beech maple forestsfurther south. Moreover, periphyton growth is always abun-dant in these streams. The NE reach is situated in the north-eastern branch of the river below the falls. This reach istypical of the lower section of the river, which has no canopycover, a gentle slope, and silty gravel. Some periphyton sam-ples were also taken from the lower river (LR), near itsmouth. This section was not sampled consistently, as it wasnot possible to locate sites with a wide spectrum of flow ve-locities. However, a few samples were taken in this rich siltyclay section for comparison with results from the NE reach.

Species collectedPeriphyton samples were examined microscopically and

found to be very uniform and composed mainly of living anddead diatoms, mainly Tabellaria spp., Eunotia spp., Naviculaspp., Synedra spp., and Gomphonema spp. At the time of sam-pling, almost no green filamentous algae (Chlorophyta) wasfound in the river and associated streams. Some species werefound in low proportion in a few samples (Tetraspora spp.,Bulbochaete spp., Draparnaldia spp., and Ulothrix spp.). Afew species of cyanobacteria also accounted for a small propor-

© 2007 NRC Canada

1372 Can. J. Fish. Aquat. Sci. Vol. 64, 2007

Rea

chTo

tal

N,

µm

olN

·L–1

NH

4+,

µm

olN

·L–1

NO

3+N

O2– ,

µm

olN

·L–1

Tota

lP,

µm

olP

·L–1

PO

4–3-P

,µ

mol

P·L

–1S

ilic

a,µ

mol

Si·L

–1H

CO

3– ,µ

equi

v.·L

–1D

IC,

µm

ol·L

–1pH

Con

d.,

µS

·cm

–1N

Cas

cade

20.7

±1.

51.

48±

0.16

2.87

±0.

390.

87±

0.18

0.42

±0.

0391

.2±

13.4

——

——

4O

nesi

me

17.1

±1.

961.

84±

0.78

2.78

±0.

391.

48±

0.36

0.53

±0.

1996

.1±

1.33

——

——

4B

igP

ool

16.1

±1.

211.

18±

0.19

3.43

±0.

491.

34±

0.20

0.38

±0.

2011

2±12

.4—

——

—4

Gra

nde

Rap

ide

–B

ards

vill

e14

.3±

0.75

1.04

±0.

173.

48±

0.29

0.92

±0.

150.

40±

0.04

91.2

±3.

4—

——

—6

Ste

.M

argu

erit

em

ean

17.1

1.40

3.14

1.15

0.43

—12

0*13

0*7.

226

18C

anad

ian

Shi

eld

mea

n†15

1.43

3.71

0.55

0.32

——

—7.

338

—

Not

e:U

npub

lishe

dda

tapr

ovid

edby

Jam

esL

uce

and

Mic

helle

Lap

oint

e,D

epar

tmen

tof

Geo

grap

hy,

McG

illU

nive

rsity

,M

ontr

éal,

Qué

bec

(rep

rese

nts

sam

plin

gov

er11

wee

ksdu

ring

June

–Aug

ust

2003

).D

IC,

diss

olve

din

orga

nic

carb

on;

Con

d.,

cond

uctiv

ity;

N,

num

ber.

*Aft

erL

acha

nce

etal

.(1

988)

,D

ICca

lcul

ated

asH

CO

3–al

kalin

ity+

KH

pCO

2(=

10µ

equi

v.·L

–1,

assu

min

gpC

O2

=3.

7×

10–4

atm

)fo

llow

ing

Stum

man

dM

orga

n(1

996)

).† A

fter

Gre

nier

etal

.20

05.

Tab

le1.

Wat

erch

emis

try

and

nutr

ient

conc

entr

atio

nsin

four

stud

yre

ache

sof

the

Ste

.M

argu

erit

eR

iver

.

tion of the communities (Gloeotrichia spp. and Phormidiumspp.). Although some fine particulate detritus (mostly leaf frag-ments) and conifer pollen grains were often present in peri-phyton samples, and these were not removed, their contributionto mass would have been negligible, and there was little doubtthat the measured periphyton signature did not include a signif-icant proportion of terrigenous material. Although it is possiblethat some of the dissolved organic matter being used by bacte-ria within the periphyton community may be of terrestrial ori-gin, their biomass is still being formed in situ and can beconsidered autochthonous in that sense.

Water velocity measurementsIn every reach, 7 to 15 sites (cross sections) were chosen

according to their overall water velocity: pool (<0.1 m·s–1),run (0.1–0.3 m·s–1), or riffle (>0.3 m·s–1). Consecutive siteswithin a reach were always different in terms of overall wa-ter velocity. Reaches varied in length and were generally≈300 m in the tributary streams, where sites were closer to-gether, and up ≈2 km in the river. Our sampling locations didnot exceed 70 cm in depth, and rock sizes varied from 5 to30 cm in diameter. Within each site, three different water ve-locities were measured using a Gurley pygmy current meter(625DF8N, wading rod suspended pygmy-type current meteroutfit with model 1100 digital indicator; Gurley PrecisionInstruments, Troy, N.Y.). Water velocity measurements weretaken 1.5 cm above the surface of the periphyton layer andin the middle of a 0.1 m2 surface area delimited by a quadratof the same size. The rocks exposed to the surface within thesampling quadrat were removed, and periphyton were com-pletely washed off the rocks with a brush and a known vol-ume of water. A subsample of this water–algae mixture wasfiltered using Whatman 47 mm glass microfibre filters andkept in the dark for subsequent chlorophyll a analysis. An-other subsample was kept for stable isotopic analysis. Allsamples were frozen at –20 °C until analysed.

δ13C and δ15N have been reported to vary seasonally (Doiet al. 2003; Finlay 2004). Although preliminary analyses hadfailed to detect differences between early- and late-summerisotopic signatures, to avoid possible biases due to smalltemporal isotopic variations and changes in the water level,all reaches were sampled over a 3-week period during mid-summer (July), when productivity is the highest. Samplingof individual reaches took no more than 2–3 consecutivedays during which the water level stayed fairly constant (ap-proximately ±2 cm).

Sample analyses

Stable isotope analysesAll samples of periphyton were oven-dried, pulverised, and

placed into tin capsules. Stable C and N isotopic analyseswere performed by mass spectrometry (Finnigan-Mat Delta-Plus continuous flow isotope-ratio mass spectrometer coupledto a Carlo-Erba elemental analyser online; G.G. Hatch IsotopeLaboratories, University of Ottawa, Ottawa, Ontario, Canada,and GEOTOP Laboratory, University of Quebec in Montreal,Quebec, Canada). The analytical precision of these apparatusis typically 1 standard deviation (SD) for C and N and is inthe range of 0.05‰ to 0.2‰, which is small relative to therange of values found in nature (Kendall and Caldwell 1998).

Stable isotopic signatures were expressed using the follow-ing standard equation:

δ δ13 15sample standardC and N 1 1000= − ×[( / ) ]R R

where R is the ratio of 13C to 12C or 15N to 14N, and the stan-dards are Pee Dee Belemnite limestone and atmospheric ni-trogen for C and N, respectively. The units are in parts perthousand (‰).

Chlorophyll a analysisChlorophyll a was extracted for 24 h from each peri-

phyton sample using 95% acetone. Chlorophyll a concentra-tions (expressed in mg·m–2) were then measured byspectrophotometry (Parsons et al. 1984), because microalgalchlorophyll a concentration is a good indicator of microalgalbiomass (Wetzel and Likens 2000).

Data analysesDescriptive statistics (mean, standard deviation) were

calculated for periphyton δ13C and δ15N, water velocity, andchlorophyll standing stock for each of the eight study reaches.Within each reach, linear regressions were calculated, and theparameters were compared among reaches. Using analysis ofcovariance (ANCOVA; general linear model in JMP-IN®,SAS Institute Inc., Cary, North Carolina), we then analyzedthe effect of velocity and chlorophyll standing stock onperiphyton signatures at the local scale (<10 m), at the inter-mediate averaging scale within reaches (10–500 m), and atthe large scale encompassing entire reaches (>5 km).

Allocating total and explained variance to different scaleintervals

Because gradients in velocity can exist both across theriver and along the river at different spatial scales, we at-tempted to partition total variance and variance explained bythe test variables velocity and chlorophyll standing stockinto increments reflecting different spatial scales. Theseanalyses of scale effects were carried out on the mainbranch, where sampling was most intensive, and involvedtwo different procedures. (i) The first was data smoothing,which combined individual periphyton samples (0.1 m2), andthe velocity and chlorophyll measurements taken with them,into groups of samples located within 10 m (site means) and500 m of each other along the river, and calculating the aver-age for each variable for each of these groupings. In a fewcases, sites belonged to more than one grouping, but nogroupings that shared more than half of their sites, werecounted in these analyses. Variances among each of thesegroupings were then calculated for each variable, and theexplanatory power of the predictors and the categoricalREACH effect in each ANCOVA analyses were determinedfor all of the grouped data sets. (ii) In the second, variableswere standardized with respect to specific scale means bysubtracting the scaled group means from each data point. Bycombining the smoothing and the standardization proce-dures, increments of total and explained variance were esti-mated for three intervals of spatial scale: (1) local scale(<10 m) that mostly reflect cross-sectional variation at apoint in the river), (2) medium scale (10–500 m), the scale atwhich river meanders produce large velocity gradientsassociated with alternating pool–run–riffle sequences, and

© 2007 NRC Canada

Rasmussen and Trudeau 1373

(3) large scale (>500 m), where variations in bed slope,channel width, and depth can give rise to velocity variations.For the analysis of local-scale effects, data from each indi-vidual sampling location were standardized to site averagesby subtracting the site mean from each data point at that site.For the analysis of medium-scale effects, site average valueswere standardized to 500 m group averages. Finally, for theanalysis of large-scale effects, 500 m group averages (un-standardized) were used for each variable.

Results

Water velocities ranged from 0 to 1.27 m·s–1, whichspans the range normally encountered in rivers andstreams (Table 2). Although the range was large in allreaches, mean velocities differed among reaches, with thefastest currents recorded in the steep, narrow upperreaches of the main branch (CA and ON), and slower cur-rents recorded in the wider, less steep sections down-stream. The lowest mean velocities were found in thesmall tributary EPT. Chlorophyll standing stock rangedfrom 0.3 to 31.1 mg·m–2, and both variables were highlyvariable within each reach (Table 2). The highest meanchlorophyll was in the NE branch, and the lowest meanswere in the steep upstream sections of the main branch.Periphyton δ13C varied from –31.5‰ to –17.7‰ over allthe reaches, and within-reach variability was also high,with ranges between 3.6‰ and 7.8‰. All reaches in themain branch, as well as the BPT reach, had similar meanδ13C between –25.3‰ and –26.7‰. The EPT reach hadthe most negative δ13C mean (–29.7‰) and the smallestrange, whereas the NE reach had the least negative δ13Cmean (–21.0‰) and the largest range. The mean δ15N ofindividual reaches were very similar and varied from1.1‰ to 3.1‰, with the highest values recorded in thelower section of the river (NE and LR).

Periphyton δ13C and δ15N versus velocity andchlorophyll standing stock

Within reaches, water velocity and periphyton δ13C (sitemeans) tended to oscillate inversely along the pool–run–riffle sequences (Fig. 2); thus when water velocity was high,δ13C was low and vice versa. Within each reach, the δ13Cversus velocity regressions were highly significant (p <0.002) (Fig. 3). The intercept values differed among reaches,ranging from –29.1‰ in EPT to –19.8‰ in the NE branch.The intercepts tended to vary in the same direction as thereach means. The reach-specific δ13C versus chlorophyll re-gressions ranged from highly significant (p < 0.001) in ON,BP, and the NE branch to nonsignificant in EPT and GR(Fig. 4). The most significant of these relationships werecurvilinear, requiring a quadratic fit with negative second-power term. Because chlorophyll standing stock tended todecline slightly with increasing velocity (Fig. 5), the inter-action term log(chlorophyll a ·velocity–1) also yielded highlysignificant regressions in each reach (Table 3). For mostreaches, multiple regressions containing velocity and chloro-phyll a terms were significant. The multiple regressionswere not significantly improved by power or logarithmictransformation of the variables, or by the inclusion of

© 2007 NRC Canada

1374 Can. J. Fish. Aquat. Sci. Vol. 64, 2007

Tri

buta

ries

Mai

nbr

anch

(mid

dle

sect

ion

from

upst

ream

todo

wns

trea

m)

Mai

nbr

anch

(low

erse

ctio

n)N

Ebr

anch

(low

erse

ctio

n)

EP

TB

PT

CA

ON

BP

GR

BA

LR

NE

δ13C

(‰)

Mea

n

(±S

D)

–29.

75

(±0.

94)

–26.

68

(±1.

19)

–25.

25

(±1.

35)

–26.

86

(±1.

78)

–26.

67

(±1.

27)

–26.

05

(±1.

53)

–26.

13

(±2.

13)

–20.

05

(±0.

40)

–21.

16

(±2.

35)

N44

3944

2327

1933

335

δ15N

(‰)

Mea

n

(±S

D)

2.05

(±1.

32)

1.11

(±0.

55)

—2.

89

(±0.

88)

2.41

(±0.

76)

2.03

(±0.

73)

1.80

(±0.

86)

5.0

(±1.

10)

3.14

(±0.

97)

N45

40—

2321

1226

339

Vel

ocit

y(m

·s–1

)M

ean

(±S

D)

0.21

(±0.

19)

0.37

(±0.

30)

0.49

(±0.

29)

0.51

(±0.

23)

0.36

(±0.

21)

0.22

(±0.

17)

0.32

(±0.

22)

0.10

(±0.

03)

0.30

(±0.

29)

N45

4245

249

2133

339

Chl

a(m

g·m

–2)

Mea

n

(±S

D)

8.06

(±2.

46)

7.24

(±5.

35)

5.34

(±3.

46)

5.21

(±3.

95)

4.63

(±5.

17)

7.27

(±5.

26)

7.48

(±3.

90)

8.9

(±3.

50)

12.2

(±7.

08)

N45

4241

2427

2133

338

Not

e:δ15

Nm

easu

rem

ents

for

the

CA

reac

hw

ere

lost

asa

resu

ltof

am

achi

nem

alfu

nctio

n.E

PT,

Epi

nette

trib

utar

y;B

PT,

Big

Pool

trib

utar

y;C

A,

Cas

cade

;O

n,O

nesi

me;

BP,

Big

Pool

;G

R,

Gra

nde

Rap

ide;

BA

,B

ards

ville

;SD

,st

anda

rdde

viat

ion;

Chl

a,ch

loro

phyl

la.

Tab

le2.

Des

crip

tive

stat

isti

csof

peri

phyt

onδ13

Can

dδ15

N,

velo

city

,an

dpe

riph

yton

biom

ass

(Chl

a)al

ong

the

Ste

.M

argu

erit

eR

iver

Sys

tem

for

each

stud

yre

ach.

second-power (quadratic) terms for chlorophyll and inter-action terms together with velocity and chlorophyll a.

Having verified that velocity was an excellent predictor ofperiphyton δ13C within reaches and that chlorophyll was atleast significant in most reaches, we tested these predictors to-gether on the entire data set using an ANCOVA with REACHas the categorical variable (Table 4). The latter ranged from–4.29 to +4.04 above the mean response of –26.15 (BA)around which the ANCOVA is centered. Both predictors, ve-locity (t value = –10.7) and chlorophyll (t value = 9.8), werehighly significant in the presence of the categorical variableREACH, which explained the baseline differences in signa-tures between reaches not explained by velocity and chloro-phyll. This model explained 85% of the variance inperiphyton δ13C across the entire data set. Although theslopes of periphyton δ13C vs. velocity and chlorophyll ob-tained in reach-specific regressions were variable, interac-tion terms (velocity × REACH and chlorophyll × REACH)were not statistically significant.

Although significant differences among reaches werepresent for periphyton δ15N, none of the variation eitheramong or within reaches could be explained by water veloc-ity or chlorophyll or any interaction between them. Of theseven reaches (note that CA samples for δ15N were lost),only BPT showed a barely significant positive, rather than anegative, linear regression with velocity. Of the sixnonsignificant relationships, half were positive (BA, GR,and ON) and half were negative (BP, EPT, and NE).

Distinguishing small-scale from medium-scale effectsVariance in periphyton δ13C increased with spatial scale,

but the greatest proportion explained (70%) was at themedium-scale increment (10–500 m) (Table 5). This reflectsthe response to the variability in velocity imposed by thepool–run–riffle sequences linked to the river meanders. At

the local scale (<10 m), the explanatory power of the predic-tors was much weaker (12%). Although the total variancewas much greater at the larger scale, it was not explainableby velocity and chlorophyll, indicating that this large-scalevariance is most likely due to hydrological and (or) biogeo-chemical differences among reaches.

The spatial pattern in δ13C observed at the medium scaleand the associated changes in water velocity across pool–run–riffle sequences are illustrated in Fig. 2, where δ13C (sitemeans) are plotted alongside water velocity against distancefor three representative main river reaches and a stream trib-utary. Each oscillation in water velocity is clearly linked toan inverse oscillation in δ13C. The sequence of alternationbetween fast and slow water takes place over scales of a fewhundreds of metres in the main river reaches and a few tensof metres in the stream tributary, and averaging at largerscales largely smooths them out. Thus the scale at which thepredictors explain the greatest proportion of variance in δ13C(Table 5) corresponds to the spatial scale of the pool–run–riffle alternation sequence along the river.

Discussion

A number of factors have been proposed to influence peri-phyton δ13C signatures, including water velocity (Finlay etal. 2002; Trudeau and Rasmussen 2003; Singer et al. 2005),chlorophyll standing stock (Hill and Middleton 2006), DICconcentrations and signatures (Finlay 2001, 2004), tempera-ture (Hill and Middleton 2006), and light (MacLeod andBarton 1998). This study of spatial patterns in the SMRshows that water velocity and chlorophyll standing stock areboth highly significant predictors of within-reach variabilityin periphyton δ13C, but that among-reach variation is morelikely related to other factors that were not measured. Themost plausible explanation for the among-reach differencesis that the δ13C signature of the DIC becomes gradually en-riched downstream, as a result of a combination of factorsincluding atmospheric signature equilibration, degassing ofexcess CO2, and the effects of instream processes such asisotopic fractionation by periphyton (Finlay 2003).

This trend is similar to the upstream–downstream trendin periphyton δ13C reported by other authors (Doucett etal. 1996; Finlay 2001). The effects of temperature andlight were not examined during this study but may alsohave contributed to unexplained variability in our models.Although our study supported the importance of previ-ously hypothesized factors, the analysis showed that theirexplanatory power depended greatly on the spatial scaleof analysis.

Effects of velocity and chlorophyll on small- andmedium-scale variation in δ13C

Analyses based on smoothing and standardization of vari-ables indicated that the explanatory power of velocity andchlorophyll was very low at the local (cross-sectional) scale(12%), despite the considerable amount of variance presentat that scale in both the dependent and independent vari-ables. The explanatory power of the predictors was by farthe highest at the medium scale (>70%), reflecting the im-portant contribution from the alternating pool–run–riffle se-quences associated with the river meander. The inability of

© 2007 NRC Canada

Rasmussen and Trudeau 1375

Fig. 2. Along-stream pattern of site mean Periphyton δ13C (solid line)and site mean water velocity (dotted line) vs. river distance for fourrepresentative reaches. For (a) Cascade (CA), (b) Bardsville (BA),and (c) northeast (NE) reaches, distance (km) was measured relativeto the mouth of the river where the Ste. Marguerite River enters BaieSte. Marguerite; for panel (d) the Epinette tributary reach (EPT), dis-tance was measured upstream from an arbitrary point.

the predictors to explain periphyton δ13C signature variationat the large scale was likely a result of the signature varia-tion associated with the pool–run–riffle alternations havingbeen smoothed out, leaving mainly among-reach variance

related mostly to source signature and not velocity or chlo-rophyll.

The reason why the velocity gradients, formed along theriver through pool–run–riffle sequences, correlate much better

© 2007 NRC Canada

1376 Can. J. Fish. Aquat. Sci. Vol. 64, 2007

Fig. 3. Periphyton δ13C regressed against water velocity (Vel) for each study reach: (a) Epinette tributary (EPT), (b) Big Pool tributary(BPT), (c) Cascade (CA), (d) Onesime (ON), (e) Big Pool (BP), (f) Grande Rapide (GR), (g) Bardsville (BA), and (h) northeast (NE).

Fig. 4. Periphyton δ13C vs. chlorophyll standing stock for each study reach: (a) Epinette tributary (EPT), (b) Big Pool tributary (BPT),(c) Cascade (CA), (d) Onesime (ON), (e) Big Pool (BP), (f) Grande Rapide (GR), (g) Bardsville (BA), and (h) northeast (NE). Statis-tically significant relationships are accompanied by a linear and (or) polynomial fitted equation and its R2 value.

with periphyton δ13C than with local cross-sectional velocitygradients, which are often just as large, probably lies in thepersistence of the gradients in time. Periphyton δ13C signa-tures become gradually enriched over weeks to months as thestanding stock builds up (Hill and Middleton 2006). Singer etal. (2005) have shown that under fluctuating velocity, flowhistory is a much stronger predictor of periphyton δ13C signa-ture than flows measured at a snapshot in time. This meansthat changes in the flow regime, caused by rises and falls inthe water level during the growing season, may alter the wa-ter velocity at different points on the cross-sectional gradientto a much different degree. For example, samples taken atlow water near the stream bank will be associated with muchlower velocities than samples taken 10 m closer to midstream.However, if the water level rises by 50 cm, both sites wouldlikely be far from the bank and their velocity regimes mightbe nearly the same. Thus, velocities measured only once maypoorly reflect the flow history at any given site. On the otherhand, sampling stations in a riffle should always be in a fastercurrent regime than stations within a pool, even though both

will experience increases and decreases in current speed asthe river rises and falls. It is likely the consistency of the rela-tive gradient that results in strong correlations between mea-sured flows and signatures.

Large-scale patterns in periphyton δ13CSteeper upstream reaches in the main branch with fastest

mean current velocities and low mean chlorophyll (CA andON) were expected to exhibit lighter periphyton δ13C signa-tures than slower downstream reaches with higher chloro-phyll standing stock, yet this was not observed. The onlyevidence that could support the effect of velocity on among-reach δ13C differences was the strongly enriched signaturesfound in the lower portion of the NE branch, a reach withlower than average velocity and high chlorophyll levels. Onthis basis, the low velocity and high biomass in EPT shouldhave produced heavy signatures, but instead, EPT periphytonhad the most depleted C signatures of all the reaches. Be-cause there were no significant relationships between thereach means of δ13C and those of velocity, periphyton bio-mass, or their interaction terms, there is no statistical basisfor among-reach differences being attributable to velocity orchlorophyll standing stock. However, only eight reacheswere studied and it is possible that some significant effectsmay have been detected with greater statistical power.

Differences in mean periphyton δ13C and intercepts be-tween reaches most likely reflect reach-specific signaturevalues of the DIC sources assimilated by the algae (Finlay2004), and the most likely sources of these differences areatmospheric signature equilibration, degassing of excessCO2, and the effects of in-stream processes such as isotopicfractionation by periphyton (Finlay 2003).

The EPT stream, with its mean of –29.7‰, had the mostdepleted periphyton δ13C, which is typical of small headwaterstreams and likely reflects the influence of soil weatheringprocesses on the isotopic signatures of DIC (Boutton 1991;Finlay 2003). In the main branch, mean periphyton δ13C esti-mates were fairly similar from one reach to another (–25.3‰ to–26.7‰). The signatures recorded in the BPT stream were notsignificantly more depleted than those of the five main branchreaches, probably because this reach was at a low elevation andnot far upstream from its mouth. In the lower reaches of the

© 2007 NRC Canada

Rasmussen and Trudeau 1377

Fig. 5. Relationship between chlorophyll standing stock andwater velocity across the whole data set.

EPT BPT CA ON BP GR BA NE

Models predicting periphyton δ13C from velocity and Chl aIntercept –29.79 –25.95 –24.83 –27.02 –26.42 –24.60 –23.83 –22.56Chl a 0.09 — 0.16 0.29 0.17 — — 0.21Velocity –3.39 –2.00 –2.60 –2.65 –2.90 –6.29 –7.28 –3.39Adjusted R2 0.46 0.20 0.37 0.72 0.78 0.48 0.53 0.65RMS 0.69 1.07 1.09 0.94 0.59 1.10 1.45 1.40

Models involving an interaction term between Chl a and velocityIntercept –32.08 –28.00 –27.59 –29.64 –28.34 –28.34 –30.33 –26.00Log(Chl a·velocity–1) 0.56 0.42 0.98 1.28 0.78 0.68 1.30 1.21Adjusted R2 0.45 0.22 0.46 0.61 0.65 0.49 0.54 0.44RMS 0.64 1.07 1.01 1.10 0.74 1.10 1.45 1.77N 44 39 41 23 27 19 33 34

Note: Only statistically significant terms were retained. EPT, Epinette tributary; BPT, Big Pool tributary; CA, Cascade; On, Onesime; BP, Big Pool;GR, Grande Rapide; BA, Bardsville; NE, northeast branch; RMS, root mean squared; N, number.

Table 3. Reach-specific regression models predicting periphyton δ13C from velocity and chlorophyll a (Chl a) and their interactionterm log(Chl a·velocity–1) (dependent variable δ13C).

SMR, mean signatures were the most enriched, and this signa-ture gradient continues seaward into the brackish waters of theSte. Marguerite Bay, where δ13C signatures are –16‰ to–18‰, and out into the marine waters of the SaguenayFjord, where δ13C signatures are –12‰ to –15‰ (Morin-ville and Rasmussen 2006b).

There are also a number of other variables not measuredin this study that might have affected the isotopic signaturesobserved and thereby contributed to residual variance in ourmodels. Although shading by shoreline vegetation is mini-mal along the main river, the light regime may have affectedthe periphyton signatures and standing stock within thesmaller tributaries. Two previous studies (MacLeod andBarton (1998) and Wienke and Fisher (1990)) had detectedmore enriched δ13C signatures at well-illuminated sites. Theeffect of chlorophyll might also be underestimated if peri-phyton had been recently abraded by spates (Hill andMiddleton 2006) or heavily grazed upon by benthos(Feminella and Hawkins 1995; Steinman 1996). We do nothave the data to test it, yet it would be useful to examinepossible effects of grazers on periphyton signatures. Differ-ences in algal photosynthetic mechanisms can also affectprimary producer δ13C (Fogel et al. 1992). Fry and Wainright(1991) have shown that cell size influences marine phyto-

plankton δ13C. Species composition has also been suggestedto increase epiphyte δ13C variability (Osmond et al. 1981).Therefore variability in the composition of the benthic com-munity might have contributed to residual variability in ourmodels.

Explained variability: δ15N versus δ13CIn most reaches, the variability of δ15N was much less than

that of δ13C, in many cases less than half. The variation ofδ15N was about the same as the residual variation of δ13C aftervelocity and biomass were accounted for. This could suggestthat it is the greater responsiveness of δ13C fractionation toboundary-layer effects (caused by flow and the periphytonbuildup) that makes it more spatially variable than δ15N. Al-though such velocity and biomass effects on δ15N were notdetected in this field study, we did detect a significant velocityeffect on periphyton δ15N in our laboratory experiment. More-over, Finlay et al. (2002), in their field study on a Californiastream, also found a relationship between algal δ15N and ve-locity very similar to that shown in our lab experiment. Fail-ure of δ15N to reflect the effects of velocity and biomass in theSMR could result from uniform boundary-layer depletion of Nto the extent that no fractionation occurred, and therefore, varia-tion in exchange flux made no difference. Thus spatial variabil-

© 2007 NRC Canada

1378 Can. J. Fish. Aquat. Sci. Vol. 64, 2007

Spatial scale

Small Medium Large

Total variance in increment 0.99 1.85 4.54Variance explained by velocity and chlorophyll a 0.12 1.30 0.00R2 0.12 0.70 0.00

Note: At the local (small) scale, individual samples were standardized by subtracting site means. Atthe medium scale, site means were standardized by subtracting means of 500 m groupings. At the larg-est scale, 500 m group means were used.

Table 5. Partitioning total and explained variance in periphyton signatures by spatial scale(small (local, ≤≈10 m), medium (10–500 m), and large (>500 m)) increments within thesix study reaches of the main Ste. Marguerite River.

Estimate SE t Prob > |t|

Intercept –26.06 0.18 –143.70 <0.0001REACH (BA) –0.23 0.19 –1.24 0.21REACH (BP) –0.19 0.21 –0.91 0.36REACH (BPT) –0.59 0.17 –3.44 0.00REACH (CA) 1.53 0.17 8.80 <0.0001REACH (EPT) –4.29 0.17 –25.29 <0.0001REACH (GR) –0.30 0.24 –1.27 0.21REACH (NE) 4.04 0.20 20.55 <0.0001Velocity (m·s–1) –3.18 0.30 –10.74 <0.0001Chl a (mg·m–2) 0.16 0.02 9.76 <0.0001

Estimate Source F ratio df Prob > F

Adjusted R2 0.84 REACH 17.48 7 <0.0001RMS 1.13 Velocity 75.37 1 <0.0001Mean of response –26.15 Chl a 55.20 1 <0.0001N 260

Note: The estimates for the REACH effect are centered around the –26.06, which is the response forthe BA reach. EPT, Epinette tributary; BPT, Big Pool tributary; CA, Cascade; On, Onesime; BP, BigPool; GR, Grande Rapide; BA, Bardsville; SE, standard error; RMS, root mean squared.

Table 4. Analysis of covariance on the entire data set with the categorical variableREACH (dependent variable, periphyton δ13C (‰)).

ity associated with velocity and biomass gradients at differentscales that greatly enhance δ13C variability would not be seen forδ15N. This idea receives some support from the fact that SMRhas lower DIN concentrations (~4 µmol·L–1) than those of theSt. Lawrence River (DIN ~ 8 µmol·L–1 after Vis et al. 1998),which was the source of the water used in our laboratory ve-locity experiment (Trudeau and Rasmussen 2003). However,we were not able to test this hypothesis directly because wedid not measure the δ15N signature of the DIN in either thewater column or the boundary layer.

An alternative explanation for our inability to explain δ15Nvariation with the hypothesized predictors could be localvariability in the N source signature from groundwater. Al-though highly heterogeneous N signatures in surface waterand groundwater, of sufficient magnitude to override the ve-locity and biomass effects, have often been observed, theyare usually anthropogenically induced and not seen in pris-tine areas such as the SMR (Heaton 1986; Cabana and Ras-mussen 1996). This combined with the overall low within-reach variability of periphyton δ15N makes this explanationdoubtful.

Implications of spatial variation in periphyton signaturesfor food web studies

The pattern of spatial variation of periphyton stable iso-tope signatures may be passed on to consumers and poten-tially provide important information about either habitatfactors (e.g., velocity or benthic productivity) or spatial in-formation about consumer resource interactions. For exam-ple, Finlay et al. (2002) found different C and N signaturesfor invertebrates sampled in riffles than in pools, which al-lowed them to infer the feeding locations of both inverte-brates and fish. Furthermore, enrichment of periphytonsignatures in pools allowed separation of pool periphytonsignatures from terrestrial detritus signatures. Morinville andRasmussen (2003) were able to detect δ13C signature differ-ences between juvenile anadromous brook trout trappedduring their outmigration and resident fish that remainedbehind. This signature difference indicated that the pre-migrant fish might be occupying faster water habitats thansimilar-aged residents, which was subsequently confirmed ina habitat study (Morinville and Rasmussen 2006a). In situa-tions where the scale of movement of consumers is large(e.g., juvenile Atlantic salmon (Salmo salar)) and the con-sumers feed on invertebrates grown in pools, as well as inriffles, the consumer might be expected to average acrossand “smooth out” the variation spectrum of its resources. Al-though this type of smoothing tends to blur the signal of theenvironmental factors operating on the resources, it can po-tentially provide important information about the spatialscale of movement exhibited by the consumer, and we mightexpect to see variation further reduced at higher trophic lev-els if the scale of spatial averaging increases with trophiclevel. If in addition to riffle–pool differences there are large-scale patterns, such as the upstream–downstream gradient inperiphyton δ13C signatures seen in this study (see also Finlay2001; Doucett et al. 1996), this could provide useful spatialinformation about higher trophic levels. For example, thesignatures of outmigrating salmon smolts might link them tothe river section in which they spent the most time. Theseare just a few ways in which spatial information contained in

periphyton signatures could be useful in studies of higher-level consumers or in the analysis of the spatial scale oftrophic interactions and of resource flows within ecosys-tems.

Sampling periphyton for stable isotopic signatures: howshould the effort be allocated?

Our results indicate that roughly half of the within-reachvariance in periphyton δ13C is a result of the combination ofwater velocity and chlorophyll standing stock and their in-fluence on fractionation via the boundary layer. In addition,this variability appears to be generated mostly via themeander-related pool–run–riffle sequences, which are mani-fested at different spatial scales in different stream systems.Superimposed on this is the even greater among-reach vari-ance most likely attributable to source signature variationsresulting from hydrological processes.

Thus, a few samples interspersed throughout the river sys-tem will provide a very poor estimate of the overall carbonsignature of a river, a tributary, a reach, or even a small sec-tion of it. Even our study, which is based on a total of 264δ13C measurements of periphyton from eight reaches of theriver system, cannot claim to have produced a very preciseestimate of the average δ13C signature for this river system.The estimate of within-reach variance for periphyton δ13Cobtained by pooling all data from this study was 2.67, andthe corresponding pooled estimate for among-reach variancewas 5.68, approximately twice the within-reach value. Ourwithin-reach means are therefore quite precise; based on anaverage of 33 samples per reach, we obtain a pooled SE ofonly 0.3‰ (i.e., (2.67·33–1)0.5). On the other hand, becauseamong-reach variance is much higher (5.68) and we sampledonly eight reaches in the study, the pooled SE of our overallmean estimate of periphyton δ13C for the whole river re-mains 0.89‰ (i.e.[(5.68·8–1) + (2.67·33–1)]0.5), assuming thatwithin- and among-reach variances are independent and ad-ditive.

The among-reach variability was not a problem for ourstudy, as the goal was to examine the relationships to ourpredictors, which were expressed mainly on a within-reachscale. However, if the goal had been to derive an average es-timate of the periphyton δ13C for the whole river system,broader sampling rather than intensive within-reach sam-pling would have been called for. Examples of situations inwhich larger scale averaging would be necessary are among-river comparative studies, either linking geochemical pat-terns to biota or perhaps attempting to link the signatures ofanadromous fish otoliths to their river of origin. Any studyof food web interactions will require a periphyton signatureaveraged over the appropriate scale of the consumer–resource interaction, which for some fish might be the localpool in which it spent most time. However, for some highlymobile fish, such as juvenile Atlantic salmon, the appropri-ate scale of signature required might extend well beyond alocal reach perhaps to the whole-river scale.

To illustrate the precision gains from allocating samplingeffort across many reaches, assume that our within- andamong-reach estimates of variance are reasonable and thatwe had sampled 15 reaches of the river with only seven sam-ples per reach. In this case, we would have obtained a stan-dard error (SE) of estimate of 0.87‰ (i.e., [(5.68·15–1) +

© 2007 NRC Canada

Rasmussen and Trudeau 1379

(2.67·7–1)]0.5), which is approximately the same as what weobtained from more than twice the number of samples(264 vs. 105). Because the among-reach variance is almosttwice the within-reach value for the SMR, any optimum al-location of sampling effort would therefore require samplingabout twice the number of reaches as the number of samplesper reach. Within reaches, the maximum amount of varianceseems to be generated at the 10–500 m scale with pool–run–riffle alternations, and thus, optimal sampling effort withinreaches would ideally be oriented toward these features.

Although our models did not explain any of the peri-phyton δ15N variance that we measured, our estimates of theaverage signatures for both the overall river and withinreaches were much more precise for this isotope than forcarbon, because the δ15N variance was six times lower thanδ13C variance over the whole river system. Again, by poolingall data from the study, for δ15N we obtain a within-reachvariance estimate of 0.88, which is less than half of the cor-responding value for δ13C, and an among-reach variance esti-mate of 0.47, more than an order of magnitude less than thecorresponding estimate for δ13C. Thus, the SE of our reachmeans for δ15N (based on an average of 29 samples perreach) was 0.16‰, and the overall SE of estimate was 0.3‰(= [(0.47·7–1) + (0.88·29–1)]0.5), which are both much moreprecise than the corresponding δ13C values. Because among-reach variance in this case is about half of the within-reachvariance, optimal allocation of sampling effort would entailabout twice as many samples taken within each reach as thenumber of reaches sampled. As no patterns were revealed re-garding variability or correlations to other variables, samplescould be taken essentially at random within reaches.

Acknowledgements

We thank Geneviève Morinville, David Browne, and othergraduate students of the McGill University Biology Depart-ment for helpful suggestions on earlier versions of the manu-script. We also thank Laurence Piche for assistance in thefield. The research was carried out at the CIRSA field sta-tion, and we are grateful for the opportunity to work thereand for the support we received from the staff. We are espe-cially grateful to James Luce and Michel Lapointe, Depart-ment of Geography, McGill University, for providing theGIS information for our study sites and for making a sum-mary of their unpublished data on water chemistry and nutri-ents available to us. Financial support was provided by aNatural Sciences and Engineering Research Council of Can-ada (NSERC) Strategic Grant to JBR. We would also like toacknowledge the useful comments of two anonymous re-viewers.

References

Boutton, T.W. 1991. Stable carbon isotope ratios of natural materi-als. II. Atmospheric, terrestrial, marine, and freshwater environ-ments. In Carbon isotope techniques. Edited by C. Coleman andB. Fry. Academic Press, San Diego, California. pp. 173–185.

Cabana, G., and Rasmussen, J.B. 1996. Comparing aquatic foodchains using nitrogen isotopes. Proc. Natl. Acad. Sci. U.S.A. 93:10 844 – 10 847.

Doi, H., Kikuchi, E., Hino, S., Itoh, T., Takagi, S., and Shikano, S.2003. Seasonal dynamics of carbon stable isotope ratios of par-

ticulate organic matter and benthic diatoms in strongly acidicLake Katanuma. Aquat. Microb. Ecol. 33: 87–94.

Doucett, R.R., Power, G., Barton, D.R., Drimmie, R.J., and Cunjak,R.A. 1996. Stable isotope analysis of nutrient pathways leading toAtlantic salmon. Can. J. Fish. Aquat. Sci. 53: 2058–2066.

Feminella, J.W., and Hawkins, C.P. 1995. Interactions between streamherbivores and periphyton: a quantitative analysis of past experi-ments. J. North Am. Benthol. Soc. 14: 465–509.

Finlay, J.C. 2001. Stable-carbon-isotope rations of river biota: impli-cations for energy flow in lotic food webs. Ecology, 82: 1052–1064.

Finlay, J.C. 2003. Controls of streamwater dissolved inorganic car-bon dynamics in a forested watershed. Biogeochemistry, 62:231–252.

Finlay, J.C. 2004. Patterns and controls of lotic algal stable isotoperatios. Limnol. Oceanogr. 49: 850–861.

Finlay, J.C., Power, M.E., and Cabana, G. 1999. Effects of watervelocity on algal carbon isotope ratios: implications for riverfood web studies. Limnol. Oceanogr. 44: 1198–1203.

Finlay, J.C., Khandwala, S., and Power, M.E. 2002. Spatial scalesof carbon flow in a river food web. Ecology, 83: 1845–1859.

Fogel, M.L., Cifuentes, L.A., Velinsky, D.J., and Sharp, J.H. 1992.Relationship of carbon availability in estuarine phytoplankton toisotopic composition. Mar. Ecol. Prog. Ser. 82: 291–300.

Fry, B., and Wainright, S.C. 1991. Diatom sources of 13-rich car-bon in marine food webs. Mar. Ecol. Prog. Ser. 76: 149–157.

Good, S.P., Dodson, J.J., Meekan, M.G., and Ryan, D.A.J. 2001.Annual variation in size-selective mortality of Atlantic salmon(Salmo salar) fry. Can. J. Fish. Aquat. Sci. 58: 1187–1195.

Grenier, M., Campeau, S., Lavoie, I., Park, Y.S., and Lek, S. 2005.Diatom reference communities in Quebec (Canada) streams basedon Kohonen self-organizing maps and multivariate analyses. Can.J. Fish. Aquat. Sci. 63: 2087–2106.

Heaton, T.H.E. 1986. Isotopic studies of nitrogen pollution in the hy-drosphere and atmosphere — a review. Chem. Geol. 59: 87–102.

Hill, W.R., and Middleton, R.G. 2006. Changes in stable isotoperatios during periphyton development. Limnol. Oceanogr. 51:2360–2369.

Fisheries and Environment Canada. 1978. Hydrological atlas ofCanada. Fisheries and Environment Canada, Ottawa, Canada.

Kendall, C., and Caldwell, E.A. 1998. Fundamentals of isotope geo-chemistry. In Isotope tracers in catchment hydrology. Edited byC. Kendall and J. McDonnell. Elsevier, Amsterdam. pp. 51–85.

Lachance, M., Brouard, D., and Walsh, G. 1988. Airborne trans-port of sulphur: impacts on chemical composition of rivers onthe north shore of the St. Lawrence River (Quebec). Water AirSoil Pollut. 39: 311–322.

MacLeod, N.A., and Barton, D.R. 1998. Effect of light intensity,water velocity, and species composition on carbon and nitrogenstable isotope ratios in periphyton. Can. J. Fish. Aquat. Sci. 55:1919–1925.

Morinville, G., and Rasmussen, J.B. 2003. Early juvenile bioenergeticdifferences between anadromous and resident brook trout(Salvelinus fontinalis). Can. J. Fish. Aquat. Sci. 60: 401–410.

Morinville,G., and Rasmussen, J.B. 2006a. Does life history vari-ability in salmonids affect habitat use by juveniles? A compari-son among streams open and closed to anadromy. J. Anim. Ecol.75: 693–704.

Morinville, G., and Rasmussen, J.B. 2006b. Marine feeding patternsof anadromous brook trout (Salvelinus fontinalis) inhabiting anestuarine river fjord. Can. J. Fish. Aquat. Sci. 63: 2011–2027.

Osmond, C.B., Valaane, N.S., Haslam, M., Votila, P., andRoksandic, Z. 1981. Comparisons of δ13C values in leaves ofaquatic macrophytes from different habitats in Britain and Fin-

© 2007 NRC Canada

1380 Can. J. Fish. Aquat. Sci. Vol. 64, 2007

land: some implications for photosynthesis processes in aquaticplants. Oecologia, 50: 117–124.

Parsons, T.R., Maita, Y., and Lalli, C.M. 1984. A manual of chemicaland biological methods for seawater analysis. Pergamon Press, Ox-ford, UK.

Singer, G.A., Panzenbock, M., Weigelhofer, G., Marchesani,C., Waringer, J., Wanek, W., and Battin, T.J. 2005. Flowhistory explains temporal and spatial variation of carbonfractionation in stream periphyton. Limnol. Oceanogr. 50:706–712.

Steinman, A.D. 1996. Effects of grazers on freshwater benthic al-gae. In Algal ecology: freshwater benthic ecosystems. Edited byJ.R. Stevenson, M.L. Bothwell, and R.L. Lowe. Academic Press,San Diego, California. pp. 341–374.

Stumm, W., and Morgan, J.J. 1996. Aquatic chemistry, chemicalequilibria and rates in natural waters. 3rd ed. John Wiley &Sons, Inc., New York.

Trudeau, V., and Rasmussen, J.B. 2003. The effect of water velocityon stable carbon and nitrogen isotope signatures of periphyton.Limnol. Oceanogr. 48: 2194–2199.

Vis, C., Hudon, C., Cattaneo, A., and Pinel-Alloul, B. 1998. Peri-phyton as an indicator of water quality in the St. LawrenceRiver (Quebec, Canada). Environ. Pollut. 101: 13–24.

Wetzel, R.G., and Likens, G.E. 2000. Limnological analyses. 3rded. Springer-Verlag, New York.

Wienke, C., and Fisher, G. 1990. Growth and stable carbon isotopecomposition of cold-water macroalgae in relation to light andtemperature. Mar. Ecol. Prog. Ser. 65: 283–292.

© 2007 NRC Canada

Rasmussen and Trudeau 1381