Slow dam passage in adult Columbia Riversalmonids associated with unsuccessfulmigration: delayed negative effects of passageobstacles or condition-dependent mortality?

Christopher C. Caudill, William R. Daigle, Matthew L. Keefer, Charles T. Boggs,Michael A. Jepson, Brian J. Burke, Richard W. Zabel, Theodore C. Bjornn, andChristopher A. Peery

Abstract: The relationships among behavior, environment, and migration success in anadromous fishes are poorly un-derstood. We monitored migration behavior at eight Columbia and Snake river dams for 18 286 adult Chinook salmon(Oncorhynchus tshawytscha) and steelhead (sea-run Oncorhynchus mykiss) over 7 years using radiotelemetry. When sta-tistically controlling for variation in flow, temperature, fisheries take, and other environmental variables, we observedthat unsuccessful individuals — those not observed to reach spawning areas — had longer passage times at nearly alldams than fish that eventually reached tributaries. In many cases, times were also longer for unsuccessful adults pass-ing through a multiple-dam reach. Four ecological mechanisms may have contributed to these patterns: (i) environmen-tal factors not accounted for in the analyses; (ii) inefficient responses by some fish to passage conditions at dams thatresulted in slowed passage, energetic depletion, and unsuccessful migration; (iii) ongoing selection for traits needed topass obstructions; and (or) (iv) passage rate was not directly linked to migration success, but rather, both resulted fromrelatively poor phenotypic condition upon river entry in unsuccessful migrants. Overall, these results illustrate the needfor a mechanistic understanding of the factors that influence migration success and the need for fitness-based criteria toassess the effects of dams on anadromous fishes.

Résumé : On connaît mal les relations entre le comportement, l’environnement et le succès de la migration chez lespoissons anadromes. Nous avons suivi par télémétrie le comportement migrateur de 18 286 saumons chinook (Oncor-hynchus tshawytscha) et truites arc-en-ciel anadromes (Oncorhynchus mykiss) adultes à huit barrages sur le fleuveColumbia et la rivière Snake sur une période de 7 ans. Une fois que nous avons tenu compte statistiquement desvariations de débit, de température, de retraits dus à la pêche et des autres variables du milieu, nous observons que lesindividus qui échouent leur migration — qui n’ont pas été observés sur les sites de fraye — prennent plus de tempspour traverser presque tous les barrages que les poissons qui éventuellement atteignent les tributaires. Dans plusieurscas, le temps de passage dans les sections avec plusieurs barrages est aussi plus long pour les adultes qui échouent leurmigration. Quatre mécanismes écologiques peuvent contribuer à ces patrons: (i) des facteurs écologiques non considérésdans les analyses, (ii) des réactions inefficaces de certains poissons aux conditions de traversée des barrages qui ontpour résultat de ralentir la traversée, d’épuiser les réserves énergétiques et de faire échouer la migration, (iii) une sélec-tion actuelle des caractéristiques nécessaires pour surmonter les obstacles et (ou) (iv) des taux de passage qui ne sontpas reliés directement au succès de la migration, les deux variables s’expliquant plutôt par une mauvaise conditionphénotypique lors de l’entrée dans la rivière chez les poissons qui échouent leur migration. Dans leur ensemble, nosrésultats indiquent qu’il est nécessaire d’obtenir une compréhension mécaniste des facteurs qui influencent le succès dela migration, ainsi que des critères basées sur la fitness pour évaluer les effets des barrages sur les poissons anadromes.

[Traduit par la Rédaction] Caudill et al. 995

Can. J. Fish Aquat. Sci. 64: 979–995 (2007) doi:10.1139/F07-065 © 2007 NRC Canada

979

Received 28 April 2006. Accepted 16 March 2007. Published on the NRC Research Press Web site at cjfas.nrc.ca on 25 July 2007.J19297

C.C. Caudill,1 W.R. Daigle, M.L. Keefer, C.T. Boggs, M.A. Jepson, T.C. Bjornn,2 and C.A. Peery. Department of Fish andWildlife Resources, College of Natural Resources, University of Idaho, Moscow, ID 83844, USA.B.J. Burke and R.W. Zabel. NOAA Fisheries Service, Northwest Science Center, 2725 Montlake Boulevard East, Seattle, WA98112, USA.

1Corresponding author (e-mail: caudill@uidaho.edu).2Deceased.

Introduction

The single greatest change to rivers worldwide since theindustrial revolution has been the construction of nearly50 000 dams higher than 15 m (Hart and Poff 2002; Poff andHart 2002; Postel and Richter 2003). Dams have long beenrecognized as obstacles to fish migration, and slowed migra-tion or “delay” at dams has been observed in many fish spe-cies, including Atlantic salmon (Salmo salar) (Karppinen etal. 2002), barbel (Barbus barbus) (Lucas and Frear 1997),American shad (Alosa sapidissima) (Moser et al. 2000), sealamprey (Petromyzon marinus) (Haro and Kynard 1997),paddlefish (Polyodon spathula) (Zigler et al. 2004), andsalmonids (Oncoryhncus spp.) of the Columbia River Basin(Williams 1998; Keefer et al. 2004a). Dams have frequentlybeen implicated in the decline of resident and anadromousfish populations, in part because of their effects on adult mi-gration (e.g., NRC 1996; Northcote 1998; Dudgeon 2003).Unfortunately, primarily because of logistical constraints,the large body of literature on adult fish passage behaviorand performance has been conducted primarily at smallscales or at single dams (e.g., Clay 1995; Hinch and Bratty2000; Bunt et al. 2001), and few studies have examined therelationship between slowed migration and eventual migra-tion success upstream (Naughton et al. 2005).

Similar to many anadromous fishes in regulated rivers,adult Chinook salmon (Oncorhynchus tshawytscha) andsteelhead (sea-run Oncorhynchus mykiss) returning tospawning tributaries of the Columbia and Snake rivers mustascend multiple dams (up to nine; Fig. 1). Construction ofthe Columbia–Snake hydrosystem has been identified as oneof several factors causing the decline and listing of morethan a dozen salmon and steelhead evolutionary significantunits (species, subspecies, geographic races, or populationsconsidered as separate entities for conservation) under theUS Endangered Species Act (NRC 1996; Ruckelshaus et al.2002). Adult fish passage facilities at Columbia and Snakeriver dams were explicitly designed for the passage ofsalmonids, and their basic design has been continually re-fined to enhance their ability to attract and pass bothsalmonids and nonsalmonid fishes (Monk et al. 1989; Bell1991; Moser et al. 2002a).

Despite the presence of fish passage facilities and effortsto improve migration conditions at dams, concern remainsthat slowed migration may have direct or indirect negativeeffects on adult salmon migration and reproductive success.In the Columbia–Snake system, most adult salmonids passeach dam within 2 days of entering the tailrace and eventu-ally reach spawning tributaries (Keefer et al. 2004a, 2005).However, 1.4%–13.7% of individuals in each run have takenmore than 5 days to pass single dams, and some fish takeseveral weeks to pass (Keefer et al. 2004a). Adult salmonidsrely on energy reserves gained in the marine environment formigration, suggesting the potential for relatively slow pas-sage to incur a fitness cost. For instance, Geist et al. (2000)estimated that migrating fall Chinook salmon in the Colum-bia River requiring more than 5 days to pass each dam mayhave insufficient energy reserves to complete spawning.Whether migratory delay at dams contributes to reduced fit-ness in anadromous fishes in general or hinders the recoveryof Columbia Basin salmonids in particular has been a linger-

ing question since the construction of dams with fishways.Determining how migration behavior at dams (as measuredby passage time) relates to upstream fate is important to un-derstanding how widespread delayed or carry-over effectsmay be in the ecology of adult anadromous fishes.

During seven migration seasons from 1996 to 2003, weradio-tagged 18 286 returning adult salmonids in four runs(defined by adult return date as spring, summer, and fallChinook salmon, and summer steelhead). Each run wascomposed of one or more evolutionary significant units. Thegeneral aim of the research has been to identify passageproblems and potential solutions. Previously, we have pre-sented general patterns of survival and migration through thehydrosystem (Keefer et al. 2005; Goniea et al. 2006; High etal. 2006) and migration behavior at individual dams andhydrosystem segments (Keefer et al. 2004a, 2004b;Naughton et al. 2005), in tributaries (Keefer et al. 2004c),and in relation to specific passage issues (e.g., Reischel andBjornn 2003; Boggs et al. 2004; Johnson et al. 2005).

In this study, we tested for an association between fishpassage time and eventual fate at two scales: passage at indi-vidual dams in the lower Columbia and Snake rivers and to-tal passage time past four dams and through three reservoirsin the lower Columbia hydrosystem. At the first scale, weused Cox proportional hazards regression (PHReg; Allison1995; Hosmer and Lemeshow 1999) to analyze passage timeat each dam in relation to eventual fate while statistically ac-counting for variation in river environment and fish traits in-somuch as possible. PHReg is particularly well suited to theanalysis of passage time data (Castro-Santos and Haro

© 2007 NRC Canada

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

Fig. 1. Map of the study region, including location of dams. Fishwere collected and tagged at Bonneville Dam. Upstream migra-tion to spawning sites was monitored as far upstream as riverkilometre (rkm) 1300 in the Snake River basin or Wells Dam onthe Columbia River using 150–170 fixed site antennas and mo-bile tracking in boats and trucks. Distances from the ColumbiaRiver mouth are given parenthetically as rkm. Abbreviations forLower Columbia River dams: BO, Bonneville; JD, John Day;TD, The Dalles; MN, McNary. Abbreviations for Snake Riverdams: IH, Ice Harbor; LM, Lower Monumental; GO, LittleGoose; GR, Lower Granite; HC, Hells Canyon (impassible); DW,Dworkshak Dam (Clearwater River).

2003), because the method readily handles cases with in-complete records by censoring and incorporates time-varying covariates — predictor variables whose values areallowed to vary during individual passage events (i.e., riverdischarge or light level). At the second scale, we used linearmodels to test whether mean passage times differed betweensuccessful and unsuccessful migrants ascending a reach ofthe lower Columbia River hydrosystem that included fourdams and three reservoirs. Overall, the analyses revealedconsistent associations between relatively long passagetimes and eventually unsuccessful migration at both scales.We used results from these two analyses to identify potentialunderlying mechanisms affecting migration success in ana-dromous fishes.

Materials and methods

Study systemThe Columbia River is the third-largest river system in

North America, draining an area of 671 000 km2, includingnearly all of Idaho and large areas of Oregon, Washington,Montana, and British Columbia (Fig. 1). Mean annual dis-charge at the river’s mouth is approximately 6650 m3·s–1.Fish returning to Snake River drainages must pass the fourlower Columbia River dams and most also pass the fourlower Snake River dams (Fig. 1). Those returning to Colum-bia River sites upstream from the Snake River confluencepass a total of four to nine dams. As fish migrate upstream,they encounter 1–2 km long tailraces below dams with tur-bulent flow caused by discharge from dam turbines andspillways. Fish must distinguish relatively low volume at-traction flows leading to fishway entrances at dam facesfrom the large discharge of turbines and spillways. Once incollection channels, fish pass through transition pools andinto ladders, which may be up to 1300 m long, gain 35 m inelevation, and contain 75 or more weirs and pools (Williams1998).

Most spring and summer Chinook salmon returning to theSnake River basin and lower Columbia River tributaries arestream type Chinook, spending their first year rearing infreshwater before migrating seaward in the spring and sum-mer of their second year. Most adults return after three win-ters at sea, with some returning after two and four winters(jacks, or precocious males returning after one winter, werenot included in this study). Adult fall Chinook salmon returnin late summer and early fall, and most spawn in themainstem Columbia and Snake rivers, especially the 70 kmHanford Reach below Priest Rapids Dam, the only remain-ing free-flowing stretch of the Columbia River accessible toanadromous fishes upstream from Bonneville Dam. Fall Chi-nook are ocean type Chinook salmon, as are some summer-run Chinook salmon, migrating to sea in their first year (age0), returning after two to five winters at sea at age three tosix. Steelhead adults may return in any month, though themajority of interior Columbia Basin steelhead enter freshwater in summer and fall and spawn the following spring.Juvenile steelhead reside in fresh water for at least 2 yearsbefore outmigrating in spring. Adult steelhead typically re-turn to spawn after one to two winters at sea, and a smallproportion (<10%) are iteroparous.

Radio-tagging, telemetry monitoring, and fateassignment

The methods used to radio-tag and monitor salmonid mi-gration and assign fates to individual fish in the Columbiaand Snake river basins have been described in detail inKeefer et al. (2004a, 2005). Briefly, fish were diverted fromthe Washington Shore fish ladder at Bonneville Dam (riverkilometre (rkm) 235) into a facility where they could be se-lected by species. During fall, we targeted “upriver bright”Chinook salmon, a group generally defined as originatingeast of the Cascade Range; however, run assignment basedon timing may have resulted in the inclusion of some ocean-type summer Chinook salmon and tule fall Chinook salmonreturning to lower Columbia River tributaries and hatcheries.Diverted fish were anesthetized, sexed, measured for forklength, and inspected for the presence of fin clips indicatingknown hatchery origin. All fish were tagged with agastrically implanted radiotransmitter (Lotek Wireless, Inc.,Newmarket, Ontario). Prior to 2000, all fish were releasedabout 9.5 km downstream of Bonneville Dam near Dodson,Oregon, or Skamania, Washington. During 2000–2002,23.2% of spring Chinook salmon, 23.9% of summer Chi-nook salmon, 35.1% of fall Chinook salmon, and 27.9% ofsteelhead were released in the Bonneville forebay to meetother study objectives; the remainder were released at thedownstream sites. In all years, secondary tags, includingpassive integrated transponder tags (PIT tags), were used toestimate rates of transmitter loss, which averaged 2.2%–4.0%, depending on species (Keefer et al. 2004d). Individ-uals with records indicating lost tags were excluded from allanalyses.

Adult migration behavior in the Columbia Basin wasmonitored using an extensive array of approximately 160 ra-dio receiver sites at dams and in tributaries (detailed inMoser et al. 2002b; Reischel and Bjornn 2003; Naughton etal. 2005). Behavior at dams was monitored using fixed aerialand underwater antennas in tailraces and fishways. Majortributaries were monitored with fixed aerial antennas neartributary mouths and with mobile tracking units attached totrucks and boats. Additional data were collected from volun-tary returns of transmitters from fisheries (US$25–$100 re-wards) and from cooperative returns from hatcheries, weirs,and spawning ground surveys operated by federal, state, andtribal agencies.

We used these data to classify individual fish as havingreached potential spawning sites (successful migrants), hav-ing unknown fates (unsuccessful migrants) as in Naughton etal. (2005), or as fisheries take in the main stem Columbia orSnake rivers. Successful migrants included salmon with te-lemetry records in known spawning tributaries, those foundas carcasses in spawning ground surveys, or those that re-turned to their hatchery of origin. Beginning in 2000, wewere able to identify straying in a subset of radio-taggedadults that had been PIT-tagged during the juvenile stage(i.e., fish of known origin). Strays were identified as known-origin adults having final records in non-natal spawning trib-utaries. Straying rates were generally low during 2000–2003(2.2% spring–summer Chinook salmon, n = 1588; 4.2% fallChinook salmon, n = 166; 6.8% steelhead, n = 1414; M.L.Keefer, C.A. Peery, J. Firehammer, and M.L. Moser, unpub-lished data), and in the context of this study, we conserva-

© 2007 NRC Canada

Caudill et al. 981

tively considered strays to be successful migrants bypresuming these individuals had reached potential spawningtributaries. Unsuccessful migrants were salmon with un-known fates whose final telemetry records were outsidespawning sites, usually at or between dams, and presumablyrepresent prespawn mortalities, though this group may haveincluded unreported harvest or fish that went undetected inmonitored tributaries. Whether fish captured in fisherieswould have successfully migrated remains unknown, andhence these fish were excluded from all analyses. We feltthese operational definitions accurately represented the fatesof the vast majority of fish for several reasons. The rate oftransmitter loss through regurgitation was low (Keefer et al.2004d). The linear nature of the study system, the largenumber of antenna sites used, and large mobile tracking ef-forts provided a high probability of detection, particularly atdams. For instance, Naughton et al. (2005) estimated theprobability of an individual passing two dams in the mid-Columbia River undetected at <1.0 × 10–7 based on the min-imum number of fixed-site receivers salmon passed at thesedams and frequencies of receiver outages. While many fishwent undetected at individual antennas, it was improbablethat misclassification of fates seriously biased our resultsoverall, because any misclassification should act to mini-mize observed differences in passage behavior between fategroups, rather than create false differences.

Data analysisFor all analyses, we defined passage time as the time

elapsed from the first detection at a tailrace antenna 0.5–2 km downstream from the face of a dam to the last recordat a fish ladder exit antenna. We conservatively used timesfor first ascents only at each dam for those fish that fell backdownstream (Boggs et al. 2004) and reascended because ofthe potential for learning or injury to affect subsequent pas-sage times. Other measures of passage time, such as first ap-proach or first entrance to ladder exit, provided qualitativelysimilar results (C.C. Caudill, unpublished data).

Passage time at individual damsWe used PHReg (Allison 1995; Hosmer and Lemeshow

1999; Castro-Santos and Haro 2003) to model the relation-ships among observed passage times at individual dams,fate, and other predictor variables using widely availablesoftware (PROC PHREG in SAS v.9, SAS Institute Inc.,Cary, North Carolina). PHReg models the probability of anevent occurring (dam passage) for an individual within avery small time interval as a passage hazard given (i) thatthe event had not occurred prior to the beginning of the timeinterval and (ii) a set of predictor variables such as river dis-charge level and temperature at the beginning of the time in-terval. Increasing passage hazard corresponds to more rapidpassage and shorter passage times. Passage hazards for dif-ferent groups are expressed as odds ratios; an odds ratio of2.0 for successful vs. unsuccessful adults indicates success-ful adults were twice as likely to pass during a time intervalas unsuccessful adults. An odds ratio of 1.0 indicates no dif-

ference in probability. We modeled passage hazard in rela-tion to fish fate and environmental factors (flow, tempera-ture, etc.) and fish traits (length, sex, etc.) known to affectmigration behavior. The primary advantages of the PHRegmethod are that it allows predictor variables to vary throughtime, the censoring of individuals, and the software iswidely available. The inability to estimate mean differencesin passage time represents the primary disadvantage of thePHReg approach. Other parametric approaches to the analy-sis of time-event data are available, but do not allow time-varying covariates, require assumptions about the form ofthe survival function (Allison 1995; Castro-Santos and Haro2003), or allow time-varying covariates, but currently do notincorporate tests among groups (Moser et al. 2005).

Prior to analysis, we identified a set of 15 candidate mod-els, each consisting of 2–14 predictor variables describingvariation in passage environment, fish traits, and individualfate (Table 1; Supplemental Appendix S13). River environ-ment was described by daily mean values of total discharge(flow), spill (the amount of water passing over the spillwayof dams rather than passing through turbines), and tempera-ture because of the potential effect of these factors onsalmon behavior and physiology. River environmental datawere obtained from Columbia River DART (2005). Fewsalmon pass ladders at night (e.g., Naughton et al. 2005),and we included a variable coding day vs. night, adjusted forseasonal changes in day length, to account for diel changesin passage behavior. During some periods with low fish pas-sage density (early spring, late fall, and winter), data fromBonneville and McNary dams and Ice Harbor and LowerGranite dams were used to estimate river values at other Co-lumbia and Snake river dams, respectively, because valuesamong dams were highly correlated (e.g., temperature atBonneville and McNary dams, 1996–2003: R2 = 0.98). Wetested for the potential of high fish densities in ladders toslow passage by including the daily dam count of allsalmonids and American shad for each dam as separate vari-ables (DART 2005). The above model covariates were time-varying for each fish; those described below were fixed.

The final fate of the fish was the primary predictor of in-terest. The Cox proportional hazards model assumes theodds ratio between groups remains constant through time.The inclusion of a fate × passage time interaction both testedfor and statistically controlled for deviation from this modelassumption (Allison 1995, p. 155). Five additional covariatesestimated fish traits: sex, fork length, origin (hatchery orwild), date of tagging as an index of seasonal differencesamong fish entering the river relatively early or late withineach run and (or) unmeasured seasonal trends in environ-mental variation, and release location (downstream ofBonneville Dam or released to the Bonneville Dam forebay).Interannual differences in river condition have a strong ef-fect on migration behavior (Boggs et al. 2004; Keefer et al.2004a, 2004b), and consequently, the analyses were strati-fied by year. The overall effect of year was tested by com-paring models with and without year effects, butunfortunately the quantitative differences among years can

© 2007 NRC Canada

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

3 Supplementary data for this article are available on the journal Web site (cjfas.nrc.ca) or may be purchased from the Depository of Unpub-lished Data, Document Delivery, CISTI, National Research Council Canada, Building M-55, 1200 Montreal Road, Ottawa, ON K1A 0R6,Canada. DUD 5182. For more information on obtaining material refer to cisti-icist.nrc-cnrc.gc.ca/irm/unpub_e.shtml.

© 2007 NRC Canada

Caudill et al. 983

Dam

Bon

nevi

lle

The

Dal

les

John

Day

McN

ary

Ice

Har

bor

Low

erM

onum

enta

lL

ittl

eG

oose

Low

erG

rani

te

Spri

ngC

hino

oksa

lmon

(Onc

orhy

nchu

sts

haw

ytsc

ha)

N38

7828

6526

2222

8716

8015

6213

3614

15C

enso

red

145

228

153

217

110

143

2346

%C

enso

red

3.7

8.0

5.8

9.5

6.5

9.2

1.7

3.3

Uns

ucce

ssfu

l64

436

827

615

187

5330

33%

Uns

ucce

ssfu

l16

.612

.810

.56.

65.

23.

42.

22.

3E

stim

ate

TO

D Odd

sra

tio

0.10

30.

106

0.06

60.

061

0.04

80.

066

0.09

20.

050

95%

CI

0.08

9,0.

120

0.08

9,0.

127

0.05

2,0.

083

0.04

7,0.

080

0.03

4,0.

067

0.04

8,0.

091

0.06

8,0.

124

0.03

5,0.

072

Flo

w Odd

sra

tio

0.99

90.

999

0.99

80.

996

0.99

30.

997

0.99

71.

002

95%

CI

0.99

6,1.

002

0.99

7,1.

001

0.99

6,1.

000

0.99

3,0.

999

0.98

9,0.

998

0.99

3,1.

002

0.99

3,1.

001

0.99

4,1.

011

Tem

pera

ture

Odd

sra

tio

1.04

21.

135

1.08

51.

130

0.98

91.

042

1.02

50.

942

95%

CI

0.86

0,1.

262

1.04

0,1.

238

1.00

5,1.

170

1.02

5,1.

245

0.91

3,1.

070

0.96

8,1.

123

0.94

1,1.

117

0.86

3,1.

029

Spi

ll Odd

sra

tio

0.98

50.

983

0.98

40.

998

0.96

80.

877

0.88

70.

847

95%

CI

0.96

8,1.

003

0.95

8,1.

008

0.94

8,1.

022

0.95

1,1.

047

0.90

9,1.

030

0.81

8,0.

941

0.82

4,0.

955

0.77

9,0.

920

Sal

mon

Odd

sra

tio

0.96

81.

052

1.06

21.

048

1.10

71.

058

1.00

01.

065

95%

CI

0.93

1,1.

006

1.00

3,1.

104

1.01

6,1.

110

1.01

3,1.

085

1.05

2,1.

164

1.01

2,1.

105

1.00

0,1.

000

1.01

5,1.

116

Sha

d Odd

sra

tio

0.17

80.

992

0.98

80.

974

1.12

31.

096

0.99

90.

070

95%

CI

0.00

1,57

.994

0.98

4,0.

999

0.97

9,0.

996

0.95

0,1.

000

1.00

5,1.

254

0.96

0,1.

251

0.99

8,1.

000

0.00

0,80

.522

Rel

ease

site

Odd

sra

tio

NA

1.43

61.

321

1.09

81.

068

0.97

01.

171

1.23

695

%C

I—

1.23

3,1.

672

1.12

7,1.

548

0.92

6,1.

302

0.89

2,1.

279

0.79

8,1.

178

0.96

2,1.

425

1.00

8,1.

514

Fate O

dds

rati

o1.

096

1.19

91.

292

1.49

41.

566

1.46

11.

112

2.22

895

%C

I0.

953,

1.25

90.

993,

1.44

71.

048,

1.59

31.

154,

1.93

41.

107,

2.21

50.

911,

2.34

50.

538,

2.29

81.

053,

4.71

5Fa

te×

Tim

eO

dds

rati

o1.

079

1.13

61.

089

1.01

11.

040

1.29

41.

449

1.20

895

%C

I1.

044,

1.11

61.

070,

1.20

71.

040,

1.14

00.

951,

1.07

40.

961,

1.12

51.

012,

1.65

30.

873,

2.40

50.

902,

1.61

8S

ex Odd

sra

tio

0.95

80.

908

0.98

00.

989

1.00

11.

026

1.08

80.

934

95%

CI

0.88

9,1.

032

0.82

8,0.

995

0.89

4,1.

074

0.89

3,1.

095

0.89

1,1.

125

0.90

3,1.

165

0.95

7,1.

236

0.82

5,1.

058

Tag

day

Odd

sra

tio

1.02

21.

012

1.00

71.

004

0.99

80.

997

1.00

71.

007

95%

CI

1.01

5,1.

029

1.00

0,1.

024

0.99

8,1.

016

0.99

5,1.

013

0.99

0,1.

006

0.98

8,1.

007

0.99

7,1.

017

0.99

4,1.

020

Tab

le1.

Mod

el-a

vera

ged

odds

rati

osan

dun

cond

itio

nal

95%

conf

iden

cein

terv

als

(CI)

from

PH

Reg

anal

yses

ofpa

ssag

eti

me

atlo

wer

Col

umbi

aR

iver

and

Sna

keR

iver

dam

s.

© 2007 NRC Canada

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

Dam

Bon

nevi

lle

The

Dal

les

John

Day

McN

ary

Ice

Har

bor

Low

erM

onum

enta

lL

ittl

eG

oose

Low

erG

rani

te

Cli

pped

Odd

sra

tio

1.00

41.

003

0.93

61.

055

0.93

01.

096

1.06

30.

938

95%

CI

0.92

8,1.

087

0.91

3,1.

102

0.84

8,1.

033

0.95

2,1.

169

0.82

2,1.

053

0.96

0,1.

251

0.92

2,1.

225

0.81

8,1.

076

Len

gth

Odd

sra

tio

0.99

80.

991

0.98

90.

987

1.00

00.

989

0.99

20.

993

95%

CI

0.99

3,1.

003

0.98

5,0.

998

0.98

2,0.

996

0.98

0,0.

994

0.99

2,1.

009

0.98

0,0.

999

0.98

2,1.

002

0.98

3,1.

003

Sum

mer

Chi

nook

salm

onN

1615

1362

1390

1075

299

283

236

244

Cen

sore

d40

4999

169

85

35

%C

enso

red

2.5

3.6

7.1

15.7

2.7

1.8

1.3

2.0

Uns

ucce

ssfu

l19

815

012

165

85

33

%U

nsuc

cess

ful

12.3

11.0

8.7

6.0

2.7

1.8

1.3

1.2

Est

imat

eT

OD Odd

sra

tio

0.11

60.

100

0.16

20.

107

0.05

70.

008

0.16

60.

054

95%

CI

0.09

2,0.

147

0.07

6,0.

131

0.12

9,0.

205

0.07

6,0.

149

0.02

5,0.

125

0.00

1,0.

069

0.09

1,0.

305

0.02

3,0.

128

Flo

w Odd

sra

tio

1.00

30.

997

0.99

81.

000

0.99

91.

003

1.00

70.

985

95%

CI

0.99

8,1.

009

0.99

5,1.

000

0.99

6,1.

001

0.99

4,1.

005

0.98

0,1.

019

0.99

0,1.

017

0.98

5,1.

029

0.96

7,1.

004

Tem

pera

ture

Odd

sra

tio

0.95

30.

938

0.97

20.

937

1.04

10.

979

0.77

20.

908

95%

CI

0.80

0,1.

136

0.85

0,1.

034

0.88

7,1.

066

0.73

7,1.

190

0.89

8,1.

207

0.86

9,1.

102

0.54

9,1.

086

0.75

5,1.

092

Spi

ll Odd

sra

tio

0.94

90.

978

0.98

20.

984

0.89

80.

950

0.85

91.

185

95%

CI

0.91

5,0.

983

0.93

4,1.

025

0.93

7,1.

028

0.92

6,1.

044

0.75

5,1.

069

0.80

5,1.

120

0.70

5,1.

046

0.86

7,1.

621

Sal

mon

Odd

sra

tio

1.01

01.

024

1.05

21.

043

2.00

01.

320

1.00

02.

772

95%

CI

0.76

6,1.

330

0.98

5,1.

065

1.01

7,1.

089

0.98

9,1.

101

1.09

4,3.

657

0.75

7,2.

301

0.99

9,1.

001

1.37

5,5.

585

Sha

d Odd

sra

tio

0.34

70.

997

1.00

01.

006

0.99

30.

964

1.00

00.

977

95%

CI

0.06

4,1.

872

0.99

4,0.

999

0.99

7,1.

003

0.99

3,1.

019

0.95

1,1.

036

0.90

9,1.

024

1.00

0,1.

000

0.44

1,2.

166

Rel

ease

site

Odd

sra

tio

NA

1.77

61.

093

1.15

91.

468

1.29

91.

011

1.49

295

%C

I—

1.41

8,2.

223

0.87

4,1.

367

0.88

5,1.

518

0.91

3,2.

362

0.79

6,2.

119

0.57

8,1.

766

0.80

8,2.

755

Fate O

dds

rati

o0.

988

0.94

62.

210

1.11

32.

402

2.46

18.

607

3.21

195

%C

I0.

759,

1.28

60.

685,

1.30

71.

571,

3.11

00.

622,

1.99

30.

575,

10.0

260.

458,

13.2

20.

387,

191.

60.

053,

195.

4Fa

te×

Tim

eO

dds

rati

o1.

304

1.47

51.

009

2.05

41.

248

0.70

20.

536

1.03

795

%C

I1.

107,

1.53

51.

074,

2.02

70.

939,

1.08

51.

181,

3.57

20.

473,

3.29

30.

235,

2.09

10.

138,

2.07

50.

067,

16.1

00

Tab

le1

(con

tinu

ed).

© 2007 NRC Canada

Caudill et al. 985

Dam

Bon

nevi

lle

The

Dal

les

John

Day

McN

ary

Ice

Har

bor

Low

erM

onum

enta

lL

ittl

eG

oose

Low

erG

rani

te

Sex O

dds

rati

o0.

981

1.07

61.

108

1.03

80.

790

0.97

81.

074

1.14

295

%C

I0.

872,

1.10

30.

945,

1.22

50.

972,

1.26

40.

882,

1.22

00.

589,

1.05

90.

720,

1.32

90.

766,

1.50

60.

819,

1.59

2Ta

gda

yO

dds

rati

o1.

012

0.99

51.

007

1.01

71.

009

0.99

51.

051

1.01

995

%C

I0.

991,

1.03

50.

969,

1.02

10.

990,

1.02

51.

004,

1.03

00.

985,

1.03

40.

976,

1.01

60.

950,

1.16

20.

977,

1.06

2C

lipp

edO

dds

rati

o1.

000

1.02

90.

968

0.98

10.

957

1.16

80.

771

0.87

695

%C

I0.

884,

1.13

10.

895,

1.18

30.

839,

1.11

80.

829,

1.16

00.

709,

1.29

10.

858,

1.58

90.

532,

1.11

80.

620,

1.23

9L

engt

hO

dds

rati

o0.

987

0.98

60.

986

0.98

50.

995

0.97

60.

984

0.98

195

%C

I0.

981,

0.99

30.

980,

0.99

20.

981,

0.99

20.

978,

0.99

30.

973,

1.01

80.

959,

0.99

40.

964,

1.00

50.

959,

1.00

4

Fal

lC

hino

oksa

lmon

N23

9815

3514

9112

2223

114

610

463

Cen

sore

d18

614

520

510

742

136

6%

Cen

sore

d7.

89.

413

.78.

818

.28.

95.

89.

5U

nsuc

cess

ful

539

248

198

8425

1812

9%

Uns

ucce

ssfu

l22

.516

.213

.36.

910

.812

.311

.514

.3E

stim

ate

TO

D Odd

sra

tio

0.11

10.

088

0.09

00.

067

0.08

10.

080

0.07

30.

170

95%

CI

0.09

3,0.

133

0.06

9,0.

111

0.07

0,0.

115

0.05

0,0.

089

0.04

1,0.

161

0.03

5,0.

185

0.02

7,0.

198

0.05

0,0.

577

Flo

w Odd

sra

tio

0.99

80.

995

0.99

50.

999

0.98

41.

000

1.08

51.

060

95%

CI

0.99

5,1.

001

0.99

0,1.

000

0.99

0,1.

000

0.99

3,1.

004

0.92

1,1.

051

0.91

8,1.

090

0.95

6,1.

232

0.86

9,1.

295

Tem

pera

ture

Odd

sra

tio

1.07

61.

111

1.12

71.

045

0.92

30.

995

0.95

11.

051

95%

CI

1.00

1,1.

157

0.93

7,1.

318

0.98

0,1.

295

0.97

9,1.

116

0.72

5,1.

174

0.81

5,1.

214

0.74

3,1.

218

0.82

3,1.

342

Spi

ll Odd

sra

tio

NA

NA

NA

NA

NA

NA

NA

NA

95%

CI

——

——

——

——

Sal

mon

Odd

sra

tio

1.00

91.

008

1.01

21.

005

1.01

61.

008

1.00

01.

249

95%

CI

0.96

8,1.

052

0.99

0,1.

026

0.98

7,1.

037

0.98

6,1.

026

0.94

3,1.

096

0.91

4,1.

112

1.00

0,1.

000

1.03

1,1.

513

Sha

d Odd

sra

tio

0.00

031

9.7

0.02

72.

775

>10

000

0.00

01.

174

0.00

095

%C

I0.

000,

>10

000

0.02

2,>

1000

00.

000,

265.

80.

011,

717.

20.

000,

>10

000

0.00

0,>

1000

00.

903,

1.52

60.

000,

>10

000

Rel

ease

site

Odd

sra

tio

NA

1.19

41.

041

0.99

01.

493

0.93

71.

764

0.44

495

%C

I—

0.99

1,1.

439

0.86

1,1.

258

0.80

6,1.

216

0.94

7,2.

354

0.50

9,1.

722

0.69

3,4.

491

0.16

5,1.

193

Tab

le1

(con

tinu

ed).

© 2007 NRC Canada

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

Dam

Bon

nevi

lle

The

Dal

les

John

Day

McN

ary

Ice

Har

bor

Low

erM

onum

enta

lL

ittl

eG

oose

Low

erG

rani

te

Fate O

dds

rati

o1.

562

1.51

12.

100

2.25

41.

385

2.21

10.

700

1.17

595

%C

I1.

280,

1.90

51.

242,

1.83

91.

491,

2.95

91.

552,

3.27

50.

653,

2.93

70.

847,

5.77

10.

175,

2.79

70.

152,

9.07

0Fa

te×

Tim

eO

dds

rati

o1.

043

0.96

81.

030

1.03

60.

613

1.03

81.

286

0.95

795

%C

I0.

999,

1.08

90.

909,

1.03

10.

983,

1.07

90.

893,

1.20

20.

295,

1.27

60.

812,

1.32

60.

348,

4.75

70.

572,

1.60

3S

ex Odd

sra

tio

0.99

01.

099

0.99

51.

013

1.26

40.

985

1.15

40.

592

95%

CI

0.88

9,1.

103

0.95

8,1.

261

0.86

0,1.

150

0.86

7,1.

184

0.82

9,1.

925

0.60

0,1.

618

0.58

9,2.

261

0.23

4,1.

493

Tag

day

Odd

sra

tio

0.99

51.

013

1.01

10.

996

1.01

60.

991

1.03

11.

000

95%

CI

0.98

3,1.

006

0.99

2,1.

034

0.99

1,1.

032

0.98

4,1.

007

0.99

2,1.

042

0.96

4,1.

019

0.99

6,1.

067

0.94

6,1.

058

Cli

pped

Odd

sra

tio

1.14

71.

007

1.00

80.

732

1.26

10.

682

0.35

60.

282

95%

CI

0.95

8,1.

375

0.79

0,1.

284

0.77

2,1.

315

0.56

1,0.

956

0.76

4,2.

082

0.34

8,1.

337

0.10

8,1.

180

0.04

7,1.

714

Len

gth

Odd

sra

tio

0.99

50.

993

0.99

30.

995

0.97

60.

989

0.95

00.

978

95%

CI

0.98

9,1.

000

0.98

6,1.

000

0.98

5,1.

000

0.98

7,1.

003

0.95

4,0.

998

0.96

2,1.

017

0.91

4,0.

986

0.93

3,1.

024

Stee

lhea

d(s

ea-r

unO

ncor

hync

hus

myk

iss)

N37

0229

7228

5120

2920

5615

8311

5711

79C

enso

red

174

8824

622

714

248

5973

%C

enso

red

4.7

3.0

8.6

11.2

6.9

3.0

5.1

6.2

Uns

ucce

ssfu

l93

260

558

233

925

416

190

78%

Uns

ucce

ssfu

l25

.220

.420

.416

.712

.410

.27.

86.

6E

stim

ate

TO

D Odd

sra

tio

0.09

00.

126

0.14

90.

160

0.12

40.

163

0.23

50.

089

95%

CI

0.07

7,0.

105

0.11

0,0.

144

0.13

1,0.

170

0.13

8,0.

185

0.10

6,0.

144

0.14

0,0.

191

0.20

1,0.

276

0.07

1,0.

112

Flo

w Odd

sra

tio

1.00

10.

998

0.99

80.

997

0.99

80.

996

0.99

81.

003

95%

CI

0.99

7,1.

006

0.99

6,1.

000

0.99

5,1.

001

0.99

3,1.

000

0.97

9,1.

016

0.98

8,1.

003

0.98

6,1.

009

0.99

5,1.

012

Tem

pera

ture

Odd

sra

tio

0.97

90.

962

1.03

51.

061

1.06

61.

043

1.04

31.

049

95%

CI

0.93

6,1.

023

0.93

5,0.

990

1.01

1,1.

059

1.02

1,1.

102

1.02

7,1.

106

1.02

3,1.

065

1.01

5,1.

071

1.02

2,1.

077

Spi

ll Odd

sra

tio

0.97

41.

008

0.95

81.

007

0.84

00.

924

0.83

61.

050

95%

CI

0.94

7,1.

001

0.96

5,1.

054

0.90

8,1.

011

0.94

1,1.

077

0.76

1,0.

927

0.58

8,1.

451

0.53

5,1.

306

0.83

7,1.

317

Sal

mon

Odd

sra

tio

1.00

21.

014

1.01

91.

023

1.04

61.

023

1.00

01.

082

95%

CI

0.93

5,1.

074

1.00

5,1.

023

1.00

2,1.

036

0.98

7,1.

059

0.95

7,1.

143

0.97

9,1.

068

1.00

0,1.

000

1.00

2,1.

168

Tab

le1

(con

tinu

ed).

© 2007 NRC Canada

Caudill et al. 987

Dam

Bon

nevi

lle

The

Dal

les

John

Day

McN

ary

Ice

Har

bor

Low

erM

onum

enta

lL

ittl

eG

oose

Low

erG

rani

te

Sha

d Odd

sra

tio

7.38

91.

002

1.00

51.

018

1.00

30.

972

1.00

00.

989

95%

CI

0.79

3,68

.870

0.99

6,1.

007

0.99

8,1.

012

0.98

0,1.

059

0.94

2,1.

067

0.90

9,1.

039

1.00

0,1.

000

0.35

4,2.

759

Rel

ease

site

Odd

sra

tio

NA

1.06

61.

122

1.01

71.

034

1.03

90.

960

1.14

195

%C

I—

0.95

0,1.

196

0.98

4,1.

278

0.87

9,1.

177

0.89

1,1.

201

0.88

3,1.

223

0.79

8,1.

155

0.91

7,1.

419

Fate O

dds

rati

o1.

182

1.07

31.

347

1.17

51.

405

1.33

31.

817

2.22

695

%C

I1.

070,

1.30

50.

952,

1.20

81.

188,

1.52

60.

949,

1.45

41.

131,

1.74

61.

016,

1.74

91.

350,

2.44

51.

543,

3.21

1Fa

te×

Tim

eO

dds

rati

o1.

049

1.04

61.

008

1.00

50.

970

1.25

90.

996

1.00

895

%C

I1.

025,

1.07

40.

996,

1.09

91.

000,

1.01

50.

980,

1.03

00.

926,

1.01

51.

075,

1.47

40.

964,

1.02

90.

968,

1.04

9S

ex Odd

sra

tio

1.05

21.

107

1.05

71.

024

1.05

11.

026

1.07

91.

053

95%

CI

0.97

5,1.

136

1.01

4,1.

207

0.96

7,1.

155

0.91

3,1.

149

0.94

5,1.

170

0.91

2,1.

153

0.93

6,1.

244

0.91

5,1.

213

Tag

day

Odd

sra

tio

1.00

61.

004

1.00

11.

008

1.00

61.

003

1.00

41.

004

95%

CI

1.00

3,1.

009

1.00

1,1.

007

0.99

8,1.

004

1.00

2,1.

013

1.00

1,1.

010

0.99

7,1.

008

0.99

8,1.

009

0.99

6,1.

011

Cli

pped

Odd

sra

tio

1.03

51.

017

1.00

61.

032

0.98

51.

020

1.04

50.

929

95%

CI

0.94

2,1.

139

0.91

6,1.

128

0.90

9,1.

114

0.90

0,1.

183

0.86

3,1.

124

0.88

6,1.

173

0.89

1,1.

226

0.78

4,1.

100

Len

gth

Odd

sra

tio

1.00

20.

997

1.01

20.

998

1.00

21.

006

1.00

31.

003

95%

CI

0.99

6,1.

009

0.99

0,1.

005

1.00

6,1.

018

0.98

9,1.

006

0.99

5,1.

009

0.99

8,1.

015

0.99

4,1.

012

0.99

4,1.

013

Not

es:

Odd

sra

tios

conf

iden

cein

terv

als

that

dono

tin

clud

e1.

0ar

egi

ven

inbo

ld.

Odd

sra

tios

estim

ate

pass

age

haza

rd(t

hepr

obab

ility

ofpa

ssag

edu

ring

any

give

ntim

ein

terv

al)

for

one

grou

pco

m-

pare

dw

ithan

othe

r.R

atio

sar

eex

pres

sed

assu

cces

sful

:uns

ucce

ssfu

l,ni

ght:d

ay,

dow

nstr

eam

:for

ebay

,m

ale:

fem

ale,

hatc

hery

:wild

.Fo

rex

ampl

e,od

dsra

tios

>1.

0fo

rfa

tein

dica

tea

high

erpa

ssag

eha

zard

(and

low

erpa

ssag

etim

e)fo

rsu

cces

sful

vers

usun

succ

essf

ulad

ults

.O

dds

ratio

sfo

rco

ntin

uous

pred

icto

rva

riab

les

estim

ate

chan

gein

haza

rdpe

run

itin

crea

sein

the

pred

icto

r.Fl

owan

dsp

illar

eex

-pr

esse

das

chan

gein

odds

per

283

m3 ·s

–1(1

000

0cu

bic

feet

per

seco

nd),

tem

pera

ture

asch

ange

inod

dspe

r°C

,sa

lmon

ids

and

shad

asch

ange

per

1000

0fi

shin

daily

ladd

erco

unt.

TO

D,

time

ofda

y(n

ight

vs.

day)

;sh

ad,

daily

dam

coun

tof

Am

eric

ansh

ad(A

losa

sapi

diss

ima)

;sa

lmon

,da

ilyda

mco

unts

ofal

lsa

lmon

ids

(Onc

orhy

nchu

ssp

p.);

NA

,no

tap

plic

able

.

Tab

le1

(con

clud

ed).

not be directly estimated using PROC PRHEG (Allison1995, p. 160). This analytical limitation also prevented testsof year × fate interactions.

We used information-theoretic techniques (Burnham andAnderson 2002) to compare among the 15 potential modelsfor each species–dam combination. Akiake’s informationcriterion (AIC) and AIC weights (wi) were used to identifythe most parsimonious model(s), given the data. Model-averaging techniques were used to calculate model-averagedodds ratios and associated unconditional confidence intervals(CIs) that incorporate uncertainty about which model is best(Burnham and Anderson 2002). Details and results of modelselection and averaging are given in Appendix S13. We con-sidered predictor variables to have explanatory power whenthe 95% CI of the odds ratio did not include 1.0.

On average, 6.7% (range: 1.3%–18.2%) of passage eventshad uncertain endpoints because of missing telemetry re-cords at the fish ladder exit antenna. These passage eventswere censored from the analysis by removing them from therisk set at the time of last observation. Operational reasonsfor incomplete records included antenna outages or passagethrough an unmonitored navigational lock. Alternatively, in-dividuals were censored because they entered a tailrace butdid not pass the dam. We assigned censoring times using thelast available record from antennas at the dam, and thesewere often at antennas near the ladder exit. We tested the ef-fect of this criterion using the sensitivity analyses describedbelow.

A critical assumption of time-event models is that censor-ing must not be informative with respect to treatment groups(Allison 1995; Hosmer and Lemeshow 1999). We observedthat fish with unknown fates were more likely to be cen-sored, violating this assumption (see Results). While the ef-fect of informative censoring on model coefficients cannotbe calculated (Allison 1995), the general effect can be as-sessed by performing sensitivity analyses where censoringtimes are assigned using different criteria and comparing theresulting estimated regression coefficients with those of thebase model. If the coefficients do not change markedly, theeffect of this violation on model conclusions is deemedsmall (Allison 1995). We used three alternative criteria inour analyses. First, we assumed that all censored fish wouldhave passed immediately after the censoring time by includ-ing these fish as uncensored and using the censoring time asthe time of ladder exit. Second, we assumed that all cen-sored fish could have remained in the risk set until the endof observation by assigning censor times equal to the longestobserved passage time. In a third analysis, we excluded apriori all fish that we did not observe to pass the dam underanalysis to directly test for any delayed upstream effects ofslow passage on fate among those fish that were known tohave passed each dam.

Total passage time in the lower Columbia Riverhydrosystem

We were also interested in the relationship between mi-gration performance and passage time past multiple dams.Specifically, we asked whether fish migration times throughthe lower Columbia River were longer for unsuccessful thanfor successful migrants, as observed in sockeye salmon(Oncorhynchus nerka) by Naughton et al. (2005). We used

the analysis of variance (ANOVA) model loge(migrationtime) = year + fate + year × fate + error to test for differ-ences in mean passage time by fate and year. Migrationtimes for each fish were calculated between the first detec-tion in the Bonneville tailrace to exit from a McNary fishladder (four dams and three reservoirs). By definition, thisanalysis excluded forebay-released fish and any fish withouta ladder exit record at McNary Dam. Passage times wereloge-transformed to improve the normality of error terms.We also calculated the relative mean difference in passagetime between fate classes for each year as (Timeunsuccessful –Timesuccessful)/Timesuccessful and then tested whether the meandifference among years differed significantly between fateclasses using one-way ANOVA. All statistical analyses wereperformed in SAS v.9 PROC PHREG or GLM (SAS Insti-tute Inc., Cary, North Carolina).

Results

Passage at individual damsMost individual fish passed each dam rapidly, and median

passage times at individual dams ranged from 0.2 to2.7 days, depending on the species and year (Fig. 2). How-ever, up to 57% of spring Chinook salmon took more than2 days and as many as 32% took more than 5 days to passindividual dams each year. Summer and fall Chinook salmonpassed more quickly than spring Chinook salmon, a patternthought to be related to temperature effects on swimmingspeed (Brett 1995) and seasonal differences in discharge(Keefer et al. 2004a). Most steelhead passed during low-discharge conditions and also rapidly passed individualdams.

The results of PHReg multimodel selection using AIC re-vealed that the full model provided the best fit to the data inthe majority of species–dam combinations, though in severalcases there was considerable uncertainty about which modelprovided the best fit to the data (e.g., all model wi < 0.4,Supplemental Appendix S13). Nonetheless, model coeffi-cients remained nearly identical among models (Supplemen-tal Appendix S13). There were important differences amongyears in passage behavior that appeared to be primarilyrelated to annual variation in mean river flow. Though wewere unable to directly assess the effects of interannualvariation on passage hazard, models not stratified by yearhad the worst AIC score in all cases (Supplemental Appen-dix S13).

Several environmental predictors provided explanatorypower for one or more runs within year. Odds ratios com-pared the probability of passage between categories withinany given time interval, or the change in odds as predictorvariables increased by one unit. Time of day was the stron-gest and most consistent effect within year across all runsand dams because few fish passed at night. Individual fishwere one-fifth to less than one-twentieth as likely to passduring night as day (Table 1). When different from even,odds ratios for flow were always less than 1.0 and indicateda 0.2%–0.67% decrease in the instantaneous probability ofpassage per 283 m3·s–1 (10 000 cubic feet per second) in-crease of flow. Spill had a similar, but stronger effect, withpassage hazard decreasing 5.1%–22.1% per 283 m3·s–1. In-creasing temperature was generally associated with faster

© 2007 NRC Canada

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

© 2007 NRC Canada

Caudill et al. 989

passage, particularly for spring Chinook salmon, which passprior to the onset of stressful temperature conditions (e.g.,≤18 °C; Table 1). Odds ratios for date of tagging indicatedthat passage hazard increased 0.4–5.1%·day–1 within yearduring the run. This pattern is consistent with other findingsthat individuals migrating late within runs tend to migratefaster, presumably to reach spawning grounds during the ap-propriate spawning period (e.g., Quinn and Adams 1996;McLean et al. 2004). In some cases, the number ofsalmonids present in the ladder was positively associatedwith passage hazard, revealing that higher densities of adultsalmonids within ladders did not slow individual passage.Conversely, at the lower Columbia River dams where largenumbers of American shad pass, passage hazards were nega-tively associated with shad numbers for spring Chinooksalmon at all dams except Bonneville Dam. Passage hazardfrequently decreased with increasing fish size in the Chi-nook salmon runs and was not consistently related to sex.Fish released downstream from Bonneville Dam passedmore rapidly at The Dalles and John Day dams than thosereleased to the Bonneville Dam forebay for most runs. Therewas no consistent association between fish origin (hatcheryvs. wild) and passage hazard.

The time-event analyses were primarily structured to testfor an association between passage time and eventual migra-tion to spawning sites, while statistically controlling for theeffects of environmental variables and fish traits. Fate wasassociated with passage hazard or there was a significant in-

teraction between fate and passage hazard for 22 out of 24(91.7%) of dam and run combinations, excluding summerand fall Chinook salmon at Snake River dams, where sam-ples sizes were small (Nunsuccessful < 25; Table 1). The oddsratio for fate ranged from 1.347 to 2.254 in 11 cases wherethere was no significant interaction between fate and pas-sage time, indicating that eventually successful fish were34.7%–225.4% more likely than unsuccessful fish to passduring any given time interval. In one-half of the dam–runcombinations (12 of 24, excluding the small samples atSnake River dams), there was a significant fate × passagetime interaction. In all cases, odds ratios were greater than1.0, indicating that successful fish became increasingly morelikely to pass during any given interval as the time sincetailrace detection increased compared with unsuccessful fish.

Fish that were eventually unsuccessful were censored sig-nificantly more frequently than successful fish in all fourruns (Supplemental Appendix S23), violating an assumptionof the Cox proportional hazard regression. The results ofthree sensitivity analyses differing in their handling of cen-sored fish suggested this violation did not seriously bias thePHReg parameters nor provide false evidence of a relation-ship between passage time and fate (Supplemental AppendixS23).

Total passage time through the lower Columbia Riverhydrosystem

Among Chinook salmon, summer Chinook were fastest

Fig. 2. Inverse cumulative passage time curves for each run analyzed by fate, all years and dams combined: (a) spring Chinook salmon(Oncorhynchus tshawytscha), (b) summer Chinook salmon, (c) fall Chinook salmon, and (d) steelhead (sea-run Oncorhynchus mykiss).Note differences in time scale. The passage time curves for unsuccessful migrants are depicted with solid lines; successful migrants areshown with broken lines. Censoring times for individual adults are indicated by open circles.

and spring Chinook had the slowest passage times throughthe lower Columbia River hydrosystem. In contrast withpassage times at individual dams, steelhead passage timesthrough the lower Columbia River hydrosystem were ap-proximately double those for Chinook salmon (Fig. 3), re-flecting extended temporary straying by steelhead intocooler tributaries during the warmest months (see Keefer etal. 2004a and High et al. 2006 for more details).

Averaged across all years, mean passage times from theBonneville tailrace to the McNary Dam forebay (235 rkmand four dams) were faster for successful migrants than forunsuccessful migrants that passed McNary Dam but did notreach spawning tributaries (spring Chinook salmon: F[13,2125] =41.73, Pfate = 0.0190, Pyear < 0.0001, Pfate × year = 0.0655;summer Chinook salmon: F[13,1348] = 10.42, Pfate < 0.0001,Pyear 0.0015, Pfate × year = 0.4059; fall Chinook salmon:F[9,1329] = 4.64, Pfate = 0.0231, Pyear 0.0198, Pfate × year =0.1657; steelhead: F[11,2433] = 7.43, Pfate = 0.0451, Pyear <0.0001, Pfate × year = 0.0645; Fig. 3). The observed point es-timates for differences between fate groups varied fromyear to year in all runs except summer Chinook salmon,

with some weak evidence of a fate × year interaction in springChinook salmon and steelhead (P ≤ 0.0655). The mean annualrelative difference in lower Columbia River passage time be-tween fate categories was also significant for all four runs, withunsuccessful passage times 4.2%–16.1% longer than those foreventually successful migrants (0.0001 < P ≤ 0.0451; Supple-mental Appendix S33). These estimates may have underesti-mated the true differences in passage time between fate groupsbecause unsuccessful adults not passing McNary Dam couldnot be included in the analysis.

Discussion

The migration success of adult Chinook salmon andsteelhead was inversely related to passage time through thedams and reservoirs of the Columbia River hydrosystem, ac-cording to our detailed analysis of radiotelemetric data. Weobserved that the majority of adult salmonids rapidly passedas many as eight dams and successfully reached their natalspawning sites. However, many fish required more than2 days to pass individual dams, some took several weeks to

© 2007 NRC Canada

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

Fig. 3. Mean passage time by fate for each species and year: (a) spring Chinook salmon (Oncorhynchus tshawytscha), (b) summerChinook salmon, (c) fall Chinook salmon, and (d) steelhead (sea-run Oncorhynchus mykiss). Unsuccessful adults are indicated by solidbars; successful adults are shown by shaded bars. Note the difference in the y axis for steelhead; passage times were approximatelytwo times longer than Chinook salmon passage times. Sample sizes given above each bar. *, p < 0.05; **, p < 0.01; ***, p < 0.001.

pass, and some did not pass at all. Adults with unknownfates — those that were not detected at spawning sites afterpassing one or more dams — had consistently longer pas-sage times at individual dams and through a multidam reachof the lower Columbia River. These patterns are striking, butcannot demonstrate an underlying mechanism(s), nor do therelationships reveal the relative importance of dams versusother factors in reducing the migration success of slowadults. However, the patterns in passage time do suggestseveral potential mechanisms that may have contributed tothe observed relationships.

Passage environment and fish traitsWithin season, passage hazards were related to time of

day, flow, spill, water temperature, densities of adult fishesin ladders, and fish length. The single strongest predictorwas time of day, because salmonids infrequently pass damsat night (Naughton et al. 2005). Slower passage at highflows probably reflected both decreased ground speed of fishswimming through higher velocity water (e.g., Hinch andRand 1998) and an increase in the searching time required tofind fishway entrances in a more turbulent tailrace environ-ment. Water was passed over dam spillways during springand summer in most years, dramatically increasing turbu-lence in tailraces and potentially accounting for the strongerassociation of passage hazard with spill than with flow. De-tailed analyses of an experiment that manipulated spill levelsin the lower Columbia River during 2000–2003 were consis-tent with those reported here (Caudill et al. 2006a).

Water temperature has dramatic effects on salmon physi-ology and behavior (Brett 1995). Cool, early spring tempera-tures (5–8 °C) compared with near optimal late springtemperatures (~15–17 °C; Brett 1995; Salinger and Ander-son 2006) probably explain the positive association betweentemperature and passage hazard for spring Chinook salmonat lower Columbia River dams. Fall Chinook salmon andsteelhead encounter warm, and in some cases stressful, tem-peratures (18–23 °C; Richter and Kolmes 2005) during Au-gust and early September and then cooler temperatures asthe run season progresses.

The consistent, negative relationship between body lengthand passage hazard was somewhat surprising because opti-mal swimming speed increases with body size (Brett 1995).However, the factors influencing optimum swim speed, mi-gration efficiency, and body size are complex (Hinch andRand 1998; Crossin et al. 2004a; Hughes 2004). Indeed,interpopulation patterns of body size suggest selection forsmaller size in long-distance migrating stocks (reviewed inQuinn 2005), though how this would relate to swim speedmeasured over smaller scales remains unclear. Clearly, thestudy of the relationships among temperature, body size,swimming speed, and river environment deserves furtherstudy.

Interestingly, passage hazards increased, corresponding toshorter passage times, as the number of concurrently passingadult salmonids increased in some models, particularly forSnake River spring Chinook salmon. This result provides noevidence for negative, density-dependent effects and mayhave resulted from the fact that the peak run presumably oc-curs during the period of optimal migration conditions, re-sulting in large numbers of fish moving quickly compared

with early or late in the run. Conversely, the number ofAmerican shad was negatively associated with passage haz-ard in spring Chinook salmon at three out of four lower Co-lumbia River dams. This result suggests the potential forhigh densities of American shad to slow the migration ofsalmonids, but could be related to the fact that shad only be-gin passing Bonneville Dam after most spring Chinooksalmon have passed (DART 2005).

Potential underlying mechanisms affecting the passagetime – fate relationship

An important question is whether the passage time – faterelationship was caused by impoundment or whether the pat-tern was primarily related to the rigors of long-distance mi-gration and would have been observed in the unmodifiedsystem. In the transformed Columbia–Snake hydrosystem,rapid passage of low-velocity reservoirs appears to compen-sate for relative slow passage through high-velocity tailracesand over dams (Keefer et al. 2004a; Naughton et al. 2005).Temperature conditions appear to be elevated in the modi-fied system because of the effects of regional climate warm-ing, changing land- and water-use patterns, and the directeffects of impoundment (Quinn and Adams 1996; Quinn etal. 1997), potentially increasing the metabolic costs of mi-gration. Spill frequently elevates dissolved gas levels abovesaturation and can cause gas bubble disease at extremesupersaturation levels (Johnson et al. 2005), and high spilllevels are associated with slowed passage (Caudill et al.2006a). In constrast, impoundment may reduce migrationcosts by reducing peak spring flows and because current fishpassage facilities may be less challenging in terms of veloc-ity barriers than the cascades and rapids of the predam mi-gration corridor. Historically, migration up the lowerColumbia River included dramatic obstacles, such as Cas-cade Rapids and the 6 m high Celilo Falls, which are bothnow inundated. Below, we outline four causal mechanismsthat may have contributed to the observed passage time –fate relationship.

First, environmental conditions during migration probablycontributed to the passage time – fate relationship. Factorsthat increase the time spent in dam tailraces and fishways,including high spill and flow levels (Caudill et al. 2006a),fallback behavior (Boggs et al. 2004), temperature (Caudillet al. 2006b; Goniea et al. 2006; High et al. 2006), and hy-draulics at the base of ladders (Naughton et al. 2006), maybe particularly important to the overall energetics of migra-tion and migration success because passage of these areas isenergetically demanding (Brown and Geist 2002; Brown etal. 2006). Generally, impoundment, hydrosystem operations,and natural climate drivers affect the environmental condi-tions encountered by migrating fishes in regulated rivers inways that probably influence migration success, including inthis study. The analysis of passage time data using thePHReg approach provides powerful, dynamic statistical con-trol of measured environmental variation when comparingfate classes, though we note that additional, unmeasured en-vironmental variation in this study that affected both passagetime and fate may have contributed to the observed patterns.

Second, variation in passage behavior among individualsmay have contributed to the observed passage time – fate re-lationship. Individual responses to complex hydraulic, tem-

© 2007 NRC Canada

Caudill et al. 991

perature, light, olfactory, and other cues during migrationare variable (Standen et al. 2004; Caudill et al. 2006b;Keefer et al. 2006). Hinch and Bratty (2000) observed thatsome sockeye salmon adults exhibited nonoptimal rapidswimming behaviors in high-flow, turbulent areas at Hell’sGate on the Fraser River. Rapid swimming was also associ-ated with unsuccessful passage. Thus, the long passagetimes and unsuccessful migration of some adults may havebeen caused by inefficient responses to the passage condi-tions encountered at dams. Whether relatively slow passageand unsuccessful migration were caused by the artificialconditions at dams or reflect underlying natural variation inbehavior remains unknown.

Third, the relationship may represent selection for physio-logical and locomotory traits needed to traverse velocity bar-riers at natural falls and artificial channel constrictions.Passage of such velocity barriers is physiologically demand-ing (Hinch and Rand 1998; Hinch and Bratty 2000; Standenet al. 2002), slowing the upstream migration of fishes in theColumbia–Snake hydrosystem (Williams 1998; Keefer et al.2004a) and in other salmonid (Karppinen et al. 2002; Laineet al. 2002) and non-salmonid fishes (e.g., Moser et al.2002a; Zigler et al. 2004). Such barriers may be impassabledepending on species and individual fish traits and condition(Moser et al. 2002b; Haro et al. 2004; Reiser et al. 2006).Under this mechanism, relatively weak swimmers would re-quire greater periods to pass individual dams and also beless likely to successfully pass other velocity barriers up-stream, including other dams.

Fourth, the relationship may have resulted from variationamong individuals in energetic and (or) physiological condi-tion upon river entry, which is determined largely by growthand development in the ocean. Individual salmon differ con-siderably in total energy content or density upon river entryboth within (Pinson 2005) and among years (Crossin et al.2004b). Adults in poor initial condition, especially thosewith relatively low energy content, may travel more slowlythrough high-gradient reaches and be less likely to completemigration. Young et al. (2006) found that among late-runFraser River sockeye salmon entering fresh water abnor-mally early, unsuccessful migrants had lower gross somaticenergy, higher plasma lactate (indicating recent anaerobicrespiration), and higher levels of reproductive hormones,suggesting that unsuccessful individuals had a smaller initialenergetic buffer against stress.

Notably, there is potential for these mechanisms to inter-act. Relatively slow migration past any project could depleteenergy reserves, potentially alter subsequent fish behaviorand orientation ability, reduce the ability to ascend laddersand traverse velocity barriers, and thereby increase the po-tential for further slowed and (or) failed migration at up-stream dams. Poor initial condition (Crossin et al. 2004a)and (or) environmental conditions (Rand and Hinch 1998)during migration could further increase the probability of afish entering this cycle. Ocean conditions prior to river entrywere considered relatively good during the study period(e.g., Scheuerell and Williams 2005), suggesting reducedmigration success and perhaps higher odds ratios for thepassage time – fate relationship following a downturn inocean conditions.

We were able to assess migration success to spawningtributaries and note that relatively slow migration in thehydrosystem may have reduced reproductive success onspawning grounds as well. Migration experience can influ-ence reproductive success (e.g., Patterson et al. 2004), adultswith longer migration times consume a greater proportion ofenergy reserves to reach spawning tributaries (Pinson 2005),and adults may hold for periods of weeks to months prior tospawning. Consequently, low initial energetic content andenergetic depletion during migration have been implicated asa contributing factor for prespawn mortality in tributaries(Gilhousen 1990; Cooke et al. 2004; Pinson 2005), an areaof increasing concern given high prespawn mortality ratesobserved in some sockeye and Chinook salmon populationsin recent years (Cooke et al. 2004; Pinson 2005).

These results represent one of the first large-scale exami-nations of the relationship between migration behavior andfate in an anadromous fish. Anadromous and catadromousfishes are thought to be particularly vulnerable to human-induced environmental change because they must have suit-able migration corridors in addition to suitable breeding,growth, and (or) overwintering–holding habitats, and be-cause river systems worldwide have been strongly altered bydamming and other human activities (Postel and Richter2003). Long-distance migrating stocks, such as the interiorsalmonid stocks studied here, may be at particular risk giventhe relatively high cost of migration, low reserves uponreaching spawning sites, long prespawn holding periods, andmarginal thermal regime encountered during migration bysome migrants. Clearly, the successful management of habi-tats for long-distance migrants, especially diadromousfishes, will require mechanistic knowledge of how initialtraits and condition, migration behavior, and environmentalconditions interact to determine migration performance andreproductive success. The results also illustrate the impor-tance of assessing the indirect effects of dams, including de-layed or carry-over effects, because successful completion ofmigration was clearly associated with passage times and be-haviors at downstream dams. Because of logistical and mon-etary constraints, the vast majority of studies examining theeffects of dams on fishes has been conducted at local scalesand has focused on short-term phenomena such as behaviorat passage structures (i.e., survival past an individual pro-ject). Efforts should be made to increase the use of fitness-based performance measures, rather than behavioral orshort-term survival criteria, for setting standards for passagefacilities and in-river conditions, because long-term popula-tion viability ultimately rests on mean fitness remaining ator above replacement.

Acknowledgements

Many, many people provided time and assistance duringthe course of this study. K. Tolotti, R. Ringe, S. Lee,T. Clabough, E. Johnson, M. Morash, T. Dick, C. Morat, M.Feeley, P. Keniry, and B. Hastings helped with field opera-tions and collection and processing of telemetry data at theUniversity of Idaho. D. Joosten, C. Nauman, and C. Wil-liams assited with the coding of telemetry records and per-formed the detailed review of records for unknown fate fish.

© 2007 NRC Canada

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

© 2007 NRC Canada

Caudill et al. 993

K. Frick, A. Matter, M. Moser, and L. Stuehrenberg ofNOAA Fisheries Service helped with data processing andmanagement. We thank the Chelan and Douglas county pub-lic utility districts for providing data from the mid-ColumbiaRiver. The US Army Corps of Engineers Portland and WallaWalla districts provided funding for this project, though theviews expressed here reflect those of the authors and do notnecessarily reflect those of the Corps. Additional fundingwas provided by Chelan and Douglas county public utilitydistricts.

References

Allison, P.D. 1995. Survival analysis using SAS: a practical guide.SAS Institute Inc., Cary, N.C.

Bell, M.C. 1991. Fisheries handbook of engineering requirementsand biological criteria. US Army Corps of Engineers, North Pa-cific Division, Portland, Ore.

Boggs, C.T., Keefer, M.L., Peery, C.A., and Bjornn, T.C. 2004.Fallback, reascension, and adjusted fishway escapement esti-mates for adult chinook salmon and steelhead at Columbia andSnake River dams. Trans. Am. Fish. Soc. 133: 932–949.

Brett, J.R. 1995. Energetics. In Physiological ecology of Pacificsalmon. Edited by C. Grott, L. Margolis, and W. Clarke. Univer-sity of British Columbia Press, Vancouver, B.C. pp. 3–68.

Brown, R.S., and Geist, D.R. 2002. Determination of swimmingspeeds and energetic demands of upriver migrating fall chinooksalmon (Oncorhynchus tshawytscha) in the Klickitat River. Re-port of Pacific Northwest National Laboratory to BonnevillePower Administration, Portland, Ore. Proj. No. 22063.

Brown, R.S., Geist, D.L., and Mesa, M.G. 2006. Use ofelectromyogram telemetry to assess swimming activity of adultspring Chinook salmon migrating past a Columbia River dam.Trans. Am. Fish. Soc. 135: 281–287.

Bunt, C.M., van Poorten, B.T., and Wong, L. 2001. Denil fishwayutilization patterns and passage of several warmwater speciesrelative to seasonal, thermal and hydraulic dynamics. Ecol.Freshw. Fish, 10(4): 212–219.

Burnham, K.P., and Anderson, D.R. 2002. Model selection andmultimodel inference: a practical information-theoretic ap-proach. 2nd ed. Springer-Verlag, New York.

Castro-Santos, T., and Haro, A. 2003. Quantifying migratory de-lay: a new application of survival analysis methods. Can. J. Fish.Aquat. Sci. 60(8): 986–996.

Caudill, C.C., Peery, C.A., Daigle, W.R., Jepson, M.A., Boggs,C.T., Bjornn, T.C., and Joosten, D. 2006a. Adult Chinooksalmon and steelhead dam passage behavior in response to ma-nipulated discharge through spillways at Bonneville Dam. Re-port for US Army Corps of Engineers, Portland District,Portland, Ore.

Caudill, C.C., Clabough, T.S., Naughton, G.P., Peery, C.A., andBurke, B.J. 2006b. Water temperatures in adult fishways atmainstem dams on the Snake and Columbia Rivers. Phase 2. Bi-ological effects. Report for US Army Corps of Engineers, WallaWalla District, Walla Walla, Wash.

Clay, C.H. 1995. Design of fishways and other fish facilities, sec-ond edition. Lewis Publishers, Boca Raton, Fla.

Cooke, S.J., Hinch, S.G., Farrell, A.P., Lapointe, M.F., Jones,S.R.M., Macdonald, J.S., Patterson, D.A., Healey, M.C., andVan Der Kraak, G. 2004. Abnormal migration timing and highen route mortality of sockeye salmon in the Fraser River, BritishColumbia. Fisheries, 29(2): 22–33.

Crossin, G.T., Hinch, S.G., Farrell, A.P., Higgs, D.A., Lotto, A.G.,

Oakes, J.D., and Healey, M.C. 2004a. Energetics and morphol-ogy of sockeye salmon: effects of upriver migratory distanceand elevation. J. Fish. Biol. 65(3): 788–810.

Crossin, G.T., Hinch, S.G., Farrell, A.P., Higgs, D.A., and Healey,M.C. 2004b. Somatic energy of sockeye salmon Oncorhynchusnerka at the onset of upriver migration: a comparison amongocean climate regimes. Fish. Oceanogr. 13(5): 345–349.

DART (Data Access in Real Time). 2005. Columbia River data ac-cess in real time [online]. Available from www.cbr.washington.edu/dart/dart.html [accessed June 2005; updated 13 April 2007].

Dudgeon, D. 2003. The contribution of scientific information tothe conservation and management of freshwater biodiversity intropical Asia. Hydrobiologia, 500: 295–314.

Geist, D.R., Abernethy, C.S., Blanton, S.L., and Cullinan, V.I.2000. The use of electromyogram telemetry to estimate energyexpenditure of adult fall chinook salmon. Trans. Am. Fish. Soc.129(1): 126–135.

Gilhousen, P. 1990. Prespawning mortalities of sockeye salmon inthe Fraser River system and possible causal factors. Int. Pac.Salmon Fish. Comm. Bull. 26.

Goniea, T.M., Keefer, M.L., Bjornn, T.C., Peery, C.A., Bennett,D.H., and Stuehrenberg, L.C. 2006. Behavioral thermoregulationand slowed migration by adult fall Chinook salmon in responseto high Columbia River water temperatures. Trans. Am. Fish.Soc. 135: 408–419.

Haro, A., and Kynard, B. 1997. Video evaluation of passage effi-ciency of American shad and sea lamprey in a modified Ice Har-bor fishway. N. Am. J. Fish. Manag. 17: 981–987.

Haro, A., Castro-Santos, T., Noreika, J., and Odeh, M. 2004.Swimming performance of upstream migrant fishes in open-channel flow: a new approach to predicting passage through ve-locity barriers. Can. J. Fish. Aquat. Sci. 61(9): 1590–1601.

Hart, D.D., and Poff, N.L. 2002. A special section on dam removaland river restoration. Bioscience, 52(8): 653–655.

High, B., Peery, C.A., and Bennett, D.H. 2006. Temporary strayingof Columbia River summer steelhead in coolwater areas and itseffect on migration rates. Trans. Am. Fish. Soc. 135: 519–528.

Hinch, S.G., and Bratty, J. 2000. Effects of swim speed and activ-ity pattern on success of adult sockeye salmon migrationthrough an area of difficult passage. Trans. Am. Fish. Soc.129(2): 598–606.

Hinch, S.G., and Rand, P.S. 1998. Swim speeds and energy use ofupriver-migrating sockeye salmon (Oncorhynchus nerka): roleof local environment and fish characteristics. Can. J. Fish.Aquat. Sci. 55(8): 1821–1831.

Hosmer, D., Jr., and Lemeshow, S. 1999. Applied survival analysis:regression modeling of time to event data. Wiley Interscience,New York.

Hughes, N.F. 2004. The wave-drag hypothesis: an explanation forsize-based lateral segregation during the upstream migration ofsalmonids. Can. J. Fish. Aquat. Sci. 61(1): 103–109.

Johnson, E.L., Clabough, T.S., Bennett, D.H., Bjornn, T.C., Peery,C.A., Caudill, C.C., and Stuehrenberg, L.C. 2005. Migrationdepths of adult spring and summer chinook salmon in the lowerColumbia and Snake Rivers in relation to dissolved gassupersaturation. Trans. Am. Fish. Soc. 134(5): 1213–1227.

Karppinen, P., Makinen, T.S., Erkinaro, J., Kostin, V.V.,Sadkovskij, R.V., Lupandin, A.I., and Kaukoranta, M. 2002. Mi-gratory and route-seeking behaviour of ascending Atlanticsalmon in the regulated River Tuloma. Hydrobiology, 483(1–3):23–30.

Keefer, M.L., Peery, C.A., Bjornn, T.C., Jepson, M.A., andStuehrenberg, L.C. 2004a. Hydrosystem, dam, and reservoir

passage rates of adult chinook salmon and steelhead in the Co-lumbia and Snake Rivers. Trans. Am. Fish. Soc. 133(6): 1413–1439.

Keefer, M.L., Peery, C.A., Jepson, M.A., Tolotti, K.R., Bjornn,T.C., and Stuehrenberg, L.C. 2004b. Stock-specific migrationtiming of adult spring–summer chinook salmon in the ColumbiaRiver basin. N. Am. J. Fish. Manag. 24(4): 1145–1162.

Keefer, Peery, M.L., C.A., Jepson, M.A., and Stuehrenberg, L.C.2004c. Upstream migration rates of radio-tagged adult chinooksalmon in riverine habitats of the Columbia River basin. J. FishBiol. 65(4): 1126–1141.

Keefer, M.L., Peery, C.A., Ringe, R.R., and Bjornn, T.C. 2004d.Regurgitation rates of intragastric radio transmitters by adultchinook salmon and steelhead during upstream migration in theColumbia and Snake Rivers. N. Am. J. Fish. Manag. 24(1): 47–54.

Keefer, M.L., Peery, C.A., Daigle, W.R., Jepson, M.A., Lee, S.R.,Boggs, C.T., Tolotti, K.R., and Burke, B.A. 2005. Escapement,harvest, and unknown loss of radio-tagged adult salmonids inthe Columbia River – Snake River hydrosystem. Can. J. Fish.Aquat. Sci. 62: 930–949.

Keefer, M.L., Caudill, C.C., Peery, C.A., and Bjornn, T.C. 2006.Route selection in a large river during the homing migration ofChinook salmon (Oncorhynchus tshawytscha). Can. J. Fish.Aquat. Sci. 63: 1752–1762.

Laine, A., Jokivirta, T., and Katopodis, C. 2002. Atlantic salmon,Salmo salar L., and sea trout, Salmo trutta L., passage in a regu-lated northern river — fishway efficiency, fish entrance and en-vironmental factors. Fish. Manag. Ecol. 9(2): 65–77.

Lucas, M., and Frear, P. 1997. Effects of a flow-gauging weir onthe migratory behvaior of adult barbel, a riverine cyprinid. J.Fish Biol. 50: 382–396.

McLean, J.E., Bentzen, P., and Quinn, T.P. 2004. Does size matter?Fitness-related factors in steelhead trout determined by geneticparentage assignment. Ecology, 85(11): 2979–2985.

Monk, B., Weaver, D., Thompson, C., and Ossiander, F. 1989. Ef-fects of flow and weir design on the passage behavior of Ameri-can shad and salmonids in an experimental fish ladder. N. Am.J. Fish. Manag. 9: 60–67.

Moser, M.L., Darazsdi, A.M., and Hall, J.R. 2000. Improving pas-sage efficiency of adult American shad at low-elevation damswith navigation locks. N. Am. J. Fish. Manag. 20(2): 376–385.

Moser, M.L., Ocker, P.A., Stuehrenberg, L.C., and Bjornn, T.C.2002a. Passage efficiency of adult pacific lampreys athydropower dams on the lower Columbia River, USA. Trans.Am. Fish. Soc. 131(5): 956–965.

Moser, M.L., Matter, A.L., Stuehrenberg, L.C., and Bjornn, T.C.2002b. Use of an extensive radio receiver network to documentPacific lamprey (Lampetra tridentata) entrance efficiency atfishways in the Lower Columbia River, USA. Hydrobiology,483(1–3): 45–53.

Moser, M.L., Zabel, R.W., Burke, B.A., Struehrenberg, L.C., andBjornn, T.C. 2005. Factors affecting adult Pacific lamprey pas-sage rates at hydropower dams: using “time to event” analysis ofradiotelemetry data. In Aquatic Telemetry: Advances and Appli-cations. Proceedings of the Fifth Conference on Fish Telemetry,9–13 June 2003, Ustica, Italy. Edited by M.T. Spedicato, G.Lembo and G. Marmulla. FAO–COISPA, Rome. pp. 61–70.

Naughton, G.P., Caudill, C.C., Keefer, M.L., Bjornn, T.C., andStruehrenberg, L.C. 2005. Late-season mortality during migra-tion of radio-tagged adult sockeye salmon (Onchorhynchusnerka) in the Columbia River. Can. J. Fish. Aquat. Sci. 62: 30–47.

Naughton, G.P., Caudill, C.C., Peery, C.A., Clabough, T.S., Jepson,M.A., Bjornn, T.C., and Stuehrenberg, L.C. 2006. Experimentalevaluation of fishway modifications on the passage behavior ofadult Chinook salmon and steelhead at Lower Granite Dam,Snake River, USA. River Res. Appl. 23: 99–111.

Northcote, T.G. 1998. Migratory behaviour of fish and its signifi-cance to movement through riverine fish passage facilities. InFish migration and fish bypass. Edited by M. Jungwirth, S.Schmutz, and S. Weiss. Fishing News Books, Oxford. pp. 3–18.

NRC (National Research Council). 1996. Upstream: salmon andsociety in the Pacific Northwest. National Academy Press,Washington, D.C.

Patterson, D.A., Macdonald, J.S., Hinch, S.G., Healey, M.C., andFarrell, A.P. 2004. The effect of exercise and captivity on energypartitioning, reproductive maturation and fertilization success inadult sockeye salmon. J. Fish Biol. 64(4): 1039–1059.

Pinson, A. 2005. Energy use, migration time, and spawning successof adult chinook salmon returning to the South Fork SalmonRiver in Central Idaho. M.Sc. thesis, Department of Fish andWildlife Resources, University of Idaho, Moscow, Idaho.

Poff, N.L., and Hart, D.D. 2002. How dams vary and why it mat-ters for the emerging science of dam removal. Bioscience, 52(8):659–668.

Postel, S., and Richter, B. 2003. Rivers for life: managing water forpeople and nature. Island Press, Washington, D.C.

Quinn, T.P. 2005. The behavior and ecology of Pacific salmon.American Fisheries Society, Bethesda, Md.

Quinn, T.P., and Adams, D.J. 1996. Environmental changes affect-ing the migratory timing of American shad and sockeye salmon.Ecology, 77(4): 1151–1162.

Quinn, T.P., Hodgson, S., and Peven, C. 1997. Temperature, flow,and the migration of adult sockeye salmon (Oncorhynchusnerka) in the Columbia River. Can. J. Fish. Aquat. Sci. 54(6):1349–1360.

Rand, P.S., and Hinch, S.G. 1998. Swim speeds and energy use ofupriver-migrating sockeye salmon (Oncorhynchus nerka): simu-lating metabolic power and assessing risk of energy depletion.Can. J. Fish. Aquat. Sci. 55(8): 1832–1841.

Reischel, T.S., and Bjornn, T.C. 2003. Influence of fishway place-ment on fallback of adult salmon at the Bonneville Dam on theColumbia River. N. Am. J. Fish. Manag. 23(4): 1215–1224.

Reiser, D.W., Huang, C.M., Beck, S., Gagner, M., and Jeanes, E.2006. Defining flow windows for upstream passage of adult ana-dromous salmonids at cascades and falls. Trans. Am. Fish. Soc.135(3): 668–679.

Richter, A., and Kolmes, S.A. 2005. Maximum temperature limitsfor chinook, coho, and chum salmon, and steelhead trout in thePacific Northwest. Rev. Fish. Sci. 13(1): 23–49.

Ruckelshaus, M.H., Levin, P., Johnson, J.B., and Kareiva, P.M.2002. The Pacific salmon wars: what science brings to the chal-lenge of recovering species. Annu. Rev. Ecol. Syst. 33: 665–706.

Salinger, D.H., and Anderson, J.J. 2006. Effects of water tempera-ture and flow on adult salmon migration swim speed and delay.Trans. Am. Fish. Soc. 135: 188–199.

Scheuerell, M.D., and Williams, J.G. 2005. Forecasting climate-induced changes in the survival of Snake River spring/summerChinook salmon Oncorhynchus tshawytscha. Fish. Oceanogr.14(6): 448–457.

Standen, E.M., Hinch, S.G., Healey, M.C., and Farrell, A.P. 2002.Energetic costs of migration through the Fraser River Canyon,British Columbia, in adult pink (Oncorhynchus gorbuscha) andsockeye (Oncorhynchus nerka) salmon as assessed by EMG te-lemetry. Can. J. Fish. Aquat. Sci. 59(11): 1809–1818.

© 2007 NRC Canada

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

Standen, E.M., Hinch, S.G., and Rand, P.S. 2004. Influence of riverspeed on path selection by migrating adult sockeye salmon(Oncorhynchus nerka). Can. J. Fish. Aquat. Sci. 61(6): 905–912.

Williams, J.G. 1998. Fish passage in the Columbia River, USA andits tributaries: problems and solutions. In Fish migration andfish bypass. Edited by M. Jungwirth, S. Schmutz, and S. Weiss.Fishing News Books, Oxford. pp. 180–191.

Young, J.L., Hinch, S.G., Cooke, S.J., Crossin, G.T., Patterson,D.A., Farrell, A.P., Van Der Kraak, G., Lotto, A.G., Lister, A.,Healey, M.C., and English, K. 2006. Physiological and energetic

correlates of en route mortality for abnormally early migratingadult sockeye salmon (Oncorhynchus nerka) in the ThompsonRiver, British Columbia. Can. J. Fish. Aquat. Sci. 63: 1067–1077.

Zigler, S.J., Dewey, M.R., Knights, B.C., Runstrom, A.L., andSteingraeber, M.T. 2004. Hydrologic and hydraulic factors af-fecting passage of paddlefish through dams in the upper Missis-sippi River. Trans. Am. Fish. Soc. 133(1): 160–172.

© 2007 NRC Canada

Caudill et al. 995