Draft...Draft 3 30 Abstract 31 We used acoustic telemetry to investigate survival of age-2 sockeye...

Draft

Quantifying Survival of Age Two Chilko Lake Sockeye

Salmon during the First 50 Days of Migration

Journal: Canadian Journal of Fisheries and Aquatic Sciences

Manuscript ID cjfas-2017-0425.R1

Manuscript Type: Article

Date Submitted by the Author: 11-Mar-2018

Complete List of Authors: Rechisky, Erin; Kintama Research Services, Porter, Aswea; Kintama Research Services Clark, Timothy; University of Tasmania, Institute of Marine and Antarctic Studies; Commonwealth Scientific and Industrial Research Organisation, Agriculture and Food Furey, Nathan; University of British Columbia, Department of Forest and

Conservation Sciences Gale, M.; University of British Columbia, Department of Forest and Conservation Sciences Hinch, Scott; University of British Columbia, Department of Forest and Conservation Sciences Welch, David; Kintama Research Services

Is the invited manuscript for consideration in a Special

Issue? : Oceans Tracking Network

Keyword: early marine survival, TELEMETRY < General, MIGRATION < General, residence time, MORTALITY < General

https://mc06.manuscriptcentral.com/cjfas-pubs

Canadian Journal of Fisheries and Aquatic Sciences

Draft

1

CJFAS-2017-0425.R1 1

Quantifying Survival of Age Two Chilko Lake Sockeye Salmon 2

during the First 50 Days of Migration 3

4

Erin L. Rechisky, Aswea D. Porter, Timothy D. Clark, Nathan B. Furey, Marika Kirstin Gale, 5

Scott G. Hinch, and David W. Welch 6

7

Erin L. Rechisky (Corresponding author): Kintama Research Services, Ltd., 4737 Vista View 8

Crescent, Nanaimo BC V9V 1N8 Canada; phone: 250-729-2600; fax: 250-729-2622; 9

[email protected] 10

Aswea D. Porter: Kintama Research Services, Ltd., 4737 Vista View Crescent, Nanaimo BC 11

V9V 1N8 Canada; [email protected] 12

Timothy D. Clark1: Pacific Salmon Ecology and Conservation Laboratory, Department of Forest 13

and Conservation Sciences, University of British Columbia, 2424 Main Mall, Vancouver BC 14

V6T 1Z4 Canada; [email protected] 15

Nathan B. Furey: Pacific Salmon Ecology and Conservation Laboratory, Department of Forest 16

and Conservation Sciences, University of British Columbia, 2424 Main Mall, Vancouver BC 17

V6T 1Z4 Canada; [email protected] 18

19

1 University of Tasmania and CSIRO Agriculture and Food, Castray Esplanade, Hobart, Tasmania, 7000

Australia

Page 1 of 65



Draft

2

Marika Kirstin Gale2: Pacific Salmon Ecology and Conservation Laboratory, Department of 20

Forest and Conservation Sciences, University of British Columbia, 2424 Main Mall, Vancouver 21

BC V6T 1Z4 Canada; [email protected] 22

Scott G. Hinch: Pacific Salmon Ecology and Conservation Laboratory, Department of Forest and 23

Conservation Sciences, University of British Columbia, 2424 Main Mall, Vancouver BC V6T 24

1Z4 Canada; [email protected] 25

David Warren Welch: Kintama Research Services, Ltd., 4737 Vista View Crescent, Nanaimo BC 26

V9V 1N8 Canada; [email protected] 27

28

29

2 Freshwater Fisheries Society of BC, 101-80 Regatta Landing, Victoria BC V9A 7S2 Canada

Page 2 of 65



Draft

3

Abstract 30

We used acoustic telemetry to investigate survival of age-2 sockeye salmon (Oncorhynchus 31

nerka) as they emigrated from Chilko Lake, BC, Canada to north-eastern Vancouver Island 32

(NEVI) from 2010-2014. We built on results reported previously in Clark et al. (2016) by 33

including an additional year of data, and by converting survival estimates into rates (distance and 34

time) to compare across disproportionate habitats. We also refined our survival estimates by 35

including individual covariates in our survival models, and by re-investigating the detection 36

efficiency of the final detection site. There was a tag burden effect in 2012 and a body size effect 37

in 2013. Excluding 2010, survival during the 35-47 day migration to NEVI (range of annual 38

mean travel time; 1044 km) ranged between 8-14%. Weekly survival rate (Swk-1

) during 39

downstream migration to the Fraser River estuary, through the central Strait of Georgia (CSOG), 40

and NEVI was 25-46%, 75-90%, and 34-64%, respectively. In addition to marked losses in 41

freshwater tributaries, sockeye also experienced high losses north of the CSOG consistent with 42

earlier results for Cultus Lake sockeye. 43

44

Page 3 of 65



Draft

4

Introduction 45

Sockeye salmon (Oncorhynchus nerka) originating from western North America have 46

experienced a widespread decline in productivity (adults produced per spawner) during the last 47

quarter century (Peterman and Dorner 2012), and for Fraser River sockeye (British Columbia, 48

Canada), there is evidence that mortality during the juvenile-to-adult stage may be responsible 49

(Peterman et al. 2010). Survival in the marine environment may be influenced by multiple 50

factors such as competition with pink salmon (Ruggerone and Connors 2015); warming ocean 51

temperatures affecting zooplankton abundance and distribution (Mackas et al. 2007), which 52

could lead to a timing-mismatch of prey availability (Chittenden et al. 2010; Cushing 1990); 53

infectious diseases (Miller et al. 2014), interactions with salmon farms (Cohen 2012); and marine 54

mammal predation, particularly by harbor seals (Thomas et al. 2016). More recently, a marine 55

heat wave which began in late 2013 and persisted through the summer of 2015 has dramatically 56

affected the Northeast Pacific region (Hu et al. 2017; Kintisch 2015). The consequences for 57

Pacific salmon have yet to be determined, but the outlook is likely not good for southern 58

populations (Healey and Healey 2011; Welch et al. 1998). Estimating juvenile survival during 59

seaward migration and during the early marine period is thus important for identifying the 60

location of survival shortfalls, and for focusing the research needed to underpin efforts to rebuild 61

populations. 62

The Chilko Lake sockeye population is one of the largest runs of sockeye salmon originating 63

from the Fraser River watershed (DFO 2016) and the juveniles, called smolts, have been 64

monitored since 1954 (Henderson and Cass 1991). On average, 1.4 million adults return to the 65

lake, and 20 million juveniles migrate from the lake (DFO 2016). The juveniles migrate rapidly 66

downstream, 650 km through freshwater tributaries and the Fraser River mainstem before 67

Page 4 of 65



Draft

5

reaching the ocean (Fig. 1). Marine survival (which in this case includes downstream survival) 68

peaked at 18% in the late 1980s and has declined to 2% in recent years (DFO 2017). A marked 69

shift (decrease) in freshwater productivity occurred in 1989, likely due to density-dependent 70

effects related to increased escapement to the lake (McKinnell 2008), but in recent years ocean 71

conditions have been implicated in unusually low (Beamish et al. 2012; Rensel et al. 2010; 72

Beamish et al. 1997) or, in the case of 2010, high (Thomson et al. 2012) adult sockeye returns. 73

Juvenile Chilko Lake sockeye salmon begin their seaward migration as smolts in the spring 74

after spending one or two winters as fry in Chilko Lake. At the onset of migration, 11-76 million 75

smolts pass through an enumeration weir at the outlet of the lake, which is deployed annually in 76

the spring by DFO to monitor the population. Most lake-type sockeye begin their seaward 77

migration after only one winter in the lake, but a small proportion remain in the lake for an 78

additional year and migrate to sea at age-2 (average of 4.3%, (Irvine and Akenhead 2013)). 79

From 1954-2013, average mean fork length (FL) ranged between 73-100 mm for age-1 smolts, 80

and between 98-160 mm for age-2 smolts (K. Benner, Fisheries and Ocean Canada, Delta, BC, 81

personal communication). Age-2 smolts were the focus of our five year study, and during that 82

time, between 1.1-10.2% (360 000 to 4.2 million) of smolts migrating from Chilko Lake were 83

age-2 (K. Benner, Fisheries and Ocean Canada, Delta, BC, personal communication). Survival to 84

adult return is similar for both age classes even though for any given year the two age classes 85

differ in brood year and in body size. Mean smolt to adult return (SAR) rates from 1960-2006 86

were 8.5% (SD=5.2) for age-1s and 9.9% (SD=7.3) for age-2s (Irvine and Akenhead 2013). 87

Lake-type sockeye salmon migrating from the Fraser River enter the Strait of Georgia (SOG) 88

from mid-April to early June, with the majority entering in early May (Preikshot et al. 2012; 89

Neville et al. 2016). Most fish then migrate north between Vancouver Island and the British 90

Page 5 of 65



Draft

6

Columbia mainland (through the Discovery Islands, Johnstone Strait, and Queen Charlotte Strait) 91

before reaching the open ocean (Groot and Margolis 1991; Groot and Cooke 1987; Welch et al. 92

2009). Travel time through the SOG can take up to seven to eight weeks for smaller juvenile 93

sockeye (113 mm average FL, Preikshot et al. 2012; Neville et al. 2016), but large, hatchery-94

reared juveniles (~180 mm FL) are able to traverse much of this distance in under three weeks 95

(Welch et al. 2009; Welch et al. 2011) migrating at close to the predicted theoretical optimum 96

migration speed of ~1 body length (BLs-1

) (Weihs 1973). Note that Welch et al. (2009) reported 97

average residence time of Cultus Lake sockeye in the Strait of Georgia as 25.6-34.1 days; 98

however, this was defined at the time as from release (including 4.0-5.6 days of travel time in the 99

lower Fraser River) to northern Vancouver Island. The mean residence time in the CSOG (from 100

the Fraser River mouth to the northern Strait of Georgia array) was actually 11-14 days, and 90% 101

of detected juveniles migrated through in 15-23 days. 102

Marine travel time from the SOG to northern Queen Charlotte Strait (near Port Hardy, BC) 103

has not been reported for smaller, wild juveniles, but for larger, hatchery Cultus Lake sockeye 104

mean travel time ranged between 9 to 16 days, and 90% of juveniles reached QCS (from NSOG) 105

between 11 and 23 days (modified from Welch et al. 2009). Because of differences in ocean-106

entry date, body size, and travel rate, juvenile Fraser River sockeye are distributed over a broad 107

area between the SOG and Prince William Sound, Alaska by late June (Beacham et al. 2014; 108

Freshwater et al. 2016). 109

Using acoustic telemetry, juvenile salmon can now be tracked for thousands of kilometers 110

across rivers, estuaries, and into the ocean (e.g. Welch et al. 2011), and these detections can be 111

used to estimate where fish go, how fast they get there, and how many survived. In turn, these 112

estimates can be used to assess the influence of environmental factors (Chittenden et al. 2010; 113

Page 6 of 65



Draft

7

Brosnan et al. 2014); to compare the performance of hatchery and wild fish (Melnychuk et al. 114

2014; Urke et al. 2013); to compare the migratory behavior of fish from different origins 115

(Lacroix 2013); to evaluate habitat use (Pinnix et al. 2013), predation (Furey et al. 2016b), and 116

disease (Jeffries et al. 2014); and to test critical hypotheses regarding the effect of dams on 117

subsequent marine survival (e.g., Rechisky et al. 2013). 118

We initiated a multi-year telemetry study using wild, age-2 Chilko Lake sockeye salmon 119

smolts in 2010 to evaluate downstream and early marine survival. Age-1 fish were excluded 120

because they were too small for the transmitters available at the time. In our companion paper 121

(Clark et al. 2016), we highlighted the surprising amount of mortality that occurs in the 122

freshwater tributaries leading to the Fraser River, and observed that smolts primarily migrate at 123

night in the tributaries, presumably to evade predators, similar to steelhead trout (Oncorhynchus 124

mykiss) (Melnychuk et al. 2007). Mortality was at least partly due to concentrations of bull trout 125

(Salvelinus confluentus) which gorge themselves on smolts near the outlet of Chilko Lake (Furey 126

et al. 2015; Furey et al. 2016a), but smolts can improve their survival in these clear waters by 127

travelling together in high densities to effectively swamp predators (Furey et al. 2016b). There is 128

also evidence that freshwater mortality was higher for smolts with immune responses indicative 129

of viral infection (Jeffries et al. 2014). 130

Although the Clark et al. (2016) study reported survival estimates from Chilko Lake to 131

northern Vancouver Island, these estimates were not scaled by distance or time which is needed 132

to fully understand landscape-specific differences. In particular, survival rates were not 133

specifically compared between the two major marine areas delineated by acoustic arrays in the 134

study: the central Strait of Georgia (CSOG, between the Fraser River mouth and the acoustic 135

array in the northern Strait of Georgia (NSOG)), and the sea along the northeast coast of 136

Page 7 of 65



Draft

8

Vancouver Island (NEVI, between the NSOG array and the Queen Charlotte Strait array (QCS)) 137

which includes the Discovery Islands, Johnstone Strait, the Broughton Archipelago, and Queen 138

Charlotte Strait (Fig. 1). 139

In this paper, we updated survival estimates reported in Clark et al. (2016) with an additional 140

year of data (2014), we aggregated estimates over all five years in three major habitats (fresh 141

water, CSOG, and NEVI), compared survival rate (per 100 km and per day) and travel time in 142

CSOG relative to NEVI, and evaluated our inferred detection probability at the final array and 143

how it may affect our interpretations. We also assess the effect of tag burden and body size on 144

survival. Finally, we draw on data from an earlier 2004-2007 Cultus Lake sockeye survival study 145

(Welch et al. 2009) to compare and discuss early marine survival over a total of nine years. 146

Methods 147

Smolt collection and tagging 148

Smolts were collected nightly by dip-net near the Chilko Lake enumeration weir between 149

22:00 and 04:00 between late April and mid-May of 2010-2014, tagged, and then held in a river-150

based netpen immediately downstream of the weir until release the following night (see Clark et 151

al. 2016). Larger smolts were selected for tagging and transferred to a flow-through holding tank; 152

smaller smolts were released back into the lake. River water temperature generally ranged 153

between 2.0-10.0° C. Negligible mortalities (<1%) were observed while the fish were in 154

captivity. The procedure for surgically implanting transmitters was reported in Clark et al. 155

(2016). 156

We used several types of uniquely identifiable VEMCO acoustic transmitters during the five- 157

year study. VEMCO V7-2L tags (7 mm diameter x 20 mm, 1.6 g in air, 69 kHz) were used in all 158

Page 8 of 65



Draft

9

years because they were compatible with the marine portions of the acoustic array, which were 159

capable of detecting 69 kHz transmissions (see Methods: Acoustic array). In 2011 and 2012, we 160

released additional groups of smolts tagged with V6-4H tags (6 mm diameter x 16.5 mm, 1.0 g in 161

air, 180 kHz) and V5-1H tags (5 mm diameter x 12 mm, 0.65 g in air, 180 kHz), respectively. 162

With these smaller tags, we could quantify survival over a broader size range by tagging smaller 163

smolts; however, they transmit on a higher frequency that was generally detectable only by the 164

river receivers. Starting in 2012, there was limited 180 kHz detection capability on the NSOG 165

array (see Methods: Acoustic array). Tag programming is reported in Table S1. 166

Each night, we selected the largest individuals for tagging in an attempt to reduce tag burden 167

(calculated as the tag-to-body mass ratio in air). In general, smolts tagged with V7 tags were 168

≥120 mm FL; however, 10 of 1 782 (<0.6%) V7 tagged fish ranged between 109-119 mm FL 169

(Table 1; Fig. 2). These small fish were tagged in 2012 when smolts migrating from the lake 170

were smaller on average, and larger fish were hard to come by since only 1.2% of the smolts 171

were age-2 in that year. Accordingly, V7 tag burden in 2012 ranged between 6-16% (in 2014, 172

when smolts were larger on average, tag burden ranged between 3-10%). V5 tags were implanted 173

into smolts ≥95 mm FL, and tag burden ranged between 5-10%. V6 tags were implanted into 174

smolts ≥115 mm FL, and tag burden ranged between 3-9%. Despite efforts to select large fish, 175

tag burdens frequently exceeded the recommended limits for salmon smolts (~6-8, Collins et al. 176

2013; Brown et al. 2010). To evaluate the effect of transmitter size on actively migrating fish, 177

we included tag burden as a covariate in the survival analysis (see Survival analysis). 178

All work involving live fish was approved by the Animal Care Committee of University of 179

British Columbia, Vancouver, BC (application # A08-0388 and A11-0215). 180

Page 9 of 65



Draft

10

Treatment groups 181

In each year of the study, a large proportion of tagged fish were released from the netpen 182

following one day of recovery (termed ‘lake released’). We released additional experimental 183

groups in some years (Table 1): we transported fish downstream (2011 and 2013) or upstream 184

(2014) from the netpen release site near the outlet of Chilko Lake, we collected tissue samples 185

from fish gills (“gill clip”; 2012 and 2013) for genomic analyses, and we tested the performance 186

of smaller acoustic tags (the V6 in 2011, and the V5 in 2012). The results from these 187

transportation and genomics studies are described and interpreted in separate papers (see Clark et 188

al 2016; Jeffries et al 2014; Furey et al 2016). The number of fish released each day is reported 189

in Figure S1. 190

Acoustic array 191

An extensive network of VEMCO acoustic receivers was used to track smolts from the 192

release site near the outlet of Chilko Lake to northern Vancouver Island— a total of 1 044 km 193

(Fig. 1). We positioned the acoustic receivers above the riverbed or seabed to form a series of 194

detection sites (referred to as ‘arrays’). Individual receivers recorded the date and time that 195

acoustic transmitters were detected, and these detections were used to estimate survival and 196

travel time to each array. 197

Within the Fraser River watershed, we deployed receivers in the Chilko River (at Henry’s 198

Bridge, 14 km downstream of the capture site (Array B), and at Siwash Bridge, near the 199

confluence of the Chilko and Chilcotin rivers (Array C), in the Chilcotin River at Farwell 200

Canyon (near the confluence of the Chilcotin and Fraser Rivers; deployed in all years but 2010, 201

Array D), and in the lower Fraser River (Mission, Derby, Fraser Mouth, Arrays E, F, G). An 202

array was also deployed downstream of the DFO fence at the netpen release site (Array A), but 203

Page 10 of 65



Draft

11

we only used data from this site in 2014 when some smolts were transported upstream. In all 204

years, 69 kHz VR2W receivers were deployed; beginning in 2011, paired 69 and 180 kHz 205

VR2W receivers were deployed at all freshwater sites to detect both tag types. 206

Three marine arrays (maintained by the Ocean Tracking Network) were used to track fish in 207

the ocean: the northern Strait of Georgia array (NSOG) and Queen Charlotte Strait array (QCS) 208

to the north of the Fraser River mouth, and the Strait of Juan de Fuca array to the south. In 2012, 209

eight of 27 69-kHz receivers (VEMCO model VR3) on the NSOG array were replaced with dual 210

frequency (69/180 kHz) VR4 receivers. Therefore, it was possible to detect V5-tagged fish 211

released in 2012 on some of the NSOG receivers (but not on any other marine arrays). We 212

deployed an additional array which extended across the continental shelf off of north-western 213

Vancouver Island in 2010 and 2011 (Lippy Point; which extended from shore to the 200 m 214

isobath in 2010 and to the 500 m isobath in 2011). The Lippy Point and Strait of Juan de Fuca 215

arrays provided evidence that very few Chilko sockeye fish migrate around the southern tip of 216

Vancouver Island (Table S2). 217

Survival analysis 218

Overview 219

We used a mark-recapture approach to estimate survival of acoustic-tagged fish where 220

detection at an acoustic receiver array equated to ‘recapture’. Estimates were calculated for each 221

year using the Cormack-Jolly-Seber (CJS) model (and variants) for live-recaptured animals 222

implemented with Program MARK (White and Burnham 1999) through the R (R Core Team 223

2014) package RMark (Laake 2012). The CJS model uses maximum-likelihood estimation to 224

derive parameter estimates (survival or φ, and detection probability or p) and their sampling 225

variance. Because the CJS model is an open model, estimates of survival represent the joint 226

Page 11 of 65



Draft

12

probability of migration and survival through each segment of the array. These estimates of 227

apparent survival are referred to as ‘survival’ for simplicity. 228

Data analysis followed a series of steps which varied slightly among years depending upon 229

the study details (Error! Reference source not found.). First, we screened the detection data to 230

exclude false positive detections, and then formed a detection history for each tagged individual. 231

Second, we assessed goodness of fit (GOF) of the data to the models. The third step applies to 232

years 2011-2014 when multiple treatment groups were released. In this step, we conducted an 233

initial assessment to determine whether treatment affected survival and if it was reasonable to 234

pool the treatment groups for analyses. Because there was little evidence of a treatment effect, in 235

step four we pooled all data from the various treatment groups so that only one survival 236

parameter was estimated for each migration segment in each year (see below for model 237

parameterization used to estimate p). In step five, we assessed the effects of fish length and tag 238

burden on survival, and then model averaged across our candidate models to produce estimates 239

of φ and p. In 2011 and 2012, some fish were implanted with 180 kHz tags that were not 240

detectable (or poorly detectable) on the ocean arrays. If these fish were included in the survival 241

models, non-detection in the ocean would have been interpreted as mortality. Therefore, we 242

repeated steps 3-5 twice in those years: we first analyzed a freshwater dataset (FW) that included 243

all tagged fish and all tag types (V5 or V6, and V7 tags), but only included freshwater detection 244

sites, and then we analyzed a freshwater-saltwater dataset (FWSW) that contained all the 245

detection sites, but only for tags that could also be detected on the ocean arrays (V7 tags). We 246

then appended the marine survival estimates from the FWSW dataset to the FW estimates so that 247

all tagged fish were used to best estimate survival. In step six, we calculated survival in 248

biologically important habitats (i.e. the Chilko, Chilcotin, and Fraser rivers, CSOG; and NEVI) 249

Page 12 of 65



Draft

13

by multiplying the segment-specific survival estimates within each habitat (e.g., we multiplied 250

Array B and C estimates to determine survival in the Chilko River) and we multiplied all 251

freshwater estimates for an overall river survival estimate. We also calculated cumulative 252

survival from release to each downstream array, 253

We converted the survival estimates for each region (river, CSOG, NEVI) to survival rates 254

per week and per 100 km of migration distance to make it possible to compare survival between 255

the regions. Finally, we estimated aggregate regional survival for all years to the Fraser River 256

mouth, NSOG and QCS. Thus, we present survival estimates in four ways: yearly survival 257

estimates within each habitat, yearly cumulative survival to each detection site, aggregated 258

survival across all years in freshwater and to marine arrays, and survival rates (per week and per 259

100 km). 260

Step 1: Data screening and organization 261

All acoustic detection data were screened for potential false positive detections. False 262

detections can occur as a result of background environmental conditions creating transmissions 263

similar to those used for telemetry, from collisions between acoustic tag transmissions reaching 264

the receiver from direct and reflected paths (‘echoes’), or from overlapping transmissions from 265

two or more tags. We used our standard criteria for acceptance: ID codes with two or more 266

detections within 0.5 hours and with more detections spaced with short intervals (<0.5 hour 267

spacing) than with long intervals (>0.5 hours spacing) were passed. Detections that failed this 268

first step were assessed individually and were accepted if the fish was detected further along the 269

migratory path after a reasonable travel time; or if the probability of tag collisions was low (i.e. if 270

there were no other tags transmitting repeatedly in the area), and travel time and migration 271

Page 13 of 65



Draft

14

behavior were reasonable. Excluded data typically formed <0.1% of the total recorded 272

detections. 273

We then compiled the screened data into capture history sequences for each year and each 274

dataset (FW and FWSW, see Overview). Capture histories use a sequence of 1s and 0s to 275

describe whether each individual fish was detected at each acoustic array. The sequence of 276

acoustic arrays that the fish encountered (Table S3) differed slightly across the years and datasets 277

because of changes in detection sites and release sites. Survival in the lower Fraser River from 278

Mission to Derby was inestimable (near 100%); therefore, we combined Derby detections with 279

Mission detections, and report as survival to Mission (the upstream site). After ocean entry, fish 280

primarily migrated north to exit the SOG. Because only three fish were detected on JDF to the 281

south (all in 2011), we focus on the vast majority (N=314) detected migrating northward, and 282

therefore removed the southern migrants from the FWSW datasets. 283

Step 2: Assumptions and GOF tests 284

Standard CJS model assumptions applied for all arrays: (1) every tagged individual of each 285

group has equal survival probability and equal probability of detection following release, (2) 286

sampling periods are instantaneous, (3) emigration is permanent, and (4) tags are not lost. 287

We used Goodness-of-fit (GOF) testing to assess assumption 1 and we assume that 288

assumptions 2-4 are generally met. We used the capture histories and the most fully 289

parameterized CJS model for each year (see Step 3 below) to test GOF of the data using the 290

bootstrapped GOF test within Program MARK with 1000 simulations. After bootstrapping, 291

there are two possible approaches to estimating over-dispersion: one approach is based on the 292

deviance and the second is based on . To be conservative, we used the higher of the two values. 293

Over-dispersion ranged between 1.20 to 1.94 for the individual years and datasets, and was 2.18 294

Page 14 of 65



Draft

15

when all years were combined. We multiplied the SEs of parameters determined in the next steps 295

by this value, and adjusted the Akaike Information Criterion (AIC) accordingly. 296

Step 3: Within-year initial assessment of treatment effect 297

Before pooling treatment groups, we investigated whether transportation (release at a 298

different time and place), gill clipping, or tag type affected survival, and whether transportation 299

affected array performance due to differences in arrival timing that the arrays (when applicable). 300

Since clipped and non-clipped fish were released together, we had no reason to believe that gill 301

clipping would affect tag detectability. We did, however, assume tag type would affect 302

performance given the differing acoustic power, frequencies, and transmission intervals used. 303

Details of the treatment groups and the within-year initial assessment are presented in the 304

Supplementary Data A. 305

Because there was little evidence that tag size, gill clipping, or transportation affected 306

survival, we pooled treatment groups to produce only one annual survival estimate for each 307

migration segment. In all but one comparison, the pooled model either had the highest AIC 308

ranking or was essentially tied for the highest ranking (Table S4), indicating survival estimates 309

were similar across treatments. The exception was in 2011 when transportation by truck 310

appeared to reduce survival in the first 13 km after release (Clark et al. 2016 Table 4). Because 311

the process of physically transporting the smolts did not reduce smolt survival, and because there 312

was no evidence of an effect of transport for the groups that were released downriver in 2011 and 313

2013, the apparent effect was likely due to random processes rather than a deterministic effect of 314

transportation itself. 315

In contrast, there was some evidence that detection probabilities were affected by 316

transportation likely due to arrival time differences for the transported groups. To account for 317

Page 15 of 65



Draft

16

this effect, we included transport as a group covariate for detection efficiency in our final models 318

(see Supplementary Data A and Step 4). 319

Step 4: Modeling annual survival and detection probability 320

Survival Probability 321

We constructed several candidate models to estimate fish survival, and to test if FL-at-322

tagging (FL) or tag-to-body mass in air ratio (tag burden) affected survival. We used the tag 323

burden parameter to investigate the potential effects of tagging related mortality; however, tag 324

burden is naturally correlated with FL since weight increases with length. For each year, we 325

modeled independent survival estimates for each migration segment (Φ(segment)). Then, we 326

hypothesized that tag burden or FL might cause a consistent shift in survival across all migration 327

segments, i.e., the effect was additive (Φ(segment+FL or segment+burden)). We also 328

hypothesized that any tag burden effects might manifest most strongly in the Chilko River 329

shortly after release as the result of the tagging process (Φ(segment+burdenChilkoR)). Finally, for 330

the freshwater-saltwater datasets, we included models with additional parameters that allowed 331

the effects of FL and tag burden on survival to be different in the ocean 332

(φ(segment+FLFW+FLocean or segment+burdenFW+burdenocean)). If the highest ranked model 333

included a covariate parameter, we examined the beta value of the covariate parameter to 334

determine the magnitude of the effect. If the upper and lower ends of confidence interval of the 335

beta value were negative, a larger covariate negatively affected survival, and likewise if the 336

confidence interval was positive, a larger covariate increased survival. 337

Detection probability 338

Page 16 of 65



Draft

17

We modeled detection probability in various ways. First, we allowed detection to vary freely 339

by array (p(site)). In the years when two tag types were used (FW datasets), we assumed that tag 340

type would also affect detection probability p(site:tag type)). Finally, because our within-year 341

initial assessment of treatment effect demonstrated that detection probability differed for fish that 342

were transported (years 2011, 2013, and 2014; see Step 3) we compared these models with ones 343

where p varied freely by release type (p(site:release type); p(site:tag type:release type)). For the 344

freshwater-saltwater datasets, we also included a model where p varied by release type in fresh 345

water but not in the ocean (transported and lake released groups were combined in the ocean to 346

increase sample size; p(siteFW:release type + siteOC)). 347

Step 5: Model selection and model averaging 348

We ran all the combinations of the specified models for survival and detection probability for 349

each year, and used AIC to assess if tag burden and FL affected survival. When there was 350

evidence of a negative effect of tag burden (which occurred in 2012), we estimated survival at 351

the 25th

percentile of tag burdens because 75% of the tagged fish in this year had tag burdens 352

greater than 7%, the recommended limit for sockeye smolts (Collins et al. 2013). We did not 353

adjust the survival estimates if there was evidence of an effect of FL. 354

We model-averaged across our candidate models to derive the final φ and p parameter 355

estimates for each migration segment (Burnham and Anderson 2002). For years 2011 and 2012, 356

when two tag types were used, we appended the estimates for the ocean segments from the 357

freshwater-saltwater dataset (69 kHz tags only) to the survival estimates for freshwater segments 358

from the freshwater dataset (69 and 180 kHz tags) to create a single set of estimates which 359

included all tagged fish in each year. 360

Page 17 of 65



Draft

18

Detection probability at the final array 361

Because survival and detection probability are confounded at the final detection site, we 362

specified a fixed value for p in order to estimate φ to QCS. For all models, we fixed the p of the 363

QCS array to 0.50 which is the value estimated at NSOG for all V7-tagged juvenile Chilko 364

sockeye combined across 2010-2014 (see Multi-year regional survival and detection probability). 365

The detection probability of NSOG is a reasonable approximation for the p at QCS because both 366

marine arrays had similar geometry, and similar low levels of gear loss. Use of a fixed value 367

resulted in reduced error in the corresponding survival estimates to QCS. The fixed value also 368

does not acknowledge possible site-specific differences in acoustic conditions or progressive 369

attenuation of output voltage and signal strength with time (migration distance) as tags drain 370

power from their silver-oxide batteries. To investigate how changes in the inferred value for p 371

would affect our conclusions, we assessed survival across a broad range (0.3 – 0.8) of possible 372

values for p (Fig. S2). For the freshwater dataset, we used the arrays in the lower Fraser River 373

delta (Array G, Lower; Fig. 1) to estimate survival to the arrays located ~10 km upriver in the 374

upper Fraser River delta (Array G, Upper); i.e., a value was not fixed. 375

Cumulative and habitat-based survival 376

To estimate cumulative survival, we multiplied the segment-specific survival estimates 377

between each detection array over the entire migration corridor from the outlet of Chilko Lake to 378

QCS. To estimate habitat survival, we multiplied segment-specific survival estimates within each 379

defined habitat. For example, we combined segment survival to Arrays B and C to provide 380

survival in the Chilko River, and we combined all freshwater sites downstream of Array D to 381

estimate survival in the Fraser River (when possible). All river arrays were use to estimate 382

survival from release to the Fraser River mouth. The variances on cumulative and habitat 383

Page 18 of 65



Draft

19

survival estimates were estimated with the delta method. For years 2011and 2012 when we 384

extracted the segment-specific survival estimates from both the freshwater and freshwater-385

saltwater datasets, we assumed that the covariance between the rivers segments and the ocean 386

segments was zero. 387

Survival rate per unit time and distance 388

The regions bounded by the acoustic arrays are not of equal distance, nor do they require 389

equal time to migrate; however, they do delimit the Fraser River basin, most of the SOG, and 390

Queen Charlotte Strait (including the Discovery Islands and Johnstone Strait). In two years 391

(2011 and 2012), we could separate tributaries (Chilko+Chilcotin rivers) from the Fraser River 392

mainstem. To assess the relative importance of different habitats, we scaled survival estimates by 393

residence time and habitat size. We calculated survival rates per week as 394

where S=survival, x= 7 days, and y=median travel time in days within each habitat. We 395

calculated survival rates per 100 km with the same formula but using x=100 km, and y=shortest 396

distance in water between the arrays in km. To estimate the variability around the rate estimates, 397

we reran the formula substituting the upper and lower 95% confidence interval on the survival 398

estimate for S. 399

Multi-year regional survival and detection probability 400

We condensed survival in the three major regions (FW, CSOG, and NEVI) and across all 401

five years of the study. This allowed us to test if survival varied by year, to increase the precision 402

of early marine survival estimates by combining years (increasing sample size), to assess 403

mortality hotspots in three major regions, and finally, to produce a detection probability at 404

Page 19 of 65



Draft

20

NSOG that we could apply to the QCS array (so that we could then estimate survival to QCS; see 405

Detection probability at the final array). 406

We simplified the multi-year analysis by using only the FWSW datasets, and by removing 407

the freshwater arrays upstream of the Fraser River mouth array from the detection history 408

sequence (Fig. 1, Table S2). We did, however, include the netpen site to remove mortality of the 409

transported fish in 2014 in the 1.3 km segment between release and the DFO counting fence. 410

We used AIC to compare the performance of several models parameterized to test the effects 411

of year on survival and detection probability. For survival we hypothesized that 1) φ varied by 412

migration segment, but not by year (φ (segment)); 2) φ varied vary freely by year for all 413

segments (φ (segment:year)), and 3) φ varied by year in all freshwater migration segments, but 414

only by segment in the ocean (φ (segmentFW:year + segmentocean)). We used the same hypotheses 415

to test if detection probability varied by year (but substituted ‘p’ for ‘φ’ and ‘site’ for ‘segment’) 416

resulting in a set of nine models across all combinations of φ and p. Note that the models were in 417

reality somewhat more complex because in some cases the treatments that were transported for 418

release below the DFO counting fence (2011 and 2013) could not be directly pooled with those 419

released at (or above) the fence. We retained these fish in the analysis, but estimated independent 420

parameters for φ in fresh water because they did not migrate over the full distance. Secondly, we 421

modeled p independently for the transported treatment in 2013 for all arrays, because detection 422

probabilities were higher for this group. This effect likely occurred because these fish were 423

transported to the lower Fraser River (saving them about seven days of travel time) and their tags 424

had not switched to the longer transmission interval by the time the fish reached NSOG, boosting 425

the detection probability (tags were programmed to switch from an average of 15 s to 60 s 426

between transmissions 14 days after activation). 427

Page 20 of 65



Draft

21

Because initial analyses showed little evidence that survival or detection probability varied 428

significantly in the ocean during the study (not shown), we used this dataset to generate multi-429

year survival estimates with more precision and we used the p at NSOG as a reasonable 430

approximation for the p at QCS. To calculate this value (0.504), we model-averaged across all 431

nine candidate hypotheses and then averaged the year-specific p values at NSOG. To obtain 432

multi-year survival estimates, we simplified the analysis by including only hypotheses where 433

survival was pooled across years (but retained all three hypotheses for p). We then reran the 434

analysis, this time specifying the fixed value for p at QCS to obtain our final estimates of 435

survival from NSOG to QCS. 436

Travel times and travel rates 437

Travel time (days), also referred to as residence time, was calculated for each fish either from 438

release to departure from the first array, or from departure from one array until departure from 439

the next array along the migratory path. These estimates could only be made for fish detected on 440

both arrays bracketing the segment. Departure was defined as last detection on each relevant 441

array. We then used these data to estimate travel rates (BLs-1

) where body lengths were as 442

measured as FL at tagging. Travel rates in kmday-1

for 2010-2013 are reported in Clark et al. 443

(2016). 444

Results 445

From 2010 to 2014 we tracked 2 181 age-2 Chilko Lake sockeye salmon to as far as the 446

north-eastern end of Vancouver Island (Array I), a distance of 1 044 km from Chilko Lake (see 447

visualization at http://kintama.com/animator/dep/Chilko2014/). Survival estimates of individual 448

treatment groups tracked to each detection site from 2010-2013 were reported in Clark et al. 449

(2016). In this paper, we combined the treatment groups because there was no negative effect of 450

Page 21 of 65



Draft

22

treatment on survival (Error! Reference source not found., Supplement A). We summarize 451

survival by habitat and region for all five years; therefore, the segment survival estimates 452

presented here are not directly comparable to estimates presented in Clark et al (2016). We also 453

refined our survival estimates by accounting for the effect of tag burden, we tested assumptions 454

regarding the detection probability on the final sub-array (QCS), and we scaled survival by 455

distance and time to understand landscape-specific differences particularly between the two 456

major marine areas: the SOG and NEVI. 457

Effect of tag burden and body size on survival 458

There was a tag burden effect in one of the five years (2012), where fish with larger burdens 459

had reduced survival. Tag burdens were generally high in that year (Table 1; Fig. 2), and models 460

that included effects for tag burden in that year accounted for 93% and 83% of model weights in 461

the FW and FWSW datasets respectively (Table 2; Supplementary Material A). There is 462

evidence that the effect was strongest soon after release in the Chilko River: the beta value for 463

the effect of tag burden on survival in the Chilko River from the highest ranked FW model was -464

0.10916 (95% CI= -0.16874 to -0.04958). Model-averaged survival estimates over a range of tag 465

burdens estimated that fish at the lowest burden in 2012 (4.8%) experienced survival as high as 466

89% from release to the first detection site vs. 73% survival for those with the highest burdens 467

(16.2%; Fig. 3). In the second migration segment, fish with the lowest burden experience 468

survival as high as 82% vs. 64% survival for those with the highest burdens. To account for this 469

effect in our final survival estimate (Table 3), we used the 25th

percentile of tag burdens (see 470

Methods, Step 5). As a result, estimated survival in the Chilko River as a whole increased from 471

61.7% to 68%. In all other years, models which included tag burden as a covariate were ranked 472

Page 22 of 65



Draft

23

slightly below the base model indicating little improvement in model fit, and thus, little to no 473

effect on survival (Table 2). 474

Larger fish survived better in 2013. For this year, models which included a FL parameter 475

accounted for 68% of the model weight (Table 2). The beta value for the effect of FL on survival 476

from the highest ranked model was 0.087 (95% CI = 0.013 to 0.161) which means the odds of 477

survival were 54% higher for each 5 mm increase in length. Cumulatively, larger fish 478

experienced survival as high as 37.4% to QCS vs. 10.7% survival for the smallest fish; however, 479

the final survival estimate for the median fish FL was only 14% to QCS in 2013 (Table S5) 480

because there were relatively few large fish. We did not account for differences in body size in 481

our survival models because we considered this to be a natural effect and not a tag effect (there 482

was no effect of tag burden in 2013). In all other years, models which included FL were ranked 483

lower than the base model indicating little improvement in model fit and no effect of FL. 484

Cumulative, habitat-based, and multi-year regional survival 485

Survival from Chilko Lake to the Fraser River mouth ranged between 22-48% (Fig. 4). 486

Survival from Chilko Lake to NSOG ranged between 16-40%, and overall survival from Chilko 487

Lake to the final marine array (QCS) ranged between 4-14% (8-14% excluding 2010; Fig. 5; 488

Table S5). 489

Survival in the Chilko River (79 km) ranged between 58-78% among years (Table 3; also see 490

Table S6 survivals between array segments). Survival in the Chilcotin River (98 km) was only 491

estimable in 2011 and 2012 and was 65% and 66%, respectively. Survival in the Fraser River 492

mainstem (nearly 500 km) in those years was 81% and 86%. In the years when we could not 493

separate survival in the Chilcotin River from survival in the Fraser River (2010, 2013 and 2014), 494

survival was 38%, 62%, and 64% for the tributary and Fraser mainstem combined. If we 495

Page 23 of 65



Draft

24

multiply survival in the Chilcotin and Fraser Rivers in 2011 and 2012 for comparison, combined 496

survival was similar to other years: 53% and 57%. Survival in the CSOG ranged between 47-497

83%. Subsequent survival along NEVI ranged between 24-53%. In four of five years, marine 498

survival was lower in the NEVI region compared to CSOG. The 2010 release group consistently 499

had the lowest survival, while the 2013 groups consistently had the highest survival. Estimates 500

presented here are not comparable to Clark et al (2016) because of differences in the survival 501

analysis. 502

Freshwater survival aggregated over 2010-2014 was 35.9% (SE=2.8%; Fig. 4a). Aggregated 503

survival through the CSOG (to NSOG) was 68.5% (SE=9.2%) and was 37.5% (SE=6.3%) for 504

NEVI. Therefore, on average, survival was approximately one-third during the downstream 505

migration (fresh water), two-thirds in the CSOG, and one-third in the NEVI region. 506

Travel times and rates 507

Travel times within each main region were remarkably consistent among years. From Chilko 508

Lake to the Fraser River mouth, travel time was short, requiring only 5.2-9.0 days (range of 509

annual means) to complete the 657 km migration through fresh water (Fig. 6). Travel times to the 510

Fraser River mouth reported in Clark et al. (2016) were slightly lower (5.1-8.4 days) because 511

travel time was calculated from arrival at one array until arrival at the next array. Ninety-five 512

percent of tagged smolts passed the lower Fraser River by the third week of May (2010=May 19, 513

2011=May16, 2012=May 21, 2013=May 10, 2014=May 18). Mean travel time from the Fraser 514

River mouth to the NSOG array ranged between 13.8-19.6 days (grand mean=15.7), and 94% of 515

all fish were detected on the NSOG array within 30 days after being detected at the Fraser River 516

mouth. From NSOG to QCS, mean travel time ranged between 9.8-16.4 days. Taken together, 517

overall travel time from release to the northern end of Vancouver Island was approximately 40 518

Page 24 of 65



Draft

25

days (range of annual mean travel time = 34.9-47.4 days; Fig. 6). All fish reached the QCS array 519

by July 1st, with the exception of one fish in 2013 (which was detected July 10

th). 520

Mean travel rate in the CSOG ranged between 0.7-1.4 BLs-1

, with a grand mean across all 521

years of 1.06 BLs-1

. From the NSOG array to the QCS array (NEVI) migration rates were faster: 522

mean travel rate ranged between 1.4-2.2 BLs-1

, with a grand mean across all years of 1.8 BLs-1

. 523

Travel rates (from 2010-2013) in kmday-1

are reported in Clark et al. (2016). 524

Survival rates 525

To better compare survival between major habitats, we scaled the survival estimates by time and 526

distance. Freshwater survival during downstream migration per 100 km ranged from 79-90% 527

S100 km-1

(Fig. 4c). In the ocean, survival per 100 km was generally highest in the CSOG 528

(range=69-88% S100 km-1

; Fig. 4c) relative to NEVI (56-69% S100 km-1

), except for 2011 529

when it was only 59% in the CSOG, but 77% for NEVI. When scaled by time, weekly freshwater 530

survival rate ranged from 25-46% Swk-1

(Fig. 4d). In the ocean, weekly survival was higher in 531

the CSOG in all five years (range=75-90% Swk-1

) relative to NEVI (range=34-64%; Fig. 4d). 532

Overall early marine survival from the Fraser River mouth to northern Vancouver Island ranged 533

from 64-74% S100 km-1

, and from 70-78% Swk-1

. Thus, overall early marine survival per 100 534

km was lower than overall freshwater survival per 100 km during downstream migration. When 535

scaled by time, however, weekly freshwater survival rate was substantially lower than the 536

weekly marine survival rate because of the high losses experienced over very few days in the 537

tributaries. 538

Detection probability at the final array 539

Using NSOG data, we inferred that p of the final array (Array I, QCS) was 0.5. To 540

investigate how changes in the value for p would affect estimated survival, survival rate (per 100 541

Page 25 of 65



Draft

26

km and per week), and cumulative survival we varied p across a broad range of possible values 542

(from 0.3 to 0.8; Fig. S2). Within the CJS model, Φ increases as p decreases, and thus if 0.5 is 543

larger than the true value of p at QCS then survival could potentially be underestimated in the 544

NEVI area. We found that the detection probability of the QCS array would have to decrease to 545

~0.3 or less in order for survival in the NEVI region to increase to that of survival in the CSOG. 546

Detection probabilities for all other sites are reported in Table S7. 547

Discussion 548

Regions of high mortality 549

Our results indicate that although there was some inter-annual variability, on average 1/3 of 550

tagged age-2 sockeye smolts survived the downstream migration, 2/3 of those survived migration 551

through the CSOG, and 1/3 of the remaining fish survived migration to the northern end of 552

Vancouver Island, with 4 – 14% of all tagged fish reaching the northern end of Vancouver Island 553

(8-14% if 2010 is excluded). 554

As identified in Clark et al. (2016), there was a steep decline in survival very soon after 555

release in the essentially pristine clear water tributaries of the Fraser River (primarily the Chilko 556

River), followed by high survival in the Fraser River. Within the early marine phase, this paper 557

identifies NEVI as another area of reduced survival. Estimated survival probabilities and survival 558

per 100 km in NEVI were either lower or equivalent to those in the CSOG, except in 2011. 559

When scaled by time, NEVI had consistently lower survival than CSOG in all years (Fig. 4d). 560

Thus, much of the mortality measured for age-2 Chilko Lake sockeye salmon during the first 1 561

000 km of migration occurs in two areas: the clear water tributaries of the Fraser River and the 562

NEVI region. 563

Page 26 of 65



Draft

27

We also found that, similar to Chilko Lake sockeye, high tributary mortality occurred for 564

Cultus Lake sockeye, suggesting wider occurrence of this phenomenon. Welch et al. (2009) 565

reported survival of acoustic-tagged Cultus Lake sockeye smolts from release at the outlet of 566

Cultus Lake to the Fraser River mouth, a distance of ~100 km, as 54-70%. When we re-567

examined the data to look at fine scale survival to the first array encountered at Mission (in the 568

lower Fraser River, only 31 km from release; Array E), nearly all of the mortality occurred in the 569

tributaries (Sweltzer Creek, the Chilliwack River, the Sumas River, and a short distance [12 km] 570

in the Fraser River, combined). For example, survival to Mission was 64% in 2004, and 57% in 571

2007, and subsequent survival in the mainstem Fraser from Mission to the Fraser River mouth 572

was 95-100%. Cultus Lake sockeye had relatively low tag burdens, and days to weeks to recover 573

from surgery prior to release so a tag effect was unlikely. Further, this pattern has been observed 574

in other salmon species as well; Melnychuk et al. (2014) reported that 7–13% of acoustic-tagged, 575

wild steelhead smolts and 30–40% of hatchery steelhead smolts died in the first 3 km of the 576

migration, accounting for ~50% of downstream mortality to the Squamish River mouth, only 577

15.9 to 27.5 km from the release points. The mechanism contributing to low tributary survival is 578

likely predation. For example, Furey et al. (2016b) demonstrated that tributary survival can vary 579

from 40% at low smolt densities to nearly 100% at high densities when predators are swamped. 580

This density-dependent response implies that the variability in cumulative survival measured to 581

the QCS array is partially dependent on the outmigration densities experienced by smolts upon 582

release. 583

Lower survival in the NEVI region was also observed in three of four years during our earlier 584

Cultus Lake hatchery sockeye study. Remarkably, mean survival of hatchery-reared Cultus Lake 585

sockeye that were acoustic-tracked through the same region in four earlier years (2004-2007) 586

Page 27 of 65



Draft

28

was also 67% through the CSOG (Welch et al. 2009); subsequent survival to the northern end of 587

Vancouver Island was 52%, somewhat higher than survival of Chilko Lake juveniles. Similar to 588

Chilko Lake sockeye, when scaled by time, this difference becomes more apparent because of 589

the higher migration speeds through the NEVI region (Fig. 7). Therefore, in eight of nine years 590

of sockeye tracking, we have evidence of lower juvenile survival in the NEVI area bounded by 591

the NSOG and QCS arrays than in the CSOG, or alternatively, juvenile survival tends to be 592

highest in the CSOG. 593

Several factors could differentially affect early marine survival in the CSOG and NEVI 594

regions. The NEVI area includes the northern-most 1/5th

of the Strait of Georgia, and continues 595

north to encompass the Discovery Islands, Johnstone Strait, the Broughton Archipelago, and 596

Queen Charlotte Strait, and it is vastly more complex than the CSOG. The area north of 597

Johnstone Strait is world-renowned for its rich underwater biodiversity (Britnell 2010), and 598

offers whale watching and other eco-tourism opportunities (Destination BC 2017). The 599

Johnstone Strait, however, has little primary production (and thus zooplankton prey) because 600

wind and currents keep it well mixed to depths well below the photic zone (Thomson 1981; 601

McKinnell et al. 2014), and this leads to decreased juvenile salmon growth rates (Journey et al. 602

2018). Following the 2009 Fraser River sockeye crash, a trophic gauntlet hypothesis was put 603

forth by McKinnell et al. (2014) which describes the extreme ocean and climate events occurring 604

in this region and Queen Charlotte Sound that may have led to poor survival of juvenile sockeye 605

in 2007, two years prior to the adults’ return. Extreme environmental conditions could lead to 606

either decreased growth and size-based selection by predators as a result, or outright starvation if 607

continued for long enough. For instance, (Tucker et al. 2016) observed that Cassin’s auklets 608

preferentially preyed on smaller salmon in poor condition in southern Queen Charlotte Sound, 609

Page 28 of 65



Draft

29

the area directly north of our study site. The SOG, on the other hand, is one of the most 610

productive inland seas. Nutrient input and spring phytoplankton bloom timing means that there is 611

an abundant prey resource pool for migrating juvenile salmon (Harrison and Mackas 2014), and 612

growth rates are higher (Journey et al. 2018). In the Discovery Islands area between the SOG and 613

Johnstone Strait, there appears to be an abundant food supply in some years (Price et al. 2013), 614

but not in others (Neville et al. 2016); McKinnell et al. (2014) discuss the potential production 615

mechanisms associated with this transition zone. 616

Predation by marine mammals, particularly pinnipeds, has gained more attention as more 617

studies reveal the preferred diets that these animals consume. The harbour seal (Phoca vitulina 618

richardsi) population in British Columbia, and particularly the SOG, has rebounded to historic 619

levels since the species was protected in 1970 (DFO 2009) and there is evidence that they feed 620

on salmon species of conservation concern, including sockeye (Thomas et al. 2016). ()Even if 621

juvenile salmon comprise only a small proportion of the total diet, this results in large numbers 622

of fish (Thomas et al. 2016; Howard et al. 2013; Chasco et al. 2017). As the NEVI region 623

includes the northern-most area of the SOG, the lower survival we estimated for NEVI could be 624

partly attributed to fish becoming be more vulnerable to predation as they are concentrated in the 625

northern SOG and narrower waterways of the Discovery Islands where there are numerous seal 626

haul outs (DFO 2009; Yurk and Trites 2000)(). 627

Finally, the NEVI area, unlike the SOG, has numerous open net-pen salmon farms, and has 628

been fraught with controversy regarding the possible effect on wild salmon (Young and 629

Matthews 2010). The potential for interaction between wild and farmed salmon was highlighted 630

in a 2012 federal inquiry into the decline of sockeye salmon (Cohen 2012). The inquiry called on 631

the Department of Fisheries and Oceans Canada (DFO) to enforce stricter regulations and 632

Page 29 of 65



Draft

30

recommended prohibiting salmon farms in the Discovery Islands region if DFO cannot 633

confidently say “the risk of serious harm [to wild salmon] is minimal” (Cohen 2012). For 634

example, farmed salmon may transmit sea lice to migrating wild salmon, possibly reducing 635

foraging success and growth rates of wild salmon (Godwin et al. 2015; Godwin et al. 2017). 636

There are also a host of viral and bacterial pathogens associated with farmed salmon which may 637

be potentially harmful to wild salmon (Johansen et al. 2011) although the transmission of disease 638

has been poorly documented. Travel times reported here indicate that juvenile salmon migrate 639

quickly through this area, but the risk of serious harm is largely unknown, emphasizing the 640

importance of evaluating the effect of salmon farm exposure times on wild salmon survival. 641

Our finding that survival is lower in NEVI is based on the inference that the detection 642

probability of the QCS array was equivalent to the average performance of the NSOG array 643

during our five year study (0.5). In our previous paper (Appendix S3 in Clark et al. 2016), we 644

used a value of 0.67, based on our prior experience using acoustic transmitters in this region 645

(Welch et al. 2011); however, we reported some evidence that factors such as weakening tag 646

batteries may affect detection rates at QCS (e.g., the number of detections per ID code was lower 647

on the QCS array than on NSOG). The lower detection rate, p, at QCS increased the survival 648

estimate in the last migration segment. For instance, in Clark et al. (2016) we reported survival 649

for the 2012 V7 lake released group from NSOG to QCS as 0.31; here, survival in the same 650

segment is reported as 0.41. This increase in survival reduced the contrast between CSOG (0.76) 651

and NEVI. Our sensitivity analysis determined that p of QCS would have to drop to ≲30% for 652

survival in the northern area to be equal or higher than in the CSOG (Fig. S2). This drop in 653

detection probability seems unlikely given the similar receiver geometry to the NSOG array, the 654

high receiver retention rates at both locations over the past decade, and generally high 655

Page 30 of 65



Draft

31

performance (Welch et al. 2011). To directly measure the detection efficiency of the QCS array 656

requires an additional array somewhere beyond this location. In our view, this should be done to 657

reduce the uncertainty in estimated survival. 658

Effect of tag burden, body size, and detection probability on survival estimates 659

We found some evidence that increased tag burden above the recommended limit of 6-7% of 660

body weight (Collins et al. 2013) makes smolts more susceptible to mortality. This may be 661

attributed to reduced swimming performance (Collins et al. 2013; Perry et al. 2013) or reduced 662

swim speed (Furey et al. 2016b) and therefore increased susceptibility to predation upon release 663

into the wild. In 2012, the only year for which we had a wide distribution of fish sizes and tag 664

burdens (Fig. 2), higher tag burden resulted in higher mortality, particularly in the Chilko River 665

soon after release. When we account for this effect in our model, survival in the Chilko River 666

increased from 61.7% to 68%. We estimate that fish at the lowest burden (4.8%) experienced 667

survival as high as 72% in the Chilko River vs. 45% for those with the highest burdens (16.2%; 668

Fig. 3). 669

In a parallel holding study, survival of dummy-tagged Chilko Lake sockeye smolts was near 670

100% during the first week and burst swimming speeds were not compromised compared with 671

non-tagged fish; however, survival decreased for all groups following transfer to salt water 672

(Clark et al. 2016, Table 2). The decrease was greater for tagged fish but because none of the fish 673

would accept food during the month in captivity the results were confounded. If survival rates in 674

the CSOG were lower relative to NEVI, this would corroborate the captive study results; 675

however, this was not the case. 676

A tag burden effect did not occur in other years despite mean tag burdens reaching nearly 677

10% in 2010 and 2011, and over 11% in 2013. The lack of an effect in other years could indicate 678

Page 31 of 65



Draft

32

that predators did not target more heavily burdened individuals; however, it is more likely due to 679

a smaller range of tag burdens with which to assess the effect, particularly on the low end, i.e. 680

most fish had high tag burdens. The relatively high tag burdens occurring in this study (Fig. 2) 681

suggest that cumulative survival of untagged juveniles may be somewhat higher than what we 682

have documented, but since there is evidence that the main effect occurs in freshwater soon after 683

release, the relative survival between CSOG and NEVI would remain unchanged. 684

Although tag burden is typically negatively correlated with FL, this is only true when using 685

one tag type, e.g., Spearman’s rank correlation for tag burden and FL for each tag type and year 686

ranged between -0.73 and -0.96. In 2012, we found a burden effect on survival but not a length 687

effect because we used the smaller V5 transmitter in addition to the V7 transmitter. Thus, the FL 688

distribution approached normal while the distribution of tag burdens was more bimodal (Fig. 2). 689

A positive effect of body size (FL) on survival was detected in only one year (2013), which 690

was surprising given the narrow range in fish size tagged in that year. Unlike the tag burden 691

effect, the model selection results indicate that FL effects extended beyond fresh water. Size-692

selective mortality is known to occur in salmonids (Quinn 2005); however, it is unclear whether 693

this is regulated during the downstream or the early marine phase. There are few controlled 694

telemetry studies which focused on the effect of body size during early migration, and these had 695

mixed results: Halfyard et al. (2013) found a body size effect in three of four years, Melnychuk 696

et al. (Melnychuk et al. 2007) found a small effect, whereas several studies found no effect 697

(Rechisky et al. 2014; Melnychuk et al. 2012; Romer et al. 2013; Newton et al. 2016)(). These 698

studies measured survival over different landscapes and distances, but there does not seem to be 699

strong support for substantial size-selective mortality during the earliest migratory phase. 700

Page 32 of 65



Draft

33

Implications for later marine survival and adult return rates 701

The estimated smolt to adult return rate (SAR) for Chilko Lake sockeye ranged between 702

approximately 2-5% for outmigration years 2010-2014 (DFO 2017). Irvine and Akenhead (2013) 703

estimated that the SAR for age-1 and age-2 juveniles is similar; hence, we assumed they are 704

similar for these outmigration years as well. In 2010, estimated juvenile survival to QCS was 705

only 4% and the adult return rate two years later was also 4%, therefore our estimate was clearly 706

too low, as it is highly unlikely that survival in the open ocean over the next two years was 707

100%. In that year, we had the smallest sample size in the time series (n=199) and the lowest 708

freshwater survival observed during our study (22%), substantially reducing the number of 709

tagged fish entering the ocean. Reduced levels of predator swamping from low numbers of co-710

migrating smolts in 2010 may partially explain the lower overall survival, but it is unclear why 711

freshwater mortality was so high in that year. From 2011 to 2014, our cumulative survival 712

estimates ranged between 8-14%, 2-3 times higher than the 2010 estimate. Although these may 713

still appear low, survival at this level could still support a healthy population if later marine 714

survival was as high as historically estimated. For instance, if marine survival following this 715

phase was 74% per year for the 1.75 years at large, as estimated by Parker (Parker 1962), then 716

survival to adult return would have roughly ranged between 4.7-8.2%. Based on our cumulative 717

survival estimates to QCS and SAR estimates from DFO, adult marine survival was between 14-718

59% (mean = 35%) for age-2 juveniles from the 2010-2014 outmigration years, substantially 719

below the Parker (1962) estimate. The most extreme example from our study is from the 2013 720

migration year: juvenile survival to QCS was 14% and the adult return rate two years later was 721

only 2%; therefore the resulting later marine survival was only 14%. If we have underestimated 722

early migration survival (to QCS), then more mortality would be allocated to the later marine 723

period. In sum, downstream and early marine survival, particularly in the tributaries and in the 724

Page 33 of 65



Draft

34

northern marine tracking area, are important contributors to determining productivity; however, 725

our data also allude to a wider marine survival problem. 726

Travel time and rate 727

Both travel time and arrival timing of acoustic-tagged smolts to the Fraser River mouth were 728

consistent with the general population of Chilko sockeye smolts which take about one week to 729

reach the lower Fraser River (at the DFO traps in Mission BC, 659 km()) and arrive in late-April 730

to early-May (Preikshot et al. 2012; Neville et al. 2016). Migration speeds of acoustic-tagged 731

fish slowed considerably once juveniles reached the SOG in early-mid May and were consistent 732

with the theoretical expectation of 1 BLs-1

(Weihs 1973) and other acoustic tagged salmonids 733

(Drenner et al. 2012). Average migration speeds then increased to nearly 2 BLs-1

as fish 734

migrated between the NSOG and QCS arrays where the migration corridors are physically 735

narrower, bounding the amount of meandering that could occur, and smolts may exhibit 736

behaviours that allow selective transport when tidal currents are favourable (Metcalfe and Arnold 737

1990). 738

Median timing of age-2 sockeye detections on the NSOG array was earlier than peak catches 739

of juvenile sockeye in DFO purse seine surveys in 2011 and 2012 in the same vicinity (Neville et 740

al. 2013). The median detection date of acoustic-tagged fish on NSOG was 29 May 2011 and 23 741

May 2012, and the last detections were on 13 June 2011 and 19 June 2012. Concurrently, 742

sockeye were being captured in purse seine samples throughout the SOG with a peak CPUE in 743

mid-June in the north. Neville et al. (2013) and Preikshot et al. (2012) both noted that larger 744

juveniles were primarily captured in the northern Strait than in southern part, and Freshwater et 745

al. (2017) demonstrated that larger fish captured during these surveys travelled at a faster rate. 746

Page 34 of 65



Draft

35

Our telemetry data support this observation as larger fish travelling at the same relative speed (in 747

BLs-1

) would reach the northern end of the SOG earlier than smaller fish. 748

Mean travel time of acoustic-tagged juvenile Chilko Lake sockeye from the Fraser River 749

mouth to the NSOG array ranged between 13.8-19.6 days, which was longer than for larger 750

(~180 mm FL) hatchery-reared Cultus Lake sockeye, which took 11-14 days to traverse the same 751

area (modified from Welch et al. (2009); see Introduction). This is logical because both groups 752

were travelling at ~1 BLs-1

. These travel times exclude the northern section of the SOG; 753

therefore, we extrapolated travel time to the northern terminus of the Strait (40 km north of the 754

NSOG array) for Chilko Lake sockeye, assuming the same migration rate of ~1 BLs-1

to 755

estimate total travel time through the Strait. This increased the estimated travel time by an 756

additional four to five days, for an estimated mean residence time for age-2 Chilko Lake sockeye 757

in the SOG of 17.5-25.0 days. 758

Preikshot et al. (2012) estimated that smaller, dominant age-1 Fraser River sockeye had a 759

considerably longer average residence time: 43-54 days (6-8 weeks) in 2008 and 2009 based on 760

trawl and seine surveys reported above, and that the Chilko Lake population was consistent with 761

this estimate. Neville et al. (2016) found that residence time of Fraser River sockeye in 2014 was 762

7-8 weeks using the same methodology. In both studies, estimates were based on smolt passage 763

at Mission, BC in the lower Fraser River (a proxy for ocean entry date) and catch rates from 764

trawl and seine surveys in the SOG, Discovery Islands, and Johnstone Strait. The residence time 765

discrepancy at least partially arises due to the size differences between the two age classes, so 766

using the same migration speed of 1 BLs-1

we estimated the theoretical residence time of age-1 767

fish assuming an average body size of 10 cm. The estimated mean travel time from the river 768

mouth to the NSOG array under these assumptions is 17-26 days, and if we include the 769

Page 35 of 65



Draft

36

additional 40 km to the Discovery Islands, our theoretical mean residence time for age-1 fish in 770

the Strait increases to 22-33 days (~3-5 weeks), significantly less than the Preikshot et al. (2012) 771

estimate, but consistent with previous estimates for sockeye made by Peterman et al. (1994), 772

Groot and Cook (1987), and Healey (1980) (see Table 1 in Preikshot et al. 2012). 773

The reason for these differences is unclear, but it is possible that age-1 fish exhibit different 774

behavior and delay the initiation of directed migration after entering the SOG, perhaps as a result 775

of individuals first migrating into Howe Sound (Groot and Cooke 1987; Welch et al. 2009) or 776

across the Strait to the Gulf Islands prior to migrating north (Neville et al. 2016; Groot and 777

Cooke 1987). Freshwater et al. (2017) estimated Chilko Lake sockeye migration rate as 0.65 778

BLs-1

based on 2011 and 2012 DFO seine surveys; therefore, age-1 fish may linger in the SOG. 779

Indeed, if we estimate a theoretical maximum residence time for age-1 fish using the variation in 780

migration rate observed for acoustic-tagged age-2 fish in the CSOG, it could take the slowest 781

individuals as long as 46-59 days to leave the Strait. Another consideration is that our telemetry-782

based estimates provide travel times for surviving fish. Fish that migrate more slowly may be 783

more vulnerable to predation and would be excluded from travel time calculations at a higher 784

rate. In any case, an important point is that if age-1 fish swim slower than age-2 fish, age-1 fish 785

may have a longer residence time in the SOG. This may ultimately benefit age-1 fish if the SOG 786

provides a protected habitat relative to NEVI, which includes the gauntlet area (Neville et al. 787

2016; McKinnell et al. 2014). 788

Interestingly, size related travel time may not be as relevant in the NEVI area. The increase 789

in travel rate from ~1 BLs-1

in the SOG to 2 BLs-1

in the NEVI area for age-2 sockeye was 790

possibly caused by strong tidal currents in the Discovery Island and Johnstone Strait. If water 791

velocity during a favorable tide drives migration rate rather than body size, then fish of different 792

Page 36 of 65



Draft

37

sizes may have similar residence times. For instance, travel times in this area were consistent for 793

sockeye which varied in size: Cultus Lake sockeye (~180 mm FL on average) took 9.3-16.1 days 794

on average to migrate from NSOG to QSC; likewise, age-2 Chilko Lake sockeye in this study 795

(40-50 mm smaller) took approximately the same amount of time (9.8-16.4 days). 796

In conclusion, much of the mortality measured for age-2 Chilko Lake sockeye salmon during 797

the first 1 000 km of migration occurred very soon after release in the near-pristine, clear water 798

tributaries of the Fraser River, and in the marine area along the north-east coast of Vancouver 799

Island. When survival is scaled by distance or time, there is evidence to support the idea that the 800

SOG provides a more protected habitat (Neville et al. 2016) than the NEVI area. 801

The very rapid migration to QCS makes it less likely that indirect mortality agents such as 802

disease or parasites, or lack of food, had sufficient time to kill juvenile salmon prior to their 803

leaving the study area and more likely that predation may be an important factor. A more robust 804

and extensive array would enable more fine-scale observations and controlled experiments (e.g., 805

Miller et al. 2011; Donaldson et al. 2014). While our data indicate that the NEVI area may be a 806

bottleneck for marine survival, our results also point to low survival in the later marine period 807

(beyond Vancouver Island) as well. 808

Acknowledgements 809

This is Publication Number 21 from the Salish Sea Marine Survival Project 810

(marinesurvivalproject.com). We thank the Pacific Salmon Foundation and anonymous donors to 811

the PSF for purchase of some acoustic tags and for logistic support. DFO Canada provided 812

additional tags and its Environmental Watch Program provided logistic support. Funding and 813

infrastructure was also provided through the Ocean Tracking Network Canada, which was 814

supported through the Natural Sciences and Engineering Research Council of Canada (NSERC) 815

Page 37 of 65



Draft

38

and the Canada Foundation for Innovation. NSERC Discovery grants to SGH also supported this 816

work. The Alfred P. Sloan and The Gordon & Betty Moore Foundation provided financial 817

support for most of the marine components of the POST system. We thank Paul Winchell 818

(Kintama) for continued array maintenance and quality assurance. NBF was supported by a 819

NSERC Vanier Graduate Scholarship and the Mitacs Accelerate program. Ken Jeffries, Matt 820

Casselman, Andrew Lotto, Marley Bassett, Collin Middleton (UBC) provided valuable field 821

assistance. We thank the Xeni Gwet’in First Nation for access to study sites, Environment 822

Canada (Jennifer MacDonald and Mark Sekela) for environmental data from the Fraser River 823

estuary as well as helping to deploy additional acoustic receivers in the estuary, and DFO for 824

providing accommodation and logistic support at the Chilko River Camp (Dennis Klassen and 825

staff), adult sockeye return rates (Sue Grant), and Chilko Lake smolt data (Keri Benner, Scott 826

Decker). 827

Literature Cited 828

Beacham, T.D., Beamish, R.J., Candy, J.R., Wallace, C., Tucker, S., Moss, J.H., and Trudel, 829

M. 2014. Stock-specific size of juvenile sockeye salmon in British Columbia waters and the Gulf 830

of Alaska. Trans. Am. Fish. Soc. 143(4): 876-889. doi: 10.1080/00028487.2014.889751. 831

Beamish, R.J., Neville, C.M., and Cass, A.J. 1997. Production of Fraser River sockeye 832

salmon (Oncorhynchus nerka) in relation to decadal-scale changes in the climate and the ocean. 833

Can. J. Fish. Aquat. Sci. 54(3): 543-554. 834

Beamish, R.J., Neville, C., Sweeting, R., and Lange, K. 2012. The synchronous failure of 835

juvenile Pacific salmon and herring production in the Strait of Georgia in 2007 and the poor 836

Page 38 of 65



Draft

39

return of sockeye salmon to the Fraser River in 2009. Marine and Coastal Fisheries. 4(1): 403-837

414. doi: 10.1080/19425120.2012.676607. 838

Britnell, J. 2010. Diving Wild in Browning Pass, British Columbia [online]. Available 839

from http://www.alertdiver.com/Diving_Wild_in_Browning_Pass_British_Columbia [accessed 840

28 September 2017]. 841

Brosnan, I.G., Welch, D.W., Rechisky, E.L., and Porter, A.D. 2014. Evaluating the influence 842

of environmental factors on yearling Chinook salmon survival in the Columbia River plume 843

(USA). Mar. Ecol. Prog. Ser. 496: 181-196. doi: 10.3354/mepsl0550. 844

Brown, R.S., Harnish, R.A., Carter, K.M., Boyd, J.W., Deters, K.A., and Eppard, M.B. 2010. 845

An evaluation of the maximum tag burden for implantation of acoustic transmitters in juvenile 846

Chinook salmon. N. Am. J. Fish. Manage. 30(2): 499-505. doi: 10.1577/M09-038.1. 847

Chasco, B., Kaplan, I.C., Thomas, A., Acevedo-GutiÃ©rrez, A., Noren, D., Ford, M.J., 848

Hanson, M.B., Scordino, J., Jeffries, S., Pearson, S., Marshall, K.N., and Ward, E.J. 2017. 849

Estimates of Chinook salmon consumption in Washington State inland waters by four marine 850

mammal predators from 1970 to 2015. Can. J. Fish. Aquat. Sci. 74(8): 1173-1194. doi: 851

10.1139/cjfas-2016-0203. 852

Chittenden, C.M., Jensen, J.L., Ewart, D., Anderson, S., Balfry, S., Downey, E., Eaves, A., 853

Saksida, S., Smith, B., Vincent, S., Welch, D., and McKinley, R. 2010. Recent salmon declines: 854

A result of lost feeding opportunities due to bad timing? PLOS One. 5(8)(8): e12423. doi: 855

10.1371/journal.pone.0012423. 856

Page 39 of 65



Draft

40

Clark, T.D., Furey, N.B., Rechisky, E.L., Gale, M.K., Jeffries, K.M., Porter, A.D., 857

Casselman, M.T., Lotto, A.G., Patterson, D.A., Cooke, S.J., Farrell, A.P., Welch, D.W., and 858

Hinch, S.G. 2016. Tracking wild sockeye salmon smolts to the ocean reveals distinct regions of 859

nocturnal movement and high mortality. Ecol. Appl. 26(4): 959-978. doi: 10.1890/15-0632. 860

Cohen, B.I. 2012. Commission of Inquiry into the Decline of Sockeye Salmon in the Fraser 861

River (Canada): The uncertain future of Fraser River sockeye. Public Works and Government 862

Services Canada, Ottawa. 863

Collins, A.L., Hinch, S.G., Welch, D.W., Cooke, S.J., and Clark, T.D. 2013. Intracoelomic 864

acoustic tagging of juvenile sockeye salmon: swimming performance, survival, and postsurgical 865

wound healing in freshwater and during a transition to seawater. Trans. Am. Fish. Soc. 142(2): 866

515-523. doi: 10.1080/00028487.2012.743928. 867

Cushing, D. 1990. Plankton production and year-class strength in fish populations - an update 868

of the match mismatch hypothesis. Adv. Mar. Biol. 26: 249-293. doi: 10.1016/S0065-869

2881(08)60202-3. 870

Destination BC. 2017. Wildlife Tours [online]. Available from 871

https://www.hellobc.com/british-columbia/things-to-do/touring-sightseeing/wildlife-tours.aspx 872

[accessed 28 September 2017]. 873

DFO. 2017. Pre-season run size forecasts for Fraser River Sockeye (Oncorhynchus nerka) 874

and Pink (O. gorbuscha) salmon in 2017. DFO Can. Sci. Advis. Sec. Sci. Resp. 2017/016, 875

Canada Dept. of Fisheries and Oceans. Pacific Region; Canadian Science Advisory Secretariat, 876

Ottawa. 877

Page 40 of 65



https://www.hellobc.com/british-columbia/things-to-do/touring-sightseeing/wildlife-tours.aspx

Draft

41

DFO. 2016. Pre-season Run Size Forecasts for Fraser River Sockeye (Oncorhynchus nerka) 878

Salmon in 2016. DFO Can. Sci. Advis. Sec. Sci. Resp. 2016/021, Canada Dept. of Fisheries and 879

Oceans. Pacific Region; Canadian Science Advisory Secretariat, Ottawa. 880

DFO. 2009. Population Assessment Pacific Harbour Seal (Phoca vitulina richardsi). DFO 881

Can. Sci. Advis. Sec. Sci. Advis. Rep. 2009/011, Canada. Dept. of Fisheries and Oceans; 882

Canadian Science Advisory Secretariat, Ottawa. 883

Donaldson, M.R., Hinch, S.G., Suski, C.D., Fisk, A.T., Heupel, M.R., and Cooke, S.J. 2014. 884

Making connections in aquatic ecosystems with acoustic telemetry monitoring. Frontiers in 885

Ecology and the Environment. 12(10): 565-573. doi: 10.1890/130283. 886

Drenner, S.M., Clark, T.D., Whitney, C.K., Martins, E.G., Cooke, S.J., and Hinch, S.G. 887

2012. A Synthesis of Tagging Studies Examining the Behaviour and Survival of Anadromous 888

Salmonids in Marine Environments. PLoS One. 7(3): e31311. doi: 889

10.1371/journal.pone.0031311. 890

Freshwater, C., Trudel, M., Beacham, T.D., Grant, S.C.H., Johnson, S.C., Neville, C.E.M., 891

Tucker, S., and Juanes, F. 2017. Effects of density during freshwater and early marine rearing on 892

juvenile sockeye salmon size, growth, and migration. Mar. Ecol. Prog. Ser. 579: 97-110. 893

Freshwater, C., Trudel, M., Beacham, T.D., Godbout, L., Neville, C.M., Tucker, S., and 894

Juanes, F. 2016. Divergent migratory behaviours associated with body size and ocean entry 895

phenology in juvenile sockeye salmon. Can. J. Fish. Aquat. Sci. 73(12): 1723-1732. doi: 896

10.1139/cjfas-2015-0425. 897

Page 41 of 65



Draft

42

Furey, N.B., Hinch, S.G., Lotto, A.G., and Beauchamp, D.A. 2015. Extensive feeding on 898

sockeye salmon Oncorhynchus nerka smolts by bull trout Salvelinus confluentus during initial 899

outmigration into a small, unregulated and inland British Columbia river. J. Fish Biol. 86(1): 900

392-401. doi: 10.1111/jfb.12567. 901

Furey, N.B., Hinch, S.G., Mesa, M.G., and Beauchamp, D.A. 2016a. Piscivorous fish exhibit 902

temperature-influenced binge feeding during an annual prey pulse. J. Anim. Ecol. 85(5): 1307-903

1317. doi: 10.1111/1365-2656.12565. 904

Furey, N.B., Hinch, S.G., Bass, A.L., Middleton, C.T., Minke-Martin, V., and Lotto, A.G. 905

2016b. Predator swamping reduces predation risk during nocturnal migration of juvenile salmon 906

in a high-mortality landscape. J. Anim. Ecol. 85(4): 948-959. doi: 10.1111/1365-2656.12528. 907

Godwin, S.C., Dill, L.M., Krkosek, M., Price, M.H.H., and Reynolds, J.D. 2017. Reduced 908

growth in wild juvenile sockeye salmon Oncorhynchus nerka infected with sea lice. J. Fish Biol. 909

91(1): 41-57. doi: 10.1111/jfb.13325. 910

Godwin, S.C., Dill, L.M., Reynolds, J.D., and Krkosek, M. 2015. Sea lice, sockeye salmon, 911

and foraging competition: lousy fish are lousy competitors. Can. J. Fish. Aquat. Sci. 72(7): 1113-912

1120. doi: 10.1139/cjfas-2014-0284. 913

Groot, C., and Margolis, L. 1991. Pacific Salmon Life Histories. UBC Press, Vancouver. 914

Groot, C., and Cooke, K. 1987. Are the migrations of juvenile and adult Fraser River sockeye 915

salmon (Oncorhynchus nerka) in near-shore waters related? In Sockeye Salmon (Oncorhynchus 916

Page 42 of 65



Draft

43

Nerka ) Population Biology and Future Management. Edited by H.D. Smith, L. Margolis and 917

C.C. Wood. Can. Spec. Publ. Fish. Aquat. Sci. 96, Ottawa, Ont. (Canada). pp. 53. 918

Halfyard, E.A., Gibson, A.F., Stokesbury, M.J., Ruzzante, D.E., and Whoriskey, F.G. 2013. 919

Correlates of estuarine survival of Atlantic salmon postsmolts from the Southern Upland, Nova 920

Scotia, Canada. Canadian Journal of Fisheries and Aquatic Sciences/Journal Canadien Des 921

Sciences Halieutiques Et Aquatiques. 70(3): 452-460. doi: 10.1139/cjfas-2012-0287. 922

Harrison, P., and Mackas, D. 2014. The Biological Oceanography. In The Sea among Us: 923

The Amazing Strait of Georgia. Edited by R.J. Beamish and G.A. McFarlane. pp. 41-65. 924

Healey, M.C. 1980. The Ecology of Juvenile Salmon in Georgia Strait, British Columbia. In 925

Salmonid Ecosystems of the North Pacific. Edited by W.J. McNeil and D.C. Himsworth. Oregon 926

State University Press and Oregon State University Sea Grant College Program, Corvallis. pp. 927

203-229. 928

Healey, M., and Healey, M. 2011. The cumulative impacts of climate change on Fraser River 929

sockeye salmon (Oncorhynchus nerka) and implications for management. Can. J. Fish. Aquat. 930

Sci. 68(4): 718; 718-737; 737. 931

Henderson, M.A., and Cass, A.J. 1991. Effect of smolt size on smolt-to-adult survival for 932

Chilko Lake sockeye-salmon (Oncorhynchus-nerka). Can. J. Fish. Aquat. Sci. 48(6): 988-994. 933

Howard, S., Lance, M.M., Jeffries, S.J., and Acevedo-Gutierrez, A. 2013. Fish consumption 934

by harbor seals (Phoca vitulina) in the San Juan Islands, Washington. Fish. Bull. 111(1): 27-41. 935

doi: http://dx.doi.org.ezproxy.library.ubc.ca/10.7755/FB.111.1.3. 936

Page 43 of 65



Draft

44

Hu, Z., Kumar, A., Jha, B., Zhu, J., and Huang, B. 2017. Persistence and Predictions of the 937

Remarkable Warm Anomaly in the Northeastern Pacific Ocean during 2014–16. J. Climate. 938

30(2): 689-702. doi: 10.1175/JCLI-D-16-0348.1. 939

Irvine, J.R., and Akenhead, S.A. 2013. Understanding smolt survival trends in sockeye 940

salmon. Marine and Coastal Fisheries. 5(1): 303-328. doi: 10.1080/19425120.2013.831002. 941

Jeffries, K.M., Hinch, S.G., Gale, M.K., Clark, T.D., Lotto, A.G., Casselman, M.T., Li, S., 942

Rechisky, E.L., Porter, A.D., Welch, D.W., and Miller, K.M. 2014. Immune response genes and 943

pathogen presence predict migration survival in wild salmon smolts. Mol. Ecol. 23(23): 5803–944

5815. doi: 10.1111/mec.12980. 945

Johansen, L.-., Jensen, I., Mikkelsen, H., Bjørn, P.-., Jansen, P.A., and Bergh, Ø. 2011. 946

Disease interaction and pathogens exchange between wild and farmed fish populations with 947

special reference to Norway. 948

Journey, M.L., Trudel, M., Young, G., and Beckman, B.R. 2018. Evidence for depressed 949

growth of juvenile Pacific salmon (Oncorhynchus) in Johnstone and Queen Charlotte Straits, 950

British Columbia. Fish. Oceanogr. 27(2): 174-183. doi: 10.1111/fog.12243. 951

Kintisch, E. 2015. 'The Blob' invades Pacific, flummoxing climate experts. Science. 952

348(6230): 17-18. 953

Laake, J. 2012. RMark: R Code for MARK Analysis. http://CRAN.R-954

project.org/package=RMark. 955

Page 44 of 65



Draft

45

Lacroix, G.L. 2013. Migratory strategies of Atlantic salmon (Salmo salar) postsmolts and 956

implications for marine survival of endangered populations. Can. J. Fish. Aquat. Sci. 70(1): 32-957

48. doi: 10.1139/cjfas-2012-0270. 958

Mackas, D., Batten, S., and Trudel, M. 2007. Effects on zooplankton of a warmer ocean: 959

Recent evidence from the Northeast Pacific. Prog. Oceanogr. 75(2): 223-252. doi: 960

10.1016/j.pocean.2007.08.010. 961

McKinnell, S. 2008. Fraser River sockeye salmon productivity and climate: A re-analysis 962

that avoids an undesirable property of Ricker's curve. Prog. Oceanogr. 77(2-3): 146-154. doi: 963

10.1016/j.pocean.2008.03.014. 964

McKinnell, S., Curchitser, E., Groot, K., Kaeriyama, M., and Trudel, M. 2014. Oceanic and 965

atmospheric extremes motivate a new hypothesis for variable marine survival of Fraser River 966

sockeye salmon. Fish. Oceanogr. 23(4): 322-341. doi: 10.1111/fog.12063. 967

Melnychuk, M.C., Welch, D.W., Walters, C.J., and Christensen, V. 2007. Riverine and early 968

ocean migration and mortality patterns of juvenile steelhead trout (Oncorhynchus mykiss) from 969

the Cheakamus River, British Columbia. Hydrobiologia. 582: 55-65. 970

Melnychuk, M.C., Walters, C.J., Christensen, V., Bothwell, M.L., and Welch, D.W. 2012. 971

Effects of solar ultraviolet radiation exposure on early ocean survival and fry-to-smolt growth of 972

juvenile salmon. Mar. Ecol. Prog. Ser. 457: 251-264. doi: 10.3354/meps09426. 973

Melnychuk, M.C., Korman, J., Hausch, S., Welch, D.W., McCubbing, D.J.F., and Walters, 974

C.J. 2014. Marine survival difference between wild and hatchery-reared steelhead trout 975

Page 45 of 65



Draft

46

determined during early downstream migration. Can. J. Fish. Aquat. Sci. 71(6): 831-846. doi: 976

10.1139/cjfas-2013-0165. 977

Metcalfe, J.D., and Arnold. 1990. The energetics of migration by selective tidal stream 978

transport: An analysis for plaice tracked in the southern North Sea. J. Mar. Biol. Assoc. U. K. 979

70(1): 149. 980

Miller, K.M., Teffer, A., Tucker, S., Li, S., Schulze, A.D., Trudel, M., Juanes, F., Tabata, A., 981

Kaukinen, K.H., Ginther, N.G., Ming, T.J., Cooke, S.J., Hipfner, J.M., Patterson, D.A., and 982

Hinch, S.G. 2014. Infectious disease, shifting climates, and opportunistic predators: cumulative 983

factors potentially impacting wild salmon declines. Evolutionary Applications. 7(7): 812-855. 984

doi: 10.1111/eva.12164. 985

Miller, K.M., Li, S., Kaukinen, K.H., Ginther, N., Hammill, E., Curtis, J.M.R., Patterson, 986

D.A., Sierocinski, T., Donnison, L., Pavlidis, P., Hinch, S.G., Hruska, K.A., Cooke, S.J., English, 987

K.K., and Farrell, A.P. 2011. Genomic signatures predict migration and spawning failure in wild 988

Canadian salmon. Science. 331(6014): 214-217. doi: 10.1126/science.1196901. 989

Neville, C.M., Johnson, S.C., Beacham, T.D., Whitehouse, T., Tadey, J., and Trudel, M. 990

2016. Initial estimates from an integrated study examining the residence period and migrating 991

timing of juvenile sockeye salmon from the Fraser River through coastal waters of British 992

Columbia. N. Pac. Anadr. Fish Comm.(6): 45-60. doi: 10.23849/npafcb6/45.60. 993

Neville, C.M., Trudel, M., Beamish, R.J., and Johnson, S.C. 2013. The early marine 994

distribution of juvenile sockeye salmon produced from the extreme low return in 2009 and the 995

Page 46 of 65



Draft

47

extreme high return in 2010. Report No. 9, North Pacific Anadromous Fish Commission, 996

Vancouver, BC. 997

Newton, M., Barry, J., Dodd, J.A., Lucas, M.C., Boylan, P., and Adams, C.E. 2016. Does 998

size matter? A test of size-specific mortality in Atlantic salmon Salmo salar smolts tagged with 999

acoustic transmitters. J. Fish Biol. 89(3): 1641-1650. doi: 10.1111/jfb.13066 [doi]. 1000

Parker, R.R. 1962. Estimations of ocean mortality rates for Pacific Salmon (Oncorhynchus). 1001

J. Fish. Res. Bd. can. 19(4): 561-589. 1002

Perry, R.W., Plumb, J.M., Fielding, S.D., Adams, N.S., and Rondorf, D.W. 2013. Comparing 1003

Effects of Transmitters within and among Populations: Application to Swimming Performance 1004

of Juvenile Chinook Salmon. Trans. Am. Fish. Soc. 142(4): 901-911. doi: 1005

10.1080/00028487.2013.788556. 1006

Peterman, R.M., Marmorek, D., Beckman, B., Bradford, M., Mantua, N., Riddell, B.E., 1007

Scheuerell, M., Staley, M., Wieckowski, K., Winton, J.R., and Wood, C.C. 2010. Synthesis of 1008

Evidence from a Workshop on the Decline of Fraser River Sockeye. June 15-17, 2010. Pacific 1009

Salmon Commission, Vancouver, BC. 1010

Peterman, R.M., Marinone, S.G., Thomson, K.A., Jardine, I.D., Crittenden, R.N., Leblond, 1011

P.H., and Walters, C.J. 1994. Simulation of juvenile sockeye salmon (Oncorhynchus nerka) 1012

migrations in the Strait of Georgia, British Columbia. Fish. Oceanogr. 3(4): 221-235. 1013

Page 47 of 65



Draft

48

Peterman, R.M., and Dorner, B. 2012. A widespread decrease in productivity of sockeye 1014

salmon (Oncorhynchus nerka) populations in western North America. Can. J. Fish. Aquat. Sci. 1015

69(8): 1255-1260. doi: 10.1139/F2012-063. 1016

Pinnix, W.D., Nelson, P.A., Stutzer, G., and Wright, K.A. 2013. Residence time and habitat 1017

use of coho salmon in Humboldt Bay, California: an acoustic telemetry study. Environ. Biol. 1018

Fishes. 96(2-3): 315-323. doi: 10.1007/s10641-012-0038-x. 1019

Preikshot, D., Beamish, R.J., Sweeting, R.M., Neville, C.M., and Beacham, T.D. 2012. The 1020

residence time of juvenile Fraser River sockeye salmon in the Strait of Georgia. Marine and 1021

Coastal Fisheries. 4(1): 438-449. doi: 10.1080/19425120.2012.683235. 1022

Price, M.H.H., Glickman, B.W., and Reynolds, J.D. 2013. Prey selectivity of Fraser River 1023

sockeye salmon during early marine migration in British Columbia. Trans. Am. Fish. Soc. 1024

142(4): 1126-1133. doi: 10.1080/00028487.2013.799517. 1025

Quinn, T.P. 2005. The behavior and ecology of Pacific salmon and trout. University of 1026

Washington Press, Seattle. 1027

R Core Team. 2014. R: A language and environment for statistical computing. R Foundation 1028

for Statistical Computing, Vienna, Austria. http://www.R-project.org. 1029

Rechisky, E.L., Welch, D.W., Porter, A.D., Hess, J.E., and Narum, S.R. 2014. Testing for 1030

delayed mortality effects in the early marine life history of Columbia River yearling Chinook 1031

salmon. Mar. Ecol. Prog. Ser. 496: 159-180. 1032

Page 48 of 65



http://www.r-project.org/

Draft

49

Rechisky, E.L., Welch, D.W., Porter, A.D., Jacobs-Scott, M.C., and Winchell, P.M. 2013. 1033

Influence of multiple dam passage on survival of juvenile Chinook salmon in the Columbia 1034

River estuary and coastal ocean Proc. Natl. Acad. Sci. USA. 110(17): 6883-6888. doi: 1035

10.1073/pnas.1219910110. 1036

Rensel, J.E.J., Haigh, N., and Tynan, T.J. 2010. Fraser river sockeye salmon marine survival 1037

decline and harmful blooms of Heterosigrna akashiwo. Harmful Algae. 10(1): 98-115. doi: 1038

10.1016/j.hal.2010.07.005. 1039

Romer, J.D., Leblanc, C.A., Clements, S., Ferguson, J.A., Kent, M.L., Noakes, D., and 1040

Schreck, C.B. 2013. Survival and behavior of juvenile steelhead trout (Oncorhynchus mykiss) in 1041

two estuaries in Oregon, USA. Environ. Biol. Fishes. 96(7): 849-863. doi: 10.1007/s10641-012-1042

0080-8. 1043

Ruggerone, G.T., and Connors, B.M. 2015. Productivity and life history of sockeye salmon 1044

in relation to competition with pink and sockeye salmon in the North Pacific Ocean. Can. J. Fish. 1045

Aquat. Sci. 72(6): 818-833. doi: 10.1139/cjfas-2014-0134. 1046

Thomas, A.C., Nelson, B.W., Lance, M.M., Deagle, B.E., and Trites, A.W. 2016. Harbour 1047

seals target juvenile salmon of conservation concern. Can. J. Fish. Aquat. Sci. 74(6): 907-921. 1048

doi: 10.1139/cjfas-2015-0558. 1049

Thomson, R.E. 1981. Oceanography of the British Columbia Coast. Canada Dept. of 1050

Fisheries and Oceans, Ottawa, ON. 1051

Page 49 of 65



Draft

50

Thomson, R.E., Beamish, R.J., Beacham, T.D., Trudel, M., Whitfield, P.H., and Hourston, 1052

R.A.S. 2012. Anomalous ocean conditions may explain the recent extreme variability in Fraser 1053

River sockeye salmon production. Marine and Coastal Fisheries. 4(1): 415-437. doi: 1054

10.1080/19425120.2012.675985. 1055

Tucker, S., Hipfner, J.M., and Trudel, M. 2016. Size- and condition-dependent predation: a 1056

seabird disproportionately targets substandard individual juvenile salmon. Ecology. 97(2): 461-1057

471. doi: 10.1890/15-0564.1. 1058

Urke, H., Kristensen, T., Ulvund, J.B., and Alfredsen, J. 2013. Riverine and fjord migration 1059

of wild and hatchery-reared Atlantic salmon smolts. Fish. Manage. Ecol. 20(6): 544-552. doi: 1060

10.1111/fme.12042. 1061

Weihs, D. 1973. Optimal Fish Cruising Speed. Nature. 245(5419): 48-50. doi: 1062

10.1038/245048a0. 1063

Welch, D.W., Ishida, Y., and Nagasawa, K. 1998. Thermal limits and ocean migrations of 1064

sockeye salmon (Oncorhynchus nerka): Long-term consequences of global warming. Can. J. 1065

Fish. Aquat. Sci. /J. can. Sci. Halieut. Aquat. 55(4): 937-948. 1066

Welch, D.W., Melnychuk, M.C., Rechisky, E.R., Porter, A.D., Jacobs, M.C., Ladouceur, A., 1067

McKinley, R.S., and Jackson, G.D. 2009. Freshwater and marine migration and survival of 1068

endangered Cultus Lake sockeye salmon (Oncorhynchus nerka) smolts using POST, a large-1069

scale acoustic telemetry array. Can. J. Fish. Aquat. Sci. 66(5): 736-750. 1070

Page 50 of 65



Draft

51

Welch, D.W., Melnychuk, M.C., Payne, J.C., Rechisky, E.L., Porter, A.D., Jackson, G.D., 1071

Ward, B.R., Vincent, S.P., Wood, C.C., and Semmens, J. 2011. In situ measurement of coastal 1072

ocean movements and survival of juvenile Pacific salmon. Proc. Natl. Acad. Sci. USA. 108(21): 1073

8708-8713. doi: 10.1073/pnas.1014044108. 1074

White, G.C., and Burnham, K.P. 1999. Program MARK: survival estimation from 1075

populations of marked animals. Bird Study. 46: 120-139. 1076

Young, N., and Matthews, R. 2010. The Aquaculture Controversy in Canada: Activism, 1077

Policy, and Contested Science. UBC Press, Vancouver, BC. 1078

Yurk, H., and Trites, A.W. 2000. Experimental attempts to reduce predation by harbor seals 1079

on out-migrating juvenile salmonids. Trans. Am. Fish. Soc. 129(6): 1360-1366. doi: 1080

10.1577/1548-8659(2000)129<1360:EATRPB>2.0.CO;2. 1081

1082

Page 51 of 65



Draft

52

Tables 1083

Table 1 1084

Summary of age-2 Chilko Lake sockeye salmon smolts captured and acoustic-tagged at the outlet of Chilko Lake, 2010-2014. 1085

Tag

Mean FL Mean weight Mean tag burden

Release type type N Release dates Transported (mm; SD) (g; SD) (%, SD)

2010

Lake released V7 199 May 2-8 No 129.8 (4.0) 17.3 (1.8) 9.4 (0.9)

2011

Lake released V6 200 Apr 29-May 9 No 127.2 (7.8) 15.5 (3.2) 6.7 (1.2)


Transport control V7 85 May 1 Yes: 0 km; 2 hrsa 133.6 (9.1) 18.1 (4.5) 9.2 (1.8)

Transport (Siwash) V7 104 May 3 Yes: 79 km; 2 hrsb 132.5 (7.0) 17.4 (3.0) 9.4 (1.5)

2012



2013

Lake released V7 203 Apr 25-29 No 123.3 (2.6) 14.5 (1.2) 11.1 (0.9)

Transport (Chilliwack) V7 229 May 8 Yes: 560 km; 10 hrsc 123.1 (2.3) 14.3 (1.0) 11.2 (0.7)

2014

Transport (upstream) V7 209 Apr 29-May 11 Yes: 1.3 km; ~1hrd 143.4 (9.0) 24.0 (4.6) 6.9 (1.4)

Lake released V7 113 Apr 28-May 10 No 148.3 (8.5) 26.7 (5.1) 6.2 (1.2) a Smolts were released at the tagging location as a control for the effect of truck-transport

b Smolts were released 350 m upstream of the Siwash Bridge array

c Smolts were released 36.5 km upstream from the Mission array

d Smolts were released 1.3 km upstream from the DFO fence

Note: The frequency of the V7 tag was 69 kHz, whereas the frequency of theV5 and V6 tags was 180 kHz (limiting their ability to be detected in

the ocean). Gill tissue samples were taken from approximately half of the fish in the V7 tagged fish in 2012 and 2013.

Page 52 of 65



Draft

53

Table 2 1086

Selection results for models with >5% model weight used to estimate survival (φ) in each 1087

migration segment and detection probability (p) at each detection site for Chilko Lake sockeye 1088

smolts. 1089

Year Dataseta φ p Npar QAICc ΔQAICc Weight

b

2010 FWSW segment site 11 528.08 0.00 0.39

segment + fl site 12 530.17 2.09 0.14

segment + burdCHILKO site 12 530.17 2.09 0.14

segment + burd site 12 530.17 2.09 0.14

segment + flFW + flOC site 13 530.71 2.63 0.11

segment + burdFW + burdOC site 13 531.11 3.04 0.09

2011 FW segment site:tag type:release type 27 2291.36 0.00 0.42

segment + burdCHILKO site:tag type:release type 28 2292.65 1.29 0.22

segment + fl site:tag type:release type 28 2293.01 1.65 0.18

segment + burd site:tag type:release type 28 2293.07 1.71 0.18

FWSW segment siteFW:release type + siteOC 21 1674.67 0.00 0.32

segment + burdCHILKO siteFW:release type + siteOC 22 1676.31 1.64 0.14

segment + fl siteFW:release type + siteOC 22 1676.42 1.76 0.13

segment + burd siteFW:release type + siteOC 22 1676.72 2.06 0.11

segment site:release type 23 1677.53 2.86 0.08

segment + burdFW + burdOC siteFW: release type + siteOC 23 1678.24 3.57 0.05

2012 FW segment + burdCHILKO site:tag type 18 1259.96 0.00 0.57

segment + burd site:tag type 18 1260.92 0.96 0.35

segment site:tag type 17 1264.83 4.87 0.05

FWSW segment + burdCHILKO site 12 1028.90 0.00 0.52

segment + burd site 12 1030.80 1.91 0.20

segment + burdFW + burdOC site 13 1032.01 3.11 0.11

segment + fl site 12 1032.56 3.66 0.08

segment site 11 1033.39 4.49 0.06

2013 FWSW segment + fl site:release type 16 1532.39 0.00 0.49

segment + flFW + flOC site:release type 17 1534.29 1.90 0.19


segment + burd site:release type 16 1536.10 3.70 0.08

segment + burdCHILKO site:release type 16 1536.68 4.28 0.06

2014 FWSW segment siteFW:release type + siteOC 17 905.88 0.00 0.27


segment + burdCHILKO siteFW:release type + siteOC 18 907.86 1.98 0.10

Page 53 of 65



Draft

54

segment + burd siteFW:release type + siteOC 18 907.87 1.99 0.10

segment + fl siteFW:release type + siteOC 18 907.95 2.07 0.10 a In years when more than one type of tag was used (2011 and 2012), we created two data sets: freshwater (FW; includes

freshwater detections from V5, V6 and V7 tags) and freshwater-saltwater (FWSW; includes only V7 detection data).

b Models with <5% model weight are not shown.

Note: We assessed whether fork length (fl) or tag burden affected survival soon after release in the Chilko River

(burdCHILKO), in fresh water (burdFW), in the ocean (burdOC), or in any migration segment (burd). We assessed whether

transportation (release type) affected detection probability. Npar= parameter count; QAICc= Akaike’s Information

Criteria with low sample size and modified for overdispersion; ΔQAICc= QAICc-QAICcmin.

Table 3 1090

Estimated survival (SE) of Chilko Lake sockeye smolts by habitat. 1091

Chilko River

(79 km)

Chilcotin River

(97 km)

Fraser River

(478 km)

Chilcotin + Fraser River

(575 km)

CSOGc

(149 km)

NEVId

(240 km)

2010 0.57 (0.05) a a

0.38 (0.06) 0.74 (0.26) 0.24 (0.16)

2011 0.63 (0.02) 0.65 (0.03) 0.86 (0.04) b

0.47 (0.11) 0.53 (0.15)

2012 0.68 (0.04) 0.66 (0.07) 0.81 (0.08) b

0.76 (0.19) 0.41 (0.14)

2013 0.78 (0.03) a a

0.62 (0.07) 0.83 (0.1) 0.36 (0.07)

2014 0.60 (0.04) a a

0.64 (0.07) 0.59 (0.13) 0.37 (0.14) a The Farwell Canyon receivers were not deployed (2010) or did not detect enough fish to estimate survival.

b Not applicable in years when we could separate Chilcotin and Fraser River survival.

c Central Strait of Georgia between the Fraser River mouth and the Northern Strait of Georgia (NSOG) array

d North-eastern Vancouver Island between the NSOG and Queen Charlotte Strait (QCS) arrays.

1092

1093

1094

Page 54 of 65



Draft

55

Figures Captions 1095

Fig. 1 1096

Map of the study area indicating the positions of acoustic receiver arrays (yellow circles and 1097

lines) and release location of acoustic-tagged smolts (pink stars). All smolts were collected at the 1098

outlet of Chilko Lake. In all years, multiple groups of smolts were released at the capture 1099

location near the Release array (Array A). In 2011, two groups of smolts were transported: one 1100

group was release near Array C and one group was release near the capture site and served as a 1101

control group for the transported group. In 2013, a group of smolts was transported to the lower 1102

Fraser River. Array D receivers were not deployed in 2010. The Lippy Point array extended 1103

across the continental shelf out to approximately the 200-m isobath in 2010, and the 500-m 1104

isobath in 2011; it was not deployed in other years. Isobaths are 200-m and 500-m depth. 1105

CSOG= central Strait of Georgia, NEVI=north-eastern Vancouver Island. Map data sources: 1106

Esri; Lakes and bathymetry: Government of Canada (Natural Resources Canada); Fraser River: 1107

Government of British Columbia (Ministry of the Environment). 1108

Fig. 2 1109

Distribution of size (FL) and tag burden of acoustic-tagged, age-2 Chilko Lake sockeye 1110

smolts. Tag burden was calculated as the tag-to-body mass ratio in air. Note the bi-modal 1111

distribution of tag burden in 2012 when two different tag sizes were used. 1112

Fig. 3 1113

Survival analysis flow chart for acoustic-tagged Chilko Lake sockeye tracked from 2010-1114

2014. *In 2011, the transport control (TC) group had higher survival than the lake released (LR) 1115

Page 55 of 65



Draft

56

group (not lower); therefore, we pooled the groups despite the difference. FW=freshwater; 1116

FWSW= freshwater saltwater; GOF=goodness of fit; T=transported; FL= fork length. 1117

Fig. 3 1118

Predicted survival when accounting for a tag burden effect in 2012 and a body size (fork 1119

length) effect in 2013. Note that the tag burden effect was strongest in the Chilko River (Henry’s 1120

Bridge and Siwash Bridge) 2012 as indicated in the model selection results. 1121

Fig. 4 1122

Comparative survival (a, b) and survival rate (c, d) of acoustic-tagged Chilko Lake sockeye 1123

smolts from 2010-2014 in three regions: fresh water (FW), the central Strait of Georgia (CSOG; 1124

between the Fraser River mouth and the northern Strait of Georgia [NSOG] array), and north-1125

east Vancouver Island (NEVI). Plot (a) shows aggregate survival estimates combined for all 1126

years; plot (b) shows the data for individual years. Error bars show 95% confidence limits. Note 1127

that much of the FW mortality occurred in the tributaries; see Clark et al 2016. FW is from 1128

release to the Fraser Mouth; CSOG is from the Fraser Mouth to the NSOG array; NEVI is from 1129

the NSOG array to the QCS array. 1130

Fig. 5 1131

Cumulative survival of acoustic-tagged Chilko Lake sockeye smolts from release at Chilko 1132

Lake to each detection site. Note that we could not separate survival in the Chilcotin and Fraser 1133

rivers in some years because either the receivers were not deployed at the confluence (2010) or 1134

detection efficiency was too poor to reliably estimate survival (2013 and 2014). SOG=Strait of 1135

Georgia, NEVI=north-east Vancouver Island. 1136

Fig. 6 1137

Page 56 of 65



Draft

57

Travel time of Chilko Lake sockeye smolts for all years combined. Only lake released smolts 1138

are included in estimates starting at the Chilko Lake release site (Release); lake released and 1139

transported smolts are pooled in the two ocean segments. FRM=Fraser River mouth, 1140

NSOG=northern Strait of Georgia array, QCS=Queen Charlotte Strait array. 1141

Fig. 7 1142

Survival rates for acoustic-tagged Cultus Lake (black circles) and Chilko Lake (red squares) 1143

sockeye in two marine areas: CSOG (central Strait of Georgia; from the Fraser River mouth to 1144

the Northern Strait of Georgia [NSOG] array) and NEVI (north-eastern Vancouver Island; the 1145

marine area which includes the Discovery Islands, Johnstone Strait and Queen Charlotte Strait 1146

bounded by the NSOG and QCS arrays). Cultus Lake sockeye rates were calculated from 1147

survival estimates in Welch et al. (2009). The dashed 1:1 line represents equal survival in the two 1148

areas. 1149

Page 57 of 65



Draft

Page 58 of 65



Draft

d2013fl$x

d201

3fl$

y

100 120 140 160

0.00

0.05

0.10

0.15

Fork Length (mm)

Den

sity

(a)

d2013burden$x

d201

3bur

den$

y

0 5 10 15 20

0.0

0.1

0.2

0.3

0.4

0.5

Tag Burden (%)

(b) 20102011201220132014

Page 59 of 65



Draft

Yes, burden

Yes, burden

No

No No No

Tre

atm

ent

gro

up

s 1

. D

ata

scre

enin

g

4.

Surv

ival

mo

del

s 2

. W

ithin

yea

r in

itia

l ass

essm

ent

5.

Mo

del

sel

ecti

on a

nd

aver

agin

g

Surv

ival

est

imat

es

V6 & V7 groups pooled

for analysis

V7 groups pooled for

analysis

Segment

varying

Ind covari-

ates (FL and tag burden)

Segment

varying

Ind covari-


Difference in

survival of

treatment

groups?

Yes*

Difference

in survival

of V7 LR

and TC?

V6 and V7

detection data

(FW data set)

V7 detection

data (FWSW

data set)

Screened detection

data

V6 lake release V7 lake release

V7 transport control

V7 transport

GOF test GOF test

Segment

varying

Ind covari-


Screened detection

data

V7 lake release

Model average

at mean FL and

mean burden

Effect of

FL or tag

burden on

survival?

GOF test

2010 2011 2012 2013

Model average

at mean FL and

mean burden

Model average

at mean FL and

mean burden

Extract river estimates

Effect of

FL or tag

burden on

survival?

Extract marine estimates

Effect of

FL or tag

burden on

survival?

V5 and V7

detection data

(FW data set)

V7 detection

data (FWSW

data set)

Screened detection

data

V5 lake release

V7 lake release

GOF test GOF test

V5 & V7 groups pooled

for analysis

Segment

varying

Ind covari-


Segment

varying

Ind covari-


Difference in

survival?

No

Model average

at mean FL and

25th percentile

burden

Model average

at mean FL and

25th percentile

burden

Extract river estimates

Effect of

FL or tag

burden on

survival?

Extract marine estimates

Effect of

FL or tag

burden on

survival?

Segment

Difference

in survival

of LR and

T?

No

Yes, FL

Segment

varying

Ind covari-


Screened detection

data

V7 lake release

V7 transport

Model average

at mean FL and

mean burden

Effect of

FL or tag

burden on

survival?

GOF test

Difference in

survival?


analysis

No

2014

Segment

varying

Ind covari-


Screened detection

data

V7 lake release

V7 transport

Model average

at mean FL and

mean burden

Effect of

FL or tag

burden on

survival?

GOF test

Difference in

survival?


analysis

No

No

2.

GO

F

Cumulative Cumulative

Habitat specific Habitat specific

Segment

Habitat specific

Segment

Cumulative

Segment

Cumulative

Habitat specific

Segment

Cumulative

Habitat specific

Page 60 of 65



DraftFraser Mouth

Farwell Mission

Henrys Bridge Siwash Bridge

6 8 10 12 14 16

6 8 10 12 14 16

25

50

75

100

25

50

75

100

25

50

75

100

Tag Burden (%)

Sur

viva

l (%

)(a)

NSOG QCS

Mission Fraser Mouth

Henrys Bridge Siwash Bridge

120 125 130 135 120 125 130 135

25

50

75

100

25

50

75

100

25

50

75

100

Fork Length (mm)

(b)Page 61 of 65



Draft0

20406080

100A

ggre

gate

d S

urvi

val (

%)

020406080

100

Sur

viva

l (%

)

2010

2011

2012

2013

2014

2010

2011

2012

2013

2014

2010

2011

2012

2013

2014

020406080

100

Sur

v pe

r 10

0 km

(%

)

2010

2011

2012

2013

2014

2010

2011

2012

2013

2014

2010

2011

2012

2013

2014

020406080

100

Sur

v pe

r w

eek

(%)

FW CSOG NEVI

2010

2011

2012

2013

2014

2010

2011

2012

2013

2014

2010

2011

2012

2013

2014

(a)

(b)

(c)

(d)

Page 62 of 65



Draft

c(10

0, x

_201

2$C

um_P

rod) ●

●

●

●

● ●

●

●

0 100 200 300 400 500 600 700 800 900 1000 1100

0102030405060708090

100 ●

●

●

● ●

●

●

●

●

●

●

● ●

●

●

Chilko Chilcotin Fraser CSOG NEVI

●

●

20102011201220132014

Sur

viva

l (%

)

Distance from Chilko Lake release site (km)

Page 63 of 65



DraftNSOG to QCS

FRM to NSOG

Release to FRM

0 10 20 30 40 50

0

50

100

150

200

250

0

5

10

15

20

0

3

6

9

12

Num

ber

of fi

sh(a) Segment

Release to QCS

Release to NSOG

Release to FRM

0 10 20 30 40 50

0

50

100

150

200

250

0

10

20

30

0

5

10

(b) Cumulative

Travel time (days)

Page 64 of 65



Draft

higher in NEVI

higher in CSOG

●

●●●

higher in NEVI

higher in CSOG

●

●

●

●

Survival per 100 kms Survival per week

0 25 50 75 100 0 25 50 75 100

0

25

50

75

100

CSOG Survival (%)

NE

VI S

urvi

val (

%)

Page 65 of 65



Draft...Draft 3 30 Abstract 31 We used acoustic telemetry to investigate survival of age-2 sockeye...

Documents

Transcript of Draft...Draft 3 30 Abstract 31 We used acoustic telemetry to investigate survival of age-2 sockeye...