Lower Coquitlam River Fish Productivity Index

Transcription

1 Coquitlam River Water Use Plan Lower Coquitlam River Fish Productivity Index COQMON#7 Study Period: Report Date: March 2008 Scott Decker, Jason Macnair, Gordon Lewis, Jody Schick and Josh Korman March 2008

2 Coquitlam River Fish Monitoring Program: Results Prepared for: BC Hydro Prepared by: Scott Decker 1 *, Jason Macnair 2, Gordon Lewis 3, Jody Schick 4, and Josh Korman 5 February Fraser Street, Kamloops B.C., V2C 3H7, decker_scott@hotmail.com Ontario St., Vancouver B.C. V5T 2Y5, jacemacnair@yahoo.ca Pipeline, Coquitlam, B.C, V3E 2X1, GMaintenance@MidBC.com Gower Point Rd. Gibsons, B.C., V0N 1V0, jschick@dccnet.com 5 Ecometric Research Inc., 3560 W 22 nd Avenue, Vancouver, BC V6S 1J3 * Author to whom correspondence should be addressed

3 ii EXECUTIVE SUMMARY As part of the Coquitlam River Water Use Plan (LB1 WUP), a long-term adaptive management study is being conducted in Coquitlam River to compare anadromous fish production under two experimental flow regimes. Fish population monitoring under the first flow regime (Treatment 1) began in 2000 and will continue until the completion of the Coquitlam Dam seismic upgrade, currently scheduled for September Fish production under Treatment 2 will be monitored for a 7-9 year period. The Coquitlam River Monitoring Program (CRMP) focuses on four anadromous species: coho, steelhead, chum and pink. During adult escapement and smolt outmigration were monitored for each species. In 2006 and 2007, we also conducted night snorkelling surveys to estimate late summer standing stocks of juvenile coho and steelhead. During , adult and smolt numbers were also monitored, but not for all species in all years. This report describes and critiques the study design and sampling methodologies for the four major components of the CRMP: adult salmon AUC escapement surveys, steelhead redd counts, juvenile standing stock surveys, and smolt outmigrant enumeration. For the last three components, primary emphasis is on work completed in 2007, but a summary of all years of available abundance data for each species and life stage is also provided. New escapement estimates are presented for adult salmon and steelhead based on the reanalysis of previous years data. Escapement and egg-to-smolt survival estimates for each species should be considered highly preliminary and will likely change as adult salmon observer efficiency and survey life data are accumulated in future years. During , frequent high flow events and associated high turbidity during the fall and winter spawning period resulted in frequently missed and partially completed surveys for adult salmon and contributed substantially to uncertainty in escapement estimates. A new maximum likelihood model that assumed a negative binomial error structure in the data was useful in incorporating counts from partial surveys, and provided good fits of predicted to observed counts for chum, pink and coho, and escapement estimates with realistic 95% confidence bounds. The model will likely require additional parameters and further development in future to allow for more complexity in arrival timing and within-year variation in the distribution of spawners among survey sections. In addition, a pilot mark-recapture study initiated in 2006 to provide Coquitlam River-specific estimates of observer efficiency and survey life will need to be expanded significantly. It is anticipated that as parameters are refined, historic estimates will be revised at the end of the review period. Results from steelhead redd surveys conducted during suggest that redd surveys are an effective method for indexing steelhead spawner abundance and potential egg deposition in Coquitlam River. A simple regression

4 iii model was useful for expanding raw counts to correct for undercounting of redds that occurred during periods of infrequent surveys when some redds were obscured by substrate before they could be detected. Steelhead spawner density varied from females/km during Survey results from 2006 and 2007 indicated that late summer populations of juvenile coho and steelhead can be effectively sampled using night snorkeling counts. Limited mark-recapture experiments in 2007 indicated relatively high snorkeling efficiency (proportion of fish present that snorkelers detect) levels for coho and steelhead fry, and steelhead parr (52%, 38% and 72%, respectively). Continued use of night snorkeling surveys in future years to estimate juvenile standing stocks, and mark-recapture experiments to calibrate snorkeling counts are recommended. For all four species, the smolt trapping program provided sufficiently precise estimates of smolt production during to meet the requirements of the management experiment. For coho, steelhead, chum, and pink, precision of the annual smolt abundance estimates ranged from ± 7%-25%, 15%-34%, 17%- 25%, and 22%-52%, respectively. Coho escapement to Coquitlam River (600 to 1,600 adults; 23 to 63 females/km) likely exceeded that necessary to seed available juvenile habitat, with the possible exception of the 2006 smolt year, for which brood escapement and smolt production were considerably lower than in other years. During , annual coho smolt numbers varied from 9,500 to 19,400 fish. Snorkeling surveys provided late summer population estimates of 32,000 and 29,000 coho fry in 2006 and 2007, respectively. Coho egg-to-smolt survival was low ( %) relative to that in other coastal streams. Redd counts suggested that steelhead escapement during ( adults; females/km) was well above that necessary to seed available juvenile habitat based on stock and recruitment data for Keogh River. Steelhead smolt output from Coquitlam River ranged from 4,400 to 7,700 fish. Snorkeling surveys provided late summer population estimates of 169,000 and 47,000 steelhead fry, and 7,100 and 14,900 steelhead parr in 2006 and 2007, respectively. Steelhead egg-to-fry survival ( %) and egg-to-parr survival (0.8%) were higher than the long-term averages for Keogh River. Adult chum escapements ranged from 14,000 to 36,000 and resulted in outmigrations of 1.3 to 4.8 million smolts. Egg-to-smolt survival ranged from 3.9% to 13.8% and averaged 7.8%, which was comparable to that for chum populations in other streams. Adult pink escapements ranged from 2,300-3,800 and resulted in outmigrations of 210, ,000 smolts. Egg-to-smolt survival ranged from 9.9% to 15.1%, which was higher than the average for pink populations in other

5 streams. Chum and pink salmon returns to Coquitlam River were greatly improved in compared to escapements in years prior to the implementation of the Treatment 1 flow regime in iv

6 v TABLE OF CONTENTS EXECUTIVE SUMMARY... ii LIST OF TABLES... vii LIST OF FIGURES... ix LIST OF APPENDICES... xii 1.0 INTRODUCTION Adult salmon escapement Adult steelhead escapement Juvenile coho and steelhead standing stock Smolt outmigrant trapping ADULT SALMON ESCAPEMENT Methods Stratified index survey design Observer efficiency, survey life and incomplete surveys Escapement model structure and parameter estimation Process Model Observation model and estimation of model parameters Results and Discussion Observer efficiency and survey life Escapement Model Adult habitat distribution Temperature Summary and recommendations for 2007 adult salmon survey Adult Winter steelhead escapement Methods Description of study area and survey methods Redd Identification Redd survey life Female escapement and egg deposition Results and Discussion Redd survey life Female escapement and egg deposition Recommendations for the 2007 survey Juvenile salmonid standing stock Methods Study design Night snorkeling counts Mark-recapture experiments to estimate snorkeling efficiency Physical characteristics of the sampling sites Standing stock estimation Results Mark-recapture experiments to estimate snorkeling efficiency Juvenile fish distribution and abundance Fish density versus habitat...81

7 4.3 Discussion Suitability of snorkeling counts for juvenile population census Fish density and standing stock estimation Recommendations for the 2007 survey SMOLT PRODUCTION Methods Coho and steelhead smolt enumeration Location and description of downstream traps Downstream trap operation Differential marking by period and initial capture location Population estimates Mark-recapture assumptions Chum and pink smolt enumeration Downstream trapping Differential marking over time Population estimates Results Off-channel sites Coquitlam River mainstem Coho Steelhead Chum Kokanee Steelhead smolt size Discussion Reliability of estimates and implications for the flow experiment Assumptions of the study design Recommendations for smolt trapping in Annual trends in FISH PRODUCTION ( ) Coho and steelhead Chum and pink REFERENCES vi

8 vii LIST OF TABLES Table 1.1 Scheduled monthly flow releases from Coquitlam Dam under Treatments 1 and 2 of the Coquitlam River Water Use Plan (BC Hydro 2003a).14 Table 2.1 Results of the 2006 mark-recapture study to estimate observer efficiency and survey life for chum salmon in Coquitlam River Table 2.2 Estimated proportion of chum, pink and coho spawning populations present at each index site (A-E) and at non-index (NI) sites (γ i ) based on yearindependent and joint-year estimation Table 2.3 Comparison of model fit based on Akaikie s second order Information Criteria (AIC c ) for two alternate approaches (year-independent and joint-year) of estimating the distribution of the chum, pink and coho salmon spawning population among survey sites in Coquitlam River. Results are shown for the Poisson escapement model only Table 2.4 Escapement estimates for chum, pink and coho salmon generated using the Poisson and negative binomial versions of the year-independent escapement model. Also shown are 95% confidence intervals, spawner densities, and standardized Pearson residuals (SDNR, the standard deviation of normalized residuals; and MAR, the median of their absolute values), which describe model fit. Uncertainty in escapements is shown as 95% confidence intervals and percent relative error (average 95% confidence interval divided by the estimate) Table 2.5 Adult spawning distribution by habitat type. Proportions shown are calculated based on counts of actively spawning fish only, during surveys when all five index sites were completed. M/S = mainstem, NOC = natural off-channel, OCR = off-channel restoration site, and OCC = off-channel sites combined Table 3.1 Survey dates with raw and estimated steelhead redd counts and adult observations for Table 3.2 Summary statistics for steelhead escapement to Coquitlam River during based on redd counts Table 4.1 Snorkeling efficiency data sources used to generate population estimates for Coquitlam River coho and steelhead parr and fry Table 4.2 Summary of habitat data for night snorkeling sampling sites in Coquitlam River....75

9 viii Table 4.3 Summary of mark-recapture results and snorkeling efficiency estimates for four sampling sites in Coquitlam River in Table 4.4 Preliminary estimates of juvenile fish density, standing stock, and 95% confidence intervals by species and age class in Coquitlam River in 2006 and Table 5.1 Description of the stratification of fish marking by location and period for coho and steelhead smolts in the Coquitlam River in The start date for each temporal marking period at each RST trap site is also shown. RST installation dates correspond to the start dates for mark group 1. RST removal dates are also given...93 Table 5.2 Summary of estimated smolt numbers and densities by species in 2007 for three off-channel sites, reaches 2-4 of the Coquitlam River mainstem and the total Coquitlam River mainstem including and excluding the off-channel sites Table 5.3 Differences in capture efficiency (proportion of marked smolts that were recaptured) for coho and steelhead from off-channel sites and the Coquitlam River mainstem at three rotary screw traps (RSTs) sites in the Coquitlam River mainstem in Stratified marking periods were pooled prior to testing (see Equation 5.1). Equal capture efficiency for mark groups was tested using Fisher s exact test. P < 0.05 indicates a significant difference in capture efficiency Table 6.1 Summary of all population estimates for all life stages and species in Coquitlam River, Survival estimates correspond to fry, parr or smolt year rather than brood year

10 ix LIST OF FIGURES Figure 1.1 Life stage periodicity chart for anadromous salmonids in Coquitlam River Figure 1.2 Map of lower Coquitlam River study area with stream reaches defined by the Coquitlam-Buntzen Water Use Plan Consultative Committee...17 Figure 2.1 Map showing adult spawning index sites A-C in the lower portion of Coquitlam River study area (reaches 1, 2a)...23 Figure 2.2 Map showing adult spawning index sites D and E, in the upper portion of Coquitlam River study area (reaches 2b, 3 and 4)...24 Figure 2.3 Predicted relationship between survey life and day of arrival in the study area for chum, pink and coho salmon in Coquitlam River Figure 2.4 Predicted 2005 chum salmon arrival and departure timing (A), predicted numbers present and observer counts expanded by observer efficiency (B), and the estimated proportion of the population that was surveyed, and observed (blue) and estimated (red and black) proportions of the population in index sites based on year-independent estimation (C). The rows at the top of graph C are, from top to bottom, raw survey counts, number of sites surveyed, and the estimate proportion of the total population that was sampled. The column of numbers at the far right shows the model-estimated proportion assigned to each site, which is constant across surveys within each year. The size of the solid blue circle represents the proportion of the observed total count at each site for each survey, with no correction for unsampled sites. The open red circles are the model predicted proportion, adjusted so that they sum to one over the sites that were sampled on each survey (to facilitate comparison with observed counts when the number of sites sampled is less than 6). The black open circles are the estimated proportions that were used in the escapement calculations, they are the same for each survey. The unadjusted and adjusted model proportions are equivalent on surveys where all 6 sites were enumerated, and for these surveys, only the black circles are visible...36 Figure 2.5 Predicted 2005 pink salmon arrival and departure timing (A), predicted numbers present and observer counts expanded by observer efficiency (B), and the estimated proportion of the population that was surveyed, and observed (blue) and estimated (red and black) proportions of the population in index sites based on year-independent estimation (C). See Figure 2.4 caption for details regarding graph C...37 Figure 2.6 Predicted 2005 coho salmon arrival and departure timing (A), predicted numbers present and observer counts expanded by observer efficiency

11 x (B), and the estimated proportion of the population that was surveyed, and observed (blue) and estimated (red and black) proportions of the population in index sites based on year-independent estimation (C). See Figure 2.4 caption for details regarding graph C Figure 2.7 Examples of relatively good fit of predicted numbers of adult salmon present to observer counts expanded by observer efficiency for the negative binomial escapement model Figure 2.8 Cases where the negative binomial escapement model produced relatively poor fits of predicted numbers of adult salmon present to observer counts expanded by observer efficiency Figure 2.9 Posterior distributions (diagonal) and correlations of escapement (esc), run timing (μ and τ) and proportion of the population at index sites A-E and non-index (ni) sites. Results are based on 500 samples drawn from 25,000 MCMC simulations based on the year-independent version of the negative binomial model applied to the 2005 coho data. The mean escapement estimate from the posterior was 747 with a 95% confidence intervals of 295. The most likely estimate (Table 2.4) was 747 adults with a 95% confidence intervals of ±295 adults Figure 3.1 Steelhead redd locations in reaches 2b-4 in Coquitlam River in 2006, which was the highest escapement year during the Coquitlam Dam is the upstream boundary of the survey area. See Figure 3.2 for redd symbol legend Figure 3.2 Steelhead redd locations in reaches 2a-2b in Coquitlam River in The downstream boundary of reach 2a is also the survey boundary...54 Figure 3.3 Discharge (cms) in Coquitlam River during steelhead spawning period in Figure 3.4 Unadjusted number of new steelhead redds observed on each survey (solid line and circles) and cumulative proportion of the total redd count observed on each annual survey (broken line amnd open circles) during 2005 and Figure 3.5 Peak counts and densities (fish/km) of adult steelhead during snorkeling surveys in (only data for complete surveys of the study area are shown). Data for were collected as part of a separate study (BCCF, Lower Mainland Branch, data on file)...61 Figure 3.6 Proportion of newly constructed steelhead redds that remained visible to observers during subsequent surveys in Coquitlam River during

12 xi Figure 4.1 Map of Coquitlam River showing juvenile standing stock study area, reach breaks and sampling sites...70 Figure 4.2 Map of Coquitlam River showing juvenile standing stock study area, reach breaks and sampling sites...77 Figure 4.3 Comparison of fish density estimates for 10 adjacent shoreline and channel-wide sites in Coquitlam River. Estimates are based on calibrated snorkeling counts...79 Figure 5.1 Map of the Coquitlam River showing constructed off-channel habitat sites, mainstem reach breaks and the locations of mainstem rotary screw traps (RSTs) in 2007 and years previous...86 Figure 5.2 Daily catches of coho smolts at downstream weirs in three offchannel sites (pooled data) and at three rotary screw trapping locations in the Coquitlam River mainstem in See Table 5.1 for start and end dates for individual trapping sites...95 Figure 5.3 Daily catches of steelhead smolts at downstream weirs in three offchannel sites (pooled data) and at three rotary screw trapping locations in the Coquitlam River mainstem in See Table 5.1 for start and end dates for individual trapping sites...96 Figure 5.4 Mean daily flows in Coquitlam River at Port Coquitlam during the smolt trapping period in (Water Survey of Canada, stn. 08MH141) Figure 5.5 Daily catches of chum smolts at two incline plane traps located at the RST2 trapping site in reach 2 and age-1 kokanee at RST2 in Coquitlam River in See Table 5.1 for start and end dates for downstream trapping Figure 5.6 Estimated capture efficiencies (across six marking periods) at three rotary screw traps (RSTs) in the Coquitlam River for marked coho and steelhead smolts from off-channel (dotted lines) and mainstem (solid lines) habitats in Dates on the horizontal axis indicate the start point for each marking period Figure 6.1 Annual numbers of coho and steelhead smolts in reach 4 of Coquitlam River Figure 6.2 Mean annual fork lengths for coho smolts and steelhead smolts and parr in different habitats in the Coquitlam River, Error bars represent ± 1 standard error

13 xii LIST OF APPENDICES Appendix 2.1 Adult chum salmon unadjusted counts during Appendix 2.2 Adult coho salmon unadjusted counts during Appendix 2.3 Adult pink salmon unadjusted raw counts during 2003 and Appendix 2.4 Mean daily temperatures in reaches 2,3, and 4 in Coquitlam River during July 1, 2005 December 31, Appendix 3.1 An example of how raw survey counts were expanded to account for redds that were completed and subsequently became undetectable between surveys (see section 3.2.1) Appendix 5.1 Summary of estimated numbers of coho, steelhead, chum and kokanee passing three RST trapping locations (not reach estimates) in the Coquitlam River mainstem in 2007, with separate estimates for each mark group. Mark group indicates the location where fish were initially captured and marked. Also shown are numbers of marked (M) and recaptured (R) smolts, estimated capture efficiencies (R/M), 95% confidence intervals, and percent relative errors Appendix 6.1 Supplemental Ministry of Environment document summarizing results for a separate electrofishing survey conducted during September

14 INTRODUCTION The lower Coquitlam River flows 17 km from the base of Coquitlam Dam to its confluence with the Fraser River. The stream was first dammed in The present dam dates from As part of Coquitlam-Buntzen Water Use Plan agreement completed in 2003 (LB1 WUP; BC Hydro 2003a), flows in the lower Coquitlam River are regulated through the Coquitlam Dam s low-level outlets that release flows from Coquitlam Reservoir. The Coquitlam Reservoir also supplies drinking water for the Greater Vancouver Regional District (GVRD) and water for power via a diversion tunnel to Buntzen Lake. Typical of lotic habitats downstream of dams, spawning and rearing habitat in lower Coquitlam River (hereafter referred to as simply Coquitlam River) has been impacted over the last hundred years by reduced gravel recruitment from upstream sources and increased sedimentation due to reduced peak flows (NHC, 2001). Several adjacent gravel pit operations adjacent to Coquitlam River also contribute large amounts of fine sediment directly to the stream. Other impacts are typical of urban streams, and include extensive channelization and dyke construction, road and bridge crossings, alteration of natural drainage patterns and discharge of pollutants. Peak, post-dam flows in Coquitlam River can exceed 200 cms, with normal flows ranging from 2-10 cms (Water Survey of Canada, Station 08MH141). Prior to June 1997, flow releases from the dam ranged from 0.06 to 0.5 cms (not including occasional spill events). From 1997 to 2007, flow releases were increased to 0.8 to 1.9 cms, depending on the time of year. This represents the current regime of two fish valves fully open, and is the baseline for an adaptive management study. Throughout the report we refer to this baseline regime as Treatment 1. The Coquitlam River historically supported all six Pacific salmon, as well as cutthroat trout (Oncorhynchus clarki) and Dolly Varden (Salvelinus malma) char. Dam construction resulted in the extirpation of summer sockeye (O. nerka), but this stock may still exist in the watershed in as resident kokanee in Coquitlam Reservoir. Other species inhabiting Coquitlam River below the dam include longnose dace (Rhinichthys cataractae), sculpins (cottus spp.), Pacific lamprey (Entosphenus tridentatus), and three-spine stickleback (Gasterosteous aculeatus). As part of the LB1 WUP, The Coquitlam-Buntzen Water Use Plan Consultative Committee (COQWUPCC) made recommendations on dam releases based on trade-offs between power, drinking water and Coquitlam River fisheries values (BC Hydro 2003). The LB1 WUP was also designed as a longterm adaptive management experiment to compare different flow regimes. The effect of different flows and other types of enhancements on the productivity of anadromous salmonid populations are often difficult to detect because of the high degree of natural variation in both freshwater and ocean survival (Keeley and

15 14 Walters 1994; Bradford 1995). Relying on a study by Higgins et al. (2002) that looked at the statistical power to detect changes in fish production in Coquitlam River under different flow regimes, the COQWUPCC selected two flow regimes for comparison: the current regime of two fish valves fully open (Treatment 1), and a new schedule of monthly flow releases prescribed by COQWUPCC (Treatment 2; Table 1.1) that attempts to improve spawning and rearing habitat conditions in Lower Coquitlam River relative to Treatment 1. Table 1.1 Scheduled monthly flow releases from Coquitlam Dam under Treatments 1 and 2 of the Coquitlam River Water Use Plan (BC Hydro 2003a). Reservoir diversion schedule (m 3 /s) Domestic water Coquitlam Dam releases Treatment Treatment 2 Target Min Target species and life Period 1 Target Min stage Jan Chinook spawning Jan Chinook incubation Feb Chinook incubation Mar Steelhead spawning Apr Steelhead spawning May Steelhead spawning Jun Steelhead parr Jul Steelhead parr Aug Steelhead parr Sep Steelhead parr Oct Chinook spawning Nov Chinook spawning Dec Chinook spawning Prior to the implementation of the monitoring program, COQWUPCC evaluated several potential flow regimes using flow-habitat models for target species and life histories, with habitat treated as a surrogate for fish productivity (BC Hydro 2003b). Habitat modelling suggested that, under Treatment 2, increased flow releases in late summer could increase the quantity and quality of juvenile rearing habitat for species with long freshwater residency periods, and that fall and spring releases could improve spawning success for all anadromous salmonids. To determine if habitat predictions translated into increased fish abundance, COQWUPCC took an empirical approach by implementing the Coquitlam River Monitoring Program (CRMP), a 16-year stock assessment program that focused on several life history stages for several species. The Treatment 1 flow regime will be evaluated for eight years ( ; monitoring did not occur in 2001), and Treatment 2 flow regime, for up to 9 years following the conclusion of Treatment 1 ( ). The initiation of Treatment 2 is currently tied to the completion of the Coquitlam Dam seismic upgrade project, which is scheduled for August 2008.

16 15 The CRMP focuses on four species: coho (O. kisutch), steelhead (O. mykiss), chum (O. keta), and pink (O. gorbuscha). Other fish species are either of too low abundance to effectively monitor, or are considered to be of lower economic, recreational, or cultural importance. Adult escapement and smolt outmigration are monitored for all four target species. In addition, beginning in 2006, fall juvenile standing stock was assessed for coho and steelhead. Coho and steelhead smolt production is the primary performance measure for the flow experiment. Coho and steelhead have lengthy freshwater residencies relative to other anadromous salmonids in Coquitlam River, and smolt production for these species was judged to be the best indicator of the effects of flow management and dam operation on freshwater production. Coho and steelhead smolt production appears to be limited by habitat carrying capacity at all but very low levels of adult escapement (Bradford and Taylor 1996; Ward and Slaney 1993). However, if adult returns are insufficient to seed available juvenile habitat, then recruitment effects may confound the relationship between smolt production and habitat. Monitoring escapement in addition to smolt production for coho and steelhead allows freshwater production to be evaluated under a scenario of recruitment-limited smolt production by substituting smolts/spawner or egg-tosmolt survival for absolute smolt production. Monitoring fall standing stock of coho and steelhead, together with smolt production, is potentially useful in addressing questions about freshwater production bottlenecks in Coquitlam River (e.g., Is overwintering habitat more important than summer rearing habitat in limiting juvenile carrying capacity?). For chum and pink, which emigrate to saltwater shortly after emergence, habitat conditions in Coquitlam River determine the quantity and quality of available spawning substrate and incubation conditions for eggs. For these species, smolt production and egg-to-smolt survival are the most important indicators of freshwater production. Figure 1.1 provides a periodicity chart for different life stages of anadromous salmonids in Coquitlam River. Species Steelhead Coho Chum Pink Chinook Month Spawning Incubation Rearing (parr) Rearing (fry) Spawning Incubation Rearing (fry) Spawning Incubation Outmigration Spawning Incubation Outmigration Spawning Incubation Rearing Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Figure 1.1 Life stage periodicity chart for anadromous salmonids in Coquitlam River. Dec

17 16 The CRMP is focused on the effects of dam releases on fish productivity in mainstem habitat in reaches 2a, 2b, 3 and 4, of Coquitlam River (Figure 1.2). This section contains the majority of productive spawning and rearing habitat in the Coquitlam River mainstem (Riley et al. 1997; Macnair 2005). The actual boundaries of the study area vary somewhat among components of the monitoring program due to sampling constraints or species distribution (see Sections ). Within reaches 2-4, spawning and rearing for steelhead, chum and pink is largely confined to the mainstem (Macnair 2005; Decker et al. 2006). Or Creek, a high gradient, nutrient-poor stream, with limited access, is the only significant tributary (Figure 1.2). There are several other tributaries, but they are very small, with accessible lengths limited to a few hundred metres. In addition to natural habitat, six large off-channel habitats, totalling about 27,000 m 2 of habitat have been constructed in reaches 2-4 (Decker and Foy 2000). The contribution of tributaries and off-channel sites to production of steelhead, chum and pink is low, but off-channel sites are used extensively by coho for spawning and rearing. Constructed off-channel habitat contributes about 50% of coho smolt production in reaches 2-4 (Decker et al. 2006). The lower reaches of several of the small natural tributaries are also used by coho for spawning. The objectives of this report are: 1) to describe the study design and sampling methodologies for each component of the CRMP, 2) to summarize all years of abundance data collected to date for each species and life stage, 3) to evaluate potential problems or shortcomings with the existing design of CRMP components, and recommend changes to be applied in future years. This report is arranged according to the four monitoring components of the CRMP: 1) adult escapement surveys for chum, pink and coho, 2) redd surveys for adult steelhead, 3) juvenile standing stock surveys for coho and steelhead, and 4) smolt outmigration trapping for all four species. Technical aspects and issues with each monitoring component are presented at the end of each section. Annual trends in abundance and survival from one life stage to the next are summarized for each species at the end of the report. The rationale for each of the four CRMP components and a description of work completed to date are provided below. 1.1 Adult salmon escapement Formal surveys of adult salmon escapement were included as a component of the Coquitlam River Monitoring Program beginning with chum and coho salmon in 2002, and pink in Live and dead Chinook are also enumerated during adult surveys, but their numbers are too low to allow for quantitative estimates. It should be noted that because adult salmon monitoring was started after smolt monitoring, estimates of smolts/spawner (or egg-to-smolt survival) for Treatment 1 will be limited to six, three and five years data for chum, pink and coho, respectively.

18 17 N R4 R3 R2B R2A R1 Figure 1.2 Map of lower Coquitlam River study area with stream reaches defined by the Coquitlam-Buntzen Water Use Plan Consultative Committee.

19 18 During , weekly total counts of live adults by shore-based observers and area-under-the-curve (AUC) methodology was used to estimate adult salmon abundance. The AUC approach requires accurate information about observer efficiency and average spawner survey life (Perrin and Irvine 1990). Values for both parameters may vary widely among stream and among years within streams (Perrin and Irvine 1990; Hilborn et al. 1990), suggesting the need to obtain stream- and year-specific estimates for each. However, to date, AUC estimates for Coquitlam River have relied on observer efficiency and survey life data from other streams. During it was not uncommon for surveys to be postponed for as long as three weeks, or for some portions of the study area to be excluded from a survey, due to poor viewing conditions. In some cases, this resulted in poorly defined run timing curves for one or more species. In this report, we present a preliminary version of a new escapement model for Coquitlam River that incorporates error in escapement estimates associated with missed and partial surveys. We outline how estimates could be further refined by collecting observer efficiency and survey life data in future years, and by further refining the model so that these error sources are incorporated more realistically. With respect to the goal of developing Coquitlam River-specific estimates of observer efficiency and residence time, the results of a mark-recapture pilot study in 2006 are also presented. Salmon escapement data and mark-recapture experiments for 2007 were not complete at the time of writing for this report, and will be presented in the 2008 annual report. 1.2 Adult steelhead escapement Assessment of adult winter steelhead escapement, in the form of redd surveys, was included as a component of the Coquitlam River Monitoring Program from 2005 to Because steelhead escapement monitoring was not included as part of the flow experiment until 2005, estimates of smolts/spawner will be available for 2007 onward, but not for the portion of the experiment. This provides only two years of data for Treatment 1. Prior to 2005, snorkeling crews conducted periodic counts of adults in some years ( ) but no attempt was made to relate these counts to actual escapement. With the exception of 1999, when redd counts were conducted in reaches 3 and 4 (see Decker and Lewis 1999), pre-2005 surveys did not include counts of steelhead redds. Because of the protracted migration and spawning period for Coquitlam River winter steelhead (4-5 months), high variation among individuals in stream residence time (Korman et al. 2002), and highly variable survey conditions within the spawning period, reliable information about residence time and observer efficiency would be needed in order to estimate escapement using counts of live steelhead and area-under-the-curve methodology (Korman et al. 2002). This was considered unfeasible given the relatively low number of adult steelhead that return to Coquitlam River and the considerable cost of collecting such information.

20 19 Alternatively, we relied on redd surveys to estimate steelhead escapement and egg deposition. Other studies have found that total redd counts are excellent predictors (i.e, R 2 values > 0.9) of steelhead escapement as estimated by direct trap counts (Freeman and Foley 1985), mark-recapture (Jacobs et al. 2002), or AUC methodology (Gallagher and Gallagher 2005). A pilot study conducted in reaches 3 and 4 in 1999 (Decker and Lewis 1999) indicated that conditions during the spring steelhead spawning period in Coquitlam River were, for the most part, well suited to redd surveys. Estimating the uncertainty in steelhead escapement estimates derived from redd counts would require the concurrent use of a second, more accurate method (e.g., resistivity counter or mark-recapture program). This is beyond the scope of the current study. Thus, estimates of steelhead estimate and egg deposition for the Coquitlam River should properly be considered indices of abundance. 1.3 Juvenile coho and steelhead standing stock In 2006 the COQWUPCC requested that a juvenile standing stock survey component be added to the Coquitlam River Monitoring Program to provide an index of annual abundance for coho and steelhead at the fry (age-0+) and parr stages (age-1+ and age-2+, steelhead only) stage. These data together with adult escapement and smolt abundance estimates, can be used to investigate freshwater production bottlenecks in Coquitlam River that may relate to specific habitat or flow issues. In early September 2006, we conducted a feasibility study to determine the best approach for collecting annual juvenile standing stock data. As part of the study, we compared two different methods of sampling fish abundance: three-pass removal electrofishing and night snorkeling counts at 20 m long enclosed sites along one shoreline. We also compared results from shoreline sites and sites that spanned the entire stream channel, using snorkeling counts only. The results suggested that sampling juvenile abundance at full channel sites using night snorkeling counts would be the most effective method for monitoring annual juvenile standing stocks in Coquitlam River (Decker et al. 2007). In 2007 this methodology was used to sample juvenile abundance, and mark-recapture experiments were conducted to correct snorkelling counts for snorkeling efficiency (the proportion of fish present that snorkelers detect). This report describes in detail the results of the 2007 juvenile standing stock survey, and summarizes preliminary population estimates for 2006 and Smolt outmigrant trapping Smolt trapping has occurred in Coquitlam River in various years since 1993 (see Decker and Lewis 2000 for a summary of earlier work). However, earlier studies were intended to compare smolt production at several constructed offchannel habitat sites to that in reach 4 of the Coquitlam River mainstem, rather than assess production in the mainstem as a whole. During , numbers of coho and steelhead smolt outmigrants were assessed for a 7.5 km long section of Coquitlam River that included reaches 3 and 4 and most of reach 2a. Smolt numbers were also assessed for individual reaches and for the four off-

21 channel sites. Chum and pink (odd years only) smolt numbers were monitored for the same section of the mainstem beginning in For the mainstem, smolt numbers were assessed using mark-recapture methodology and rotary screw traps (coho and steelhead) and incline plane traps (chum and pink). Fullspan downstream weirs were used at off-channel sites. This report describes in detail the results of the 2007 smolt trapping program and summarizes population estimates for all species and reaches for

22 ADULT SALMON ESCAPEMENT 2.1 Methods Escapement is often estimated by obtaining repeat counts of the number of fish present over the spawning migration. The proportion of fish observed (observer efficiency) and the proportion of the total run that is present must both be estimated on each survey to determine the total escapement. The total number of fish present on a survey is simply the difference between the cumulative arrivals and departures by the survey date. Departure schedule will be determined based on the arrival schedule and the length of time spawners remain in the survey area (survey life). The proportion of the run that is present on any survey date can therefore be estimated from data on at least 2 of the 3 run timing components: arrival timing; survey life; and departure timing. Analytical approaches for estimating escapement from repeat count data have advanced considerably from the original Area-Under-the-Curve (AUC) methodology (e.g. English et al. 1992). Hilborn et al. (1999) used a maximum likelihood approach to estimate escapement and arrival timing parameters by assuming that survey life was constant, and that, on average, all fish present in the survey area were counted. Korman et al. (2002) estimated escapement from repeat mark-recapture experiments in conjunction with more flexible arrival timing and survey life models. Escapement estimates will be uncertain if there are no post peak counts (Hilborn et al. 1999, Adkison and Su 2001), or if peak and post peak surveys occur during periods of low catchability (Korman et al. 2002). In these situations, the possibility of a large number of fish entering at the peak or late in the run cannot be discounted in the estimation process because there is little information about arrival timing in the repeat count data. This problem becomes more severe with increases in the flexibility of arrival timing and survey life model structure. Su et al. (2001) used a hierarchical Bayesian model to borrow information on arrival timing from years when it was better defined to improve estimates of escapement and timing in years when few or no counts were made after the peak of escapement. The approach has utility for escapement programs with many years of data and where timing is well defined in at least some years, but, given the limited number of years of survey data at present, is less useful for Coquitlam River Stratified index survey design Returning spawners to the Coquitlam River have been enumerated by stream walk surveys conducted on an annual basis since 2002 for chum and coho, and during odds years starting in 2003 for pink. For adult salmon, the study area extends downstream from Coquitlam Dam to the downstream boundary of reach 1 at the Maple Creek confluence, encompassing reaches 1-4 in their entirety (Figure 1.2). Reach 0 (Fraser River confluence to Maple Creek) was excluded as it contains little spawning habitat and because fish entering the

23 22 Hoy/Scott Creek system often hold in this reach and could be confused with fish destined for upper reaches in Coquitlam River. Due to the length of the study area (approximately 12.8 km), and the concentration of spawning activity within specific sections, sampling efficiency was improved by stratifying the survey to focus on five key areas hereafter referred to as index sites A-E (Figures 2.1, 2.2). Irvine et al. (1992) demonstrated that using a stratified index design to select areas to conduct visual surveys for adult coho provided accurate estimates of escapement at a lower cost than more intensive methods such as mark-recapture or operation of counting fences. Coquitlam River index sites were originally developed from spawning distribution maps developed as part of the LB1 WUP. The boundaries of these sites were then refined during , as better information about spawning distributions was obtained. The five index sites have a collective length of approximately 9 km, or 63% of the length of the study area for adult salmon. However, they account for a higher percentage of the total fish present during any one survey because they encompass the majority of available spawning habitat. Included as part of the index sites were mainstem areas, natural side-channels and braids, and constructed off-channel habitat. To account for spawners that were present in the study area, but not at one of the five index sites, on several occasions each year, the survey was extended to all wetted habitats (mainstem areas, side-channels, braids, and constructed off-channel habitat) in the entire study area. Data from complete surveys were used to compute expansion factors that were applied to counts for surveys that included only the index sites; for each complete survey, the expansion factor was computed as the total count for the entire study area divided by the aggregate count for the five index sites. The goal was to complete three full surveys of the study area for each species (with some dates providing full surveys for more than one species). The surveys were scheduled to capture early, peak, and late portions of the spawning period for each species, to account for possible temporal variation in the proportion of spawners in non-index sites. Considerable overlap exists for the spawning periods of chum (mid October early December), pink (mid September late October) and coho (mid-october mid January) in Coquitlam River. To address this, we conducted concurrent counts for whichever species were present during a particular survey. To capture the entire spawning period for each species, we conducted weekly surveys that began roughly in the third week of September and ended in the second week of January. The date of the first survey each year was adjusted in an attempt to capture the beginning of the spawning period for pink, and the date of the last survey was adjusted to capture the end of the spawning period for coho.

24 23 Site C. N Site C Site B. Site B Temperature Monitoring Site Restoration Site Site A 0m 250m 500m 750m Scale is approximate Figure 2.1 Map showing adult spawning index sites A-C in the lower portion of Coquitlam River study area (reaches 1, 2a).

25 24 N Site E Site D. Site D Adult Tagging Site Temperature Monitoring Site Restoration Site 0m 250m 500m 750m Scale is approximate Figure 2.2 Map showing adult spawning index sites D and E, in the upper portion of Coquitlam River study area (reaches 2b, 3 and 4).

26 25 Spawner surveys were performed by a crew of two people walking in an upstream direction, with one person on either side of the river. The survey crew minimized the likelihood of making duplicate counts by regularly discussing which portions of the river channel each person was responsible for. Surveyors carried walking staffs that they used to probe under cutbanks and LWD accumulations in order to detect fish that were not in plain view. Total numbers of live and dead adults were recorded during each survey, but only data for live fish were used to estimate escapement. In most cases, a survey of the five index sites was completed over two days Observer efficiency, survey life and incomplete surveys To obtain preliminary estimates of adult salmon escapements for we used methods of estimating or approximating observer efficiency that differed from those specified in the terms of reference for the Coquitlam River monitoring program (Leake 2005). The terms of reference specified that average observer efficiency would be estimated for each species by having the survey crew periodically conduct counts of the same stream sections independent of one another, the difference in counts providing an estimate of the error in observer efficiency. While this approach does provide estimates of a portion of the total error resulting from inter-observer variation, total error in observer efficiency is likely to be underestimated because other sources of error in observer efficiency, particularly variation in streamflow and water clarity, and variation in fish behaviour (e.g, staging versus spawning), are not accounted for. To reliably estimate observer efficiency requires some means of independently estimating the proportion of fish actually seen by observers. Mark-recapture experiments are commonly used for this purpose. In 2006, the feasibility of using mark-recapture methodology to obtain estimates of observer efficiency and survey life specific to Coquitlam River was investigated. On several occasions during the peak of chum spawning period, we attempted to capture and tag chum at several capture locations using two different capture methods, beach seining and a portable upstream fyke net. Standard Petersen disc tags were used to tag fish, with different colours used to distinguish temporal mark groups. To estimate observer efficiency, we attempted to conduct a complete survey of the study area within two days of a tagging event so that the number of tags lost to mortality would be minimized. For each mark group, we attempted to estimate survey life based on the decline in number of tagged live adults seen over each successive survey following the tagging event, but too few surveys were completed subsequent to tagging events in 2006 to provide meaningful estimates of survey life. Frequent high flow events and associated high turbidity during the fall and winter spawning period also contribute to the uncertainty of salmon escapement

27 26 estimates in Coquitlam River. During it was not uncommon for surveys to be postponed for as long as three weeks, or for some portions of the study area to be excluded from a survey, due to poor viewing conditions. In some cases, this resulted in poorly defined run timing curves for one or more species. The CRMP Terms of Reference and previous analyses of spawner survey data for Coquitlam River (Macnair 2003, 2004, 2005, 2006) do not explicitly consider the effect of missed or partial surveys on escapement estimates. In this report, we present a preliminary version of a new escapement model for Coquitlam River that incorporates error in escapement estimates associated with missed and partial surveys. The model was applied to all years of available data for each species Escapement model structure and parameter estimation The escapement model consists of two main elements: i) a simple process model predicts the number of fish present on each day of the run and the departure schedule based on the total escapement and parametric relationships simulating arrival timing and survey life, and ii) an observation model simulates the number of fish counted at index (sites A-E) and non-index (site=ni) sites based on the predicted numbers present, observer efficiency, and the proportion of the population that is present at each site (hereafter referred to as siteproportions). The process model is very close to the one developed in Korman et al. (2002) and differs only in the form of the survey life relationship. We developed two versions of the escapement model (referred to as the Poisson and negative binomial models) to estimate process and observation parameters. The Poisson model minimizes a negative log-likelihood value by comparing predicted and observed counts at each site, whereas the negative binomial model minimizes a negative binomial function. The Poisson model relies on the assumption that the ratio of the variance of the predicted mean count to the predicted mean count is one, or, in other words, the data are not over-dispersed. The negative binomial model makes no such assumption, and is better suited to data that are over-dispersed Process Model The proportion of the total escapement entering the survey area on day t (PA t ) of the simulated run is predicted by a beta distribution, τ 1 β 1 PA t = φt (1 φt ) (2.1) where, τ and β are parameters of the beta distribution and φ t represents the proportional day of the run for day i, ranging from 0 to 1, on the assumed first (t=1) and last (t=t) day, respectively. τ is the precision of run timing with smaller values representing a low and constant rate of arrival over the duration of the run, and larger values representing a shorter and more concentrated arrival

28 27 timing. The beta distribution is reparameterized so that β is calculated based on estimates of the day when the peak arrival rate occurs (μ) and the precision of arrival timing, using the transformation: τ 1 β = + 2 τ μ T (2.2) For chum, pink and coho, survey life, the number of days a fish spends in the survey area, is normally longer for fish that arrive earlier in the spawning period (Perrin and Irvine 1990; Su et al. 2001). We predicted survey life using a decaying exponential relationship, SL t = (2.3) st λce λ where SL t is the mean survey life for a fish entering on day t, λ c is the maximum mean survey life, and λ s is the slope of the relationship. The mean departure day for a fish arriving on day t is simply d t = t + SLt. This equation differs from the one recommended in the terms of reference for the Coquitlam River Monitoring Program (Leake 2005, p ), which was developed for winter-run steelhead (Korman et al. 2002). Equation 2.3 is derived from an equation used to estimate survey life for pink salmon (Su et al. 2001) and is more appropriate for fall spawning salmon species, which have a more truncated spawning period as compared to steelhead. The proportion of fish that arrive on day t and depart on day tt is predicted from a normal distribution with mean d t and standard deviation σ t, PAD t,tt ~ Normal(tt, d t, σ t ) (2.4) We assume that error around the date of entry-mean survey life relationship is lognormally distributed, thus σ t =λ v *SL t, where λ v is the coefficient of variation in mean survey life. PAD values are standardized so that proportions across all departure days for each arrival day sum to 1, that is, all fish must exit the survey area by the assumed last day of the run. As a fish can obviously not depart before it arrives, PAD t,tt = 0 for tt < t. The proportion of fish departing on each day (PD t ) is computed from, = * (2.5) PD t PAt PADt, tt t

29 28 Note that departure timing depends on both arrival timing and the survey life relationship that defines PAD. Finally, the number of fish present in the survey area on each day (U t ) is the product of the total escapement (E) and the difference between the cumulative arrivals and departures on that day, U t t t = E PA 1 1 PD (2.6) The difference between the cumulative values of PA and PD on any date represents the proportion of the total run that is present Observation model and estimation of model parameters Escapement (E) and arrival timing parameters (μ, τ), and those defining the proportion of the total population that is present at site= i for a survey conducted at time= t (γ t,i ) are jointly estimated by assuming that the count data arise from either Poisson or negative binomial distributions. In this analysis we assume that the site-proportions (γ t,i ) do not vary across surveys, so γ t,i reduces to γ i. The Poisson model, u t, i ~ t, i i i Poisson( η γ U ) (2.7) determines the probability of counting u t,i fish for a survey at time= t at site= i given a prediction of the number that would be counted. The prediction is simply the product of the predicted proportion present at the site (γ i ), the observer efficiency (η i,t ), and the predicted total number of fish present in the river (U i ). The negative binomial model, u NegBin( η γ U, ) (2.8) t, i ~ t, i i i k uses the same observations and predictions as the Poisson model, but requires the estimation of an over-dispersion parameter k. The over-dispersion parameter can be expressed in terms of the variance (σ 2 ) associated with a predicted mean count (m= η i,t γ t U i ) and the predicted mean count (Hilborn and Mangel 1997), k 2 m σ m = 2 (2.9)

30 29 The variance must exceed the mean for a negative binomial distribution to apply so k must be positive. A very small k value indicates that the variance greatly exceeds the mean (i.e., a large variance-to-mean ratio) resulting in a distribution with a long right-hand tail. Also note that as σ 2 m, k and the negative binomial distribution is approximated by the Poisson. In some cases, estimates of k can be very small and result in unrealistic escapement estimates. A penalty function is used to constrain k by penalizing very large variance-to-mean ratios. Variance-to-mean rations are computed from the predicted counts and the estimate of k from equation 2.9. We then assume that the log of the variance-to-mean ratios are normally distributed with mean, μ v and standard deviation σ v, 2 σ i, t Log( γ η U i i, t i ) ~ N( μ, σ ) v v (2.10) A large value for σ v (e.g. σ v =10) results in very little penalty on the varianceto-mean ratios. In this situation the over-dispersion parameter is determined only by the data. As σ v is reduced, there are higher penalties for larger variance-tomean ratios. We assume that μ v =0, that is, the expected variance-to-mean ratio, is 1. Thus, as σ v is reduced, the negative binomial becomes increasingly like a Poisson model and estimates of k will increase. Most likely parameter estimates (MLEs) were computed using the AD model builder software (Otter Research 2004). The sum of the predicted proportions ata-site (γ i ) must be one as we assume that all areas that can potentially hold fish in the survey area are surveyed. We use the penalty function, i 2 pen = 1000 * γ i 1 (2.11) i to constrain γ = 1. The log-likelihoods returned from the Poisson i (equation 2.8 with k fixed to a large value) or negative binomial (k estimated) models are summed across all survey periods and sites where data is available, and added to the variance-to-mean and γ i =1 penalties. Given the limited amount of mark-recapture data available for chum in Coquitlam River, observer efficiency for chum was assumed to be constant and approximated as the average across all estimates obtained in Observer efficiency for pink and coho in Coquitlam River was assumed to be similar to that for chum. Survey life parameters (λ c, λ s, λ v ) were fixed and approximated by manually adjusting the values so that survey life at the peak of spawning was similar to average survey life values for chum pink and coho in other streams (chum: SL peak = 11.9 days, λ c =25, λ s =0.02, λ v =0.65; pink: SL peak = 16.5 days, λ c =40, λ s =0.03, λ v =0.65; coho:

31 30 SL peak = 11.4 days, λ c =34, λ s =0.013, λ v =0.65; Perrin and Irvine 1990; Su et al. 2001). For each species, the coefficient of variation for survey life (λ v ) was also based on results from other studies. The standard deviation of the variance-tomean ratio penalty was set to a large value (σ v =100) unless otherwise specified. Two alternate hierarchies were used to estimate multi-year time series of escapements. In the first case, all model parameters were estimated independently for each year. In this analysis, we examined two years of data for pink salmon (2003, 2005), and five years for chum and coho ( ). For the year-independent model, 10 parameters must be estimated each year. A total of 20, 50, and 50 parameters must therefore be estimated across all years of available data for pink, chum, and coho salmon, respectively. If a Poisson model is used in the estimation, k is fixed at a large value, and the number of parameters estimated for each year is reduced by one. In the second case, siteproportion and over-dispersion parameters were jointly estimated across years (γ yr,i and k yr reduce to γ i and k), while escapement and arrival timing parameters were not. In this case 13, 22, and 22 parameters must be estimated for pink, chum, and coho salmon, respectively (or 12, 21, and 21 if the Poisson model is used). The potential advantage of estimating site-proportion jointly across years is that there is that the relative distribution of spawners among sites can be estimated using all available data. However, if site-proportion varies significant among years, model fit will be poorer than if site-proportion was estimated for each year independently. We used the second order Akaikie Information Criterion (AIC c ) to compare the out-of-sample predicted power of year-independent and joint-year escapement models. AIC c scores were computed as the sum of twice the deviance (D) at the maximum likelihood estimate (D= -2 * negative log likelihood) and the number of parameters adjusted by a correction factor that accounts for small sample size (Burnham and Anderson 2002). The AIC c criterion describes the trade-off between model fit (D) and complexity (the number of parameters). Models with the lowest scores are those with the fewest number of parameters that still provide an adequate representation of the data. Models with too few parameters will provide poorer fits to the data because they will fail to recognize important structure. Models with too many parameters may fit the data better, but will have variances that are needlessly large and may identify spurious structures. The difference between the lowest AIC c score across a set of candidate models and the score for each model, referred to as Δ AIC, measures the extent of support for each model given the data. We use the Δ AIC guidelines for model support provided by Burnham and Anderson (2002): Δ AIC <2 (substantial/strong support); 4<Δ AIC <7 (weak support); Δ AIC >10 (essentially no support). Sample size was computed as the sum of the number of sites where fish were enumerated over the survey season for each species.

32 31 For the Poisson and the negative binomial versions of the escapement model, model fit associated with the most likely estimates was compared using a variety of diagnostics. Predictions of the numbers present for each day (U t ) of the run were compared to the expanded survey-specific counts ( Uˆ t ). The expanded counts were computed from, u i t i Uˆ t = for all sites visited on survey t. (2.12) γ i t, i η, i For the Poisson model, Pearson standardized residuals of site- and surveyspecific predictions were computed from, r t, i u η γ U t, i t, i i i = (2.13) σ t, i where, σ t,i is the standard deviation of the predicted count, which is the estimated mean count (η i,t γ i U t ). In the case of the negative binomial model, Pearson standardized residuals were computed from the mean and the over-dispersion term using equation 2.9. A negative residual indicates that the observed count is greater than predicted one. Residuals were examined graphically to reveal potential biases. We relied on residual statistics to determine which error model, the Poisson or the negative binomial, was the most appropriate for the data at hand. As the standardized Pearson residuals should be normally distributed, their standard deviation should be one (SDNR=1) and the median of their absolute values should be 0.67 (MAR=0.67). Residual statistics substantially above these values indicate that there is more variance than expected by the assumed error distribution. High residual statistics values could be caused by inadequacies in the process model (e.g. arrival timing not perfectly described by a beta distribution), or an inappropriate error assumption (e.g., the error is not Poisson distributed). As our likelihood assumes no process error (i.e, it assumes a smooth, unimodal run curve), we interpreted deviations from the expected residual statistics as an indication that the error model was incorrect. Asymptotic estimates of the standard deviation for the MLEs were computed using the Delta method. Joint posterior distributions of model parameters were estimated using Markov Chain Monte Carlo simulation (MCMC). We used uniform priors for all model parameters except k and selected bounds that were wide enough to have no effect on the estimation of the joint posteriors. Posterior distributions of parameters were estimated by taking every 25 th MCMC simulation from a total 25,000 trials.

33 Results and Discussion Unadjusted survey counts from all surveys in all years are shown for chum, pink and coho in Appendices The period of peak spawning varied from the first week of October for pink, to the last week of October for chum, to the second week of December for coho. Pink salmon were already present in the spawning area at the time of the first survey in both 2003 and As well, in both years, data for pink were sparse in the later half of the spawning period because of surveys on account of high caused flows. For chum, the beginning and end of the spawning period each year were well defined. The beginning of the spawning period was well defined for coho, but in the majority of years, some coho were still present during the final survey. For modeling purposes, the maximum length of the spawning period for chum, pink, and coho, was constrained to 113 days (September 9 January 2), 75 days (September 5 November 19), and 130 days (September 19 January 27), respectively Observer efficiency and survey life In 2006, chum salmon were captured and tagged on seven days during the spawning period (Table 2.1). Due to low number of fish captured on individual days, tagged fish from October 24 and October were treated as one tag group. There were then four distinct tag groups in total, all representing the peak or pre-peak portion of the spawning period. Although no sex ratio data were collected, tagging was likely biased towards males, as the percentage of females in the tagged populations was only 18% to 36%. Table 2.1 Results of the 2006 mark-recapture study to estimate observer efficiency and survey life for chum salmon in Coquitlam River. Tagging Tag Tagging Recovery Duration Marks Recovered % location group date date (days) (M ) (R ) R/M females d/s of site A 1 Oct 17 Oct % 18% d/s of site A 1 Oct 17 Oct 31-Nov % 18% site C 2 Oct 19 Oct % 30% site C 2 Oct 19 Oct 31-Nov % 30% site C 3 Oct 20 Oct % 36% site C 3 Oct 20 Oct 31-Nov % 36% site C 4 Oct 24, Oct 31-Nov % 44% site C 4 Oct 24, Oct 31-Nov % 44% In 2006, we were unable to locate an effective tagging site for chum near the downstream boundary of the study area. With the exception of the October 17 tag group, which were captured at downstream of the study area at the Scott/Hoy Creek confluence, all tag groups were captured at a site located at about the midpoint of the study area, 8 km upstream of the downstream boundary. We did not attempt to estimate survey life for chum because, given that the tagging location was in the middle of spawning area, the number of days tagged fish

34 33 were present in the study area prior to tagging was not known, but may have been considerable, and estimates of survey life would have been biased low as a result. Alternatively, survey life was estimated for chum, pink and coho using the date of entry-mean survey life model described in equations 2.3 and 2.4, with parameter values derived from other studies (Figure 2.3; Perrin and Irvine 1990; Su et al. 2001). To estimate observer efficiency for each chum tag group, we attempted to conduct a complete survey of the study area within two days of tagging, so that the number of tags lost to mortality would be nominal. However, this was only possible for the October 19 (47 adults) and October 20 (42 adults) tag groups because, for other tag groups, stream visibility deteriorated rapidly after tagging occurred. A complete survey of the study area was completed on October Thirty-two and 29 tags were resighted for the October 19 and October 20 tag groups, respectively, providing two similar estimates of observer efficiency for chum (68% and 69%, respectively). Based on these results, we assumed an observer efficiency value of for chum, pink and coho for all surveys in all years Survey life (days) Coho Pink Chum Day of run when fish enters study area Figure 2.3 Predicted relationship between survey life and day of arrival in the study area for chum, pink and coho salmon in Coquitlam River Escapement Model For the Poisson escapement model, the information theoretic comparison of year-independent and joint-year models for estimated site-proportions clearly

35 34 demonstrated that year-independent models had higher out-of-sample predictive power. There was enough inter annual variation in the site-proportion estimates (Table 2.2) that the poorer fits of the joint models was not compensated by the reduction in parameters (Table 2.3). ΔAIC c values were in the hundreds and indicated no support for jointly estimated models (Table 2.3). The yearindependent models provided a significant improvement in fit (P values <0.001) relative to the joint-year models. The negative binomial model yielded similar results with respect to the relative performance of the year-independent and jointyear methods of estimating site-proportions (results not shown). Given the poor fit to the data for the joint-year method of estimating site-proportions, we focused only on the year-independent method with respect to escapement estimation. Table 2.2 Estimated proportion of chum, pink and coho spawning populations present at each index site (A-E) and at non-index (NI) sites (γ i ) based on year-independent and jointyear estimation. Species Site Joint Chum A B C D E NI Pink A B C D E NI Coho A B C D E NI

36 35 Table 2.3 Comparison of model fit based on Akaikie s second order Information Criteria (AIC c ) for two alternate approaches (year-independent and joint-year) of estimating the distribution of the chum, pink and coho salmon spawning population among survey sites in Coquitlam River. Results are shown for the Poisson escapement model only. Species Test statistic Independent Joint Pink Deviance (D) Parameters (k) AIC c ΔAIC 376 Chum Deviance (D) Parameters (k) AIC c ΔAIC 4087 Coho Deviance (D) Parameters (k) AIC c ΔAIC 715 The Poisson escapement model provided reasonably good fits of predicted to observed weekly counts of chum, pink and coho (observed counts being expanded for observer efficiency and unsurveyed sites). As examples, we show fits of the year-independent model to the 2005 data for each species (Figures ). Although the data showed some evidence of changes in site-proportions over the course of the spawning period within individual years, the model, which assumes no variation over surveys within years, generally provided a good fit. Site-proportion estimates are more heavily influenced by large counts. There is little penalty to the likelihood incurred by incorrectly predicting the site-proportion when there are few fish counted. As a result, the model provided good fits to the site-proportions for critical surveys that had high numbers of fish, and, correspondingly, had a large influence on escapement estimates. Despite the reasonably good fits of predicted to observed weekly counts, an analysis of standardized Pearson residuals indicated that there was more variance then expected if error was Poisson distributed (i.e., the data were overdispersed). For the Poisson model, SDNR and MAR values exceeded 1.0 and 0.67 by a substantial margin (Table 2.4), indicating poor model fit. By contrast, residual statistics for the negative binomial model were near the expected values of 1.0 for SDNR and 0.67 for MAR, indicating that the negative binomial error distribution was appropriate for the data (Table 2.4). Assuming a negative binomial model error distribution as opposed to a Poisson error distribution is essentially recognizing greater uncertainty in the survey data.

37 36 Proportion of Run a Arrivals Departures 0.00 Oct Nov Dec Number adults present b Predicted Expanded Count Oct Nov Dec NI c E 0.05 Site D 0.19 C 0.19 B 0.06 A Sep 12-Oct 1-Nov 16-Nov 4-Dec 17-Dec 28-Dec Figure 2.4 Predicted 2005 chum salmon arrival and departure timing (A), predicted numbers present and observer counts expanded by observer efficiency (B), and the estimated proportion of the population that was surveyed, and observed (blue) and estimated (red and black) proportions of the population in index sites based on yearindependent estimation (C). The rows at the top of graph C are, from top to bottom, raw survey counts, number of sites surveyed, and the estimate proportion of the total population that was sampled. The column of numbers at the far right shows the modelestimated proportion assigned to each site, which is constant across surveys within each year. The size of the solid blue circle represents the proportion of the observed total count at each site for each survey, with no correction for unsampled sites. The open red circles are the model predicted proportion, adjusted so that they sum to one over the sites that were sampled on each survey (to facilitate comparison with observed counts when the number of sites sampled is less than 6). The black open circles are the estimated proportions that were used in the escapement calculations, they are the same for each survey. The unadjusted and adjusted model proportions are equivalent on surveys where all 6 sites were enumerated, and for these surveys, only the black circles are visible.

38 37 Proportion of Run a Arrivals Departures Sep 14 Oct 04 Oct 24 Nov 13 Number adults present b Predicted Expanded Count Sep 14 Oct 04 Oct 24 Nov 13 NI c Site E D C B A Sep 5-Oct 12-Oct 24-Oct 1-Nov 9-Nov 16-Nov Figure 2.5 Predicted 2005 pink salmon arrival and departure timing (A), predicted numbers present and observer counts expanded by observer efficiency (B), and the estimated proportion of the population that was surveyed, and observed (blue) and estimated (red and black) proportions of the population in index sites based on year-independent estimation (C). See Figure 2.4 caption for details regarding graph C.

39 38 Proportion of Run a Arrivals Departures Sep Nov Jan Number adults present b Predicted Expanded Count Sep Nov Jan NI c Site E D C B A Sep 12-Oct 01-Nov 16-Nov 04-Dec 17-Dec 28-Dec Figure 2.6 Predicted 2005 coho salmon arrival and departure timing (A), predicted numbers present and observer counts expanded by observer efficiency (B), and the estimated proportion of the population that was surveyed, and observed (blue) and estimated (red and black) proportions of the population in index sites based on yearindependent estimation (C). See Figure 2.4 caption for details regarding graph C.

40 39 Table 2.4 Escapement estimates for chum, pink and coho salmon generated using the Poisson and negative binomial versions of the year-independent escapement model. Also shown are 95% confidence intervals, spawner densities, and standardized Pearson residuals (SDNR, the standard deviation of normalized residuals; and MAR, the median of their absolute values), which describe model fit. Uncertainty in escapements is shown as 95% confidence intervals and percent relative error (average 95% confidence interval divided by the estimate). Species Model Parameter Chum Poisson Escapement 13,262 16,711 22,061 16,576 36,828 ± 95% CI , ,766 ± CI (%) 4.8% 4.7% 4.6% 4.9% 4.8% Adults/km 1,036 1,306 1,724 1,295 2,877 SDNR MAR Negative Escapement 13,938 19,772 20,868 21,717 65,685 binomial 95% CI 3,877 9,740 4,031 9,988 29,878 ± CI (%) 27.8% 49.3% 19.3% 46.0% 45.5% Adults/km 1,089 1,545 1,630 1,697 5,132 SDNR MAR Pink Poisson Escapement - 3,817-2,349-95% CI ± CI (%) - 5.0% - 6.2% - Adults/km SDNR MAR Negative Escapement - 2,686-1,814 - binomial 95% CI - 1, ± CI (%) % % - Adults/km SDNR MAR Coho Poisson Escapement 1, , % CI ± CI (%) 6.8% 8.2% 8.1% 9.9% 12.1% Adults/km SDNR MAR Negative Escapement 2,390 1,466 1, binomial 95% CI 2,465 1, ± CI (%) 103.1% 91.4% 49.7% 39.5% 41.6% Adults/km SDNR MAR The negative binomial model produced parameter estimates that were confounded in some cases, but overall the degree of confounding was not excessive. We show the correlations among values from the joint posterior distributions for coho data from 2005 as a typical example (Figure 2.9).

41 40 Escapement estimates increased with the mode of arrival timing (see plot for esc versus μ in Figure 2.9). Because coho are the last of the three species to spawn, there were few surveys conducted at the very end of the coho run, so there is less information on run-timing compared to that for the other species. Later run timing for coho results in higher probabilities of large runs, regardless of the actual count data. The effect of this confounding is not severe, even in the case of coho, because the model adjusts run timing so that no fish are present by the terminal day, T, which was generally chosen to occur shortly after the last survey. The assumption that the run is complete by day T helps reduce uncertainty, but there is probably substantial uncertainty about this assumption, at least for coho. A trade-off between the mode and precision of arrival timing is also apparent in the data (see μ vs. τ plot in Fig. 4). In order to explain the data, the run must arrive within a relatively narrow window when arrival timing is relatively late, or visa-versa. In the majority of cases, the negative binomial model provided similar or better fits of predicted to observed counts compared to the Poisson model (e.g., Figure 2.7), but there were several exceptions to this, notably pink in 2003 and 2005, coho in 2002 and 2004, and chum in 2006 (Figure 2.8). In the case of pink, poor fits were likely the result of uncertainty as to when the run began, and Number adults present a Chum 2002 Oct Nov Dec Number adults present a Coho 2005 Sep Nov Jan Figure 2.7 Examples of relatively good fit of predicted numbers of adult salmon present to observer counts expanded by observer efficiency for the negative binomial escapement model.

42 41 variation in within-year site-proportions. Pink were already present in the study area at the time of the first survey in 2003 and With respect to siteproportions, as the pink run progressed, the distribution of spawners shifted from downstream to upstream areas (Figure 2.5). Variation in site-proportions likely also contributed to the poor fit for the coho data. In 2002 and 2004 coho tended to show limited use of lower sites (A and B) and the most upstream (E) site early in the spawning period, but the most upstream site became increasingly used during later surveys (Figure 2.6). In 2006, high flows precluded mainstem surveys for over four weeks, (October 31-November 30). Peak spawning for chum salmon likely occurred at the beginning of this period (Appendix 2.1). The lack of survey data following a expanded peak count of 16,688 chum on October 31 led the negative binomial model to assign a low likelihood to this count (despite the fact that all index and non-index were surveyed under excellent conditions), resulting in a predicted count of 33,098 chum for October 31 (Figure 2.8), which was likely biased high. In cases where the negative binomial model provided a poor fit to the data, the two most likely causes were inadequacies in the process model, which assumed no error (i.e., a smooth, unimodal run curve) or the site-proportions model, which assumed constant site-proportions within years. Additional parameters should be added to future models to allow for more flexibility in the run timing curve (e.g., multiple modes). With more years data, it may also be possible to model within-year trends in site-proportions (e.g., pink and coho shifting to upper sites later in the spawning period). A summary of most likely escapement estimates and run-timing parameters for all species and years for the Poisson and negative binomial models is provided in Table 2.4. Chum, pink and coho escapement estimates ranged from 13,000-66,000, 1,800-3,800, and 500-2,400 across years and models, respectively. The difference in escapement estimates for the two models ranged from 5% to 78% and averaged 29%, with no consistent trend of higher estimates for one model versus the other. Uncertainty in escapement estimates was unrealistically low for the Poisson model (95% confidence intervals: ±5% to 12% of the estimate, Table 2.4) because error in count data exceeded that assumed by the model (i.e., variance to mean ratio > 1). Uncertainty in escapement estimates was substantially higher for the negative binomial model (±19% to 103% of the estimate, Table 2.4). However, even the negative binomial model likely underestimated uncertainty because it assumed no uncertainty in observer efficiency and the final day of the run (T), and relied on data from other studies to estimate variance in survey life. Escapement estimates are highly sensitive to these assumptions. For example, we recomputed the 2005 coho escapement estimate at an observer efficiency (η) of 0.4 as opposed to The escapement at η=0.4 was 1,400, 2-fold higher than the estimate for η=0.68. Escapement increased from 747 to 1,200 when the survey life constant (λ c ) was changed from 34 to 25 days. Observer efficiency and survey life data is as important to obtaining reliable estimates of escapement as the weekly count

43 42 Number adults present a Pink 2003 Sep 14 Oct 04 Oct 24 Nov 13 Number adults present a Pink 2005 Sep 14 Oct 04 Oct 24 Nov 13 Number adults present a Coho 2002 Sep Nov Jan Number adults present a Coho 2004 Sep Nov Jan Number adults present a Chum 2006 Oct Nov Dec Figure 2.8 Cases where the negative binomial escapement model produced relatively poor fits of predicted numbers of adult salmon present to observer counts expanded by observer efficiency.

44 esc μ τ a b c d e Coho 2005 ni nb_k Figure 2.9 Posterior distributions (diagonal) and correlations of escapement (esc), run timing (μ and τ) and proportion of the population at index sites A-E and non-index (ni) sites. Results are based on 500 samples drawn from 25,000 MCMC simulations based on the year-independent version of the negative binomial model applied to the 2005 coho data. The mean escapement estimate from the posterior was 747 with a 95% confidence intervals of 295. The most likely estimate (Table 2.4) was 747 adults with a 95% confidence intervals of ±295 adults.

45 44 data, and should be made a priority in future years. For the purposes of this report, we used escapement estimates derived from the negative binomial model as preliminary values for Table 6.1 in the summary section. However, in the case of chum in 2006, pink in 2003 and 2005, and coho in 2002 and 2004, we used estimates derived from the Poisson model (shaded values in Table 2.4) on account of the previously discussed poor fit of the negative binomial model in these cases Adult habitat distribution Chum salmon in particular show a preference for mainstem spawning habitat in the Coquitlam River. This preference has been noted in many studies for chum salmon in medium-sized rivers (Salo, 1991). In addition, adult chum show a preference for spawning in the lower reaches of the Coquitlam River, (Table 2.5 shows an average of 60% of adult chum spawning in index sites A-C during ). This may be an adaptation in species with a brief freshwater residency period. The short distance to estuarine habitat and the tendency of chum smolts to migrate at night may combine to reduce outmigration mortality. (Salo 1991). Spawning gravels are also more abundant in the lower reaches of Coquitlam River. Table 2.5 Adult spawning distribution by habitat type. Proportions shown are calculated based on counts of actively spawning fish only, during surveys when all five index sites were completed. M/S = mainstem, NOC = natural off-channel, OCR = off-channel restoration site, and OCC = off-channel sites combined. Species Habitat Average Chum M/S NOC OCR OCC Pink M/S NOC OCR OCC Coho M/S NOC OCR OCC Chinook M/S NOC OCR OCC

46 45 Pink salmon also have a brief freshwater residency period, but unlike chum, pink spawners made greater use of spawning areas in upper reaches of Coquitlam River (Table 2.5). Habitat segregation may be an adaptation to interspecies competition for spawning habitat that acts to reduce incidence of redd superimposition (Fukishima 1998). Pink salmon also made much greater use of off-channel sites than did chum (45% in 2003 and 35% in 2005; Table 2.5). Coho salmon showed a preference for off-channel spawning areas in Coquitlam River, but the proportion of spawners using off-channel sites declined during (Table 2.5). Changes in spawning habitat use beginning in 2005 (coinciding with dam construction and dewatering of the Grant s Tomb offchannel site) coincided with a shift in spawning away from off-channel sites, to greater use of the mainstem. In 2005 and 2006, only 51% and 20% of coho were observed spawning in off-channel sites (natural and constructed), respectively, compared to 68% and 73% in 2003 and 2004, respectively. Lower use of offchannel sites by coho in 2005 coincided with low flows during the peak of spawning (<2.0 cms at Port Coquitlam during December 1-19), resulting in poor access, but use of off-channel sites was lower still in 2006, despite relatively high flows in December (4-35 cms). Among-year differences in the proportion of coho spawning in off-channel habitat is probably best explained by the operational closure of the Grant s Tomb off-channel site in 2005 and 2006; this site was used extensively in previous years (and again in 2007, when it was reopened; J. Macnair, unpublished data). Poor access to off-channel sites may have lead to spawning in areas of lower suitability in the mainstem. Spawner access to offchannel sites will likely improve with higher flows under Treatment 2. Evidence of impeded migration for spawning adults was not apparent at any time during the surveys during Treatment 1. Fish arriving during the late summer low flow period (which in 2006 lasted until the middle of October), were observed at all index sites. However, observations by the survey crew suggest that low flows may limit access to restoration sites and natural offchannel habitat in some instances. During a low flow period in 2005, pink did not enter off-channel sites until October 2, and in 2006, chum avoided off-channel sites until October Temperature Optimal temperatures range from C for coho, to C for chum, to C for pink (Jeffres 2003, and Pauley 1988). Stream temperatures in Coquitlam River were within the optimal range for chum and coho during their mid October late November and November January spawning periods, respectively (Appendix 2.4). The same was true for pink during their late September late October spawning period, with the exception of reach 4, where daily temperatures averaged about 16 C in late September. Optimal temperatures during the incubation stage range from between 4.4 and 14 C for all species (Koski 1975).

47 Summary and recommendations for 2007 adult salmon survey Our analysis of salmon escapement data for Coquitlam River differed from that recommended in the CRMP Terms of Reference in several aspects: 1. Whereas, the CRMP-TOR recommended that observer efficiency being approximated based on the professional judgement of the survey crews, we relied on preliminary results of a mark-recapture study conducted in Coquitlam River in The equation we used to estimate survey life relative for fish entering the survey area at specific points in the survey period was derived from a model for pink salmon (Su et al. 2001) rather than the Korman et al. (2002) survey life model for winter steelhead recommended by the CRMP-TOR. 3. Our escapement model incorporated error in escapement estimates associated with missed and partial surveys, whereas the CRMP-TOR and previous analyses of spawner survey data for Coquitlam River did not explicitly consider the effect of missed or partial surveys on escapement estimates 4. Whereas the CRMP-TOR suggests methods for obtaining estimates of uncertainty for parameter estimates and for generating 95% confidence intervals for escapement estimates, given the lack of Coquitlam Riverspecific observer efficiency and survey life data, we did not attempt to quantify the uncertainty in our estimates. However, it may be possible to do this retroactively as observer efficiency and survey life data are collected in future years. We recommend that the CRMP-TOR be updated to incorporate the additions/modifications described above. We also recommend the following: 5. Reconnaissance surveys should be conducted prior to the beginning of the pink run and following the completion of the coho run in order to confirm the absence of these species from the study area. This is needed to reduce uncertainty in the arrival and departure timing models. Pink access assessments are normally conducted during the low-flow period in late August early September period as a separate requirement of the Coquitlam-Buntzen WUP, but could also serve as reconnaissance surveys to determine the start date of the pink run. 6. The mark-recapture study should be expanded in 2007 and both chum and pink salmon should be included. A suitable tagging location is required near the downstream boundary of the study area. As many tagging events as possible should be conducted, and these should ideally be distributed throughout the spawning period for each species. Every effort should be made to conduct a complete survey of the study area within one or two

48 47 days of each tagging event to provide estimates of observer efficiency. To provide information about survey life, additional surveys should be conducted following a tagging event, ideally at 3-4 day intervals. 7. The negative binomial escapement model should be further refined as follows: a) additional parameters should be included in the process model to allow for more complexity in arrival timing, b) the feasibility of modeling within-year variation in site-proportions should examined, and c) if the 2007 mark-recapture data warrants it, the model should be expanded to allow for estimates of Coquitlam River-specific variance and temporal patterns in observer efficiency as input. 8. During Treatment 2 target flow releases from Coquitlam Dam during the salmon spawning period steelhead spawning will increase to 2-6 cms compared to 1 cms during Treatment 1 (Table 1.1), but this is not expected to affect observer efficiency or survey life substantially. During much of the fall and winter, streamflows and water clarity in Coquitlam River are determined largely by natural inflows from Or Creek and sporadic controlled and uncontrolled spills from Coquitlam Dam; releases of very low turbidity water from the dam will have little influence on survey conditions during this period.

49 48 Appendix 2.1 Adult chum salmon unadjusted counts during No. sites Unadjusted count of the number of adults present Year Date Run day surveyed site A site B site C site D site E non-index Oct Oct Oct Oct Oct Nov Nov Nov Nov Dec Sep Sep Sep Oct Oct Oct Nov Nov Nov Nov Dec Dec Dec Dec Sep Oct Oct Oct Oct Nov Nov Nov Nov Dec Dec Dec Sep Oct Oct Oct Nov Nov Nov Nov Dec Dec Dec Dec Dec Sep Oct Oct Oct Oct Nov

50 49 Appendix 2.2 Adult coho salmon unadjusted counts during No. sites Unadjusted count of the number of adults present Year Date Run day surveyed site A site B site C site D site E non-index Oct Oct Oct Oct Oct Nov Nov Nov Nov Dec Dec Dec Dec Jan Sep Oct Oct Oct Nov Nov Nov Nov Dec Dec Dec Dec Jan Sep Oct Oct Oct Oct Nov Nov Nov Dec Dec Dec Dec Sep Oct Oct Oct Nov Nov Nov Nov Dec Dec Dec Dec Dec Jan Sep Oct Oct Oct Oct Nov Nov Dec Dec Dec Dec Jan

51 50 Appendix 2.3 Adult pink salmon unadjusted raw counts during 2003 and No. sites Unadjusted count of the number of adults present Year Date Run day surveyed site A site B site C site D site E non-index Sep Sep Sep Oct Oct Oct Nov Nov Nov Sep Oct Oct Oct Nov Nov Nov

52 51 Appendix 2.4 Mean daily temperatures in reaches 2,3, and 4 in Coquitlam River during July 1, 2005 December 31, Reach Temperature Celcius /1/05 10/1/05 10/31/05 11/30/05 12/30/05 1/29/06 2/28/06 3/30/06 4/29/06 5/29/06 6/28/06 7/28/06 8/27/06 9/26/06 10/26/06 11/25/06 12/25/06 24 Reach Temperature Celcius /1/05 10/1/05 10/31/05 11/30/05 12/30/05 1/29/06 2/28/06 3/30/06 4/29/06 5/29/06 6/28/06 7/28/06 8/27/06 9/26/06 10/26/06 11/25/06 12/25/06 24 Reach Temperature Celcius /1/05 10/1/05 10/31/05 11/30/05 12/30/05 1/29/06 2/28/06 3/30/06 4/29/06 5/29/06 6/28/06 7/28/06 8/27/06 9/26/06 10/26/06 11/25/06 12/25/06

53 ADULT WINTER STEELHEAD ESCAPEMENT 3.1 Methods To approximate adult winter steelhead abundance and potential egg deposition in Coquitlam River during , we conducted periodic redd surveys to estimate the cumulative number of redds constructed during the spawning period, and then used empirical data from other studies of coastal winter steelhead to approximate the number of redds constructed by each female, the average sex ratio, and mean fecundity per female (see section 3.1.4). Variation in counts among observers was not investigated, but was minimized by having the same two-person crew conduct all surveys. We used redd survey life data from the Coquitlam River to adjust our redd counts upwards to account for redds that were constructed and then became undetectable in the intervals between surveys (see section 3.1.3) Description of study area and survey methods For steelhead redds, the study area extended approximately 10.8 km from Coquitlam Dam downstream to Patricia Footbridge, and included reaches 2a, 2b 3, and 4 (Figures 3.1, 3.2). Reach 1 was omitted as minimal steelhead spawning occurs there. Bi-weekly surveys were scheduled from March until June, but the interval between surveys was often greater due to sustained periods of poor visibility. We attempted to conduct surveys just prior to high flow events in order to minimize the number of new redds becoming obscured by substrate movement before they could be detected. Owing to the length of the study area, each survey was completed over a two-day period. Redd surveys were conducted by two trained technicians that were familiar with steelhead spawning locations in the Coquitlam River and had considerable experience identifying steelhead redds. During each survey, one crewmember wore a dry suit and snorkeling gear and searched for redds in deep water, while the other wore chest waders and searched for redds in shallow water along the banks. The shoreline observer marked the location of each redd detected with numbered flags and a global positioning system (GPS) to prevent double counting on future surveys, and to provide estimates of redd survey life (see section 3.1.3). Additional data collected for each redd included width and length, specific location within the stream channel, and average substrate size. The crew also recorded the number of live adults observed on each survey, along with their location, and, if possible, their approximate fork length, sex, and whether they lacked an adipose fin indicating hatchery origin. Data for live adults were not used to estimate escapement.

54 53 Reach 4 Reach 3 Reach 2b See Inset Fig. 3.3 Scale 0m 500m 1000m Figure 3.1 Steelhead redd locations in reaches 2b-4 in Coquitlam River in 2006, which was the highest escapement year during the Coquitlam Dam is the upstream boundary of the survey area. See Figure 3.2 for redd symbol legend.

55 54 Reach 2b RST Site March 13, redds sites 8 No. redds 10 Only half of survey area completed due to sediment input from gravel pits. April 13, redd sites 30 No. redds 37 Incomplete due to rainfall Reach 2a April 19, redd sites 94 No. redds 248 May 13-15, redd sites 40 No. redds 82 June 12, 14, redd sites 17 No. redds 31 One dot represents one redd site. One redd site may represent more than one redd. Scale 0m 500m 1000m Figure 3.2 Steelhead redd locations in reaches 2a-2b in Coquitlam River in The downstream boundary of reach 2a is also the survey boundary.

56 Redd Identification Redds were identified as approximately dish-shaped excavations in the bed material, often of brighter appearance than surrounding substrates, accompanied by a deposit beginning in the excavated pit and spilling out of it in a downstream direction. Disturbances in the bed material caused by fish were discriminated from natural scour by: i) the presence of tail stroke marks; ii) an over-steepened (as opposed to smooth) pit wall often accompanied by perched substrate that could be easily dislodged down into the pit, and often demarcated by sand deposited in the velocity break caused by the front wall; iii) excavation marks alongside the front portion of the deposit demarcating the pit associated with earlier egg laying events; and iv) a highly characteristic overall shape that included a backstop of gravel deposited onto the unexcavated substrates, a deposit made up of gravels continuous with this backstop and continuing upstream into the pit, and a pit typically broader than the deposit and of a circular shape resulting from the sweeping of gravels from all sides to cover the eggs (in a portion of redds gravels are swept into the pit from only one side, often a shallow gravel bar on the shore side). A second important determination was whether fish had actually spawned at a location where an excavation had been started. Test digs were considered to be pits, often small, accompanied by substrate mounded up on the unexcavated bed material downstream but with no substrate swept into the pit itself, which would denote at least one egg deposition event. In the case of a test dig determination the mound of gravels would typically be short and narrow around the downstream side of a relatively small pit. Potential test digs were tagged and re-examined on subsequent surveys to determine if they had been further developed into actual redds. Redds constructed by resident cutthroat or rainbow trout or lamprey were distinguished from steelhead redds by their considerably smaller size, lack of a large deposit downstream of the pit, and a conical or bowl shape rather than a rectangular shape. In areas of limited gravel or high redd abundance, or where spawning site selection is highly specific, superimposition of redds can occur (Baxter and McPhail 1996). For our study, in such instances the redd count was based on a subjective evaluation, with the most recent complete redd(s) counted and the disturbed remains of prior redds being estimated in relation to it. A greatly extended deposit length (subjectively evaluated to be at least twice the length of a typical deposit length) was grounds to consider whether a second female had made use of the pit created by a first to construct a separate redd. Fortunately, such cases usually represent a very small proportion of the total number of redds present.

57 Redd survey life In most cases, steelhead redds can be readily detected upon initial construction, but over time, they become undetectable as they are obscured by scour or deposition, regrowth of periphyton, or superimposition of new redds. Thus, survey frequency is an important consideration in designing redds surveys, particularly for streams like Coquitlam River, where moderately high flow events can occur during the steelhead spawning period. If the length of time between surveys exceeds average redd survey life, than undercounting will occur. Freymond and Foley (1985) reported winter steelhead redds remaining easily identifiable for a period of 14 to 30 days in coastal Washington streams. Based on five years data from several coastal Oregon streams, Jacob et al. (2002) concluded that, on average, 95% of winter steelhead redds remain visible one week after completion, while 86% remain visible after two weeks. The CRMP-TOR (Leake 2005) recommended that during 2005 and 2006 weekly surveys be conducted during a three-week period that captured peak spawning, and biweekly surveys during the remainder of the spawning period. The assumption here was that none of the redds constructed during the threeweek period of weekly surveys would go undetected, and that, during this period, the proportion of redds from the first survey that were resighted on the second survey, but not the third survey, would provide a reasonable estimate of the proportion of redds constructed and then lost during the two-week interval between regular bi-weekly surveys. In 2007, the study was modified such that the goal was to complete weekly surveys during the majority of the spawning period (early April mid May) and bi-weekly surveys during the remainder. However, survey goals could not be achieve in any of the three years. Poor visibility conditions during limited the number of survey windows, such that the actual interval between surveys ranged from one to five weeks (Table 3.1). This necessitated an alternate means of estimating redd survey life. Since numbered flags were used to identify new redds (or groups of redds) during each survey, we were able track the percentage of these redds that remained visible on each subsequent survey (e.g., of 100 new redds observed on survey 1, 80, 30, 15, and 0 redds might remain visible on surveys 2-5, respectively). Using this information, we constructed a simple linear regression model to predict Y, the proportion of redds that remained visible x i days after construction. There were insufficient data to test whether redd longevity was dependent on initial date of construction. We treated the number of redds initially detected on survey i that remained on surveys i+1, i+2 as independent observations and pooled observations from surveys i-, recognizing that repeated observations of redds from survey i on surveys i+1, i+2 are not truly independent. We tested for possible differences in regression slopes among years (ANCOVA), with the intention of pooling all years data if sloped were similar. In order to correct the raw redd counts for survey i+1 to include lost redds, we assumed that redd construction was uniformly distributed over time between

58 57 survey i+1 and survey i, and divided the number of new redds observed on survey i+1 by the number of days since survey i, to arrive at number of redds constructed on each day between the two surveys that were still detectable on survey i+1. The regression equation was then used to predict the total number of redds constructed on each day, and these values were summed to estimate the total number of new redds constructed after survey i and before survey i+1 (see Appendix 3.1 for an example). Table 3.1 Survey dates with raw and estimated steelhead redd counts and adult observations for Raw count Estimated # # Live adults Year Survey date of new redds new redds observed Mar Apr Apr May Jun Total peak = Feb Mar Apr May Jun Total peak = Mar Apr Apr Apr May May Jun Total peak = 45 1 Redd survey incomplete due to poor conditions 2 Live adult totals incomplete, 3 Redd totals from aborted April 13 survey added to April 19 survey.

59 Female escapement and egg deposition The objective of the steelhead redd survey component is to allow smolt production to be related to spawning effort. Estimating the total number redds is therefore as or more useful for our purposes than estimating total escapement, since redd numbers are a more direct measure of spawning effort and egg deposition than adult numbers. However, the number of recruits per spawner is commonly expressed as the number of smolts per female. Following this convention, we converted our estimates of total redd abundance to total female abundance by relying on empirical estimates of the average number of redds per female for winter steelhead in Pacific coastal streams. Jacobs et al. (202) compared total redd counts to accurate estimates of female escapement for four streams over three years using total counts at full-span upstream fences, or at upstream fences coupled with intensive mark-recapture methodology. Gallagher and Gallagher (2005) reported redds/female values for winter steelhead in several streams, but their estimates were based on mark-recapture and AUC estimates that were themselves highly imprecise. Freeman and Foley (1985) reported the average the number of redds per adult in Snow Creek, Washington, but not the average number per female. In the Jacobs et al. (2002) study, the number of redds per female ranged from 0.75 to 1.63 an averaged 1.2, with relatively little variably among years for individual streams. We used this value (1.2 redds/female) to covert total redd numbers to female escapement. The total number of adult female steelhead in the surveyed portion of Coquitlam River (N) was approximated as: 1 N = RPF n x i i= 1 (3.1) where x i is the cumulative number of new redds summed across n surveys and RPF is the number of redds per individual female spawner (1.2). We used the redd survey life model to adjust values for x i to account for the estimated number of new redds that were constructed following survey i, but then became undetectable prior to survey i+1. Estimating egg-to-smolt-survival requires an estimate of average fecundity per female for Coquitlam River steelhead. As this was not available, we used average fecundity for winter steelhead in the Keogh River on northern Vancouver Island (3,700 eggs/female, Ward and Slaney 1993) to approximate that for Coquitlam River steelhead. We also generated estimates of total adult escapement by assuming a 1:1 sex ratio, which is an average value for coastal winter steelhead (Jacobs et al 2001). 3.2 Results and Discussion Redd surveys were conducted in Coquitlam River from late March to early June in 2005, from mid February to mid June in 2006, and early March to mid June in 2007 (Table 3.1). The frequency and magnitude of high flows during the

60 and 2006 survey periods was relatively moderate. In contrast, during 2007, mean daily flows remained above 10 cms for most of March, with a peak flow of 118 cms occurring on March 11. However, during April and May, when most of spawning occurred (Figure 3.3), flows seldom exceeded 10 cms during In all years, surveys were conducted at flows of between 2-4 cms. Frequent poor stream visibility conditions limited the frequency of surveys in all years (see Section 3.2.1) Feb 19-Mar 8-Apr 28-Apr 18-May 7-Jun 27-Jun Feb 19-Mar 8-Apr 28-Apr 18-May 7-Jun 27-Jun Feb 19-Mar 8-Apr 28-Apr 18-May 7-Jun 27-Jun Figure 3.3 Discharge (cms) in Coquitlam River during steelhead spawning period in

61 60 Adult steelhead were present on the first survey in 2006 and 2007 (February 15 and March 2, respectively, Table 3.1, Figure 3.4), but spawning had not yet begun. Spawning was already underway at the time of the first survey in 2005 (March 24). During each year, about 85% of spawning appeared to occur over a six-week period from early April to mid May (Figure 3.4). Spawning steelhead preferred mainstem habitat to natural side channel and constructed off-channel habitat by a large margin. Of the total number of redds observed in 2005 and 2006, 78% and 94% were in the mainstem, respectively. Average redd size was about 2 m 2 during all years. Misidentification of resident trout or lamprey redds as steelhead redds did not appear an issue, as the former were much smaller than steelhead redds, and, in the case of trout, spawning was largely complete prior to the beginning of steelhead spawning. 550 Cumulative number redds Feb 4-Mar 24-Mar 13-Apr 3-May 23-May 12-Jun Figure 3.4 Cumulative proportion of the total steelhead redd count observed over time during the , and 2007 surveys. On the March 13 and April 13 survey dates in 2006, only a portion of the study area was surveyed (reaches 4, 2a and 2b on March 13; 50% of reach 2b on April 13). Ten and 37 new redds were observed on the March 13 and April 13 surveys, respectively. We approximated the number of new redds in reaches 3 and 4 and the remainder of reach 2b that were missed on March 13 as follows: i) for each of the remaining (complete) surveys in 2006 (April 19, May 13, June 12), we calculated the proportion of the total number of redds counted that were from the section surveyed on March 13, and ii) we divided the March 13 redd count

62 61 (10 redds) by the average of these proportions (0.314) to arrive at 32 redds (Table 3.1). In other words, we assumed that relative spawning effort among sections in 2006 was constant over time. The April 13 survey occurred only six days before a complete survey on April 19, so we simply added new redds counted on April 13 to the 248 new redds counted on the April 19 survey (Table 3.1). In 2006 the maximum number of live adults observed on a single survey was 95 (Table 3.1), which was four-fold higher the peak count in 2005 (22 adults), and double the peak count in 2007 (45 adults). However, counts of live adults were incomplete on several surveys in 2005 (Table 3.1). By comparison, during , when snorkel counts of adult steelhead occurred as part of a larger survey of steelhead escapement in BC Lower Mainland streams (BCCF, Lower Mainland Branch, data on file), the maximum number of adult steelhead observed on any one survey ranged from (Figure 3.5). However, values shown in Figure 3.5 do not necessarily reflect the annual trend in actual Total adult count Adults / km 10 Adult Steelhead Count Adults observed / km 2 0 Figure 3.5 Peak counts and densities (fish/km) of adult steelhead during snorkeling surveys in (only data for complete surveys of the study area are shown). Data for were collected as part of a separate study (BCCF, Lower Mainland Branch, data on file).

63 62 escapement. Unadjusted peak counts of winter steelhead are often poorly correlated with actual escapement (Korman et al. 2002). Counts of live adult steelhead in Coquitlam River varied more within years (average of CVs for within-year counts during was 0.58) than among years (CV for mean annual live counts for was 0.53) Redd survey life The ability of the survey crew to re-identify older redds that were originally observed on a previous survey was strongly dependent on the number of days that had elapsed since the previous survey (Figure 3.6). The data suggested that all redds remained detectable for up to seven days. Beyond seven days, there was a linear decline over time in the proportion of redds that remained detectable, such that, after days, 100% of redds were no longer detectable. In 2005, 2006 and 2007, the mean and range for the interval between surveys were 18 and 9-29 days, 29 and days, and 17 and 9-34 days, respectively (Table 3.1). To model redd survey life, we first assumed that all redds remained detectable after seven days and that 100% of redds were undetectable after 45 days. We then developed a simple linear regression model to predict the proportion of redds no longer detectable 8-45 days after completion (% redds lost = x number days since previous survey , n = 25, R 2 = 0.81, P < ). The model predicted a rate of redd loss ranging from 22% after 14 days, to 43% after 21, to 65% after 28 days. These estimates were comparable to other published values. Gallagher and Gallagher observed redd loss rates of 27% and 75% after 14 and 28 days, respectively, for winter steelhead in coastal streams in California. For Oregon coastal streams, Jacob et al. (2002) observed winter steelhead redd loss rates of 14%, 33%, and 50% after 14, 21, and 28 days, respectively. When corrected for undectected redds, total redd counts in 2005, 2006 and 2007 counts increased by 14%, 21% and 8%, respectively (Table 3.1). However, these corrections were based on the assumption that the distribution of redd construction over time was uniform (Section 3.1.3), which is probably unrealistic, as it does not take into account the shape of the spawning timing curve (Korman et al. 2002). Because surveys were often infrequent during the major period of spawning, particularly during 2005 and 2006, it was not possible to accurately model the distribution of spawning over time. As more information about spawning timing is gathered in future years, a more realistic redd survey life model may be possible, which could be used to reanalyse the 2005 and 2006 data. Alternatively, it may be possible to develop an AUC model, similar to that used to estimate adult salmon escapement to Coquitlam River, to estimate annual steelhead redd numbers, with redd survey life analogous to survey life (residence time) for live adult salmon.

64 63 Proportion of redds no longer visiblesss Number of days from last survey Figure 3.6 Proportion of newly constructed steelhead redds that remained visible to observers during subsequent surveys in Coquitlam River during Female escapement and egg deposition Based on redd counts and an assumed value of 1.2 redds/female, estimated adult female escapement in 2006 (434 females, Table 3.2) was double that in 2005 (202 females), and four-fold that in 2007 (134 females). Annual spawner density for the Coquitlam River as a whole ranged from females/km (Table 3.2). Among reaches and years, female spawner density ranged from 5-60 females/km (Table 3.2). In each year, female spawner density was comparable among reaches 2b, 3 and, and several-fold lower in reach 2a. The exception was in 2007, when spawner density in reach 4 was unusually low. The actual distribution of steelhead redds within the study area during 2006, the year of highest escapement, is shown in Figures 3.1 and 3.2. In the Keogh River an egg deposition of 13,300 eggs/km was estimated as the minimum required for saturating juvenile rearing habitat (derived from Ward and Slaney 1993). Egg deposition in Coquitlam River exceeded this value by a factor of 5.2 in 2005, by a factor of 11.2 in 2006, and by a factor of 3.5 in 2007 (Table 3.2), suggesting that, even if a somewhat higher egg deposition is

65 64 Table 3.2 Summary statistics for steelhead escapement to Coquitlam River during based on redd counts. Total number Total female Females Total egg Eggs Total adult Year Reach of redds spawners /km deposition /km escapement a ,000 24, b ,000 79, , , , , Total ,000 69, a ,000 52, b , , , , , , Total ,606, , a ,000 19, b ,000 63, , , ,000 24, Total ,000 46, required for Coquitlam River due to better growing conditions and younger mean smolt age (Chaput et al. 1998) compared to that in Keogh River, the minimum level of egg deposition required for saturating juvenile rearing habitat was still more than met. 3.3 Recommendations for the 2007 survey The Coquitlam River is well suited to conducting steelhead redd surveys. The only major problem encountered was sustained periods of poor water visibility in 2005 and 2006 during the major 6-7 week spawning period from late March until early May. During Treatment 2 target flow releases from Coquitlam Dam during steelhead spawning will increase to 3-4 cms compared to 1 cms during Treatment 1 (Table 1.1), but is unlikely that this will reduce the ability of surveyors to detect redds because water released from the dam is extremely low in turbidity. During it was usually increases in water turbidity from downstream gravel mining operations or much higher spate flows originating in Or Creek that precluded redd surveys.

66 65 The CRMP-TOR originally recommended that redd surveys being conducted every two weeks, with surveys conducted on three consecutive weeks during the peak of the spawning period to provide an estimate of the number of redds constructed and lost before they could be counted during the normal two-week interval. However, experience from 2005 and 2006 indicated that in many years this will not be possible due to poor survey conditions. As an alternative, we recommend that surveys be scheduled weekly from late March through the first half of May in order to minimize the number of new redds that are constructed and lost between surveys. Every effort should be made to take advantage of periods of clear water conditions during this period, with surveys being conducted less than one week if necessary to avoid missed surveys. Jacobs et al. (2002) recommended conducting steelhead redd surveys in Oregon coastal streams on a 7-10 day recurrence interval. If resources are limited, surveys in Coquitlam River should be delayed until the third week of March to allow for more surveys during the peak period, since relatively few redds were constructed prior to late March during We also recommend the continued use and development of the redd life model described in this report (or an alternative, see Section 3.2.1) in place of the model proposed in the CRMP-TOR. Finally, in future years the number of survey crews should be increased from one to two so that the entire study area can be surveyed in one day, thereby decreasing the risk of a survey being aborted due to the onset of poor conditions. Incomplete redd surveys were generally not a problem during , but if were in future years as a result of poor survey conditions, it may be possible to extrapolate redd numbers from surveyed to un-surveyed sections using a similar approach to that taken for estimating adult salmon escapement (see Section 2).

67 66 Appendix 3.1 An example of how raw survey counts were expanded to account for redds that were completed and subsequently became undetectable between surveys (see section 3.2.1). May 13, 2006 redd survey Total # new redds observed 82 Number days from previous survey 24 Number of redds constructed per day 3.42 assuming uniform distribution of spawning over time Redd survey life equation for days 8-24 Proportion redds lost = *day (assuming no redds lost for days 1-7) Day Loss rate Adjusted # redds Total new redds adjusted for redd survey life

68 JUVENILE SALMONID STANDING STOCK In 2006 the COQWUP CC requested that a juvenile standing stock survey component be added to the monitoring program to provide estimates of total abundance for coho and steelhead fry (age-0+), and steelhead parr (age-1+ and 2+) in the Coquitlam River mainstem. The existing smolt trapping program provides smolt abundance data only for these species. In early September 2006, we conducted a feasibility study to determine the best approach for collecting annual juvenile standing stock data. The study compared three different sampling methods, three-pass removal electrofishing at 20 m long enclosed sites along one shoreline, night snorkeling counts within the electrofishing sites, and night snorkeling counts at sites that extended across the entire stream channel, at 10 sampling locations distributed throughout reaches 2a,2b, and 3. Reach 4 was excluded from the feasibility study in Guided by the results from 2006 (Decker et al. 2007), in 2007 we proceeded with a juvenile survey that was limited to night snorkeling counts at full channel sites. In 2007 the survey was expanded to include reach 4. A multi-year markrecapture study was also initiated to provide estimates of snorkeling efficiency (proportion of total fish at a site that snorkelers detect), which is necessary to expand raw snorkeling counts to population estimates. In 2007 we also conducted a separate electrofishing survey with the assistance of Ron Ptolemy (MOE stock assessment) to further evaluate the effectiveness of mutiple-pass electrofishing as a method for estimating abundance and assessing fish-habitat relationships in Coquitlam River. The results of this work are summarized in a separate management brief (Ptolemy 2007) which is included as Appendix 6.1 at the end of this document). 4.1 Methods Study design In 2007 the study area included 10.3 km of mainstem habitat from Coquitlam Dam downstream to Patricia Footbridge just upstream of Lougheed highway (Figure 4.1). Sampling sites were chosen using a simple (unstratified) systematic sampling design (SSS). Sampling was not stratified by reach or habitat type on account of the limited number of sites sampled. In 2007 the 10 sites selected in 2006 using the SSS design were re-sampled, and an additional two sites added in reach 4 to maintain a uniform sampling interval of 0.85 km (12 sites in total; Figure 4.1). Natural and man-made off-channel habitats in Coquitlam River were not surveyed, and juvenile fish populations in these habitats are therefore not included in juvenile standing stock values reported in this section and in Section 6. To correctly locate sampling sites in the field, we used a hand-held GPS unit to determine the straight-line distance from Patricia Footbridge to Coquitlam Dam, and divided this distance by the number of target sites to obtain a uniform

69 68 sampling interval. The downstream boundary of each site was then located according to the appropriate pre-determined distance from Patricia Footbridge. Each site was 25 m in length and spanned the entire stream channel. If the stream was split into two or more wetted channels at the selected site location, the entire wetted width of all channels was surveyed as part of the 25 m site to ensure that the site accurately represented available habitat for a particular channel cross-section. The standing stock survey was scheduled for early September when precipitation is normally low and regulated discharge from Coquitlam Dam is about 1 cms Night snorkeling counts Snorkeling counts of juvenile salmonids were performed once at each site by a two-person crew. Counts were performed at night because numerous studies have shown that daytime concealment behaviour is common in juvenile salmonids (e.g, Bradford and Higgins 2000 and references therein), and likely depends on factors such as temperature, time-of-day, season and habitat. We limited snorkeling surveys to a four-hour period beginning 0.5 hours after dusk. We based this on Bradford and Higgins (2000) finding that, throughout the year, the highest counts of juvenile salmonids during a 24-hour period were consistently recorded during a 3-4 hour period after dusk. To illuminate the sampling sites at night, snorkelers used handheld dive lights that cast diffuse rather than direct beams to minimize the disturbance to fish. Snorkelers surveyed the stream's entire wetted width, with each snorkeler entering the site at its downstream end and systematically sweeping in an upstream direction the area between his bank and the agreed upon mid-point of the site. Regular communication between snorkelers was essential to avoid duplicating counts, particularly in the instances where fish were present in mid-channel areas. Some studies have shown that snorkeler counts are not effective for estimating the abundance of age-0+ salmonids (Griffith 1981; Campbell and Neuner 1985; Hillman et al. 1992), principally because they occupy shallow (< 20 cm deep) near-shore habitats that are difficult for snorkelers to survey from an underwater, off-shore position. To address this, during snorkel surveys, snorkelers delineated areas that were too shallow to view from an underwater position, and, following the completion of an underwater search of the remainder of the site, conducted a separate foot survey of these areas with mask removed. We found that at night small fish along the stream margin remained relatively stationary and could be identified to species and size class, and, if necessary, could be captured with a small net to confirm observations. At sites where these shallow areas were not well delineated from the rest of the site and the risk of double counting fish was apparent, the two snorkelers worked parallel to one another, with one person searching shallow near-shore areas, and the other searching adjacent off-shore areas. Each person communicated movements of detected fish to the other. This procedure was then repeated for the other half of the site. Other studies have shown that streamside visual counts can be

70 69 excellent predictors of abundance estimates when calibrated using more accurate methods (Bozek and Rahel 1991; Decker and Hagen 2006). Using underwater notebooks, snorkelers tallied numbers of steelhead and coho fry and visually estimated the fork lengths of individual steelhead parr. Parr forklength data were collected to aid in estimating age class (1+ or 2+). To

71 Figure 4.1 Map of Coquitlam River showing juvenile standing stock study area, reach breaks and sampling sites. 70

72 71 estimate steelhead parr length, snorkelers drew ruled scales on the cover of their notebooks. Snorkelers were typically able to hold the notebooks within 30 cm of a fish to measure its length without disturbing it Mark-recapture experiments to estimate snorkeling efficiency To derive population estimates for individual sampling sites from snorkeling counts, snorkeling efficiency (proportion of total fish at a site that snorkelers detect) must be estimated. In 2007 we initiated a mark-recapture study for this purpose. The mark recapture study will continue during future years of the monitoring program, with approximately four sites included each year, until enough data are obtained to provide a reliable model of snorkeling efficiency. By distributing the mark-recapture experiments over several years and equally among the 12 annual sampling sites, bias resulting from differences in snorkeling efficiency among years or habitat types will should be minimized. To estimate snorkeling efficiency discretely for each target species/age class at a sampling site, one night prior to conducting the normal snorkeling survey as described above, a single snorkeler captured and marked fish throughout the site using one or two large aquarium nets affixed to handles of approximately 80 cm in length. The snorkeler searched for and captured fish throughout the site, with the goal of obtaining marked individuals each for the three primary species/age groups (coho fry, steelhead fry, and steelhead parr). Minimizing disturbance to marked and unmarked fish was a primary goal of the marking methodology. Captured fish were handed to a second crew member on shore, who immediately measured the fish (fork length to nearest 5 mm), marked it, and returned it to its original location, once the snorkeler had moved to a new location. Anticipating that snorkeling efficiency would differ for smaller and larger juvenile steelhead over the size range occurring in Coquitlam River (Decker and Hagen 2006; Hagen et al. 2005), we used colour-coded tags to obtain five discrete mark groups for steelhead (<50 mm, mm, mm, mm, and > 150 mm). The smaller two length classes represent age-0+ fry, while the larger three represent age-1+ and 2+ parr. Marking consisted of inserting a custom-made tag into the fish s back at the insertion of the dorsal fin. Tags consisted of size barbed fish hooks (size 16 for fish > 170 mm fork length, size 18 for fish mm, and size 20 for fish < 70 mm) with a length (8-15 mm depending on fish size) of coloured plastic chenille attached at the hook eye with heat shrink tubing (Hagen et al. 2005). We adjusted the length of the coloured chenille in accordance to fish size, so that during the snorkeling survey a mark could be readily detect on a fish, without increasing the likelihood of a marked fish being seen relative to an unmarked one. Captured fish were not anaesthetized because of uncertainty about behavioural effects from the anesthetic, and because they could be measured and marked without being removed from the net. During the underwater survey the following evening, marked and unmarked fish were recorded separately, so that snorkeling

73 72 efficiency could be estimated (snorkeling efficiency = marked fish seen / fish marked, or R/M). This type of mark-recapture study assumes a closed population, whereas our sites were not enclosed. Over sufficiently short time periods, however, and if study animals restrict their movements to a defined area, physically open sites can be treated as closed without introducing significant bias (Pollock 1982; Bohlin et al. 1989; Mitro and Zale 2002). We chose to conduct the underwater surveys 24 hours after marking because we considered this to the shortest time period that would still allow fish to recover from marking and complete a diurnal cycle of movement and redistribution within the site, but would minimize movement from the site. We investigated the assumption of site closure by surveying an additional distance of approximately half the site length adjoining both the upstream and downstream site boundaries, so that the total distance surveyed for marks was approximately two times the length of the original site where fish were marked. Marked fish that had moved beyond the original site boundaries were recorded separately. The number of marked fish that emigrated from the original site was estimated as the number of marks observed in the adjoining sections divided by R/M Physical characteristics of the sampling sites To describe the physical characteristics of the sampling sites, we conducted basic surveys. At each site, depth was measured at five stations along each of three transects spanning the width of the site. Stations were uniformly-spaced along transects, and transects were uniformly-spaced along the length of the site. We also recorded maximum depth, substrate composition (boulder, cobble, gravel, and fines as percentages of the site area), D90 and D50 (diameters of substrate particles for which 90% and 50%, respectively, of the site area consist of smaller particles), site length, site width, cover (categories included: overhead vegetation, turbulence, deep water and boulder as percentages of the site area, undercut bank as a percentage of the combined length of the stream banks, and the total area of the site covered by wood debris). Other information collected for each site included location (UTMs), and water quality parameters (water temperature, ph, and total alkalinity taken at the time of sampling at each site). Although, the limited number of sites sampled (n 12) and the necessity of adjusting significance tests for multiple comparisons meant that statistical power was low, we used Pearson correlations to test for significant relationships between fish density and habitat variables that have been cited as important factors in determining fish distribution in other studies (mean depth, D90, % fine substrate, and total cover as a scaled index from 0-1) Standing stock estimation With respect to the sampling design for the standing stock survey, error in the estimation of mean fish density and standing stock is the result of both first stage or process error (spatial variation in fish abundance from site to site) and second

74 73 stage or measurement error. Measurement error is the variation in snorkeling efficiency among sites that results from differences in fish behaviour, site characteristics, visibility, snorkeler experience, etc. With respect to first stage error, stream fish populations are usually distributed non-normally, with spatial distribution patterns sometimes varying from year-to-year. To address the problem of sparse or non-normally distributed fish density data, we used a nonparametric bootstrap procedure (Efron and Tibshirani 1993) to estimate variance in fish density among sites. The bootstrap method relies on repeated sampling of the actual data, so no assumption need be made about the underlying probability distribution of the data. For each species/age class of interest, mean fish density per linear m (D) and standing stock (SS) were computed for the surveyed portion of Coquitlam River simply as point estimates from the sample, since point estimates generally perform better than bootstrap estimates of the median or mean (Efron and Tibshirani 1993): D n i = 1 = n β N i= 1 i L i (4.1) SS = D (4.2) L Coquitlam where: i = sample site n i = the number of sites for which data were collected N i = snorkeling count at one of n sample sites randomly selected by the bootstrap model. D shoreline = fish/m 2 at the shoreline electrofishing site L j = length (m) for site i Length Coquitlam = total stream length for the study area (10,300 m) and n i i = 1 β = n (4.3) i = 1 R M i where:

75 74 R i = number of marked fish detected by snorkelers at site i M i = number of fish marked at site i We obtained 95% confidence intervals for D and SS by bootstrapping the data. Bootstrap iterations were computed using an algorithm written for R, an open-source statistical program. For each species/age class category, 5000 iterations of equations were computed with the bootstrap model choosing sample sites (N i ) with equal probability and with replacement. For each iteration, the number of sites selected was set equal to the actual number of sites sampled (12). Upper and lower 95% confidence limits for D and SS were estimated as the 97.5% and 2.5% percentiles, respectively, from the cumulative distribution of the 5000 bootstrap iterations (Efron and Tibshirani 1993). In cases where the median of the bootstrap estimates differed from the point estimate, we used the difference between the estimates to develop bias-corrected confidence limits (Haddon 2001). To describe the precision of the standing stock estimates, we used percent relative error, which can be defined as the average half confidence interval (UCL minus LCL divided by two) divided by the mean and expressed as a percentage (Krebs 1999). Measurement error (snorkeling efficiency) was incorporated in the 95% confidence intervals for the population estimates by including a Monte Carlo procedure within the bootstrap routine as follows: As each site was selected by the bootstrap algorithm during an iteration, β was varied stochastically using a function that randomly selected a snorkeling efficiency value from a normal distribution with a mean equal to β and a standard deviation derived from the estimates of β at mark-recapture sites i-n. However, the number of markrecapture sites completed in 2007 was low (n=4), and this would result in overestimates of SD(β) for each species/age class. We used 2007 markrecapture data from Coquitlam River to estimate mean values for β in equation 4.3, but, to estimate error in β, we substituted SD(β) values derived similar markrecapture experiments in other streams (values were standardized between studies by taking CV values for other studies and converting them to SD for Coquitlam River based on CV = SD/mean). For steelhead parr, we relied on a dataset from 51 sites distributed among seven streams in the lower Thompson River in the BC southern interior (Hagen et al. 2005; Table 4.1). For Coquitlam River coho we used a dataset for age-0+ Chinook from 40 sites distributed among five streams in the lower Thompson River (Table 4.1; Decker and Hagen 2007). For steelhead fry we used a dataset for Gerrard rainbow trout at 19 sites in the Lardeau River, a medium-sized stream in the West Kootenays (Decker and Hagen 2006; Table 4.1). It is important to note that standing stock values and there confidence intervals for 2006 and 2007 will differ in future years reports, as estimates of β are refined by additional mark-recapture experiments, as Coquitlam River-specific estimates of SD(β) are substituted for values from other streams, and as site-specific habitat or environmental variables (e.g., temperature, mean depth, etc.) are incorporated in snorkeling efficiency models, if they are found to significant predictors of snorkeling efficiency.

76 75 Table 4.1 Snorkeling efficiency data sources used to generate population estimates for Coquitlam River coho and steelhead parr and fry. Coquitlam River Data source Snorkeling efficiency species/age class location species mean CV # sites # streams steelhead parr Thompson tribs steelhead parr steelhead fry Lardeau River rainbow trout fry coho fry Thompson tribs Chinook fry Results In 2006 and 2007 the juvenile standing stock survey was completed during September 6-14, and September 6-20, respectively, at flows of m 3 /s and m 3 /s, respectively ( 18%-32% and 18%-37% post-flow regulation MAD, respectively; WSC, station 08MH002, Port Coquitlam). Water temperatures at the time of the night snorkeler surveys ranged from 17 C to 19 C during 2006, and from 14 C to 20 C during Conductivity was measured at each site in 2006 and ranged from 10 μs/cm in reach 3 to 30 μs/cm in reach 2 downstream. Total alkalinity was consistently low in both years (8-14 mg/l; Table 4.2). In 2006 turbidity was low at all sites (0.4 to 2.8 NTU), allowing for more than adequate underwater visibility for conducting snorkeling counts (> 4 m measured horizontally underwater). In 2007, horizontal underwater visibility was 4 m or greater at sites upstream of the gravel mining sites (and at site 0.55 near Port Coquitlam; Table 4.2), but releases of sediment- laden water from the gravel mines caused frequent periods of poor underwater visibility at downstream sites, and it was necessary to reschedule surveys on several occasions. Downstream of the gravel pits, four sites (1.25, 195, 2.65, and 3.35) were surveyed at an underwater visibility of 3 m. Site 4.05, immediately downstream of the gravel mining operations was not surveyed due to insufficient visibility. Table 4.2 Summary of habitat data for night snorkeling sampling sites in Coquitlam River. Upstream Site Mean Max. Total Cond- Site distance area thalweg depth D90 D50 Fines Gravel Cobble Boulder alkalinity uctivity no. (km) (m 2 ) depth (m) (m) (m) (m) (%) (%) (%) (%) (mg/l) (μs/cm)

77 76 Table 4.3 Summary of mark-recapture results and snorkeling efficiency estimates for four sampling sites in Coquitlam River in Fish _95% No. of marks resighted fork length Marks Resighted Snorkeling Confidence limit in adjacent sections to class (mm) Site (M) marks (R) efficiency SD Lower Upper original marking site Coho fry all all Steelhead fry < all all all all Steelhead parr all all > all all all

78 Mark-recapture experiments to estimate snorkeling efficiency In 2007, we opportunistically conducted mark-recapture experiments at four sites upstream of the gravel pits while waiting for water quality to improve below. Because regular surveys had already been made at several sites above the gravel pits prior to the mark-recapture work, we added 2 additional sites (5.15, 5.35) downstream of Or Creek for mark-recapture purposes only (data from these sites did not contribute to estimates of mean fish density). In total, 50 coho fry, 40 steelhead fry, and 61 steelhead parr were marked at the four markrecapture sites (Table 4.3). Based on detection of marked fish by snorkelers during the survey 24 hours after marking, maximum likelihood estimates of mean snorkeling efficiency differed significantly for coho fry, steelhead fry and steelhead parr, (52%, 38%, and 72%, respectively; Pearson chi-square, χ=12.34, df=2, P=0.002; Table 4.3). For steelhead, the data also suggested a curvilinear, dome-shaped relationship between snorkeling efficiency and fish length (Figure 4.2); snorkeling efficiency increased sharply among length categories from <50 mm to mm, and then declined modestly for larger length classes. However, sample sizes of marked fish were low when the data are partitioned by length class, making this result uncertain. Testing of the statistical significance of fish length and other factors on snorkeling efficiency awaits additional markrecapture experiments in future study years. 100% Snorkeling capture efficiency 80% 60% 40% 20% % Coho fry < > 150 Steelhead length class (mm) Figure 4.2 Maximum likelihood estimates of mean snorkeling efficiency for juvenile coho and steelhead by fork length class (steelhead only) at four sites in the Coquitlam River in Values above bars are total numbers of resighted marked fish for each category.

79 78 Numbers of marked fish resighted by snorkelers in upstream and downstream sections adjacent to mark-recapture sites supported the assumption that populations within mark-recapture sample sites can be considered effectively closed. During snorkeler surveys at mark-recapture sites one night after marking had occurred, only two marked coho, one marked steelhead fry, and two marked steelhead parr were detected in adjacent upstream and downstream sections as opposed to the original marking site. This equated to total emigration rates of 6.2%, 8.3%, and 3.3% for marked coho steelhead fry and steelhead parr, respectively. Table 4.4 Preliminary estimates of juvenile fish density, standing stock, and 95% confidence intervals by species and age class in Coquitlam River in 2006 and Density Standing Lower Upper ± Species/age class Year (fish/100 m 2 ) stock 95% CI 95% CI 95% CI coho (0+) ,031 12,044 66,044 84% coho (0+) ,795 15,528 48,993 58% steelhead (0+) , , ,953 45% steelhead (0+) ,081 20,687 82,396 66% steelhead (1+) ,728 2,992 9,117 53% steelhead (1+) ,992 8,406 18,287 38% steelhead (2+) , ,817 88% steelhead (2+) , ,230 66% steelhead (all parr) ,114 3,502 11,581 57% steelhead (all parr) ,879 9,680 21,107 38% Juvenile fish distribution and abundance Coho fry abundance in Coquitlam River was similar in 2006 and Mean densities were 14.3 and 12.9 fish/100m 2 (Table 4.4), respectively, and standing stocks were 32,000 and 29,000, respectively. In 2006 (when sampling sites extended from reach 2a to reach 3, Figure 1), there was no apparent longitudinal pattern in coho density among sites (Figure 4.3); in 2007 (when sites extended from reach 2a to reach 4), coho density was positively correlated with distance from the stream mouth (R=0.73, P=0.005, n=13; Figure 4.3). Steelhead fry abundance in 2006 (75.7 fish/100m 2 ; Table 4.4) was three-fold higher than in 2007 (21.1 fish/100m 2 ). Standing stocks of steelhead fry were 169,000 and 47,000 in 2006 and 2007, respectively (Table 4.4). In both years,

80 79 peak steelhead fry densities were observed at sampling sites in the middle portion of the study area (11-16 upstream of the stream mouth; Figure 4.3) Steelhead fry Fish / 100 m Steelhead parr Coho fry Distance from stream mouth (km) Figure 4.3 Comparison of fish density estimates for 10 adjacent shoreline and channelwide sites in Coquitlam River. Estimates are based on calibrated snorkeling counts.

81 80

82 81 Abundance of age-1+ steelhead parr increased 2.3-fold in 2007 (13,000 parr, 5.8 parr/100m 2 ; Table 4.4) compared to 2006 (5,800 parr, 2.6 parr/100m 2 ). Higher age-1+ parr abundance in 2007 corresponded to relatively high fry abundance in Age-2+ parr numbers were similar in 2006 and 2007 (1,400 and 1,900 parr, respectively; Table 4.4). Densities of age-1+ and 2+ parr combined were 3.2 parr/100m 2 in 2006 and 6.7 parr/100m 2 in 2007, with age-1+ parr representing 81% and 87% of the total parr standing stock in 2006 and 2007, respectively. There was no apparent longitudinal pattern in steelhead parr density among sites in 2006 (Figure 4.3), whereas at a higher overall density in 2007, parr density was consistently higher in the upper river (13-18 km from stream mouth, Figure 4.3; reaches 3,4 and upper portion of 2b, Figure 4.1) Fish density versus habitat All species/age classes exhibited high site-to-site variation in fish abundance (Figures 4.3), but this variation was not strongly dependent any one habitat variable. When P-values were corrected for multiple comparisons, correlations between coho fry, steelhead fry or steelhead parr density and a subset of habitat variables most likely determine habitat suitability (mean depth, D90, % fine substrate, and total cover as a scaled index from 0-1) were all non-significant. The positive correlation between steelhead parr density and D90 was the only case where R exceeded 0.50 in both years (2006: R=0.57; 2007: R=0.45). Site-specific habitat data are summarized in Table Discussion Suitability of snorkeling counts for juvenile population census Results from 2006 and 2007 suggest that night snorkeling is an efficient and effective method for surveying juvenile abundance in the Coquitlam River mainstem. Snorkelers found that they could readily survey all habitats used by juvenile fish, and in no instance was it necessary to relocate a systematically selected site on account of difficult survey conditions. Typical late summer flows and water quality conditions in Coquitlam River under Treatment 1 and 2 likely provide adequate underwater visibility (> 4 m) to conduct snorkeling surveys. During most of September 2007 underwater visibility deteriorated to less than 4 m in reaches 2a and 2b as a result of frequent releases of sediment-laden water from several nearby gravel quarries. In future years, the survey crew will liaise with gravel mine operators to schedule sediment releases to avoid impacts to BC Hydro monitoring activities. Although, in 2007, snorkeling surveys were conducted at several sites at a visibility level of about 3 m, Hagen et al. (2005) found no relationship between snorkeling efficiency and visibility at visibility levels of 3 m or greater. Preliminary mark-recapture experiments conducted in 2007 demonstrated that snorkeling efficiency or capture probability (proportion of fish at a site observed by snorkelers) was relatively high, particularly for steelhead parr (72%). This is an important result since the accuracy and precision of any sampling

83 82 method is directly and positively related to capture probability (i.e., snorkeling efficiency; Otis et al. 1978). The close similarity in estimates of mean snorkeling efficiency for coho and steelhead fry and steelhead parr in this study and in other studies (see Tables 4.1 and 4.3), despite the limited number (4) of markrecapture experiments completed in Coquitlam River, suggests that the sampling method may be relatively robust to the effects of differing habitat conditions on capture probability. Once we have completed enough mark-recapture experiments, we will incorporate a Coquitlam River-specific model for snorkelling efficiency in the bootstrap model for estimating standing stock. Factors likely to affect snorkeling efficiency (e.g., fush size class, site area, mean depth, temperature, etc.) would be included as independent predictor variables if they were found to explain a significant amount of variance in snorkeling efficiency (Thurow et al. 2006). A key assumption of the mark-recapture methodology was that marked and unmarked fish had equal probabilities of being seen by snorkelers the night following marking. Testing for this type of bias was beyond the scope of this study, but results from a similar study of steelhead parr in the Thompson watershed (Hagen et al. 2005; in prep.) suggests that fish marked by snorkelers can provide unbiased estimates of nighttime snorkeling efficiency. In that study two independent calibration methods were used: mark-recapture using snorkelers to capture fish for marking, similar to this study, and three-pass electrofishing removal estimates. After the removal estimates were adjusted for expected negative bias associated with declining capture probabilities (15-25% in good estimates of juvenile salmonid abundance, see Bohlin and Cowx 1990), the removal data provided an estimate of snorkeling efficiency (66%-72%) that compared favourably with the mark-recapture estimate of 70% (Hagen et al in prep.). In our study, the sampling protocol for snorkeling-captured fish was designed to minimize the behavioural effects on fish from handling and marking. Fish were captured in a relatively low impact manner (hand nets), were not anaesthetized prior to marking, were released into the same location that they had been captured from (or first seen in), and were allowed a 24-hour recovery period prior to the snorkeler count. Snorkelers noted that, after 24 hours, marked fish occupied comparable locations to unmarked ones and behaved in a similar way. A second assumption of our mark-recapture methodology was that the populations were closed between marking and resighting events. While our sites were not enclosed, we treated the fish populations within as being closed over the 24-hour period between marking and the snorkeler survey. The incidence of parr emigrating from the original marking sites was relatively low, consistent with our assumption of site closure. Of the 85 fish marked by snorkelers that were resighted, only five had moved to an adjacent upstream or downstream stream section. Moreover, the marked fish that had moved from the original marking site had remained within a short distance of the original site boundaries. We could not detect the number of fish that had moved beyond the adjacent sections, but

84 83 given that mark-recovery rates within the original sites were relatively high, and that we expanded the sites during the recapture survey to account for small-scale movement, a significant violation of the assumption of population closure was unlikely Fish density and standing stock estimation Riley et al. (1997) surveyed juvenile abundance in the Coquitlam River in 1997, prior to the installation of the fish flow valves and the implementation of the existing flow regime. Although their sampling methodology differed from ours (three-pass electrofishing), lower flows allowed them to extend sites across the entire wetted width of the channel, similar to our channel-wide snorkeling sites. Comparing the results of the two studies would suggest that mean densities of coho fry in the Coquitlam River mainstem in 2006 and 2007 (13-14 fish/100 m 2 ) were nearly three times that in 1997 (5 fish/100 m 2, Riley et al. 1997). Steelhead fry densities were six- and two-fold higher in 2006 and 2007 (76 and 21 fish/100 m 2, respectively) compared to 1997 (12 fish/100 m 2 ). Steelhead parr densities were six- and thirteen-fold higher in 2006 and 2007 (3.2 and 6.7 fish/100 m 2, respectively) compared to 1997 (0.5 fish/100 m 2 ). Electrofishing removal estimates obtained in 1997 were likely biased-low as a result of low conductivity and difficulty in electrofishing in deeper mid-channel habitats (Riley et al. 1997), thus exaggerating the apparent increases in standing stock from 1997 to Nevertheless, these increases in fish density are too large to be explained by negative bias in electrofishing depletion estimates (Bohlin and Sundstrom 1977; Peterson et al. 2004). Under the Treatment 1 flow regime (dam releases from cms, Table 1.2), juvenile standing stocks in Coquitlam River in appear to have increased substantially compared to those in 1997 under the previous flow regime (dam releases from 0.06 to 0.5 cms). With respect to juvenile abundance in Coquitlam River relative to that in other streams, we relied on published values summarized by Satterthwaite (2002). Steelhead fry density in Coquitlam River in 2006 (76 fish/100 m 2 ) exceeded values for the majority of other streams, while density in 2007 (21 fish/100 m 2 ) was somewhat lower than that reported for other streams. For example, Hume and Parkinson (1987) considered 30 steelhead fry/100 m 2 to be about average in BC coastal streams. Ward and Slaney (1993) reported that steelhead fry densities in Keogh River averaged 34 fish/100 m 2 one month after emergence. Although brood escapement for the 2006 steelhead fry cohort was quite high (see section 3), and a linear relationship adult and juvenile steelhead abundance has been observed elsewhere (Ward and Slaney 1993), the high fry density observed in 2006 nevertheless suggests that under current conditions the Coquitlam River is capable of supporting relatively high fry densities compared to other streams. Steelhead parr densities in Coquitlam River in 2006 and 2007 (3.2 and 6.7 fish/100 m 2, respectively) fell roughly within the middle of a range of steelhead

85 84 parr densities summarized by Satterthwaite (2002), and were also comparable to that in tributaries of the Thompson River in the BC southern interior ( fish/100 m 2 ; Hagen et al. 2005). However, some of the streams included in Satterthwaite s (2002) review had steelhead parr densities that were considerably higher (up to 20 fish/100 m 2 ) than that observed in Coquitlam River. Coho densities in Coquitlam River in (13-14 fish/100 m 2 ) was much lower than the range of mean coho densities at annual index sites in 15 other Lower Mainland streams ( fish/100 m 2, respectively; DFO, data on file), although it should be said that these streams were considerably smaller, and were sampled at sites chosen to represent good coho habitat. It should also be noted that constructed off-channel habitat contributes about half of coho smolt production in Coquitlam River, and numbers of coho fry from off-channel areas were not included in our estimate for the mainstem. Comparison with other streams suggests the Coquitlam River may be a more productive stream for steelhead than coho, which is not surprising given its relatively high gradient and large substrate. 4.4 Recommendations for the 2007 survey Results from 2006 and 2007 indicate that juvenile salmonid abundance can be sampled effectively in all habitat types in Coquitlam River using night snorkeling counts, we recommend that juvenile surveys continue to be conducted annually, guided by the following general criteria. 1. As much as possible, the same 12 sites should be sampled each year. 2. Sampling sites should be 25 m in length, span, should span the entire wetted width of the stream, and where the stream is braided, all channels within the 25 m length of stream should be surveyed. 3. Mark-recapture experiments should by conducted each year at a subset of the sampling sites ( 4 sites each year) to develop a Coquitlam Riverspecific model of snorkeling efficiency. Across years, mark-recapture experiments should be equally distributed among the 12 permanent sampling sites. Mark-recapture methodology should follow that described in Hagen et al. (2005) and Decker and Hagen (2006). 4. Because the principle objective of the juvenile standing stock survey is to estimate juvenile abundance (as opposed to exploring fish abundance/habitat relationships), we recommend that available resources be concentrated on maximizing sampling effort and calibrating the sampling method rather than collecting extensive habitat data.

86 SMOLT PRODUCTION 5.1 Methods Coho and steelhead smolt enumeration In 2007, downstream migrating coho and steelhead smolts were captured at the outlets of three constructed off-channel sites (Or Creek, Archery Pond and Overland Channel; Figure 5.1), and at four RSTs in the Coquitlam River mainstem (Figure 5.1). A fourth off-channel site, Grant s Tomb, which was trapped in previous years, was dewatered during to allow for improvements to the Coquitlam Dam, and did not contribute to smolt production. We used total counts at full-span weirs to estimate smolt abundance at the offchannel sites, and mark-recapture data collected at rotary screw traps (RSTs) to estimate smolt numbers for the three mainstem reaches and for the entire Coquitlam River mainstem upstream of Port Coquitlam Location and description of downstream traps Because of the limited number of sites that possess adequate water depth and velocity and site security) for RST operation, RSTs were not installed at the downstream reach boundaries (Figure 5.1). Ideally, RST trapping would have conducted at the downstream end of reach 1 at Port Coquitlam (the upper limit of tidal influence) to properly estimate smolt populations for the entire Coquitlam River. Until 2005, our lowermost trapping site was located just downstream of the upper boundary of reach 2a, 5.1 km upstream of downstream boundary of reach 1 1 (Figure 5.1), as this was the lowermost site suitable for RST operation. In 2006, the lowermost trapping site was moved 600 m downstream because a high water event infilled the former trapping site and provided a new location downstream. The 2006 site was again used in This increased the length of the lowermost section, which includes most of reach 2b and the upper portion of reach 2a, and which we refer to in this report as reach 2, to 3.2 km in 2006 and 2007 compared to 2.6 km in previous years. To approximate smolt production for reach 1 and the portion of reach 2a downstream of the lowermost trapping site (4.5 km of habitat in 2007), we extrapolated smolt density estimates from the 3.2 km section described above as reach 2. Limitations and uncertainties of this approach are addressed in Section We refer to the 2.7 km long section between RST3 and RST4 as reach 3 (Figure 5.1), but it should be noted that this section also includes the upper 900 m portion of reach 2b. The fourth RST (RST4) was installed at the bridge to the RCMP Rifle Range, 100 m upstream of the downstream end of reach 4, and sampled a 1.6 km section of reach 4 directly below the Coquitlam Dam (Figure 5.1) 2. 1 The remaining portion of the Coquitlam River between reach 1 and the Fraser River is influenced by a combination of tidal and Fraser River influences (Riley et al. 1997). 2 Prior to 2002, a full-span downstream weir was used in place of an RST in reach 4 (see Decker and Lewis 2000).

87 86 N Grant's Tomb R4 RST4 Or Creek Ponds R3 Archery Pond RST3 Overland Channel RST2 ( ) R2B RST2 ( ) R2A R1 Figure 5.1 Map of the Coquitlam River showing constructed off-channel habitat sites, mainstem reach breaks and the locations of mainstem rotary screw traps (RSTs) in 2007 and years previous.

88 87 All three off-channel sites are located between RST2 and RST4 (Figure 5.1). Enumeration of smolts from these off-channel sites was necessary for two reasons: 1) to distinguish between smolt production in constructed off-channel habitat that is largely unaffected by flow releases from the dam and production in natural mainstem habitat that is directly affected by flow releases; and 2) to provide additional marked smolts to improve the precision of smolt abundance estimates for downstream mainstem reaches. In 2007, full-span downstream weirs (Conlin and Tutty 1979) were used to capture juveniles outmigrating from the three off-channel sites (see Decker 1998 for a detailed description of offchannel sites in the Coquitlam River and the design of downstream weirs used to capture fish emigrating from these sites). Two of the sites, Or Creek and Archery Pond, have single outlet channels. The third site, Overland Channel, consists of two ponds each with a separate outlet. We used two weirs at Overland Channel Downstream trap operation Similar to previous years, in 2007, two RSTs (RST2.1 and RST2.2; Figure 5.1) were operated in close proximity to one another at the reach 2 trapping site to increase total capture efficiency at this site. In this report we will refer to this trapping location as simply RST2. The off-channel weirs and the mainstem RSTs were operated continuously from mid-april until mid-june (Table 5.1). All juvenile fish captured at the weirs and RSTs were anesthetized in a bath of dilute clove oil dissolved in ethanol, identified to species and counted. All smolts were measured for fork length (nearest mm), and unmarked smolts were given a unique fin clip identifying capture period and location (see section ). Every effort was made to minimize the stress on fish during the sampling and marking process, and, once recovered, fish were immediately released. We assumed that all downstream migrating coho larger than 60 mm forklength were smolts. We assumed that all age-2 and older steelhead ( mm in length) were seaward migrating smolts, while smaller yearling fish would remain in the river for at least one more year (see section for a discussion of this assumption). We examined frequency histograms of steelhead fork length to verify that 120 mm was an appropriate length cut-off for age-1 and age-2 fish. We recorded daily catches of steelhead parr (< 120 mm) caught at each downstream trapping site, but, because there was no way of knowing what proportion of the total parr population these downstream migrants represented, we did not attempt to estimate parr populations by mark-recapture. Conversely, it was reasonable to assume that all smolts were downstream migrants Differential marking by period and initial capture location As in previous years, we estimated smolt abundance in mainstem reaches of the Coquitlam River using a stratified mark-recapture method (Arnason et al. 1996). Significant temporal variation in capture efficiency (% of marked smolts recovered) is common when mark-recapture methods are used to estimate the

89 88 abundance of a migrating population (Seber 1982), and stratifying marking by period allows for unbiased estimates when temporal variation in capture efficiency is expected. To provide distinct mark groups over time, all unmarked coho and steelhead smolts captured at the off-channel weirs and the upstream RSTs (RST3, RST4) were differentially batch-marked according to date of initial capture (six mark groups; Table 5.1). A unique mark consisted of a small clip at one of several fin locations. The duration of the marking period was determined with the objective of achieving a minimum recapture target of 40 coho smolts from each group at each RST (10 recaptures for steelhead smolts). We monitored daily catch totals to meet this target and relied on observations of migration patterns in previous years to plan strata duration. We also began new strata if discharge changed significantly, since capture efficiency can be affected by discharge. While almost all unmarked smolts originated from the mainstem, a large proportion of marked coho smolts originated from off-channel sites. This is of concern because previous work in the Coquitlam River has shown significant differences in capture efficiency for smolts originating from these two habitat types (Decker and Lewis 2000; Decker et al. 2003), suggesting that estimates based on combined marked populations could be biased. To address this, in addition to the mark given to identity capture period, smolts were given a second unique mark identifying their original capture location (see Table 5.1 and paragraph below). By separately analyzing marking and recovery data for these different mark groups, we were able to generate several independent estimates of the number of smolts passing the same RST. For example, independent estimates of steelhead smolt abundance at RST2 could be generated using four different mark groups (off-channel, RST2, RST3 and RST4). Stratification of marking by location was achieved by administering unique fin-clip marks to unmarked smolts captured at the off-channel weirs, RST2 (steelhead only), RST3, and RST4 (Table 5.1). Although RST2 was the lowermost trapping location (Figure 5.1), unmarked steelhead captured there were given a unique second mark so that they could be released upstream ( 1 km upstream) rather than downstream in order to increase the potential number of recaptures at RST2. For the uppermost RST site (RST4; Figure 5.1), marked populations of coho and steelhead were provided by releasing smolts 1 km upstream of the trapping site after being marked. Since the precision of a mark-recapture estimate improves with the number of smolts marked, it is advantageous to generate estimates based on pooled data for different mark groups. To decide which spatial mark groups could be included in the final mark-recapture dataset for a particular RST, we used the following rationale and procedures: 1. Although we were not able to test this (see section ), we assumed that capture efficiency for unmarked smolts from the mainstem would be

90 89 better approximated by observed capture efficiency for marked mainstem smolts than by capture efficiency for marked off-channel smolts. 2. Using Fisher s exact test, we then tested whether overall capture efficiency (pooled data for temporal mark groups) differed (P < 0.05) for marked smolts from the off-channel and mainstem areas. For example, capture efficiencies (CE) for off-channel and mainstem smolts at RST2 were computed as: 6 i 6 i R M off channel, i off channel, i and 6 i 6 i R M RST 2, RST 3, RST 4, i RST 2, RST 3, RST 4, i (5.1) where R off-channel,i = number of marked off-channel smolts from marking period i that were recaptured at RST2 M off-channel,i = number of off-channel smolts marked during marking period i R = number of marked mainstem smolts (all mainstem trapping locations summed) from marking period i that were recaptured at RST2 M RST 2, RST 3, RST 4, i = number of mainstem smolts that were marked during marking period i 3. If we failed to detect a difference in CE, all mark groups were included in the dataset used to compute the final mark-recapture estimate. On the other hand, if a difference was detected, the final dataset was limited to pooled data for the mainstem mark groups only. 4. Finally, we showed each independent estimate of abundance (and its 95% confidence interval) that could be generated for each species for each RST (Appendix 5.1). This was done to give the reader a sense of how much an estimate or its precision varied depending on which mark group or combination of groups was used Population estimates For the three off-channel sites where full-span weirs were operated, in the absence of evidence to the contrary, we assumed a CE of 100% for each weir, and used the total number of smolts captured to estimate smolt production. For mainstem reaches of the Coquitlam River, numbers of smolts passing each RST were estimated using a maximum likelihood (ML) model developed by Darroch (1961) and modified by Plante (1990) for stratified mark-recapture data.

91 90 In this study, smolts captured and marked at the weirs or upstream RSTs constituted the marking sample and smolts recovered at an RST represented the recovery sample. With stratified mark-recapture methodology, both the marking and recovery samples are stratified. All smolt population estimates and confidence intervals were computed using a software package that is available to the public (SPAS, http// ~popan/). A description of the ML estimator and the use of the SPAS software is provided by Arnason et al. (1996). In general, we delineated six marking and recovery periods (Table 5.1), although in some cases, it was necessary to pool strata to avoid small sample and numeric problems that may prevent the maximum likelihood iterations from converging. When pooling strata, we followed the recommendations of Arnason et al. (1996). If numbers of marked and recaptured smolts in the majority of strata were too low to use the stratified estimator, data from all marking and recovery periods were pooled and the standard pooled Petersen estimator for unstratified data was used (see Arnason et al and for a discussion of the problems associated with pooling sparse data). To estimate smolt abundance for a particular reach (N reach ), we computed an estimate for the RST at the downstream end of that reach, and then subtracted, from this estimate, the estimate for the next RST upstream 1 : N reach 2 = N RST2 - N RST3 (5.2) N reach 3 = N RST3 - N RST4 (5.3) N reach 4 = N RST4 (5.4) To compute 95% confidence intervals for N reach 2 and N reach 3, we summed variances for all RST or minnow-trapping mark-recapture estimates affecting N reach 2 or N reach 3. For example: ( N ) = ± 1. Var( N ) Var( N ) 95% CI reach 2 96 RST 2 + RST 3 ± (5.5) Since reach 4 is the uppermost reach, the variance of population estimates is not affected by the uncertainty of mark-recapture estimates for other trapping sites: ( N ) = 1. Var( N ) 95% CI reach 4 ± 96 RST 4 ± (5.6) 1 In computing smolt numbers for each reach, smolt numbers for off-channel sites with weirs are not subtracted because all fish from these sites were marked and excluded.

92 91 To estimate coho and steelhead smolt production for the Coquitlam River mainstem, we extrapolated estimated smolt densities in reach 2 to the 4.5 km of mainstem habitat between the lowermost RST (RST2) and the downstream boundary of reach 1, and added these values to population estimates for RST2: N mainstem = N RST2 + N reach km /Length reach 2 (5.7) The 95% confidence interval for the estimate of Coquitlam River mainstem smolt production would be: ( N ) = ± 1.96 Var( N ) Var( N ) 95% CI mainstem RST 2 + reach2 ± (5.8) In equation 5.8, the Var ( N reach2 ) term is added to reflect the uncertainty associated with extrapolating smolt densities in reach 2 in order to estimate smolt abundance for the 4.5 km of mainstem habitat between RST2 and the downstream boundary of reach 1 1. The estimate and 95% confidence interval for total smolt numbers for the Coquitlam River study area including the four off-channel sites 2 were computed as: N total = N RST2 + (N reach km / Length reach 2 ) + N Off-channel (5.9) ( N ) = ± 1.96 Var( N ) Var( N ) 95% CI total RST 2 + reach2 ± (5.10) Mark-recapture assumptions We evaluated the assumption of population closure subjectively by plotting a frequency histogram of daily smolt catches for each weir or RST and then comparing the numbers of smolts captured at the beginning and end of the trapping period to captures during the peak of the migration. Very low catches at the tails of the trapping period relative to catches during the peak were taken as an indication that most smolts emigrated during the trapping period. We assumed 100% mark retention and 0% marking-induced mortality based on two earlier studies using similar marking procedures (Decker 1998; Decker and Lewis 1999). With respect to the assumption of equal capture efficiency for marked 1 The validity of this approach is discussed in section This is an underestimate of total smolt numbers in the Coquitlam River since there are two major off-channel sites (Oxbow Pond, Grist Channel) and 5.8 km of marginal productivity mainstem habitat between the lower boundary of reach 1 and the Fraser River confluence (Figure 1).

93 92 and unmarked smolts, we assumed marking did not change CE at the RSTs, but we did not test this directly. To do so would require that there be more than one potential recapture event for individual fish with similar effort for each trapping period (Seber 1982). In our study, individual fish may be recaptured at more than one RST site, but trapping effort is not equal among sites because the efficiency of each RST depends on its location. The steps taken to address potential differences in CE between marked and unmarked smolts are described in section ). With respect to the assumptions of constant CE and proportions of marked to unmarked smolts over time, the use of a stratified markrecapture design minimizes or avoids violations of these two assumptions by stratifying both the marking and recovery periods. We limited the time period during which CE and the proportion of marked to unmarked smolts were assumed to be constant to less than 15 days for most strata (Table 5.1). We attempted to minimize variation in CE and proportions of marked to unmarked smolts within each stratum by timing the marking and recovery strata to reflect suspected changes in these parameters Chum and pink smolt enumeration Only chum salmon smolts were present in Coquitlam River in spring 2007, since fall returns of adult pink salmon to the Fraser River occur predominately in odd years Downstream trapping To estimate chum smolt outmigrant numbers for the Coquitlam River, we relied on methodology developed by Cope (2002) for the Alouette River. Two incline plane traps (IPTs) were operated in reach 2 at the same location as the RSTs (RST2 site; Figure 1). IPTs were used in place of RSTs to capture chum because the RSTs were fitted with large diameter screening ( 15 mm) to achieve capture efficiency for larger coho and steelhead smolts at higher flows, and this larger screening allowed most chum to escape Differential marking over time To generate temporally stratified mark-recapture estimates, a single day s catch of chum smolts were periodically marked and released 1 km upstream of the IPTs. This differs from the approach taken for coho and steelhead smolts in that marking is not continuous. We distributed marking events at least three days apart to allow for all marked smolts from one group to pass the IPTs before the next group was released (very few marked fish were captured beyond three days of the marking date). This provided temporally stratified data without the need for different marks. We mass-marked chum smolts by placing them using in a solution of Bismark brown Y, a vital stain (Deacon 1963), and water (1:100,000 concentration) for one hour. Adequate oxygen level within the solution was maintained using bottled oxygen and a flow meter. Smolts were held in a live box and released at dusk to reduce predation. Mortalities prior to release were noted and subtracted

94 93 from the count for each mark group. Mark loss was not assessed, but Deacon (1963) suggests that smolts marked with Bismark brown are readily identifiable for at least 5 days following staining. Daily captures of chum were individually sorted from other fry species (coho, Chinook and steelhead) and counted and inspected for marks. Table 5.1 Description of the stratification of fish marking by location and period for coho and steelhead smolts in the Coquitlam River in The start date for each temporal marking period at each RST trap site is also shown. RST installation dates correspond to the start dates for mark group 1. RST removal dates are also given. Downstream Mark stratification Mark stratification by period Traps RST trapping site by location mark 1 mark 2 mark 3 mark 4 mark 5 mark 6 remove Reach 2 RST 2.1 mark E 13-Apr 05-May 14-May 17-May 24-May 01-Jun 15-Jun Reach 2 RST 2.2 mark E 14-Apr 05-May 14-May 17-May 24-May 01-Jun 15-Jun Reach 3 RST mark D 13-Apr 05-May 14-May 17-May 24-May 01-Jun 15-Jun Reach 4 RST mark B 13-Apr 05-May 14-May 17-May 24-May 01-Jun 15-Jun Overland Ponds mark A 13-Apr 05-May 14-May 17-May 24-May 01-Jun 15-Jun Or Creek Ponds mark A 14-Apr 05-May 14-May 17-May 24-May 01-Jun 15-Jun Archery Pond mark A 13-Apr 05-May 14-May 17-May 24-May 01-Jun 15-Jun Population estimates The population estimate and 95% confidence interval for chum smolts passing the IPTs in reach 2 was computed using the same methodology as that for other species at the RSTs (ML estimator for stratified data; see section ). To estimate chum smolt production for the entire Coquitlam River study area, we extrapolated smolt density for the area (m 2 ) sampled above RST2 (including off-channel sites) to the area of the 4.5 km section of reach 2 downstream of RST2. The validity of this approach is discussed in section N Area Coquitlam River study area total = N IPT (5.11) AreaCoquitlam River upstream of RST Results AreaCoquitlam River area ± 95% CI ( N ) ( ) study total = ± CI N IPT (5.12) Area Coquitlam River upstream of RST Off-channel sites Daily catches of coho and steelhead smolts at the off-channel weirs at the beginning and end of the trapping period were very low compared to catches

95 94 during the peak of the migration (Figures 5.2 and 5.3). Therefore, we assumed that population closure was largely met, and that captures at the weirs accurately represented total smolt output. Observed mortality was < 1% for all target species at the off-channel weirs. No incidents of weir failure or fish leakage were apparent at the Archery Pond, Or Creek, or Overland Channel sites. In 2007 an aggregate total of 5,517 coho were captured at the downstream weirs as they outmigrated from Overland Channel, Or Creek and Archery Pond off-channel sites (Table 5.2). Mean weighted density of coho in the off-channel sites was 23 smolts/100 m 2. Of the off-channel sites, Overland Channel produced the highest number and density of outmigrant coho smolts (2,681 smolts and 60 smolts/100 m 2, respectively; Table 5.2). A total of only 53 steelhead smolts outmigrated from the three off-channel sites in 2007 (Table 5.2). Mean weighted density of steelhead smolts for the off-channel sites was 0.2 smolts/100 m Coquitlam River mainstem With the exception of a three-day period of high flows from April (peak flow = 28 cms), discharge in the Coquitlam River was relatively stable during the spring 2007 trapping period (4-10 m 3 /s, Figure 5.4), Overall, observed trap and handling mortality was < 1% for coho and steelhead captured in RSTs, and 8% for chum captured in IPTs. For coho and steelhead, daily catches at the beginning and end of the trapping period were very low compared to catches during the peak of the migration (Figures 5.2 and 5.3), suggesting that population closure was largely met. Considerable numbers of chum were captured at the very beginning of the trapping period (Figure 5.5), suggesting that trapping period likely did not encompass the entire migration period, which would result in an underestimate of total smolt output. Stratified mark-recapture data (catch tables) are not presented in this report, but are available for all species, mark groups and sites, along with a summary of which marking and recovery strata were pooled (if any) in generating population estimates, from Dave Hunter at BC Hydro (BC Hydro, data on file). Appendix 5.1 provides a summary of mark recapture statistics for each species and mark group and the estimates of the number of smolts passing each RST Coho Using recovery data for marked smolts that were captured at RST4 and released upstream (mean capture efficiency = 40%, Appendix 5.1), we obtained an estimate of 969 coho smolts (95% confidence interval: ± 102 smolts; Table 5.2) emigrating from reach 4.

96 A. Off-channel Apr 28-Apr 12-May 26-May 9-Jun Numbers per day B. RST 2.1 and Apr 28-Apr 12-May 26-May 9-Jun C. RST 3 14-Apr 28-Apr 12-May 26-May 9-Jun D. RST 4 14-Apr 28-Apr 12-May 26-May 9-Jun Date Figure 5.2 Daily catches of coho smolts at downstream weirs in three off-channel sites (pooled data) and at three rotary screw trapping locations in the Coquitlam River mainstem in See Table 5.1 for start and end dates for individual trapping sites.

97 96 15 A. Off-channel Apr 28-Apr 12-May 26-May 9-Jun Numbers per day B. RST 2.1 and 2.2 April 8 and Apr 28-Apr 12-May 26-May 9-Jun C. RST 3 14-Apr 28-Apr 12-May 26-May 9-Jun D. RST 4 14-Apr 28-Apr 12-May 26-May 9-Jun Date Figure 5.3 Daily catches of steelhead smolts at downstream weirs in three off-channel sites (pooled data) and at three rotary screw trapping locations in the Coquitlam River mainstem in See Table 5.1 for start and end dates for individual trapping sites.

98 Mean daily flow (cms) Apr 17-Apr 1-May 15-May 29-May 12-Jun 26-Jun Date Figure 5.4 Mean daily flows in Coquitlam River at Port Coquitlam during the smolt trapping period in (Water Survey of Canada, stn. 08MH141). At RST3, CE was similar for marked off-channel and mainstem coho smolts (CE for pooled strata was 28% for both mark groups, Table 5.3, Figure 5.6). Therefore, we combined data for the two mark groups to generate a population estimate of 1,086 smolts (± 237; Table 5.2) for reach 3. At RST2, we combined the mark-recapture data for mainstem and offchannel smolts since CE was similar for the two groups (mean CE was 38% for both mark groups; Table 5.3, Figure 5.6). The resultant estimate for reach 2 was 815 ± 278 coho smolts (Table 5.2). In 2007, based on combined data for the off-channel and mainstem mark groups, the estimated number of coho smolts outmigrating from the mainstem of the Coquitlam River was 4,016 ± 331 (9,533 ± 331 smolts including smolts from the three off-channel sites, Table 5.2). Average coho smolt density in the Coquitlam River mainstem was 1.8 smolts/100 m 2 (3.9 smolts/100 m 2 in the Coquitlam River including the off-channel sites, Table 5.2). Areal coho density was highest in reach 4 (5.0 smolts/100m 2 ; Table 5.2), followed by reach 3 (2.3 smolts/100m 2 ), and reach 2 (1.0 smolts /100 m 2 ). Precision ranged from ± 3% for the estimate for the entire study area including off-channel sites, to ± 34% for the smolt estimate for reach 2 (Table 5.2).

99 98 Table 5.2 Summary of estimated smolt numbers and densities by species in 2007 for three off-channel sites, reaches 2-4 of the Coquitlam River mainstem and the total Coquitlam River mainstem including and excluding the off-channel sites. Length Area Density Site (km) (m 2 ) N smolts CI (+/-) CI (%) (no./100m 2 ) (no./km) Coho Off-channel sites Grant's Tomb Or Creek - 13,336 2, Archery Pond - 5, Overland Channel - 4,500 2, Total - 23,636 5, Mainstem Reach 2, Coquitlam River , % Reach 3, Coquitlam River ,920 1, % Reach 4, Coquitlam River , % Total ,530 4, % Coquitlam R.incl. off-channel sites ,166 9, % Steelhead Off-channel sites Grant's Tomb Or Creek site - 13, Archery Pond - 5, Overland Channel - 4, Total - 23, Mainstem Reach 2, Coquitlam River ,778 1,251 1,099 88% Reach 3, Coquitlam River , % Reach 4, Coquitlam River , % Total ,530 4,374 1,443 33% Coquitlam R.incl. off-channel sites ,166 4,427 1,443 33% Chum Coquitlam R.incl. off-channel sites ,166 4,414, ,404 19% 1, ,894 Kokanee Number past RST 2 n/a n/a n/a n/a Number past RST 3 n/a n/a n/a n/a 1 Includes reaches 2-4 and an additional 4.5 km mainstem section downstream of the lowermost RST Steelhead Based on recovery data for marked smolts that were captured at RST4 and released upstream (mean CE = 19%, Appendix 5.1), the estimated the number of steelhead smolts for reach 4 was 824 (95% confidence interval: ± 235 smolts; Table 5.2). Because only 29 marked smolts were recaptured in total (Appendix 5.1), it was necessary to use the pooled Petersen estimator (all release and recovery strata pooled), and this may have resulted in a biased estimate because capture efficiency was highly variable over time (Figure 5.6).

100 99 Numbers per day Chum 4-Apr 18-Apr 2-May 16-May 30-May 13-Jun Kokanee 4-Apr 18-Apr 2-May 16-May 30-May 13-Jun Date Figure 5.5 Daily catches of chum smolts at two incline plane traps located at the RST2 trapping site in reach 2 and age-1 kokanee at RST2 in Coquitlam River in See Table 5.1 for start and end dates for downstream trapping. At RST3, CE was significantly higher for the off-channel mark group compared to the mainstem group (23% and 10%, respectively, P = 0.036; Table 5.3). As well, the numbers of recaptures for individual release strata were too low to produce an estimate using the Darroch ML estimator. Therefore, we used only the mainstem mark group and the pooled Petersen estimator (all release and recovery strata pooled) to obtain an estimate of 540 smolts in reach 3 (± 624 smolts, Table 5.2). This estimate may be biased given the high temporal variation in CE for mainstem steelhead smolts at RST3 (Figure 5.6). At RST2, pooled CE was higher for off-channel smolts compared to mainstem smolts (36% and 16%, respectively, Fisher s exact test, P = 0.002; Table 5.3), and we therefore used only the mainstem mark group to generate a population estimate of 1,251 steelhead smolts for reach 2 (± 1,099, Table 5.2).

101 100 RST capture efficiency (CE) RST capture efficiency (CE) 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Steelhead RST 2 off-channel RST 2 mainstem RST 3 off-channel RST 3 mainstem RST 4 mainstem 13-Apr 5-May 14-May 17-May 24-May 1-Jun Coho 13-Apr 5-May 14-May 17-May 24-May 1-Jun Figure 5.6 Estimated capture efficiencies (across six marking periods) at three rotary screw traps (RSTs) in the Coquitlam River for marked coho and steelhead smolts from offchannel (dotted lines) and mainstem (solid lines) habitats in Dates on the horizontal axis indicate the start point for each marking period.

102 101 Based on the mainstem mark group data, the estimated number of steelhead smolts outmigrating from the Coquitlam River mainstem was 4,374 ±1,443 (4,427 ± 1,443 smolts when off-channel sites were included, Table 5.2). Average steelhead density in the Coquitlam River mainstem was 2.0 smolts/100 m 2 (1.8 smolts/100 m 2 in the Coquitlam River including the off-channel sites, Table 5.2). For steelhead, both areal and linear smolt density was higher in reach 4 than in reaches 2 and 3 (Table 5.2). The precision of the abundance estimates ranged from ± 29% for reach 4, to ± 116% for reach 3 (Table 5.2). Table 5.3 Differences in capture efficiency (proportion of marked smolts that were recaptured) for coho and steelhead from off-channel sites and the Coquitlam River mainstem at three rotary screw traps (RSTs) sites in the Coquitlam River mainstem in Stratified marking periods were pooled prior to testing (see Equation 5.1). Equal capture efficiency for mark groups was tested using Fisher s exact test. P < 0.05 indicates a significant difference in capture efficiency. Capture efficiency Mainstem mark group Off-channel mark group Fisher's exact test (P) Species Recapture site Coho RST Coho RST Steelhead RST Steelhead RST Chum During 2007, the IPTs were operated from April 4 to May 30. Chum were batch-marked on 12 separate occasions. Capture efficiency for chum smolts at the IPTs ranged from 1% to 14% and averaged 8% across the 12 marking strata (Appendix 5.1). Due to moderate high flows ( 10 cms), the IPTS were not operated on two days during the trapping period (April 10 and May 21, Figure 5.4). Considerably higher flows (peak 28 cms) during a separate 3-day period (April 26-28) during the peak of the migration also prevented trap operation. To approximate the number of migrant chum for days with missing data, we first estimated the number of migrants for days when trapping occurred by factoring daily catches by appropriate CE values. We then used linear interpolation to estimate the number of migrants for missing days. An estimated 3.1 million chum (± 600,000; Appendix 5.1) migrated past the trapping site in reach 2. Approximated numbers of migrants for missing days represented 17% of this estimate. Extrapolation of this value to include the remaining 4.5 of the study area downstream yielded an estimate for the Coquitlam River (including off-channel sites) of 4.4 million chum (± 850,000, Table 5.2) or 1,800 chum/100 m 2 (370,000 smolts/km; Table 5.2).

103 Kokanee In total, 172 kokanee were captured at RST2 and RST3 in These fish ranged in size from 58 mm to 101 mm, and were likely all age-1 fish. Kokanee were not marked in 2007 since the mortality rate during mark-recapture in 2006 was considered excessive (25%, Decker et al. 2006). Instead, we assumed that mean CE for kokanee was the same as that for coho smolts (RST 2: 38%; RST3: 28%), given their roughly similar size range. Based on the pooled Petersen estimator, approximations of the number of kokanee migrating past RST3 and RST2 were 196 and 191, respectively Steelhead Numbers of Fish Steelhead smolt size We 20assumed all steelhead mm in fork length to be smolts. As in previous years, 120 mm corresponded to the minima between two defined modes representing 0 age-1 and age-2 and older juveniles, respectively (Figure 5.7). Fish larger than 230 mm had the general appearance of resident rainbow trout (i.e., cryptic colouring, heavily spotting) as opposed to smolts (bright silver), and some were sexually mature. Fork Length (mm) Figure 5.7 Length-frequency distribution for steelhead captured in the Coquitlam River in 2007 (data pooled for all trap sites). Arrows indicate fork length cut-offs for each ageclass. 5.3 Discussion Reliability of estimates and implications for the flow experiment Higgins et al. (2002) demonstrated that the statistical power to detect differences in fish production in the Coquitlam River under different flow regimes was strongly influenced by the precision of annual estimates of smolt abundance.

104 103 Specifically, they showed that power (β) decreases significantly over a range of increasing observation error (σ sm,o in their paper) for estimates of smolt abundance from about 0.1 to 0.5 σ sm,o (Figure 5, p. 18 in their paper). Expressed as a 95% confidence interval, values for σ sm,o of 0.1 to 0.5 are equivalent to levels of precision of ± 20% to ±110% of the estimate. For coho, during , the precision of the smolt abundance estimate for the Coquitlam River mainstem (95% CI = ± 7% to ± 14%) bettered the theoretical optimal value of σ sm,o 0.1 (± 20%). The exception was 2004 when precision was slightly less than optimal (± 25% or σ sm,o 0.13). For steelhead, precision of the mainstem estimate during approached the theoretical optimum, ranging from ± 15% to ± 34%. Assuming that coho and steelhead smolt populations can be estimated with similar precision in future years, the current smolt trapping program in Coquitlam River should be adequate to meet requirements for the proposed flow experiment. The length of the trapping period and the trap configurations and locations for coho and steelhead was appropriate in The 2007 study design should be repeated in In 2007, the estimated precision of the abundance estimates for chum smolts (± 19% Table 5.2) was at the theoretical optimum and similar to previous years (± 17% to ± 25% for chum during ). However, the estimate of 4.4 million chum for 2007 was likely biased-low as a result of three missed trapping days during the peak of the run (April 26-28). Chum numbers for these days (and two other missed days) were approximated by linear interpolation, but, given that flows during April (13-28 cms) were considerably higher than those on days previous and following (4-7 cms), trap capture efficiency would likely have been lower and the actual number of migrants higher than the values derived by linear interpolation. This bias may have been substantial considering that the number of smolts approximated for the five missing days represented 17% of the total estimate. The estimate was also bias-low as a result of chum migrating prior to beginning of the trapping period, given that significant numbers of smolts were already migrating during the first week of trapping (Figure 5.5). However, considering that estimated numbers of smolts in the first 10 days of trapping represented only about 6% of the total estimate, this failure to meet the assumption of population closure was probably a less important source of bias. During the last several years, we have found that using IPTS to capture smolts is very inefficient at flows greater than about 8-10 cms. These traps do not have self-cleaning drums like RSTS and are prone to debris-clogging at higher flows. While flows generally remain below 8-10 cms for most of the chum migration period, the high frequency of trap checking and cleaning at higher flows makes it difficult or impossible to operate the traps continuously or to complete mark-recapture experiments to estimate capture efficiency, as was the case in Moreover, base flows from Coquitlam Dam during Treatment 2 will result in an increased frequency of flows > 10 cms. To address this source of bias, if possible, we recommend replacing the IPTS in 2008 with two small-sized RSTs

105 104 (i.e., four foot drum) equipped with fine mesh screening. These would require less maintenance and provide more effective trapping at higher flows. As well, numbers of marked chum for each mark-recapture experiment should remain high in future years to compensate for relatively low capture efficiency (1%-14%), and 10 or more marking experiments should be conducted Assumptions of the study design In 2007, similar to previous years, the study design relied on extrapolation of smolt densities in reach 2 (reach 2b and a portion of reach 2a as defined by the COQWUP CC; Figure 5.1) to provide estimates of smolt numbers in the remaining portion of the Coquitlam River study area below RST2 (reaches 1 and 2a). This may have resulted in biased estimates of overall smolt numbers for the Coquitlam River depending on how actual smolt densities in the lower 4.5 km section compared to those in reach 2b. This potential bias could be substantial. For example, extrapolating relatively high steelhead smolt density in reach 2b in 2007 to the 4.5 km section downstream, resulted in an estimate of 3,975 steelhead smolts for the Coquitlam River mainstem (Table 5.2) based on 2,449 smolts passing RST2 (Appendix 5.1). This suggests that the unsampled lower 4.5 km section produced 40% of mainstem steelhead, despite relatively low densities of steelhead redds (Figure 3.2) and parr (Figure 4.3). In evaluating the effect of the flow experiment on fish production, this potential source of bias can be avoided by simply limiting analyses of smolt production or recruits per spawner to the portion of Coquitlam River between the dam and RST2. This section of Coquitlam River contains the majority of productive juvenile habitat in the Coquitlam River mainstem (Riley et al. 1997). Moreover, for all four species included in the monitoring program (coho, steelhead, chum, and pink), the majority of spawning occurs upstream of RST2 (e.g., coho: mean = 88%; steelhead: 87% in 2006; chum, mean = 55%; pink: 2003, 2005 mean = 69%). Separate escapement estimates are given in Table 6.1 for the section of Coquitlam River upstream of RST2 as well for the entire area included in surveys of adult salmon and steelhead redds. We assumed all two year and older steelhead ( mm in length) were smolts, yet a small proportion of smaller fish in this size range were likely steelhead parr that were dispersing to downstream habitats and would not have smolted until the following year at age-3 (Withers 1966). As well, some of the larger fish in this size range were likely small resident rainbow trout (at several trapping sites, we captured smolt-sized fish that were heavily spotted rather than silvery and exhibited signs of sexual maturity). Regardless, this was not likely an important source of bias. The vast majority of steelhead that were captured and recorded as smolts were silvery in appearance (e.g., >97% were silvery in appearance in 2002 and 2005 when physical characteristics were categorized for all steelhead captured). Moreover, the average fork length of steelhead smolts during study years from 1996 to 2007 varied from 154 mm to 171 mm, which is

106 105 in good agreement with mean length at ocean entry for steelhead stocks in the North Pacific (mean = 160 mm, CV = 10%-15%; Burgner et al. 1992). 5.4 Recommendations for smolt trapping in 2008 For the purpose of assessing smolt production and smolts/spawner in Coquitlam River under Treatments 1 and 2, we recommend limiting the study area to the portion of the mainstem between Coquitlam Dam and RST2 to avoid bias associated with extrapolating smolt densities upstream of RST2 to unsampled reaches below. The design of the adult salmon and steelhead redd surveys are such that it will be possible to obtain discrete estimates of spawner numbers upstream of RST2. For coho and steelhead, results to date suggest that the downstream trapping program in its current form is adequate for the purposes of generating sufficiently precise and reliable estimates of smolt abundance to meet CRMP objectives. For chum and pink smolts, we recommend replacing the IPTS in 2008 with two small-sized RSTs. As well, numbers of marked chum for each mark-recapture experiment should remain high in future years to compensate for relatively low capture efficiency (1%-14%), and 10 or more marking experiments should be conducted.

107 106 Appendix 5.1 Summary of estimated numbers of coho, steelhead, chum and kokanee passing three RST trapping locations (not reach estimates) in the Coquitlam River mainstem in 2007, with separate estimates for each mark group. Mark group indicates the location where fish were initially captured and marked. Also shown are numbers of marked (M) and recaptured (R) smolts, estimated capture efficiencies (R/M), 95% confidence intervals, and percent relative errors. Mark Capture Species Site group(s) M R U efficiency N smolts CI (+/-) CI (%) Coho RST 2 RST , , % RST 2 RST , , % RST 2 mainstem , , % RST 2 off-channel 5,418 2,089 3, , % RST 2 all 6,212 2,371 3, , % RST 3 mainstem , % RST 3 off-channel 2, , , % RST 3 all 3, , , % RST 4 mainstem % RST 4 off-channel RST 4 all Steelhead RST 2 RST , % RST 2 RST , % RST 2 RST ,742 2,226 47% RST 2 mainstem , % RST 2 off-channel , % RST 2 all , % RST 3 mainstem , % RST 3 off-channel % RST 3 all , % RST 4 mainstem % RST 4 off-channel RST 4 all Chum IPT@RST2 IPT@RST2 18,419 1, , ,099, ,470 19% Kokanee RST 2 all RST 3 all

108 ANNUAL TRENDS IN FISH PRODUCTION ( ) 6.1 Coho and steelhead During the number of coho smolts outmigrating from Coquitlam River mainstem ranged from 4,000 to 14,000 (mean = 8,043 smolts; Table 6.1). For mainstem and off-channel sites habitat combined, annual coho smolt production was less variable (9,500-19,400, Table 6.1) and, on average, nearly double that of mainstem production alone (15,100 smolts). In reach 4, where annual downstream trapping has occurred over a longer time period ( ), coho smolt numbers peaked in 2000, and declined in each year thereafter (Figure 6.1). Smolt numbers_ Coho Steelhead Figure 6.1 Annual numbers of coho and steelhead smolts in reach 4 of Coquitlam River. The three constructed off-channel habitats included in the study 1, which represent about 10% of available habitat in the Coquitlam River study area, supported from 26% to 71% of the overwintering coho smolt population, with the off-channel sites contributing a higher proportion during (mean = 60%) than in (mean = 32%). This trend was not associated with a higher proportion of spawning in restored off-channel sites for the cohorts compared to the cohorts; in recent years, the proportion of adult coho returning to Coquitlam River that spawned in constructed offchannel habitat has declined steadily from 53% in 2003 to 13% in 2006 (Table 2.5). During , the mean density of coho smolts in the mainstem portion of the study area ranged from 1.8 to 6.3 smolts/100m 2, which was several times lower than that in the off-channel sites (22.6 to 44.9 smolts/100m 2 ). While constructed off-channel habitat may represent relatively productive coho habitat in Coquitlam River, smolt densities in Coquitlam River off-channel sites were 1 There are seven major off-channel habitat sites in Coquitlam River, four in the smolt study area, including Grant s Tomb, which was dewatered in 2005 to facilitate repairs to Coquitlam Dam, and three downstream of the study area.

109 108 below average densities reported for constructed side-channels and ponds in other Pacific Northwest streams (67 and 69 smolts/100m 2, respectively; Koning and Keeley 1997). Although the accuracy of escapement estimates was highly uncertain due to a lack of information about observer efficiency and survey life, spawner survey data for suggest that, in most cases, recent coho escapement levels were, theoretically, more than adequate to seed available juvenile habitat. For coho, estimated spawner densities in ranged from 45 to 125 fish/km (Table 2.4), or 23 to 63 females/km, assuming a 1:1 sex ratio. The lowest estimate of 23 females/km in 2006 approaches the minimum threshold of 19 females/km necessary to achieve maximum coho smolt production in an average coastal stream (Bradford and Myers 2000). The high estimate of 63 females/km in the Coquitlam River in 2005 exceeds the threshold by a factor of three. That said, similar brood escapements in Coquitlam River during (1,500-1,600 spawners, Table 6.1) produced similar numbers of smolts in (14,100-19,000), whereas, the 2005 brood escapement, which was about half that in previous years (750 spawners) produced about 35% less smolts (9,500 smolts). This contradicts the assumption that escapement was adequate to fully seed juvenile habitat in every year. Smolt output in 2008 will be the product of an even lower escapement than 2005 (570 spawners) and will provide additional insight concerning the stock and recruitment relationship for coho under Treatment 1. It is important to note that only mainstem length was included in calculations of coho spawner density in Table 2.4. Coquitlam River contains a large amount of constructed off-channel habitat, which is used extensively by juvenile coho. If off-channel rearing habitat is also considered, it is possible that more than 19 females/km are required to seed juvenile habitat. Coho egg-to-smolt survival remained consistently low ( %) for the smolt years (Table 6.1), with the highest value associated with the lowest escapement (750 spawners in 2005). By comparison, the average eggto-smolt survival rate for coho populations in nine other Pacific coastal streams was considerably higher (1.5%, ±1 SD of 0.7%-3.0%, Bradford 1995). It should noted, however, that high uncertainty in coho escapement estimates (Table 2.4) directly affects egg-to-smolt survival estimates, thus, atypically low egg-to-smolt survival estimates for Coquitlam River coho may have be an artefact of biasedhigh observer efficiency or biased-low survey life estimates for adults. During the number of steelhead smolts outmigrating from Coquitlam River mainstem ranged from 4,400 to 7,700 (mean = 6,400 smolts; Table 6.1). Similar to the result for coho, 2007 produced the fewest number of steelhead smolts for the Coquitlam River as a whole. In reach 4 there was a positive trend in steelhead smolt numbers during (Figure 6.2), the opposite result to that for coho in recent years. Nearly all steelhead smolts ( 99%) originated from the mainstem as opposed to the three off-channel sites in the study area. Steelhead smolt densities in the mainstem averaged 2.9

110 109 smolts/100m 2 (range = smolts/100m 2 ), which exceeded the provincial steelhead biostandard of 2.0 smolts/100m 2 (Tautz et al. 1992). Mean steelhead parr densities in Coquitlam River in late summer (3.2 and 6.7 fish/100 m 2, respectively) were comparable to values reported in other studies (see Section 4.3.2). Late summer steelhead fry densities in Coquitlam River were higher than the average for other streams in 2007, and somewhat lower in 2007 (see Section 4.3.2). Table 6.1 Summary of all population estimates for all life stages and species in Coquitlam River, Survival estimates correspond to fry, parr or smolt year rather than brood year. Year Life stage Species Adult escapement chum 13,938 19,772 20,868 21,717 36,828 pink - 3,817-2,349 - coho 1,594 1,466 1, steelhead (female) steelhead (total) Fall standing stock coho 32,031 28,795 steelhead fry 169,250 47,081 steelhead parr 7,114 14,879 Smolt production chum (ms+off-chan) 2,070,000 1,490,000 1,310,000 4,800,000 4,410,000 pink (ms+off-chan) - 520, ,000 - coho (mainstem) 14,030 8,112 12,413 5,209 4,429 8,091 4,016 coho (ms+off-chan) 19,378 12,581 19,017 14,138 15,033 16,003 9,533 steelhead (ms) 6,304 4,799 6,926 7,561 7,677 7,378 4,374 Egg-to-fall fry survival 1 steelhead 10.5% 9.5% Egg-to-parr survival (1+) steelhead 0.8% 0.8% Egg-to-smolt survival 1 chum 9.3% 4.7% 3.9% 13.8% 7.5% pink % - 9.9% - coho 0.6% 0.7% 0.7% 0.9% 1 Assuming a 1:1 sex ratio for all species and average fecundity values of 3,200, 1,800, 3000, and 3,700 eggs/female for chum, pink, coho, and steelhead (Groot and Margolis 1991; Ward and Slaney 1993). For steelhead, estimated egg deposition ranged from 46,000 and 149,000 eggs/km during (Table 3.2). In the Keogh River 13,300 eggs/km was estimated as the minimum required for saturating juvenile rearing habitat for winter steelhead (derived from Ward and Slaney 1993). Even if a somewhat higher egg deposition is required for Coquitlam River due to a longer growing season and a younger mean smolt age (Chaput et al. 1998), the minimum level

111 110 of egg deposition required to saturate juvenile rearing habitat has likely been more than met for the smolt years. Based on redd survey and juvenile standing stock data, steelhead egg-to-fry survivals in 2006 and 2007 were 10.5% and 9.5%, respectively (Table 6.1). These values were relatively high compared to a long-term average of 6.5% for Keogh River steelhead (range = 1.8%-11.5%; Ward and Slaney 1993). Egg-to-parr survival was 0.8% for both the 2006 and 2007 age-1+ cohorts (Table 6.1), which was slightly higher than the average for two years data for the Keogh River (0.65%, derived from Ward and Slaney 1993). An estimate of the number age-3 smolts produced in 2008 from the 2005 brood escapement is necessary before the first estimate of egg-to-smolt survival for steelhead in Coquitlam River can be obtained. Following a period of relative stability during , the mean length of coho smolts and age-1 steelhead parr in the Coquitlam River mainstem (spring downstream migrants) increased significantly during , and then levelled off in (Figure 6.2). The relatively small size of migrant steelhead parr in 2007 in reaches 2 and 3 was an exception to this trend. It is possible that increased flow releases from Coquitlam Dam starting in 1997 with the implementation of the Treatment 1 flow regime improved growth rates, but coho smolt and steelhead parr length also increased post-1997 in off-channel sites (Figure 6.2), where flows were less affected by dam releases. The increased size of coho smolts and steelhead parr in recent years may also be related to increased foraging opportunities for the eggs, carcasses and fry of chum and pink salmon. Chum returns to Coquitlam River increased from a few hundred fish prior to 1997 to an average of 23,000 during (Table 6.1). In contrast to coho smolts and steelhead parr, the mean size of steelhead smolts has not increased in recent years. Steelhead smolts migrating from reach 4 were larger in compared to , and there was no consistent pattern in steelhead smolt size during in reaches 2 and 3 of the mainstem, or at the off-channel sites (Figure 6.2). Contrasting results for steelhead parr and smolts is not unexpected given that, for steelhead and Atlantic salmon (Salmo salar), age at smolting is primarily determined by body size, with higher growth rates resulting in younger smolts (Ward and Slaney 1993; Chaput et al. 1998). Steelhead smolts in the Coquitlam River are a mix of age-2 and age-3 fish, thus, rather than increasing mean size, increased growth rates may have shifted the age structure towards a higher proportion of younger and smaller age-2 smolts. Fertilization of the Keogh River resulted in a shift from age-3 to age-2 smolts and a substantial increase in overall smolt production (Ward and Slaney 1993). During scale samples were collected from steelhead to assess length-at-age and mean smolt age, but scale analysis has yet to be completed. There were some consistent among-reach differences in the density and mean length of coho and steelhead. During areal density of coho

112 flow increase Coho Off-channel sites Reach 4 Reach 2,3 Overland channel Steelhead smolt Off-channel sites Reach 4 Reach 2, Steelhead parr Year Off-channel sites Reach 4 Reach 2,3 Figure 6.2 Mean annual fork lengths for coho smolts and steelhead smolts and parr in different habitats in the Coquitlam River, Error bars represent ± 1 standard error.

113 112 smolts was generally greatest in reach 4, exhibiting a downstream decline from reach 4 to reach 2 (detailed summary tables for 2007 smolt trapping program are available upon request). Areal steelhead density was highest in reach 4 in all years except 2000, but this was largely due to greater wetted width in downstream reaches. In many cases steelhead population estimates for individual reaches were highly uncertain due to low numbers of marked and recovered fish, or, in the case of downstream reaches, compounding error, as estimates were derived from the difference in the number of smolts passing two RST locations (see Decker et al. 2006). 6.2 Chum and pink As for coho, escapement and egg-to-smolt survival estimates for chum and pink should be considered highly preliminary and will likely change as adult salmon observed efficiency and survey life data are collected in future years. Nevertheless, for the period of record, existing data suggest that adult returns of chum salmon to Coquitlam River have ranged from 14,000 to 37,000, while smolt production has ranged from 1.3 to 4.8 million (Table 6.1). During , chum egg-to-smolt survival ranged from 3.9% to 13.8.% (mean = 7.8%; Table 6.1). This was similar to the average egg-to-smolt survival rate reported by Bradford (1995) for nine chum populations (6.7% ±1 standard deviation 3.3%-13.5%). The lowest record of chum egg-to-smolt survival in Coquitlam River (3.9%) occurred in 2005 following a winter of frequent high flows (mean daily discharge at Port Coquitlam exceeded 60 cms on 12 days during the winter of , WSC, data on file). However, egg-to-smolt survival was two-fold higher in 2007 (7.5%), despite flows in excess of 60 cms on 24 days during the previous winter. Thirty-eight hundred and 2,300 pink salmon returned to Coquitlam River in 2003 and 2005, respectively, resulting in outmigrations of 520,000 and 210,000 smolts in 2004 and 2006 (Table 6.1). Egg-to-smolt survival was 15.1% in 2004 and 9.9% in These values were higher than the average egg-to-smolt survival rate of 7.4% (±1 standard deviation 3.2%-17.4%) reported for pink salmon populations in 18 streams (Bradford 1995).. Overall, chum and pink salmon returns to Coquitlam River were markedly improved in compared to previous years. Chum salmon escapement was not rigorously assessed until 2002, but qualitative surveys by DFO field staff over the last several decades suggest that total escapement was typically less than 1000 adults prior to the implementation of the Treatment 1 flow regime in Pink salmon were successfully reintroduced to Coquitlam River in 1995 following their extirpation in the 1960 s. Increased minimum flows in Coquitlam River beginning in 1997 likely improved migration and spawning conditions for chum and pink. Further increases in minimum flow under Treatment 2 may further improve access and migration, particularly for pink salmon, since this species makes extensive use of several off-channel sites in Coquitlam River that are difficult for adults to access at current base flows. Several high flow events

114 113 during (up to 200 cms at Port Coquitlam in October 2003 and November 2006) may also have improved spawning habitat for chum and pink by scouring large amounts of fine substrate that had accumulated in reaches 2a, 2b and 3 as a result of nearby gravel extraction, erosion of a large clay bank in Or Creek, and several years of relatively modest peak flows prior to As well, in the case of chum, escapement has increased recently in many lower Fraser streams as a result of reduced commercial exploitation (Godbout et al. 2004), and this has likely influenced returns to Coquitlam River as well.

115 REFERENCES Adkison, M.D. and Z. Su A comparison of salmon escapement estimates using a hierarchical Bayesian approach versus separate maximum likelihood estimation of each year s return. Can. J. Fish. Aquat. Sci. 58: Arnason, A.N., C.W. Kirby, C.J. Swartz, J.R, Irvine Computer analysis of data from stratified mark-recovery experiments for estimation of salmon escapements and other populations. Can. Tech. Rep. Fish. Aquat. Sci BC Hydro. 2003a. Addendum to the Coquitlam Dam Water Use Plan Consultative Report. Prepared for the BC Hydro Coquitlam Dam Water Use Plan project, Burnaby, BC. BC Hydro. 2003b. Lower Coquitlam River 2003 instream flow needs assessment interim report on transect data collection. Prepared for the BC Hydro Coquitlam Dam Water Use Plan project, Burnaby, BC. Bohlin, T., and B. Sundstrom Influence of unequal catchability on population estimates using the Lincoln Index and the depletion method applied to electro-fishing. Oikos 28: Bozek, M. A., F.J. Rahel Comparison of streamside visual counts to electrofishing estimates of Colorado River cutthroat trout fry and adults. N. Amer. J. Fish. Mange. 11: Bradford, M. J Comparative review of Pacific salmon survival rates. Can. J. Fish. Aquat. Sci. 52: Bradford, M. J., G. C. Taylor, et al Empirical review of coho salmon smolt abundance and the prediction of smolt production at the regional level. Trans. Amer. Fish. Soc. 126: Bradford, M. J., R. A. Myers, et al Reference points for coho salmon (Oncorhynchus kisutch) harvest rates and escapement goals based on freshwater production. Can. J. Fish. Aquat. Sci. 57: Burnham, K.P., and D.R. Anderson Model selection and multimodel inference, 2nd ed. Springer. 454 pp. Campbell, R. F. and J. H. Neuner Seasonal and diurnal shifts in habitat utilized by resident rainbow trout (Salmo gairderi) observed in western Washington Cascade Mountain streams. Pages in F. W. Olson, R. G. White, and R. H. Hamre, editors. Symposium on Small Hydropower and Fisheries. American Fisheries Society, Western Division and Engineering Section, Bethesda, Maryland.

116 115 Chaput, G., J. Allard, F. Caron, J. B. Dempson, C. C. Mullins, and M. F. O Connell River-specific target spawning requirements for Atlantic salmon (Salmo salar) based on a generalized smolt production model. Can. J. Fish. Aquat. Sci. 55: Chaput, G. and five co-authors River-specific target spawning requirements for Atlantic salmon (Salmo salar) based on a generalized smolt production model. Can. J. Fish. Aquat. Sci. 55: Conlin, K. and B.D. Tutty Juvenile field trapping manual. Fish. and Mar. Serv. Man. Rep. No Cope, S Alouette River salmonid smolt migration enumeration. Unpublished report prepared for Alouette River Management Committee and BC Hydro by Westslope Fisheries, Cranbrook BC. Darroch, J.N The two-sample capture-recapture census when tagging and sampling are stratified. Biometrika 48: Decker, S Influence of off-channel habitat restoration and other enhancement on the abundance and distribution of salmonids in the Coquitlam River. Unpublished report prepared for B.C. Hydro Power Facilities, Burnaby, B.C. 35 p. Decker, S., and G. Lewis Response of salmonids to off-channel habitat restoration and increased minimum flows in the Coquitlam River. Unpublished report prepared for B.C. Hydro Power Facilities, Burnaby, B.C., December p. Decker, S., and G. Lewis Fish Response to Increased Minimum Flows and Habitat Restoration in the Coquitlam River: a 5 Year Review. Unpublished report prepared for B.C. Hydro Power Facilities, Burnaby, B.C., December p. Decker, S., J. Schick, and G. Lewis results for the Coquitlam River salmonid smolt enumeration program. Unpublished report prepared for B.C. Hydro Power Facilities, Burnaby, B.C., April p. Decker, S., J. Schick, and G. Lewis Coquitlam River salmonid smolt enumeration program: 2006 results. Unpublished report prepared for B.C. Hydro Power Facilities, Burnaby, B.C., January p. Decker. A.S., and M. Foy The contribution of restored off-channel habitat to smolt production in the Coquitlam River. British Columbia Ministry of Environment, Watershed Restoration Technical Bulletin 4(3):22-24.

117 116 Decker, A.S., and J. Hagen Juvenile stock assessment of Gerrard rainbow trout in the Lardeau/Duncan system: 1st year progress report, Prepared for Habitat Conservation Trust Fund, Victoria, British Columbia, September p. Decker, A.S., and J. Hagen The distribution and abundance of juvenile Chinook salmon in the lower Thompson River basin in relation to spawner abundance and habitat characteristics. Unpublished report prepared for the Pacific Salmon Commission, June pages. Decker, S., J. MacNair, G. Lewis, and J. Korman Coquitlam River Fish Monitoring Program: Results. Unpublished report prepared for B.C. Hydro Power Facilities, Burnaby, B.C., June p. Efron, B. and R. Tibshirani An introduction to the bootstrap. New York, Chapman and Hall. English, K.K., Bocking, R.C., and Irvine, J.R A robust procedure for estimating salmon escapement based on the area-under-the-curve method. Can. J. Fish. Aquat. Sci. 59: Freymond, B and S. Foley Wild steelhead: Spawning escapement estimates for Bolt Case Area Rivers Washington State Game Department, Fisheries Management Division. Report No pp. Fukushima. M. and W. Smoker. Determinants of stream life, spawning efficiency, and spawning habitat in pink salmon in the Auke Lake system, Alaska. Can. J. Fish. Aquat. Sci. 54: Gallagher, S.P., and C.M. Gallagher Discrimination of Chinook salmon, coho salmon, and steelhead redds and evaluation of the use of redd data for estimating escapement in several unregulated streams in northern California. N. Amer. J. Fish. Mange. 25: Gelman, A., Carlin, J.B. Stern, H.S., and D.B. Rubin Bayesian data analysis. Chapman and Hall/CRC. 668 pp. Godbout, L., J.R. Irvine, D. Bailey, P. Van Will, and C. McConnell Stock status of wild chum salmon (Oncorhynchus keta Walbaum) Returning to British Columbia s central coast and Johnstone and Georgia Straits (excluding the Fraser River). Canadian Science Advisory Secretariat Research Document 2004/07. Griffith, J. S Estimation of the age-frequency distribution of streamdwelling trout by underwater observation. Prog. Fish. Cult. 43:51-52.

118 117 Groot, C., and L. Margolis Pacific salmon life histories. UBC Press, Vancouver BC. Hagen, J., A. S. Decker, and R. G. Bison Calibrated rapid assessment methods for estimating total steelhead parr abundance in the Thompson River system, British Columbia interim report. Consultant Report Prepared for BC Ministry of Water, Land, and Air Protection, Kamloops BC. Hetrick, N. J., and M. J. Nemeth Survey of coho salmon runs on the Pacific Coast of the Alaska Peninsula and Becharof National Wildlife Refuges, 1994, with estimates of escapement for two small streams in 1995 and U. S. Fish and Wildlife Service, King Salmon Fish and Wildlife Field Office, Alaska Fisheries Technical Report 63, King Salmon, Alaska. Higgins, P.S., J. Korman, and M.J. Bradford Statistical power of monitoring inferences derived from experimental flow comparisons planned for Coquitlam-Buntzen Water Use Plan. B.C. Hydro internal report. Burnaby, B.C. 20 p. Hillborn, R., B.G. Bue, and S. Sharr Estimating spawning escapements from periodic counts: a comparison of methods. Can. J. Fish. Aquat. Sci. 56: Hillman, T. W., J. W. Mullan, and J. S. Griffith Accuracy of underwater counts of juvenile chinook salmon, coho salmon, and steelhead. N. Amer. J. Fish. Manage. 12: Hume, J.M.B., and E.A. Parkinson Effect of density on the survival, growth, and dispersal of steelhead trout fry (Salmo gairdneri). Can. J. Fish. Aquat. Sci. 44: Irvine, J.R., R.C. Bocking, K.K. English, and M. Labelle Can. J. Fish. Aquat. Sci. 49: Koski, K.V The survival and fitness of two stocks of chum salmon from egg deposition to emergence in a controlled stream environment at Big Beef Creek. Doctoral Dissertation, University of Washington, Washington. Jacobs S., J. Firman, G. Susac, D. Stewart and J. Weybright Status of Oregon coastal stocks of anadromous salmonids, and ; Monitoring Program Report Number OPSW-ODFW , Oregon Department of Fish and Wildlife, Portland, Oregon. Keeley, E. R. and C. J. Walters The British Columbia Watershed Restoration Program: summary of the experimental design, monitoring and

119 118 restoration techniques workshop. Vancouver, B.C., B.C. Ministry of Environment, Lands and Parks. Kintama Research Corporation, Assessment of Coquitlam Reservoir salmon migration pathways. Project #05.Co.02, final report, November 14, Prepared for BC Hydro Fish & Wildlife Bridge Coastal Restoration Program, Burnaby, BC. Korman, J., R.N.M. Ahrens, P.S. Higgins, C.J. Walters Effects of observer efficiency, arrival timing, and survey life on estimates of escapement for steelhead trout (Oncorhynchus mykiss) derived from repeat mark and recapture experiments. Can. J. Fish. Aqaut. Sci. 59: Korman, J, Melville, C.C., and P.S. Higgins Integrating multiple sources of data on migratory timing and catchability to estimate escapement for steelhead trout (Oncorhynchus mykiss). Submitted to CJFAS, Nov Koning, C.W., and E.R. Keeley Salmonid biostandards for estimating production benefits of fish habitat rehabilitation techniques. In P.A. Slaney and D. Zaldokas. [eds.] Fish habitat rehabilitation procedures. Watershed Restoration Technical Circular No. 9. Krebs, C.J Ecological Methodology. Benjamin/Cummings, Menlo Park. 620 pp. Leake, A.C Coquitlam-Buntzen Water Use Plan Terms of Reference: 7COQMON#7 - Lower Coquitlam River Fish Productivity Index. BC Hydro, Water Use Plan Monitoring Report, November 28, Leland, R.F Anadromous game fish investigations in Washington, Progress Report, July 1, 1995-June 30, Washington Department of Fish and Wildlife Fish Management Program. Olympia, WA. pp Macnair, J Coquitlam River water use plan: 2003 chum fry enumeration. Unpublished report prepared for BC Generation Sustainability and Aboriginal Relations, Burnaby BC by Living Resources Environmental Services. Macnair, J Coquitlam River adult salmon escapement survey. Prepared for BC Generation Sustainability and Aboriginal Relations, Burnaby BC by Living Resources Environmental Services, 21 p. Macnair, J Coquitlam River adult salmon escapement survey. Prepared for BC Generation Sustainability and Aboriginal Relations, Burnaby BC by Living Resources Environmental Services, 29 p.

120 119 Macnair, J Progress Report: 2005 Coquitlam River adult salmon escapement survey. Prepared for BC Generation Sustainability and Aboriginal Relations, Burnaby BC by Living Resources Environmental Services, pp. 36 Manske, M, and C.J. Schwarz Estimates of stream residence time and escapement based on capture-recapture data. Can. J. Fish. Aquat. Sci. 57: Myers, J, Historical population structure of Willamette and lower Columbia River basin Pacific salmonids. National Marine Fisheries Service, Washington Dept. of Fish and Wildlife.188 p. Northwest Hydraulic Consultants Ltd Coquitlam River channel morphology and substrate condition study. Unpublished report prepared for BC Hydro and Power Authority, August pages + appendices. Otis, D. L., K. P. Burnham, G. C. White, and D. R. Anderson Statistical inference from capture data on closed animal populations. Wild. Monogr. 62: Otter Research Ltd An introduction to AD Model Builder version for use in nonlinear modeling and statistics. 194 pp. Report available from otterrsch.com. Perrin C.J., J.R. Irvine A review of survey life estimates as they apply to the area-under-the-curve method for estimating the spawning escapement of Pacific salmon. Can. Tech. Rep. Of Fish and Aquat. Sci. No Peterson, J.T., R.F. Thurow, and J.W. Guzevich An evaluation of multipass electrofishing for estimating the abundance of stream-dwelling salmonids. Trans. Amer. Fish. Soc. 113: Peterman, R. M Statistical power analysis can improve fisheries research and management. Can. J. Fish. Aquat. Sci. 47:2-15. Plante, N Estimation de la taillee d une population animale a l aide d un modele de capture-recapture avec stratification. M.Sc. thesis, Universite Laval, Quebec. Ptolemy, R.A. Trial electrofishing and training using shore-based units; basic findings and observations for Coquitlam River sampling of riffle/rapid habitats during September 10-11, Management brief prepared for the Coquitlam-Buntzen Water Use Plan Consultative Committee, November 21, (included as Appendix 6.1 in this document).

121 120 Riley, S.C., J. Korman, S. Decker Coquitlam River salmonid stock assessment. Unpublished draft report prepared for B.C. Hydro, Burnaby, B.C. 20 p. Salo, E.O Life history of chum salmon, Oncorhynchus keta. In Groot, C., and L. Margolis (eds.), Pacific salmon life histories, p Univ. B. C. Press, Vancouver, B. C., Canada. Satterthwaite, T.D Population health goals and assessment methods for steelhead in the Klamath mountains province. Monitoring Program Report Number OPSW-ODFW , Oregon Department of Fish and Wildlife, Portland, OR. Seber, G.A.F The estimation of animal abundance. 2nd ed. Griffin, London. Smoker, W.W., A.J. Gharrett, M.S. Stekoll, and S.G. Taylor Genetic variation of fecundity and egg size in anadromous pink salmon Oncorhynchus gorbuscha Alaska Fishery Research Bulletin 7: Su, Z., Adkison, M.D., and Van Alen, B.W A hierarchical Bayesian model for estimating historical salmon escapement and escapement timing. Can. J. Fish. Aquat. Sci. 58: Tautz, A.F., B.R. Ward, and R. A. Ptolemy Steelhead trout productivity and stream carrying capacity. PSARC Working Paper S Thurow, R. F., J. Peterson, J.W. Guzevich Utility and validation of day and night snorkel counts for estimating bull trout abundance in first- to third-order streams. N. Amer. J. Fish. Mange. 26: Ward, B. R. and P. A. Slaney Egg-to-smolt survival and fry-to-smolt density dependence of Keogh River steelhead trout. Can. Spec. Publ. Fish. Aquat. Sci. 118.

122 121 Appendix 6.1 Supplemental Ministry of Environment document summarizing results for a separate electrofishing survey conducted during September ******************************************************************************************** Report to Coquitlam River WUP Files November 21, 2007 Author: Ronald A. Ptolemy, R.P.Bio. Topic: Trial electrofishing and training using shore-based units; basic findings and observations for Coquitlam River sampling of riffle/rapid habitats during September 10-11, 2007 Goals: 1. Falsify the notion that electrofishing is ineffectual for accurately quantifying juvenile steelhead abundance due to low conductivity. 2. Train two WUP survey contract crews in the large river electrofishing protocol and techniques used by Ministry of Environment (Fisheries). 3. Provide a more recent stock assessment of fish abundance in fast-water mesohabitats and contrast this to earlier survey results. Stream Environment and Sampling Context: Stream conductivity ranged from 10.6 µs/cm at River km (Coho Creek) to 40.3 µs/cm at Railway Bridge. This is a normal state for regulated summer baseflows and the water quality gradient has been previously reported. Stream flow at PoCo WSC gauge on September 12, 2007 was 1115 L/s at 1000 hr. Stream stage was 7.70 m and the WSC flow estimate was 1600 L/s. Flows in percent mean annual discharge (mad) were about 4%mad. Typically, flows below 10%mad are classified as severe. This maybe an artefact of a confined channel with highly reduced stream width; observed width at Site 1 was 15 m while the predicted width is about double at 30 m. Riffles would be expected to be impacted (reduced in width and depth) at this stage. However riffle habitats were judged to be healthy at this flow.

123 122 Flows immediately prior to the survey were higher (~2000 L/s) and had dropped to flows near 1100 L/s. See daily flow plot. Winter in the lower Coquitlam River (as elsewhere in southwest BC) was fairly severe with several floods to about 200 cms. This may have a negative influence on steelhead parr numbers should the fry and parr be vulnerable to scour/displacement mortality. Yearling parr captured in 2007 are 2006 brood year fish (fry in 2006). Early September 2007 summer baseflows. Winter flood events during in the lower Coquitlam River.