Duncan Dam Bull Trout Passage Monitoring

Duncan Dam Project Water Use Plan Duncan Dam Bull Trout Passage Monitoring Reference: DDMMON-6 An assessment of the effect of the weir on the proportion of small bull trout accessing the flip bucket Study Period: 1995 to 2008 Joseph L. Thorley. Poisson Consulting Ltd. 4216 Shasheen Road, Nelson, BC, V1L 6X1. March 10, 2009

Introduction In 1967, the construction of the Duncan Dam created a barrier that prevented fluvial and adfluvial bull trout (Salvelinus confluentus) from returning to their natal tributaries to spawn. Fortunately, the dam operator at that time devised a method of allowing bull trout to pass through the dam. However, the possibility remains that the dam is a selective barrier favouring the passage of larger bull trout over smaller individuals. In the long term such a selective force might lead to life history trait evolution (Haugen et al. 2008). The transfer of bull trout through the dam is described in O Brien (1999). In short, in order to be transferred through the dam, a fish must first ascend a waterfall cascading over the edge of the flip bucket a structure designed to direct discharge from the lower level operating gates away from the base of the dam. The height of the waterfall depends on the level of the Duncan River but can be 1.5 to 2 m (Hagen 2003). In 1994, following concerns that smaller bull trout were generally less successful at jumping into the flip bucket, a two stage fish ladder (the weir) was installed below Low Level Outlet Gate #2 (LLOG2). The weir was also used in at least 6 of the 14 subsequent years. A primary objective of the Duncan Dam Water Use Plan Consultative Committee (DDM WUP CC) report is to maximize fish abundance and diversity in the Duncan River and Duncan Reservoir within the operating potential of Duncan Dam (BC Hydro 2005). An associated subobjective is to minimize any fish passage issues in the mainstem. Based on this sub objective the Lower Duncan River Bull Trout Passage Monitoring Program (DDMMON#6) was developed (BC Hydro 2008). DDMMON#6 is a 10 year monitoring program that is aiming to test the following two hypotheses: H01: The weir as currently operated does not increase the size range or number of adult bull trout accessing the flip bucket at Duncan Dam. H02: There are no feasible alternative designs or operations that would improve operational flexibility or weir performance. Refuting the first hypothesis will result in a program shift to addressing the second hypothesis (BC Hydro 2008). This memorandum report represents a statistical test of the first hypothesis. More specifically it models the extent to which the presence of the weir increases the ability of the smaller bull trout to jump into the flip bucket relative to their larger conspecifics. Literature Review As bull trout ascend steep creeks and streams to spawn (McPhail and Baxter 1996; McPhail 2007) they might be expected to have prodigious swimming and jumping capabilities. However, the only swimming performance data on bull trout concern the prolonged (critical) swimming speeds of individuals with a fork length of less than 40 cm (Mesa et al. 2004). The bull trout transferred above the Duncan Dam are often greater than 60 cm and sometimes Page 1

greater than 90 cm in fork length and the parameter of interest at Duncan Dam is the jump height as opposed to the prolonged swimming speed. A conclusion of the Mesa et al. (2004) study was that the estimated critical swimming speeds of bull trout with a fork length of between 30 cm and 40 cm compare favorably with those of rainbow trout (Oncorhynchus mykiss). One approach might therefore be to use the jumping capabilities of rainbow trout to predict whether smaller bull trout might be less successful at gaining access to the flip bucket. However, not only is using data from one species to predict the capabilities of another considered ill advised (Mesa et al. 2004) but performance depends on numerous other factors including life history stage, plunge pool depth (Stuart 1962), water temperature, condition factor (Kondratieff and Myrick 2006) and motivation (Binder and Stevens 2004). Consequently, any inference about the likely benefit of the weir for smaller bull trout would be highly uncertain at best. Methods Given the complexities of predicting the jumping behaviour of bull trout, the most reliable method for determining whether or not the weir downstream of LLOG2 at Duncan Dam is a size selective barrier for bull trout is to analyze the historical length frequency data. A previous unpublished analysis of the data by BC Hydro concluded that presence of the weir has no statistical effect on the size of the bull trout entering the flip bucket. However, the response variable in the preliminary analysis was the mean size of the bull trout enumerated in the flip bucket. The response variable in the current analysis is the proportion of bull trout below a size threshold which is a potentially more sensitive metric of size selectively. In addition, the current analysis predicts the effect of the weir on the proportion of small bull trout whilst taking into account the long term trend, the seasonal pattern, and the non independence (pseudo replication) associated with multiple estimates within the same year. The size threshold below which the presence of the weir might be differentially facilitating bull trout access was determined by examining the length frequencies plots with and without the weir (Figures 1 and 2). Based on visual examination, the relative frequency of bull trout less than 65 cm appeared to be higher with the weir. Consequently, 65 cm was considered the threshold below which bull trout were considered small. It is worth noting that in most other systems, a 64 cm bull trout would be considered a large fish. Due to its abundant kokanee (Oncorhynchus nerka) population and large size, Kootenay Lake produces exceptionally large bull trout. The proportion of bull trout smaller than the threshold of 65 cm was calculated for each individual transfer event for which bull trout in the flip bucket were measured and data were available. The number of measured bull trout is generally less than the number of bull trout enumerated in the flip bucket. This is because in order to be measured the fish must be caught using a seine net but typically some fish evade capture. The analysis assumes that the sample of measured fish is effectively random with respect to size. Fish that were recorded as coming from any location other than the flip bucket (i.e. fish weir) were excluded from the analysis. Page 2

Without Weir Number of Bull Trout 0 50 100 150 200 40 50 60 70 80 90 100 Length (cm) Figure 1. The length frequency histogram for bull trout measured in the flip bucket without the weir. With Weir Number of Bull Trout 0 100 200 300 400 40 50 60 70 80 90 100 Length (cm) Figure 2. The length frequency histogram for bull trout measured in the flip bucket with the weir. Page 3

Weir was included in the analysis as a binary linear explanatory variable, i.e., present or absent (Dobson and Barnett 2008). The jump height, i.e., the distance fish would have to jump to get into the flip bucket without the presence of the weir, is measured by a Dam Operator positioned on the edge of the flip bucket. However, to date, this measurement has only been recorded on 11 occasions. Another data source for tail water elevations is approximated by the river height on a staff gauge in the tail water area adjacent the discharge channel (T. Oussoren pers. comm.). However, the gauge height has only been recorded intermittently from December 2006 onwards as a component of the Dam Operators weekly inspections. Consequently it was necessary to examine the other daily discharge data for the Duncan River to determine if they could be used as a proxy. A scatterplot matrix of the gauge height (Gauge), the total daily discharge from Duncan Dam (NPR), the total daily discharge from the Lower Level Outlet Gates (LLOG) and the total daily discharge 1 km below the confluence with the Lardeau (Confluence) is plotted in Figure 3. Examination of the scatterplot matrix indicates that although all four measures are positively correlated, the discharge below the confluence is the best predictor of the river height in the tail water. The explanation for this finding is that during freshet, the high Lardeau River flows and the typically low Duncan Dam discharge result in a backwatering effect to the base of the dam (T. Oussoren pers. comm.). 0 100 300 0 200 400 600 Gauge 543.0 544.0 0 100 300 NPR LLOG 0 100 200 300 0 200 400 600 Confluence 543.0 544.0 0 100 200 300 Figure 3. A scatterplot matrix of the gauge height (Gauge), the total daily discharge from Duncan Dam (NRP), the total daily discharge from the Lower Level Outlet Gates (LLOG) and the total daily discharge 1 km below the confluence with the Lardeau (Confluence). Page 4

Generally, transfer events occurred at biweekly intervals during which time bull trout are able to jump into and out of the flip bucket. In the absence of any information on bull trout residence time in the flip bucket the discharge associated with each transfer event was assumed to be the mean discharge 1 km below the confluence with the Lardeau during the week prior to the transfer (T. Oussoren pers. comm.). As well as the discharge, the year and day of the year of the transfer were included in the analysis to control for any long term trends or seasonal patterns in the proportion of small bull trout, respectively. The input data are tabulated in Appendix A. The response was the proportion of small bull trout which was determined by classifying each fish as either less than the threshold of 65 cm in length or greater than or equal to the threshold. As the response was a probability it was modelled using a binomial (generalized) model with the logistic link function (Collett 2003). To account for the marginal overdispersion in the data, a quasi likelihood approach was adopted (Collett 2003). The year, day of the year and discharge in contrast were all modelled as additive predictors using thin plate regression splines to capture any non linearity in the long term trends, seasonal patterns or relationship with discharge (Wood 2006). There were often multiple transfers in the same year. As transfers within the same year were strongly correlated, the non independence due to repeated measures within years was controlled for by including year as a random effect, i.e., a mixed model was used (Pinheiro and Bates 2000). As the model included a random effect, it was fitted by maximizing the restricted log likelihood (Pinheiro and Bates 2000). Following the general recommendations of Bradford et al. (2005), the inferred influence of the weir on the passage of small versus large bull trout was presented as an effect size with confidence intervals. The size of the effect of the weir was expressed in terms of the percent change in the number of small (less than 65 cm) bull trout in the flip bucket that would be required to produce the change in the response if the number of large bull trout remained constant. The effect size was plotted with both 95% and 80% confidence intervals as the former is standard but the latter may be more useful for management purposes. All analyses were performed using R 2.8.1 (R Development Core Team 2008). Results The effect of weir on the proportion of small bull trout in the flip bucket was significant at the 2% level (Table 1). Partial model output is presented in Appendix B. When expressed in terms of the effect size, the model estimated that the presence of the weir increased the number of small trout in the flip bucket by 71% although the actual value could lie between 6% and 174% with 95% confidence and between 28% and 128% with 80% confidence (Figure 4). Page 5

Table 1. The significance of the estimated relationships between the explanatory variables in the final model and the proportion of small (<65 cm) bull trout in the flip bucket. Variable P value Weir <0.02 Year <0.0001 Day of the Year <<0.0001 Discharge <0.01 Effect (%) -100-50 0 50 100 150 200 250 95% 80% Figure 4. The estimated effect of the weir expressed as the percent change in the number of small (<65 cm) bull trout in the flip bucket with 95% and 80% confidence intervals. Year, day of the year and discharge were also statistically significant at the 1% level or less (Table 1). More specifically, the proportion of small bull char in the flip bucket was suppressed between the end of the 1990s and the beginning of the current millennium (Figure 5) and increased strongly with day of the year (Figure 6). Contrary to expectations the proportion of small bull trout decreased with increasing discharge (Figure 7). The plotted points represent the within year residual variation, i.e., the remaining variation after controlling for the effects of the other explanatory variables and between year variation. Page 6

Proportion Small (< 65 cm) BT 0.0 0.2 0.4 0.6 0.8 1.0 With Weir Without Weir 1996 1998 2000 2002 2004 2006 2008 Year Figure 5. The estimated relationship between the proportion of small (<65 cm) bull trout in the flip bucket and year with and without the weir for the 200 th day of the year and discharge of 200 cms. The points represent the within year residual variation. Proportion Small (< 65 cm) Bull Trout 0.0 0.2 0.4 0.6 0.8 1.0 With Weir Without Weir Jun Jul Aug Sep Month Figure 6. The estimated relationship between the proportion of small (<65 cm) bull trout in the flip bucket and day of the year (plotted by month) with and without the weir for the year 2000 and a discharge of 200 cms. The points represent the within year residual variation. Page 7

Proportion Small (< 65 cm) Bull Trout 0.0 0.2 0.4 0.6 0.8 1.0 With Weir Without Weir 100 200 300 400 500 Mean Discharge (cms) Figure 7. The estimated relationship between the proportion of small (<65 cm) bull trout in the flip bucket and discharge with and without the weir for the 200 th day of the year 2000. The points represent the within year residual variation. Discussion The current data suggest that the presence of the weir increases the ability of smaller bull trout to access the flip bucket relative to larger individuals. According to the model, the increase is somewhere between 28% and 128% with 80% confidence. It is important to note that even the lower 80% confidence interval represents a substantial increase in abundance. Nevertheless, some caution in interpreting the results is required. With the exception of 1999, all the modelled years for which the weir was installed occurred prior to 2002 (Figure 5). Consequently, the years in which the weir was installed tended to occur in a different decade to the years in which the weir was absent. The model attempts to correct for any long term changes in either the proportion of small bull trout approaching the dam or physical parameters affecting jumping ability through a long term trend. However, due to the decadal separation, the long term trend is partially confounded with the absence of the weir. One or more years of data with the weir installed for part of the year would help to discriminate between the effects of the weir and any decadal changes and increase the certainty concerning the validity of the model. The decline in the proportion of smaller fish between the end of the 1990s and the beginning of the current millennium is noteworthy although the cause is unclear. It may be due to a Page 8

reduction in the Kootenay Lake Fertilization Program s nutrient inputs between 1997 and 2000 that resulted in a decrease in kokanee abundance (Schindler et al. 2007). The analysis also indicates that as the season progresses the proportion of small bull trout in the flip bucket tends to increase suggesting that on average larger bull trout have an earlier run timing than smaller individuals. The BC Hydro protocol is to install the weir during low flow years when the jump height is expected to be a problem for the smaller bull trout (T. Oussoren pers. comm.). Although this means the data might be biased, the bias would be conservative in the sense that it will tend to mask the effect of the weir. The mean weekly discharge below the confluence with the Lardeau River was included in the model in an attempt to account for the variation in jump height. However, the proportion of small bull trout decreased with increasing discharge suggesting that some other discharge related factor is affecting the number of smaller bull trout entering the flip bucket. The analysis suggests that without the weir the dam is size selective and that installing the weir reduces the size bias. It should, however, be noted that some uncertainty surrounds the current analysis as the presence or absence of the weir is partially confounded with any interdecadal changes. Furthermore, even if completely valid, the analysis does not demonstrate that installing the weir completely eliminates any size bias. In order to demonstrate that dam passage is not size selective a study would have to be conducted to demonstrate that smaller bull trout approaching the dam have the same probability of being successfully transferred into the reservoir as larger individuals. The probability of successful transfer could be estimated by radio or acoustic tracking of bull trout ascending the Duncan River. Recommendations If the numbers and sizes of bull trout accessing the flip bucket are to be monitored long term then a level logger should be installed at the base of the lower level operating gates to provide a more accurate measure of the jump height. A temperature logger should also be installed to measure the water temperature in the plunge pool below the flip bucket. One or more additional years of data with the weir installed for part of the year should be collected to help discriminate between the effects of the weir and any decadal changes. Page 9

Closure This memorandum is to the best of my knowledge accurate and correct. If you have any questions regarding its contents please contact the undersigned. Dr. Joseph Thorley, R.P.Bio. Poisson Consulting Ltd. Fish Population Biologist Page 10

References BC Hydro (2005). Consultative Committee Report: Duncan Dam Water Use Plan. Prepared for the Consultative Committee for the Duncan Dam Water Use Plan. BC Hydro (2008). Duncan Dam Bull Trout Passage Monitoring Terms of Reference. Duncan Dam Water Use Plan Monitoring Program. Binder, T. and E. Stevens (2004). Appetitive Conditioning Technique Reveals Behavioural Limits to Passage Performance in Fishes Environmental Biology of Fishes 71: 1573 5133. Bradford, M. J., J. Korman and P. S. Higgins (2005). Using Confidence Intervals to Estimate the Response of Salmon Populations (Oncorhynchus Spp.) to Experimental Habitat Alterations. Canadian Journal of Fisheries and Aquatic Sciences 62: 2716 2726. Collett, D. (2003). Modelling Binary Data. Second Edition. Boca Raton, Florida, Chapmand & Hall/CRC. Dobson, A. J. and A. G. Barnett (2008). An Introduction to Generalized Linear Models. Third Edition. Boca Raton, Florida, CRC Press. Hagen, J. (2003). Precision of Bull Trout Escapement Estimates at Duncan Dam Relative to Sampling Intensity, and a Discussion of Factors Influencing Transfer Success. Haugen, T. O., P. Aass, N. C. Stenseth and L. A. Vollestad (2008). Changes in Selection and Evolutionary Responses in Migratory Brown Trout Following the Construction of a Fish Ladder. Evolutionary Applications 1: 319 335. Kondratieff, M. C. and C. A. Myrick (2006). How High Can Brook Trout Jump? A Laboratory Evaluation of Brook Trout Jumping Performance. Transactions of the American Fisheries Society 135: 361 370. McPhail, J. D. (2007). The Freshwater Fishes of British Columbia. Edmonton, University of Alberta Press. McPhail, J. D. and J. S. Baxter (1996). A Review of Bull Trout (Salvelinus Confluentus) Life History and Habitat Use in Relation to Compensation and Improvement Opportunities. Fisheries Management Report No. 104. Mesa, M. G., L. K. Weiland and G. B. Zydlewski (2004). Critical Swimming Speeds of Wild Bull Trout. Northwest Science 78: 59 65. O'Brien, D. S. (1999). The Duncan Bull Trout Telemetry Project (1995 1997). For Columbia Basin Fish & Wildlife Compensation Program. Pinheiro, J. C. and D. M. Bates (2000). Mixed Effects Models in S and S Plus. New York, Springer. Page 11

R Development Core Team (2008). R: A Language and Environment for Statistical Computing, Vienna, Austria. Schindler, E. U., H. Andrusak, K. I. Ashley, G. F. Andrusak, L. Vidmanic, D. Sebastian, G. Scholten, P. Woodruff, J. Stockner, F. Pick, L. M. Ley and P. B. Hamilton (2007). Kootenay Lake Fertilization Experiment Year 14 (North Arm) and Year 2 (South Arm) (2005) Report. Fisheries Project Report No. RD 122. Fish and Wildlife Compensation Program Columbia Basin and Kootenai Tribe of Idaho. Fish and Wildlife Science and Allocation, Ministry of Environment, Province of British Columbia. Stuart, T. A. (1962). The Leaping Behaviour of Salmon and Trout at Falls and Obstructions. Freshwater and Salmon Fisheries Research. Wood, S. N. (2006). Generalized Additive Models: An Introduction with R. Boca Raton, Florida, Chapman & Hall/CRC. Page 12

Appendix A Table 2. The input data. Transfer Number Weir Year Day of the YearDischargeFish MeasuredProportion Small 199504 Yes 1995 159 237.6 86 0.40 199506 Yes 1995 173 179.3 93 0.47 199507 Yes 1995 187 146.0 92 0.57 199509 Yes 1995 200 194.0 35 0.63 199510 Yes 1995 208 168.6 6 0.50 199602 Yes 1996 158 184.1 31 0.48 199603 Yes 1996 166 270.9 42 0.38 199604 Yes 1996 185 201.6 22 0.59 199605 Yes 1996 192 260.4 60 0.58 199608 Yes 1996 220 442.9 1 0.00 199609 Yes 1996 227 338.4 50 0.66 199702 Yes 1997 163 271.0 45 0.31 199703 Yes 1997 176 257.3 25 0.44 199704 Yes 1997 185 165.7 29 0.48 199705 Yes 1997 190 223.4 6 0.17 199706 Yes 1997 204 546.9 25 0.32 199708 Yes 1997 225 347.6 15 0.40 199709 Yes 1997 237 206.0 12 0.50 199710 Yes 1997 247 172.9 24 0.63 199801 Yes 1998 134 192.3 55 0.18 199802 Yes 1998 148 163.3 177 0.35 199803 Yes 1998 161 158.0 102 0.47 199804 Yes 1998 174 132.3 67 0.54 199805 Yes 1998 189 110.0 43 0.44 199806 Yes 1998 203 131.7 141 0.60 199807 Yes 1998 217 272.1 24 0.13 199808 Yes 1998 231 202.7 70 0.47 199809 Yes 1998 245 180.0 26 0.65 199810 Yes 1998 259 222.4 9 0.22 199811 Yes 1998 271 218.3 2 1.00 199901 No 1999 144 195.1 30 0.07 199903 No 1999 174 300.0 84 0.24 199904 No 1999 190 189.0 126 0.27 199905 No 1999 202 217.6 54 0.30 200001 Yes 2000 131 122.4 37 0.24 200002 Yes 2000 147 132.0 49 0.22 200003 Yes 2000 164 179.9 96 0.28 200006 Yes 2000 201 182.0 45 0.29 200008 Yes 2000 236 206.1 13 0.31 Page 13

200010 Yes 2000 264 328.6 8 0.50 200101 Yes 2001 150 161.0 51 0.37 200103 Yes 2001 178 127.4 70 0.26 200104 Yes 2001 193 127.0 45 0.60 200105 Yes 2001 206 113.5 15 0.53 200106 Yes 2001 220 209.9 31 0.58 200301 No 2003 149 118.1 6 0.33 200302 No 2003 163 217.3 40 0.23 200303 No 2003 176 219.1 47 0.26 200304 No 2003 190 148.0 47 0.38 200305 No 2003 204 104.7 2 0.00 200306 No 2003 218 189.6 28 0.36 200307 No 2003 232 170.9 25 0.44 200401 No 2004 147 133.4 25 0.24 200402 No 2004 161 145.0 30 0.13 200404 No 2004 189 159.7 20 0.35 200405 No 2004 203 93.7 30 0.27 200509 No 2005 256 228.4 33 0.36 200706 No 2007 234 185.6 23 0.52 200802 No 2008 169 164.9 52 0.42 200805 No 2008 198 179.3 55 0.49 200808 No 2008 232 234.9 15 0.73 Page 14

Appendix B The output of the quasi binomial (generalized) additive mixed model fitted in R 2.8.1. Family: quasibinomial Link function: logit Formula: Prop ~ Weir + s(doy) + s(year) + s(discharge) Parametric coefficients: Estimate Std. Error t value Pr(> t ) (Intercept) -0.7216 0.1478-4.882 9.63e-06 *** Weir 0.5343 0.2037 2.623 0.0113 * --- Signif. codes: 0 *** 0.001 ** 0.01 * 0.05. 0.1 1 Approximate significance of smooth terms: edf Ref.df F p-value s(doy) 1.000 1.500 27.928 7.88e-08 *** s(year) 2.754 3.254 6.311 0.000699 *** s(discharge) 1.000 1.500 6.046 0.008368 ** --- Signif. codes: 0 *** 0.001 ** 0.01 * 0.05. 0.1 1 R-sq.(adj) = 0.934 Scale est. = 1.3616 n = 61 Page 15