BPA Fish and Wildlife FY 1997 Proposal

Section 1. Administrative
Section 2. Narrative
Section 3. Budget

see CBFWA and BPA funding recommendations

Section 1. Administrative

Title of project
Analysis of Smolt Migration Rate & Survival Based on Pit-Tag Recoveries

BPA project number   5500100

Business name of agency, institution or organization requesting funding
S.P. Cramer & Associates

Sponsor type   OR-Consultant

Proposal contact person or principal investigator

BPA technical contact   Deborah Docherty, EWN 503/230-4458

Biological opinion ID   13.f, 13.g, 16, 17

NWPPC Program number   5.0F.4

Short description
Analysis of the comprehensive PIT tag detections from all sources to estimate spill effectiveness, in-season changes in smolt survival and migration rate, effects of flow pulses. Simulate biases in alternative methods for estimating collection efficiency and survival.

Project start year   1997    End year   1999

Start of operation and/or maintenance   1998

Project development phase   Implementation

Section 2. Narrative

Related projects
Project Relationship to this project (No. 5500100)
8332300 Tagging and detection of PIT tags that contribute to database analyzed in this project
8401400 PIT tag detections at Bonneville and John Day will be included in this analysis
9102800 Tagging and detection of PIT tags that contribute to database analyzed in this project
8910700 Tagging and detection of PIT tags that contribute to database analyzed in this project
9105100 Analysis limited in scope. This projects expands data set and river stretches considered.
9302900 Tagging and detection of PIT tags that contribute to database analyzed in this project
9008000 Ptagis Database will be information source for this project
9300800 This project will enable correlation of daily or weekly survival rates to gas saturation levels
9301200 This project will enable daily juvenile abundance estimates and correlation to passage survival
8741302 This project will enable estimates of passage survival during outmigration of CWT groups.
8910800 Migration rate and survival estimates from this project will be used in CRiSP model calibrations.

Project history
Initial efforts for this project in 1995 were funded independently by the Direct Service Industries (DSIs). Start-up year cost was approximately $70,000. Little or no funding will be available from the DSIs to continue this project. The project was initially intended to determine if the PIT tag database was sufficient to enable estimation of survival during 1-day to 1-week intervals within the outmigration season. Once this was found to be possible, the analyses were expanded to develop survival estimates from Lower Granite to McNary Dam for each daily cohort starting at Lower Granite. Two analytic methods for estimating collection efficiency and survival were compared and found to deviate in the answers they provided for different portions of the year. Potential biases in the different estimation methods, including the Jolly-Seber method commonly used by NMFS and FPC were explored, and some biases were found in each method investigated. Further exploration of biases created by variable collection efficiencies, variable survival rates within daily cohorts, and preferential route selection by individual fish were recommended. Some sources of bias may be substantial.

The first year of work revealed that the database is vast and that we have just begun to scratch the surface of possible analyses. Ongoing analyses, other than this project, have focused on the results of specific tagging projects, but have not analyzed the comprehensive set of data. Comparisons we made of fish size, migration timing, and migration rate indicated these characteristics were similar between most PIT-tagged groups throughout the Snake River Basin. Thus, the initial analyses suggest that tagged fish from variety of sources can be aggregated to achieve large sample sizes, and preform analyses between a number of different detection points.

Biological results achieved
' A new analytical method for estimating survival of PIT-tagged fish was developed that enables use of a larger data set and division of detections into more discrete time intervals than the traditional Jolly-Seber method.
' Methods were developed to estimate variance and confidence intervals around the survival estimates
' Survival of yearling chinook smolts from Lower Granite Dam to McNary Dam was estimated for daily cohorts within 95% confidence intervals of �10% for most of the outmigration season in 1995.
' Survival dropped sharply for daily cohorts passing Lower Granite Dam in mid May, and rose sharply again in late May.
' Sharp changes in survival correlated with changes in cumulative exposure of smolts to gas saturation, but not with fish size, travel time, stock composition, or river temperature.
' Median travel time changed little over a wide range of flows.

Annual reports and technical papers
Cramer, S.P. February, 1996. Seasonal changes in survival of yearling chinook smolts emigrating through the Snake River in 1995 as estimated from detections of PIT tags. Final Report, prepared for Direct Service Industries. S.P. Cramer & Associates, Gresham, Oregon

Cramer, S.P. November, 1995. Response to comments by Fish Passage Center on �Assessment of the effects of spill on survival of anadromous salmonids in the Columbia Basin.� Draft report prepared for Direct Service Industries. S.P. Cramer & Associates, Gresham, Oregon

Cramer, S.P., C.F. Willis and K.L. Witty. September 1995. Assessment of effects of spill on survival of anadromous salmonids in the Columbia Basin. Progress Report prepared for Direct Service Industries. S.P. Cramer & Associates, Gresham, Oregon.

Neeley, D. February 1996. Estimated variance of survival estimate Si=Sj(dij/rj)/Ni. Draft Report prepared for Direct Service Industries. S.P. Cramer & Associates, Gresham, Oregon.

Management implications
' Factors influencing smolt survival can be determined more precisely than ever before because survival can be estimated with large sample sizes and short time intervals.
' River flows and dam operations were identified that resulted in passage survivals of about 90% for the entire distance form Lower Granite to McNary Dam.
' Spill must be managed to limit the cumulative exposure of smolts to dissolved gas supersaturation through several projects. Smolts survived well through supersaturation levels up to 130% below Ice Harbor Dam only when supersaturation was less than 115% in the tailraces of upstream dams.
' The general correlation of smolt travel time to river flow does not hold under some conditions, so further exploration of the factors influencing travel time is needed.
' Travel time cannot be used as a surrogate measure of smolt survival
' Traditional methods for estimating smolt survival may be strongly biased, so the nature and magnitude of these biases needs thorough evaluation.

Specific measureable objectives
Objective 1: Estimate the proportion of smolts that pass over the spillway at each collector dam for varying proportions of spill.

Objective 2: Quantify the relationship of smolt survival rates to specific aspects of system operations

Objective 3: Quantify the relationship of smolt migration rates to specific aspects of system operations

Objective 4: Determine if pulses in flow Que rapid outmigration of smolts, such that survival is enhanced

Objective 5: Describe the magnitude of bias generated by the Jolly-Seber method and by the time-stratified method in estimates of collection efficiency and passage survival when several of the common variables influencing these parameters are varying simultaneously.

Testable hypothesis
Objective 1.
A. FGE remains constant throughout the migration season at each collector dam
B. Variation in FGE is not related to fish size, river temperature, turbidity other physical variable
C. The proportion of smolts passing over the spillway is equal to the proportion flow spilled.
D. Variation in the proportion of smolts passing over the spillway is not related to fish size, river temperature, turbidity other physical variable
Objective 2
A. Smolt survival rate does not vary between days or weeks
B Smolt survival rate does not vary between dams
C. Variation in smolt survival rate is not related to river flow, fish size, river temperature, turbidity, dissolved gas supersaturation, or other physical variable.
D. Variation in smolt survival rate is not related to spill rate.
E. Smolt survival is not related to migration rate or water particle speed.
F. There is no mortality due to passage through turbines or over spillways that is delayed until passage of two or three dams downstream.
Objective 3
A. Smolt migration rate does not vary between days or weeks
B Smolt migration rate does not vary between dams
C. Smolt migration rate is not related to water particle speed.
D. Variation in smolt migration rate is not related to spill rate.
E. Variation in smolt migration rate is not related to fish size, river temperature, turbidity, dissolved gas supersaturation, or other physical variable.

Objective 4
A. Migration rate does not change before, during, and after a pulse in flow.
B. Changes of less than 20% in flow do not stimulate migration rate.
C. Slow, medium and fast migrating individuals of a cohort all respond similarly to a sharp change in flow
D. Migration rates of fish that show and initial stimulatory response to a pulse in flow do not slow down when flow stops changing.
E. The stimulatory response of smolts to a pulse in flow is not altered by the magnitude of the pulse, the duration of the peak flow, the rate of change in flow, or the elapsed time since the last pulse in flow.
F. Passage survival does not improve among smolts that show a stimulatory response to a pulse in flow.

Objective 5.
Collection efficiency and passage survival from PIT tag detections are not biased under the following conditions:
A collection efficiencies at the two recovery locations are both varying
B survival rate varies between the fast and slow migrants within the same marked group,
C. fish collected at one dam are more prone to be collected at the next dam,
These hypotheses would be tested for each estimation method.

Underlying assumptions or critical constraints
' The PIT-tag detectors will continue to function properly at collector dams
' A large number of PIT-tagged juveniles will continue to be released each year.
' Environmental conditions will vary sufficiently within and between years that factors influencing collection efficiency, migration rate, and survival will become detectable by standard statistical techniques.
' PIT-tagged fish behave similarly to untagged fish.

Methods
Objective 1: Estimate the proportion of smolts that pass over the spillway at each collector dam for varying proportions of spill.


Detections of PIT tags at any point downstream of a given dam provide an opportunity to determine the proportion of these fish that were detected as they passed a dam of interest upstream. Because there were more than 2,500 Snake River chinook smolts with PIT tags detected as they passed John Day Dam in 1995, we can estimate Fish Passage Efficiency at least on a weekly basis for McNary Dam. Sample sizes were substantially larger at upstream collector dams where FPE has the potential to be estimated daily. For example, over 74,000 PIT- tagged chinook yearlings were detected at Lower Monumental Dam, over 79,000 were detected at Little Goose Dam, and over 35,000 were detected at Lower Granite Dam.

Depending on the number of PIT-tagged fish available each year, we may be able to estimate passage efficiency for separate races and species, including:

' Yearling chinook
' Subyearling chinook
' Steelhead

In order to estimate the proportion of smolts passing over the spill, we must also estimate the proportion that pass through the turbines. PIT-tag data are available at the Snake River collector dams during periods of no spill, and PIT-tag detection rates at such times provide an estimate of Fish Guidance Efficiency (FGE). Our analysis of PIT-tag recoveries at McNary Dam in 1995 indicates the FGE was constant through most of the season, such that most variation in detection rates could be assigned to fish passing with the spill. Thus, once we estimate FGE with the turbines in standard operating mode, we can use that estimate and the number of fish collected to estimate the number that passed through the turbines. Variation in FGE will cause some scatter in the correlation of spill effectiveness to spill volume, but this variation is likely to be small in comparison to the variation caused by spill volume.


Task 1.1 Determine the FGE for the juvenile collection facilities at Lower Granite, Little Goose, Lower Monumental, and McNary during standard operating conditions

There was no spill at Lower Granite and Lower Monumental Dams until May 10 in 1994, so PIT-tag detections prior to this time can be used to estimate FGE. On the other hand, water has been spilled throughout the spring outmigration in both 1994 and 1995 at Little Goose, so FGE will have to be estimated through an iterative or simulation process in combination with spill efficiency.

Task 1.2 Determine the proportion of smolts that pass over the spillway at each dam

Once FGE has been estimated, then the number of fish that passed through the turbines can be estimated based on the number collected, and the number of fish that passed over the spill and be estimated by subtraction. That is:

# Through Turbines = (#Collected/FGE) x (1-FGE)
#Spilled = (#Collected/Proportion Collected) - #Collected - # Through Turbines

The proportion spilled can then be estimate from the number spilled divided by the total passage of PIT tags.

Proportion Spilled = #Spilled/(#Collected + #Turbines + #Spilled)

These calculations would be repeated for each day or week of PIT-tag release (depending on sample size) and for each race or species of fish.

Task 1.3 Quantify the effects of spill rate on Fish Passage Efficiency.

In order examine the factors that influence Fish Passage Efficiency through the spill, we must first line up the daily or weekly estimates of spill efficiency with the data on possible influencing factors, such that the dates coincide. To do this, we would estimate the day or week that most of the PIT-tagged fish passed the dam of interest. This date would be estimated in one of two ways, depending on which method provided the least variance in passage date. One method would be to use fish released on a given date, and determine the mode date of arrival at the dam of interest downstream. The second method would be to use fish detected on a given day(s) at the next dam downstream and determine the mode date of passage at the dam of interest upstream.

Once we have lined up the data on collection efficiency and influencing factors, all for corresponding dates, we would employ several analytic techniques to determine their relationship. Multiple regression would be used to examine additive and multiplicative relationships. For example, regression could be used to quantify the combined effects of spill rate and turbidity on passage efficiency. Analysis of variance (ANOVA) could be used to examine step functions, such as differences in passage efficiency between mid Columbia fish (Washington shore) and Snake River fish (Oregon shore) approaching McNary Dam. Simulation analysis with submodels could be used in cases were several complex influences were believed important. This might be the case if different turbine outages or gate settings were thought to have substantial influence on passage efficiency.

Objective 2: Quantify the relationship of smolt survival rates to specific aspects of system operations

Task 2.1 Estimate smolt survival rates within discrete river sections and time intervals.

We are most likely to identify what and where the key sources of smolt mortality are if we can partition survival estimates into the smallest space and time increments possible. Daily estimates of survival for spring chinook smolts are possible between Lower Granite to Little Goose, Little Goose to Lower Monumental, and Lower Monumental to McNary. Season-long estimates are possible from McNary to John Day, and perhaps from John Day to Bonneville, based on detections at Bonneville and in the estuary. Estimates between various locations and for various time intervals are also possible for subyearling chinook, steelhead, and sockeye.

Task 2.2 Estimate delayed mortality of smolts due to passage over the spill or through the turbines.

This could be accomplished by comparing relative detection rates of special PIT-tag groups at a sequence of recovery points downstream. NMFS has been releasing PIT-tag groups into the spillways and into the turbines at Snake River dams, and estimating their relative survival to Lower Monumental Dam. There is some evidence that mortality associated with these passage routes may be delayed rather than immediate. For example, fish passing over the spillway may experience high descaling rates, that later translate to increased mortality. One way of testing for delayed mortality (which NMFS has not fully pursued) is to compare relative detection rates of different released types (eg. spill vs. bypass) at sequential points downstream. There were substantial numbers of recoveries at McNary and John Day dams that could be examined. There were even several hundred detections in 1995 at Bonneville and in the estuary that could be examined.

Task 2.3 Estimate smolt survival relative to spill rate at each dam

Smolt survival rate can be examined for changes before and after sharp changes in spill, either by using ANOVA or regression. ANOVA would be especially useful for comparing survival immediately before and after the initiation of controlled spill each year.

Task 2.4 Estimate smolt survival relative to flow and water-particle speed between each dam.
Flow and water-particle speed are the focus of many operational alternatives proposed to benefit smolt survival. Again, smolt survival can be examined for changes before and after sharp changes in flow or water-particle speed, either by using ANOVA or regression.

Task 2.4 Estimate smolt survival relative to migration rate between each dam

Migration rate has often been used as a surrogate indicator of smolt survival. Data are now available to test the hypothesis that the two are linked, and this hypothesis can be tested between several dams and for yearling chinook, subyearling chinook, steelhead, and sockeye. Several recent tests of the hypothesis indicate that it is false, at least for the river conditions tested. We should determine if there are conditions under which it is true.


Objective 3: Quantify the relationship of smolt migration rates to specific aspects of system operations
The tasks required to answer this objective are parallel to those for Objective 2 on survival rate. Only the task titles are listed below. The analyses would be similar to those for tasks under Objective 2, but the variable of interest would be migration rate rather than survival.

Task 3.1 Estimate smolt travel time within discrete river sections and time intervals.


Task 3.2 Estimate smolt travel time relative to spill rate at each dam


Task 3.3 Estimate smolt travel time relative to flow and water-particle speed between each dam.


Objective 4: Determine if pulses in flow Que rapid outmigration of smolts, such that survival is enhanced
If pulses in flow can be used to Que rapid outmigration of smolts, then survival might be increased directly by the fishes response, or indirectly by fish-friendly operational changes timed to coincide with the stimulated migration. The comprehensive set of PIT tag data offers the capability to perform opportunistic analyses of smolt behavior following sharp changes in flow that have occurred in the past and will occur in the future. We have data from a number of west coast streams indicating that sharp increases in flow act as temporary stimuli to downstream migration of chinook. For example, the increasing migration rates of yearling chinook in the Snake River during late April and early May of 1995 coincided with a period of rapidly increasing flow, and migration rates dropped after flow plateaued. As another example, a doubling of flow in the Snake River during July 1994 stimulated a sharp increase in outmigration of juvenile chinook, but the number of outmigrants passing Lower Granite Dam dropped back to nominal levels within a few days, despite continued high flows. There were numerous PIT-tagged fish detected at Lower Granite Dam during July of 1994, so an analysis of their migration and survival rates before, during, and after the pulse in flow is possible.

Task 4.1 Compare juvenile migration rates before, during, and after sharp changes in flow.

We know that catches of juvenile chinook migrants increase dramatically at dams following some sharp increases in flow, but we need to determine if these fish are truly stimulated to quickly migrate to sea. There is evidence from studies in the Rogue River that some fish are stimulated by a pulse in flow to migrate to sea, while other fish move varying distance downstream and then stop or slow down to resume rearing. We would compare frequency distributions of migration rates between fish passing a given dam immediately before, during and after sharp increases in flow to determine if slow, fast, or all migrants alike show the same response to a pulse flow stimulus.

Task 4.2 Compare migration rates at sequential points downstream for fish which show an initial stimulus to migrate from a sharp increase in flow.

We would compare migration rates at subsequent times and downstream dams for groups of fish that passed Lower Granite Dam in response to a flow pulse. We need to determine how long and how far downstream the stimulus to migration lasts.

Task 4.3 Quantify the characteristics of flow pulses that determine their effectiveness at stimulating migration.

Juvenile migrants appear to respond to some sharp increases in flow but not to others. We would compare several characteristics of the flow events between effective and ineffective pulses, including magnitude of flow increase, rapidity of flow increase, base flow, duration of flow increase, elapsed time since last sharp increase or decrease in flow.

Task 4.4 Determine if smolt survival improves among groups that show a migratory response to a sharp increase in flow.

Analyses under this task would be analogous to those for migration rate under tasks 4.1 and 4.2. We would be seeking to quantify the temporal and spatial increase in survival among fish that were stimulated to migrate by a pulse in flow (How long and how far downstream does the effect last?). Most analyses would be likely be ANOVA between groups that responded to a pulse flow and groups that preceded or followed them.

Objective 5: Describe the magnitude of bias generated by the Jolly-Seber method and by the time-stratified method in estimates of collection efficiency and passage survival when several of the common variables influencing these parameters are varying simultaneously.

Our preliminary examination of the Jolly-Seber and time-stratified estimation methods indicated that simultaneous variation in collection efficiencies at the two recovery dams would bias estimates of survival. Further exploration of biases created by variable collection efficiencies, variable survival rates within daily cohorts, and preferential route selection by individual fish are needed.

Task 5.1 Construct a simulation model that calculates actual detections of PIT tags at each collector dam under a variety of scenarios for daily survival rates, collection efficiencies, migration rates, and passage route preferences.

Task 5.2 Calculate bias sensitivity curves for each estimation method and each variable examined.

Brief schedule of activities
First 2 mo. Complete simulations to compare bias and confidence intervals for alternative estimation models
Ongoing Participate in coordination and review meetings regarding mainstem migration of juveniles.
Ongoing Sort the Ptagis database into data sets useful for testing designated hypotheses
Jun-Sep Perform statistical analysis of data sets
Oct Prepare rough draft and distribute for peer review
Nov Annual Report

Biological need
' Determine the best means to optimize juvenile passage survival
' Identify system operations and management actions that negatively impact passage survival

Critical uncertainties
Are the sources of bias to be examined in Objective 5 substantial, and if so, can we develop methods to avoid or compensate for those biases?

Summary of expected outcome
We should be able to delineate the set of water management and dam operating conditions that produce the highest smolt survival.

Dependencies/opportunities for cooperation
The magnitude of the database depends on two broad sets of actions by others:
1. the number of fish tagged at various locations is dependent on others
2. the proportion of fish detected at each dam is dependent on the installation and regular operation of PIT tag detectors
Enhancement of detection capability at John Day, Bonneville, and the estuary will greatly increase the size of the database and the river reaches for which survival can be estimated.

Risks

Monitoring activity
We will aggressively seek peer review by:
' Participating in regional coordination and planning meetings regarding juvenile passage survival
' Scheduling meetings in early summer to review our findings with other groups working on smolt survival, including NBS, NMFS, FPC, IDFG, and ODFW.
' Distributing progress reports for peer review
' Hiring statistical consultants for critical review

Section 3. Budget

* For most projects, Authorized is the amount recommended by CBFWA and the Council. Planned is amount currently allocated. Contracted is the amount obligated to date of printout.

Funding recommendations

CBFWA funding review group   Mainstem

Recommendation    Tier 3 - do not fund