I "fifth;

2W9

This is to certify that the
thesis entitled

EVALUATION OF FIXED-LOCATION, RIVERINE
HYDROACOUSTICS FOR ENUMERATING OUT-
MIGRATING CHINOOK SALMON SMOLTS

presented by

Jeremy Dean Price

has been accepted towards fulfillment
of the requirements for the

MS. ree in \ Fisheries and Wildlife

 

 

M/ZZUW

Major Professor’s Signature

7ft Eerie 7c§§5

 

Date

MSU is an Affirmative Action/Equal Opportunity Institution

, fi——-—-—— w v.—

LIBRARY
Michigan State
University

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE

DATE DUE

DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2/05 p:/CIRC/DateDue.indd-p.1

 

EVALUATION OF FIXED-LOCATION, RIVERINE
HYDROACOUSTICS FOR ENUMERATING OUT-
MIGRATING CHINOOK SALMON SMOLTS
By

Jeremy Dean Price

A THESIS
Submitted to
Michigan State University
in partial fulfillment of the requirement
for the degree of
MASTER OF SCIENCE
Department of Fisheries and Wildlife

2005

ABSTRACT

EVALUATION OF FIXED-LOCATION, RIVERINE HYDROACOUSTICS FOR
ENUMERATING OUT-MIGRATING CHINOOK SALMON SMOLTS

By
Jeremy Dean Price
Limited information regarding natural chinook production in Lake Michigan
tributaries complicates management of this economically important fishery. Various
techniques have been used to quantify smolt production in Michigan streams, each
having strengths and weaknesses in its application. Fixed-location, riverine
' hydroacoustics are proposed as an additional method for quantifying chinook
production in the Muskegon River, Michigan.
Evaluation of the hydroacoustic systems and processing software performance
‘ was qualitatively achieved through spatio-temporal comparisons of results from visual
inspection of echOgrams and data processing results. This comparison revealed a
perceived inability to detect targets and create tracks where many were expected and a
propensity towards creating fish tracks from noise echoes.
Passage estimates were compared from concurrent hydroacoustic and auger-
.type, smolt trapping surveys conducted in 2001 and 2002. Simple linear regression was
used to examine the relationship between trap and acoustic daily passage estimates.
The relationship between corresponding daily passage results from 2001, 2002, and
both years combined proved to be statistically insignificant (ANOVA, p = 0.94, 0.59,
and 0.61 respectively). The methods used herein may be worthy of further development

for use on smaller streams. However, use in their current form for management would

be premature.

For my parents, who set the bar high for me and gave me everything I needed to
achieve my goals. For my wife, Kathy, whose love and friendship I could never be
without. And finally, for my children. You are my hope, my joy, and my inspiration.

May each of your lives be full of the same.

iii

ACKNOWLEDGEMENTS

I would like to thank the Great Lakes Fishery Trust for funding this research. I
would also like to thank my graduate committee Dr. Doran Mason, Dr. Daniel Hayes, Dr.
Ed Rutherford, and Dr. Jan Stevenson for each of their roles in obtaining my degree and
furthering my professional development. I would especially like to express my
appreciation to Dr. Mason for this opportunity and his confidence in my work.

I would like to thank my entire family for their endless support. Thanks to my
brother, Jim, for talking me off the ledge at a critical time. You’re my man! Special
thanks to my wife, Kathy, for her patience in this process. I never could have finished
this without your help. I owe you big!

Last, but not least, I would like to thank God for the abilities and opportunities
that I’ve been given in life. You’ve blessed me far beyond what any man deserves

despite my inadequacies. Thank you for everything.

iv

TABLE OF CONTENTS
List of Tables ....................................................................................
List of Figures ...................................................................................
Key to Symbols or Abbreviations ............................................................
Chapter 1 - A Brief History of Chinook Salmon in the Great Lakes ....................
Introduction .................................................................................
Problem Statement ....... - ....................................................................
Project Objectives ...........................................................................

Chapter 2 - Evaluation of Fixed-Location, Riverine, Hydroacoustics for
Enumerating Out-migrating Chinook Salmon Smolts ...............................

Introduction .................................................................................
Methods ......................................................................................

Study site .................................................................................

Site selection .............................................................................

Data collection ...........................................................................

Beammapping..................... ......................................................

Data processing ..........................................................................

Performance analysis ...................................................................

Iterative parameter perturbation ......................................................

Passage estimates ........................................................................

Results .......................................................................................

Beam Mapping ..........................................................................

vii

viii

ix

1

10

10

11

l4

l6

17

22

28

28

32

32

Chapter 2 (continued) ...........................................................................

Results (continued) ..........................................................................
Performance analysis .................................................................... 34
Passage Estimates ........................................................................ 39

Discussion .................................................................................... 44
Gear Comparisons ................................................................... 44
Performance Analysis ................................................................... 45
Passage Estimates ........................................................................ 46
Fixed river acoustics: Assumptions and problems ............................. 47

Conclusions ................................................................................... 48

Future Research .............................................................................. 49

Chapter 3 -The Future of Fixed-location, Riverine Hydroacoustics in Great Lakes

Fisheries Management ...................................................................... 51
Project Summary ............................................................................. 52
Conclusions ................................................................................... 55
Appendix .......................................................................................... 57

. Literature Cited .................................................................................. 65

vi

Table 1.
Table 2.
Table 3.

Table 4.

Table 5.
Table 6.
Table 7.

Table 8.

Table 9.

LIST OF TABLES

List of parameters and values used in processing and analyzing hydroacoustic
data ....................................................................................... 18

Minimum and maximum length of smolts captured in the trap survey
(Rutherford, unpublished data) and the corresponding TS threshold

values .................................................................................... 30
Percentages of each result code's occurrence by range bin ...................... 34
Daily smolt passage estimates from acoustic and trapping studies. . . .. . . . . ....39

Correlations and significance from comparisons of daily passage estimates

. from trap and acoustic results ........................................................ 42

Contribution of post-processing filters to total number of tracks filtered for

2001 ..................... . ............................................................... 58
Contribution of post-processing filters to total number of tracks filtered for
2002 ..................................................................................... 59
Species composition from smolt trap by trap day in 2001 ...................... 6O
Species composition from smolt trap by trap day in 2002 ...................... 62

vii

Figure 1.
Figure 2.
Figure 3.
. Figure 4.
Figure 5.

Figure 6.

Figure 7.

Figure 8. .

Figure 9.

Figure 10.

Figure 11.

Figure 12.

Figure 13.

Figure. 14.

Figure 15.

Figure 16.

Figure 17.

Figure 18.

LIST OF FIGURES

Aerial photo showing location of hydroacoustic and trap deployments... . . . . l 2

Cross-sectional profile of acoustic sites used in 2001 and 2002.................14
Pipe flame upon which the transducers were mounted............................14
Basic descriptive measurements of a single echo ................................. 20

Decision tree for assigning result codes to each track or anticipated track....24

Geometry of direction of travel measurements in the horizontal plane (aerial

view) ..................................................................................... 29
Rationale behind calculation of whole river estimates ........................... 31
Cross-sectional coverage of the ensonified region ................................ 33
Number of categorical results per volume sampled by range bin..... . .........35

Absolute numbers of results from the four range gating manipulations ...... 36

Ratio of bad to good results for each iteration of the range gating parameter
manipulation ............................................................................ 37

Ratio of omitted to good results for each iteration of the range gating
parameter manipulation ................................................................ 38

Regression results from comparisons of daily passage estimates from trap
and acoustic surveys ................................................................... 40

Comparison of daily passage estimates from trap and acoustic surveys from
2001 with an outlier removed ............................................... ' ......... 41

Mean normalized hourly passage rates from 2001 and 2002 acoustic surveys
(shading from approximation of sunset to sunrise) ............................... 43

Noise levels typically observed in this study with chinook and steelhead TS
shown for reference .................................................................. 55

Temperature and discharge from 2001 as measured at Croton Dam (data
courtesy of US Geological Survey) ................................................. 63

Temperature and discharge from 2002 as measured at Croton Dam (data
courtesy of US Geological Survey) ................................................. 64

viii

KEY TO SYMBOLS OR ABBREVIATIONS

Symbol/ Meaning/U nit Symbol/ Meaning/Unit
Abbreviation Abbreviation
cm ............. centimeters mm ............ millimeters
dB ............. decibels ms ............. millisecond
~ ft ............... feet m/s ............. meters per second
FTA ........... Fish Tracking Algorithm uPa ............ micropascals
GB ............ Gigabytes PLDL ......... Pulse Length
Determination Level
in ............... inches
STDA. . . . . . Single Target Detection
kHz ............ kilohertz Algorithm
m ............... meters TS ............. Target Strength
max ............ maxnnum TL ............. Total Length
mm ............ minimum

ix

 

 

 

 

juidflfiflvr! .Llpjnqud‘g- .

Chapter 1

A Brief History of Chinook Salmon
in the Great Lakes

Introduction

Chinook salmon Onchorhyncus tshauytscha were first introduced into the Great
Lakes in 1873 (Emery 1985). However, it was not until 1967 that a rigorous stocking
program was developed to establish chinook as a major predator in the Great Lakes
ecosystem (Carl 1982). The purpose of the 1960’s reintroductions was to control the
overabundant alewife Alosa psuedoharengus, an exotic invader whose population
exploded following the decline of lake trout Salvelinus namaycush in the Great Lakes and
to establish a substantial sport fishery (Tody & Tanner 1966; Emery 1985).

Chinook are pacific salmon which are generally anadromous in their native range,
spending most of their lives in the ocean and returning to freshwater streams to spawn.
Great Lakes chinook are actually potamodromous, meaning they spend the majority of
their lives in freshwater lakes. There are two variants of the chinook life cycle: the ocean
type and the stream type. These variants are distinguished by the residence time of
juveniles in their natal stream with the ocean type characterized by relative short stream
residence during the early life history stages while stream type remain in streams in
excess of one year after hatching. In general, Great Lakes chinook are of the “ocean
type”. A typical ocean-type adult chinook in the Great Lakes basin returns to its natal
stream to spawn in the fall. Juveniles hatch, emerge, and out-migrate as fingerlings
within a year of spawning. In the Muskegon River, the majority of smolts have usually
emigrated by the end of June. This life cycle is well suited for successful reproduction in
streams that provide marginal thermal habitat during the late summer months such as the

tributaries of southern Lake Michigan.

Today, the results of the stocking efforts of the late 1960’s are felt strongly in the
Great Lakes region. The Great Lakes support a sport fishery of immense economic
value. The fishery generated approximately $2 billion for the regional economy in 1985
(T alhelrn 1988). A substantial portion of that number can be attributed directly to salmon
angler expenditures. At nine index ports in the Michigan waters of Lake Michigan alone,
salmon fishermen accounted for more than 25 million angler hours and nearly 3.2 million
salmonines harvested from 1985 to 1994 (Rakoczy & Svoboda 1997). Chinook salmon
accounted for nearly 44% of the total salmonine harvest during that period (Rakoczy &
Svoboda 1997). l

The successfirl development of the sport fishery can be partially attributed to the
species itself. Chinook salmon adapted quickly to the Great Lakes ecosystem and soon
began to reproduce naturally in Lake Michigan tributaries. Carl (1982) estimated that in
1979 natural recruits contributed 23% of the fishery. Hesse (1994) later estimated that
30% of age I and age 11 chinook harvested in 1992 and 1993 were naturally produced. In
1993 and 1994, Rutherford et a1. (1999) estimated that between 76% and 96% of adult
chinook returning to the Muskegon and Manistee Rivers to spawn were non-stocked fish.
. These studies show that natural salmon production is occurring and is on the rise. Further
increases in natural production of chinook are expected with new guidelines imposed
upon hydropower facilities (i.e. run-of-the-river discharge; O’Neal 1997). Such
guidelines have improved the quality of nursery habitat below the dams and should only
enhance the ability of salmon reproduce naturally in impounded rivers such as the

Muskegon and Manistee.

Beginning in the late 1980’s, angler harvest of chinook salmon and angler hours
targeting salmonines declined precipitously in Lake Michigan (Rakoczy & Svoboda
1997). Returns from stocked fish plummeted approximately 70% between the 1984 and
1985 year-classes to the 1989 and 1990 year-classes (Rakoczy & Svoboda 1997).
Bacterial Kidney Disease (BKD) Renibacterium salmoninarum was identified as playing
a key role in the increased mortality observed (Nelson & Hnath 1990). Factors
contributing to the increased susceptibility of salmon to BKD included: increased salmon
density, reduced prey abundance (specifically alewife), and parasitic stress (Holey et a1.
1998). Evidence suggested that predator density had exceeded sustainable levels and that

stocking rates should be reduced (Holey et a1. 1998; Wesley 1996).

Problem Statement

Fishery managers continue to stock chinook into the tributaries of Lake Michigan
without the benefit of knowing the magnitude of the contribution of natural recruitment
to the system in a given year. The resulting time lag could have detrimental effects upon
the trophic balance within the lake due to a very strong or very weak naturally produced
year class. Ideally, managers could alter stocking rates based on current prey abundance
with the knowledge of natural inputs of predators into the system and reduce the need for
a reactive management strategy. The need for a method of quickly and accurately
estimating natural predator inputs to the system is evident.

With the collapse of salmon in the late 1980’s and early 1990’s, we have come to
understand the potential impacts of overly ambitious stocking practices. The Fish-

Community objectives for Lake Michigan (Eshenroder et a1. 1995) outline several

guiding principles for decision-making regarding the future of the lake’s fisheries
resources. Two of these principles have tremendous implications for stocking policies.
The first of which is “recognizing the limits on lake productivity.” In the discussion of
this principle, the authors highlight excessive stocking as a practice that may lead to
trophic imbalance and high levels of mortality. The second principle is “enhanced
natural reproduction of desirable introduced fishes.” Here the authors acknowledge the
importance of natural feedback mechanisms and the drawbacks of time lags associated
with their artificial counterparts. Both principles underline the importance of predator

inputs into the system and, thus, an understanding of all sources of those inputs.

Project Objectives

Given the economic value of the Lake Michigan salmon fishery and the need for
improved information regarding natural recruitment occurring in Lake Michigan
tributaries, we are proposing a novel method for estimating natural production of chinook
salmon. Here I evaluate fixed, riverine, hydroacoustic technology for enumerating
naturally produced out migrant chinook salmon smolts. For this evaluation, we have
selected the Muskegon River, a large tributary to Lake Michigan in central, lower
‘Michigan. The Muskegon was selected as our study river to complement data being
collected in related studies and because the Muskegon is suspected to be a major
contributor of chinook to the Lake Michigan fish community (Carl 1980; Carl 1982;

O’Neal 1997). The specific objectives of this thesis are:

1.) To comprehensively evaluate fixed-location riverine hydroacoustic methods for
enumerating out migrant chinook salmon smolts.
2.) To eStimate the number of naturally produced chinook salmon smolts contributed

to Lake Michigan from the Muskegon River.

In chapter two, these objectives will first be addressed through comparisons of visual
inspection and software processing results to determine the ability of the method to detect
and track chinook smolts as they pass the deployment site. Also in chapter two,
hydroacoustic estimates of daily smolt passage will be compared with daily passage
estimates from a mark- recapture smolt trapping study that was conducted concurrently.
The results of this should provide a mode of comparison by which to evaluate the

application of hydroacoustics for quantifying smolt production and passage.

Chapter 2

Evaluation of Fixed-Location, Riverine,
Hydroacoustics for Enumerating Out-mi grating
Chinook Salmon Smolts

Introduction

Laurentian Great Lakes support a recreational sport fishery of vast economic
importance, however; this billion-dollar fishery (Talhehn 1988) is not completely self-
sustaining. Fishery managers artificially augment the abundance of the system’s top
predators through stocking of hatchery-raised salmonids. Hatchery production
capabilities, rather than the system’s carrying capacity, have determined the number of
salmon stocked. Public demand can also lead to increases in hatchery production and
stocking rates. However, natural predator populations often exhibit a numerical response
to fluctuations in the abundance of prey available in the system through density
dependent feedback mechanism (Lotka 1925, Volterra 1926, Carpenter and Kitchell
1998). Without consideration of prey abundance, system perturbations from stocking can
prove to be excessive and result in piscivore populations exceeding the carrying capacity
of the system. This can lead to catastrophic mortality in the salmonid populations and
tremendous impacts to the recreational fishery. One such decline in the late 1980’s was
attributed to bacterial kidney disease Renibacterium salmoninarum (Nelson and Hnath
1990). In recent years, fishery managers have considered prey abundance when
determining stocking rates. However, other unknowns can contribute to instability within
the system. Specifically, increases in natural salmonid production coupled with stocking
rates based solely on prey abundance could easily lead to an overabundance of predators.
Thus, a better understanding of natural salmonid production is necessary for efficient
management of the fishery as well as maintaining the overall health of the system.

Several methods are used to quantify the natural recruitment of salmonids in the

Great Lakes basin. Two common methods are pre-smolt electro-fishing surveys and out-

migrant smolt-trapping surveys. While these methods have been used extensively and
evaluated on several occasions (Carl 1982, Newcomb and Coon 2001, Kennen et al.
- 1994, Thedinga et al. 1994, Peterson et al. 2004), the potential biases associated with
capture efficiencies in each method highlight the need for additional sampling options.
Another approach to measure the natural recruitment/production of smolts is fixed-
location riverine acoustics. We have proposed that fixed-location riverine hydroacoustics
can be used to quantify smolt out-migration. This type of survey has a similar concept to
that of the smolt trapping survey. Out-migrant smolts are counted as they move past a
point along the migration route to the outlet. However, hydroacoustics have the ability to
sample a much larger volume of water to provide increased coverage of the river cross-
section relative to that of the smolt trap. Other benefits of using hydroacoustics include
remote sensing of smolts (i.e. no handling stress/mortality), limited obstruction of
waterways, less labor intensive than the aforementioned methods, and the availability of
high resolution temporal migration data. Also, while the system does require daily
monitoring, deployments can be set up to allow remote monitoring of the system and its
components.

Fixed-location riverine hydroacoustics have long been used in a wide range of
stream sizes to measure spawning migrations of adult salmon (Johnston and Steig 1995;
Ransom et al. 1998). It has also been shown to be an effective method of estimating the
abundance of salmon migrating upstream (Enzenhofer et al. 1998; Ransom et al. 1998).
Therefore, we propose to use split-beam hydroacoustic technology deployed in a fixed

location to estimate the abundance of out-migrating smolts. We have chosen split-beam

technology over single and dual beam technology because of the superior fish tracking

capabilities of the split-beam system.

Methods

Study Site

The Muskegon is located in north-central Michigan and flows westward to Lake
Michigan. The river is 212 miles long and drains an area of 2,350 square miles. Croton
Dam, a hydroelectric facility at the town of Croton, restricts passage of potarnodromous
fish to the lower 45 miles of the Muskegon.

The Muskegon River fish community below Croton Dam is comprised of cold,
cool, and warmwater fishes. This stretch supports good sport fisheries for walleye,
brown and rainbow trout, chinook salmon, and steelhead (O’Neal 1997). Annual
stockings of chinook salmon in the lower Muskegon have ranged from 97,000 to 530,000
since the program’s inception in 1967 (O’Neal 1997, www.michigandnr.com). Over the
last five years, stocking rates have not exceeded 150,000 (www.michigandnr.com).

The Muskegon River was selected as the river of study due to its large natural
contribution to the Lake Michigan salmon population. The Muskegon was found to be
one of the most significant producers of chinook salmon in the Lake Michigan Basin
(Carl 1980; Carl 1982; O’Neal 1997). Average annual smolt production in the Muskegon

is thought to be approximately 350,000 (O’Neal 1997).

10

Site Selection

Site selection for the acoustic deployment was based on several criteria, which
included prOxirnity to the smolt trap, current velocity, cross-sectional profile, and
security.

The smolt trap was deployed from the B-31 bridge crossing of the Muskegon on
the Muskegon/Newaygo County line (Figure 1). Mortality incurred along the river by the
out-migrant smolt population necessitated minimizing the distance between the trap and
the acoustic deployment. This ensured maximum comparability between passage

estimates derived from trap and acoustic data.

Current velocity is an important factor in the success or failure of a riverine
acoustic survey. Very high flow rates tend to entrain air bubbles, decreasing the signal-
to-noise ratio, and thus decreasing the probability of detecting a target and the precision
and accuracy of angular position estimates (MacLennan & Simmonds 1992; Keiser et al.
2000). Rapid flow can also result in targets passing through the acoustical beam quickly
so that multiple detections of the same target are unlikely. Multiple detections are
necessary to determine the target’s direction of travel and velocity, which are crucial for
filtering fish tracks by direction of travel. In very slow flow conditions, sedentary
behavior of non-target species is likely to occur in areas of low flow and would likely
bias smolt estimates high, due to the inclusion of non-target fish traces (Ransom et al.
1998)

The cross-sectional profile of each site was qualitatively measured using a
Humminbird® 200 kHz fish finder. The most desirable channel characteristics are a

narrow channel with gradual gradation of smooth strearnbed to the

11

 

.§Eo_§ an as oesaoeaco sass 33% 33a .33 a saw

mitlt... r: .. L . .r

mad 0
eoSEQE .8 28w

“55.632... 8:. o

z €058.80 9.8.63. D

12

thalweg (Ransom et a1. 1998). The streambed must slope down to the thalweg at slightly
greater than the angle at which the sound beam opens on the vertical axis. Streambed
slopes greater than the vertical angle of the sound beam will result in a smaller proportion
of the channel’s cross-section being sampled. If the angle is less than that value, the
substrate will interfere with the beam creating long echo traces at the point of
intersection. Objects jutting up from the streambed into the beam will also return an echo
(Ransom et al. 1998). While echoes created from these issues are easily identified due to
their long traces on the echogram, they must be excluded from the analysis. As a result,
any targets passing through the beam at that range would also be excluded potentially
biasing passage estimates low.

. Given the extensive recreational use of the river, site security was also a concern. .
To minimize the likelihood of vandalism or theft of equipment, a site was sought in close
proximity to multiple homes.

Two separate sites were used for the 2001 and 2002 sample seasons. Each year,
after obtaining landowner permission for the selected location, a 4’ x 7’ storage shed was
constructed to house the land based portion of the acoustic equipment. In 2001, a
location meeting these criteria was found approximately 1.70 km downstream of the B-31
bridge crossing (Figure 1). In 2002, an alternative location 0.27 km downstream of the
B-31 bridge. Both sites were similar in usable width, maximum depth, and cross

sectional profile (Figure 2).

13

 

Range from transducer (m) l
0 5 10 15 20 25 30 35

 

E —w
J: — - — 2001 .
§ —2002
Q

 

 

 

 

2.5

 

 

 

Figure 2. Cross-sectional profile of acoustic sites used in 2001 and 2002.

Data collection

We used a 200 kHz split-beam (3° x 6° elliptical beam) digital echosounder
(Biosonics DE 6000) to measure acoustic size, swimming direction and velocity, and
numbers of fish passing the plane sampled by the echosounder. The echosounder had a
225.5 dB source level (re uPa @ 1m), pulse duration of 0.4 ms and was set at 7 pings per
second. The transducer was mounted on a dual axis rotator with the rotator mounted on a
We constructed of 1/2 in. galvanized pipe (Figure 3). The frame, rotator, and
transducer were placed close to the riverbank ..
in approximately 1 m of water. The rotator
was operated using a control panel (located in
the shed) to aim the transducer towards the

center of the stream and perpendicular to the

     

Figure 3. Pipe frame upon which the
transducers were mounted.

 

main flow. The vertical aiming of the
transducer was tuned to maximize range and

volume of water sampled.

The system was checked daily in mid to late afternoon. At this time, data were
transferred from the laptop to a 120 GB portable hard drive. A deep cycle 12-volt marine
battery was used to power the system. The battery was changed daily with a fully
charged replacement. The dual axis rotators were also readjusted minimize the signal to
noise ratio in the first 20 meters of range.

In May 2001, a major flooding event delayed permanent system deployment until
6-June. Data were collected without significant interruption through 24-June. In all
approximately 30 GB of hydroacoustic data were collected.

In 2002 system malfunction and the subsequent repair delayed data collection
until 25-May. Data were then collected from 25-May until 9-June, when further
equipment failures resulted in a 17-day postponement. During this period, the river
receded substantially, thus requiring the shed to be relocated before data collection could
resume. To avoid the risk of obstacles imposed by future stage changes in the river, we
secured the shed in a 22 it. johnboat and secured the johnboat using a 3-pt mooring. Data
acquisition resumed with this new deployment on 26-June and continued until the system
was removed on 2-July.

To limit the volume of data collected, a systematic sampling strategy was adopted
for the 2002 season. The echosounder was configured to collect data for a two minute
burst. The system would then rest for four minutes between bursts. As a result, we
collected 20 minutes of data per hour during the 2002 season. '

Additional technical difficulties were encountered during data collection in 2002.
Repeated failures of the dual axis rotators and occasional battery depletions resulted in

further reductions in the amount of quality data collected. In all, quality data were

15

collected for only seven “full” days and six partial days. This translated to approximately
13 GB of hydroacoustic data.

A VEMCO Minilog TR was deployed prior to commencement of acoustic data
collection in 2002 to provide stream temperature data necessary for data file processing.

The Minilog was configured to record the temperature at 10-minute increments.

Beam mapping

An in situ beam mapping procedure was conducted to determine the extent and
location of coverage of the river cross-section. During this process a 38.1-mm tungsten
carbide reference sphere was passed through the acoustic beam on a plane corresponding
to 0.9 meters below the water surface. Following the same procedures used to process
field: data, single target data were exported from Echoview, a hydroacoustic data
processing software package, and analyzed. The two-dimensional (range and vertical)
location of 63 echoes was plotted in relation to the transducer and beam axis. The range
of these echoes varied from 10.68 to 19.73 meters. Least squares regression was applied
to the range and vertical distance off axis data to determine the location of the line
marking the intersection of the 0.9-meter plane and the vertical plane of the beam axis.
The regression was constrained to ensure that when plotted relative to the river’s cross-
section, the position of the transducer was accurate relative to the 0.9-meter plane that the
reference sphere was passed through. The beam axis was then geometrically rectified
relative to the surface of the water. The upper and lower boundaries of the beam (:1 .5)
were then plotted to show the coverage of the top l-meter of the river. The proportion of

the cross sections coverage was calculated for four 5-meter increment totaling 20 meters

16

from the location of the transducer. These proportions allowed for adjustment of passage
estimates based on differences in coverage at varying range. Data from this procedure

was also used to calibrate the acoustic data during the processing phase.

Data processing

Data were processed using SonarData Echoview® Version 3.0.80
(www.sonardata.com). The same general procedure was used for all aspects of
processing. Raw data files were opened in Echoview. Split-beam data files
automatically have three variables associated with them upon opening. These are TS,
Angular Positions, and Sv variables. When analyzing single target data, the former two
are the variables of interest. Each of these variables was calibrated with data collection
settings, sound speed, . and the absorption coefficient. Following calibration of the raw
variables, a single target variable was created using Echoview’s Single Target Detection
Algorithm (STDA).

The STDA is a process used to filter echoes based on user-defined criteria. These
criteria include: Target Strength thresholds, Pulse Length Determination Level (PLDL),
minimum/maximum normalized pulse width, maximum beam compensation, and
maximum standard deviation of major and minor-axis angles (Table 1). Echoes passing
all criteria are stored in the single target variable for further analysis.

Target Strength (TS). TS, or acoustic size, is a measure of the reflectivity of an
individual target and is measured in decibels (dB). Reflectivity has been correlated with

length inanumerous species of fish. From length/T S relationships we can estimate the

17

.Eme 0:383an 9:393 new 963005 5 com: mo:_m> new £29:qu Lo El. .F cam...

 

 

 

 

 

 

 

 

 

 

 

 

 

 

N Ammcav 9on5... comEom awe 83::me
. mé mo_mc< m_x< 8.6.2 Co .505 BE
N Ammcav x09... m c_ mmcE Co .onEaz E:E_c__2
N 59... m c_ fleece 065m .6 59:32 82552
n... mm_mc< m_x<,.oc_s_ ho 5.60.5 BS.
c 32925 ace 9.5 .
o 93925 we .
me o6 cozmmcanoo Emom E:E_xm_2
o4 82925 85m
on 82%; m3. .252
mé 59.3 3.3.“. umN=m§oz E:E_xm_2
on 839m; $2 8.55.
E ed 8951 £54 £5. .m_x< mag coacmaxm mEn. veg—2
ad 593 main. uoN__m§oz E:E_c_s_
E mo Aomcmmv :o_m=_oxm mczmw mmcmm
E 3 E5 .55. .m_x< as: 8590 958 Beam
me o.o .93 Sang—€98 59.3 0min.
88.0 seam .m_x< .52 .m_x< as: 2mm
80; 39mm .82 .52 .m_x< as; 932 me o. B- 2885 we
m:_m> .muoEEma o:_m> .295ch

 

 

 

 

 

836829“ :ozooaoo 39a... 29.5

 

882:2?— Efitouz use—out. not

 

 

l8

length of a fish creating an echo that we observe or filter echoes based on the expected
length of a target species.

Length/side-aspect TS relationship for chinook salmon smolts for use with a 200
kHz system are currently unavailable. Efforts to determine such a relationship as part of
this project failed due to laboratory tank limitations. However, Lilja et al. (2000)
developed a side-aspect TS relationship for brown trout Salmo trutta (29.0 - 63.0 cm)

and Atlantic salmon Salmo salar (30.0 - 119.0 cm) for a 200 kHz system:

TS = 26.2 * Log10(TL) - 73.8

where TL is total length in cm. While the lengths of fish used to create this relationship
were substantially larger than those observed in this field study, no other side-aspect TS
relationship for any salmonid was found for use with a 200 kHz system. Therefore, this
relationship was used for TS thresholding in this study.

PLDL & Min/Max. Normalized Pulse Length. Pulse length is the measure of the
duration of a sound pulse in ms. The normalized pulse length is the length of the
- received echo pulse divided by the transmitted pulse length (Echoview® 3.0.80 Help
File). This value is used to ensure the echo received is fi'om a single target. Echoes with
pulse lengths shorter than the propagated signal are noise. Echoes with pulse lengths
substantially larger than the propagated signal are due to multiple targets at similar ranges
and should not be included in the analysis (MacLennan & Simmonds 1992). Echoes with
a Normalized Pulse Length between 0.8 and 1.5 were deemed to be of adequate quality

for further candidature. The PLDL is simply the number of decibels down from the peak

19

that an echo’s pulse length is measured on the oscilloscope '(Figure 4). The PLDL was
set to 6 dB. This value is high enough to ensure complete formation of the echo envelope
but low enough to allow targets to be detected where background noise might interfere
with detection.

Maximum Beam Compensation. As targets move away from the acoustic axis,
beam pattern effects reduce the amount of sound that they reflect. Split-beam
hydroacoustics can account for this loss through beam compensation. By limiting the

beam compensation, low quality targets from beyond the nominal beam angles can be

 

 

Echo TS

 

PLDL .

TS (dB)

 

-- ---—--------—- -------I

 

 

I Pulse length , I

 

 

 

Time

 

 

 

Figure 4. Basic descriptive measurements of a single echo.

20

excluded from the analysis. Difficulties in detecting single targets led to slight loosening
on this criterion. This value was set at 6 dB for this analysis.

Standard Deviations of Axis Angles. Each single target passing the above criteria
will have a value describing level of precision associated with its estimates of minor and
major axis angles. The precision of these estimates is highly affected by the presence of
background noise in the data. Therefore, in a noisy environment like we have in riverine
systems, the standard deviations of the angular position estimates will be relatively high.
In spite of this, a value of 1.5 was selected for each of these parameters. This value was
the lowest value at which adequate numbers of single targets could be detected for fish
tracking analysis.

After all single fish targets were identified; a Fish Tracking Algorithm (FT A) was
applied. The FT A analyzes each single target variable ping by ping. When the algorithm
encounters a single target, a fish track is opened. Echoview then searches for echoes that
are likely to be associated with the opening target by creating an ellipsoid to predict the
3-dimensional location of the target on the next ping. If one or more targets fall within
that ellipsoid, one can be added to the track. If multiple targets are identified within the
ellipse, weighting parameters determine which target is added to the track. These
weighting parameters include major and minor axis range (i.e. the proximity to the
predicted location of the next sequential target), TS, and ping gap (i.e. the number of
sequential pings without an echo). Once the new target is added, the algorithm repeats
this process for the next ping. If no target is identified, the algorithm moves to the next
ping, increasing the dimensions of the ellipsoid by a user-specified percentage. This

process continues until the maximum ping gap parameter is exceeded. When the

21

maximum ping gap has been reached, the track is closed. Tracks including a number of
targets and pings greater than or equal to the minimum number of targets and pings
required for a track are defined as fish track regions. Tracks not meeting the criteria are

discarded.

Performance Analysis

Performance of the FTA was analyzed to pararneterize the single target detection
and fish tracking algorithms, to qualitatively evaluate the performance of the software,
and to provide baseline results that could be applied to the overall processing results to
estimate smolt total passage. The process involved:

1.) Trace definition

2.) Data processing

3.) Evaluation of processing results

4.) Adjustment of algorithm parameters

Trace definition. A trace is defined as series of echoes from a sequence of pings
that seem to progress in a predictable manner. Since a trace is made up of multiple
echoes, the data were evaluated at the trace and echo levels concurrently. Thirty raw data
echograms were randomly selected fiom the data collected from May to July of 2002 and
fish traces were visually identified based on characteristics of the echoes and various
aspects of trace morphology including pulse width and TS. Pulse width was evaluated
in a manner similar to that portion of the STDA. The TS threshold on the echograms was

set to exclude echoes from targets too small to be out-migrant smolts after adding the

22

maximum beam compensation. Echoes with TS values falling below this threshold are
not visible on the echograms. Thus, they were not identified in the trace definition
process. Echoes meeting the visual inspection requirements were considered candidates
for trace definition.

Complexities associated with integrating three-dimensional data into the trace
definition process prevent filtering of traces by direction of travel. Thus, no
consideration was given to angular data in the trace definition process of the performance
analysis.

The first range consideration is the maximum change in echo range on sequential
pings. Out-mi grant chinook generally travel downstream with minimal lateral movement
(relative to the direction of flow; Stables & Kautsky 2000). In the trace definition
process, echoes showing a high degree of lateral movement relative to each other were
only defined as traces if they exhibited a high degree of predictability in that movement.

Range is also used in conjunction with temporal data in determining the minimum
number of pings required for formation of a trace. The minimum number of echoes that
can constitute a trace is two. The presence of two or more echoes allows for estimation
of distance traveled estimates, swimming speed, and direction of travel.

In addition to the minimum number of targets required, there is also a maximum
number of targets allowable. Traces in excess of 20 pings (approximately 3 seconds) are
unlikely to be smolts exhibiting migratory behavior and were not defined as traces. This
type of trace is commonly caused by shifting of the river substrate.

Sequences of echoes that met the aforementioned guidelines were outlined using a

polygon. This polygon served as a marker region for spatio-temporal comparisons

23

 

 

 

 

 

 

 

 

 

 

 

 

Automated
Processing
Results

 

 

 

 

 

 

  
   

 

 
  

 

 

 

 

 

 

 

 

 

 

N° Visual
Inspection
Results

Fish

Track

Expected?
1

Yes No

 

 

 

 

 

   

  
   
 
   

What
Failed?

Single Target
Detection
Algorithm

Tracking
Algorithm

 

 

 

 

 

Yes
.nm‘g‘m Created?
Results
Fish
Track
Expected?
Yes No
Track
Correct?
What
Y” NO Failed?
4. 7.
or 9
Process
- User
. Autgrrp‘altion Error

 

 

 

 

 

Figure 5. Decision tree for assigning result codes to each track or anticipated

track.

24

 

 

between the expected fish tracks identified in the trace ID process and observed fish
tracks from the application of the software’s algorithms.

Data processing. Data Processing was accomplished in the manner previously
outlined. Following the complete inspection and identification of all suspected traces
within the raw echogram, the STDA and the FTA were applied to each data file
individually.

Evaluation of results. Following the application of the FTA to the single target
variable, the raw echograms with the trace polygons and fish tracks overlaid, were
inspected to evaluate the performance of software algorithms relative to the expectations
defined during the trace definition process. Process result codes are assigned to each
user—defined polygon indicating a potential trace. Result codes are also assigned to any
fish tracks that were created on areas of the echo gram where tracks were not anticipated.
The total number of occurrences for each code was then determined for each data file,
and the results were evaluated relative to visual expectations (Figure 5). The list of
possible result codes included the following:

0 Code 0 — A fish track was correctly identified within the trace polygon. This
indicates adequate performance of the entire process relative to the visual

\

inspection by the user.

0 Code 1 — The STDA failed to identify multiple targets within the trace polygon
created by the user. Therefore, no track could be created.

0 Code 2 - The STDA identified multiple targets within a trace polygon, but a fish

track was not created due to poor STDA performance. This code is generally

25

reserved for longer traces where few targets are identified and the maximum ping
gap is exceeded

Code 3 — The STDA identified multiple targets within a trace polygon, but a fish
track was not created due to poor FTA performance. This code applies to traces
that have multiple targets within the maximum ping gap, but the FTA fails to
create a track based on target exclusion.

Code 4 — The FTA created multiple tracks within a single trace polygon. This is
usually due to the maximum ping gap being exceeded between two groups of
single targets within the same trace.

Code 5 — The FTA created a track in an unanticipated location using targets that
seem to be unrelated. Use of this code is based on the user’s visual inspection of
the raw data and overlaid fish track following data processing.

Code 6 — The FTA created a track in an unanticipated location. However, in this
case, the user failed to identify a potential track during the initial inspection of the
raw data echogram. Therefore, the FTA performed correctly.

Code 7 - The FTA created a fish track in which non-trace targets were selected
over trace targets occurring on the same ping.

Code 8 - The FTA failed to create a fish track due to the exclusion of targets
because of borders (i.e. temporal bounding, bad data, etc.).

Code 9 - The FTA created a fish track in location as expected with multiple
targets within the trace polygon. However, targets external to the trace polygon

were also included.

26

In each file, result codes were recorded for each trace outlined in the trace
identification step and all fish tracks created by the software that were not deemed to be
associated with an outlined trace. Proportions of the total number of result codes were
calculated for each individual result code. The arithmetic mean was calculated for each
code’s proportion across all data files. The mean proportions were used to determine
what algorithm failed most often and why the perceived failure occurred.

These ten result codes were also divided into three broader groups to simplify
performance evaluation. These categories were: Good (codes 0 & 6), Bad (codes 4, 5, 7,
& 9), and Omitted (codes 1, 2, 3, & 8). “Good” results occur when the algorithms
properly identify a fish track. “Bad” results occur when fish tracks are created from or
include what appear to be noise echoes. “Omitted” results occur when a trace is
identified during the visual inspection process, but no fish track is formed. Proportions of
the total number of result codes for each category were calculated for each data file.
These proportions were used to evaluate the general performance of the entire process
and for comparison across different algorithm parameterizations.

Glaring differences in data quality at different ranges were apparent. Thus, the
data were binned out into four 5-meter increments. Names of these increments
correspond to their maximum range (i.e. 10 to 15 meter increment is called the 15 meter
range bin). ANOVA was used to determine if there was a significant effect of range bin

on the numbers of categorical results.

27

Iterative Parameter Perturbation

Results obtained fi'om the above processes were used to modify parameters in a
feedback loop to improve the performance of the FT A. While time constraints prevented
most parameters to be iteratively tuned, the process was used to adjust the “range gating”
parameter. This parameter restricts the maximum allowable change in range between
sequential targets in a track. The starting value was 0.3 meters. Iterations subsequent to
the initial run involved perturbations of 0.1 meter. The proportions of categorical results
were compared across iterations. The overall goal was maximize the proportion of good

results while minimizing the proportion of bad and omitted.

. Passage Estimates

Passage estimates for days with more than 10 hours of data fi‘om both years were
‘ obtained using the parameters from the performance analysis. Each day was processed
using the procedure described in the data processing section above. Following the
completion of the data processing, fish track regions were exported fiom Echoview. Fish
track data that were exported and used in analysis include: date, time, mean TS, mean
target range, horizontal direction of travel, and tortuosity.

Additional filtering was applied to the processed data to determine is targets were
likely to be out-migrating chinook smolts. Studies regarding out-migrant smolt behavior
- provide some guidance on selecting filtering criteria. Passive drifting and active
swimming are important components of smolt migration behavior (Fried et al. 1978;
Fangstam 1993; Lacroix and McCurdy 1996). Fangstam (1993) observed that active

swimming was only used by migratory fish approximately 10% of the time.

28

Additionally, active swimming speeds of 0.11 to 0.23 m/s are predicted for smolts of the
size produced in the Muskegon River. Swimming speeds in this range are not
discemable from passive drifting due to the poor precision of angular position estimates
for individual echoes. As a result, swimming speed was not used in the filtering process.
Another filter considered was swimming direction. Movements of migratory
smolts are strongly oriented towards their destination (Groot 1972; Fangstam 1993;
Stables & Kautsky 2000). Generally migratory smolts orient themselves parallel to the
current of the stream during out-migration (Stables & Kautsky 2000). Therefore,
direction of travel is likely to be a good filtering criteria. The variable providing the most
information regarding the direction of travel relative to up/downstream is “horizontal

direction”. This variable is measured in degrees and is determined for each track by

 

drawing a straight line between the first and last
target in the track. As deployed, objects

moving from upstream (right) to downstream
270 90

 

(left) while looking across the river from the
transducer position exhibit a horizontal
direction of 270° (Figure 6). Filter thresholds
Flow for each day-range bin combination were

. determined by creating histograms of horizontal

direction. Binning of the histograms was set to

 

 

5° increments. The increment exhibiting the

 

Figure 6. Geometry of direction of most fish tracks was determined to be the
travel measurements in the

horizontal plane (aerial view). direction of flow. Each day-range bin

29

combination was then filtered to exclude targets outside i45° of this value. Tracks
passing this filter are assumed to be moving primarily in a downstream direction.

Tortuosity is another variable dealing with orientation and the deliberate
movements associated with migratory behavior. Tortuosity (unitless) is the sum of the
distances in a track divided by the distance from the first to last targets in the track,
measured in 3 dimensional space. Tracks with only two targets in them have a tortuosity
of 1. Tracks with three or more targets are likely to have a tortuosity greater than 1. The
deliberate movements associated with migratory behavior are unlikely to yield
' tortuosities substantially larger than 1. Thus, tracks with tortuosities greater than 1.5
were filtered from the results.

The final filtering criterion used was mean TS. The side-aspect length/T S
relationship developed by Lilja et al. (2000) for trout was used to calculate daily filtering
values. The range of lengths for chinook smolts collected in the auger-trap was used for

daily upper and lower TS thresholds (T able 2). The trap-catch on 26-May

 

 

 

 

 

 

 

 

 

 

*lncludes minimum and maximum chinook length from the
previous and following day.

Table 2. Minimum and maximum length of smolts
captured in the trap survey (Rutherford, unpublished

data) and the corresponding TS threshold values.

Chinook Lengths (mm) Chinook TS (dB) consrsted 0f only three
Date Minimum Maximum Minimum Maximum chinook. Thus, the minimum
SIZE/2002‘ 58 86 -53.80 -49.32
5/29/2002 64 101 -52.68 47.49 and maximum length of
5/30/2002 52 90 -55.04 48.80 . .

ChlIlOOk from the prevrous and

5/31/2002 65 90 -52.50 -48.80
6/5/2002 52 88 -55.04 49.05 subsequent days were used.
6/6/2002 54 83 -54.61 -49.72
617/2002 56 96 -5420 48.06 The “Ported fiSh

tracks were filtered according
to the criteria description

above. Following the filtering

process, total numbers of fish tracks passing the criteria for each day were summed. The
temporal and spatial distributions of tracks exported were used to calculate whole river
passage estimates for each day.

To estimate total number of fish out-migrating, the number of tracks detected for
each S-meter increment was multiplied by a constant to estimate the total number of
tracks that would be detected if the top meter of the water column were completely
ensonified for that range. Daily track totals were calculated for the following ranges: 0 to
5 m, 5 to 10 m, 10 to 15 m, and 15 to 20 m. An estimate for the entire river cross-section

was also calculated by extrapolating the 0 to 15 m results to the corresponding distances

 

 

 

 

 

 

 

 

 

 

 

 

 

from the opposite shore (Figure 7).
Range from transducer (m)
0 5 10 15 20 25 30 35
A B C D = C = B = A
a:
’9’.-
D \ /
\¥
D
Range of Data Collection Whole River Passage = 2A +23 + 2C +D

 

 

 

Figure 7. Rationale behind calculation of whole river estimates.

Since less than 24 hours of data were collected on each day, summation of the
number of tracks observed would yield daily passage estimates that were biased low due

to non-continuous temporal coverage. Thus, it was necessary to calculate a mean hourly

31

passage rate for each day to arrive at a total daily passage estimate for comparison to the
results of the trap study.

Data limitations constrained the statistical analyses that could be done. The
statistically low number of days for trap and hydroacoustic estimate comparison dictates
that a very simplified regression model be used. Thus, simple linear regression was used
to determine the relationship between hydroacoustic and smolt trap estimates.

Mean normalized hourly passage rates across all days were calculated for each
year to determine if any detectable diel migration pattern existed. These quantities were
calculated by dividing each hourly passage estimate by the maximum hourly passage for
the corresponding day. The possible range of resulting values was 0 to 1, with 1
representing the hour of maximum passage for each day. Results for each hour of the day

(i.e. 0100, 0200, etc.) were then averaged across all days to determine if a trend existed.

RESULTS

Beam Mapping

The transducer was aimed downward approximately 2° to maximize the
signal:noise ratio at extended ranges. Accounting for the vertical beam angles, the beam
sampled 13% ofthe top 1 mat the 0 - 5 m range, 38.61% for the >5 - 10 m range, 44.62%
for the >10 - 15 m range, and 40.46% for the >15 - 20 m range (Figure 8). Height of the
beam (vertical distance ensonified) increased linearly with range, while volume sampled
within the upper 1 m only increased slightly between the second and third range

increment and decreased from the third to the fourth range increment.

32

 

 

 

 

------- 3foot1ino
—— — Beam Isis

 

 

 

 

 

 

 

 

Distance from Transducer (m)

 

Figure 8. Cross-sectional coverage of the ensonified region.

33

 

Performance Analysis

There was a statistically significant effect of range increment on numbers of Good
(ANOVA, p < 0.0001), Bad (ANOVA, p < 0.0001), and Omitted (ANOVA, p < 0.0001)
tracks per 1000 m3 (Figure 9). Substantial decreases in the density of Good and Omitted
results with increasing range were evident, while only a slight increase was observed in
Bad results was observed

A total of four runs of the individual parameter perturbation were necessary to
optimize the “range gating” parameter of the FTA. Only slight decreases in total
numbers of results were observed as the range gating parameter decreased (Figure 10).
Since the loss of results was minimal across iterations, valid comparisons of ratios for
categorical results could be made to evaluate the software’s performance. As range
increased, the range gating parameter had an increasingly positive effect on the ratio Bad
to Good results (Figure 11). Little effect of the perturbation was observed on Omitted to
Good results across iterations (Figure 12).

Proportions of result code occurrences were used to evaluate the performance of
the STDA and the FTA across range bins (Table 3). Result codes 0 and 6 indicate good
performance of both algorithms. In general, these codes decrease with increasing range.

Codes 1 and 2 indicate perceived failures in the STDA due to a lack of target

 

 

 

 

 

 

Range Percentages of Result Code Occurrences
Bin 0 1 2 3 4 5 6 7 8 9
5 30.8 53.4 0.0 0.0 0.0 0.0 15.7 0.0 0.0 0.0
10 21.5 58.1 0.7 0.2 0.1 1.5 17.9 0.0 0.0 0.1
15 15.2 61.5 1.6 0.0 0.0 10.0 11.4 0.0 0.1 0.1
20 12.7 39.0 2.6 0.0 0.8 28.7 16.1 0.0 0.0 0.1

 

 

 

 

 

 

 

 

 

 

Table 3. Percentages of each result code's occurrence by range bin (0.1 m iteration of
range gating manipulation).

34

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

8 3‘
E
8 2.5 ~
2
t 2.. lGood
a
.g IBad
g 1'5 “ ClOmitted
a 1-
to
Q
I! 0.5 -

0 T r -

Oto5 5to 10 10to 15 15to 20

Range Bin (m)

 

Figure 9. Number of categorical results per volume sampled by range
bin (0.1 m iteration of range gating manipulation).

35

 

 

 

1 000

 

900 4
800 ~
700 -
600 -
500 ~
400 -
300 -
200 —
100 -

# of Results

 

02

 

 

 

0.4

Range Gating Parameter Value (m)

 

 

 

0.3

 

 

 

0.2

 

 

lows
I5to10
D10to 15
E115t020

 

 

 

 

0.1

 

Figure 10. Absolute numbers of results from the four range gating manipulations.

36

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

3
2.5 _
2- . I0m5
a ”T I5m10
1.5 ~ ‘ f D10to15
2 , Li: [315l020
1~ - I '. —1
05" i . i J-[_lz “N
0 _ :.;' I I I I
0A 03 02 DJ

Range Gating Parameter

 

Figure 11. Ratio of bad to good results for each iteration of the range gating
parameter manipulation.

37

 

 

 

 

 

 

 

 

 

0.4 0.3 0.2 0.1
Range Gating Parameter

Figure 12. Ratio of omitted to good results for each iteration of the
range gating parameter manipulation.

38

 

 

I0t05

l5to 10
D10to 15
E115t020

 

 

 

 

detection. Loosening of STDA criteria reduced these percentages. The percentage of
these codes increased to the 15-meter range bin and decreased slightly at 20 meters. Code
3 indicates an FTA parameterization that was too restrictive to create a track associated
with an identified trace. Code 3 was only observed twice in the entire performance
analysis. Codes 7 and 9 generally indicate perceived failure caused by a loosely
parameterized FTA. No code 7 occurrences and a low percentage of occurrences for
code 9 was observed in the performance analysis. Codes 4 and 5 can indicate
shortcomings of both algorithms and generally appeared to be a combination of STDA
and PTA failures. A restrictive STDA and FTA characterize code 4, while code 5

indicates a loose parameterization of both. Occurrences of code 4 were sparse. Code 5

 

 

increased with increasing range. Finally, .

Acoustic Trap

. . . Passa e Passa e

code 8 rs an artifact of data bounding and Date Estimagte Estimagte
. 6/8 9,028 4,000
was only observed once in the performance 5,10 7,554 3,351
. 6/11 5,773 3,135
analysrs. 6/12 7,707 3,838
‘- 6/13 7,287 3,189
8 6/14 5,945 5,459
. N 6/19 3,781 1,243
Passage Estrmates 6/20 5,400 1,838
_ . 6/21 2,610 919
In 2001, acoustic daily passage 6122 6,155 1,405
. . 6/23 14,791 865
estimates ranged from a maxrmum of 5,25 9,319 317
, , 5/29 15,991 556
14,791 on June 23 to a nnmmum of 2,610 N 5,30 13,330 1,349
. . 8 5131 12,732 3,095
on June 21 (Table 4). Acoustic darly N 5/5 20,757 1,825
. 6/6 13,062 5,793
passage estimates from 2002 ranged from a 6/7 20 232 4,920

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

maximum of 20,767 on June 5 to a Table 4. Daily smolt passage estimates
from acoustic and trapping studies.

minimum of 9,319 on May 26. Across

39

 

 

 

 

 

a 15000 -_—1
'6
E 2001
9 12000 «
.9
I“
.§
3 9000 — .
1D .0 0
D)
g 8000 - .. , ,
3
o. . = +
.g 3000 4 . y 0.02537x 6778.1
3 R =0.0006
O
0
< 0 I I
0 5000 10000 15000

Trap Passage Estimate (Smolts)

 

 

24000

 

20000 - ’ 0 2002

18000 - /

12000 ~

8000 - y = 0.4889x + 13829

2 _
4000 - R -0.0622

Acoustic Passage Estimate (Smolts)
0
O
O

 

 

 

0 1 1
0 8000 16000 ' 24000

Trap Passage Estimate (Smolts)

 

Figure 13. Regression results from comparisons of daily passage estimates from trap
and acoustic surveys.

40

 

 

 

Acoustic Passage Estimate (Smolts)

 

 

20m 1 y = 0.882x + 3630.9
R2 = 0.4453

 

 

 

0 . . . .
0 2000 4000 6000 8000 10000

Trap Passage Estimate (Smolts)

 

 

Figure 14. Comparison of daily passage estimates from trap and acoustic
surveys from 2001 with an outlier removed.

41

both years, acoustic daily passage estimates ranged from slightly greater than 1 to nearly
30 times the estimates yielded by the smolt trap study. In general, daily estimate
comparisons between methods from 2002 were much more variable than those from 2001
(Figure 13). While the comparison of 2002 daily acoustic estimates and trap estimates of
smolt passage appeared to be positively correlated, the correlation was not statistically
significant (ANOVA, P = 0.59; Table 5). There was also no significant correlation
- between the acoustic and trap estimates from 2001 (ANOVA, P = 0.94; Table 5).
However, with the removal of a single outlier (23-June) from the 2001 comparison, a
statistically significant positive correlation was achieved (ANOVA, P = 0.035; Table 5).
The decision to exclude this data point was made solely on its position relative to the
other days’ comparisons. No apparent change in discharge or data quality was observed.

The slope fiom the regression equation resulting from the 2001 comparison with the

 

 

 

 

 

 

 

 

 

 

 

 

 

P-value '
Year B (i 95% 0.1.) n r r2 (ANOVA)
2001 0.054 (1.599) 11 0.0253 0.0006 0.9410
2001* 0.882 (0.802) 10 0.6673 0.4453 0.0350
2002 0.489 (2.182) 7 0.2494 0.0622 0.5897
Combined* 0.419 (1.727) 17 0.1323 0.0175 0.6128
" outlier removed

Table 5. Correlations and significance from comparisons of daily passage estimates
from trap and acoustic results.

outlier removed was 0.882, and the intercept was 3,630.9 (Figure 14). A final
comparison of the pooled results from 2001 and 2002 (outlier removed) also proved to be

insignificant (ANOVA, P = 0.61; Table 5).

42

 

2001

9.0.0.0
0109x100

9.0.0
MOI-h

Normalized passage rate

Q
'_.s

 

O

0000
0200
0400

885885888

Local Time

 

2002

Nomalized passage rate
D
U'l

99.01:.
AMU‘

 

D

D
D
D
D

0200
0400
0000
0000
1000
1200
1400
1600
1800

2000
2200

 

Local time

 

 

Figure 15. Mean normalized hourly passage rates from 2001 and 2002 acoustic
surveys (shading from approximation of sunset to sunrise).

No obvious trends were detected upon examination of the mean normalized
hourly passage rates for either year (Figure 15). Passage rates did appear to be more
stable during the nighttime hours during the nighttime hours than the day. However, data

were collected for fewer days during the late morning and early afiernoon hours due to

43

battery depletions. The reduced sample size for these hours probably led to the perceived

increase in variability.

Discussion

Gear Comparisons

The Muskegon River is Michigan’s largest contributor of naturally produced
smolts to Lake Michigan. This is supported by numerous surveys since the first work
done by Carl (1982) in the late 1970’s. The electrofishing estimates made by Rutherford
from 2000 and 2001 are comparable to previous estimates of pre-smolt abundance and
lend credence to the smolt trapping methods used in those years. However, concerns still
remain regarding the precision and accuracy of estimates from both methods.

The pre-smolt electrofishing survey used only five sample sites to obtain the
average smolt density that is extrapolated to reach the total pre-smolt abundance estimate.
This low samples size incites questions regarding the precision of the estimate.
Additionally, the sites were not chosen randomly fi'om all possible sites within the
nursery area. While the survey may provide an index of pre-smolt abundance across
years, the validity of the overall estimate is questionable.

The smolt trapping survey is also subject to questions of precision and bias. In
larger rivers such as the Muskegon, a very small percentage of the river’s cross section is
sampled. The magnitude of this percentage varies with changes in river stage. As a
result, in years of highly variable river stage, the probability of capturing a given smolt is
likely to be variable as well. Additionally, auger-type traps are operated by the pressure

exerted by flowing water. As the flow rate decreases, the trap turns slower and provides

44

more of an Opportunity for migrating smolts to avoid capture. Depending on the stage
changes, flow rates, and the timing of the migration, a real potential exists for the daily

estimates to be biased high or low.

Performance Analysis

The results of the performance analysis indicate shortcomings in the parameters
used in both the STDA and the F TA. It should be noted that all data used in the
performance analysis was collected in 2002 when daily acoustic passage estimates were
extremely variable relative to daily trap passage estimates. However, the performance
analysis does provide some insight into the problems of 2002.

Foremost, the perceived inability of the STDA to identify multiple targets in
many visually identified fish traces caused relatively high numbers of omitted results.
The parameters of the STDA were selected to limit the influence of debris such as tree
branches and twigs. Any object passing through the beam is likely to create a trace.
Without perfect knowledge of both target and non-target objects passing through the
acoustic beam, much uncertainty remains regarding the performance of the STDA.
Human error in the trace identification process probably led to inclusion of non-target
debris traces that were in turn filtered out by the STDA. The magnitude of such error
remains a mystery.

Shortcomings of the FTA, while still subject to the same target uncertainties as
those of the STDA, can be more concretely identified. The primary error associated with
the FTA was the creation of tracks where none had been identified visually. These tracks

were often spatially erratic. This irregular pattern was easily identified in the visual

45

inspections, as the change in range from ping to ping was generally unpredictable. This
conclusion is also supported by the results of the range gating parameter perturbation
where the ratio of bad to good results decreased by approximately half at each of the

range bins as the range gating parameter was adjusted from 0.4 m to 0.1 m.

Passage Estimates

The performance analysis results indicate that software was very limited in its
ability detect the tracks identified in the visual inspection. Yet, the daily passage
- estimates derived from the acoustic methods used in this study were still consistently
higher than estimates made by Rutherford. Acoustic passage estimates from 2001 ranged
from 1.09 to 17.1 times higher than the corresponding trapping estimates. However, with
the removal of the outlier, the upper end of that range is lowered to 4.38 (mean = 2.50).
Acoustic passage estimates from 2002 ranged from 2.25 to 29.4 times higher than the
corresponding trapping estimates (mean = 12.8).

The generally low magnitude of the differences between daily passage estimates
from acoustic and trapping methods from 2001 shows some promise for the methods
used herein. Additional optimism can be drawn from the significant correlation between
the 2001 estimates. However, the variable results from 2002 raise concerns for the
immediate implementation of these methods for management purposes. More in depth
analysis of this variability will be necessary. River channel morphology and equipment
limitations led to gross simplifications in calculating whole river passage estimates. The
, channel cross-sections for the acoustic sites used in 2001 and 2002 were asymmetrical.

However, data cable length, decreased ping rate from running multiple transducers, and

46

the use of a single hydroacoustic system, prevented the deployment of transducers from
both riverbanks. As a result, it was necessary to extrapolate results from the half of the
channel nearest the transducer in a symmetrical manner to the far half of the channel to
estimate total passage. The most probable effect of this strategy is an underestimation of
passage on the far side of the river channel where currents are faster and smolts are more

likely to out-migrate.

Fixed River Acoustics: Assumptions and Problems

Successfully executing an acoustic survey such as this relies on two primary
assumptions. The first is that the hydroacoustic system is consistently able to detect the
target species for filtering. This is especially crucial when data on direction of travel and
swimming speed data are necessary. Given the inherent noise associated with riverine
environments (i.e. turbulence, debris, non-target species, etc.), detectability issues are a
significant hurdle to overcome when dealing with small targets such as chinook salmon
smolts. Low signal-to-noise ratio substantially limited the system’s ability to detect
smolts at extended ranges. Therefore in the center of the river channel, an area likely to
pass a large portion of migratory smolts, the probability of detecting a given smolt was
low.

The second assumption is that the data processing phase can effectively remove,
or at least account for, non-target objects without affecting numbers of target traces.
Undoubtedly, this assumption is also violated at some level. Through improved
parameterization the impact of these violations can be minimized but never eliminated.

The regression techniques used in the comparison can account for non-target tracks

47

through the intercept. This implies that non-target track numbers are consistent across
days. When river stage and discharge is stable, this may be a valid assumption.
However, in the days following a rain event, river stage, discharge, and debris load can
all change significantly.

One drawback to the use of hydroacoustic surveys of this type is the need for
ground-truthing data. Hydroacoustics cannot be used as a standalone method for
quantifying smolt passage. Some additional form of verification that identifies objects
passing through the acoustic beam will continue to be necessary. These data can take the
form of video surveillance or trapping. Differences in fish communities between streams
or between stretches within a stream may result in significant variability. Thus, it may be

necessary to reevaluate parameters and deployment configurations to ensure accurate

results.

Conclusions

Each of the methods used to quantify smolt production in the Muskegon River has
problems associated with its use. However, pre-smolt electrofishing and smolt trapping
surveys continue to be the best methods for estimating natural 'smolt production at this
time. While these methods do have biases associated with their use, they have been
documented and can be qualitatively considered when evaluating the quality of the
estimates they provide. Currently, unknowns surrounding the use of hydroacoustic
methods for estimating

This study has shown that fixed location riverine hydroacoustics does have

promise as a method for enumerating out-migrating salmonids. Detection issues may

48

limit the use of hydroacoustics methods for enumerating out-migrant chinook. However,
in low gradient sections of stream where turbulence is reduced, this ability is likely to be
enhanced. This method holds much promise in applications where stream-type,
anadromous salmonids such as coho, steelhead, and Atlantic salmon are targeted.
Species residing in rivers for a year or more following emergence attain much larger sizes
prior to their seaward migration. As a result, they exhibit much larger side aspect TS and

greatly improve the signal:noise ratio that impedes the use of acoustics on smaller

species.

Future Research

Given the results found in this study, fixed-location, riverine hydroacoustics could
potentially become a useful tool for fishery managers concerned with natural salmonid
production in streams. Further development of data collection and analysis methods will
be necessary but also appears to be warranted.

Foremost, development of side-aspect TS relationships for chinook or other target
species would improve track-filtering capabilities. Such advances could translate to
more precise estimates of daily smolt passage if detectability issues can be addressed.

Additionally, as technology improves, investigation into alternate deployment
configurations should also enhance acoustic capabilities. Use of multiple or multiplexing
systems deployed to allow data acquisition from both sides of the channel will increase
coverage of the stream’s cross-section and eliminate the uncertainty of extrapolating

results from one half of the channel to the other. A bottom deployed, stationary, up-

49

looking transducer may also improve coverage and smolt detectability in the center of the
river where bank deployed transducers provide data of marginal quality.

Finally, development of fixed-location, riverine, hydroacoustic survey methods
should be expanded to other streams and rivers in the Great Lakes basin. Although it is
the largest contributor, the Muskegon River is one of many Great Lakes tributaries with
naturally reproducing populations of chinook salmon. To gain acceptance as a legitimate
technique for quantifying smolt production, this method will require testing on numerous

streams throughout the Great Lakes watershed.

50

Chapter 3

The Future of F ixed-location, Riverine
Hydroacoustics in Great Lakes

Fisheries Management

51

Project Summary

This study has shown significant obstacles associated with chinook smolt surveys.
The primary obstacle identified is detectability of chinook smolts and is due to their small
size at emigration. While with improved site selection and deployment strategies
hydroacoustic surveys of chinook smolt passage may be possible, the detectability issue
will continue to complicate the process.

With this in mind, what contributions can this method make towards managing
Great Lakes fisheries? Results from this study and published literature point to several
possibilities. Specifically, riverine hydroacoustics could be applied to quantify spawning
stock during adult salmon spawning runs, provide smolt production estimates of coho and
steelhead, and potentially provide information for management of non-game species.

Adult salmon returns. Fixed-location, riverine hydroacoustics have provided
sound information for use in management of salmonid fisheries on the west coast of
North America for over a decade (Ransom et a1. 1998). Typical applications of the
method are usually implemented to quantify escapement of adults in efforts to maintain
adequate spawning stock biomass. Although rivers of the west coast generally exhibit a '
high gradient and, as a result, are acoustically noisy, the method is well suited for
counting adult salmon. Adult salmon migrations consist of fish that exhibit target
strength (TS) in the range of -30 dB and higher. With appropriate site selection, such a
TS is easily distinguishable from background noise levels. Additionally, the advent of
split beam hydroacoustics allows for three-dimensional tracking capabilities (Ransom et

al. 1998; MacLennan and Simmonds 1993). ,Thus, fish migrating upstream are easily

52

distinguished from drifting debris. These factors enhance the ability of hydroacoustic
methods to quantify adult salmonid abundance in rivers.

With the current state of technology, hydroacoustics will be best applied to
quantifying adult returns to natal streams. Results from adult passage surveys can at
minimum be considered a reliable index of abundance. With the prior development of
‘ stock-recruitment models for streams making significant natural smolt contributions,
adult returns can be used to make preliminary estimates of smolt production nearly six
months in advance of stocking. This information combined with the status of Lake
Michigan predator and forage populations can be used to make adjustments to stocking
rates and reallocate hatchery resources for optimal production of all species. Stocking
rates can then be fine tuned with updated information on pre-smolt abundance
immediately prior to the planting of fish.

The benefits of this type of management strategy are twofold. Foremost,
management actions shift away fiom relatively static stocking rates and reacting to
natural inputs the year after they are added to the lake. Instead, lake conditions and
natural smolt contributions are considered prior to stocking, thus, circumventing the time
lag associated with the reactive strategy. As a result, lake conditions should remain more
stable, decreasing concerns of predator overpopulation forage depletion. Secondly,

Smolt production. Substantially less work has been conducted targeting out-
migrating salmon smolts. Obstacles such as small size at out-migration (e.g. low TS) and
direction of travel make distinguishing smolts from noise and other non-target objects
more difficult as noise levels increase. As a result, high gradient streams are not well

suited for this application of the method, and its development has been relatively slow.

53

However, managers working in low gradient streams like the lower Muskegon could
potentially benefit fiom out-migrant work. Such work will probably be limited to
“stream type” species (i.e. steelhead and coho) that out-migrate at much larger size than
chinook.

Development of techniques for enumerating out-migrant coho and steelhead is not
only promising, but also justified. One of the major difficulties encountered during this
study was detectability of the small chinook smolts. The TS corresponding to the
observed range of chinook lengths in the Muskegon was approximately —55 dB to -47.5
dB. The background noise levels observed in this study gradually increased with range
until it exceeded even the high end of this range (Figure 1). Newcomb and Coon (2001)
found that the average length of steelhead smolts emigrating from the Betsie River,
Michigan was approximately 195 mm. Using the TS-length relationship of Lilja et al.
(2001), a smolt this size would exhibit a TS of approximately -40 dB. This is well above
background noise levels and bodes well for steelhead smolt detection (Figure 1).

In addition to improved detectability, coho and steelhead abundance in the Lake
Michigan system also provide support for further development. Northwest lower
. Michigan streams have been known to consistently produce significant numbers of wild
coho and steelhead smolts since the late 1970’s (Carl 1982, Seelbach and Whelan 1988,
Seelbach 1993). Additionally, these species contribute a sizable portion of the Lake
Michigan sport fishery. In 1999, the combined coho and steelhead catch from mandatory
charter boat reporting of trips in Michigan waters was 21,916 compared to 28,689 for
. chinook (Rakoczy and Russell 2003a). In a similar comparison of 1998 reporting, echo

and steelhead catch exceeded chinook catch by more than 2,000 (Rakoczy and Russell

54

Team
40 -30 -20 -10

A A
W v ‘v

 

 

-—- Observed noise level
Chinook TS range

— Steelhead TS (Average length from
Newcomb and Coon 2001)

 

 

 

 

 

Range (meters)

 

 

 

 

 

 

 

 

 

Figure 16. Noise levels typically observed in this study with chinook and
steelhead TS shown for reference.

2003b). At these levels of relative abundance, coho and steelhead can significantly
influence the trophic balance of the lake.

Non-game species management. Just as it is important to understand the
reproductive dynamics of important game and commercial species like salmon and trout,
this information is important for conservation of certain non-game species as well.
Hydroacoustic surveys could potentially aid in the management of other potamodromous

species such as lake sturgeon, which return periodically to their natal streams to spawn.

Conclusions

While the outlook for using riverine hydroacoustics as a method to quantify

chinook smolt production is marginal at this time, several other applications hold promise

55

for the technique. Use of hydroacoustics in quantifying adult returns, smolt production of
alternate species, and non-game species management is feasible and could be

implemented with a reasonable amount of deve10pment.

56

APPENDIX

57

 

% of Filtered Tracks

 

 

 

 

 

 

 

 

 

 

 

 

58

Date Rama Bin Total Tracks Number Lower TS Upper TS Direction Tortuosity Process
618101 0 - 5 618 346 0.0 84.1 13.9 2.0 0.0
5 - 10 1,622 1,350 0.0 74.7 22.7 1.7 0.8
10 - 15 9,832 8,708 0.0 77.3 18.4 2.6 1.7
15 - 20 13,638 13,000 0.0 91.2 7.1 0.8 0.9
6110101 0 - 5 367 215 0.0 99.5 0.0 0.5 0.0
5 - 10 2,897 2,421 0.0 86.5 11.8 1.7 0.0
10 - 15 12,506 11,487 0.0 82.6 14.9 2.4 0.0
15 - 20 9,712 9,464 0.0 95.9 3.8 0.4 0.0
6111101 0 - 5 978 720 0.1 78.9 20.3 0.7 0.0
5 - 10 13,105 11,612 0.0 75.8 21.7 2.5 0.0
10 - 15 34,582 32,938 0.0 84.1 13.9 2.0 0.0
15 - 20 24,605 24,285 0.0 93.6 6.0 0.4 0.0
6112101 0 - 5 1,112 828 0.5 58.0 36.8 4.3 0.4
5 - 10 9,652 7,982 0.2 73.0 23.1 3.2 0.6
10 - 15 34,042 30,162 0.1 73.7 21.2 4.4 0.6
15 - 20 16,632 16,248 0.0 91.1 7.6 0.6 0.8
6113101 0 - 5 1,307 927 0.0 52.0 42.7 4.5 0.8
5 - 10 10,127 8,717 0.0 76.1 19.9 2.9 1.1
10 - 15 24,268 21,648 0.0 74.8 19.5 4.2 1.5
15 — 20 11,708 11,514 0.0 93.4 5.6 0.4 0.5
6114101 0 - 5 690 446 0.0 84.5 14.8 0.7 0.0
5 - 10 5,516 4,827 0.0 80.9 17.2 1.9 0.0
10 - 15 20,312 18,644 0.0 82.2 15.2 2.5 0.0
15 -'20 15,834 15,581 0.0 94.7 4.9 0.4 0.0
611 9101 0— 5 796 622 3.5 70.1 22.7 1.3 2.4
5 - 10 9,330 8,421 1.1 80.9 15.0 2.4 0.6
10 - 15 11,509 11,183 0.2 92.6 6.3 0.7 0.2
15 - 20 8,012 7,969 0.0 98.3 1.4 0.1 0.3
6121101 0 - 5 494 302 0.0 82.5 16.2 1.0 0.3
- 5 - 10 6,492 5,298 0.9 70.0 24.3 4.5 0.3
10 - 15 7.636 7,230 0.1 86.3 11.8 1.6 0.2
15 - 20 8,353 8,227 0.0 95.6 4.2 0.2 0.0
6122101 0 - 5 433 347 5.2 85.3 9.2 0.3 0.0
5 - 10 5,823 5,230 4.4 80.1 13.3 2.2 0.0
10 - 15 9,928 9,592 1.6 90.1 7.6 0.7 0.0
15 - 20 9,769 9,699 _ 0.6 96.3 2.6 0.2 0.2
6123101 0 - 5 991 818 0.0 92.1 7.3 0.1 0.5
5 - 10 6.091 5,575 0.0 89.3 9.3 0.9 0.4
10 - 15 1,224 1,213 0.0 97.7 2.1 0.1 0.2
15 - 20 1,388 1,369 0.0 96.3 3.4 0.0 0.3 .
6124/01 0 - 5 2,309 1,834 6.1 82.2 9.7 1.7 0.3
5 - 10 5,521 4,843 3.7 83.3 10.9 1.7 0.4
10 - 15 1,492 1.468 0.9 93.5 4.6 0.4 0.6
15 - 20 L545 1,506 0.7 87.7 8.9 1.3 1.4
Table 6. Contribution of post-processing filters to total number of tracks filtered for 2001.

 

 

% of Filtered Tracks

 

 

 

 

 

 

 

 

 

 

 

Range Total Number Lower TS Upper TS Direction Process
Date Bin Tracks Filtered Threshhold Threshhold of Travel Tortuosity Error
5128102 0-5 112 80 0.0 63.8 36.3 0.0 0.0
5 - 10 879 683 0.2 56.1 36.3 7.4 0.0
10 - 15 1,229 1,098 0.0 63.0 30.0 7.0 0.0
15 - 20 2,304 2,204 0.0 78.3 18.4 3.3 0.0
5129102 0-5 203 105 4.8 32.4 54.3 8.6 0.0
5 - 10 941 908 2.1 20.0 70.0 7.8 0.0
10 - 15 1,342 1,095 0.0 38.1 48.9 13.0 0.0
15 - 20 1,614 1,482 0.4 51.7 41.9 5.9 0.1
5130102 0-5 241 149 0.0 59.7 38.3 2.0 ' 0.0
5 - 10 1,245 1,004 0.0 57.3 37.5 5.2 0.0
10 - 15 3.304 3,122 0.0 72.6 24.0 3.4 0.0
15 - 20 6,918 8.843 0.0 94.4 5.3 0.4 0.0
5131102 0-5 267 168 3.0 51.8 41.1 4.2 0.0
5 - 10 1.088 888 2.8 51.2 41.5 4.8 0.0
10 - 15 982 899 1.0 84.4 30.5 4.1 0.0
15 - 20 1,171 1,124 0.1 77.7 21.0 1.2 0.0
815102 0-5 221 146 0.0 85.1 32.2 2.7 0.0
5 - 10 1,572 1,176 0.0 59.3 34.1 8.8 0.0
10 - 15 2,280 2.141 0.0 70.1 27.1 2.8 0.0
15 - 20 1,724 1,646 0.0 78.7 21.9 1.4 0.0
bis/02 0-5 262 185 0.0 72.4 27.0 0.5 0.0
5 - 10 1.448 1,144 0.0 88.4 28.8 3.0 0.0
10 - 15 1,755 .1 .584 0.0 73.5 23.1 3.3 0.0
15 - 20 2,017 1,893 0.0 81.8 15.3 2.9 0.0
817102 0-5 257 128 0.0 48.9 48.4 4.7 '0.0
5 - 10 1.088 717 0.0 38.8 52.7 8.5 0.0
10 — 15 1,610 1.353 0.0 43.2 43.7 13.2 0.0
15 - 20 2,524 2,252 0.0 55.4 38.2 8.4 0.0

 

Table 7. Contribution of post-processing filters to total number of tracks filtered for 2002.

59

 

 

 

 

 

 

 

 

 

 

 

 

 

Date Species Number“ lMinimum Length (cm)*illlaximum Length (cm)"
51812001 Chinook salmon 74 5.4 9.4
Steelhead (parr) 17 - <3.5
[Notropis sp. 1 - -
611012001 hinook salmon 62 6.5 8.9
teelhead (parr) 13 - <3.5
lewife 1 - 1 3.1
hite perch 1 - 16.2
Notropis spp. - - -
Bluntnose minnow - - -
Rock bass - - -
811112001 Chinook salmon 58 - -
p11212001 Chinook salmon 71 - -
I811312001 Chinook salmon 59 5.0 10.6
Wotmpis sp. 3 9.8 11.9
Rock bass 3 4.5 16.4
Shorthead redhorse 2 35.0 39.0
Newife 1 - -
Steelhead (smolt) 1 - 20.9
Fathead minnow 1 - 7.5
611 412001 Chinook salmon 101 5.3 9.6
Steelhead (parr) 30 - <3.5
Alewife 4 13.8 17.1
Shorthead redhorse 3 34.0 42.0
Greater redhorse 2 38.0 44.0
Steelhead (smolt) 2 23.4 24.2
511912001 Chinook salmon 23 6.4 9.3
Alewife 2 13.5 16.0
Bowfin 1 - 7.0
Rock bass - - -
Fathead minnow - - -
Notropis sp. - - -
612012001 Chinook salmon 34 6.0 ‘ 10.2
6121/2001 ' Chinook salmon 17 7.0 9.0

 

 

 

 

 

 

*Abundance and length data were not collected for all species on all days.

Table 8. Species composition from smolt trap by trap day in 2001.

60

 

 

 

 

 

Date Species Number‘ lMinimum Length (cm)* Maximum Length
(cm)*
612212001 Chinook salmon 26 5.5 8.5
612312001 Chinook salmon 16 6.7 _ 8.9
Steelhead (parr) - - <3.5
Largemouth bass (fry) - - -
otropis sp. - - -
ellow perch - - -

 

 

 

 

 

 

*Abundance and length data were not collected for all species on all days.

Table 8 (continued). Species composition from smolt trap by trap day in 2001.

61

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Date Species Number Minimum Length (cm) Maximum Length (cm)

512612002 otropis spp. 5 3.6 9.9
hinook Salmon 4 7.0 7.5

ock bass 3 3.6 5.6

rown trout 1 - 16.2

512912002 hinook salmon 7 8.4 10.1
ock bass 5 3.7 4.9

rown trout 4 19.6 20.9

horthead redhorse 3 44.5 47.0

otropis spp. 2 4.7 7.2

reek chub 2 9.2 11.2

luntnose minnow 1 - ' 4.7

Iuegill 1 - 3.3

513012002 hinook salmon 17 5.2 9.0
otropis spp. 10 4.1 8.8

teelhead (smolt) 3 16.9 21.8

ock bass 3 3.2 4.2

rown trout 2 - 23.2

ite sucker 2 38.9 47.6

ainbow darter 1 - 3.2

513112002 Rhinook salmon 39 6.5 9.0
01101118 SP0 9 3.6 10.1
Rock bass 3 4.4 14.4
Brown trout 1 - 16.4

51512002 hinook salmon 23 5.2 8.8
otropis spp. 6 6.5 8.5

ock bass 1 - 4.0

teelhead (parr) 1 - 3.0
ea lamprey 1 - 45.5

61612002 hinook salmon 73 5.4 8.3
otropis spp. 13 2.9 9.9
lack redhorse 1 - 43.4

617/2002 hinook salmon 62 5.2 9.6
teelhead (parr) 13 2.8 3.3

otropis spp. 9 4.1 9.7
ock bass 6 2.1 13.2
Iuegill 1 - 10.5
horthead redhorse 1 - 32.7

 

Table 9. Species composition from smolt trap by trap day in 2002.

62

 

 

25

 

 

 

 

 

 

 

7,000
6,000 -
7" ---'~- 20
8 5,000 -
m 6
'I -\ —i— 15 v
‘2: 4,000 :“vd - 5
I f E
a 3.0001 4- 10 e
a a
S l-
g 2.000 -\/-\j
5 Discharge -_ 5
1.000 - - - - Temperature '
o l l I r l I T j 0
‘- 00 In N O) In N O) (D
B i?) E Q Q 6 E S Q‘
ID ID 1.0 (D (O (O

 

 

Figure 17. Temperature and discharge from 2001 as measured at Croton Dam (data
courtesy of US Geological Survey).

63

 

 

25

 

g

Discharge (th * Sec”)

.3

Discharge

- - - Temperature

    

1- 20

Temperature (C)

 

 

 

08

5/1

5/8 ‘
5115 ‘

5l22 ‘

5/29 ‘

6/5 ‘

6/12 ‘

6/19 "

6/26 ‘

 

Figure 18. Temperature and discharge from 2002 as measured at Croton Dam (data

courtesy of US Geological Survey).

64

 

LITERATURE CITED

Carl, LM. 1980. Aspects of the p0pulation ecology of chinook salmon in Lake Michigan
tributaries. Ph.D. thesis, University of Michigan, Ann Arbor, MI.

Carl, LM. 1982. Natural reproduction of coho salmon and chinook salmon in some
Michigan streams. North American Journal of Fisheries Management. 2: 375-380.

Carpenter, SR and JF Kitchell, eds. 1989. The Trophic Cascade in Lakes. Cambridge
University Press, Cambridge. 385 pp.

Echoview® Help File. Software version 3.0.80. Hobart, Tasmania, Australia.

Emery, L. 1985. Review of fish species introduced into the Great Lakes, 1819-1974.
Great Lakes Fishery Commission Technical Report No. 45. Ann Arbor, MI.

Enzenhofer, HJ, N Olsen, & TJ Mulligan. 1998. F ixed-location riverine hydroacoustics
as a method of enumerating migrating adult Pacific salmon: comparison of split-beam
acousticsvs. visual counting. Aquatic Living Resources 11(2): 61-74.

Eshenroder, RL, ME Holey, TK Gorenflo, and RD Clark, Jr. 1995. Fish-community
objectives for Lake Michigan. Great Lakes Fisheries Commission Special Publication
95-3.

Fiingstam, H. 1993. Individual downstream swimming speed during the natural smolting
period among young of Baltic salmon (Salmo salar). Canadian Journal of Zoology
71:1782-1786.

Fried, SM, JD McCleave, & GW LaBar. 1978. Seaward migration of hatchery-reared
Atlantic salmon, Salmo salar, smolts in the Penobscot River estuary, Maine: riverine
movements. Journal of the Fisheries Research Board of Canada 35:76-87.

Groot, C. 1972. Migration of yearling sockeye salmon (Oncorhynchus nerka) as
determined by time-lapse photography of sonar observations. Journal of the Fisheries
Research Board of Canada 29:1431-1444.

Hesse, JA. 1994. Contribution of hatchery and natural chinook salmon to the eastern
Lake Michigan sport fishery, 1992-1993. MS Thesis, Michigan State University, East
Lansing. 82pp.

Holey, ME, RF Elliot, SV Marcquenski, JG Hnath, and KD Smith. 1998. Chinook

salmon epizootics in Lake Michigan: possible contributing factors and management
implications. Journal of Aquatic Animal Health. 10:202-210.

65

Johnston, SV, & TW Steig. 1995. Evaluation of fixed-location techniques for
monitoring upstream migrating adult salmonids (Oncorhynchus spp.) in riverine

environments. ICES International Symposium on Fisheries and Plankton Acoustics,
June12-16, Aberdeen, Scotland, 15 pp.

Keiser, R, T Mulligan, & J Ehrenberg. 2000. Observation and explanation of systematic
split-beam angle measurement errors. Aquatic Living Resources 13:275-281.

Kennen, JG, SJ Wisniewski, NH Ringler, & HM Hawkins. 1994. Application and
modification of an auger trap to quantify emigrating fishes in Lake Ontario tributaries.
North American Journal of Fisheries Management 14:828-836.

Lacroix, GL & P McCurdy. 1996. Migratory behaviour of post-smolt Atlantic salmon
during the initial stages of seaward migration. Journal of Fish Biology 49: 1086-1 101.

Lilja, J, T] Marjoméiki, R Riikonen, & J Jurvelius. 2000. Side-aspect target strength of
Atlantic salmon (Salmo salar), brown trout (Salmo trutta), Whitefish (Coregonus
lavaretus), and pike (Esox lucius). Aquatic Living Resources 13:355-360.

Lotka, AJ. 1925. Elements of physical biology. Reprinted 1956. New York: Dover.

MacLennan, DN & EJ Simmonds. Fisheries Acoustics. (London: Chapman & Hall,
1992)

Nelson, DD and JG Hnath. 1990. Lake Michigan chinook salmon mortality — 1989.
Michigan Department of Natural Resources Fisheries Technical Report No. 90-4. Ann
Arbor, Michigan.

Newcomb, TJ and TG Coon. 2001. Evaluation of three methods for estimating numbers
of steelhead smolts emigrating from Great Lakes tributaries. North American Journal of
Fisheries Management 21: 548-560.

O’Neal, RP. 1997. Muskegon River Watershed Assessment. Michigan Department of
Natural Resources Fisheries Special Report No. 19. Ann Arbor, Michigan.

Peterson, J T, RF Thurow, and JW Guzevich. 2004. An evaluation of multipass
electrofishing for estimating the abundance of stream dwelling salmonids. Transactions
of the American Fisheries Society 133: 462-475.

Rakoczy, GP, and RF Svoboda. 1997. Sportfishing catch and efi°ort from the Michigan
waters of Lakes Michigan, Huron, Erie, and Superior, April], 1994 — March 31, 1995.
Michigan Department of Natural Resources Fisheries Technical Report No. 97-4. Ann
Arbor, Michigan. '

66

Rakoczy, GP and D Wesander-Russell. 2003a. Charter boat catch and effort from
Michigan waters of the Great Lakes, 1999. Michigan Department of Natural Resources
Fisheries Technical Report No. 2002-3. Ann Arbor, Michigan.

Rakoczy, GP and D Wesander-Russell. 2003b. Charter boat catch and effort from
Michigan waters of the Great Lakes, 1998. Michigan Department of Natural Resources
Fisheries Technical Report No. 2001-4. Ann Arbor, Michigan.

Ransom, BH, SV Johnston, & TW Steig. 1998. Review on monitoring adult salmonid
(Oncorhynchus and Salmo spp.) escapement using fixed-location split-beam
hydroacoustics. Fisheries Research 35:33-42.

Rutherford, ES, S Thorrold, A Woldt, D Swank, and N Godby. 1999. Abundance,
production, and harvest of anadromous salmonids in the Manistee, Muskegon, and Au
Sable Rivers. Progress Report to the Michigan Department of Natural Resources,
Contract #23 1663.

Seelbach, PW and GW Whelan. 1988. Identification and contribution of wild and
hatchery steelhead stocks in Lake Michigan Tributaries. Transactions of the American
Fisheries Society 117: 444 - 451.

Seelbach, PW. 1993. Population biology of steelhead in a stable-flow, low gradient
tributary of Lake Michigan. Transactions of the American Fisheries Society 122: 1 79-
1 98.

Stables, TB, & GA Kautsky. 2000. Evaluation of a stimulus-response method for
distinguishing out-migrant salmonids from drifting debris for sonar counts in the Trinity
River, California. Aquatic Living Resources 13:341-347.

Talhelrn, DR 1988. Economics of Great Lakes fisheries: a 1985 assessment. Great
Lakes Fish. Comm. Tech. Report 54.

Thedinga, JF, ML Murphy, SW Johnson, JM Lorenz, & KV Koski. 1994. Determination
of salmonid smolt yield with rotary-screw traps in the Situk River, Alaska, to predict
effects of glacial flooding. North American Journal of Fisheries Management 14:837-
851.

Tody, WH, and HA Tanner. 1966. Coho salmon for the Great Lakes. Fish Management
Report No. 1. Fisheries Division — Michigan Department of Natural Resources, Lansing,
Michigan. .

Volterra, V. 1926. Fluctuations in the abundance of a species considered
mathematically. Nature 188:558-560. 1

67

Wesley, J K. 1996. Age and growth of chinook salmon in Lake Michigan: verification,
current analysis, and past trends. Michigan Department of Natural Resources Fisheries
Research Report No. 2029. Ann Arbor, Michigan.

68

 

   

    
 

   

NNNNNNNNNNNNNNN

lllllllllllllill11111111111111

3 1293 0 845