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

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 cJCIFICJDanDuopGS-p. 15

 

STRATEGIES FOR THE ASSESSMENT OF FISH SPECIES COMPOSITION
IN GREAT LAKES STREAMS

By

Katherine LeConte Smith

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

2003

ABSTRACT

STRATEGIES FOR THE ASSESSMENT OF FISH SPECIES COMPOSITION
IN GREAT LAKES STREAMS

By

Katherine LeConte Smith

Fishery managers are considering the use of low-head barrier dams in as many as
100 Great Lakes tributary streams for sea lamprey control purposes. Barriers prevent
reproductive migrations of the exotic sea lamprey but may also impact non-target species.
Complete species lists for barrier candidate streams are a prerequisite for ensuring that
barriers do not cause the loss of species of concern. Here, I examine (1) the amount of
sampling effort required to detect a target percentage of the estimated number of species
in the watershed, (2) strategies to increase the efficiency of species inventories, and (3)
the adequacy of exiting survey data for characterizing species composition in Great Lakes
streams. I found that sampling effort requirements increased exponentially as the
targeted percentage of estimated species richness rose. At the stream reach level,
sampling an intermediate length of stream maximized rates of species accumulation. At
the watershed level, allocating sampling effort to higher order sections of the watershed
increased rates of species accumulation, however, headwater species were likely to be
consistently missed when more than 70 % of sampling effort was allocated to higher
order strata. Existing lists compiled from multiple, repeated surveys contained more
species than field lists, but when only a few one-time surveys were available existing lists
were considerably smaller than field lists. In addition. we found existing data may
contain errors and older lists may not reflect the current species composition of the

stream.

ACKNOWLEDGEMENTS

Funding and support for this research was provided by the Great Lakes Fishery
Commission. The United States Fish and Wildlife Service and the Canadian Department
of Fisheries and Oceans provided additional data, advice, and support. I would also like
to thank all the participants of the barrier workshops for contributing additional advice
and suggestions. Special thanks go to the project Co-PIs Rob McLaughlin and Nick
Mandrak for advice and edits and to Deb, Dak, Jason, and the Joshes for all extra data
collection. In addition, this research would not have been possible without the
exceptional field help-Thanks Nate, Steve, Jesse, Sandra, Randy, and Dustin for all the
long field days.

The students in the Department of Fisheries and Wildlife at Michigan State
University deserve special thanks for making Lansing a wonderful place to live and work.
I particularly want to thank the Jones/Bence lab- Mike J ., Jim, Amy, Amber, Heather,
Mike W., Norine, Emily, Brian, and Wen-for great suggestions, a fun environment, and
teaching me that stats jokes can be funny. Thanks to the entire faculty for a friendly,
challenging, and exciting environment and to Dr. Bill Taylor for creating great
opportunities.

My committee Drs. Hayes and Lupi provided valuable insights into this project’s
development and were exceptionally generous with their time and advice. My utmost
thanks goes to Dr. Mike Jones for being such a great mentor and friend. His advice and
assistance greatly improved this research. Finally, I owe more than I can say to my

family- Furrnan and Rosalyn, and sister Caroline-who encouraged and supported me.

TABLE OF CONTENTS

LIST OF TABLES ............................................................................................................. vi
LIST OF FIGURES ............................................................................................................ ix
INTRODUCTION ............................................................................................................... 1
Objectives ........................................................................................................................ 6
Literature Cited ............................................................................................................... 7
CHAPTER ONE
EVALUATION OF HISTORICAL DATA: A COMPARISON OF INTENSIVE
SURVEY DATA WITH EXISTING SURVEY DATA ................................................... 10
Introduction ................................................................................................................... 10
Methods ......................................................................................................................... 14
Study Sites ................................................................................................................. 14
Intensive Field Survey ............................................................................................... 14
Survey Completeness ................................................................................................ 16
Existing survey data .................................................................................................. 16
Comparisons .............................................................................................................. 18
Results ........................................................................................................................... 19
Discussion ..................................................................................................................... 21
Conclusions ................................................................................................................... 24
Literature Cited ............................................................................................................. 25
CHAPTER TWO
SAMPLING EFFORT REQUIRMENTS FOR CHARACTERIZING FISH SPECIES
COMPOSITION IN GREAT LAKES WATERSHEDS .................................................. 41
Introduction ................................................................................................................... 41
Methods ......................................................................................................................... 45
Study sites ................................................................................................................. 45
Sampling locations .................................................................................................... 45
Sampling .................................................................................................................... 46
Statistical Methods .................................................................................................... 46
Results ........................................................................................................................... 49
Model Choice ............................................................................................................ 50
Sampling Effort Requirements .................................................................................. 51
Influencing Factors .................................................................................................... 51
Discussion ..................................................................................................................... 52
Conclusions ................................................................................................................... 56
Literature CIted ............................................................................................................. 58
CHAPTER THREE
OPTIMIZATION OF SAMPLING EFFORT FOR FISH FAUNAL INVENTORIES IN
MICHIGAN WATERSHEDS ........................................................................................... 69

iv

Introduction ................................................................................................................... 69

Methods ......................................................................................................................... 7 1
Results ........................................................................................................................... 74
Discussion ..................................................................................................................... 77
Literature Cited ............................................................................................................. 81
MANAGEMENT IMPLICATIONS .............................................................. 96

LIST OF TABLES
CHAPTER ONE

Table 1. Existing survey data sources and sampling descriptions. Electrofishing is
abbreviated as P-EF (backpack units) and B-EF (boat unit). River number l=Au
Gres, 2=Little Pigeon, 3=Harlow, 4=Sucker, 5=Boyne, 6: Big Otter, 7=Conneaut,
8=Raccoon, 9= Grindstone. ...................................................................................... 28

Table 2. Field survey completeness. Rivers with an * indicate the estimate is a lower
bound because the species accumulation curve was not clearly reaching an
asymptote .................................................................................................................. 29

Table 3. Number of species unique to existing list A and the field list in a pairwise
comparison. ............................................................................................................... 29

Table 4. Contribution of annual sampling to the size of species lists. This analysis was
only conducted for the three rivers with two years of field sampling and the six
rivers with data provided by the USFWS (the only source of multiple repeated data
including dates of capture for all individuals). ......................................................... 30

Table 5. Effect of survey age on quality of existing species lists. The date species were
last detected on list A was recorded and species were grouped into the three date
categories below. The percentages of species from each date category that were not
redetected in our 2002 field survey are given below ................................................ 30

Table 6. Probable errors in existing species lists. Species on the existing list A but not on
the regional species list were identified as possible errors. A few of these species
were detected and confirmed during our 2002 field survey, however, the remaining

species (column three) were probable errors. ........................................................... 31
Table 7. Summary of the advantages and disadvantages of each species list type. .......... 32
Appendix A. Summary of species occurrences on the regional, existing, and field species

lists. ........................................................................................................................... 37
CHAPTER TWO

Table 1. Description of models. All models exhibit a general non-sigmoidal increasing
shape and are described in Ratkowsky (1990) or Flather (1996). Variable S = the
number of species detected, X = the number of reaches sampled, and A, B, C, and
D= parameters. .......................................................................................................... 62

Table 2: Sampling characteristics by river. Columns from left to right indicate the
number of species observed (Sobs) and estimated (Sest), difference between Sest and

vi

Sobs, standard error of the estimate of Sest, the percentage of Sest observed during the
field survey, and the total number of reaches sampled per river. ............................. 62

Table 3: Model fit by river. Goodness of fit characteristics are indicated by the corrected
AIC values. Corrected AIC values are a maximum likelihood estimate of model fit
corrected for number of parameters. A lower number indicates a better fit (e.g.,
larger negative values indicate best fit). Average rank and range of ranks for each
model and river based on corrected AIC values are presented in the last two rows. 63

Table 4. Comparison of the number of reaches needed to obtain 80, 90, 95, and 100 % of

Sest using models 11, 10, and 6. The best model fit for each river is listed in column
two. N indicates sampling effort requirements could not be estimated because the

model reached an asymptote lower than Sest. Likewise, 200+ indicates a very high
and unrealistic estimate of sampling effort requirements. ........................................ 64

Table 5. Comparison of two factors influencing sampling effort requirements: choice of
model and species richness estimator. The table presents the effects of two
techniques to estimate species richness -the j ackknife estimator (J) and the
asymptotic estimator(A)- on estimates of number of reaches needed to detect 95 %
of the estimated species richness. The average estimate of species richness
estimator choice is compared to the average estimate of model choice in the bottom
two rows. The difference between average number of reaches estimated from the
two estimators is less than the difference in the average number of reaches estimated

with different models ................................................................................................ 65
CHAPTER THREE
Table 1. Sampling effort allocation rules .......................................................................... 83

Table 2. Species distribution by stream order strata. Total is the proportion of the total
number of species in the watershed found in each strata, Rare refers to the
percentage of rare species (those found in only one reach), and Unique refers to the
number of species only found in one of the three strata. ‘Not 111’ refers to the
number of species found in either strata I and II but not in strata 111. Table 3.
Average sampling costs (hours). Fixed refers to the average of travel, set up, and
clear down time combined and variable refers to the average of combined sampling
and fish processing time ............................................................................................ 84

Table 3. Average sampling costs (hours). Fixed refers to the average of travel, set up,
and clear down time combined and variable refers to the average of combined

sampling and fish processing time. ........................................................................... 85

Table 4. Number of sample units per reach that yields the maximum rate of species
accumulation. More than one value is listed when rates were equivalent. .............. 86

vii

Table 5. Number of species missed in all simulations given different effort allocation
scenarios. Total sampling effort is based on approximately 25 % of original
sampling time. ........................................................................................................... 86

Appendix A. Number of times in the 100 simulations that a species is missed given for

each allocation rule. Results are based on approximately 25 % of original sampling
time. .......................................................................................................................... 92

viii

LIST OF FIGURES
CHAPTER ONE

Figure 1. Study stream locations. Note: streams not drawn to scale. ............................. 33

Figure 2. Species accumulation curves for all nine study streams. The cumulative
number of species is plotted against the number of reaches sampled ....................... 34

Figure 3. Comparison of the number of species on the field list and existing lists A and B
by river. List A includes species found in all existing surveys, field list includes
only the species detected in our 2002 field survey, and list B includes species
detected in all existing surveys except the sea lamprey control surveys. River
number 1=Au Gres, 2=Little Pigeon, 3=Harlow, 4=Sucker, 5=Big Otter, 6= Boyne,

7=Conneaut, 8=Raccoon, 9= Grindstone .................................................................. 35
Figure 4. Relationship between watershed size and estimated species richness. ............ 36
CHAPTER TWO
Figure 1. Map of study streams. Note: streams not drawn to scale. ................................. 66

Figure 2. Species accumulation curves for nine Great Lakes streams. River 1=Au Gres,
2=Little Pigeon, 3=Harlow, 4=Sucker, 5=Big Otter, 6= Boyne, 7=Conneaut,
8=Raccoon, 9= Grindstone. The cumulative number of species is plotted against the
number of reaches sampled ....................................................................................... 67

Figure 3. Sampling effort requirements to achieve a targeted percentage of estimated
species richness. The amount and range of sampling effort, in reaches, needed to

detect 80, 90, 95, and 100 % of Sest is presented for seven study streams. .............. 68
CHAPTER THREE
Figure 1. Map of study streams ........................................................................................ 87

Figure 2 A&B. Example of the average number of species found per hour given
strategies of sampling 1 to 10 sampling units per reach. .......................................... 88

Figure 3. Effects of skipping sample units on rates of species accumulation. The
character 0 refers to sampling consecutive sample units, x refers to skipping one

sample units, and + refers to skiping two sample units. Skipping one or two sample
units does not consistently improve rates of species accumulation over sampling
consecutive sample units ........................................................................................... 89

Figure 4. Number of species found under each sampling effort allocation rule. As the

allocation rule increases the amount of effort allocated to third order strata increases
from 34 % to 100 %. ................................................................................................. 9O

ix

Figure 5. Rates of species accumulation by season. Characters + refers to summer
sampling only, x refers to spring sampling only, and 0 refers to both spring and
summer sampling. In Stream 1 and 3, rates of species accumulation are greatest in
summer. In Rivers 2 and 4, rates are greatest in spring. Rates from combined spring
and summer sampling are always intermediate. ....................................................... 91

INTRODUCTION

Sea lamprey (Petromyzon marinas) were first observed in the upper Great Lakes
in the 1930’s. By the mid 1940’s, sea lamprey were causing extensive ecological
changes, the most obvious of which was the practical extirpation of large native predators
such as lake trout (Smith et al. 1980). By the late 1950’s, lake trout were effectively
eliminated in Lakes Huron and Michigan (Holey et al. 1995). Faced with the drastic
decline of one of the world’s largest freshwater fisheries, the United States and Canada
created the Great Lakes Fishery Commission (GLFC) in 1956 to control sea lamprey and
coordinate fisheries research (Fetterolf 1980). The cooperative sea lamprey control
program that resulted was one of the first successful exotic species control programs.
Although successful, the program has required considerable on-going investment and
innovation (Christie and Goddard In Press).

The GLFC has focused control efforts in streams, where lamprey are concentrated
as adults or vulnerable during their larval stage. rather than in the Great Lakes
themselves, where lamprey are widely dispersed. In streams, sea lamprey can be
controlled by either blocking spawning migrations, with barriers or traps, or by
chemically treating streams to kill the larvae which burrow in stream sediments for up to
7 years (Manion and McLain 1971). Early control methods included a variety of lamprey
barriers developed to prevent adult spawning migrations. Although effective in many
streams, these early barriers were expensive to operate and were subject to failure during
flood events when they were often damaged or submerged (Smith et al. 1980). Electrical
barriers, which were operated seasonally and thought to affect lamprey exclusively, were

also tested. They were not widely adopted because they proved dangerous to humans and

deterred spawning migrations of non-target fish (McLain 1957). In addition, due to
uncertainties in migration timing, early and late-migrating lampreys were often
unimpeded when barriers were not operational (Adams per. com).

These difficulties prompted managers to consider chemical control methods.
Over 6000 chemicals were tested for treatment of ammocetes in streams. Of these, TF M
(3-triflouromethyl-4-nitrophenol) was chosen because it is highly selective to sea lamprey
and does not accumulate in the environment. TFM proved extremely effective, reducing
sea lamprey populations to 10 % of historical levels (Smith et a1. 1980) and since 1958,
has been used as the primary control method. Studies have not proven TFM to have any
lasting negative effects on non-target fish or the aquatic environment; however, negative
public sentiment, uncertain supply, and high costs have prompted the GLFC to seek
alternative sea lamprey control methods (Christie and Goddard In Press).

These concerns prompted the GLFC to set goals to reduce TFM usage by 50 %
through the use of alternative controls such as sterile-male release and improved low
head barriers (GLFC 1992). Sterile male release has no known effects on non-target
species or the environment, however, its efficacy is unknown and it is currently only used
in the St. Mary’s River (Jones et al. In Press). Barriers are currently the only proven
alternative to lampricide treatments and are likely to be an integral part of any future
control program (Lavis et al. In Press).

Barriers are a financially attractive alternative to chemical controls but may pose
environmental risks. Benefits of barriers include: more efficient control on streams
where physical characteristics make chemical treatments difficult, expensive, or

ineffective; potential savings in treatment costs and man hours; reduced dependency on

lampricides; and reduced quantity of chemicals added to the environment. However, as
barriers are slated for over 90 streams (Lavis et al. In Press), the GLFC wants to
determine the possible non-target effects of barriers on streams and their fish
communities.

The ecological effects of large dams have been studied extensively, but much
remains unknown about the effects of smaller non-impounding dams. Such barriers do
not cause direct mortality, although they may indirectly affect stream fish communities
by blocking fish movement and altering stream habitat (Pringle 1997). A recent study of
47 streams in the Great Lakes basin concluded that low-head barrier dams do have effects
on stream fish communities around the Great Lakes (Dodd et al. In Press; Porto et al.
1999). These studies demonstrated that (1) fish species richness declines from
downstream to upstream more rapidly in streams with barriers than in reference streams
and (2) species richness shows a distinct peak just downstream of the barriers possibly
reflecting a bottleneck for fish movements (Noakes et al. 2000). Porto et al. (1999) also
presented evidence from mark-recapture data that barriers cause restrictions to fish
movement. Dodd et al. (In Press) suggested that habitat alteration associated with
barriers contributes to fish community effects and that some species are more vulnerable
to barriers than others. Analysis of historical data and longitudinal comparisons of fish
communities in streams with natural barriers indicated that impacts of low head dams
occurred over decades (Noakes et al. 2000). Although a variety of barriers, including
low-head, velocity, seasonal inflatable, and electric barriers are available, the basic
question of how fragmentation due to sea lamprey barriers will affect aquatic

communities over the long-term remains unanswered.

In light of these considerations, the GLFC adopted an interim policy for barrier
placement in December of 1999 with the stipulation that the decision to construct barriers

must consider possible effects on non-target species. In this policy, the GLFC states:

"Barriers will only be sited on those streams where it is anticipated that
replacement of lampricide treatment with a barrier does not detract from
the biological integrity of the candidate stream or of the ecoregion where

the candidate stream is located."

This policy statement includes: no loss in species richness in the ecoregion where the
candidate stream is located; no fragmentation of, or reduction at the edges of, species
ranges; and no ecoregion effect that causes a change in the species status in the Great
Lakes. Furthermore, the GLF C views barriers as an alternative to the established method
of chemical control and, as such, a barrier should not result in a greater net environmental
impact than lampricide treatments if it is to replace chemical methods of control
(McLaughlin et al. In Press).

Although this policy may seem straightforward, its application is challenging.
There are over 450 tributaries to the Great Lakes that produce sea lamprey (Pearce et al.
1980). Data on the fish communities in many of these streams are lacking or incomplete.
In addition, we have limited knowledge of the natural history characteristics that
influence a species’ vulnerability to barriers for the majority of Great Lakes fishes.
Therefore, managers must judge the effects of proposed barriers with limited data or by

developing alternative methods to predict or minimize these effects.

Implementation of the GLF C barrier policy depends on the ability of managers to
determine species composition in banier candidate streams. This question is of special
importance because the effects of barriers on individual species, especially endangered or
threatened species, remain largely unknown. Furthermore, many barriers are proposed
for larger, environmentally sensitive streams where approval will be most difficult (Lavis
et al. In Press). If the GLFC adopts a “no impact” policy for threatened and endangered
species, it is vital to know with certainty if species of concern occur in a candidate
stream.

As budgets are limited for such activities, an understanding of the adequacy of
historical data and the sampling effort requirements for watershed-level fish community
characterization are essential. This evaluation reflects on the feasibility of the barrier
policy itself and is essential for prioritizing candidate streams. Given that ascertaining
species composition of candidate streams is a necessary component of this evaluation,
determining the most efficient sampling strategy within a steam or watershed will further
increase the cost-effectiveness of implementing the barrier policy.

Studies have been conducted in many regions of the US to determine sampling
effort requirements for fish species inventories at the stream-reach-level only (Hughes et
al. 2002, Patton et al. 2000, Bowen and Freeman 1998, Paller 1995, Angermeir and
Smogor 1995, Lyons 1992). Although landscape-level inventory sampling effort
requirements have been examined for a broad range of taxa (Moreno and Halffier 2000,
Gimaret-Carpentier et al.1998, Dobyns 1997, F lather 1996), we know of no studies that

have examined sampling effort requirements to determine fish species composition in a

watershed. Furthermore, survey data exist for most streams although the completeness of

resulting species lists is unclear.

OBJECTIVES

We extensively surveyed nine Great Lakes streams between May and August
2002 to; (1) compare the species lists from these intensive field surveys to species lists
obtained from existing survey data; (2) determine the sampling effort needed to
determine a target percentage of estimated species richness our study streams and
determine if environmental factors, such as watershed size, influence sampling effort
requirements; and (3) develop a protocol for the efficient assessment of species
composition in Great Lakes streams by examining the allocation of effort within a

watershed or stream.

LITERATURE CITED

Angerrneier, P. L. and Smogor, R. A. 1995. Estimating number of species and relative
abundances in stream-fish communities: Effects of sampling effort and discontinuous
spatial distributions. Canadian Journal of Fisheries and Aquatic Sciences 56: 739-47.

Bowen, Z. H. and Freeman, M. C. 1998. Sampling effort and estimates of species
richness based on propositioned area electrofisher samples. North American Journal of
Fisheries Management 18: 144-153.

Christie, G. C. and Goddard, C. I. In Press. Advances in the integrated management of
sea lampreys in the Great Lakes. Canadian Journal of Fisheries and Aquatic Sciences.

Dobyns, J. R. 1997. Effects of sampling intensity on the collection of spider species and
the estimation of species richness. Pest Management and Sampling 26: 150-61.

Dodd, H. R., Hayes, D. B., Baylis, J. R., Carl, L. M., Goldstein, J. D., McLaughlin, R.
M., Noakes, R. L., David, L. G., Porto, L. M., and Jones, M. L. In Press. Low-head
lamprey barriers dam impacts on stream habitat and fish communities in the Great Lakes
basin. Canadian Journal of Fisheries and Aquatic Sciences.

Fetterolf, C. M. 1980. Why a Great Lakes Fishery Commission and why a Sea Lamprey
International Symposium. Canadian Journal of Fisheries and Aquatic Sciences 37: 1588—
93.

Flather, C. H. 1996. Fitting species-accumulation functions and assessing regional land
use impacts on avian diversity. Journal of Biogeography 23: 155-168.

Gimaret-Carpentier, C. R., Pelissier, R., Pascal, J. P., and Houllier, F. 1998. Sampling
strategies for the assessment of tree species diversity. Journal of Vegetation Science 9:
161-72.

Great Lakes Fishery Commission. 1992. Strategic vision of the Great Lakes Fishery
Commission for the decades of the 19905. Great Lakes Fishery Commission. Ann
Arbor, MI

Holey, M. E., Rybicki, R. W., Eck, G. W., Brown, E. H., Marsden, J. E., Lavis, D. S.,
Twohey, M. B., Trudeau, T. N., and Horrall, R. M. 1995. Progress toward lake trout
restoration in Lake Michigan. Proceedings of the International Conference on

Restoration of Lake Trout in the Laurentian Great Lakes. Journal of Great Lakes
Research 21 Suppl. 1: 128-151.

Hughes, R. M., Kaufmann, P. R., Herlihy, A. T., Intelmann, S. S., Corbett, S. C.,
Arbogast, M. C., and Hjort, R. C. 2002. Electrofishing distance needed to estimate fish

species richness in rafiable Oregon rivers. North American Journal of Fisheries
Management 22: 1229-40.

Jones, M. L., Bergstedt, R. A., Twohey, M. B., Fodale, M. F., Cuddy, D. W., and Slade,
J. W. In Press. Compensatory mechanisms in Great Lakes sea lamprey populations:
Implications for alternative control strategies. Canadian Journal of Fisheries and Aquatic
Sciences.

Lavis, D. S., Hallett, A., Koon, E. M., and McAuley, T. In Press. History of and
advances in barriers as an alternative method to suppress sea lampreys in the Great
Lakes. Canadian Journal of Fisheries and Aquatic Sciences.

Lyons, J. 1992. The length of stream to sample with a towed electrofishing unit when
fish species richness is estimated. North American Journal of Fisheries Management 12:
198-203.

Manion, P. J. and McLain, A. L. 1971. Biology of larval sea lampreys of the 1960 year
class, isolated in the big Garlic River, Michigan, 1960-65. Great Lakes Fishery
Commission.

McLain, A. L. 1957. The control of upstream movement of fish with pulsed direct
current. Transactions of the American Fisheries Society 86: 269-284.

McLaughlin, R. L., Marsden, J. E., and Hayes, D. B. In Press. Achieving the benefits of
sea lamprey control while minimizing effects on nontarget species: conceptual synthesis
and proposed policy. Canadian Journal of Fisheries and Aquatic Sciences

Moreno, C. E. and Halffier, G. 2000. Assessing the completeness of bat biodiversity
inventories using species accumulation curves. Journal of Applied Ecology 37: 149-58.

Noakes, D., McLaughlin, R., Baylis, J ., Carl, L., Hayes, D., and Randall, R. 2000.
Biological impact of low-head barrier dams. Great Lakes Fishery Commission project
completion report.

Paller, M. H. 1995. Relationships among number of fish species sampled, reach length
surveyed, and sampling effort in South Carolina coastal plain streams. North American
Journal of Fisheries Management 15: 110-120.

Patton, T. M., Hubert, W. A., Rahel, F. J ., and Gerow, K. G. 2000. Effort needed to
estimate species richness in small streams on the Great Plains in Wyoming. North
American Journal of Fisheries Management 20: 394-398.

Pearce, W. A., Braem, R. A., Dustin, S. M., and Tibbles, J. J. 1980. Sea Lamprey in the
lower Great Lakes. Canadian Journal of Fisheries and Aquatic Sciences 37: 1802-1810.

Porto, L. M., McLaughlin, R. L., and Noakes, D. L. 1999. Low-head barrier dams
restrict the movement of fishes in two Great Lakes streams. North American Journal of
Fisheries Management 19: 1028-36.

Pringle, CM. 1997. Exploring how disturbance is transmitted upstream: Going against
the flow. Journal of the North American Benthological Society 16: 425-38.

Smith, B. R. and Tibbles, J. J. 1980. Sea lamprey (Petromyzon marinus ) in lakes Huron,
Michigan, and Superior: History of invasion and control, 1936-78. Canadian Journal of
Fisheries and Aquatic Sciences 37: 1780-1801.

CHAPTER ONE

EVALUATION OF HISTORICAL DATA: A COMPARISON OF INTENSIVE
SURVEY DATA WITH EXISTING SURVEY DATA

INTRODUCTION

Human uses of aquatic systems have placed tremendous stresses on aquatic fauna.
As a result, these species are proportionally more threatened than terrestrial species
(Shaffer er al. 1998; Master 1991; Master 1990). More than 20 % of the world’s known
freshwater fish are extinct or imperiled (Argent et al. 2003). In the United States, 303
species (37 %) of freshwater fish are at risk of extinction and 17 species have already
gone extinct (Nature Conservancy 2003). North America harbors the greatest diversity of
freshwater fish in the world (Matthews and Robinson 1998); however, these species are
among the most imperiled taxa in the United States (Braun et al. 2000).

Because fish communities may be impacted by both upstream and downstream
disturbances (Pringle 1997), landscape-level threats require management and
conservation actions at a watershed level (Fausch er al. 2002). Landscape-level threats
such as non-point source pollution, habitat and hydrological alterations, and exotic
species introductions are the primary threats to freshwater species (Braun et al. 2000;
Richter et al. 1997). Management strategies to address these landscape level threats and
protect fish species diversity require knowledge of the fauna present in individual
watersheds. Data on species occurrences are generally available and presumed to be
accurate at the regional scale due to ichthyofaunal surveys conducted to establish species
distribution maps (Page and Burr 1991). These lists, however, rarely allow accurate
assessment of species occurrence for individual watersheds. Data on species occurrence

in individual watersheds can be obtained from natural resource management agencies,

10

museum collections, and previous research reported in the literature, but these sources
can differ in temporal and spatial scale, accuracy of species identification, ease of data
access, and survey methodology and purpose. Species lists obtained from natural
resource management agencies are often based on relatively recent collections, but
identification of non-game species is generally a low priority and voucher specimens are
not often retained. Species lists from previous research are likely more to be accurate but
are often time-consuming to identify and collect (especially in the case of raw data) and
are usually only available for a small number of streams. Museum records generally
include voucher specimens but such records are often quite old and species composition
of the stream may have changed since the original data were collected.

A compilation of existing data is generally a less expensive approach than an
intensive field survey for assessing species composition of a watershed, but such
compilations are frequently hampered by geographical and temporal gaps, non-
standardized sampling protocols, insufficient or unknown sampling effort, and sampling
biases (Ponder et al. 2001; McLaughlin et al. 2001). For example, the completeness of
an existing survey often cannot be quantitatively assessed (De Silva and Medellin 2001);
existing data may be more complete for some locations and habitats than others; rare
species are more likely to be missing from existing species lists than common species
(Green and Young 1993); and species composition may change over time (Bowen and
Freeman 1998; Fairchild et al. 1998). Despite these drawbacks, conservation and
management actions are often made based on available data because comprehensive field

surveys are often prohibitively expensive (Oliver and Beattie 1996; Kremen 1994).

ll

Guidelines for the use of existing species lists are desirable for environmental
assessments, conservation prioritization, and ecological studies (Blackburn and Gaston
1998; Vemhes and Younes 1993). Comparisons of historical records with current
surveys have been used to document species declines (Quinn and Kwak 2003; Shaffer et
al. 1998; Shaffer et al. 1998; Winston and Taylor 1991). These data have also been used
to identify areas of high endemism and diversity for conservation reserves (Scott et al.
1993). However, the use of potentially biased or incomplete data sets can lead to
inappropriate management or conservation actions (Gomez de Silva and Medellin 2001;
Blackburn and Gaston 1998; Remsen 1994; Vemhes and Younes 1993). Because of this,
the limitations and biases of using existing data to compile watershed-level species lists
should be clear. Evaluation of the adequacy of existing data should lead to better
management decisions or at least a better appreciation of the risks associated with the use
of these data.

In the Great Lakes basin, fishery managers are considering the use of low-head
barrier dams in as many as 100 Great Lakes tributary streams for sea lamprey control
purposes. Barriers prevent reproductive migrations of the exotic sea lamprey but may
also have effects on non-target species (Dodd et al. In Press; Porto et al. 1999), so the
Great Lakes Fishery Commission, who manages the sea lamprey control program, has
developed an interim policy to ensure that barriers do not cause the loss of listed or
locally rare species. A complete species list for barrier candidate streams is a prerequisite
for effective application of the policy. Completing an intensive survey of all candidate

streams adds considerably to the cost of using this control method, so the use of existing

12

survey data is desirable. However, the completeness of species lists complied from
existing survey data will determine the adequacy of this alternative, less costly approach.

For some regions and taxa, strategies have been developed to evaluate the
completeness or adequacy of existing species lists. Ponder et al. (2001) conducted a
spatial analysis of sampling effort and for several taxa tested the robustness of species
distributions predicted from existing data, using bootstrapping techniques to assess the
adequacy of background sampling effort. For Mexican Birds, De Silva and Medellin
(2001) determined species likely to be found in all habitats which they called
“omnipresent taxa.” Existing species lists that lack omnipresent taxa or with fewer than
the minimum number of species were categorized as seriously incomplete. Conroy and
Noon (1996) used decision theory to incorporate the uncertainty in the completeness of
species lists into conservation decisions. Kodric-Brown and Brown (1993), corrected for
sampling biases in an existing data set (Glover and Sim 1978) to illustrate the effects of
biased and incomplete data on the conclusions and conservation recommendations in a
previous study.

To assess the adequacy of existing species lists in Great Lakes streams, we
compiled fish species lists for nine Great Lakes streams from all historical and federal
and state management agency surveys. These compiled species lists represent the readily
obtainable data on species composition in the absence of an extensive field survey. We
compared these existing species lists to data collected from our own intensive field
surveys of the same watersheds. Our objectives were to evaluate the adequacy of these
data sets and provide recommendation for the use of existing data for characterizing

stream fish communities.

13

METHODS
Study Sites

For these comparisons, we selected nine Great Lakes tributary stream systems that
differed in location within the basin, physical characteristics, and fish community types

(Figure 1). Watershed area ranged from 23.6 km2 to 432.8 km2 and streams were third to

fifth order (Strahler 1952). Six of the nine streams were shallow enough to be sampled
with a backpack electrofisher at all locations while the three largest streams required the
use of a boat shocker in deeper areas. Two streams drained into Lake Superior, two into
Lake Huron, three into Lake Erie, and two into Lake Ontario. Fish assemblages ranged
from highly diverse cold/coollwann-water communities (>60 species) to low-diversity

cold-water communities (<20 species).

Intensive Field Survey

 

Sampling was stratified by stream order so that the wide range of stream habitat
types present in the watershed were represented. All streams were delineated by order
into first to fifth order stretches of stream (Strahler 1952) and a minimum of six stretches
was randomly selected from each order type. One access point at each stretch was
randomly selected for sampling and at each access point we sampled a reach of 30 mean
stream-widths in length, according to recommendations for Wisconsin streams (Lyons
1992). To detect seasonal changes in species composition, we sampled in both spring
(May and June) and summer (July and August) of 2002. In addition, preliminary
sampling was conducted in 2001 on three study streams. Sampling in 2001 was not
conducted in a standardized manner and these lists were used only to identify species

found in only one of two sampling years.

14

All reaches were sampled by backpack electrofishing (Model 12-B Smith-Root,
Inc.). Sampling consisted of a single pass and reaches were not blocked. A three-person
field crew conducted sampling in a upstream direction, moving back and forth across the
river while moving the anode in a continuous ‘m’ pattern with effort made to cover the
entire stream reach area. We focused sampling effort in areas where fish were most
likely to be found (e. g., around large woody debris) and moved quickly through
unproductive areas (e.g., sandy open areas). Captured fish were held in large containers
for later identification and subsequent release. Fish that could not be identified in the
field were preserved in a 10 % formaldehyde solution and returned to the lab for
identification. Dr. Tom Coon and the Royal Ontario Museum confirmed identifications
of all uncommon species. Lamprey species were excluded from our analysis because of
concerns over field identification of larval lampreys.

We sampled non-wadeable reaches by boat electrofishing using the following
depth-stratified point sampling technique. Reaches were divided into 15 sample units,
each measuring one stream-width in length. At the beginning and end of each sample
unit, a point at each of the following depths was sampled for 10 seconds: 0 m, 0.75 m, 1.5
m, 2.25 m off both banks. Points at the 0 m contour were fished by “nosing” the anode
into shore. Sampling was conducted in an upstream to downstream direction. After
completing point sampling, we sampled the entire length of both shores by boating
slowly upstream as close to the shore as possible, while continually shocking and

collecting fish.

15

Survey Completeness

We generated a species accumulation curve for each river to assess sampling
sufficiency. To produce these curves, we developed a simulation program which
randomly chose, without replacement, a permutation of the sequence of all reaches, and
recorded the cumulative number of species found versus the number of reaches sampled.
We ran 1000 simulations and generated a curve representing the average number of
species found versus the number of reaches sampled. To calculate the estimated species
richness for each river, we used the first order jackknife technique, often recommended
as one of the least biased and most precise estimators (Boulinier et al. 1998; Palmer
1990; Heltshe and Forrester 1983). We used these estimates of species richness to
calculate the pr0portion of estimated species richness that we observed in our field

surveys.

Existing survey data

To compile existing species list for our study rivers, we contacted all state and
federal natural resource management agencies for each watershed (Table 1). These
agencies collected fish species during game species abundance surveys, sea lamprey
control activities, index of biological integrity assessments, or general surveys. When
available, we also obtained species occurrence records from historical or museum
databases. However, these data were not available for all streams. At least two sources
of existing survey data were found for all nine rivers.

Information on the survey date and location, purpose, gear type(s) used, and
collection of voucher specimens was requested with all existing survey data or species

lists. In some cases, some survey details were not available. Most surveys were

16

conducted by electrofishing; voucher specimens were generally only retained in historical
and museum collections; and some surveys included exact dates while others were only
able to provide a summary of species detected over a large time period (Table 1). Some
surveys targeted all species while others targeted only a single species and recorded by-
catch. One source of data, provided by the United States Fish and Wildlife Service
(USFWS) and the Canadian Department of Fisheries and Oceans (DFO) from sea
lamprey control activities, was unique in that these lists contained multiple years of
repeated survey data from surveys conducted over as many as 45 years. The USFWS
was able to provide exact dates that all species were detected while the DFO provided a
summary of species detected over the past 18 years. These sea lamprey control lists
were, on average, much larger than species lists from other sources, but are only available
on sea lamprey-producing streams and so do not represent widely available existing
survey data.

We compiled a regional species list, a field list, and two existing species lists for
each of our study rivers to facilitate our analysis. Regional lists were compiled from
historical collections from each stream’s greater watershed and represent the possible
species pool for each study stream (Mandrak per. com.). Species detected during our
2002 field survey were recorded on the field lists. The first existing species list, (list A),
was compiled from all existing survey data combined. Because sea lamprey surveys are
a unique data source only available on sea lamprey producing streams, we also created a
second existing species list (lists B) that excludes this source. On our existing species
lists, the most recent date that a species was detected in any survey was recorded for each

river.

17

Comparisons

To examine the relative completeness of existing and field survey data, field lists
were compared to existing species lists A and B. The number of species unique to either
the field or existing lists A was also noted. To determine if differences between lists
were affected by the number of years sampling was conducted, we examined the number
of species found in only one survey year for existing and field lists. Multiple years of
repeated survey data containing the dates that all species were detected were available for
the six streams surveyed by the USFWS only. For these streams, we calculated the
average number of years that surveys were conducted and the average number of species
found per annual survey. The proportion of species found in only one survey year was
compared to the total number of species found in all survey years combined. For the
three streams with 2001 and 2002 field survey data, we recorded the number of species
unique to only one of two survey years.

To determine if existing lists were comparable in quality to species lists compiled
from an intensive field survey, we examine our existing species lists for errors or
problems arising from the age of the existing survey data. Errors in existing data may
arise because identification of non-target species was not a priority for many existing
surveys. In addition, species identifications often cannot be confirmed in existing lists
because voucher specimens were generally not collected. To detect species which may
have been misidentified or mis-recorded, we compared existing lists to the regional
species lists. Our rationale was that if a species is found on the existing species lists but
not on the regional watershed lists it was likely misidentified in the existing survey. To

examine the effect of the age of species lists, we grouped the species in existing list A as

18

last detected before 1960, between 1960 and 1984, and after 1984 and determined the
percentage of species from each group that was not redetected in our 2002 field survey.
If species composition of the river had changed since existing surveys were conducted,
we would expect to find a larger percentage of species in the older age categories that
were not redetected in our field survey.

Identifying a relationship between watershed characteristics and species richness,
might allow inadequate species lists, those with too few species given its watershed
characteristics, to be identified. Several studies have documented a relationship between
watershed size, regional fish species richness, and species richness of a watershed
(Argent et al. 2003; Matthews and Robinson 1998). To examine if the adequacy of
historical lists might be assessed a priori, we examined correlations between (1)
watershed size and the estimated species richness of the watershed and (2) the number of

species in watershed and the regional species richness.

RESULTS

Our field sampling appeared to capture the majority of species estimated to be
present in the study rivers, ranging from 84 to 94 % of estimated species richness (Table
2). Species accumulation curves were not clearly approaching an asymptote in four of
the nine study streams, however (Figure 2). Estimates of the proportion of species
detected for these rivers should be considered an upper estimate because the jackknife
estimator of species richness tends to underestimate species richness with low sample

sizes and can estimate species richness for the years and seasons sampled only (Colwell

and Coddington 1994; Palmer 1990).

19

Comparing the number of species on field lists to existing lists A and B, we found
that when multiple repeated survey data were available, existing lists included more
species than lists from our intensive field survey. List A, which included the multiple
repeated sea lamprey control data, yielded a median of 9 more species (Figure 3) than
field lists. In a pairwise comparison of list A to the field list for each stream, both lists
contained some unique species (Table 3). When multiple repeated survey data were not
available, however, field survey data were more complete than existing data in almost all
cases. List B, which represent widely available data, included a median of 6 fewer
species than our field lists (Figure 3).

Repeated annual sampling added many species to both field and existing species
lists. Comparing our 2001 and 2002 field species lists, we found an average of 10 species
per river unique to only one of two sampling years (Table 4). Existing lists were also
greatly increased by repeated annual sampling. The USFWS sea lamprey control surveys
were conducted an average of 31 times per river over the past 50 years. An average of
7.5 species were found during one annual survey compared to an average of 45.5 species
for all survey years combined.

Although existing lists may in some cases be larger than field survey lists, we
found two potential problems with the use of existing data: (1) older lists do not appear
to well represent current species composition of the study streams and (2) several species
were likely misidentified in existing lists. Of the species detected before 1960, an
average of 69 percent were not redetected in our 2002 field survey; an average of 48

percent of those detected between 1960 and 1984 were not redetected in our 2002 field

20

survey (Table 5); and an average of 29 percent of those last detected after 1984 were not
redetected (Table 5).

Several species on existing species lists were not found on the regional species
list, suggesting possible errors in existing data (Appendix A). We found as many as 12,
with an average of 4.1 species per river recorded on the existing lists but not on the
regional lists (Table 6). An average of 3 of these species were not detected in the field
surveys so were likely either misidentified, a recording error, or previously unknown in
the watershed. However, the remaining 1.1 species per river not found on the regional
list were positively identified in our field survey and so suggest the regional list is not
complete.

In our nine study streams, we found that the size of the watershed, but not
regional species richness, might be used to predict the number of species expected in a
watershed. Watershed size was positively correlated to the estimated species richness of
the study streams (Figure 4, r2=0.614, F=1 1.122, p=0.0125, df=8); however, the regional
species richness was not correlated with the estimated species richness of the streams

(r2=0.463, F= 1.904, p=0.210, df=8).
DISCUSSION

Not surprisingly, there are substantial drawbacks to using Species lists compiled
from existing survey data (Table 7). When only data from infrequent surveys were
available, existing species lists were generally much less complete than lists from an
intensive field survey. However, when data from multiple, repeated surveys were

available, existing data provided a larger species list than was obtained from a single-year

21

intensive field survey. Existing data should, however, be used with caution because (1)
species composition of the stream may have changed since surveys were conducted and
(2) errors in existing data are probable.

The number of survey sources, gears used, purpose of survey, number of years,
and seasons sampled varied among existing surveys and rivers resulting in differences in
completeness of existing lists among rivers. For example, sea lamprey surveys carried
out by the USFWS were conducted every two to five years, contained records from
approximately 1955 to the present, and all species were recorded individually. Similar
surveys conducted by the DFO could only provide observations from 1986 to the present
and resulting species lists were on average much smaller than USFWS lists. Conversely,
a historical database was available for Canadian streams, while in the US historical data
are organized in individual museum collections and databases. For many smaller
streams, only two data sources per river were available, while on larger streams three to
four sources were available. In addition, some species lists were quite current while for
other lists, more than half of the species were last detected before 1985. For example, in
Grindstone Creek nearly half of the species from list A were last recorded before 1960.

Given these differences among lists and the financial advantages of using existing
data, a method to evaluate the completeness of existing Species list is desirable.
Estimating the likely species richness of a watershed, based on its physical and
geographical characteristics, provides a possible method for identifying species lists
which are substantially incomplete. For example, a species list with considerably fewer
species than predicted by the streams watershed characteristics would likely be

incomplete. (Matthews and Robinson 1998) found a positive and significant species-area

22

slope for Arkansas streams. In addition, they found local species richness to be
significantly, although only slightly positively, related to regional species richness. In
our study, we found a positive, significant relationship between watershed size and
estimated species richness. However, the number of species in the regional list, a
measure of regional species richness, was not correlated to the estimated species richness
of our streams. A larger sample size might allow for identification of additional
watershed characteristics correlated to species richness.

Even when the completeness of an existing list can be assessed a priori,
limitations of existing data should be considered. Older lists are likely to be less valuable
since they may not reflect the current species composition of the stream. In addition,
existing species lists will likely contain biases or errors because identification of all
species was not a priority of most existing surveys. In our experience, existing lists often
missed many difficult to identify species such as cyprinids and redhorse suckers. These
species may have been caught during surveys but not identified or recorded because they
are time consuming to identify and not of particular importance to the goals of past
surveys. Furthermore, our existing lists contained several species that were probable
recording errors or misidentifications. In the Au Gres River, as many as seven species
were probable errors.

Our results indicate that even regional species lists may not be complete. These
lists are based on historical surveys and are thought to provide a comprehensive species
list for a region. However, we found species on existing lists that we collected and

vouchered in our field survey that were not included on the regional species lists. This

23

finding indicates that although the regional watershed lists are often considered to include

all species present in the region, even these lists may miss some species.

CONCULSIONS

A comprehensive field survey can provide an accurate species list for an
individual watershed and allow verification of all species captured. However, field
surveys may miss species that only use the stream on a seasonal, annual, or short-tenn
basis. In our field surveys, we most often failed to detect seasonally migrating and lotic
species. Existing lists with multiple repeated surveys may contain more species than an
intensive one-time field survey list because they include species detected during a range
of sampling years and seasons or species that only use the river intermittently. However,
these lists may be of limited usefulness if they are comprised from surveys over 15 years
old or contain errors. Existing lists without multiple repeated surveys were on average
much less complete than field lists and, in general, did not provide an adequate
characterization of species composition.

In cases when repeated surveys are not available, existing survey data are more
than 15 years old, or many species on the list are out of their known range, existing data
are obviously inadequate and a field survey is needed. Comparing the number of species
expected from a streams watershed size might further allow identification of obviously
incomplete species lists. When existing data are likely to be relatively complete and
accurate, managers need to consider if information gained from an additional field survey

will yield information that justified this cost.

24

LITERATURE CITED

Argent, D. G., Bishop, J. A., Stauffer, J. R. Jr., Carline, R. F., and Myers, W. L. 2003.
Predicting freshwater fish distribution using landscape-level variables. Fisheries Research
60: 17-32.

Blackburn, T. M. and Gaston, K. J. 1998. Some methodological issues in macroecology.
American Naturalist 151: 68-83.

Boulinier, T., Nichols, J. D., Sauer, J. R., Hines, J. E., and Pollock, K. H. 1998.
Estimating species richness: The importance of heterogeneity in species detectability.
Ecology 79: 1018-1028.

Bowen, Z. H. and Freeman, M. C. 1998. Sampling effort and estimates of species
richness based on prepositioned area electrofisher samples. North American Journal of
Fisheries Management 18: 144-153.

Braun, D. P., Bach, L. B., Ciruna K.A., and Warner, A. T. Watershed 2000. Watershed-
scale abatement of threats to freshwater biodiversity: The Nature Conservancy's
Freshwater Initiative. 2000.

Colwell, R. K. and Coddington, J. A. 1994. Estimating terrestrial biodiversity through

extrapolation. Philosophical Transactions of the Royal Society of London Series B-
Biological Sciences 345: 101-118.

Conroy, M. J. and Noon, B. R. 1996. Mapping of species richness for conservation of
biological diversity: Conceptual and methodological issues. Ecological Applications 6:
763-773.

De Silva, H. G. and Medellin, R. A. 2001. Evaluating completeness of species lists for
conservation and macroecology: A case study of Mexican land birds. Conservation
Biology 15: 1384-1395.

Dodd, H. R., Hayes, D. B., Baylis, J. R., Carl, L. M., Goldstein, J. D., McLaughlin, R.
M., Noakes, R. L., David, L. G., Porto, L. M., and Jones, M. L. In Press. Low-head
lamprey barriers darn impacts on stream habitat and fish communities in the Great Lakes
basin. Canadian Journal of Fisheries and Aquatic Sciences.

Fairchild, G. W., Horwitz, R. J ., Nieman, D. A., Boyer, M. R., and Knorr, D. F. 1998.
Spatial variation and historical change in fish communities of the Schuylkill River
drainage, Southeast Pennsylvania. American Midland Naturalist 139: 282-295.

Fausch, K. D., Torgersen, C. E., Baxter, C. V., and Li, H. W. 2002. Landscapes to
riverscapes: Bridging the gap between research and conservation of stream fishes.
BioScience 52: 483-498.

Glover, C. J. M. and Sim, T. C. 1978. A survey of central Australian ichthyology.
Australian Zoologist 19: 245-256.

25

Gomez de Silva, H. and Medellin, R. A. 2001. Evaluating completeness of species lists
for conservation and macroecology: A case of Mexican land birds. Conservation Biology
15: 1384-1395.

Heltshe, J. F. and Forrester, N. E. 1983. Estimating species richness using the j ackknife
procedure. Biometrics 39: 1-1 1.

Kodric-Brown, A. and Brown, J. H. 1993. Incomplete data sets in community ecology
and biogeography: A cautionary tale. Ecological Applications 3: 736-742.

Kremen, C. 1994. Biological inventory using target taxa: A case study of the butterflies
of Madagascar. Ecological Applications 4: 407-422.

Lyons, J. 1992. The length of stream to sample with a towed electrofishing unit when fish
species richness is estimated. North American Journal of Fisheries Management 12: 198-
203.

Mandrak, N. Canadian Department of Fisheries and Oceans, Burlington Ontario.

Master, L. L. 1990. The imperiled status of North American aquatic animals. Biodiversity
Network News 3. 1-2.

Master, L. L. 1991. Assessing threats and setting priorities for conservation. Conservation
Biology 9: 101-118.

Matthews, W. J. and Robinson, H. W. 1998. Influence of drainage connectivity, drainage
area and regional species richness on fishes of the interior highlands in Arkansas.
American Midland Naturalist 139: 1-19.

McLaughlin, R. T., Carl, L., Middel, T., Ross, M., Noakes, D. L. G., Hayes, D. B., and
Baylis, J. R. 2001. Potentials and pitfalls of integrating data from diverse sources:
Lessons from a historical database for Great Lakes stream fishes. Fisheries 26: 14-23.

Oliver, 1. and Beattie, A. J. 1996. Designing a cost-effective invertebrate survey: A test
of methods for rapid assessment of biodiversity. Ecological Applications 6: 594-607.

Palmer, M. W. 1990. The estimation of species richness by extrapolation. Ecology 71:
1195-1198.

Ponder, W. E, Carter, G. A., F lemons, P., and Chapman, R. R. 2001. Evaluation of
museum collection data for use in biodiversity assessment. Conservation Biology 15:
648-657.

Porto, L. M., McLaughlin, R. L., and Noakes, D. L. 1999. Low-head barrier dams restrict
the movement of fishes in two Great Lakes streams. North American Journal of Fisheries
Management 19: 1028-36.

Pringle, C. M. 1997. Exploring how disturbance is transmitted upstream: Going against

26

the flow. Journal of the North American Benthological Society 16: 425-438.

Quinn, J. W. and Kwak, T. J. 2003. Fish assemblage changes in an Ozark River alter
impoundment: A long-term perspective. Transactions of the American Fisheries Society
132: 110-119.

Remsen, J. A. 1994. Use and misuse of birdlists in community ecology and conservation.
AUK 111: 225-227.

Richter, B. D., Braun, D. P., Mendelson, M. A., and Master, L. L. 1997. Threats to
imperiled freshwater fauna. Conservation Biology 11: 1081-1093.

Scott, J. M., Davis, F ., Csuti, B., Noss, R., Butterfield, B., Groves, C., Anderson, H.,
Caicco, S., D'Erchia, R, Edwards, T. C. Jr., Ulliman, J ., and Wright, R. G. 1993. Gap

analysis: A geographic approach to protection of biological diversity. Wildlife
Monographs 123: 1-41.

Shaffer, H. 8., Fisher, R. N., and Davidson, C. 1998. The role of natural history
collections in documenting species declines. TREE 13: 27-30.

Vemhes, J. R. and Younes, T. 1993. Inventorying and monitoring biodiversity under the
Diversitas programme. Biology International 27: 3-14.

Winston, M. R. and Taylor, C. M. 1991. Upstream extirpation of four minnow species
due to damming of a prairie stream. Transactions of the American Fisheries Society 120:
98-105.

27

 

:288800

 

oZ mmi >033. 6:233 bfiaflb 33 w «mom 8: :2» «893883.
8:388:00 _S:68:o:>:m
oz mmi 868888.. 8:05: 33 o E 828395 38m 80> 302
0 Z mum .mmi 888?. 889808 5:8 33.33 \. >o:ow< :ocoeoi _S:08:o:>:m 020
58383 83m :Bowmw::o> 80a 82—888: >n 8863:: 28m :Boswgn; a.
o8om mmd @2888 £88-? 8:: 8o8mmomwm 86on :8: Bow 5 88:83“ #8532 Lo EoEtaaoQ 2:0
5:30
02 ”id 808883 >233 :8?» 888m 53 _ _S:08:o:>:m Co 868830Q 58:22
8:86:00
8.» 088 >083 38:38 :mc 80¢ $888 2888mm: 9%.: a >m£o>5£o~ 58383 =68o0
835 83385
8> 088 8 m>o>8m >:8:o>8 fie 80¢ £888 32.852 momfl o .m 8385 83: can 8&3 380
oz 088 a. mud >033. 835. 8883 mowfiéonofi o .m 88:80”— 1852.? >8ng 083:0
28825 38>qu 8300
:2 mid ans; 86836me :5 8898 .68:8 >898: mom =a 888932 a .o .m :5 8:0ng .8 8688305 8:885
>033 380 808883 88.83%
oZ .mmé .mmi 86on 38am .8 808:: St: >033 388% 32-83 v .m .m J 88¢ Z («c 8088309 58:22
£888 888 838mm
8> 088 m>o>8m 88:38 sud 80¢ 8:88 32:88: Bo.“ a a. Roma— v .m .N J :88on 8E :meomE we £8383
£8825 8838: w 4.
:2 mid .8881 80833.8 :5 $888 408:8 >688:— aom =« 8808.32 .v .m .m 4 338m 8:235 :5 5E 35m 3:83
3:26:00 803 ”80>:
88230 >o>8m 8 m>o>8m :03? :8 8a:
Eunoso> :83 880 >o>8m .8 988.30 w88: 88C @0285 8:: >985 wEExo w82>8m >o:ow<

 

6:98:8U Hm .:oooomMHw .Sao::ooun .o:>om No .530 Emnm coo—05m”:
.Botwmnm .:oowE 22qu £80 3‘": 528:: :33 .SE._. 622.882 05 mo 888:3: 38:8 388$ 8 :88 28825
.383 mm:— ::m 988.83 mud mm 3338an a wEficoboo—m 8:28:08: 32:88 :5 88:8 8% >083 859m A 2an

28

Table 2. Field survey completeness. Rivers with an * indicate the estimate is a lower
bound because the species accumulation curve was not clearly reaching an asymptote.

 

 

River Observed number Estimated number % of sest observed
0f Species of species (Sest) in field survey
Au Gres 60 62.9 95
Little Pigeon 32 35.9 89
Harlow 18 20.9 86
Sucker 22 23.9 92
Big Otter* 43 49.9 88
Boyne* 30 35.7 84
Conneaut“ 53 62.7 85
Raccoon“ 25 29.8 84
Grindstone 38 42.8 89

 

Table 3. Number of species unique to existing list A and the field list in a pairwise
comparison.

 

River Number of Number of
species unique species unique
to existing list A to the field list

 

 

Au Gres 20 11
Little Pigeon 12 13
Harlow 17 1
Sucker 10 1
Big Otter 19 3
Boyne 6 7
Conneaut 29 2
Raccoon 11 8
Grindstone 9 10
Average 14.8 6.2

 

29

Table 4. Contribution of annual sampling to the size of species lists. This analysis was
only conducted for the three rivers with two years of field sampling and the six rivers
with data provided by the USFWS (the only source of multiple repeated data including
dates of capture for all individuals).

 

 

 

Number of years Average number Number of species Number of species

a survey was of species found only detected during only detected during
River conducted per survey the 2001 field survey the 2002 field survey
Au Gres 33 14 5 6
Little Pigeon 26 5 2 2
Harlow 42 5 - -
Sucker 45 7 - -
Conneaut 24 10 10 5
Raccoon 18 4 - -
Average 31.3 7.5 5.7 4.3

 

Table 5. Effect of survey age on quality of existing species lists. The date species were
last detected on list A was recorded and species were grouped into the three date
categories below. The percentages of species from each date category that were not
redetected in our 2002 field survey are given below.

 

Percentage of species not redetected in our 2002 field
survey that were last detected on existing surveys:

 

 

 

River Total number of After Between 1984- Before
species on list 1984 1960 1960

Au Gres 69 25 57 100
Little Pigeon 31 31 18 0
Harlow 36 42 45 100
Sucker 3 8 41 50 -
Big Otter 60 15 33 100
Boyne 29 20 24 -
Conneaut 8O 34 100 -
Raccoon 27 30 100 -
Grindstone 39 16 O 44 .
Average 42.33 29 48 69

 

30

Table 6. Probable errors in existing species lists. Species on the existing list A but not on
the regional species list were identified as possible errors. A few of these species were
detected and confirmed during our 2002 field survey, however, the remaining species
(column three) were probable errors.

 

Detected on existing lists but not Number of these species not

 

 

River on the regional list detected in our field survey
Au Gres 12 9
Little Pigeon 2 1
Harlow 3 2
Sucker 4 4
Big Otter 0 0
Boyne 0 0
Conneaut 9 7
Raccoon 2 2
Grindstone 5 2
Averge 4.1 3

 

31

Table 7. Summary of the advantages and disadvantages of each species list type.

 

List type

Advantages

Disadvantages

 

Regional
species lists

Provides a list of all species

which might be present in
the stream

Even this list may not be complete

 

 

 

Field lists Completeness of list may be Cannot capture annual variation in species
assessed composition, without multiple years of
. sampling
Represents current spec1es
composition of the stream
Survey can be designed to
prevent biases
Verification of species
identification possible
Existing Captures seasonal and Species composition of streams may have
lists annual variation in species changed since early surveys
comprised composition ,
ofmultiple M t . . May contain errors
re eated ay con am more spec1es , ,
surrve s than a field survey list Can not quantitatively evaluate
y completeness of list
Sampling may not include all sections of
the watershed
Data are not available for all streams
Existing Survey data is widely No method exists to quantitatively
lists from available evaluate completeness of list.
few surveys

Sampling may not include all sections of
the watershed

Lists are often very incomplete

 

32

Grindstone

 

Raccoon
Conneaut

Figure 1. Study stream locations. Note: streams not drawn to scale.

33

    

 

 

 

60 — --:::=:::::::========°
(I) r5a¢::' . - >
G.) 90’ . t ’ : 3 , f
.6 50fl 6” .tytr’yi
(D ,, , ,' /
0 by.” '7777.77' ' v
"5 40" ,, 0 Au Gres
:— , e9?“ ' . .
a) >< LIttIe Pigeon
Q n
E 30 - + Harlow
2 -. A Sucker
'5‘ T J I;’” 5‘ ‘ <l Bovne
'3 > Conneaut
g D Raccoon
0 <> Grindstone
F l l l

 

O 10 20 3O 4O 50
Number of Reaches Sampled

Figure 2. Species accumulation curves for all nine study streams. The cumulative
number of species is plotted against the number of reaches sampled.

34

‘ 90 ._
l 80 ,___ w, -w-# —7— ____ W...— _.__.._. #__L
70 1

 

 

 

Number of Species

 

 

 

 

 

 

 

 

 

 

l
l
, . r 3 . .__ _. l
1 2 3 4 5 6 7 8 9 j
, Species List by River :

 

 

Figure 3. Comparison of the number of species on the field list and existing lists A and B
by river. List A includes species found in all existing surveys, field list includes only the
species detected in our 2002 field survey, and list B includes species detected in all
existing surveys except the sea lamprey control surveys. River number 1=Au Gres,
2=Little Pigeon, 3=Harlow, 4=Sucker, 5=Big Otter, 6= Boyne, 7=Conneaut, 8=Raccoon,
9= Grindstone.

35

707

 

a - -
g 60—
.1:
.2
Ct
.8 50‘ '
0
(D
C?)- e
U 40-—
0 O.
'6
E
-5 30- -
NJ
20 . l l l l l

 

O 100 200 300 400 500
Watershed Size (km2)

Figure 4. Relationship between watershed size and estimated species richness.

36

Appendix A. Summary of species occurrences on the regional, existing, and field species
lists.
Legend: The number 1 indicates presence and 0 indicates absence on a list.
Cell contents

 

Species List

Regional 0
Existing 0
Field

 

OOH}
.—-

t—‘OO’TI

440C)

 

 

Species

 

.jAu': Little Harlow Sucker B 6 Big _3C0n- Rac- Grind?
(common name) Gres Pigeon ' ”‘ Otter _eaBt coon ne
Alewife , A . E A A '
American Eel : '- A
Arctic Grayling - . A
Atlantic Salmon ' ' A "
Banded Killifish C A
Bigeye Chub
Bigmouth Buffalo
Black Bullhead
Black Crappie
Black Redhorse
Blackchin Shiner
Blacknose Dace
Blacknose Shiner
Blackside Darter
Bluegill
Bluntnose Minnow
Bowfin
Brassy Minnow
Bridle Shiner

  

    

 

O
>>>>>>

>OU
>0

 

Um>UU

 

  

U>UUUUU>
w
O>U

>UU>

 

   

Brindled Madtom ”y A * [D A
Brook Silverside ‘ F ' A ' A f A B A
Brook Stickleback D D D _ D D ,. D F

Brook Trout D D D ‘ D D ' D E E
Brown Bullhead B C '13} - A 9 B’ B D D
Brown Trout D G Bf A D D C B
Burbot A A D D A B _ 1_
Central Mudminnow D D D D ' D D D A
Central Stoneroller D A A D D
Chain Pickerel 5*; ,

Channel Catfish A ' D . D A
Channel Darter A A A A A A
Chinook Salmon B F B '- B7 D B A A

37

Cisco

Coho Salmon
Common Carp
Common Shiner
Creek Chub
Creek Chubsucker
Cutlips Minnow
Deepwater Sculpin
E. Sand Darter

E. Silvery Minnow
Emerald Shiner
Fallfish

Fantail Darter

F athead Minnow
F inescale Dace
Freshwater Drum
Ghost Shiner
Gizzard Shad
Golden Redhorse
Golden Shiner
Goldfish

Grass Pickerel
Grass Pike
Greater Redhorse
Green Sunfish
Greenside Darter
Homyhead Chub
Iowa Darter
Johnny Darter
Lake Chub

Lake Chubsucker
Lake Sturgeon
Lake Trout

Lake Whitefish
Largemouth Bass
Least Darter
Logperch
Longear Sunfish
Longnose Dace
Longnose Gar
Longnose Sucker
Mimic Shiner
Mooneye
Mottled Sculpin

tTlUUUlTi>

’TJOUU U

DUO

{'11

mmUUUmUO

mUU>U>U

U

>w>U>U > UU> '11.}

>

CD

>Umm

w C33>>

53>

38

>ww>

w>>>>

U

UUU>§>

U

wwOU>U>U>

U>O

>o>>wc>>>> U>w>ow>

> OUU>>

>U>

U>>>>> >UO> >> >>

>>>D

U

UCDUW

mU>QUU >QUUUU>UU1>U U >

U

>>U Ow

UU>w

>

>

>U>O>>mOU

>>

>ED CD

U>>

U>UUO>>

OUUCDCJ

03>

U

>>>U

Muskellunge
Ninespine
Stickleback

N. Hog Sucker
Northern Pike

N. Redbelly Dace
Pearl Dace

Pink Salmon
Pugnose Shiner
Pumpkinseed
Quillback
Rainbow Darter
Rainbow Smelt
Rainbow Trout
Redfin Shiner
Redside Dace
River Chub

Rock Bass
Rosyface Shiner
Round goby
Round Whitefish
S. RedBelly Dace
Sand Shiner
Sauger

Shorthead Redhorse
Silver Chub
Silver Redhorse
Silver Shiner
Silveij aw Minnow
Slimy Sculpin
Smallmouth Bass
Smallmouth Buffalo
Sockeye Salmon
Spoonhead Sculpin
Spotfin Shiner
Spottail Shiner
Spotted Gar
Stonecat

Striped Shiner
Tadpole Madtom
Tessellated Darter
Threespine
Stickleback
Trout-perch

U CCU [Tl U3> U '11me U>U Um EDUUU

w

0300

03>

>Uw

03>

39

CDIIJUUJITJ m

U

>U

21>

TIUUUI>>U>I>UU>UIJUU>U

w>U>OU> U> U>U>>

> >

>>> 00w U>U>U >wo>> >>

O

01>

>>>

WCD

11>

DUOUUITJUUSUUU

U>CD>

DUO UwO

UUU'JUJU

>0

0>U>> U>UwU

>>>>

O>>

>00

0'11

Walleye

White Bass
White Crappie
White Perch
White Sucker
Yellow Bullhead
Yellow Perch

ODD

OJ’TIU

U>UUD>UI>

>>

>Z>U>

UUUCDUJUJCU

>OU>>>>

U>U

 

40

CHAPTER TWO
SAMPLING EFFORT REQUIREMENTS FOR CHARACTERIZING
FISH SPECIES COMPOSITION IN GREAT LAKES WATERSHEDS
INTRODUCTION

Fish are among the worlds most threatened taxa, with 37 % percent of US.
freshwater species at risk of extinction (Braun et al. 2000). Stream fish are particularly
threatened by watershed level threats such as land-use changes, dams, and water
withdrawals (Richter et al. 1997). Managers need research and information targeted at
the watershed level to address such large-scale human disturbances (Fausch et al. 2002;
Oliver and Beattie 1996).

Watershed-level fish species inventories are often necessary for conservation
planning, environmental assessments, and testing ecological hypotheses. Despite these
needs, examination of sampling effort requirements for stream fish inventories have
targeted only the stream-reach level rather than entire watersheds (Hughes et al. 2002;
Cao et al. 2001; Patton et al. 2000; Bowen and Freeman 1998; Peterson and Rabeni
1995; Angermeier and Smogor 1995; Lyons 1992). An understanding of sampling effort
requirements is necessary to ensure that species inventories are sufficiently complete,
comparable among locations, and cost effective.

The growing interest in the status and conservation of native fish species
necessitates accurate and complete inventories for conservation planning. Often,
conservation resources may be directed at watersheds with the greatest number of
species. Failure to consider the effects of sampling effort on species richness estimates

may result in overemphasis on watersheds that have experienced greater sampling effort

41

and have larger detected, rather than true, species richness. For example, Nelson et al.
(1990) found that areas with relatively high numbers of endemic species were strongly
correlated with sampling intensity.

Environmental assessments and planning also require watershed-level species
inventories. Fish are quite mobile and may be affected by disturbances at distance within
the watershed. For example, downstream watershed development projects, such as dams,
may impact headwater species (Quinn and Kwak 2003; F airchild et al. 1998; Pringle
1997; Winston and Taylor 1991). Knowledge of the species composition in the entire
watershed is necessary to determine if species are present which may be impacted by
such activities. For example, stream development may proceed if species of concern are
presumed to be absent from the stream. However, if the species is present and has simply
not been detected because of inadequate sampling effort, stream modifications may result
in the species’ extirpation from the watershed. Species that are rare in a watershed are
both difficult to detect and potentially vulnerable to ecosystem change.

Ecological studies often require complete watershed-level species inventories.
Inventories are necessary for assessing factors that influence species distributions,
determining environmental correlates of species occurrence (Neave et al. 1992), and for.
studying community interactions (Taylor et al. 2001). Species inventories are also
valuable for documenting range changes, which may be of particular importance given
growing concern over the effects of climate change on aquatic species (Coddington et al.
1991; Gillison and Brewer 1985).

Sampling effort guidelines are a critical aspect of inventorying species and

estimating species richness. Complete inventories require enough sampling effort to

42

ensure that species are not missed; however, excessive sampling wastes resources while
yielding few additional species. Furthermore, the level of completion required for an
inventory determines sampling effort requirements. For some community-level research,
it may not be necessary that the rarest species are detected; however, for other purposes,
such as environmental assessments, detailed knowledge of community composition is
desirable (Cao et al. 2001; Cao et a1. 1998; Vemhes and Younes 1993; Angermeier and
Karr 1986). Guidelines regarding the amount of sampling effort required to characterize
a target percentage of species present in a watershed are needed to develop efficient
sampling protocols and best allocate resources.

The relationship between sampling effort and number of species detected in an
inventory follows a positive monotonic asymptotic curve because many species are found
in the first samples while additional sampling yields a decreasing number of species not
previously detected. This function is often referred to as a species accumulation curve.
Species accumulation curves can provide a quantitative method for assessing
completeness of fauna] inventories, estimating total species richness, and developing
guidelines for the minimum sampling effort needed to reach a target level of inventory
completeness (Moreno and Halffler 2001).

Species accumulation curves have been used to guide landscape-level species
inventories and to develop sampling guidelines for a variety of taxa including plants,
birds, insects, and algae (Moreno and Halffier 2000; Hellmann and Fowler 1999;
Gimaret—Carpentier et al. 1998; Neave et al. 1997; Dobyns 1997; F lather 1996; Soberon
and Llorente 1993). Guidelines for fish inventories, however, have only been developed

for the steam-reach level (Hughes et al. 2002; Patton et al. 2000; Bowen and Freeman

43

1998; Angermeier and Smogor 1995; Lyons 1992). While reach-level sampling
guidelines are needed for some community level research, monitoring, and ecosystem
health assessments (Cao et al. 2002; Cao et al. 2001; Peterson and Rabeni 1995;
Angermeier and Karr 1986), watershed-level guidelines are needed for conservation
prioritization and environmental assessments (Fausch et al. 2002; Braun et al. 2000;
Richter et al. 1997; Vemhes and Younes 1993).

The shape of a species accumulation curve depends on relative abundance and
distribution of species, so the best model to describe the curve will depend on the taxa,
environment, or scale (Colwell and Coddington 1994). They suggested testing all
reasonable models and choosing the model of ‘best’ fit to characterize the relationship
between sampling effort and number of species detected. However, few studies have
quantitatively examined the best model to fit species accumulation curves (see Flather
1996). Choosing the most appropriate model is central to assessing completeness of
species inventories and developing sampling effort guidelines. This information is
particularly important at a watershed level where species inventories may be very costly.

In our study, we sought to: (1) determine the best model to describe fish species
accumulation at a watershed level; (2) use this model to quantify sampling effort needed
to characterize species composition at the watershed level; (3) examine the effects of
model choice on estimates of sampling effort requirements; and (4) determine if
environmental characteristics of our study streams influence sampling effort

requirements.

44

METHODS

Study sites
We selected nine Great Lakes tributary stream systems that differed in location

within the basin, physical characteristics, and fish community types (Figure 1).
Watershed area ranged from 23.6 km2 to 432.8 km2 and streams were third to fifth order.

Six of the nine streams were shallow enough to be sampled with a backpack electrofisher
at all locations, while the three largest streams required the use of a boat shocker in
deeper areas. In this analysis, we only examined wadeable (first through third order)
sections of the watershed to facilitate comparisons among streams. Two streams drained
into Lake Superior, two into Lake Huron, three into Lake Erie, and two into Lake
Ontario. Fish assemblages ranged from highly diverse cold/cool/warm water

communities (>60 species) to low-diversity coldwater communities (<20 species).

Sampling locations

All streams were delineated using 1250,000 scale topographical maps and
stretches of streams were classified as first to fifth order stretches of stream according to
the Strahler method (Strahler 1954). To ensure proportional representation, we grouped
stretches of stream by order and partitioned long stretches so they were not
underrepresented in the selection process. A minimum of six stretches of stream were
randomly selected from each order type. Within each selected stretch, we determined all
access points from the map and randomly selected one access point at which to conduct
sampling. Lyons (1992) recommended sampling a length of approximately 30 times the
mean stream width to characterize fish communities at the reach-level in warm-water

Wisconsin streams. Accordingly, at each access point we sampled a reach of 30 mean

45

stream widths in length. Wadeable reaches ranged from 28 m to less than 1m in width
and we sampled between 20 and 39 reaches of stream per river. To account for seasonal
variability in species composition, all rivers included spring (May and June) and summer
(July and August) sampling periods. Equal effort was allocated to each stream-order

group to include differences in species composition throughout the watershed.

Sampling

We sampled wadeable reaches by backpack electrofishing (Model 12-B Smith-
Root, Inc.). To maximize capture efficiency, we tested and adjusted settings outside the
sampling site. Given our study objectives, sampling consisted of a single pass and
reaches were not blocked. A three-person field crew conducted sampling in a upstream
direction, moving back and forth across the river while moving the anode in a continuous
‘m’ pattern with effort made to cover the entire stream reach area. We focused sampling
effort in areas where fish were most likely to be found (eg. around large woody debris)
and moved quickly through unproductive areas (eg. sandy open areas). Captured fish
were held in large containers for later identification and subsequent release. Fish that
could not be identified in the field were preserved in a 10 % formaldehyde solution and
returned to the lab for identification. Dr. Tom Coon and the Royal Ontario Museum
verified identifications of uncommon species. Lamprey species were excluded from our

analysis because of concerns over field identification of larval lampreys.

Statistical Methods

We developed a Monte Carlo simulation program to generate a species

accumulation curve for each river. Randomization of sample order is needed for

46

organisms which are patchily distributed and sampled in a spatially random manner
(Flather 1996). This smoothing technique prevents heterogeneity among samples and the
order in which samples are added from affecting the curve shape (Reynolds 1996). To
generate a curve our program randomly chose, without replacement, one permutation of
the sequence of all reaches and recorded the cumulative number of species found versus
the number of reaches sampled. We repeated this process 1000 times and used the results
to generate a curve that represented the average number of species found versus the
number of reaches sampled.

Three methods exist for species richness estimations: (1) extrapolation from
species accumulation curves; (2) non-parametric methods based on species distributions
among samples; and (3) parametric models of relative abundance (Walther and Martin
2001; Colwell and Coddington 1994; Palmer 1990). Non-parametric methods are most
often recommended for estimation of species richness, because the estimated asymptote
and slope of species accumulation curves differ depending on species abundance patterns
(Gomez de Silva and Medellin 2001; Soberon and Llorente 1993; Palmer 1990).

Most estimators will underestimate species richness for small sample sizes and no
estimator is universally accepted as best. However, the first order jackknife technique is
often recommended as one of the least biased and most precise estimators (Hellmann and
Fowler 1999; Boulinier et al. 1998; Palmer 1990 and derived in Heltshe and Forrester

1983) and is often used in the fisheries inventory literature (Bowen and Freeman 1998).

This technique estimates species richness (865‘) as a function of the number of observed

species (Sobs), the number of species found in only one sample (r), and the number of

samples taken (n) as:

47

Sest: Sobs + r(n-1)/n
We used this technique to calculate the estimated species richness ($65!) from our data.

We then determined the percentage of Scst actually observed (Sobs) in each survey as

Sobs/Sest-

As suggested by Colwell and Coddington (1994), we tested several plausible
models to determine the model that best characterized the species accumulation function
in our streams. We compared 11 non—linear models (Table 1) commonly used in the
ecological and species inventory literature (Bowen and Freeman 1998; Flather 1996;
Colwell and Coddington 1994; Soberon and Llorente 1993). Although Soberon and
Llorente (1993) suggest using asymptotic curves for well-known taxa and non-
asymptotic curves for poorly known taxa, we tried both types. Models were fit with AD
Model Builder (Version 6.0.2, Otter Research Limited) using a maximum likelihood
technique. We compared among models using corrected Akaike’s Information Criterion
(AIC) which accounts for the number of parameters in quantifying model fit (Bumham
and Anderson 2002). For each river, each model was given a rank from best to worst, 1
to 11 respectively, based on its corrected AIC values. The average rank and range of
rank was then calculated for each model.

Using the model which best fit the data for each river, we estimated the amount of

sampling effort, presented as number of reaches, necessary to detect 80, 90, 95, and 100

% of Scst for each river. We examined our species accumulation curves to determine the

average number of reaches necessary to detect 80 % of Scst. To make sampling effort

48

requirement estimations for detecting greater than 80 % of Sest it was often necessary to

extrapolate beyond the range of our observations.
Because many models can fit the data reasonably well, but may make very
different predictions (Soberon and Llorente 1993), we selected the three best models and

compared their predictions of sampling effort requirements, represented as the number of
reaches needed to detect a percentage of $65,. In addition to the choice of a model, the

choice of species richness estimator may affect sampling effort estimates. To examine

the effect of species richness estimator choice, we compared sampling effort estimates

using each model’s projected asymptote (Sest-A) to estimates obtained using the
J ackknife estimator of species richness (Scst).

We examined relationships between watershed size (kmz), species composition,
and sampling effort requirements using multiple linear regression. Species composition
characteristics included the estimated species richness of the watershed (865,) and the

number of uncommon species, defined as the number of species detected in only one

reach in a watershed. We measured sampling effort requirements as the number of
reaches needed to detect 80 %, of Sest, which was within the range of data for all rivers in

the analysis.

RESULTS

We observed between 18 and 55 species in our nine study rivers (Table 2). The
observed number of species ranged from 78 % to 95 ”/0 of Sest with Sest exceeding the

observed number of species by between 2 and 6 species (Table 2). As reflected in

49

percentage of Scst detected, species accumulation curves more nearly reached an

asymptote in some rivers than others (Figure 2). Because of high variance in Sest and
poor model fit, we excluded from our analysis two rivers (Harlow and Boyne) where less

than 80 % of Scst was observed to ensure our data were sufficient for developing

sampling effort guidelines. Excluding these two rivers, standard error in Sest ranged

from 1.8 to 5.9 species.

Model Choice

Of the 11 models considered in our analysis, two parameter models generally fit
the data poorly while three and four parameter models (6, 10 and 11) best fit the data
according to the corrected AIC values for all rivers (Table 3). Although model 11 was
the best fit for five of the seven rivers, its average rank was 2.9. Model 10 was the best
model for only two rivers; however, its average rank was the best at 1.9. Model 6 was
the second or third best fit for all rivers, with an average rank of 2.6. Plots of residuals
versus sampling effort showed less trending for models 11, 10 and 6 than for the other
models.

Within the range of the data, the best three models (11, 10, and 6) produced

similar estimates of sampling effort requirements (e. g., number of reaches needed to

obtain 80 % of $65,). However, for extrapolation beyond the range of the data (e. g., 95-

100 % 865,, Model 11 consistently predicted a greater number of reaches than did Model

10 or 6, which predicted very similar estimates to one another (Table 4). Related to this

50

trend, the asymptote of Model 11 was less than Scst in all but one river, so could not be

used to estimate the number of reaches needed to find 100 % of Scst for most rivers.

Sampling Effort Requirements

The best model for each river was used to estimate the sampling effort required to
detect 80, 90, 95, and 100 % of Sest (Table 4). We estimated between 9 and 25 reaches,
with an average of 15.1 reaches, were needed to detect 80 % of Sest. To detect 90 % of

865,, the estimate ranged from 17 to 49, with an average of 30 reaches. To find 95 % of

Sest, between 30 and 98 reaches, with an average 49 reaches, should be sampled. River

5 and 1 always produced the most extreme estimates of sampling effort requirements.

Excluding these rivers, estimates of sampling requirements to detect 80, 90, and 95 % of
Scst were much more similar between rivers, only differing by 5, 7, and 12 reaches,
respectively.

To determine 100 % of 865,, we used model 10 for all rivers, because using model
11 this value could not be estimated for rivers 2, 4, 5, 7 and 8 because model 11 reached
an asymptote lower than Sest. We estimated that between 76 and 151 reaches, with an

average of 119, should be sampled to determine 100 % of Sest. Using model 10, the

number of reaches needed to detect 100 °/o of Scst could not be estimated only in the case

of River 8.

Influencing Factors

The choice of species richness estimator had less effect than model choice on

estimates of sampling effort required to achieve 95 % of Sest. The difference in sampling

51

effort estimates obtained from two different species richness estimators, the Jackknife
(SCSI) or Asymptotic (Sest-A), was less than the difference between estimates from models
10 and 11 (Table 5). Average sampling effort estimates using model 10 and 11 differed
by 22 reaches, while the average difference in estimates using Sest and SCSI-A was only 8

reaches. We were not able to use Model 6 in this comparison, because it is a non-
asymptotic function.

River characteristics did not significantly influence estimates of the number of

reaches required to determine 80 °/o of Sest. The multiple linear regression was not
significant (r2: 0.772, F= 3.393, p= 0.171, df= 6) indicating that watershed size (kmz),

estimated species richness, and the number of uncommon species did not obviously affect

sampling effort requirements in our study streams.

DISCUSSION

Extrapolation from any model is risky; however, extrapolation from species
accumulation curves has worked well for determining sampling effort requirements in
previous studies (Hughes et al. 2002; Moreno and Halff’ter 2000; Bowen and Freeman
1998; Lyons 1992). We found that, on average, 15, 29, 49, and 119 randomly selected

reaches of stream stratified by order should be sampled to detect 80, 90, 95, and 100 % of
Sest, respectively. On our study rivers, sampling requirements would involve
approximately 3, 5, 9, and 20 field-days (with a three person crew) of sampling,
respectively. As the desired percentage of Sest increases, however, the range among

streams in the number of reaches required also increased (Figure 3). Thus these averaged

52

estimates should be used as a rough planning guide because sampling time will be
influenced by travel time, river access, crew experience, and size of the stream.
Environmental correlates were not useful in predicting the number of reaches

required to detect 80 % of Sest. In our study, the size of the watershed was not

significantly correlated with sampling effort, indicating that watershed size may not

influence the number of reaches needed. In fact, one of our largest rivers required the
least number of reaches to detect 80, 90, and 95 % of Sest. Similar results have been

obtained at the reach-level, where many studies have found that stream width is not
correlated with sampling effort requirements (Hughes et al. 2002; Lyons 1992) and some
studies have even reported a negative relationship between channel size and sampling
effort requirements (Paller 1995). Although larger rivers may not require more reaches,
sampling such rivers will require more time because reach length is based on stream
width.

It is not possible to ensure that all species in a watershed are detected; however,
we believe our sampling was reasonable for assessing sampling effort requirements in
Great Lakes watersheds. Streams that were insufficiently sampled were excluded from
our analysis and species accumulation curves for all rivers used in our analysis were
approaching an asymptote, minimizing the chance of biases in our jackknife estimates of

species richness (Hellmann and Fowler 1999; Palmer 1990). The uncertainty in our
estimates of $65,, was not correlated with sampling effort requirements (r2= 0.096, df= 6,

F: 0.531, p= 0.499), suggesting that our analysis was not influenced by completeness of

our field survey.

53

Good model fit and knowledge of the system are necessary for choosing an
appropriate model for extrapolation. Comparing the fit of several models improved our
ability to accurately estimate sampling effort requirements. For example, we found that
two parameter models provide a poor fit to the data and often exhibit biases at the
extremes of the curves. Bias, particularly at the extreme of the data range may result in

considerable overestimates when extrapolating sampling effort requirements for detecting
a high percentage of Sest. However, many models used in species inventory studies

have demonstrated such biases (Soberon and Llorente 1993).

In a comparison of models for characterizing species accumulation curves for
breeding bird surveys, Flather (1996) found models 11 and 10 used here, to be the first
and second best fit within the range of data, respectively. Of the models we tested, we
found model 10 to be the best model for extrapolating sampling effort requirements from
species accumulation curves for stream fish. Although it provided the best fit for only
two of the seven rivers, model 10 best fit the data overall with an average rank of 1.9, and

was always at least the third best model for all rivers. Model 11, the best fit for five of

seven rivers, ranked as low as sixth for the other two rivers. In addition, model 11
reached an asymptote very rapidly, and its asymptote parameter was often less than Sest.
As a result, model 11 could not be used to estimate sampling effort requirements for 100
% of Sest and often produced estimates that appeared unreasonably high by visual

inspection of species accumulation curves.
Species accumulation curves provide a method for quantitative assessment of
fauna] inventories (Colwell and Coddington 1994; Soberon and Llorente 1993) allowing

measurement of the percent of species detected, development of sampling effort

54

guidelines, and comparisons among watersheds. The use of species accumulation curves
for species inventories is widely accepted; however, there are several limitations to the
effectiveness of this technique. No species richness estimator or model to characterize
curves is universally accepted (Bunge and Fitzpatrick 1993). Although the first order

j ackknife estimator has performed well in many studies (Hellmann and Fowler 1999;
Palmer 1990) it is not possible to ensure that it is the best estimator without knowing the
true species richness of a watershed. Sampling estimates using model 10, on average the
model of best fit, appeared reasonable; however, as for species richness estimators, there
is no guarantee that the model selected to best fit a species accumulation curve is the best
model for extrapolation. We found that use of a different estimator had a smaller effect
on sampling effort estimates than varying the model choice, indicating that model choice
may be a more important factor in developing sampling effort guidelines.

Even with proper guidelines and extensive sampling effort it is never possible to
ensure that all species in a large area have been detected (Remsen 1994). This concern is
especially relevant in streams where fish species are patchily distributed (Angermeier and
Smogor 1995) and rare species may only be located in one small area the watershed.
Given the impossibility of proving the absence of rare species (Cam et al. 2000), it may
be prudent to assume that a species of concern, known to occur in the surrounding region,
is present in the watershed or to target sampling for detecting species of concern.

Techniques of species richness estimation and modeling of species accumulation
curves are only valid for assessing survey completeness for the seasons or years sampled
and for the species that are susceptible to the sampling methods and gears used.

Although electrofishing is considered the least biased and most efficient sampling

55

method for stream fishes, it is inefficient for detecting some cryptic fishes (Quinn and
Kwak 2003; Reynolds 1996). Species that are unsusceptible to electrofishing will be
unaccounted for using this technique. In addition, species which are present in the
watershed only for a few days during spawning runs would not be detected by most
standardized sampling protocols. Likewise, extremely rare species may only be detected
in a standardized protocol if their specific habitats are randomly selected or their habitats
are specifically targeted. Depending on the inventory objectives, using additional gears
or targeting rare species, species that use the stream on only a seasonal or annual basis, or
species not susceptible to electrofishing may improve detection and increase the rate of
species accumulation (Conroy and Noon 1996) but at the cost of inventory
standardization. Watershed-level sampling effort guidelines would greatly benefit from
studies addressing rates of species accumulation in larger watersheds, the effects of
multiple gears, and effects of additional seasonal sampling and annual sampling on rates

of species accumulation.

CONCLUSIONS

We found that an average of 15, 29, 49, and 119 randomly selected, reaches of
stream stratified by stream order, should on average be sufficient to detect 80, 90, 95, and
100 % of species present in Great Lakes watershed. However, these estimates varied by
river and should be used as a rough planning guide. While river size does not obviously
affect sampling effort requirements in number of reaches, sampling streams with wider
channel widths will increase sampling time. Because sampling effort requirements will

vary by river, to prevent under or over-sampling, researchers should plot species

56

accumulation curves in the field and adjust sampling effort to achieve specific objectives

of inventory completeness.

57

LITERATURE CITED

Angermeier, P. L. and Karr, J. R. 1986. Applying an index of biological integrity based
on stream-fish communities: Considerations in sampling and interpretation. North
American Journal of Fisheries Management 6: 418-429.

Angermeier, P. L. and Smogor, R. A. 1995. Estimating number of species and relative
abundances in stream-fish communities: Effects of sampling effort and discontinuous
spatial distributions. Canadian Journal of Fishery and Aquatic Sciences 52: 936—949.

Blackburn, T. M. and Gaston, K. J. 1998. Some methodological issues in macroecology.
American Naturalist 151: 68-83.

Boulinier, T., Nichols, J. D., Sauer, J. R., Hines, J. E., and Pollock, K. H. 1998.
Estimating species richness: The importance of heterogeneity in species detectability.
Ecology 79: 1018-1028.

Bowen, Z. H. and Freeman, M. C. 1998. Sampling effort and estimates of species
richness based on prepositioned area electrofisher samples. North American Journal of
Fisheries Management 18: 144-153.

Braun, D. P., Bach, L. B., Ciruna K.A., and Warner, A. T. Watershed 2000. Watershed-
scale abatement of threats to freshwater biodiversity: The Nature Conservancy's
Freshwater Initiative. 2000.

Bunge, J. and Fitzpatrick, M. 1993. Estimating the number of species - a review. Journal
of the American Statistical Association 88: 364-373.

Bumham, K. P. and Anderson, D. R. 2002. Model selection and multimodel inference: A
practical information-theoretic approach. Springer, New York.

Cam, E., Nichols, J. D., Hines, J. E., and Sauer, J. R. 2000. Inferences about nested
subsets structure when not all species are detected. Oikos 91: 428-434.

Cao, Y., Larsen, D. P., and Hughes, R. M. 2001. Evaluating sampling sufficiency in fish
assemblage surveys: A similarity-based approach. Canadian Journal of Fisheries and
Aquatic Sciences 58: 1782-1793.

Cao, Y., Larsen, D. P., Hughes, R. M., Angermeier, P. L., and Patton, T. M. 2002.
Sampling effort affects multivariate comparisons of stream assemblages. Journal of the
North American Benthological Society 21: 701-714.

Cao, Y., Larsen, D. P., and Thome, R. 2001. Rare species in multivariate analysis for
bioassessment: Some considerations. Journal of the North American Benthological
Society 20: 144-153.

Cao, Y., Williams, D., and Williams, N. 1998. How important are rare species in aquatic

58

community ecology and bioassessment? Limnology and Oceanography 43: 1403-1409.

Cao, Y., Williams, D. D., and Larsen, D. P. 2002. Comparison of ecological
communities: The problem of sample representativeness. Ecological Monographs 72: 41 -
56.

Coddington, J. A., Griswold, C. E., Silva Davila, D., and Penaranda, E. 1991. Designing
and testing sampling protocols to estimate biodiversity in tropical ecosystems.
Dioscorides Press, Portland, Oregon p. 1048.

Colwell, R. K. and Coddington, J. A. 1994. Estimating terrestrial biodiversity through
extrapolation. Philosophical Transactions of the Royal Society of London Series B-
Biological Sciences. 345: 101-118.

Conroy, M. J. and Noon, B. R. 1996. Mapping of species richness for conservation of
biological diversity: Conceptual and methodological issues. Ecological Applications 6:
763-773.

Dobyns, J. R. 1997. Effects of sampling intensity on the collection of spider (Araneae)
species and the estimation of species richness. Environmental Entomology 26: 150-162.

F airchild, G. W., Horwitz, R. J., Nieman, D. A., Boyer, M. R., and Knorr, D. F. 1998.
Spatial variation and historical change in fish communities of the Schuylkill River
drainage, Southeast Pennsylvania. American Midland Naturalist 139: 282-295.

F ausch, K. D., Torgersen, C. E., Baxter, C. V., and Li, H. W. 2002. Landscapes to
riverscapes: Bridging the gap between research and conservation of stream fishes.
BioScience 52: 483-498.

Flather, C. H. 1996. Fitting species-accumulation functions and assessing regional land
use impacts on avian diversity. Journal of Biogeography 23: 155-168.

Gillison, A. N. and Brewer, K. R. W. 1985. The use of gradient directed transects or

gradsects in natural resources surveys. Journal of Environmental Management 20: 103-
127.

Gimaret-Carpentier, C., Pelissier, R., Pascal, J. P., and Houllier, F. 1998. Sampling
strategies for the assessment of tree species diversity. Journal of Vegetation Science 9:
161-172.

Gomez de Silva, H. and Medellin, R. A. 2001. Evaluating completeness of species lists
for conservation and macroecology: A case of Mexican land birds. Conservation Biology
15:1384-1395.

Hellmann, J. J. and Fowler, G. W. 1999. Bias, precision, and accuracy of four measures
of species richness. Ecological Applications 9: 824-834.

Heltshe, J. F. and Forrester, N. E. 1983. Estimating species richness using the jackknife

59

procedure. Biometrics 39: 1-11.

Hughes, R. M., Kaufmann, P. R., Herlihy, A. T., Intelmann, S. S., Corbett, S. C.,
Arbogast, M. C., and Hjort, R. C. 2002. Electrofishing distance needed to estimate fish

species richness in raftable Oregon rivers. North American Journal of Fisheries
Management 22: 1229—1240.

Lyons, J. 1992. The length of stream to sample with a towed electrofishing unit when fish
species richness is estimated. North American Journal of Fisheries Management. 12: 198-
203.

Moreno, C. E. and Halffter, G. 2000. Assessing the completeness of bat biodiversity
inventories using species accumulation curves. Journal of Applied Ecology 37: 149-158.

Moreno, C. E. and Halffter, G. 2001. On the measure of sampling effort used in species
accumulation curves. Journal of Applied Ecology 38: 487-490.

Neave, H. M., Cunningham, R. B., Norton T. W., and Nix, H. A. 1992. Evaluation of
field sampling strategies for estimating species richness by Monte Carlo methods.
Mathematics and Computers in Simulation 33: 391-396.

Neave, H. M., Cunningham, R. 3., Norton T. W., and Nix, H. A. 1997. Preliminary
evaluation of sampling strategies to estimate the species richness of diurnal, terrestrial
bird using Monte Carlo simulation. Ecological Modeling 95: ”-27

Nelson, B. W., Ferreira, C. A. C., da Silva, M. F ., and Kawasaki, M. L. 1990. Endemism
centres, refugia and botanical collection density in Brasillian Amazonia. Nature 345:
714-716.

Oliver, 1. and Beattie, A. J. 1996. Designing a cost-effective invertebrate survey: A test
of methods for rapid assessment of biodiversity. Ecological Applications 6: 594-607.

Paller, M. H. 1995. Relationships among number of fish species sampled, reach length
surveyed, and sampling effort in South Carolina coastal plain streams. North American
Journal of Fisheries Management 15: 110-120.

Palmer, M. W. 1990. The estimation of species richness by extrapolation. Ecology 71:
1195-1198.

Patton, T. M. , Hubert, W. A., Rahel, F. J ., and Gerow, K. G. 2000. Effort needed to
estimate species richness in small streams on the Great Plains in Wyoming. North
American Journal of Fisheries Management 20. 394-398

Peterson, J. T. and Rabeni, C. F. 1995. Optimizing sampling effort for sampling
warrnwater stream fish communities. North American Journal of Fisheries Management
15: 528-541.

Pringle, C. M. 1997. Exploring how disturbance is transmitted upstream: Going against

60

the flow. Journal of the North American Benthological Society 16: 425-438.

Quinn, J. W. and Kwak, T. J. 2003. Fish assemblage changes in an Ozark River after
impoundment: A long-term perspective. Transactions of the American Fisheries Society
132:110-119.

Remsen, J. A. 1994. Use and misuse of birdlists in community ecology and conservation.
AUK 111: 225-227.

Reynolds, J. B. 1996. Electrofishing. In Fisheries Techniques. American Fisheries
Society, Bethesda, Maryland p. 221.

Richter, B. D., Braun, D. P., Mendelson, M. A., and Master, L. L. 1997. Threats to
imperiled fi'eshwater fauna. Conservation Biology 11: 1081-1093.

Soberon, J. and Llorente, J. 1993. The use of species accumulation functions for the
prediction of species richness. Conservation Biology 17: 480-488.

Taylor, B. W., Mcintosh, A. R., and Peckarsky, B. L. 2001. Sampling stream
invertebrates using electroshocking techniques: Implications for basic and applied
research. Canadian Journal of Fisheries and Aquatic Sciences 58: 437-445.

Vemhes, J. R. and Younes, T. 1993. Inventorying and monitoring biodiversity under the
Diversitas programme. Biology International 27: 3-14.

Walther, B. A. and Martin, J. L. 2001. Species richness estimation of bird communities:
How to control for sampling effort? Ibis 143: 413-419.

Winston, M. R. and Taylor, C. M. 1991. Upstream extirpation of four minnow species

due to damming of a prairie stream. Transactions of the American Fisheries Society 120:
98-105.

61

Table 1. Description of models. All models exhibit a general non-sigmoidal increasing
shape and are described in Ratkowsky (1990) or Flather (1996). Variable S = the number
of species detected, X = the number of reaches sampled, and A, B, C, and D= parameters.

 

 

Model Function Parameters Asymptote
1 S=AX/(1+BX) 2 Yes
2 S=AXAB 2 Yes
3 S=A+BLogX 2 Yes
4 S=A/(1+B*X"-C) 3 No
5 S=AX"(BXC) 3 No
6 S= (A+BX)/(1+CX) 3 No
7 S= (A/B)( l -(e"(CX)) 3 No
8 S=A(1-e"(-B*X-C) 3 Yes
9 S=A(1-e"-BX)"C 3 Yes
10 S=A(1-(1+(X/ B)"C)"-D) 4 Yes
1 1 S=A[1-e"(-(B(X-C)"D))] 4 Yes

 

Table 2: Sampling characteristics by river. Columns from left to right indicate the
number of species observed (Sobs) and estimated (Sest), difference between Sest and Sobs,

standard error of the estimate of Sest, the percentage of Sest observed during the field
survey, and the total number of reaches sampled per river.

 

 

River sobs sest Sest'sobs Standard Percentage Reaches
error of Sest of Sest Sampled
Au Gres 55 57.9 2.9 2.7 95 39
Little Pigeon 32 35.9 3.9 3.3 89 28
Harlow 18 22.8 4.8 7.7 79 25
Sucker 21 22.9 1.9 1.8 92 33
Big Otter 24 28.8 4.8 4.0 83 30
Boyne 30 35.7 5.7 9.6 78 20
Conneaut 35 39.8 4.8 5.5 88 22
Raccoon 25 29.8 4.8 5.9 84 . 22
Grindstone 34 38.8 4.8 3.7 88 23

 

62

 

 

 

 

63

c; m; elm m-~ 9v YN CTN. SA: 5% fl Tm ~ Ta mxcmmho owcwm
ad a; :4 5.0 may 9N o.w v.3 o6 5w 2: 353:0 owSo><
QmT NE- m.mm- mHm- 54%. _.mm. 0.3 ~13 Won #63 Q: ocowmcctO
WE- m.mm- Dom- QRN Ndm ofiv- wd Eh de- Qwfi Woo :oooomm
N. K- Wmm- ©.wN- m.m- adv- msm- fimm 06w m6 Ywm Wow 38550
ONNT o4 _ T m.wm- v.34 ©.wm- mdo- Ndm oéw ndm- Rum N3 330 Em
9N2- Woo- mdn- _.N~ v.wm- Woo- wsm v.53 v.0 adv v.2: gov—05m
m.mm- mdm- m. _ m- 04m 93- 9mm- _.mm 5.2: wé flow 562 comma 2E..—
N._ T ogw- mdm- «mm 0.2- 5.21 NAON mdmm mdo— v.3; 56$ v.80 =<

: 2 a w n o w v m N M 53m

$602

 

$39 92 as 65 E c.8585

8a 82? UH< $62.80 no woman 53.“ 38 6on some HE 5.ch mo owed.“ 98 SEE owSo>< AE Emma 83me 82? gamma:
B32 swov E Stop a 8:865 5955 832 < 286:8ch go Benz: .38 889:8 E 3on mo 038:8 @0958:— 825388
a 8m 83g UH< “3808.80 .82? U~< “388:8 65 .3 332?: 8a 8553330 E Mo 36580 .82." 3 E 682 ”m 633.

 

 

 

 

 

mwéw 0: <2 93 :3 Wow mdm Pom 2r. v.2 mm Z ES - owfio><
+oom f _ Z we om Z cm mm m m 3 3 E o _ ocowmnctw
mo +oom Z 3 cm 3 R mm em M: C C 2 cocoowm
K E. 2 mm am am mm mm om 2 2 2 I 58500
v0 02 Z we on mo 3 co av mm mm mm _ _ 698m
m2 2; +oom ow 3 co mm mm mm 3 2 2 _ _ coxozm
on ma Z «V 3 mm om om cm 2 2 2 2 ~5va 635
mm _m_ Z on OM cm 2 C S m a a 3 RED s<
mom o\o o2 amom .xo 3 wow o\o om Hmom o\o cm :6on 83M

0 E Z c S 2 o 2 I c o_ : “mom

Home. 382

 

.mEoEEEUE town wEEEwm mo 638:8 33:88:

was nwE bo> m 8:865 +oom 6&38—3 .aom 5.5 $32 Boafixmm am 3:282 $608 2: 3283 338:3 on
8c 258 8588252 tote mezufimm 8:8me Z .95 2828 E 3%: fl haze some Sm E 3on “won 9:. .o cum.
.3 .2 2332: mai— mom («0 o\o o3 can do do .ow Efino 9 @338: meson»: .«o 538:: 65 mo «53.8800 .v 2an

64

Table 5. Comparison of two factors influencing sampling effort requirements: choice of
model and species richness estimator. The table presents the effects of two techniques to
estimate species richness -the jackknife estimator (J) and the asymptotic estimator(A)- on
estimates of number of reaches needed to detect 95 % of the estimated species richness.
The average estimate of species richness estimator choice is compared to the average
estimate of model choice in the bottom two rows. The difference between average
number of reaches estimated from the two estimators is less than the difference in the
average number of reaches estimated with different models.

 

 

 

 

River Estimator Model 11 Model 10
Au Gres A 22 34

J 30 30
Little Pigeon A 40 91

J 53 47
Sucker A 33 72

J 60 49
Boyne A 57 131

J 98 7O
Conneaut A 29 88

J 59 39
Raccoon A 40 57

J 65 54
Grindstone A 26 155

J N 50
Model average 11= 47.08 10= 69.07
Estimator average A = 62.5 J = 54.15

 

65

    

Sucker

Grindstone

Raccoon
Conneaut

Figure 1. Map of study streams. Note: streams not drawn to scale.

66

 

Cumulative Number of Species

 

 

 

O l l l l
0 1O 20 3O 40
Number of Reaches Sampled

Figure 2. Species accumulation curves for nine Great Lakes streams. River 1=Au Gres,
2=Little Pigeon, 3=Harlow, 4=Sucker, 5=Bi g Otter, 6= Boyne, 7=Conneaut, 8=Raccoon,
9= Grindstone. The cumulative number of species is plotted against the number of
reaches sampled.

67

200 —

 

‘0
g
81504 3.
2
E .
0
8100-- X RIVER
95 . 1
2 . ' x 2
a)
.Q _ 4 + 4
g 50 5 A 5
Z A e . V 7
i ' < 8
O I I I I I ' 9

 

80 90 95 100
Targeted "/0 of Estimated Species Richness

Figure 3. Sampling effort requirements to achieve a targeted percentage of estimated
species richness. The amount and range of sampling effort, in reaches, needed to detect

80, 90, 95, and 100 % of Sest is presented for seven study streams.

68

CHAPTER THREE
OPTIMIZATION OF SAMPLING EFFORT FOR FISH F AUNAL INVENTORIES
IN MICHIGAN WATERSHEDS
INTRODUCTION

Fish species composition data are a prerequisite for informed stream management
and are increasingly being used for stream management (F ausch et al. 2002; Peterson and
Rabeni 1995). However, systematic surveys of non-game fish have begun in earnest only
in the past 15 years. Particular emphasis has been placed on species inventories and
sampling effort requirements to determine fish species composition at a stream-reach
level (Hughes et al. 2002; Cao et al. 2001; Patton et al. 2000; Bowen and Freeman 1998;
Peterson and Rabeni 1995; Angermeier and Smogor 1995; Lyons 1992) but information
at this scale can be insufficient for addressing large-scale disturbances and preserving
critical habitats at the landscape level (Fausch et a1. 2002). Fish species inventories at the
watershed level can be used for environmental assessment, conservation prioritization,
management decisions, baseline species distribution data, and ecological hypothesis
testing (Vemhes and Younes 1993).

Since management and conservation decisions are often made without adequate
surveys due to lack of time, funds, and sampling guidelines (Fausch et al. 2002; Oliver
and Beattie 1996; Kremen 1994), developing sampling guidelines which reduce the cost
of species inventories should improve the quality of these surveys and the increase the
quantity of species occurrence data. Development of efficient, comprehensive stream
fish inventories should consider: (1) possible biases due to gear or methods, (2) adequacy

of sampling effort, (3) differences in species composition among stream orders, habitat

69

types, and season, and (4) efficiency of sampling effort distribution. Examining
sampling adequacy and minimizing biases helps to ensure inventory completeness.
Inventory efficiency can be improved by examining how differences in species
composition in space and time can affect the rates of species accumulation in field
surveys. Recent studies have evaluated strategies to increase cost effectiveness and set
guidelines for species inventories, demonstrating that stratifying sampling effort,
sampling along diversity gradients, and adjusting sample unit size and sampling method
may increase efficiency of species inventories (Fisher 1999; Gimaret-Carpentier et a1.
1998; Neave et al. 1997; Longino and Colwell 1997; Neave et a1. 1992; Coddington et al.
1991; Gillison and Brewer 1985).

To develop sampling effort guidelines for fish species inventories in Great Lakes
streams, we sampled between 20 and 39 stream reaches on each of nine study streams.
Stream reaches were 30 stream widths in length and randomly distributed throughout the
watershed, stratified by stream order. We found that 95 % of species present in a
watershed could be detected by sampling an average of 49 reaches (Chapter 2). Here we
use data from four of these streams to examine strategies for allocating samples among
space to minimize sampling costs, increase the rate of species accumulation, and improve
inventory efficiency.

Since it is not feasible to evaluate the relative efficiency of alternative sampling
strategies by comparing them in field surveys, we used simulation methods to resample
our survey data to examine rates of species accumulation under different strategies of
sampling effort allocation. Our objectives were to evaluate: (1) the trade-off between the

length of stream sampled at each reach and the number of reaches which can be sampled;

70

(2) the optimum distribution of sampling effort (hours) between watershed stratum
(stream order categories); and (3) the species most likely to be missed given various

strategies for effort allocation in the watershed.

METHODS
Two Lake Huron and two Lake Superior tributary stream systems were
extensively sampled between May and August 2002 (Figure 1). These watersheds

differed in physical characteristics and fish community types. Watershed size ranged
from 23.6 km2 to 392.6 kmz, although all streams were a maximum of third order. All

streams were shallow enough to be sampled comprehensively with a backpack
electrofisher, although small stretches of stream near the mouths required use of a boat
electrofisher in three of the four streams. To facilitate comparisons among streams, data
from these small non-wadeable sections of the watershed were excluded from this
analysis.

Sampling was stratified by stream order to examine the relative efficiency of
watershed-level sampling effort allocation strategies. All streams were delineated using
1:50,000 scale topographical maps and stretches of streams were classified as first to
third order stretches of stream according to the Strahler method (Strahler 1954). To
ensure proportional representation, we grouped stretches of stream by order and
partitioned long stretches so they were not underrepresented in the selection process. A
minimum of eight stretches of stream were randomly selected from each order type.
Within each selected stretch, we determined all access points from the map and randomly

selected one access point at which to conduct sampling. Lyons (1992) recommended

71

sampling a length of approximately 30 times the mean stream width to characterize fish
communities at the reach-level in warm-water Wisconsin streams. Accordingly, at each
access point we sampled a reach of 30 mean stream widths in length. Reaches were
subdivided into 10 three-stream-width sample units so that species accumulation rates at
the reach level could be examined. To examine the effect of seasonal variability on rates
of species accumulation, we sampled in both the spring (May and June) and the summer
(July and August).

All reaches were sampled by backpack electrofishing (Model 12-B Smith-Root,
Inc.). Sampling consisted of a single pass and reaches were not blocked. A three-person
field crew conducted sampling in an upstream direction, moving back and forth across
the river while moving the anode in a continuous ‘m’ pattern, in an effort to cover the
entire stream reach area. We focused sampling efforts on areas where fish were most
likely to be found (e. g., around large woody debris) and moved quickly through
unproductive areas (e. g., sandy open areas). Captured fish were held in large containers
for later identification and subsequent release. Fish that could not be identified in the
field were preserved in 10 % formaldehyde solution and returned to the laboratory for
identification. Dr. Tom Coon verified identification of all uncommon species.

To compare the costs (in hours) of sampling strategies, we recorded travel, setup,
clearout, sampling, and fish identification times. Travel, setup, and clearout were
considered fixed costs of sampling a reach while sampling and fish identification times,
which depend on the number of sample units sampled at a reach, were considered
variable costs. Fixed costs were averaged by river and variable costs, which depend on

the size of the stream, were averaged by river and strata.

72

Stream fish diversity is generally greater in higher order sections of the watershed
(Fairchild et al. 1998; Paller 1994); however, species present in the often fast moving,
colder headwater branches of a stream may not be found in the warmer, turbid areas near
the mouth of the stream. To describe watershed-level patterns of species distribution, for
each stratum we calculated the proportion of the total number of species and the
percentage of rare species, defined as those species only found in one reach. In addition,
we recorded the number of species that were found in only one stratum.

Species accumulation curves are a tool that allows comparison of species
accumulation rates and completeness of an inventory (Longino and Colwell 1997). For
comparison of sampling strategies, we developed a simulation program to generate
species accumulation curves for our rivers. The program randomly chose, without
replacement, a single permutation of the sequence of all reaches, and recorded the
cumulative number of species found versus the number of reaches sampled. By
repeating the simulations 1000 times, we generated a curve representing the average
number of species found versus the number of reaches sampled. These smoothed species
accumulation curves were then used to compare: (1) the rate of species accumulation
considering only spring samples, spring and summer samples combined, and only
summer samples; and (2) the rate of species accumulation achieved by sampling every
second sample unit, every third sample unit, or consecutive sample units.

We modified the program to examine trade-offs between the number of sample
units sampled at each reach and the number of reaches sampled given a fixed amount of
time. We calculated and plotted the average number of species found per sampling hour

when only the first, the first to third, first to fifth, first to seventh, and all ten sample units

73

were sampled. The strategy which demonstrated the greatest rate of species
accumulation was recorded as the optimal number of sample units per reach. We
compared these results to the optimal number of sample units per reach when travel time
between reaches was set to zero to illustrate the importance of travel time costs.

After determining the optimal length (i.e., number of sample units) of reach to
sample, we further modified our program to examine allocation of sampling effort at the
watershed level. We fixed the time available for sampling at approximately 25 % of the
total sampling time originally spent in each river so that we could resample within the
bounds of our data. We then calculated the rate of species accumulation for different
allocations of sampling time among stream order strata ( Table 1). For each river and
allocation rule, average fixed and variable costs for each stratum were used to calculate
the number of reaches that could be sampled in each stratum. The determined number of
reaches from each strata were then randomly selected and the total number of species
found recorded. This process was randomized 100 times and the average of the number
of species detected was recorded. To identify species most often missed under a given
effort allocation rule, we determined the number of times each species was missed in the
100 randomizations. For each allocation rule, the number of species missed in all 100

simulations was recorded.

RESULTS
In all rivers, third order strata contained a greater proportion of the total number
of species, a greater percentage of rare species, and more species found in only one

stratum (Table 2). Between 87 and 95 %, with an average of 9O ”/0 of all species detected

74

were found in third order strata compared to an average of 50 % in first order strata. On
average, more than 2.5 times as many species were unique to third order strata compared
to the other strata. In addition, the greatest percentage of rare species, those only detected
in one reach, were found in third order strata.

Variable costs, however, were generally greater in third order strata (Table 3).
The time required to sample one sample unit in third order strata was on average 2.2 and
1.25 times greater than for first and second order strata, respectively. More than two
thirds of variable costs was sampling time while fish processing time comprised the
remainder. Fixed time varied by river as river size and access will affect travel time.
Actual driving time was slightly more than half of fixed costs while set up and clear out
time comprised the remainder.

When fixed costs are excluded, sampling only one sample unit per reach
maximized rates of species accumulation in all cases (Figure 2a). However, including
travel time, the optimum number of sample units per reach varied by strata and river
(Figure 2b). An intermediate number of sample units per reach, generally between 3 and
7, most often maximized rates of species accumulation. In third order strata, sampling
approximately 4 sample units per reach was on average the best strategy. In first and
second order strata, approximately 6 sample units per reach maximized rates of species
accumulation (Table 4).

Varying the location of sample units within a reach did not further improve rates
of species accumulation. Rates of species accumulation achieved when sampling
consecutive sample units were similar to rates achieved by skipping one or two sample

units at a reach (Figure 3). While no strategy maximized rates of species accumulation

75

across all rivers, sampling consecutive sample units was either the first or second best
strategy for all rivers. Furthermore, skipping sample units should slightly increase total
sampling time at a reach by requiring additional time and moving equipment upstream.

At the watershed level, allocating a greater amount of effort to third order strata
increased rates of species accumulation (Figure 4). As the allocation rule increased from
1 to 8, the effort allocated to third order strata increased from 34 to 100 %. As the
amount of effort in third order strata approached lOO %, the number of species detected
also on average increased. The only exception was river 2, where number of species
detected appeared to increase only slightly and inconsistently as effort allocated to third
order strata increased. This result may be due to the occurrence of many high diversity
wetland areas in the second order stratum of this river. Allocating 50 % of effort to third
order strata and the remaining effort to either first or second order strata, exclusively,
(rule 9 and 10) did not consistently increase rates of species accumulation for all rivers
indicating that, on average, there is no benefit to allocating effort unevenly between first
and second order strata.

Up to 70 % of sampling could be allocated to third order strata before headwater
species were consistently missed in the simulations. As effort to third order was
increased from 34 to 70 %, more species were detected on average with no change in the
number of species missed in all simulations. However, as effort to third order strata
increased from 80 to 100 %, an increasing number of species were consistently missed
(Table 5, Appendix A).

Improving sampling effort allocation at the reach and watershed level

improved efficiency, however, focusing sampling effort in only one season, either spring

76

or summer, did not consistently increase rates of species accumulation across rivers
(Figure 4). In two cases, sampling only in spring maximized rates of species
accumulation while in the other two rivers, summer sampling alone was the optimal
strategy. A combination of spring and summer sampling was always an intermediate
option. In addition, variable costs in spring were on average 22 % greater than in
summer. An average of 5.75 species per river were unique to only one season with only

21 % of these species found in spring.

DISCUSSION

Species richness was greatest in the third order strata where we found the
greatest proportion of the total number species, the greatest percentage of rare species,
and the greatest number of species found in only one stratum. Since third order strata
were on average wider, deeper, and contained more fish to process, variable costs were
also greatest in this stratum. However, several species were only found in the headwaters
or only rarely found in third order strata. Moving from an allocation of 34 to 70 % of
sampling effort in third order strata increased rates of species accumulation with little
effect on the number of species always missed. However, increasing sampling effort
allocation to between 80 and 100 % in third order strata continued to increase the rate of
species accumulation but at the cost of consistently missing a greater number of
headwater species. This result indicates that most headwater species can be detected with
relatively little sampling effort but once sampling falls below a critical number of reaches

in the headwaters, many species will be missed.

77

Allocation of sampling effort should depend on the time available and the
purpose of the survey. For surveys where detecting as many species as possible is
desirable, such as for environmental assessments, we found that allocating 60 or 70 % of
sampling effort in third order strata maximized rates of species accumulation without
increasing the chance that headwater species will be consistently missed. For rapid
inventories where time is limited and maximizing the number of species detected is the
primary objective, allocating 80 to 100 % of sampling effort in third order strata may be
preferable. We found that focusing the vast majority of sampling effort to third order
strata maximized the average number of species detected; however, some headwater
species were consistently missed with this strategy.

At the reach level our results were not as clear. Across our study streams
sampling 6 sample units per reach in first and second order strata and 4 in third order
strata on average maximized rates of species accumulation. However, these results
varied considerably at both the watershed and strata levels. These differences are likely a
result of disparity in the amount of variability in species composition among sample units
at a reach and between reaches in a stratum. In general, sampling fewer sampling units
per reach will improve rates of species accumulation in streams with high variability
between reaches. Rates of species accumulation might be further increased by sampling
many sample units per reach in streams with high travel costs, such as streams with poor
access. Similarly, if a standardized protocol is not essential, at highly diverse reaches,
continuing to sample more sample units may increase inventory efficiency. However,
when standardization is important, based on our surveys, sampling an intermediate

number of channel widths was in general the best strategy. Skipping sample units did not

78

appear to increase rates of species accumulation. Since sampling new habitats, rather
than sample units at a greater distance may increase detection of new species. However,
targeted sampling may affect survey standardization and ability to compare species
accumulation data among watersheds.

When possible, sampling across seasons is desirable because some species
are unique to each season, however, if time limits the survey to one season, summer
sampling appears preferable to spring sampling. More species were found on average in
summer than spring, possibly because warmer temperatures support a higher diversity of
species. Sampling costs were also lower in summer possibly due to lower water levels
which decrease sampling time and may increase ease of species detection. In addition,
repeated annual sampling would likely impact both the number of species detected and
rates of species accumulation, as noted in Chapter One.

Our results suggest that the efficiency of fish species inventories can be
improved by optimizing allocation of sampling at the reach and watershed level. But
there may be other strategies to further increase inventory efficiency. In a simulation
study, Austin and Adomeit (1991) found that ‘informed’ sampling, where participants
with knowledge of the system could select habitats for sampling, yielded the greatest
number of species per effort. Targeted sampling would likely yield additional sampling
in fish inventories, as many rare species are habitat specific and patchily distributed in the
watershed. Studies of tropical insect inventories have concluded that switching gears
more easily detects some species and inventory efficiency varies by gear used (Fisher
1999; Longino and Colwell 1997). For stream fish inventories, use of additional gears

(and boat sampling at the mouth of rivers) would likely further improve rates of species

79

accumulation and number of species detected. Net or trap sampling may be particularly
effective for sampling migratory species in spring and fall. Future stream fish inventory
optimization studies should examine rates of species accumulation using additional gears,
especially in larger deeper streams where electrofishing is likely to be least effective
(Reynolds 1996). Additional stratification by habitat types or surrounding land-use types
may further improve stream fish inventory efficiency, although stratification by habitat

type may be difficult in chanalized or low gradient streams.

80

LITERATURE CITED

Angermeier, P. L. and Smogor, R. A. 1995. Estimating number of species and relative
abundances in stream-fish communities: Effects of sampling effort and discontinuous
spatial distributions. Canadian Journal of Fishery and Aquatic Sciences 52: 936-949.

Austin, M. P. and Adomeit E. M. 1991. Sampling strategies by Simulation. CSIRO,
Australia p. 167.

Bowen, Z. H. and Freeman, M. C. 1998. Sampling effort and estimates of species
richness based on prepositioned area electrofisher samples. North American Journal of
Fisheries Management 18: 144-153.

Cao, Y., Larsen, D. P., and Hughes, R. M. 2001. Evaluating sampling sufficiency in fish
assemblage surveys: A similarity-based approach. Canadian Journal of Fisheries and
Aquatic Sciences 58: 1782-1793.

Coddington, J. A., Griswold, c. E., Silva Davila, 1)., and Penaranda, E. 1991. Designing
and testing sampling protocols to estimate biodiversity in tropical ecosystems.
Dioscorides Press, Portland, Oregon. p. 1048.

Fairchild, G. W., Horwitz, R. J ., Nieman, D. A., Boyer, M. R., and Knorr, D. F. 1998.
Spatial variation and historical change in fish communities of the Schuylkill River
drainage, Southeast Pennsylvania. American Midland Naturalist 139: 282-295.

F ausch, K. D., Torgersen, C. E., Baxter, C. V., and Li, H. W. 2002. Landscapes to
riverscapes: Bridging the gap between research and conservation of stream fishes.
BioScience 52: 483-498.

Fisher, B. L. 1999. Improving inventory efficiency: A case study of leaf-litter ant
diversity in Madagascar. Ecological Applications 9: 714-731.

Gillison, A. N. and Brewer, K. R. W. 1985. The use of gradient directed transects or

gradsects in natural resources surveys. Journal of Environmental Management 20: 103-
127.

Gimaret-Carpentier, C., Pelissier, R., Pascal, J. P., and Houllier, F. 1998. Sampling
strategies for the assessment of tree species diversity. Journal of Vegetation Science 9:
161-172.

Hughes, R. M. , Kaufmann, P. R., Herlihy, A. T., Intelmann, S. S., Corbett, S. C.,
Arbogast, M. C., and Hjort, R. C. 2002. Electrofishing distance needed to estimate fish

species richness in raflable Oregon rivers. North American Journal of Fisheries
Management 22: 1229-1240.

Kremen, C. 1994. Biological inventory using target taxa: A case study of the butterflies
of Madagascar. Ecological Applications 4: 407—422.

81

Longino, J. T. and Colwell, R. K. 1997. Biodiversity assessment using structured
inventory: Capturing the ant fauna of a tropical rain forest. Ecological Applications 7:
1263-1277.

Lyons, J. 1992. The length of stream to sample with a towed electrofishing unit when fish
species richness is estimated. North American Journal of Fisheries Management 12: 198-
203.

Neave, H. M., Cunningham, R. B., Norton T. W., and Nix, H. A. 1992. Evaluation of
field sampling strategies for estimating species richness by Monte Carlo methods.
Mathematics and Computers in Simulation 33: 391-396.

Neave, H. M., Cunningham, R. B., Norton T. W., and Nix, H. A. 1997. Preliminary
evaluation of sampling strategies to estimate the species richness of diurnal, terrestrial
bird using Monte Carlo simulation. Ecological Modeling 95: 17-27.

Oliver, 1. and Beattie, A. J. 1996. Designing a cost-effective invertebrate survey: A test
of methods for rapid assessment of biodiversity. Ecological Applications 6: 594-607.

Paller, M. H. 1994. Relationships between fish assemblage structure and stream order in
South-Carolina coastal-plains streams. Transactions of the American Fisheries Society
123: 150-161.

Patton, T. M. , Hubert, W. A., Rahel, F. J ., and Gerow, K. G. 2000. Effort needed to
estimate species richness in small streams on the Great Plains in Wyoming. North
American Journal of Fisheries Management 20: 394-398.

Peterson, J. T. and Rabeni, C. F. 1995. Optimizing sampling effort for sampling
warmwater stream fish communities. North American Journal of Fisheries Management
15: 528-541.

Reynolds, J. B. 1996. Electrofishing. In Fisheries Techniques. American Fisheries
Society, Bethesda, Maryland p. 221.

Vemhes, J. R. and Younes, T. 1993. Inventorying and monitoring biodiversity under the
Diversitas programme. Biology International 27: 3-14.

82

Table 1. Sampling effort allocation rules.

 

 

 

Allocation Percentage of Time in Stream Order Strata
Rule I II III
1 33 33 34
2 30 30 40
3 25 25 50
4 20 20 60
5 15 15 70
6 10 10 80
7 5 5 90
8 0 100
9 50 0 50
10 0 50 50

 

83

 

om.m - i F - - N - i v i i N - . :202
ad med omd v md mad a mud and w mNO mad 2 mod nwd .:
md m...o Nod .. md 05.9 o cod mmd F cod mod o 00.0 mod __

md ..N.o omd o od mvd o mNd mmd _. mud mmd _. mmd cod _

 

039:3 0.3”. .38 029:: 0.0”. .30... 030E: 0.6m. .90... 0:95.. 0.5m. .90... 039:: 0.0.0. .33 0.050 c.
0m0.0..>|< ..0v..0:.lw 310:0... cgi 0.3.". m0. ...I0..:< 0:30“.

 

 

.0>.m

 

.E 88.0 E :0: :5

= “E: H 80:0 050 E .058 00.00% .0 09:8: 0:. 8 0.30: .E .02. .886 00E: 0:: m0 0:0 E 2:8 3:0 0060mm :0 0:80:
05 0. $0.3: 0:3:3 :50 .3000: 0:0 3:0 E 08:8 0005. 0.080% 0:: 1.0 0wma:00:0: 05 2 E80: 03% .8050 £000 E 3:0,:
00:30:03 0:. E 00.00% m0 89:5: .90: 0:: m0 83:80:: 05 m. .88. .805 020 800:6 .3 5:2:me 3.00% .N 030,—.

84

 

0.: _..mm 0.9. 0.0m N: 0.5 0N? 0.00 ...mm m.mm .__

 

 

 

0.3 :00 0.0: 0.00 0.: 0:0 00: 0.00 0.: 0.00 __
0.0 :00 0.0 0.00 0.0 0. :0 0.0 0.00 0.0: 0.00 _
0_00_.0> 00:0“. 030...; BE 0_00_:0> 00x0. 0_00_.0> 00x0. 030005 00x0. 0:000
mg 0:00.000 Elan 8%.: 00:... meo|z<
:0>.m

 

.0E: wE00000:0 :00 0:0 3:00:00 00:3800 :0 0w0:0>0 0:: 0: 0:000: 0.:0::0>
0:0 00:5800 0E: :300 :00.0 0:0 .0: :00 ..0>0:: .:0 0w0:0>0 0:: 0: 0:000: 000:“: .0520 0:000 wEEE00 0w0:0>< .m 030:.

85

Table 4. Number of sample units per reach that yields the maximum rate of species
accumulation. More than one value is listed when rates were equivalent.

 

 

 

River
Strata Au Gres Little Pigeon Harlow Sucker Average
I 5 10 3, 1 7, 10 6.0
II 5, 3 10 5, 10 3, 5,10 6.4
III 5, 3 3, 5,7 1, 3,5 3, 5 4.0

 

Table 5. Number of species missed in all simulations given different effort allocation

scenarios. Total sampling effort is based on approximately 25 % of original sampling
time.

 

% of effort allocated

to third order strata 34 4O 50 60 7O 80 90 100
Number of species

never detected 1 0 O 1 O 4 11 13

86

 
 

   

Sucker

Little ._
Pigeon

Au Gres

 

Figure 1. Map of study streams.

87

:000: :00
0::: wEEE00 o— 0: : wEEE00 .:0 00:w0:0::0 :0>_w :00: :00 0:00.: 006000 .:0 :0:E:: 0w0:0>0 0:: (:0 0.08005 .maxq. m 0:03”:

 

 

 
    

08.: .0220 5.3 .0 06F .25.; 39:03 .<
230T— ”.301
com 000 com com 00.. 0 com 000 com com cow 0
. _ . . _ Ow . . . h . O_‘
0.. D
N. d
0 + :00 :00
m x
P o
:000:\0:::: 0.0::00
.:0:0::::Z 0

tom tom

segoeds J0 JeqwnN eAneInanQ

 

 

tom tom

88

30-

01
O
l

River 1 River 2

X0

A

O

I

x 0
x

X

Cumulative Number of Specres
8
l
O
X 9
Cumulative Number of Specres
E

x-E)

20“ 5E++
+

x9

*0
X+O

N
O
l
00

 

+>0

 

 

 

_x
O
..
O
_.
_.
_.

10 15 0 5 10 15
20:

._s
01
l
O
01

River 3 River 4

x+0
x40
X+O

6
5
— 6
15 256

x+O
X+O

10—

410
N—O
“-0
+ or

OX

Cumulative Number of Specres
+10
Cumulative Number of Species
8
l
40X
+ 0<
+

 

 

 

0 r l I 0 r l
0 5 10 15 o 5 1o 15

Reach Reach

 

—

Figure 3. Effects of skipping sample units on rates of species accumulation. The
character 0 refers to sampling consecutive sample units, x refers to skipping one sample

units, and + refers to skiping two sample units. Skipping one or two sample units does
not consistently improve rates of species accumulation over sampling consecutive sample
units.

89

Number of Species
N 00 00 (.0 0) g
(0 O —I N 00
l J I l 1 J

N
on
1

 

River 1

 

N
\l

O

101

Number of Species

.r
1

I I I I I
2 3 4 5 6
Effort Allocation Rule

River 3

III
789

 

 

0

I
1

I I I I I I I
2 3 4 5 6 7 8
Effort Allocation Rules

I
9

13-

Number of Species
2 a
l l

.3
o
I

 

River 2

 

O

11.0-

.3

.0

0|
1

10.0:

Number of Species

5°
0:
1

I
1

I I I I I I I fl
2 3 4 5 6 7 8 9
Effort Allocation Rule

River 4

 

 

9.0
0

I
1

IIIIIIII
23456789

Effort Allocation Rule

Figure 4. Number of species found under each sampling effort allocation rule. As the
allocation rule increases the amount of effort allocated to third order strata increases from
34 % to 100 %.

90

60— 30-
River 1

River 2

20—

Cumulative Number of Species
Cumulative Number of Species

 

 

 

 

 

 

 

 

O I I I I I I I I I
012345678910
20- '
0 River3 (n River4
.9 .9
U 0
8 8
a) a)
“510— “515—
(T) b
.D .0
E E
D 3
z z
.“2’ 5— .310—
E E
:5 3
E E
:5 D
O 0
O I I I l I I I I I I 5 I l I l
0 1 2 3 4 5 6 7 8 9 10 0 4 8 12 16
Reach Reach

Figure 5. Rates of species accumulation by season. Characters + refers to summer
sampling only, x refers to spring sampling only, and 0 refers to both spring and summer
sampling. In Stream 1 and 3, rates of species accumulation are greatest in summer. In
Rivers 2 and 4, rates are greatest in spring. Rates from combined spring and summer
sampling are always intermediate.

91

Appendix A. Number of times in the 100 simulations that a species is missed given for
each allocation rule. Results are based on approximately 25 % of original sampling time.

 

 

 

Rule

River Species Name Avg 1 2 3 4 5 6 7 8 9 10
1 Black Bullhead 70.4 63 70 64 72 72 73 77 69 56 79
2 Black Bullhead 94.5 98 96 91 98 89 93 90 92 97 96
1 Black Crappie 92.8 93 97 94 97 96 93 94 92 98 86
1 Black Redhorse 83.5 95 90 89 80 8373 81 74 85 85
1 Blacknose Dace 1.8 4 1 0 0 1'0 3 4 0 1
2 Blacknose Dace 2.9 1 3 2 2 3 7 2 2 4 2
4 Blacknose Dace 2.9 7 4 4 3 4 5 1 2 2 ‘ 0
'1 Blacknose Shiner 19.6 27 30 24 31 23 14 16 18 12 23
2 Blacknose Shiner 76.8 71 81 74 79 75 83 79 65 78 74
4 Blacknose Shiner 68.5 68 74 67 81 54 70 59 67 79 59
1 Blackside Darter 25.0 42 36 25 18 18 15 14 14 31 31
2 Blackside Darter 33.0 52 54 40 39 24 20 12 13 48 39
1 Bluegill 12.5 29 15 18 23 10 3 6 3 10 20
3 Bluegill 74.4 84 84 71 71 82 69 59 70 80 74
1 Bluntnose Minnow 8.3 25 7 9 9 2 6 4 5 8 6
2 Bluntnose Minnow 90.9 90 97 92 92 97 94 92 90 83 95
3 Bluntnose Minnow 76.0 78 77 82 73 76 73 71 59 85 81

' 10 10
1 Brassy Minnow 70.0 52 49 56 65 73 70 0 0 6856
2 Brassy Minnow 89.8 88 90 81 79 89 97 90 93 94 75
1 Brook Silverside 96.1 99 96 96 98 96 94 95 94 98 93
1 Brook Stickleback 52.4 25 16 31 37 61 64 98 96 5 85
2 Brook Stickleback 52.5 18 42 34 34 34 78 81 71 34 52
3 Brook Stickleback 60.4 81 75 69 62 56 41 57 43 67 65
4 Brook Stickleback 30.0 32 32 30 32 31 34 23 27 23 39

10 10
1 Brook Trout 86.6 75 81 78 89 91 90 0 O 81 79

10 10 10
2 Brook Trout 89.3 81 80 85 76 86 0 0 0 83 85
3 BrookTrout 0.1 0 0 0 0 O 0 0 1 0 0
4 BrookTrout 1.8 0 0 0 1 0 1 1 5 1 0
2 Brown Bullhead 94.8 94 98 96 96 95 94 95 84 99 97
1 Brown Trout 76.8 68 65 65 77 75 74 92 90 90 59

92

A—k-D-NAN—‘A-k-waA-bwk)

—L

N

A-hODN-hw

ANAAANA

Brown Trout

Burbot

Burbot

Central Mudminnow
Central Mudminnow
Central Mudminnow
Central Mudminnow
Central Stoneroller
Common Carp
Common Shiner
Common Shiner
Creek Chub

Creek Chub

Creek Chub
Emerald Shiner
Fantail Darter

Fathead Minnow
Fathead Minnow

Fathead Minnow
Fathead Minnow
Finescale Dace
Finescale Dace
Finescale Dace
Golden Redhorse

Golden Shiner
Golden Shiner
Golden Shiner
Greater Redhorse
Green Sunfish
Green Sunfish
Homyhead Chub

Iowa Darter

66.4
87.6
77.8
1.3
15.6
57.8
9.9
3.4
31.4
0.0
32.6
0.0
0.1
60.3
82.1
4.6

82.0
96.1

93.3
89.4
70.8
49.5
86.8
20.8

96.9
82.1
95.9
86.3
1.9
38.5
21.3

77.6

.73

96

88
89
37
38
91
43

94
89
98
92

48
4O

60

93

75
91
88

13
71

50

40

56
83

70
92

89
93
60
31
94
37

94
89
97
92

57
32

64

75
94
83

10
57

43

29

67
89

75

88

89
94
60
28
92
29

96
79
92
90

29
23

71

76
92
81

62
14

28

23

61
82

83

88

91
90
68
31
91
24

99
78
96
86

29
19

74

63
85
78

57
10

31

21

60
70

85

96

94
92
61
62
86
1 8

98
86
96
84

25
14

88

61

83

74

31

46

23

43

65
84

90

'10

92
90
97
57
84
1 0

99
87
96
86

31
14

94
91
92
76
81

10

77
94
76

18

12
10

61 65
81 90
67 76

26 2
43 58
10 6

N
0.)
O5

24 32

57 75
76 83

10
0 6O
10 10

10

86 92
91 46
88 21
82 88

10

72 90
92 98
73 96
0 2
24 48

5 31
10

890054

76
92
83

17
65
18

40

17

32
80

84

81

94
83
73
37
90
29

96
75
98
89

43
21

87

N—t—s-booN—k—l-bOO-tN—th—x-t com—am

N ._L

NAAN-b 00

.h

Iowa Darter

Johnny Darter
Johnny Darter
Johnny Darter

Lake Chubsucker
Largemouth Bass
Largemouth Bass
Largemouth Bass
Logperch
Logperch
Longnose Dace
Longnose Dace
Longnose Dace

Longnose Gar

Mottled Sculpin
Mottled Sculpin
Mottled Sculpin
Mottled Sculpin _
Northern Hog Sucker
Northern Pike
Northern Pike

Northern Redbelly
Dace

Northern Redbelly
Dace
Northern Redbelly
Dace
Northern Redbelly
Dace

Pearl Dace
Pearl Dace
Pumpkinseed
Pumpkinseed

Pumpkinseed
Rainbow Darter
Rainbow Trout

90.5
0.0
60.8
52.5

88:3
4L6

9545
3513
50:3
8015
16:3
2249

5313

96.0
43.9
11.6
5.9
0.0
3.5
45.6
94.8

85.3
35.0
63.3

7045
831
6115
15-4
4543

9813
‘10
50:3

95
0
75

63_

75C

1 7
96
47
78
88

38~

40
73
99
37

16,

13
0
10
63
98

82

7

43

65
65
62
32
51

98
5

94

90
0
78
67

84
9
9.8
56
65
8.8
20
40
68
96
34
25
15
0
10
51
98

71

16
44

71
70
65
17
55

88
0
58
58

89
1
99
46
5O
81
20
32
50
94
39
6
6
0
1
48
96

79

21

43

68
70
73
9
44

91
0
68
51

86
5
98
35
51
82
19
28
54
94
52
15
4
0
6
48
96

79
20
47

68
71
71
23

47
10

’97 99 c)
2 0 2
54 44 46 54

90

58
57

96

94-

42
52
70

18
51
96
51

O-b

38
97

86

14

77

68
82
61
21
39

99

47

90
0
52
42

92

93

‘25

43
83

40
93

38

12

46

98

92.

73
72

76
98
62
10
46
10

41

86
0
42

41
10

95
18
34
68

13
31
98
55

A

36

92
10

59

88

68
93
57
15

38
10

55

85

48‘

26
10

93
12

29 48

68
10
10

42

96
56

A

31

94
10

59
10

74
90
53

35
10

53

34 ,
18.

27
62

98*

52

CO)

52

>94

'62

53

57
91
68
14
57
10

58

89

30

36

80
65
57
17
41

95

37

AAANAAAAQN AAAAwaAAhQQ

#w-ANA-hODN-x—t

Rainbow Trout
Rainbow Trout
River Chub
Rock Bass
Rock Bass
Rock Bass

Rock Bass
Rosyface Shiner

Round goby
Sand Shiner
Sea Lamprey

Sea Lamprey
Sea Lamprey
Sea Lamprey

Shorthead Redhorse .

Slimy Sculpin

Smallmouth Bass
Smallmouth Bass

Spotfin Shiner
Spottail Shiner

‘ Stonecat . ‘
Walleye (Yellow

Pickerel) +_ ,.
White Sucker
White Sucker
White Sucker
White Sucker
Yellow Bullhead
Yellow Bullhead
Yellow Perch
Yellow Perch
Yellow Perch

1.0
0.9
1.9

2.5 ‘

39.9
94.8
89.3

43.0,

45.6
31.3
78.5

94.1
6315
28:1
781
9313
2213
7813
413

6613
5913

96.1

3.5
41.9
38.3
62.6
66.6
81.5
38.8
61.8
22.1

mN-h-b

95
85
67
70
57
88

82
83
43
80
94
37
84
1 5
81

76
10

'12

57

52

81
89
86
50
65
26

93
74
44
82
96
40
93
1 0
77
62

97

54
51
65
77
94
39
61
24

AAOA
ANOO

50 41
97 95
90 89
40 44
49 43
37 29
83 86

93 95
61 68
31 41
80 81
90 95
19 27
84 81

71 69
55 64

99 96

43 4O
43 37
63 69
67 63
83 82
43 41
47 53
20 25

O-ido

40

82
34
39
27
71

93
63
26
71

92_
14‘
85‘

65
58

95

36
40
61

76
35
67
21

NOOO

91
89
31
37
18

79
10

59
25
76
90
14
75

55
47

94

50
30
59
60
84
34
71
25

4000

96
90
29
30
18

74
10

50
12
83
94
1 5
69

56
51

94

31
28
43
60
77
38
61
18

COCO
#NO@

20 43
98 95
98 98
26 48
29 58
£3 43

67 76

1O

41 65
13 40
67 81
94 90

68 83

56 75
43 65

94 99

26 38
18 50

56 77'

49 63
72 76
31 38
58 43
17 29

0000

92
73
50
50
37
80

.99

71
24
83
97

‘23

82.

72
76

97

47

58
71
80
43
76
12

 

95

MANAGEMENT IMPLICATIONS

Implementation of the GLFC barrier policy depends on the ability of managers to
determine species composition in barrier candidate streams. As budgets are limited for
such activities, an understanding of the adequacy of historical data is essential. In
addition, estimates of sampling effort requirements for watershed-level fish community
characterization are necessary to determine the feasibility of the barrier policy and to
prioritize candidate streams. Increasing the efficiency of fish species inventories will
further increase the cost-effectiveness of implementing the barrier policy. Below, I
summarize the recommendation and findings from this evaluation of strategies for
assessing fish species composition in Great Lakes streams.

Existing species lists from multiple repeated surveys contained more species than
our intensive field survey list, but in general were not as reliable because species
composition of the stream may have changed since existing surveys were conducted and
errors, due to the goals of existing surveys, are probable. Given our findings from this
historical data evaluation, existing survey data should only be used when: (1) lists are no
more than 10 to 15 years old, (2) lists are compiled from multiple years of repeated
surveys, (3) the watershed was representatively sampled, (4) sampling was conducted in
spring and summer, and (5) the number of species on existing lists is comparable to the
estimated species richness based on watershed size. It is unlikely that data meeting these
criteria are available for most watersheds.

Sampling effort requirements to detect a majority of the estimated species
richness of a watershed were relatively small; however, sampling requirements to detect

the majority of species were extensive. Sampling an average of 15, 29, 49, and 119

96

randomly selected reaches of stream, stratified by stream order should on average be
sufficient to detect 80, 90, 95, and 100 %, respectively, of species present in Great Lakes
watersheds. Because sampling effort requirements will vary by river, to prevent under or
over-sampling, field crews should plot species accumulation curves in the field and adjust
sampling effort to achieve specific objectives of inventory completeness. Because
species of concern found in the region will likely be the rarest and most easily missed
species in a survey, these species may be identified from the regional list and specifically
targeted. Based on the difficulty of detecting the last 10 % of species, random sampling
to detect 9O % of estimated species richness should be conducted in addition to such
targeted sampling.

Sampling efficiency can be further improved by optimally allocating
sampling effort in the watershed. We found that (1) sampling an intermediate number of
stream widths at a reach and (2) focusing sampling effort in higher order sections of the
watershed can improve rates of species accumulation. Allocating 7O % of sampling
effort increases the average number of species detected without increasing the number of
headwater species consistently missed. Given our results, field crews should sample
reaches of approximately 12 stream widths in length in third order strata and 18 stream
widths in first and second order strata. If a reach is particularly diverse or unique,
additional sampling may be beneficial. At the watershed level, crews should focus up to
70 % of sampling time to third or higher order strata. Distributing this sampling effort

over multiple years is desirable.

97