EPA REGION VIII
Bioassessment Workshop:
Analysis of Biological Data
The College Inn Conference Center
Boulder, CO
September 20,21, and 22,1995
Steering Committee:
Loren Bahls, Montana Department of Health & Environmental Sciences
Mike Barbour, US Environmental Protection Agency
Chris Faulkner, US Environmental Protection Agency
Susan Foster, Thorne Ecological Institute
Jeroen Gerritsen, Tetra Tech, Inc.
Susan Jackson, US Environmental Protection Agency
Phil Johnson, US Environmental Protection Agency
Bob McConnell, Colorado Department of Public Health & Environment
Toney Ott, US Environmental Protection Agency
Bill Wuerthele, US Environmental Protection Agency
Workshop Organizer
Thorne Ecological Institute
5398 Manhattan Circle, Suite 120
Boulder, CO 80303
(303) 499-3647
-------
257
|2"2
0*335
EPA REGION VIII BIO ASSESSMENT WORKSHOP:
ANALYSIS OF BIOLOGICAL DATA
September 20,21 and 22,1995
The College Inn Conference Center, University of Colorado, Boulder, CO
Wprinpsriav. Sentwnhpr 20
1:00 p.m. WELCOME, Betty Lou Carpenter, Thome Ecological Institute, Boulder, CO and Phil Johnson, U.S.
EPA Region VIII, Denver, CO
SESSION I: AN OVERVIEW OF DATA ANALYSIS AND RELATED ISSUES
1:15 p.m. Multimetric Approaches to Bioassessment, Jeroen Gerritsen, Tetra Tech, Inc., Owings Mills, MD
2:IS p.m. Bioassessment Using Predictive Multivariate Models: Clarity, Not Smoke and Mirrors, Robert
Bailey, Dept. of Zoology, University of Western Ontario, London, Ontario
3:15 p.m. BREAK
3:30 p.m. Data Management Issues and GIS Approaches for Biological Assessment, Tony Selle, U.S. EPA
Region VIII, Denver, CO
Philadelphia, PA
9:00 a.m. Trend Detection in Biological Monitoring: Evaluating Patterns of Community Similarity, Donald
Charles, Academy of Natural Sciences, Philadelphia, PA
9:30 a.m. Invertebrate Assemblages and Regional Classification: The Interior Columbia Basin as a Case
Study in Scaling Up, Judith Li, Oregon State University, Corvallis, OR
10:00 a.m. BREAK
10:15 a.m. Bioassessment Using Macroinvertebrates in Low Gradient Reaches of the South Platte River, Jill
Miruer, Dept. of Earth Resources, Colorado StateUniversity, Ft. Collins, CO
10:45 a.m. Developing Biological Criteria for Pacific Northwest Streams, Leska Fore, Institute for
Environmental Studies, University of Washington, Seattle, WA
11:15 a.m. Developing Bioassessment Protocols for Montana Wetlands, Randy Apfelbeck, Montana Dept. of
Environmental Quality, Helena, MT
1
-------
11:45 a.m. DISCUSSION
12:15 p.m. LUNCH
1:15 p.m. An Index of Biological Integrity for the Red River Ecoregion: A Biological" Desert.", Eric
Pearson, North Dakota Dept. of Health, Bismarck, ND
1:45 p.m. Update on the Use of Macroinvertebrates in Montana Stream Bioassessments, Bob Bukantis,
Montana Dept. of Environmental Quality, Helena, MT
2:15 p.m. Use of Lake Benthos as a Bioassessment Tool, Malcolm Butler, Dept. of Zoology, North Dakota State
University, Fargo, ND
2:45 p.m. BREAK
3:00 p.m. Assessing Effects of Metals on Benthic Macroinvertebrate Communities in Rocky Mountain
Streams, Peter Kiffney, Dept. of Fishery and Wildlife Biology, Colorado State University, Fl Collins,
CO
3:30 p.m. The Ordination of Benthic Invertebrate Communities in the South Platte River Basin in Relation
to Environmental Factors, Cathy Tate, U.S. Geological Survey, Denver, CO
4:00 p.m. DISCUSSION AND SUMMARY
5:00 p.m. ADJOURN
Friday. September 22
REGION VIII WORKING GROUP MEETING
(A facilitated discussion among EPA, States and Tribes examining challenges to program implementation)
8:00 a.m. Introduction
8:15 a.m. Sample Collection and Analysis Needs for Region VIII States and Tribes
9:00 a.m. Database Needs within Region VIQ and Possible Solutions
9:45 a.m. BREAK
10:00 a.m. Data Analysis Needs within Region vm and Possible Solutions
10:45 a.m. Maximizing Program Effectiveness through Interagency Coordination: Defining Roles for EPA
Region VIII, the States and the Tribes
11:30 a.m. Next Steps and Action Items
12:30 p.m. ADJOURN
Organized by Thorne Ecological Institute
Thorne Ecological Institute is a non-profit environmental education organization founded in 19SS. Tfcl promotes positive
change by encouraging individual behaviors based on ecological principles that achieve environmental, economic and social
harmony. For more information call (303) 499-3647.
2
-------
MULTIMETRIC APPROACHES TO BIOASSESSMENT
Jeroeti Gerritsen
Tetra Tech, Inc.
10045 Red Run Blvd., Suite 110
Owings Mills, MD 21117
NOTES:
The last decade has seen the development and increasing application of
multimetric biological assessment methods, with the purpose of expressing and
interpreting whether a water body is similar to an undisturbed reference state.
Multimetric indices are additive; i.e., the sum of several measurements or
calculated variables, termed metrics, obtained from sampling an aquatic
biological assemblage. A metric is an ecological attribute of the assemblage,
and should be responsive to perturbation or pollution. The multimetric
approach was designed to be simple to implement for routine, operational
monitoring, although establishment of a program requires substantial sampling
and analysis effort.
Assessments are based on comparison of biological metrics at a site to their
expected value under regional reference conditions. Reference conditions are not
ad hoc special cases, such as upstream or paired reference sites, but should
reflect regional conditions, and regional variability, under minimal human
disturbance. Reference conditions are established by identifying sites that
meet selected criteria for minimal disturbance, and characterizing their
biological condition. A critical element in characterization of reference
conditions is appropriate classification of site types, to ensure that apples are
kept separate from oranges. A common approach to classification has been
geographic categories such as ecoregion or biogeographic province, and
continuous covariates such as stream size or salinity.
Metrics are selected based on low variability in the reference population, and
responsiveness to disturbance or pollution. Metrics are scored on a common scale
and the index is the sum of the selected metrics. Metrics are commonly scored on
an ordinal scale of 1,3,5, or as a percent of the reference value. Experience with
the multimetric approach has shown it to be reliable for detecting impairment
of water bodies. Successful application requires a step-by-step progression to
establish regional reference conditions, to select appropriate metrics, and to
test the component metrics on a regional basis. Significant unresolved issues of
biological assessment include spatial and temporal variability, and optimal
sampling to minimize the effects of variability.
3
-------
Jeroen Gerritsen
Tetra Tech, Inc.
Multimetric Approaches to
Biological Assessment
"It
i.i
Biological Criteria and
Bioassessments
Severely impaired
Non-impaired
Slightly impaired
Moderately impaired
¦ Biological criteria,
determined from the
reference condition,
are used as the
measuring slick for
biological integrity
• Biological assessment
scores are compared
to these criteria to
determine if a site is
impaired
1.2
Traditional Statistical Tests
SSSSSSi "• ¦¦¦¦ ¦
Are these means
significantly different?
1.3
Standards and Criteria
Standard
X
1
1.4
* y ' . Does the mean
significantly exceed the
standard?
Bioassessment
reference distribution
75%
5% /
' / '
\ 95%
\ 1
1
' • V
i
i i"—
1.5
A ^
Is this observation (or mean of
several observations) different
from the reference population?
Assessment Alternatives
~ Single indicator
* Multimetric index (this talk)
Multivariate models (Bob Bailey)
~ Indicator species models (Don Charles)
~ Discriminant function models
(in use in Maine)
1.6
4
-------
Multimetric Approach
»- Development of an index that is the sum
of several metrics, or indicators, of
biological condition
~ The value of each component metric and
the resultant index is based on reference
condition
1.7
Simple and Stupid Index; The index
is the sum of several metrics:
LiL.!
-4- -5
5
-J- ¦
1 3 +
i
i
[a] supporting
—
partially
supporting
not
supporting
1.9
POTENTIAL ATTRIBUTES THAT MAY SERVE AS METRICS
COMMUNITY
TAXONOMIC
INDIVIDUAL
BIOLOGICAL
STRUCTURE
COMPOSITION
CONDITION
PROCESSES
TAX A
IDENTITY
DISEASE
TROPHIC
RICHNESS
STRUCTURE
RELATIVE
SENSITIVITY
ANOMALIES
NUMBERS.
ABUNDANCE
(intolerance)
BIOMASS
DOMINANCE
RARE OR
CONTAMINANT
PROCESS
ENDANGERED
LEVELS
RATES
KEY TAXA
(production.
METABOLIC
predalton,
RATE
recruitment)
1
BIOLOGICAL ASSESSMENT
1.11
Definition
~ A metric is a characteristic of the biota
that changes in some predictable way
with increased human influence.
1.8
Metrics
~ Measurements to determine biological
integrity should:
-be relevant to societal concerns
-be responsive to environmental stresses
-have low uncertainty
-be cost-effective
-be environmentally benign to measure
1.10
Multimetric Approach
~ The strength of using multiple metrics
(attributes) is the ability to integrate
information from individual, population,
community, zoogeographic, and
ecosystem level of a single assemblage
into a single ecologically based index of
the quality of a water resource.
1.12
5
-------
Multimetric Approach
Multimetric bioassessment is not a
ready-made, off-the-shelf instrument. It
is an approach that must be modified to
specific regional conditions before it can
be applied.
1.13
Topics
~ Reference characterization
~ Metric selection
~ Index development
~ Performance and unresolved issues
1.14
Reference Conditions for
Biological Assessment
>- Reference Conditions: Expectations on
the state of biological communities in the
absence of anthropogenic disturbance
and pollution.
»• Reference Sites: Real sites that form a
database for characterizing reference
conditions. These sites should be
minimally impaired by human pollution
and disturbance.
Characterization of Reference
Conditions
*¦ Present-day reference sites
~ Paleo data
»- Historic data
»¦ Expert consensus (SWAG)
2.1
2.2
Reference Site Criteria
~ All within category
~ Least impacted within the context of the
ecoregion or class
~ Least impacted must be determined by direct
evidence of human activity, not by biological
responses
-population
-roads
-discharges
-structures (canals, dikes, armor)
-hydrological modification
2.3
How Undisturbed is
Undisturbed?
Requires decision on acceptability
•May differ among geographic regions
•Relative scale of impairment
2.4
6
-------
Regionalization and Preliminary
Classification
- The intent of classification is to identify
groups of sites that under ideal condition
would have comparable biological
communities.
~ Classification should rely on those
characteristics of sites that are intrinsic,
or natural, and not the result human
activities.
2.5
Testing the a priori
classification
.. ,
~- graphic analysis
~ ANOVA, MANOVA and discriminant
analysis
~ Ordination
2.7
Graphical Analysis
~ The fundamental problem of biological
assessment is not to determine whether two
populations (or samples) have a different mean,
rather to determine whether an individual site is
a member of the least-impaired reference
population.
»- Rather than statistical methods of analysis
testing whether two or more populations have
different means, the entire distribution of a
metric is effectively displayed with a
box-and-whisker plot.
_^9
Classification Approaches
~ Two fundamental approaches exist for
classification: a priori and a posteriori.
- a priori consists of developing logical rules
for classification based on observed
patterns in the characteristics of the objects
(e.g., classifying lakes on ecoregion, surface
area, and maximum depth)
-a posteriori develops groups from a
database of observations from the sites and
is restricted to those sites in the database
2.6
Florida Ecoregions
2.8
Distribution of EPT Taxa Metric
Among Florida Subregional Reference
Sites
2.10
7
-------
Ecoregion
Ecoregion
2.11
2.12
Distribution of EPT Taxa
in Aggregated Ecoregions
mtmi
20
16
J 12
£
LU 8
;H o j
4
§§1
I ~T~
EO:?6S4.j8K •
iitiii
:? T ¦ |
0
6575A
Panhandle
75BCD 75EF
Peninsula Northeast
2.13
Metric Selection
~ Preliminary selection (reference sites)
-low variability
-scope for detection of change
~ Testing responsiveness with impaired
sites
¦ Metrics that are too variable within the
reference sites are unlikely to be
effective for assessment.
•A measure of metric variability is the
ratio of the interquartile range to the
distance between the lower quartile and
the minimum possible value of the
metric .
3.1
3.2
8
-------
Assessing Candidate Metrics
A--
«• Metrics that have
-sr high values under
U '
reference
^ (unimpaired)
conditions.
-
~ Metrics that have low
=~ values under
Y--1
reference conditions.
3.3
Candidate Macroinvertebrate Metrics to
be Used for Site Classification and
Discrimination
Richness
Composition
Tolerance
Trophic
Measures
Measures
Measures
Measures
# of Total Taxa
Shannon-
Florida Index
% Collector-
Wiener Index
Gatherers
EPT Index
% Class 1 and
% Dominant
Class 2
% Collector-
#of
Taxon
Filterers
Chironomldae
Hilsenhoff
Taxa
% Diptera
Biotic Index
% Shredders
# of
% Crustacean
Crustacean +
+ Mollusc
Mollusc Taxa
3.4
Metric Evaluation and Index
Development
gJK-JK ¦:« : ¦
~ Selection of metrics and development of a
multimetric index requires a test data set
composed of reference sites and nonreference
(test) sites that might be impaired or that simply do
not meet the criteria for reference sites.
~ Ideally, the test sites should include at least some
sites that are severely impaired by different
stressors.
~ Reference condition characterization used only the
reference site data; metric evaluation and index
development use both reference and test site data.
3.5
Metric Response
mrnsmsmtm
¦ Response of metrics to stresses is evaluated
by comparison of reference sites to test sites
using box-and
¦whisker plots of the distribution.
I
I
a
I
T
IMRURED
TEST
SITES
UNKNOWN
TEST
SITES
3.6
ill
liilill
i-sil:
Tu»mk |
;:;X>
Reference Other Impaired Reference Other Impaired
6575A 75BCD
3.7
Discriminatory Power of the
Macroinvertebrate Metrics
3.8
Pjp N?t Collections
Strongest
Waakast
L nHMHnill ; I I
EPT Ms
Total Tna
Shannon Indn
Honda Index
X Htenars
• Ous/Mol
X Dominants
KDiptara
* CruvMoll
%Shr»ddars
# Chron. T«a
% Galhtrars
9
-------
Metric Response to Increasing
Biological Condition
Cala^ory
Mafric
Raaponta
Rldmatt Maaiurat
# of Total Taoca
+
EPT Indax
*
# of CNronomidaa Taxa
*
# ofCmttaoMn ~ MoUuse Taw
+
Competition Miiiufti
Sh«nnoo-Wian*r Max
% Dominant Taxon
% Diptara
% Cruatacaan ~ Mollusc
*
3.9
Metric Response to Increasing
Biological Condition
Catagory
Marie
Raapenaa
Tolaranca Maaaurat
Florida Indax
% Claai 1 and Claaa 2
HilaanhofT Biovc Indax
~
Trophic Mawuraa
% CoNactor-Galharara
% CoRactor-Ftttarara
%Shraddar«
~
3.10
Muttimetric Index Development
¦Index scoring alternatives
¦ Examples
4.1
sssssssssKssssKSBBsassssKi
Index Development
- Following classification and
characterization of the reference
condition, metrics are evaluated for
suitability in a multimetric index.
- Suitable metrics are those that respond
in a predictable way to stressors on the
system and that have low noise or
variability.
4.2
^ Scoring and Index Development
•» Combining unlike measurements is
possible only when the values have been
standardized by a transformation
through which measurements become
unitless.
*¦ Standardization of these measurements
into a logical progression of scores is
the typical means for comparing and
interpreting unlike metric values.
4.3
Basis of Bioassessment Scores
*¦ unimpaired »- population
reference sites distribution
MfTKC
VALUE
X
** *
xm 3
REFERENCE SCORMQ
OUSTRSUTION
4.4
METRIC
vaujc
REFERENCE SCORING
Distroution
10
-------
Scoring and Index Development
•-The scoring method should reflect how well the
reference sites represent unimpaired conditions.
If reference sites are unimpaired and considered
to be representative, bisection is recommended.
•¦This method assumes that the reference sites
are representative of relatively unimpaired
conditions and that the metric distribution
reflects natural variation of the metric.
~ A value above the cut off is then assumed to be
similar to reference conditions.
4.5
Scoring and Index Development
~ Choice of scoring method should be based on
confidence in the reference sites, rather than
on the method that will produce the most
conservative or most liberal scoring.
•¦ If confidence is high that reference sites are
representative of relatively unimpaired
conditions, then the lower percentile cutoff and
bisection are preferred.
•-If confidence is low, then bisection below the
95th percentile is preferred.
4.7
Assessment
»> Detection of impairment, as deviation
from reference condition
-Real sites
-Impaired scenarios
4.9
Scoring and Index Development
~The bisection method is best for scoring in regions
where impacts might be so pervasive that nearly all
reference sites are thought to be impacted or for
assessment of reservoirs where reference sites
cannot be defined.
* In bisection, it is assumed that at least some of the
lakes attain an excellent value for the metric, but that
many reference lakes are impaired and hence the
lower limit of the reference distribution is not known.
• The 95th percentile is thus taken as the 'best' value
and the range is trisected below it.
4.6
Scoring Criteria for the Core Metrics as
Determined by the 25th Percentile of the Metric
Values for the Two Aggregated Subecoregions
Metric
Panhandle
Peninsula
5
3
l
5
3
1
# of Taxa
*31
1630
0-15
*27
14-26
0-13
EFTTaxa
*7
4-6
03
*4
03
Crust + Mol Taxa
-
-
*4
03
XDomTmon
*20
>20
<37
>37
% Diptera
*38
>38
532
>32
% Cms + Moil
-
-
-
*16
0-15
Florida Index
£18
9-17
06
*7
4-6
03
%F3terers
*12
7-11
06
*8
4-7
03
% Shredders
*10
0-9
*13
0-12
Range of
Aggregated Score
7-29
933
4.8
Site Criteria
•-Reference Sites
-Stream watershed wholly within subecoregion.
-minimal human disturbance (land use,
streamside activities) relative to subecoregion.
•-Impaired Sites
-From point source monitoring program.
-Criteria:
~low DO (<2ppm); or
-toxic discharge; or
- evidence of nonpoint source pollution.
4.10
11
-------
Impairment Scenarios
-Scenario 1: Toxic substance similar to
pesticide. Extirpation of 50% of species of
hemimetabolous insects and crustaceans,
selected randomly (emphemeroptera,
plecoptera, odonata, hemiptera, crustacea).
-Scenario 2: Toxic substance similar to
metals. Extirpation of 50% of species that
are suspension feeders or deposit feeders.
Includes representatives of most insect
orders and non-insect phyla.
4.11
Discrimination of the Invertebrate
Stream Index for Florida
2bf
e>
?4
V
v
22
—
?(}
CO
o
18
16
o
CO
14
B
12
~—
10
4.12
Reference Other Impaired Reference Oltier Impaired
6575A 75BCD
Bioassessment Decision Criteria
Example"' Bibfogicai Concfrtiori
Cntena (a) Cateaorv
Attributes
>83% Non-impaired
Optimum communrty structure
(composition & dominance) for
stream size and habitat quality in
ecoregion.
54-79% Slightly impaired
Composition (species richness)
lower than expected due to loss
of some intolerant forms
Percent contribution of tolerant
forms increases.
21-50% Moderately impaired
Fewer species due to loss of
important forms. Reduction in
EPT index.
<17% Severely impaired
(a) percent ot refer once
Few species present If high
densities of organisms, then
dominated by one or two taxa.
4.13
Research Issues
~ Index performance
~ Spatial and seasonal variability
~ Optimal field methods
~ Annual variability
~ Optimal taxonomic level
5.1
Spatial Variability
~ within habitat
-(meters)
~ among habitats
-(meters to tens of meters)
~ among sampling sites within stream
-(100s to 1000s of meters)
~ among streams within class
-(kilometer and up)
5.2
Temporal Variability
~ Seasonal: requires selection of index
period when target organism populations
are least variable; i.e., before or after
major recruitment or spawning events.
~ Annual: requires annual monitoring of
subset sites to characterized annual
variability.
5.3
12
-------
Optimal Sampling
•¦Replication is expensive! Allocation of sampling
effort requires careful definition of the
questions being asked of the monitoring
program, and of the sampling unit.
»A sample observation seeks to characterize its
sampling unit, and should be done in a way to
minimize the variability of that characterization.
Typically, a composite sample form multiple
habitats or multiple net hauls is the most
cost-effective way to characterize the unit.
5.4
Repeated Observations
»• It is usually more cost-effective to
sample more units (sites) than to repeat
samples at a site, but there are
important exceptions:
~ Exceptions
-Compliance or attainment studies of index
sites
-Quality control replication to estimate
measurement error (typically 10%)
13
-------
BIOASSESSMENT USING PREDICTIVE MULTIVARIATE MODELS:
CLARITY, NOT SMOKE AND MIRRORS
Robert Bailey
Department of Zoology
The University of Western Ontario
London, Ontario, Canada N6A 5B7
NOTES:
The primary consideration in bioassessment should be that "good is variable."
Regardless of the temporal and spatial scale of the assessment, we cannot
arbitrarily decree that the organisms we find at one particular site are
representative of a healthy community. It is also unrealistic to test real
communities against a hypothetical community that only exists in a
benthologist's fantasy streamwalk. Thus, the first step in our approach is to
define criteria for Reference sites, and then sample them and describe the
variation among their biological communities. We also describe the habitat of
these same Reference sites, and quantify correlations between the habitat and
the community. Ultimately, we need to assess one or more 'Test" sites. We can
use correlations between the habitat and community of the Reference sites to
predict the structure of a Test site's community from its habitat. This answers
the question posed in most bioassessment studies... "Is the community at the Test
site close to what we would expect if it was one of the Reference sites?"
Comparing the predicted to the actual community allows us to assess human
impacts. If we see a community similar to that predicted, the site "passes." If
the community is quite different from the prediction, the site "fails."
None of this approach necessarily requires multivariate statistics, but it takes
more than one variable to adequately describe either a community or its
habitat. Once we decide which variables to use for such a description (e.g.,
family-level abundances or biomasses for communities; particle-size
distribution, flow for habitat), we let their variation and covariation describe
the structure. If s risky to count on arbitrary functions of the variables (indices)
to do the job, but they may be useful in helping to better describe the structure
after it has been revealed using multivariate analysis of the original
variables. So far, researchers using the predictive multivariate approach
have used some form of cluster analysis to describe the structure of Reference
site communities. They have then contrasted the habitats of groups of sites
with similar communities using Discriminant Functions Analysis (DFA).
Finally, they've taken the predictive equations from the DFA, along with
habitat data from a Test site, to predict which group the Test site's community
would be in if it was "healthy."
The approach has been used in streams across the United Kingdom (RIVPACS),
the North American Great Lakes (BEAST), and Yukon streams (no acronym
yet!). I'll look at some results from each of these studies to illustrate and
provoke discussion about the techniques.
14
-------
R
tn
eft
0>
S
C3
s
R
T
R
R
"universally
good number
of mayflies to
have"
R
Depth
-------
Yukon Strearns" ANDERSON CK„ (INDEX=1„34)
0.8
0.6
0.4
0.2
0.0
CT>
ANDa
FLDa
FLUa
RFNa
RFUa
UILa
• BLBa
-CRUa
¦PlUa
¦ CRDa
-PI Da
- RFDa
-DRDa
-MOMa
-MODa
- DRfla
-RUSa
-riUDa
-DRUa
- MCGa
-riOUa
-------
Yukon Streams: FISHER CK
-------
Australian Journal of Ecology (1995) 20, 198-219
Biological guidelines for freshwater sediment based on
BEnthic Assessment of SedimenT (the BEAST) using a
multivariate approach for predicting biological state
TREFOR B. REYNOLDSON,1 R. C. BAILEY,2 K. E. DAY1 AND
R. H. NORRIS3
1National Water Research Institute, Environment Canada CCIW, Burlington,
Ontario L7R 4A6, Canada, department of Zoology, University of Western
Ontario, London, Ontario, Canada and ^University of Canberra Water Research
Centre, Canberra, Australia
Abstract This paper describes the first results for an alternative approach to the development
of sediment quality criteria in the nearshore areas of the Laurentian Great Lakes. The approach
is derived from methods developed in the United Kingdom for establishing predictive
relationships between macroinvertebrate fauna and the physico-chemistry of riverine environ-
ments. The technique involves a multivariate statistical approach using (i) data on the structure
of benthic invertebrate communities, (ii) functional responses (survival, growth and repro-
duction) in four sediment toxicity tests (bioassays) with benthic invertebrates; and (iii) selected
environmental variables at 96 reference ('clean') sites in the nearshore areas of all five Great
Lakes (Lakes Superior, Huron, Erie, Ontario and Michigan). Two pattern recognition tech-
niques (using the computer software package patn) are employed in the analysis: cluster
analysis and ordination. The ordination vector scores from the original axes of the pattern
analysis are correlated (using CORR in SAS) with environmental variables which are anticipated
to be least affected by anthropogenic activities (e.g. alkalinity, depth, silt, sodium etc.). Multiple
discriminant analysis (MD A) is used to relate the site groupings from the pattern analysis to the
environmental variables and to generate a model that can be used to predict community
assemblages and functional responses at new sites with unknown but potential contamination.
The predicted community assemblages and functional responses are then compared with the
actual benthic communities and responses at a site, and the need for remedial action is
determined.
The predictive capability of the discriminant model was confirmed by performing several
validation runs on subsets of the data. An example of the use of the model for sediment in
Collingwood Bay (an area of concern designated by the IJC in Georgian Bay, Lake Huron) is
presented and the technique is shown to be more precise in determining the need for
remediation than the currently used provincial sediment quality criteria based on Screening
Level Concentration (SLC) and laboratory toxicity tests. The ultimate goal of the study is the
development of a method to determine the need for, and the success of, remedial action and to
predict what benthic communities should look like at a site if it were clean and what responses
of organisms in sediment toxicity tests constitute an acceptable end-point.
Key words: benthic, communities, criteria, invertebrates, multivariate analysis, sediment,
water quality.
INTRODUCTION
Environmental managers and regulatory decision makers
have traditionally set water and sediment quality guide-
lines based on chemical concentrations. The primary
Accepted for publication October 1994.
advantage of a chemical approach is the apparent ease of
simple numerical comparison of concentrations of chem-
icals found in environmental matrices with levels of these
same compounds known to cause a toxic response in
biota. However, the chemical approach has been criticized
in recent years because it frequently fails to achieve its
objectives (Cairns & van der Schalie 1980; Long &
-------
PREDICTING STATE OF FRESHWATER SEDIMENT 199
Chapman 1985; Chapman 1986; Chapman 1990) or
because it is so excessively rigorous that it has limited
value (Painter 1992; Zarull & Reynoldson 1993).
The purpose of environmental assessment and manage-
ment is, ultimately, the maintenance of biological
integrity; thus, we suggest that the setting of water and
sediment quality objectives should involve the use of
biological criteria rather than chemical surrogates. Until
recendy, the development of numeric biological objectives
was considered too difficult due to the temporal and
spatial variability inherent in biological systems. However,
over the past 10 years, methods developed in the United
Kingdom (Wright et al. 1984; Moss et al. 1987; Armitage
et al. 1987; Ormerod & Edwards 1987) and elsewhere
(Corkum & Currie 1987; Johnson & Wiederholm 1989)
have demonstrated the ability to predict the community
structure of benthic invertebrates in clean (or 'uncon-
taminated') sites using simple habitat and water quality
descriptors. This approach allows appropriate site-specific
biological objectives to be set for ecosystems from
measured habitat characteristics and also provides an
appropriate reference for determining when degradation
at a site due to anthropogenic contamination is occurring.
The acceptance by regulatory agencies of biological water
and sediment quality objectives has been slow but is now
being given serious consideration as shown by current
work in Canada (Reynoldson & Zarull 1993), the USA
(Hunsaker & Carpenter 1990) and the United Kingdom
(the rivpacs method; Wright et al. 1984) and recent
initiatives in Australia (R. H. Norris, pers. comm.).
This paper describes the development of biological
objectives for sediments in nearshore habitats in the
North American Great Lakes using a modification of the
technique developed in the UK (Wright et al. 1984;
Furse et al. 1984; Armitage et al. 1987). A large data base
is being assembled from reference sites in Lakes Ontario,
Erie, Michigan, Superior and Huron and includes infor-
mation on: (i) the structure of the benthic invertebrate
communities; (ii) measured environmental variables and
(iii) the responses of four species of benthic invertebrates
(Hyalella azteca, Chironomus riparius, Hexagenia spp.
and Tubifex tubifex) exposed in the laboratory to sediment
collected from the same sites. Benthic invertebrates were
selected as the most appropriate biological indicators
because they are most directly associated with contam-
inants in sediments through their feeding and behavioral
activities. Laboratory sediment testing was included in
addition to estimates of benthic invertebrate community
structure, to identify the sediment rather than the water
column or other physical disturbances as the cause of the
observed effect at any given site. These data are being
used to develop numeric biological sediment objectives
for the Great Lakes and the results presented in this
paper incorporate data collected from 96 out of a potential
250 reference sites sampled during 1991-93.
METHODS
Reference sites
The study area encompasses the entire basin of the Great
Lakes. To ensure the range of habitat characteristics
were adequately represented, a preliminary list of 250
sites were identified and stratified among 17 ecoregions
described by Wickware and Rubic (1989) for the Great
Lakes. These ecoregions are defined from characteristics
such as climate, vegetation, bedrock geology, flora and
fauna et cetera. The reference sites were selected to
represent 'unpolluted' conditions within an ecodistrict
and the inclusion of each site required it to meet the
following criteria: the site be located well away (>10
km) from known discharges as described in the Ontario
Intake and Outfall Atlas (Ontario Ministry of Environ-
ment 1990); the site be located within 2 km of the shore
and at a depth of less than 30 m (with the exception of
Lake Michigan) and; the site be known or suspected to
have a fine-grained substrate. The sites were sampled
over a 3 year period (1991-93). This paper presents
preliminary results from the first 2 years of the study (i.e.
50 sites from 1991 and 46 sites from 1992; Fig. 1).
General
The location of each site was established in the field
using either Loran C or a hand-held Geographical
Positioning System (GPS). At each reference site, samples
were taken of sediment, water and pore-water for chemical
and physical analysis; in addition, samples were collected
for the determination of the community structure of
benthic macroinvertebrates and for laboratory sediment
bioassays with selected species of benthic invertebrates.
Each site was sampled once in late summer or early fall
over a 3 year period. In addition, a sub-set of sites (10%)
were sampled in each of the three field years, and four
sites have been sampled monthly over 2 years. This will
allow a subsequent determination of the effects of both
annual and seasonal variation on the outcome of the
predictions of community structure and toxicity. This
will be the subject of later publications.
Environmental variables
A list of the variables measured at a site is presented in
Table 1. Samples for water chemistry were taken using a
Van Dorn sampler from 0.5 m above the sediment-water
interface. A 1L sample was stored at 4°C prior to
analysis of total phosphorus, Kjeldahl nitrogen, nitrate-
nitrite and alkalinity at the National Water Research
Laboratory in Burlington, Ontario, Canada. Measure-
ments of pH, dissolved oxygen and temperature were
made in the field.
-------
200 T. B. REYNOLDSON ET AL.
Sediment and sediment pore-water were characterized
from samples taken from a mini-box core. The mini-box
core takes a 40 x 40 cm section of sediment to a depth of
25-30 cm. Samples for geochemical analysis were taken
from the surface 2 cm of the box core. After sampling the
sediment was homogenized in a glass dish with a nalgene
spoon. The sample was divided as follows.
(1) An aliquot of sediment for organic contaminants
-—r
ol
100 Km
N
Fig. 1. Location of reference sites in the Great Lakes. (•) Sites included tif analysis; (O) sites sampled since analysis.
TsMe 1. Summary of measured environmental variables and abbreviations used
Field
Water (mg/L)
Sediment (ug/g dry wt)
(6 variables)
(5 variables)
(32 variables)
Latitude (LAT)
Alkalinity (AKW)
Silica (SI)
Cobalt
Longitude (LON)
T. phosphorus (TPW)
Titanium
Nickel
Water depth (m) (DP)
Kjeldahl Nitrogen (TKN)
Aluminium (Al)
Copper
Oxygen (mg/L) (OXW)
Nitrate-nitrite (NOW)
Iron (Fe)
Zinc
Bottom temperature (°C) (TMW)
Ammonia
Manganese (Mn)
Arsenic
pH (PHW)
Magn—lum (Mg)
Strontium
Calcium (Ca)
Yttrium
Sodium (Na)
Molybdenum
Potassium (K)..
Silver
T. Nitrogen (TN)
Cadmium
T. Phosphorus (TP)
Tin
T. Org. Carbon (TOC)
Lead
Loss on ignition (LOI)
% Gravel (GR)
Selenium
% Sand (SN)
Vanadium (V)
% Silt (SL)
Chromium
% Clay (CL)
Variables in bold were considered for use as predictors in MDA.
-------
PREDICTING STATE OF FRESHWATER SEDIMENT
201
was placed into a hexane-prewashed glass bottle with a
hexane rinsed aluminium foil liner. Samples were sealed
and stored frozen (or at 4°C in the field) for subsequent
freeze-drying and storage. These samples were not
normally analysed but were archived in the event of a site
being suspected as contaminated.
(2) Samples for the determination of particle size
distribution were placed into a plastic pill jar and stored
at ambient temperature in the field. Upon return to the
laboratory, samples were freeze-dried and analysed
following the method of Duncan and LaHaie (1979).
(3) The remaining sediment in the glass dish was
stored in a 500 mL plastic container at 4°C in the field
and shipped to the National Water Research Laboratory
for freeze drying and analyses for metals, major ions and
nutrients.
Invertebrate community structure
Samples for the identification and enumeration of benthic
invertebrates were taken by inserting five 10 cm plexiglass
tubes (internal diameter 6.6 cm) into the sediment in the
box core. Each core tube was considered to be a replicate
sample unit and was removed and the contents placed
into a plastic bag and kept cool until sieved. The contents
of each bag were sieved through a 250 mesh in the
field as quickly as possible after sampling. If sieving
could not be done in the field, 4% formalin was added to
the bag and the replicate samples were stored at 4°C and
sieved as-soon as possible thereafter. After sieving the
samples were placed in plastic vials (50 mL) and preserved
with 4% formalin. Replicates with large amounts of
organic material were placed in larger containers and
again preserved with 4% formalin. After 24 h the formalin
was replaced by ethanol.
Samples were sorted with a low power stereo micro-
scope and identified to species or genus level where
possible. As required (Chironomidae and Oligochaeta)
slide mounts were made for high power microscopic
identification. Appropriate identification guides were used
and voucher specimens of all identified specimens were
submitted to experts for confirmation. The confirmed
voucher specimens are being maintained as a reference
collection.
Sediment toxicity
A mini-ponar sampler was used to obtain five replicate
field samples of sediment for laboratory bioassays with
four species of invertebrates. Each replicate sample was
placed in a plastic bag and held at 4°C until tests could
be conducted.
Tests were conducted, in sets of six to seven, over a
period of approximately 6 months. A clean control sedi-
ment from the Canadian Wildlife Bird Sanctuary, Long
Point, Lake Erie was also tested with each set of samples
to provide biological quality assurance. Complete details
of the culture of organisms and conditions for each
toxicity test with C. riparius and T. tubifex are described
elsewhere (Reynoldson et al. 1991; Day et al. 1994;
Reynoldson et al. 1994). Culture of H. azteca was
conducted according to the procedure described in
Borgmann et al. (1989). Eggs of the mayfly, Hexagenia
spp. (both H. limbata and H. rigida), were collected
during late June and July in 1991 according to the
method of Hanes and Ciborowski (1992) and organisms
were cultured using the procedure of Bedard et al.
(1992). Tests with H. azteca, C. riparius and T. tubifex
were conducted in 250 mL glass beakers containing
60-100 mL of sieved (500 /urn mesh), homogenized
sediment with approximately 100-140 mL of overlying
carbon-filtered, dechlorinated and aerated Lake Ontario
water (pH 7.8-8.3, conductivity 439-578 /nohms/cm,
hardness 119 to 137 mg/L). Tests with the mayfly,
Hexagenia, were conducted in 1 L glass jars with 150 mL
of test sediment and 850 mL overlying water. The
sediment was allowed to settle for 24 h prior to addition
of the test organisms. Tests were initiated with the
random addition of 15 organisms per beaker for H. azteca
and C. riparius, 10 organisms per jar for Hexagenia spp.
and four organisms per beaker for T. tubifex. Juveniles of
H. azteca were 3 to 7 days old at test initiation; C. riparius
larvae were first instars and were approximately 3 days
post-oviposition; Hexagenia nymphs were 1.5 to 2 months
old (approximately 5 to 10 mg wet weight) and T. tubifex
adults were 8 to 9 weeks old. Tests were conducted at
23±1°C with a 16L:8D photoperiod (T. tubifex 24 h
dark). Tests were static with the periodic addition of
distilled water to replace water lost due to evaporation.
Each beaker was covered with a plastic petri dish with a
central hole for aeration using a Pasteur pipette and air
line. Dissolved oxygen concentrations and pH were
measured at the beginning, middle and end of each
exposure period. Tests were terminated after 10 days for
C. riparius, 21 days for Hexagenia and 28 days for
H. azteca and T. tubifex by passing the sediment samples
through a 500 um mesh sieve. Sediment from the
T. tubifex test was passed through an additional 250 fim
mesh sieve at test completion. End-points measured in
the tests were survival and growth for C. riparius,
Hexagenia spp. and H. azteca and for T. tubifex survival
and production of cocoons and young. Mean dry weights
of H. azteca, C. riparius and Hexagenia spp. were deter-
mined after drying the surviving animals from each
treatment replicate as a group to a constant weight in a
drying oven (60°C).
Data analysis
The analytical strategy used is similar to that proposed
by Wright et al. (1984). Pattern) (analysis was used to
describe the biological structure of the data at the
-------
202 T. B. REYNOLDSON ET AL.
reference sites and correlation and multiple discriminant
analysis (MD A) to relate the observed biological structure
to the environmental characteristics.
Classification of biological data
The biological structure of the data was examined using
two pattern recognition techniques, cluster analysis and
ordination. The mean values from the five replicates for
the species counts were used as descriptors of ttie benthic
invertebrate community. These community data were
not transformed and the raw scores were used as we
considered numeric differences to be important com-
munity descriptors. The Bray and Curtis association
measure was used because it performs consistently well
in a variety of tests and simulations on different types of
data (Faith et al. 1987). Clustering of the reference sites
was done using an agglomerative hierarchical fusion
method with unweighted pair group mean averages
(UPGMA). The appropriate number of groups was
selected by examining the group structure and, par-
ticularly, the spatial location of the groups in ordination
space. Ordination was used to reduce the variables
required to identify the structure of the data. A multi-
dimensional scaling (MDS) method of ordination was
used (i.e. Semi-Strong-Hybrid multidimensional scaling;
Belbin 1991). Multi-dimensional scaling methods use
metric and non-metric rank order rather than metric
information and thus provide a robust relationship with
ecological distance and do not assume a linear relation-
ship, which is an inherent assumption in some dissimi-
larity measures used by other ordination techniques
(Faith et al. 1987). This is of particular value when
relating ordination scores to environmental characteristics.
All clustering and ordination was done using patn, a
pattern analysis software package developed by CSIRO
in Australia (Belbin 1993).
The relationship with the biological data was examined
by correlation analysis of environmental characteristics
with ordination axis vector scores using the procedure
CORR in sas.
Prediction of biological groupings
Based on the results from correlation analysis, a suite of
environmental variables were selected for use in multiple
discriminant analysis (MDA) to relate the biological site
groupings to the environmental characteristics of the
sites. The sas version of MDA was used with raw
environmental data to generate discriminant scores, and
to predict the probability of group membership.
Validation of discriminant model
To test the predictive capability of the discriminant
model, five validation runs were performed. In each
validation test, ten sites were randomly removed from
the reference data set. The discriminant model was
calculated for the remaining sites and tested on the ten
removed sites. The predicted grouping and its probability
were compared with the actual group identified from the
initial classification analysis.
Testing of model: Collingwood Harbour case study
As a demonstration of the practical application of this
approach, it was used to determine the need for remedial
dredging in Collingwood Harbour, Georgian Bay, Lake
Huron. This harbour has been identified as an 'Area of
Concern' (International Joint Commission 1987) in part
because of sediments defined as contaminated using
-chemical guidelines (Persaud et al. 1992). Twenty-five
sites were sampled and compared with the reference
sites.
Correlation of biological data with environmental
characteristics
Of the 43 environmental variables measured in this
study, 25 were examined for their relationship with the
biological structure of the data (Table 1). We excluded
those variables most likely to be influenced by anthro-
pogenic activity, particularly those associated with sedi-
ment contamination. Thus, all the metals were excluded
from consideration as potential predictor variables. The
variables used were general descriptors of sediment type
such as the major elements, particle size and oiganic
material as a potential indicator of nutritive quality.
These together with physical attributes such as water
depth and general water chemistry were considered to be
the most appropriate general habitat descriptors that are
not as subject to modification from human activity.
RESULTS
Classification of sites
At 93 reference sites, 103 species and 44 genera were
identified. Certain groups were not identified below
higher taxonomic levels, such as the Porifera, Platy-
helminthes and Empididae. The most diverse taxonomic
groups were the Chironomidae (43 genera), the Oligo-
chaeta (37 species) and the Mollusca (36 species). Because
of the large number of tajft (1%) that were available foT
classification analysis we reduced the data set by includ-
ing only those taxa with an abundance equal or greater to
0.05% of the total number of organisms (excluding the
Porifera). This is because large numbers of rare species
tend to add now *>»-—*—¦' " aion analysis.
-------
PREDICTING STATE OF FRESHWATER SEDIMENT 203
(a)
0.0176
0104
0105
010C
010?
0301
5908
oioa
5000
0111
0112
0114
Oils
0113
0300
5911
0312
0313
0302
0303 -
0109
0305
1302
0307
0300
0«01
1«01
0«02
1«03
1212
1303
1211
1213
1 602
1403
2201
2517
2510
1406
1407
2200
1600
2516
580 3
5902
5903
1 405
2510
2511
5602
0110
0403
0404
0306
0405
0304
0310
0311
5501
5502
5504
saos
5904
5806
5901
5906
5909
5600
5806
seio
5801
5802
5807
5503
5710
5507
5508
5509
5703
5804
5611
5705
5706
5505
5701
sao»
5702
5506
5708
5511
5707
5510
5711
5900
(b)
0104
0302
0105
0312
0106
1302
0303
1407
2510
1603
0107
0405
0109
0404
0110
5802
0305
0403
5504
0108
0112
0115
5508
5701
5507
5509
0111
2511
5506
5702
5710
0300
0301
5610
5703
0307
0308
2201
0313
1213
1303
1406
2200
1211
1600
2518
1212
1403
1601
1405
2516
2517
1602
0113
0304
5510
5511
5502
5505
5906
5911
5907
5901
5909
0114
0602
5807
0601
5810
0306
0311
5808
5503
5903
5706
5711
5811
5501
5809
5904
5602
5705
5900
0310
5800
5801
5803
5908
5707
5806
5902
2411
5708
5600
5804
5805
Fig. 2. Dendrogram of reference sites based on (a) invertebrate community structure and (b) toxicity.
Table 2. Geographic distribution of sites in five groups from classification of benthic invertebrate community structure
Gpl
(19 sites)
Gp2
(14 sites)
Gp3
(16 sites)
Gp4
(8 sites)
Gp 5
(36 sites)
Lake Ontario
Lake Erie
Lake Huron
Georgian Bay
North Channel
Lake Superior
Lake Michigan
16
36
-------
204 T. B. REYNOLDSON £7" AL.
This reduced the number of taxa used in the analysis to
55 (Appendix I).
The results of the cluster analysis for the 93 sites are
shown in Fig. 2a. The first group of sites to be dis-
tinguished are 36 sites, identified as Gp 5, from Lake
Michigan. The remaining sites fall into four groups,
beyond which the structure breaks down and small
groups of sites begin to form. The geographical dis-
tribution (Table 2) of the sites in these five groups
suggests that there is a strong spatial signal in the
observed grouping. The sites forming Gp 1 are pre-
dominantly from mesotrophic Lake Erie together with
three shallow sites from Lake Michigan. Eight of the
nine Lake Erie sites not included in Gp 1, are from Long
Table 3. List of species that occur in at least 50% of sites in a group, ordered by Helming frequency of occurrence
Gp 1
Gp2
Gp 3
Gp4
Gp 5
Tubificidae (co h)
Tubificidae (c h)
Procladius spp.
Pisidium spp.
Pisidium casertanum
Porifera
Spirosperma ferox*
Dreissema polymorpha*
Platyhelminthes
Limnodrilus hcffmeisteri*
Chironomus spp.
Auiodrilus pigueti
Porifera
Tubifidae (co h)
Procladius spp.
Cryptochiroaomus spp.
Chironomus spp.
Pisidium casertanum
Tubificidae (c bar)
Tanvtarsus spp.*
Valvata tricarinata*
Auiodrilus pigueti
Platyhelminthes
Pisidium spp.
Rolypedium spp.
Tubificidae (co h)
Tubificidae (c h)
Procladius spp.
Pisidium casertanum
Diporeia hoyi
Micropsectra spp.*
Porifera
Procladius spp.
Chironomus spp.
Dicrotendipes spp.*
Microtendipes spp.*
Cryptochironomus spp.
Tubificidae (co h)
Physella spp*
Polypedium spp.
Pisidtum casertanum
Pisidium nitidum*
Endochironomus spp*
Pseudochironomus spp.*
Pisidium spp*
Auiodrilus pigueti
Ammcola limosa*
Diporeia hoyi
Stylodrilus heringianus*
Pisidium spp.
Vejdovskella intermedia*
Platyhelminthes
Heterotrissoclodius spp.*
Pisidium casertanum
Tubificidae (co h)
Tubificidae (c h)
Species in bold have >70% occurrence. Taxa marked with * are unique to that group of sites.
1.5
1
0.5
CN
O 0
"5
-0.5
-1
-1.5
~
' *\
-1.5 -1 -0.5 0 0.5 1 1.5
Vector 1
Fig. 3. Ordination of reference sites based on invertebrate community structure KM r.ni./mr „
• vv; up j, up (~} Gp 3; (!) Gp 4; (i
-1.5 -1 -0.5 0 0.5 1 1.5
Vector 1
)Gd5.
-------
PREDICTING STATE OF FRESHWATER SEDIMENT
205
Point Bay and are shallow inner bay sites classified as Gp
4. Group 4 also includes three shallow sites from
Presque'Ile Bay in Lake Ontario. The Gp 3 sites are
mostly from the oligotrophic North Channel of Lake
Huron and Lake Superior. Group 2 includes southern
Georgian Bay sites together with a few Lake Erie and
Lake Ontario sites.
The more common species present in each of the five
groups, those found at 70% and 50% of the sites in a
group, are shown in Table 3. In Gps 1 and 3, the
Tubificidae are the most frequently occurring taxa and in
Gp 1, which represents the more mesotrophic sites, the
oligochaete species Spirosperma ferox, Limnodrilus
hoffmeisteri and Aulodrilus pigueti are commonly found,
as are Chironomus spp. These species are all typically
associated with greater amounts of organic material. In
contrast, Gp 3 and, particularly, Gp 5 have species that
are characteristic of oligotrophic conditions; for example,
the amphipod Diporeia hoyi and the lumbriculid worm
Stylodrilus heringianus are important components of the
benthic assemblage at the sites which make up these two
groups. The shallow sites (Gp 4) are characterized by a
community dominated by the Porifera (sponges) and a
large number of chironomid species. Some taxa are
common to all the groups, notably the Tubificidae and
Pisidium casertanum, and the presence of these organisms
in a group is not a useful indicator; however, their
abundance does vary between groups. Other taxa are
restricted to single groups: Spirosperma ferox, Limnodrilus
hoffmeisteri and Dreissenia polymorpha are common in
Gp 1 only; Tanytarsus spp. and Valvata tricarinata in Gp
2; Microspectra spp. in Gp 3; several chironomid species
and molluscs in Gp 4; and Stylodrilus heringianus and
Heterotrissocladius spp. in Gp 5.
The location of the reference sites in ordination space
is shown in Fig. 3. With the exception of the sites
forming Gp 3, there is good discrimination on the first
two ordination vectors. The Gp 3 sites are separated on
the third vector. The relative contribution of the species
to the ordination vectors has been determined by principal
axis correlation (Table 4). Those taxa contributing most
to the pattern of site distribution are indicated by arrows
in Fig. 3; the direction of the arrow indicates the direction
of the loading and the length of the arrow the importance
of the taxa. Sites with low scores on vector 1 and located
on the left of the plot (Gp 5) are dominated by the
presence of D. hoyi and 5. heringianus. Those sites
scoring low on vector two have communities dominated
Table 4. Correlation of taxa with vectors from SSH ordination
Taxa r
S. heringianus 0.7465
Tubificidae A 0.7447
D. hoyi 0.7446
Porifera 0.6718
Cryptockironomus spp. 0.5082
Micropsectra spp. 0.4633
Procladius spp. 0.4626
Tubificidae B 0.4378
Gammarus pseudolimnaeus 0.4266
5. lacustris 0.4229
Dicrotendipes spp. 0.4157
Dreissenia polymorpha 0.4146
V. intermedia 0.4126
A. pigueti 0.4071
Microtendipes spp. 0.4058
M. speciosa 0.4009
Psectrocladius spp. 0.3912
Physella sp. 0.3840
Polypedium spp. 0.3821
Endochironomus spp. 0.3816
5. ferox 03784
Platyhelminthes 0.3680
Coelotanyopus spp. 0.3485
A. limosa 0.3359
P. nitidum 0.3348
Enchytraediae 0.3312
C. intermedius 0.3279
V. tricarinata 0.3271
Taxa
Heterotrissocladius spp.
0.3203
N. variabilis
0.3076
P. moldaviensis
0.2956
Chironomus spp.
0.2941
A. pluriseta
0.2838
P. vejodoskyi
0.2747
V. piscinalis
0.2747
L. hoffmeisteri
0.2727
Glypotendipes spp.
0.2711
A. lomondi
0.2666
Q. multisetosus
0.2663
Pisidium spp.
0.2661
Cryptotendipes spp.
0.2604
P. casertanum
0.2592
S. josmae
0.2446
Cladopelma spp.
0.2401
Chaoborus spp.
0.2309
P. henslotoanum
0.2193
Stictochironomus spp.
0.2147
T. tubifex
0.2022
Tanypus spp.
0.1913
Demicryptockironomus spp.
0.1837
H. americana
0.1572
M. securis
0.1347
P. compressum
0.1144
Caecidotea spp.
0.0537
-------
206 T. B. REYNOLDSON ET AL.
by sponges and Procladius spp. While tubificid worms
are found at many sites they are most abundant at those
sites scoring high on vectors 1 and 3 (Gps 1 and 2).
The same approach was used to define the pattern in
the physiological responses of the test organisms from
the chronic sediment bioassays. The similarity scores in
the dendrogram (Fig. 2b) show that there is less structure
in these data compared to that found in the community
structure. Examination of the physiological data suggests
that a three-cluster solution is most appropriate.'Beyond
this, it was not possible to relate the biological groupings
to the environmental data. In fact, even at this level,
there is little difference between the groups for several of
the test end-points (Table 5); for example, survival of
C. riparius and Hexagenia spp., growth of C. riparius and
cocoon production of T. tubifex. The end-points which
appear sensitive to site differences are survival of
H. azteca, which was very low in the five sites comprising
Gp 3, and growth of both H. azteca and Hexagenia spp.,
which is reduced at the Gp 3 sites. Conversely, pro-
duction of young by T. tubifex is highest at these same
sites and reduced in the Gp 1 sites.
Table 5. Toxicity test end-points in the three reference site groups defined by classification analysis
End-point
Overall
Gp 1 («
= 53)
Gp 2 (ft
= 38)
Gp 3(«
= 5)
C. riparius
% Survival
Growth (mg dry wt)
82.9
0.34
(8.6)
(0.08)
80.2
0.33
"(7.8)
(0.08)
86.6
0.37
(8.6)
(0.08)
82.9
0.30
(6.2)
(0.04)
H. azteca
% Survival
Growth (mg dry wt)
83.7
0.50
(19.0)
(0.14)
88.7
0.51
(11.8)
(0.11)
84.7
0.50
(12.9)
(0.17)
24.5
0.29
(22.4)
(0.11)
Hexagenia
% Survival
Growth (mg dry wt)
96.9
3.5
(4.1)
(3.6)
97.4
3.07
(3.4)
(2.56)
96.4
4.43
(4.8)
(4.64)
95.5
1.49
(4.6)
(0.31)
T. tubifex
Coccoons
Young
34.9
87.7
(5.8)
(40.0)
32.8
58.5
(6.1)
(23.5)
37.6
124.0
(4.2)
(23.0)
35.8
122.2
(2.8)
(24.9)
Values are means with SD in parentheses.
1.5
1
cn 0.5
o
I' 0
-o4
-1
-1.5
-3
o
aD> ^OO^O
- <& A
u ' - <<*>
~~ y®0 o
~ c®
~ § °
~ 0
-2 -1 0 1
Vector 1
-3 -2-10 1
Vector 1
Fig. 4. Ordination of reference sites based on sediment toxicity. (0) Gp 1; (O) Gp 2; (~) Gp 3.
-------
PREDICTING STATE OF FRESHWATER SEDIMENT 207
Tabic 6. Correlation coefficients between selected environmental variables and ordination vectors from community structure
Vector 1
Vector 2
Vector 3
TMW
0.6456
SI
0.3122
AKW
0.6990
TKN
0.5760
V
0.2905
DP
0.4669
SL
0.5541
K
0.2757
PHW
0.4656
NA
0.5035
NOW
0.2427
NOW
0.4188
AL
0.4903
AL
0.2322
MG
0.2597
TOC
0.3642
NA
0.1807
K
0.1984
TN
0.3617
TMW
0.1761
OXW
0.1624
CA
0.3557
FE
0.1527
CA
0.1237
LOI
0.3265
GR
0.1222
SN
0.0964
FE
0.1747
TPW
0.1145
MN
0.0729
V
0.1528
CY
0.0964
LOI
0.0683
CY
0.1358
SN
0.021
SI
0.0086
TPW
0.0763
MN
0.0143
CY
0.0014
K
0.0736
OXW
0.0128
GR
-0.0134
GR
0.0649
MG
0.0017
TP
-0.0208
PHW
0.062
TPO
-0.0002
SL
-0.1376
MN
-0.055
DP
-0.1094
FE
-0.1622
TP
-0.103
SL
-0.1114
TPW
-0.2192
MG
-0.16
TKN
-0.1881
TKN
-0.2781
oxw
- 0.377
AKW
-0.2331
TOC
-0.2959
SI
-0.464
PHW
— 0.2922
TN
-0.3149
SN
-0.465
TOC
-0.3088
AL
-0.3271
AKW
-0.476
TN
-0.3169
V
-0.4156
NOW
-0.711
LOI
-0.4446
NA
- 0.5347
DP
-0.8073
CA
-0.4815
TMW
-0.6761
Each vector ranked from high positive to high negative; variables in bold P< 0.001.
Table 7. Variables correlated (P<0.01) with three ordination
vectors from toxicity end-points in rank order
Vector 1 Vector 2 Vector 3
Alkalinity (w)
Alkalinity (w) Vanadium
Sodium
Aluminium
Aluminium
Sodium
Depth
Temperature (w)
Temperature (w)
Depth
Vanadium
Sand
pH (w)
Oxygen (w)
Iron
Vanadium
Nitrate (w)
Kjeldahl Nitrogen (w)
Silt
Iron
Silica
Clay
(w) Indicates those variables measured in water.
The ordination of the toxicity data matrix shows a
strong axis on the first vector on which the three groups
of sites can be separated (Fig. 4a). Sites in the lower left
hand corner on vectors 1 and 2 have low Hyalella
survivorship and high Tubifex reproduction and at the
top right of the same plot Tubifex reproduction is lower.
Correlation with environmental characteristics
To establish the relationship between community struc-
ture, toxicity and the environmental characteristics of
the sites, correlation coefficients and probabilities were
calculated for the 25 variables considered to be the most
useful in developing a predictive model (Table 1) and the
vector scores from ordination.
The results of the correlation with the first community
structure vector (Fig. 3) showed a negative correlation
with depth (r = - 0.80733; Table 6) and nitrate-nitrite,
and a positive correlation with temperature, Kjeldahl
nitrogen and silt. None of the measured variables was
well correlated with the second community vector (Table
6). On the third community vector, which discriminates
the Gp 5 sites, water temperature (and depth), sodium
and alkalinity were important. In the ordination plot
(Fig. 3a), sites toward the left on vector 1 tend to
represent deeper, more oligotrophic conditions and sites
toward the bottom and right of the plot represent
shallower, warmer and more mesotrophic conditions.
The variables best correlated with the toxicity ordin-
ation vectors were the major ions. Surprisingly, particle
size and the organic carbon content were not correlated
with the toxicity ordination vectors (Table 7). Water
chemistry descriptors, particularly alkalinity, were well
-------
208 T. B. REYNOLDSON ET AL.
correlated with the bioassay end-points, although the
test water is dechlorinated Lake Ontario water. The five
sites forming Gp 3 were distinguished by being deeper
(mean depth 67 m cf. 20.1 and 38.0 m) with a very low
silt content (mean 8.7%) compared with the other two
site groups, which have mean silt contents of 24.2% and
33.2%.
Prediction and validation of biological groups
Of the 25 variables examined, only those strongly related
(P< 0.0001) to the ordination vectors from the com-
munity structure were considered for use in discriminant
analysis for predicting group membership derived from
the species data. Water temperature (TMW) was also
Table 8. Environmental variables used to predict community and
toxicity group membership
Community predictors
Depth
Nitrate (w)
Silt
Aluminium
Calcium
LOI
Alkalinity (w)
Sodium
pH (w)
Bioassay predictors
Silica
Aluminium
Iron
Manganese
Calcium
Sodium
Potassium
Phosphate
LOI
Total Nitrogen
Total Organic Carbon
Sand
Silt
Clay
Vanadium
Depth
Alkalinity (w)
Feeding regime
(w) Indicates those variables measured in water.
• discarded as it is dependent on the time of sampling and
was highly correlated with depth (DP).
The use of nine variables (Table 8) in the discriminant
model produced the best prediction of group membership
from community structure data (Table 9). Results showed
that 79 of 91 (86.8%) sites were correctly predicted by the
nine variables. The discriminant model had the most
difficulty in classifying the Gp 2 sites (64.3% correct, 9 of
14) with four of the five incorrectly classified sites being
predicted as Gp 1. Group 1 is adjacent to Gp 2 in
ordination space and there is considerable overlap in the
two groups (Fig. 3). Examination of the discriminant
functions (Fig. 5) shows a clear separation of the groups
based on community structure. The first discriminant
function explained 73.6% of the variance with the greatest
contributions being made by nitrate, depth and calcium,
respectively. On the second function, which explains a
further 14.8% of the variance, loss on ignition (%LOI)
and alkalinity were major contributors.
The mean values for the nine environmental variables,
for each of the groups, are shown in Table 10. The sites
forming Gp 4 are shallow and have the greatest silt
content and per cent loss on ignition (LOI). These sites
are dominated by the presence of sponges (Porifera)
which are found at densities of over 1900 per 100 cm2.
These shallow sites also have the greatest number of
frequently occurring organisms (Table 4), particularly
chiionomids. The deep Gp 5 sites are characterized by
the presence of the amphipod, Diporeia hoyi, and the
lumbriculid oligochaete, Siylodrilus heringianus, both of
which are typical of more oligotrophic conditions in the
Great Lakes.
The same approach was used in classifying the site
groups based on toxicity from the environmental data.
The measured environmental variables (Table 9) were
less able to discriminate between the groups derived
from the toxicity test end-points. Only 70.2% of the sites
were correctly identified using 18 variables compared
Table 9. The number and percentage of 91 sites predicted to the correct biological group using discriminant analysis with
vdriflhlcs
environmental variables
Group member
Community structure
1
2
3
4
5
Predicted Gp I Predicted Gp 2 Predicted Gp 3 Predicted Gp 4 Predicted Gp 5
19 (100%)
4
3
0
1
9 (64.3%)
1
0
U (73.3%)
0
(75%)
0
0
0
0
34 (97.1%)
Toxicity
1
2
3
43 (82.7%)
6 (16.2%)
0
5 (9.6%)
25 (67.6%)
0
4 (7.7%)
6 (16.2%)
3 (100%)
-------
PREDICTING STATE OF FRESHWATER SEDIMENT 209
Table 10. Mean (SD) for nine environmental variables from five groups of sites grouped by benthic invertebrate community structure
Variable
Gp 1
Gp 2
Gp 3
Gp4
Gp5
Depth m
15.9 (9.3)
8.4 (3.9)
7.2 (4.8)
1.8 (0.6)
57.9 (22.0)
Nitrate (w) mg L"1
0.22 (0.08)
0.10 (0.08)
0.20 (0.07)
0.01 (0.00)
0.33 (0.07)
Silt%
40.3 (19.8)
42.9 (27.0)
27.7 (24.5)
50.7 (24.6)
13.2 (19.4)
Aluminium fig g~'
9.5 (2.2)
10.9 (2.1)
9.2 (3.4)
8.0 (2.1)
6.1 (2.3)
Calcium fig g'1
8.1 (3.3)
6.8 (5.6)
3.6 (3.0)
15.8 (9.4)
5.5 (3.5)
Loss on ignition %
12.4 (2.8)
13.5 (6.8)
7.3 (4.3)
21.9 (11.4)
11.1 (6.4)
Alkalinity mg L"1
92.5 (7.3)
82.2 (11.2)
64.3 (25.5)
78.6 (10.4)
111.2 (4.3)
Sodium fig g~>
1.26 (0.39)
1.8 (0.6)
1.8 (0.8)
1.5 (0.8)
0.8 (0.2)
pH(w)
8.1 (0.2)
7.6 (0.5)
7.3 (0.5)
8.2 (0.1)
7.9 (0.6)
(w) Indicates those variables measured in water.
I "1
E
o _2
v>
5
•
Gp 4
Eutrophic
silly & shallow
Gp 2
•
GpS
•
Gp 1
-
Oligotrophia
sancy & deep
•
Gp 3
Table 11. Summary of five validation runs of discriminant analysis
site predictions
-4 -2' 0 2 4
Discriminant function 1 (73.6%)
Fig. 5. Community structure site group means in discriminant
space.
with 86.8% for the groups based on community structure
using nine variables (Table 8). Again, variables such as
water column alkalinity and depth, which correlated well
with the ordination vectors, were important contributors
to the discriminant functions. This poorer discrimination
between the toxicity based groups is primarily due to the
greater degree of similarity between them.
While these predictions provide an estimate of the
ability of a suite of environmental variables to predict
biological structure, a more realistic indicator of predictive
capability is provided by using a separate set of test sites.
Ten sites were randomly removed from the reference
data base. A discriminant model based on the remaining
reference sites was developed. The environmental vari-
ables for the 10 test sites were then substituted and a
biological grouping predicted which was compared with
the actual grouping from the original classification. This
procedure was repeated separately five times for both the
community structure and toxicity data matrices.
For the community validation, between 80 and 100%
of the sites were correctly predicted from run to run with
an overall average of 90% correct (Table 11). In each of
the five runs, the Gp 1 sites were always correctly
predicted. In each of Gps 2, 3 and 4, one site was
Run
Correct (%) Group
Correct
Community structure
1
80
1
15 of 15
(100%)
2
90
2
4 of 5
(80%)
3
80
3
5 of 6
(83%)
4
100
4
6 of 7
(86%)
5
100
5
15 of 17
(88%)
Overall
90
Toxicity
1
80
1
22 of 29
(76%)
2
80
2
11 of 20
(55%)
3
4
70
Ml
3
1 of 1
(100%)
5
w
30
Overall
68
incorrectly predicted overall. The Gp 5 sites were
correctly identified 88% of the time. These data suggest
high confidence in an assemblage of organisms being
correcdy predicted at a new site.
The results of the validation runs on the basis of the
bioassay responses (Table-11) confirm their reduced
predictive ability as only 68% of the test sites were
correctly assigned. The discriminant model had the
greatest difficulty in predicting the Gp 2 sites (55%
correct). To determine the implications of this, we have
compared the observed result at the incorrectly predicted
sites with the average for the correct group and the
average for the predicted group (Table 12). Chironomid
survival showed little difference between the three groups;
for example, the Gp 1 sites all fell within the actual Gp 1
range and within the range of the predicted group. The
Gp 2 sites were also within the Gp 2 range and, with the
exception of site 5808, were within the range of the
predicted group. Survivorship in site 5808 was slightly
higher than for the predicted group. Chironomid growth
-------
210 T. B. REYNOLDSON ET AL.
Table 12. Values of bioassay endpoints for those sites incorrectly predicted and expected values for each test endpoint
Expected
CRSU
CRGW
HLSU
HLGW
HASU
HAGW
TTCC
TTYG
Gp 1
65-96
0.17-0.49
91-100
0.51-5.63
65-100
0.29-0.73
27-39
35-82
Gp 2
70-100
0.21-0.53
87-100
0.00-9.07
59-100
0.16-0.84
33-42
99-147
Gp 3
70-95
0.22-0.38
86-100
1.18-1.80
0-69
0.07-0.51
33-39
97-147
Gp 1 sites
Predicted to
0106
2
80
0.28
96.7
3.83
86.7
0.50
37.4
70.0
0107
2
77.8
0.20
93.3
5.17
98.2
0.62
39.7
93.0
0108
2
88.9
0.21
100
5.28
97.8
0.47
37.7
79.3
5506
3
75.0
0.25
98.0
1.70
60.0
0.49
34.0
82.0
5507
2
88.0
0.31
90.0
2.09
96.7
0.48
31.0
90.0
5710
3
92.0
0.31
100
1.25
69.3
0.36
38.0
83.0
Gp 2 sites
Expected
0310
1
91.1
0.52
100
15.43
95.6
0.68
47.0
153.7
0311
1
95.6
0.36
96.7
13.99
91.1
0.61
40.3
118.3
5502
3
90.7
0.33
94.0
3.15
92.0
0.46
38.0
113.0
5705
3
92.0
0.33
84.0
1.66
78.7
0.12
38.0
136.0
5807
3
85.3
0.44
100
1.48
83.1
0.54
41.0
100.0
5808
1
97.3
0.40
96.0
2.10
83.1
0.79
40.0
120.0
Expected value ranges are
2 standard deviations about the mean, and 1 standard deviation about the for TTCC and TTYG
Fig. 6. Location of sampling sites in Collingwood Harbour.
was within the expected range or higher (OJiO, 5807,
5808), with the exception of site 0107 in which growth
was lower than expected for the predicted group (Gp 2)
but not for the actual jjroup (Gp 1). This is a case where
a site could be deemed toxic when in fact it is not. For
Hexagema survival, only site 5705 is lower than predicted.
-------
PREDICTING STATE OF FRESHWATER SEDIMENT 211
Hexagenia growth is the most variable of the end-points.
None of the Gp 1 sites is affected by the incorrect
predictions. Of the Gp 2 sites, 0310 and 0311 show much
greater growth than expected but this would not
necessarily produce a toxic designation. Survival of
H. azteca is an end-point that could provide a false toxic
designation if not correctly predicted to Gp 3. The one
Gp 3 site in the validation tests was correctly identified
and site 5506 would not be considered toxic as it was
predicted to Gp 3, although the survivorship of Hexagenia
is lower than expected for Gp 1.
Application in Collingwood Harbour
Collingwood Harbour is located in Collingwood Bay,
Georgian Bay, Lake Huron (Fig. 6) and has been identi-
fied as an area of concern by the International Joint
Commission (International Joint Commission 1987)
partly because of sediment contamination by various
metals and partly based on eutrophication. As part of a
remedial programme for the harbour, sediment removal
is being considered. To define the extent of the sediment
contamination, it was decided to compare the more
generally used chemical criteria with the biological
approach developed in this paper.
Based on sediment chemistry and the Province of
Ontario's sediment quality criteria, the harbour was
heavily contaminated by metals (Table 13), with three
sites in the east boat slip (6706,6070 and 6708) and one
site in the west boat slip (6709) exceeding the Ontario
Ministry of Energy and Environment's (Persaud et al.
1992) severe effects criteria for copper, zinc, lead, arsenic
and iron. Furthermore, all the sites in both the boat slips
and the outer harbour exceeded low effect concentrations
for at least eight sediment variables and several sites for
all 12 variables for which criteria have been established.
This prompted the Canadian federal government (En-
vironment Canada) and the Province of Ontario to
consider removal of the contaminated material. However,
the large area of removal and the anticipated cost
prompted examination of the biological significance of
the contamination and the biological objective? developed
in this project were used to help make a managerial
decision.
Table 13. Sediment chemistry for contaminants for which guidelines are available
Total
N
Low
Severe
Site
6703
6704
6705
6706
6707
6708
6709
6710
6711
6712
6713
6714
6715
6716
6717
6718
6719
6720
6721
6722
6723
6724
6725
6726
6727
550
4800
975
2205
2042
1524
1244
1364
1825
2441
2211
1661
2499
2439
2371
2299
2431
2534
2128
2082
2239
2315
2194
2230
2293
2368
2305
Total
P
600
2000
1095
1290
1335
1230
945
1080
1200
1515
1560
607
866
821
801
1032
850
988
902
1124
779
949
1029
868
998
630
852
TOC
1
10
Total
Fe (%)
Mn
460
1100
1.8 1.6 522
2.1 1.5 480
3.5 2.2 570
2.3 11.2 594
1.8 14.4 642
2.1 8.9 646
2.4 5.0 1114
2.7 3.0 647
2.6 2.2 583
2.1 1.5 355
2.6 1.7 491
2.4 1.7 495
2.5 1.6 460
3.1 1.6 405
2.9 1.5 444
2.6 .1.6 453
2.3 1.8 487
2.8 2.0 480
2.1 1.7 472
2.2 1.8 489
2.4 1.9 525
2.3 1.8 507
2.4 2.2 508
2.7 1.7 464
2.3 2.0 526
Cr
Ni
Cu
Zn
As
Cd
Pb
26
16
16
120
6
0.6
31
110
75
110
820
33
10
230
25
21
37
129
2.5
0.6
76
24
21
35
123
2.5
0.5
63
29
24
101
411
2.5
1.2
329
65
33
2835
10780
105
3.9
7.9
85
36
4170
13943
137
4.9
974
58
31
2121
7527
70
4.2
724
47
27
893
3154
25
1.8
430
40
29
201
669
2.5
1.5
200
32
25
109
380
2.5
0.4
149
18
12
49
216
7
0.4
105
23
19
36
145
12
<0.2
71
22
20
35
141
9
0.3
71
20
19
32
125
<5
0.6
57
25
17
65
247
6
0.4
88
20
17
35
151
10
0.3
55
21
20
34
138
8
0.5
59
23
22
50
161
13
0.3
94
25
23
54
204
11
0.3
115
22
20
36
145
11
0.8
68
23
21
39
155
18
0.54
92
25
23
41
165
12
<0.2
91
25
23
41
169
<5
0.4
92
26
22
96
467
12
0.5
111
20
19
54
198
10
<0.2
195
26
26
45
182
15
0.9
105
Values in ppm dry wt.
-------
212
T. B. REYNOLDSON ET AL.
Table 14. Probability of Collingwood Harbour sites being a member
of one of five community groups, using MDA with nine environ-
mental variables
SITE
Gp 1
Gp2
Gp 3
Gp4
Gp 5
Predicted
Gp
6703
0.791
0.006
0.202
0.000
0.000
1
6704
0.826
0.006
0.168
0.000
0.000
1
6705
0.906
0.000
0.087
0.000
0.006
1
6706
0.885
0.000
0.113
0.000
0.001
1
6707
0.748
0.000
0.242
0.000
0.010
1
6708
0.924
0.000
0.071
0.000
0.004
1
6709
0.893
0.009
0.098
0.000
0.000
1
6710
0.839
0.000
0.158
0.000
0.002
1
6711
0.002
0.904
0.001
0.093
0.000
6712
0.993
0.003
0.005
0.000
0.000
1
6713
0.987
0.008
0.004
0.000
0.000
1
6714
0.987
0.010
0.002
0.000
0.000
1
6715
0.974
0.016
0.009
0.001
0.000
1
6716
0.769
0.095
0.095
0.042
0.000
1
6717
0.972
0.017
0.010
0.001
0.000
1
6718
0.933
0.032
0.032
0.003
0.000
1
6719
0.989
0.009
0.001
0.000
0.000
1
6720
0.991
0.007
0.002
0.000
0.000
1
6721
0.986
0.009
0.004
0.000
0.000
1
6722
0.993
0.006
0.001
0.000
0.000
1
6723
0.994
0.005
0.001
0.000
0.000
1
6724
0.992
0.006
0.002
0.000
0.000
1
6725
0.990
0.007
0.003
0.000
0.000
6726
0.965
0.027
0.001
0.007
0.000
1
6727
0.992
0.007
0.001
0.000
0.000
1
Using the nine predictor variables selected previously
(Table 8) and the reference site data matrix to develop
discriminant equations, we predicted community assem-
blages for each of the 25 sites in Collingwood Harbour
(Table 14) for which chemical data were available. With
the exception of site 6711, all the sites were predicted as
having a Gp 1 community assemblage (see Table 4). This
type of community is represented by the Tubificidae and
the chironomid, Procladius spp., as the most common
organisms, but several other species are also frequently
found (Table 4).
To determine whether the observed community was,
in fact, similar to the predicted community, we repeated
the ordination and examined the location of the Colling-
wood Harbour sites relative to the Gp 1 reference sites in
ordination space (Fig. 7). The results show that of the 24
sites predicted as having a Gp 1 community most fall
within the range of variation found in the reference sites
that comprise Gp 1. Exceptions were sites 6708 and 6717
on the second vector (Fig. 7a), and sites 6706,6707,6708
and 6709 on the third vector. One site (6711) was
predicted as having a community represented by Gp 2.
This site was well outside the range observed in the
reference sites (Fig. 7c) on vector 2.
Using the environmental data, we also predicted site
groupings (Table 15) for the expected responses in
sediment bioassays and sites were classified as being
members of either Gp 1 or Gp 2. Survival and growth of
1.5
1
™ 0.5
•2 0
o w
o
o
g -0.5
-1
-1.5
-2
Gp 1
5708,
a
-------
PREDICTING STATE OF FRESHWATER SEDIMENT 213
Table 15. Collingwood Harbour sites, predicted group and mean values for test end-points
Site
ToGp
CRSU
CRGW
HXSU
HXGW
HASU
HAGW
TTCC
TTYG
88.0
0.38
98.0
6.87
89.3
0.72
22.4
22.0
81.3
0.38
98.0
8.11
94.7
0.75
24.4
22.8
80.0
0.43
100.0
5.78
90.0
0.66
21.8
14.6
72.0
0.39
98.0
5.62
93.3
0.50
27.2
22.4
86.6
0.33
100.0
6.17
90.7
0.42
25.8
28.8
82.6
0.36
94.0
5.35
94.7
0.53
24.4
22.3
68.0
0.46
100.0
3.86
94.7
0.53
36.4
42.4
78.6
0.40
100.0
5.04
88.0
0.60
31.8
42.2
85.3
0.40
100.0
4.56
84.0
0.50
29.0
55.8
89.3
0.52
80.0
10.12
94.7
0.70
44.7
171.7
76.0
0.51
100.0
10.93
88.0
0.66
43.5
125.5
86.6
0.61
100.0
10.59
84.0
0.74
45.6
135.4
85.3
0.49
100.0
10.94
82.7
0.76
43.0
126.0
90.6
0.65
100.0
10.58
89.3
0.79
NA
NA
89.3
0.64
94.0
10.40
86.7
0.87
43.4
153.6
86.6
0.66
98.0
11.16
93.3
0.82
46.6
146.8
88.0
0.45
100.0
8.56
92.0
0.62
40.0
74.2
90.6
0.56
100.0
6.70
89.3
0.74
43.8
36.6
84.0
0.58
98.0
9.97
89.3
0.74
45.0
117.2
88.0
0.55
98.0
9.11
77.3
0.71
42.8
74.2
96.0
0.35
100.0
7.98
90.7
0.71
39.2
42.8
76.0
0.51
98.0
8.53
90.7
0.66
42.4
62.4
90.6
0.37
84.0
8.34
94.7
0.61
38.4
77.8
92.0
92.0
0.34
96.0
7.98
72.0
0.75
39.4
48.8
0.38
98.0
6.28
85.3
0.56
37.8
45.8
6703
6704
6705
6706
6707
6708
6709
6710
6711
6712
6713
6714
6715
6716
6717
6718
6719
6720
6721
6722
6723
6724
6725
6726
6727
Values in italic are below the range expected for the Gp from Table 10.
1.5
1
CM
S 0.5
o
g 0
-0.5
-1
-1.5
a
Gp1
-
^}
. i i
Vector 1
-3 -2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2
Vector 1
1.5
F*. o™,™™ of H-bo. - -i»"(0)'^:o>'Gp:211(+>'
Harbour.
M
-------
214 T. B. REYNOLDSON ET AL.
C. riparius and H. azteca were not significantly reduced
below levels in all the reference sites and thus, these
end-points did not show evidence of toxicity at any of the
sites. In the case of the amphipod, H. azteca, two sites
showed slightly more growth than expected (6704 and
6717). Similarly nine sites had greater levels of growth of
Hexagenia spp. than expected for the range of sediments
in the bay. Two sites had slighdy reduced survival (6712
and 6725). The effects of sediment contamination were
most notable in the data for reproduction of T. tubifex.
All the sites in the east slip and 1-2 sites in the outer
harbour showed both reduced production of cocoons and
total young. Several sites in the southeast corner of the
outer harbour also showed a reduction in the number of
young tubificids produced.
When plotted in ordination space (Fig. 8), the sites
predicted as being Gp 1 all fell within the boundary of
the reference sites. However, a number of sites predicted
as Gp 2 (Fig. 8c,d) fell outside the reference site boundary,
viz. 6719,6720, 6723,6724,6725,6726, 6727 on vector 2,
and 6720, 6722, 6723, 6724, 6726 and 6727 on vector 3.
These are the same sites that had reduced production of
young tubificids (Table 15). While the reduction in
production of young could suggest a low level of chronic
toxicity at these outer harbour sites, the fact that
reproduction does occur and that tubificid oligochaetes
are abundant at these same sites in the harbour suggests
another mechanism may be responsible. The chemical
data do not indicate higher levels of contaminants at
these specific Gp 2 sites (Table 13). However, these sites
did have a slightly higher clay content (mean 25%)
compared with the other harbour sites (19% clay).
Tubificid worms are known to be sensitive to higher clay
contents (Reynoldson et al. 1991).
Sediment remediation may not be warranted at all
harbour sites in Collingwood Bay based on the data from
this analysis. The spatial pattern (Fig. 9) from the analysis
of community structure shows good correspondence with
the chemical data. For example, the in situ benthic
community structure was only outside the range of what
should be present if the sediment were clean at sites
where the levels of metals exceed the severe effects levels
based on sediment quality criteria. Only one site in the
outer harbour was identified as being outside the range
of our reference data base (i.e. 6717) but the levels of
contaminants at this site were all below the severe effects
limits and this outlier can be attributed to the high
numbers of Porifera at this site rather than toxicity.
The results from the laboratory tests do not indicate a
sediment toxicity problem in the harbour with the
exception of the results for reproduction of the oligochaete
worm, T. tubifex. Total production of young is reduced
in the east boat slip, at three sites in the outer harbour
and in the southeast corner of the harbour. The reduction
(b)
-------
PREDICTING STATE OF FRESHWATER SEDIMENT
215
in reproduction in the east slip may be attributed to the
high levels of metals at these sites. Oligochaetes are
known to be particularly sensitive to metal contamination
(Hynes 1960; Brinkhurst & Jamieson 1971; Aston 1973;
Chapman et al. 1980; Wachs 1980). In addition, the
number of tubificids in the field-collected benthic inver-
tebrate samples at these same sites were very low.
Based on the above results, removal of contaminated
sediment is warranted in the boat slips. The remainder of
Collingwood Harbour cannot be considered as having a
degraded benthic community despite the fact that several
of the provincial chemical sediment criteria are exceeded
for metals. There is very little weight of evidence in the
laboratory toxicity test data for an extensive remedial
action programme in the outer harbour. The reduction
in total production of young in tubificids occurs at only a
few sites and this effect may not be in response to
toxicity but rather to physical characteristics of the
sediment. This case study emphasizes the value of a
combined laboratory and field approach using both
chemical and biological data in interpreting the effects.
DISCUSSION
This study was undertaken to provide an alternative to
environmental guidelines and criteria for sediments in
the Laurentian Great Lakes based solely on comparisons
of bulk chemical concentrations of contaminants in sedi-
ments to levels of these same compounds known to cause
a toxic response in biota (International Joint Commission
1988; Persaud et al. 1992). In our view, the results
provide a more relevant and realistic method for deter-
mining environmental impact. We believe that ecosystem
integrity is primarily a biological concern. The past trend
of developing decision-making criteria based on chemical
concentrations has primarily been in response to the
inability of biologists to provide environmental managers
with the information they require to make decisions.
However, the development of standardized methodologies
and multivariate statistical analysis, which allows pre-
diction of biological responses based on simple environ-
mental variables, have been revolutionary in the
application of biological data to the environmental
decision-making process. We view this paper as a demon-
stration of one approach to the use of biological data in
decision making.
Comparison with other multivariate studies
A number of other studies have demonstrated the ability
to predict the structure of benthic invertebrate com-
munities from a set of environmental variables (Furse et
al. 1984; Wright et al. 1984; Armitage et al. 1987;
Corkum & Currie 1987; Moss et al. 1987; Ormerod &
Edwards 1987; Johnson & Wiederholm 1989). Most of
the studies predicting community assemblages have been
conducted in lotic systems and their prediction accuracy
is in the range of 68.9 to 79.6% (Reynoldson & Metcalfe-
Smith 1992). The only other example of this method
being used in lake systems is the work of Johnson and
Wiederholm (1989) who showed that they could correctly
predict 90% of the benthic invertebrate assemblages in a
set of Swedish lakes using variables such as depth, silica,
bicarbonate and phytoplankton volume. From our data,
we were similarly able to correctly predict the community
structure more than 86% of the time. Furthermore, as far
as we know, this is the first time that an attempt has been
made to examine and predict both structural (com-
munities) and functional (survival, growth and repro-
duction) biological attributes.
Variation in functional responses
While exposure of benthic invertebrates to whole sedi-
ments is frequently used in the assessment of contamin-
ation, there has been little examination of the natural
range of responses in clean sediments with a variety of
geochemical characteristics. The data from this study
show that there is considerable variation in the measured
end-points to sediment attributes, and this is dependent
on the species of invertebrate used in the bioassay and
the response being measured. For example, C. riparius is
very robust in its response to sediment type and neither
growth nor survival are notably affected by sediment
quality or nutrition. Similarly, H. azteca shows low
variation in growth but survival is more variable.
Chiconomus riparius and H. azteca were both fed over the
period of the bioassay with fish food flakes (Nutrafin*")
and therefore are less likely to respond to the nutritive
quality of the various sediments. Ankley et al. (1994)
have also shown that addition of exogenous food to test
sediments with H. azteca and C. riparius significantly
reduced variability in the test end-points. It is surprising,
however, that there was little correlation in response to
particle size distribution, mineralogical composition or
organic carbon content.
Hexagenia spp. showed the least variation in survival
but growth was highly variable. This variation is likely
due to the fact that Hexagenia were not fed during the
laboratory tests and the nutritive quality of the sediment
may have a greater influence on the growth of the
organisms than expected. We suspect that a lack of
exogenous food also explains the greater variability in the
reproductive end-points measured in the oligochaete test
with T. tubifex, and total numbers of young oligochaetes
has been shown to be sensitive to the amount of available
food in the sediment, as measured by organic content
(Reynoldson et al. 1991).
-------
216 T. B. REYNOLDSON ET AL.
Comparison with other chemical and biological
approaches
The use of this approach in Collingwood Harbour illus-
trates its advantages over the more traditional assessment
methods that rely on chemical guidelines. Comparison of
the concentrations of metals measured in sediments
collected from the harbour with provincial sediment
quality criteria (Persaud el al. 1992) demonstrates the
inability of the criteria to determine non-impact. In
addition, such criteria were only capable of providing a
gradient in terms of high and low concentrations) for an
array of contaminants.
The biological approach on the other hand was able to
specify where remediation was required and where con-
tamination was not eliciting a biological response. In
fact, there was good concordance between the chemical
and biological data for the most severely contaminated
sites (i.e. both the in situ data on community structure
and the data from the laboratory toxicity tests indicated
toxicity). The most contaminated sites (6706,6707,6708,
6709; Table 13) had the most depauperate communities
with few organisms present, particularly the Oligochaeta.
These same sites also had lower than expected T. tubijex
reproduction in the chronic assays (Table 15).
Other sites in the outer harbour with lower contaminant
concentrations required the biological data to define
whether an impact was occurring. Two sites exceeded
the severe effects level based on chemical criteria yet
biological effects were not demonstrated at these sites.
Site 6710, which had a high copper concentration (210
ppm) and site 6705 with high lead concentration (329
ppm) had reduced numbers of organisms but were within
the range of variation found in reference sites. No
evidence of toxicity was demonstrated in laboratory tests
at these sites. None of the sites which exceeded any of
the low effect criteria demonstrated any divergence from
an expected community as defined by the reference sites.
The toxicity data demonstrate the importance of using
a range of species rather than relying on a single species
of invertebrate. In the Collingwood study, the only
bioassay that showed any negative response was repro-
duction in the oligochaete, T. tubifex. The other three
species showed no divergence from results found in clean
sediment. This contradicts the commonly held belief
that oligochaete worms are insensitive to contamination.
The only other unexpected divergence was the enhanced
growth of Hexagenia at several sites in the outer harbour.
This may be attributed to a slight effect of eutrophication
(G. Krantzberg, pers. comm. 1994).
We view these data as providing excellent validation of
a mi tvariate approach for establishing referenet or
nomin-i conditions for biological variables which can be
used in a practical manner to assist in management
decisions. In the case of Collingwood Harbour we would
fully support the remediation of the east and west boat
slips. However, the outer harbour appears to be un-
impacted and may require no remedial activity as both
the benthic community structure and the results of
chronic bioassays in four invertebrate species do not
distinguish these sites from reference conditions. This is
clearly something that chemical criteria were unable to
do.
These data from 96 sites are the first portion of a data
set that will finally consist of more than 250 sites and
thus have even greater power for impact resolution.
While this approach has numerous merits, particularly
its ability to incorporate normal variability in predicting
biological state and to define ecological targets, there are
some disadvantages. The approach is initially labour
intensive, requiring the acquisition of a large data base.
This data base is necessary to ensure that the range of
potential natural variation is captured by the reference
data set and that the appropriate statistical degrees of
freedom are available for performing the multivariate
algorithms. Further, the model can only be applied to
test sites within the range encompassed by the reference
sites. Lastly, it is not a method that is intuitively easy to
comprehend and thus methods will be required to transfer
the technology to potential users.
Finally, there are a number of issues that have yet to
be addressed. The grouping methodology is somewhat
arbitrary. The degree to which the selected number of
groups represent true community assemblages is un-
known. The need to identify groups is a requirement of
the analytical method selected; discriminant analysis
requires a grouping variable in order to predict group
membership from a set of variables. In this case we used
two criteria to define group membership. First, the
structure of the data, selecting as many groups as possible
(five) before the structure began to be lost and many
small groups of few sites formed. Second, the distribution
of the selected groups; the strong geographic correlation
between the groups and the lakes suggests that these are
meaningful groupings. However, further investigation of
"an appropriate a priori method for group definition is
desirable. A possible approach could be to use replicate
samples from sites and set the number of groups by the
point at which replicate samples we retained within a
group. The majority of the reference sites have only been
visited on one occasion. The robustness of the reference
state must be established and the temporal trajectory of a
site in ordination space must be determined. This will
require resolution of both seasonal and annual trajectories.
At present the majority of sites have been sampled in the
fall and this may require sampling of test sites to be
restricted to that season. The appropriate taxonomic
level for incorporation intdthe model requires further
determination. For practical considerations, it would be
advantageous if a reduced taxonomic effort provided
sufficient predictive capability for the model to be effec-
tive. This would ultimately mate this method more
-------
PREDICTING STATE OF FRESHWATER SEDIMENT
217
attractive. Identification to the taxonomic level of species
or genus is a major effort and if family or higher
taxonomic levels of identifications are shown to be
adequate in their diagnostic ability and predictability,
then considerable time, effort and expense will be saved.
Similarly, it is unreasonable to expect all investigators to
use the same equipment as used in this study. Therefore,
the effects of sampler type on the prediction of sites will
also need to be established. These issues are currently
under investigation.
CONCLUSIONS
We are confident that the preliminary results of this
study demonstrate the ability to develop a reference
data base that can provide meaningful information on
community assemblages of benthic invertebrates at a
'clean' site and can also predict the responses of selected
invertebrates to natural sediments. These biological
attributes can be predicted from a relatively small set of
environmental characteristics. The ultimate objective is
to develop a set of numerical guidelines based on bio-
logical attributes that can be used in making appropriate
management decisions.
ACK NOWLEDGEMENTS
The authors wish to thank Mr Griffin Sherbin, Great
Lakes CleanUp Fund, and Dr Steven Lozano of the US
EPA, Duluth, MN, provided partial financial support to
this study. In particular, we would like to thank Susan
Humphrey of Environment Canada who has been a
stalwart financial and moral supporter. We also acknow-
ledge Dan Faith and Dr Norris Sr for their comments on
this paper. Finally, the assistance of Craig Logan, Danielle
Milani, Cheryl Clarke, Leeanne Gris and Scott Kirby is
gratefully acknowledged.
REFERENCES
Ankiey G. T., Benoit D. A., Baiogh J. C., Reynoldson T. B„ Day
K. E. & Hoke R. A. (1994) Evaluation of potential confounding
factors in sediment toxicity tests with three freshwater benthic
invertebrates. Erwir. Toxicol. Chem. 13,627-35.
Armitage P. D., Gunn R. J. M., Furse, M. T., Wright J. F. & Moss
D. (1987) The use of prediction to assess macroinvertebrate
response to river regulation. Hydrobtolopa 144, 25-32.
Aston R. J. (1973) Tubificids and water quality: A review. Erwir.
Pollut. 5, 1-10.
Bedard D., Hayton A. & Persaud D. (!992) Ontario Ministry of
Environment laboratory sediment biological testing protocol.
Water Resources Branch, Ontario Ministry of Environment,
Toronto, Ontario, 26 pp.
Belbin L. (1991) Semi-strong hybrid scaling, a new ordination
algorithm. J. Veg, Sci. 2,491-6.
Belbin L. (1993) patn, pattern analysis package. Division of Wildlife
and Ecology, CSIRO, Australia.
Borgmann U., Ralph K. M. & Norwood W. P. (1989) Toxicity test
procedures for Hyaiella azteca and chronic toxicity of cadmium
and pentachlorophenoi to H. azteca, Gammarus fasciatus and
Daphnia magna. Arch. Erwir. Ctmtam. Toxicol. 18, 756-64.
Brinichurst R. O. & Jamieson B. G. M. (1971) Aquatic Oligochaeta
of the World. University of Toronto Press, Toronto, 860pp.
Cairns J. J. & van der Schalie W. H. (1980) Biological monitoring
Pan 1: Early warning systems. Water Res. 14, 1179-96.
Chapman P. M. (1986) Sediment quality criteria from the sediment
quality triad: An example. Envir. Toxicol. Chem. 5, 957-64.
Chapman P. M. (1990) The sediment quality triad approach to
determining pollution-induced degradation. Sci. Total Envir.
97/98,815-25.
Chapman P. M., Churchland L. M., Thomson P. A.-& Michnowsky
E. (1980) Heavy metal studies with oligochaetes. In Aquatic
Oligochaete Biology (eds R. O. Brinkhurst & D. G. Cook).
Plenum Press, New York.
Corkum L. D. & Currie D. C. (1987) Distributional patterns of
immature Simuliidae (Diptera) in northwestern North America.
Freshwat. Biol. 17,201-21.
Day K. E., Kirby R. S. & Reynoldson T. B. (1994) Sexual
dimorphism in Chironomus riparius (Meigen): Impact on
interpretation of growth in whole sediment toxicity tests.
Envir. Toxicol. Chem. 13, 35-9.
Duncan G. A. & LaHaie G. G. (1979) Size analysis procedures
used in the sedimentology laboratory. NWRI, CCIW,
Hydraulics Division Manual. September, 1979.
Faith D. P., Minchin P. R. & Belbin L. (1987) Compositional
dissimilarity as a robust measure of ecological distance: A
theoretical model and computer simulations. Vegetation 69,
57-68.
Furse M. T., Moss D., Wright J. F. & Armitage P. D. (1984) The
influence of seasonal and taxonomic factors on the ordination
and classification of running-water sites in Great Britain and
on the prediction of their macro-invertebrate communities.
Freshwat. Biol. 14, 257-80.
Hanes E. C. & Ciborowski J. H. (1992) Effects of density and food
limitation on size variation and mortality of larval Hexagenia
rigida (Ephemeroptera: Ephemeridae). Can. J. ZooL 70,
1824-32.
Hunsaker C. T. & Carpenter D. E. (eds) (1990) Ecological indicators
for the environmental monitoring and assessment program.
EPA 600/3-90/060. U.S. Environmental Protection Agency,
Office of Research and Development, Research Triangle Park,
North Carolina.
Hynes H. B. N. (1960) The Biology of Polluted Waters Liverpool
University Press, UK 202 pp.
International Joint Commission (1987) Report on Great Lakes
water quality. Report to the International Joint Commission,
Great Lakes Water Quality Board. Windsor, Ontario. 236 pp.
International Joint Commission (1988) Procedures for the assess-
ment of contaminated sediment problems in the Great Lakes.
Report to the Great Lakes Water Quality Board. Windsor,
Ontario. 140 pp.
Johnson R. K. & Wiederholm T. (1989) Classification and ordination
of profundal macroinvertebrate communities in nutrient poor,
oligo-mesohumic lakes in relation to environmental data.
Freshwat. Biol. 21, 375-86.
Long E. R. & Chapman P. M. (1985) A sediment quality triad:
Measures of sediment contamination, toxicity and infaunal
community composition in Puget Sound. Mar. Poll. Bull. 16,
405-15.
-------
218 T. B. REYNOLDSON ET AL.
assessment of aquatic ecosystem health using benthic inver-
tebrates. J- Aquai. Ecosystem Health 1: 295-308.
Reynoldson T. B. & Zarull M. A. (1993) An approach to the
development of biological sediment guidelines. In Ecological
Integrity and the Management of Ecosystems (eds G. Francis,
J. Kay & S. Woodley) St Lucie Press, Florida.
Reynoldson T. B., Day K. E., Clarke C. & Milani D. (1994) Effect
of indigenous animals on chronic endpoints in freshwater
sediment toxicity. Errnir. Toxicol. Chem. 13, 973-7.
Wachs B. (1980) Bioakkumulation und biogene retention von
metallen in Fliessgewasseren. Munchener Beitrage zur
Abwasser-, Fischerei- und Flussbiologie 32, 387 -414.
Wickware G. M. & Rubic C. D. A. (1989) Ecoregions of Ontario.
Environment Canada, Ecological Land Classification Series
No. 26. 37pp.
Wright J. F., Moss D., Armitage P. D. & Furse M. T. (1984) A
preliminary classification of running-water sites in Great Britain
based on macro-invertebrate species and the prediction of
community type using environmental data. Freshmtt. Biol. 14,
221-56.
Zarull M. A. & Reynoldson T, B. (1992) A management strategy for
contaminated sediments: Assessment and remediation. Water
Poll. Res. J. Canada 27, 871-82.
Appendix I. Great Lakes Benthic Community Structure Study: Species list
Gastropoda
Bithyniidae
Bithynia lentaculata
Hydrobiidae
Amnicola limosa (Amn lim)
A. walkeri
Marstonia decepta
Probytkinella lacustris
Hydrobiidae immatures
Lymnaeidae
Fossaria obrussa
Physidae
Physella mtegra
P. species (Phy spp)
Planorbidae
Armiger crista
Gyraulus circumstriatus
G. deflectus
Helisoma anceps
Prcrmenetus exacuous
Valvatidae
Valvata iemisi
V. piscinalis (Val pis)
V. sincera
K tricarinata (Val tri)
Viviparidae
Campeloma decision
Unknown spp (damaged)
Pelycepoda
Sphaeridae
Pisidium cascrtanum (Pis c:
P. compression (Pis com)
P. ferruginevm
P. bcnslowaaum (Pis ben)
P. mtidum (Pis nit)
P. ventricosum
P. unknown (Pis unk)
Moss D., Furse M. T., Wright J. F. & Armitage P. D. (1987) The
prediction of the macroinvertebrate fauna of unpolluted
running-water sites in Great Britain using environmental data.
Fresfrwat. Biol. 17,41-52.
Ontario Ministry of Environment (1990) The Canadian Great Lakes
basin intake outfall atlas, 8 vols (ed. M. Griffiths). Water Re-
sources Branch, Toronto, Ontario.
Ormerod, S. J. & Edwards R. W. (1987) The ordination and
classification of macroinvertebrate assemblages in the catchment
of the River Wye in relation to environmental factors. Fres/mat.
Biol. 17,533-46.
Painter S. (1992) Regional variability in sediment background
metal concentrations and the Ontario sediment Quality Guide-
lines, NWRI Report No. 92-85. Environment Canada, Bur-
lington, Ontario.
Persaud D., Jaagumagi R. & Hayton A. (1992) Guidelines for the
protection and management of aquatic sediment quality in
Ontario. Water Resources Branch, Ontario Ministry of En-
vironment. Toronto, Ontario, 23pp.
Reynoldson T. B., Thompson S. P. & Bamsey J. .L. (1991) A
sediment bioassay using the tubifirid oligochaete worm Tubifex
tubifex. Emir. Toxicol. Chem. 10, 1061-72.
Reynoldson T. B. & Metcalfe-Smith J. (1992) An overview of the
Sphaerum miidum
S. simile
S. striatum
S. unknown
Musculim partinium (Mus par)
M. securis
M. iramversum
Unionidae
Elliptic) camplanaia
Lampris radiaia
Dreissenidae
Dreissena pofymorpha (Dre pol)
Diptera
Chironomidae
Chironomus (Chi spp)
Cladopelma (Cpe spp)
Clonotonytarsus
Cryptochironomus (Cch spp)
Cryptoteadipes (Cte spp)
D/crotendipes (Die spp)
Dcmicryptochironamus (Dem spp)
Endochirooomus (End spp)
Glypotendipes (Gly spp)
Harmschia
Micropscctra (Mps spp)
Microiendipes
Nilothauma
Pagastiella
Parachironomus
Paracladoplema
Paralauterborwella
Paratendipes
Poiypcdium (Pol spfli
Pseudockironomus (Pse spp)
Stietocbirooomus (Sti spp)
Tanytars us (Tan spp)
Tribelos
-------
PREDICTING STATE OF FRESHWATER SEDIMENT
219
Appendix I. continued
Stmpellina
Zavreliella
Unknown chironominae
Potthastia
Protattypus
Coryntmeura
Cricocopus
Epoicricotopus
Heterotrissocladius (Het spp)
Nanocladius
Parakiefferiella
Psectrocladius
Unknown orthocladinae
Ablabesmyia
Clinoianypus
Coelotanypus (Coe spp)
Larsia
Prociadius (Pro spp)
Tanypus
Monodiamesia
Ceratopogonidae
Bezzia sp.
Mallochohelea sp.
Probezzia sp.
Chaoboridae
Chaoborus sp. (Cha spp)
Empididae
Ephmeroptera
Ephemeridae
Hexagenia limbaia
Caenidae
Caenis sp.
Colembola
Trichoptera
Polyccntropodidae
Polycentropus sp.
Phylocentropus sp.
Helicopsychidae
Helicopsyche
Leptoceridae
Leptocerus ameriumus
Mystacides sp.
Nectopsyche sp.
Oecetis sp.
Molannidae
Molatma sp.
Hydroptilidae
Agraylea sp.
Polychaeta
Sabellidae
Manayunkia speciosa (Man spp)
Oligochaeta
Lumbriclidae
Edipidrilus lacustris
Lumbriadus variegatus
Stylodrilus heringiaaus (Sty her)
Enchytrcidae
Naididae
Arcteotuds lomondi (Arc lom)
Amphichaeia leydigi
Chaetogaster diaphanus
Nais barbata
N. elinguis
N. pseudobtusa
N. simplex
Dero digitate (Der dig)
Pristina leidyi
Prisiinella acuminata
Specaria josinae (Spe jos)
Stylaria lacustris (Sty lac)
Uncmais uncinaia
OpUdtmais serpentina
Piguetiella nrichiganensis
Vejdovskyella intermedia (Vej int)
Tubifiddae
Immatures with hair chaetae (A) (Imm chr)
Immatures without hair chaetae (B) (Imm coh)
Auiodrilus americona
A. limnobms
A.pigueti (Aulpig)
A. pluriseta (Aul plu)
Branchiura somerbyi
Ilyodrilus templetoni
Limnodrilm claparedianus
L. cervix
L. hoffmeisteri (Lim hoQ
L. profundicola
Potamotkrix bedoti
P. moldaviensis (Pot mol)
P. vejdovskyi (Pot vej)
Quastadrilus multisetosus (Qui mul)
Spirosperma ferox (Spi fer)
Tasserkidriks hessleri
Tubifex tubifcx (Tub tub)
Hirudinea
Glossiphoniidae
Alboglassophonia heteroclita
Gloiobdella elongata
Helobdtlla stagrutlis
Piscicolidae
Myzobdella lugubris
Platybeiminthes (Platy)
Isopoda
Asellidae
Caecidotea racovitzai
C. intermedins (Cae int)
C. sp. (Cae spp)
Amphipoda
Gammaridae
Gammarus pscudolimnacus (Gam pse)
Haustoriidae
Diaporeia boyi (Dia hoy)
Taliridae
Hyllela azteca
Coelenterata
Hydridae
Hydra americana (Hyd ame)
Porifera (Porif)
Tardigrada
Macrobiotidae
Dactylobiotus
Note those taxa used in classification analysis are shown in bold.
-------
DATA MANAGEMENT ISSUES AND GIS APPROACHES
FOR BIOLOGICAL ASSESSMENT
Tony SelJe
U.S. Environmental Protection Agency, Region VIII
999 18th Street, Suite 500
Denver, CO 80202
NOTES'-
Within the federal government, environmental samples are traditionally
collected for a single purpose, i.e. to answer important questions about a specific
area of concern. Unfortunately, once serving that purpose the data
immediately becomes devalued because no plan or mechanism is put in place to
extend the data use beyond the original purpose.
This presentation will focus on how we can make the biological data collected
retain its value and increase its utility over time by applying sound data
management principles. We will discuss developing and implementing a data
management plan as part of the monitoring or sampling plan, integrating field
and laboratory data with ancillary data, and managing data and information
for the widest possible array of applications and audiences.
We will then take a more detailed look at one application, the use of
Geographic Information Systems (GIS) for biological assessment. We will
discuss the various ways GIS can be employed, including logistical support,
data management, data development, analysis, and data/information display
and distribution.
18
-------
Why Do We Collect Biological Data ?
# To determine what lives (or doesn't live)
at the sample site
# To create biota-relevant environmental
health metrics
# To measure changes to the biota over time
EE\ R8 Bioassessment Workshop, T. Selle 8128195
(jftl
-------
What Do We Do With Biologic Data ?
Problem specific analysis for species
Site/area assessment for environmental health
Trends/change detection for environmental and
species health
EBi R8 Bioassessment Workshop, T. Selle 8/28/95
20
-------
What Don't We Dn (TJvjnlM ?
# Store data in a common electronic format
• Integrate sampled data
• Make it available to everyone else
• Use the data for assessment other
than original purpose
EBi R8 Bioassessment Workshop, T. Selle 8128195
fJBLl
%. 9
^ PROTfe0
21
-------
How Do We Change?
Integrated Data ManaeementJ
11
-------
Data Management Plan
Is Part of Sampling Plan
% Details sample integration
# Details data standardization (IDs, Locations)
# Details data distribution
# Describes data dictionaries and other metadata
EE1R8 Bioassessment Workshop, T. SeUe 8/28/95
23
-------
Data Management Plan
Sample Integration
Field Measures
Lab Chemistry
Fish
site-id
sample-id
site-id
sample-id
lab-id
latitude
lab-id
fish-id
lab-id
longitude
fish-id
habit-id
site-id
~
site-id
sample-id
latitude
lab-id
longitude
habit-id
Macro Invertibrates
Habitat
EBU R8 Bioassessment Workshop, T. Selle 8128195
24
-------
Tntegmting Compu
And Applications
ER1R8 Bioassessment Workshop, T. Selle 8128195
25
-------
Institutional Changes Needed
9 Dedicate to long term data use
• Resources for network/application integration
• Resources for integrated data management
• Commit to using data for decision-making
26
-------
WhatlsA G
How Does
EPA R8 Bioassessment Workshop, T. Selle 8/28/95
-------
WHAT MAKES GIS UNIQUE?
1. All data is physically indexed in space.
2. GIS can be used throughout a project lifecycle
A. PLANNING / DEVELOPMENT
B. DESKTOP SENARIOS / FINE-TUNING OF PLANS
C. IMPLIMENTATION
D. ANALYSIS
E. REPORTING
F. DECISION SUPPORT
28
-------
Fig. i. Br. John Snow's map (1855) of deaths from cholera in the Broad Street
area of London in September 1834
29
-------
m i ( i i •
fVJ)f* , ^ •" Number of Element Occurances
I —""' \r_ . i/'iy: ^i/u7 A\ • J Within Conservation Sites
^7^1 /yj C f U ( in the South Platte River Basin
1 ]. ' ~ • «^* •» / J \ \ \ |V J
r « ff» «r I / y j 1 \ ' » 1 I
C.t . /f - f / ^ \ 5 \ HH Greater than 100
1 Ni ^ ^jr / J y ar»yy n* w»» *****J^y»
k ' Hv i-^5. • / 1 1 ' V j J o 25 to 50
1 ^ 1 ¦ | ^ ^ \ J I / J _ ^ t.......¦.¦ j v * S>yi¦S^uVitf iltTiT^i^K) lTuSTihi |*<|*">f>
I I * 1^ -« L^f ^' * ^ % \ [ / J I ^ wTL'wwfi'iiK wMlS >£"«*>
1 1 V i8%3t C\r I -»jAiI/i i. s 1 ¦ ¦ ¦¦ n ms rnummt* ** iKwuffwn n» wmmn #
1 Juain-A3 u/ V rs*:^^ » v ' ^Jf V ~ » I 1 * V"* 1 , «y »*¦ w w< w>yy«fa wi ¦*»>¦»¦> wm
#$ - • \„fa/y I-)v At r i i ¦ u»tk»s tfsM-mm.-.
1
-------
DRAFT
1994 Region VIII REMAP Sampling Sites
Fish Sampling Species and Totals
Legend
-------
CLEAR CREEK BASIN, COLORADO
-------
w
H
O
2
n
>
n
sc
>
z
o
m
00
n
tn
5=
>
H
m
n
x
(D
2
w
H
73
<
oo
CO
O
O
ADIRONDACK PARK
DRAINAGE LAKES
PALEO INFERRED CHANGE IN pH WITH
PRECIP. AND ELEVATION OVERLAYED
A pH
~ •
~ •
A O
A O
A ©
-.75 or more
-.75 to -.50
-.50 to -.25
-.25 to 0
0 to .25
.25 or more
Preeip. contour interval 5 cm/year
Elevation greater than 600 meters
Figure 11-43a.
Maps of paleolimnological inference of historical changes in (A) pH, (B) ANC,-„ (C) Al,, and (D) DOC for Adirondack
drainage lakes included in the PIRLA-I and PIRLA-II studies.
-------
USE OF PERIPHYTON AS A BIOASSESSMENT TOOL
Donald Charles
Patrick Center for Environmental Research
The Academy of Natural Sciences
1900 Benjamin Franklin Parkway
Philadelphia, PA 19103
NOTES:
Analysis of algal assemblages is becoming an increasingly important tool for
bioassessment and management of our nation's streams, rivers, lakes and
wetlands. Both of the national programs that monitor aquatic ecosystems,
USGS's NAWQA and EPA's EMAP-SW, have algal components. Montana,
Kentucky, and Idaho have programs to monitor streams and rivers; other states
are showing an interest. Several regional monitoring programs also exist (e.g.
City of Austin, TX), and many individual sites have been studied. Large scale
paleolimnological studies of diatom remains in lake sediments have been
implemented to assess recent and long-term ecosystem changes in response to a
variety of anthropogenic stresses (acidic deposition, land-use change).
There ate many advantages of algae as aquatic indicators. Algae are a
primary food base for aquatic ecosystems. They are widely distributed in most
types of habitats. There are thousands of taxa and very large numbers of
individuals can be collected easily; assemblages contain considerable ecological
information. Taxa are identifiable to the lowest taxonomic level (in the case of
diatoms). Distributions of most taxa are closely correlated with water
chemistry and other environmental characteristics. Algae respond rapidly to
change. Samples of assemblages preserve for long periods of time (especially
diatoms) and take little storage space. Overall, analysis of algae is cost
effective compared with other groups of organisms.
In general, algae are better indicators of water chemistry characteristics than
fish and benthic invertebrates. A program with all three groups of organisms
provides a balanced set of indicators.
Many advances in the past 5 -10 years have increased the potential indicator
value of algal assemblages. Advances include better taxonomy, more and
higher quality ecological data, availability of computer software and
hardware to store and analyze large data sets, and advanced multivariate ami
statistical.methods to obtain the most ecological information from assemblage
data.
Until recently, algal results were expressed as relatively simple metrics,
including percent of indicator taxa and ecological groups, indices based on
ecological groups, diversity, and percent similarity. New applications of
Canonical Correspondence Analysis and inference models based on weighted
averaging are providing both improved understanding of ecological conditions
and more effective bioassessment indicators.
34
-------
TREND DETECTION IN BIOLOGICAL MONITORING:
EVALUATING PATTERNS OF COMMUNITY SIMILARITY
E.L. Silldorff, P. Brusilovskly and D.D. Hart
Presented by Donald Charles
Patrick Center for Environmental Research
The Academy of Natural Sciences
1900 Benjamin Franklin Parkway
Philadelphia, PA 19103
NOTES:
Repeated biological monitoring of particular water bodies provides a powerful
basis for determining whether ecological conditions are improving or
deteriorating. It is widely recognized that such biological assessments should
focus on many different components of the aquatic community. Some of the
traditional metrics used in bioassessment, however, can be relatively
insensitive to temporal changes in community characteristics that may in turn
reflect significant variations in environmental quality. Even if a community
exhibits a relatively stable level of species diversity, for example, it may be
undergoing marked variations in species composition. To supplement currant
methods for evaluating trends, we describe an approach for quantifying spatial
and temporal variations in community compositions based on similar analyses.
This approach uses Autosimilarity Analysis (a technique that evaluates
temporal variations in similarity at a single site) in conjunction with
Synchrosimilarity Analysis (a technique that evaluates temporal variations
in similarity between two sites) to evaluate community trends. We demonstrate
how this approach can quantify trends in community composition relative to
reference sites, and we illustrate the approach using a long-term ecological
database which describes variations in aquatic macroinvertebrates of the
Savannah River.
35
-------
Some Recent References on Use of Periphyton for Monitoring Rivers and Streams
Source: D. Charles, Academy of Natural Sciences of Philadelphia, 9/95
Whitton, B.A., E. Rott, and G. Friedrich. 1991. Use of algae for monitoring rivers. E. Rott,
Innsbruck, Austria. 193 pages.
Aloi, J.E. 1990. A critical review of recent freshwater periphyton field methods. Canadian
Journal of Fisheries and Aquatic Sciences, 47:656-670.
Bahls, L.L. 1993. Periphyton bioassessment methods for Montana streams. Water Quality
Bureau, Dept. of Health and Environmental Sciences, Helena, MT.
Lynch, R.A. 199?. Development of rapid bioassessment protocols for Oklahoma utilizing
characteristics of the diatom community. Oklahoma Conservation Commission, Oklahoma
City, OK. 106 pages.
McCormick, P.V. and J. Cairns. 1994. Algae as indicators of environmental change. Journal of
Applied Phycology, 6:509-526.
Mills, M.R., G. Beck, J. Brumley, SJvl. Call, J. Grubbs, R. Houp, L. Metzmeier, and K. Smathers.
1993. Methods for assessing biological integrity of surface waters. Kentucky Department for
Environmental Protection, Division of Water, Water Quality Branch, Ecological Support
Section, Frankfort, KY. 139 pages (includes chapter on algae).
Porter, S.D., T.F. Cuffney, M.E. Gurtz, and M.R. Meador. 1993. Methods for collecting algal
samples as part of the national water quality assessment program. Open-File ReDort 93-409
U.S. Geological Survey, Raleigh, NC. 39 pages.
Retd, M.A., J.C. Tibby, D. Penny, and P.A. Gell. 1995. The use of diatoms to assess past and
present water quality. Australian Journal of Ecology, 20:57-64.
Rosen, B.H. 1994. Use of periphyton in the development of biocriteria. Pages 209-215 In WS
Davis and T.P. Simon, editors. Biological Assessment and Criteria. Tools for Water Resource '
Planning and Decision Making. Lewis Publishers, Boca Raton.
*12
-------
INVERTEBRATE ASSEMBLAGES AND REGIONAL CLASSIFICATION:
THE INTERIOR COLUMBIA BASIN AS A CASE STUDY IN SCALING UP
Judith Li
Department of Fisheries and Wildlife
Oregon State University
Corvallis, OR 97331
NOTES-.
We chose five assemblage measures to capture essential qualities of
macroinvertebrate populations of the Inner Columbia River Basin. Taxa
richness, EPT taxa, EPT/Chironomids, total abundance, and percentage
represented by the most dominant taxa, were used for our multivariate
analyses. Despite the variations in methodology and other limitations of
survey data made available to us, the database was extensive and suitable for
comparisons of these metrics. Principal components analysis proved to be a
useful tool in making distinctions between stream assemblages in the Blue
Mountain, Eastern Cascade and Columbia Plateau ecoregions. Comparisons of
taxa among streams and ecoregions were made by combining summary taxa
tables and DECORANA analyses of the ten most dominant taxa for each
stream. Surprisingly few taxa were ubiquitous, and a limited number of taxa
were identified as characteristic of particular ecoregions. Whereas few
distinctions between streams in the High Desert ecoregion of southern Oregon
and Idaho were apparent using assemblage metrics, this was the region where
most unique taxa were found. These combined techniques provided a
preliminary template of assemblage distributions scaled to an expansive inter-
regional landbase, readily modified by new surveys and more detailed
information.
37
-------
I/O COM PtrtrTB Benthic Invertebrates
' Source: "A: Survey of Eastside Ecosystem Benthic Invertebrates" J. Li, K. Wright, and J. Furnish 1995.
-------
II T^R-AP ~T . Benthic Invertebrates
' Source: TV Survey of Easts/de Ecosystem Benthic Invertebrates" J. LI, K. Wright, and J. Furnish 1995.
LEGEND
/V State Boundaries
A/ Ecological Rep Units
* Quality N/A
A Riverine
* Low Diversity, Flashy
* Low Diversity,
High Desert
* Moderate Diversity
* High Diversity,
Cool Habitat
NOKI
s
ICBEMP 1995
soimiraflf cascades *
*
-------
Eastern
Cascades
Columbia
Plateau
Northern
Rockies
Blue
Mountains
High
Desert
AVERAGE ASSEMBLAGE MEASURES
Based on Principal Components Analysis
Total Total Taxa EPT
Number/m2 per stream
6707 37 18
2881 23 11
5635 25 13
4236 42 23
5936 26 10
ETP/C % Dominant HBI
Taxa
8.5 39.0 3.6
5.5 43.7 2.7
36.0 4.3
5.4 28.4 3.9
5.5 43.4 5.0
EPT:Ephemeroptera,Plecoptera & Trichoptera; C:Chironomidae
HBI:Hilsenhoff Biotic Index
-------
r
Map of stream clusters for the Blue Mountains ecoregion.
1 = Low Quality
2 = Best Quality
3 = Averages
4 = Good Quality
Washington
Idaho
t
-------
Clusters of streams plotted from PCA analysis within the Blue Mountains ecoregion.
Domtax -.829, Taxa .809, EPT .662
-------
REORGANIZED ASSEMBLAGE CLASSIFICATIONS
REGION TAXA TOTAL HBI EPT/C EPT DOMTAX
RIVERINE
COLUMBIA PLATEAU 6 49 2 70.1
LOW DIVERSITY,FLASHY
EASTERN CASCADES 22 8422 4.88 2.33 8 67.79
COLUMBIA PLATEAU 23 5892 8 53.2
LOW DIVERSITY,HIGH DESERT OR DISTURBED
BLUE MOUNTAINS 23 4688 2.4 8 45.0
COLUMBIA PLATEAU 22 1879 11 43.8
NORTHERN ROCKIES 20 1404 4.13 11 39.1
HIGH DESERT 25 4.76 10 37.4
COLUMBIA PLATEAU 31 7968 12 33.7
MODERA TE DIVERSITY, GOOD HABITA T
EASTERN CASCADES 38 4433 3.85 3.92 17 36.3
NORTHERN ROCKIES 38 17790 4.57 21 32.8
BLUE MOUNTAINS 42 4314 5.35 23 29.0
BLUE MOUNTAINS 41 5850 9.1 22 28.4
HIGH DIVERSITY, COOL HABITAT
EASTERN CASCADES 38 3915 1.9 20.3 24 27.8
COLUMBIA PLATEAU 32 652 21 24.2
EASTERN CASCADES 54 10561 3.77 2.12 26 23.5
BLUE MOUNTAINS 59 5941 3.44 30 18.2
-------
Unique Taxa in Each Ecoregion
Based on ten most common taxa per stream
Eastern Columbia Northern Blue High
Rockies Plateau Rockies Mountains Desert
Number Taxa 64 47 60 60 117
per Ecoregion
Number Taxa 14 4 11 4 35
Unique to Each
Ecoregion
Percent of Basinwide 23 10 18 7 30
Unique Taxa
-------
WIDESPREAD and COMMON TAXA
TAXA Entire Blue Northern Columbia
Basin Mountains Rockies Plateau
Baetis
Chironomidae
Cinygmula
Rhyacophila
Hydropsyche
Optioservus
•k
*
•k *
* *
k "k
k •k
Widespread# among top 10 taxa in more than 1/2 streams within a region
CHARACTERISTIC TAXA
Watershed
Yoraperla
Cinygmula
Paraleptophlebia
Eastern Cascades: Crater Lake
Columbia Plateau: Disturbed streams
Drunella
Seratella
Epeorus
Rithrogena
Blue Mountains: Grande Ronde
Blue Mountains: High elevation streams
Eastern Cascades: High elevation streams
-------
Figure 2. Taxa abundance curves for two streams in western Oregon, illustrating log series abundance. Each symbol
denotes abundance of a particular taxon. Arrow is drawn to show percent of total abundance represented by 10 taxa.
to
o
0
10
20 30 40 50
Rank of Individual Taxa
60
70
-------
BIOASSESSMENT USING MACROINVERTEBRATES
IN LOW-GRADIENT REACHES OF THE SOUTH PLATTE RIVER
fill Minter
Department of Earth Resources
Colorado State University
Fort Collins, CO 80523
NOTES:
Biomonitoring in low-gradient rivers with shifting-sand substrates has not been
fully addressed. This research examined the composition and variability of
benthic communities from substrates representing three major habitats in low-
gradient reaches of the South Platte River in Colorado and Nebraska. Benthic
macroinvertebrates were collected during the summer season, 1993, from
shifting-sand, submerged woody debris, and rock or rock-like material at six
stations representing sixth and seventh order streams. This research also
examined the effects of nutrients and habitat quality on those benthic
communities and whether a particular substrate or combination of substrates is
more appropriate for distinguishing between stations of differing water
quality.
Fifty-seven taxa were collected from all substrates. Thirty-three taxa occurred
in the sand substrate and 47 taxa occurred on wood and rock substrates. The
proportion of major macroinvertebrate groups varied among substrate and
station. Chiromonids predominated sand substrates except for one station
impacted by wastewater discharge where Oligochaeta was the predominant
group. Wood substrates contained a larger proportion of other Diptera,
mayflies (Ephemeroptera) and caddisflies (Trichoptera) and contained veiy
few Oligochaeta. The major groups found on rock substrates were similar to
wood substrates, but there was a larger proportion of mayflies and fewer
chironomids at reference stations. The number of taxa per replicate was less
variable than macroinvertebrate abundance for all substrates and stations. In
general, rock substrates were less variable than sand or wood, but clear patterns
were not observed. Combining all six stations, 8-40 sand samples would be
necessary to estimate number of taxa ± 10% of the mean compared to 1-16 wood
samples and 1-14 rock samples.
Twelve taxa accounted for 91.0% to 99.5% of all organisms collected from any
single replicate. Taxa examined individually from wood and rock substrates
showed similar changes in patterns of abundance along a longitudinal gradient,
and were dissimilar to the sand substrate. Canonical discriminant analysis,
based on the abundance of the 12 dominant taxa, was used to examine the
overlap and separation of stations for each substrate. Replicates from each
station grouped together and were clearly separated from other stations. The
dominant taxa responsible for this separation varied by substrate. There was a
consistent pattern for all three substrates of the three less impacted stations
being grouped together and separated from the three more impacted stations.
Correlations were conducted between macroinvertebrate communities,
represented by canonical axes 1 and 2, nutrient levels, and measures of habitat
quality. High correlations <0.77 to 0.96) were found between macroinvertebrate
communities and levels of ammonia, nitrite and nitrate, and phosphorous as
well as habitat quality (e.g. riparian and bank structure). High correlations
varied by substrate and may indicate those abiotic factors responsible for
38
-------
shaping macroinvertebrate communities. The results of this research suggest
that the choice of substrate or combination of substrates for biomonitoring in
low-gradient streams may be dependent upon which abiotic factors are of most
interest.
39
-------
DEVELOPING BIOLOGICAL CRITERIA FOR
PACIFIC NORTHWEST STREAMS
Leska Fore and James Karr
Institute for Environmental Studies
Box 352200
University of Washington
Seattle, WA 98195
NOTES:
The goal of biological monitoring is to evaluate the effect of human activities
on biological resources. In this study, we linked human activities across
landscapes to specific changes in assemblages of benthic macroinvertebrates in
streams that drain those landscapes. We used data from second to fourth-order
streams in southwestern Oregon to test approximately 30 hypotheses about how
macroinvertebrates respond to several common human actions, especially
logging and associated road construction. We found 11 attributes of
macroinvertebrate assemblages (or metrics) to be reliable indicators of
degradation. We outline methods for selecting reliable attributes of the
assemblage to monitor changes in water-resource quality.
We constructed a multimetric index (benthic index of biological integrity, or B-
IBI) from metrics that distinguished degraded stream sites from minimally
degraded sites. Using an independent data set, B-IBI scores were significantly
lower for streams whose watersheds were more degraded by human activities.
We also tested rapid bioassessment protocol (RBP) III as modified by Oregon
Department of Environmental Quality. RBP III failed to detect differences
among sites to which B-IBI responded. In a comparison of multivariate and
multimetric approaches to explore pattern in multidimensional data, we found
that principal components analysis responded primarily to the absence of
species—that is, zeros in the data matrix.
The collection of data for biological monitoring is not a goal in itself; data
should be collected to answer specific questions about the management of
environmental resources. Simple statistical tests are appropriate in some
situations, such as when the goal is to regulate a polluter by assessing the sites
upstream and downstream of a point source. But when a biologist or statistician
reports a significant difference, an alert audience will ask, "How different?"
and, "In what?" In a monitoring context, we are less interested in testing for
statistical significance than we are in comparisons, but they also provide a
yardstick that can be used to rank sites, for example, to determine which sites
are the best candidates for restoration.
40
-------
Draft manuscript submitted to
Journal of the North American Benthological Society
July 20,1995
A Benthic Index of Biotic Integrity for
Streams in the Pacific Northwest
Leska S. Fore and James R. Kan-
Institute for Environmental Studies, Box 352200
University of Washington
Seattle, WA 98195
Robert W. Wisseman
Aquatic Biology Associates
3490 NW Deer Run Rd.
Corvallis, OR 97330
41
-------
Introduction
Ik. PaciSc Nonbwest is on the hint of losing its salmon ^ 19Mi ^ ,
salmon fishery was closed along the West Coast ten Mexico » Washing™ because of low returns of spawning
*M* Th' ^"Saim0n and^ fisbeiies(TO™, Society I993) represetltsM
annual economic loss in excess of $1 billion and 60,000jobs (Pacific Rivers Council 1995). Ibis (rend is simply die
most recent chapter inahislory of decline and loss of regional aquatic resources thai spans die continent (Huglies and
Noss 1992, Allan and Flecker 1993) -Adandc salmon C^tofiaaJat) of New England rivers, spon and ~
species in the Midwest (Kan- et al 1985), mi native fishes of California (Moyle and Williams 1990), Salmon and
aquatic organisms are disappearing because dams, timber harvest agriculture, urbanization, and other human
actions alter watersheds and degrade rivers that support populations of aquatic species.
Widespread recognition of this degradation has stimulated numerous efforts to improve our ability to track
the condition of aquatic ecosystems (Davis and Simon 1995). Comprehensive, multimetric indexes (Barbour et al.
1995) were first developed in the Midwest for use with fishes (Kair 1981. Fausch et al. 1984, Kair et al. 1986) and
modified for use with invertebrates (Ohio EPA 1988, Piafkin et al. 1989, Kerans and Kair 1994, DeShon 1995).
Although the original index of biotic integrity (IBI) has been adapted for fish in the Willamette River (a large river in
Oregon - Hughes and Gammon 1987), broad application of a fish IBI to the Pacific Northwest is restricted by three
problems: 1) streams in the Pacific Northwest have few species of fish; 2) many fish spend a significant portion of
their life cycle in marine waters; and 3) salmonids, the dominant taxa in most streams, are affected by harvesting and
stocking.
In contrast, streams west of the Cascades host several hundred easily identifiable taxa of macroinvertebrates
that complete their life cycle in freshwater. The variety of taxa and knowledge of their natural history makes it
possible to distinguish different types and levels of human-induced degradation. Thus, we evaluate the biotic integrity
of streams based on analysis of the benthic invertebrate assemblage.
In this paper, we test and modify the benthic IBI (B-IBI) developed for the Tennessee Valley (Kerans and
Kair 1994). We test hypotheses about invertebrate responses to degradation and construct an index with reliable
statistical properties. Further, we test the resultant B-IBI on other data sets and compare B-IBI with a similar index.
Rapid Bioassessment Protocol in (Piafkin et al. 1989) as modified by Oregon Department of Environmental Quality
(DEQ) (Mulvey et al. 1992). We also compare the multimetric B-IBI with multivariate approaches to biotic
assessment
Methods
Study location
As is typical of many monitoring studies, our analysis of existing data sets was retrospective. Of the data
available from Alaska, Idaho, Oregon, and Washington (EPA's Region 10), we chose streams from southwestern
Oregon for initial studies (Table 1) because a large number of sites had been sampled over three years and land-use
42
-------
data were sane of the best available. Macroinvertebrates were collected from mid-order streams within the drainages
of the Umpqua and Rogue rivers in the western foothills of the Cascade Mountains south of Eugene, Oregon.
Watershed area ranged from 15 to 400 km2 with a median area of 54 km2. Original vegetation was primarily Douglas
Fir fPsmirintsiMra mermesii).
Our study area was located in the Umpqua National Forest and in the Roseburg and Medford Districts of the
Bureau of Management, the latter in a checkerboard pattern with private owners. Within the National Forest,
clearcuts have increased the frequency and intensity of runoff which has simplified channel morphology and
increased debris torrents (USFS 1992). Logging and road-building have increased sediment and significantly
widened stream channels since 1937 (Dose and Roper 1994). Watersheds under BLM and private ownership are
mnriifiwi by timber harvest as well as grazing, agriculture, and urbanization.
Data sets
An ideal data set for testing metrics would include sites that were physically identical and differed only in
their degree of human influence. Data available for the Pacific Northwest were not ideal, and so we had to first
identify a subset of the data that included a large enough number of physically similar sites. We controlled for the
effect of large-scale geographic features by selecting sites within southwest Oregon. We minimized seasonal effects
with fall sampling and for year effects by testing within a single year. We selected second and third-order streams and
excluded sites from the Umpqua River because the mainstem is a much larger river system with increased taxa
richness, different species composition, and markedly different biological dynamics. We divided the data into subsets
for each phase of index development according to year and location.
Metric testing -1990 Umpqua - We tested hypotheses about benthic invertebrate responses to timber harvest
and road building with data from 45 sites collected in 1990 from the Umpqua National Forest.
Metric validation -1992 Umpqua - We validated patterns observed in 1990 metric scores with data from 24
Umpqua sites collected in 1992.
Index testing -1991 Umpqua and Roseburg - We compared B-IBI scores of 20 Umpqua sites with 30 site
scores from the Roseburg area. The different levels of human activity in these two areas made them
iflffal for a comparison of B-IBI and RBR
Data collection
Tnvwtfthfates- Macroinvertebrates were collected with a 500 micron kicknet. All individuals collected were
identified and counted, unless the sample was greater than 1000 individuals, in which case a subsample was
processed. Insects were in the benthic assemblages, and were usually identified to genus or species. Level
of taxonomic identifications for non-insects varied from species to phyla. Invertebrates were collected in the fall
because the largest number of species are typically present at this time of year and are, in general, as large as they get
after feeding all summer.
A single in watershed yielded one collection of invertebrates for each stream. We analyzed
43
-------
riffle data for this study for three reasons: I) riffles are easier to define operationally than pools or margins; (2) riffles
axe more uniform than other stream microenvironments; and (3) riffles have high current velocity and shallow depth,
facilitating sampling with kicknets or Surber samplers. Pools provide information that riffles do not and there is some
evidence that they are degraded by sediment before riffles (Kerans et al. 1992); however, our priority was to develop
a sampling protocol that could be easily applied by others.
Land use. Elevation, location of roads and urban areas, location of private and federal land, and stream size
were obtained from USGS topographic maps. Estimates of total watershed area, watershed area logged, and 0f
road within each watershed were provided by the U. S. Forest Service (USFS) for Umpqua sites. Unfortunately,
information about the type, location and date of timber harvest was not available. For instream condition of sites
within the Umpqua NF, we asked USFS hydrologist Mikeal Jones to rank streams based on observations of the
riparian corridor, scream bed, bank stability, and the influence of road building and logging on the stream channel. No
water chemistry ffam were available. For BLM sites west of the National Forest, specific information was available
for only a few watersheds.
Data analysis
Our goal was to extract relevant biological pattern from taxa lists and counts of macroinvertebrates and
relate those patterns to human-induced changes in the watersheds. We evaluated data sets with a multimetric
approach (Kair et al. 1986, Kerans and Karr 1994, Barbour et al. 1995) and a multivariate approach (principal
components analysis) (Norris and George 1993) to identify patterns in macroinvertebrate assemblages.
index development. For the multimetric analysis, metrics started out as hypotheses about how invertebrates
respond to disturbance. The hypotheses we tested came from published literature (Cummins et al. 1989, Karr and
Kerans 1991), existing protocols and multimetric indexes (Plafkin et al. 1989, Hayslip 1993, Kerans and Karr 1994,
DeShon 1995), and our £eld experience. We grouped sites based on instream condition, level of logging, and other
important human activities into three groups ~ least, moderately, and most degraded. Instead of statistical correlation
we used graphical analysis to judge candidate metrics. Metrics that showed little or no overlap between most
degraded and least degraded sites and bad moderately degraded sites arrayed mostly in between (Figure 1) were
and tested again with data from a subsequent year.
Metrics from three general classes: taxonomic richness and composition, tolerance and intolerance, and
population attributes were selected and combined into an index by transforming the metric values to a score of 5
(similar to or deviated slightly from undisturbed condition), 3 (moderate deviation), or 1 (strong deviation from
undisturbed condition) (Karr 1981,1991). Hie sum of the metric scores gave an index score for each site. We tested
B-EBI by comparing index scores for sites from two areas (Umpqua NF and Roseburg BLM) experiencing different
levels of human influence.
We tested the Oregon DEQ version of the RBP because it resolves some of the problems of die original RBP
m. Ratio metrics in the original RBP, such as scrapers/filterers and EPT/chironomids, are replaced with metrics such
44
-------
as percentage scrapers and percentage filterers. The original RBP calculated scoring criteria as a percentage of a
reference site which we could not do because reference sites were not sampled, hi general, setting scoring criteria as
a percentage of a reference site is a poor approach because it fails to recognize that reference condition is more
meaningful when defined as a range of values rather than as a single value. We used the scoring criteria developed by
Oregon DEQ (Mulvey et al. 1992).
Statistrfl] considerations- We could have used a statistical test (e.g., Mann-Whitney test for two independent
samples) to decide if the most and least degraded sites were significantly different Instead, we based our decisions on
graphical analysis because it provided more insight into the biology than a simple p-value could. For example, we
were interested in whether scores for best and worst sites clumped together tightly or were spread more evenly.
Graphs allowed us to examine each metric's range and evaluate where along the continuum of degradation the metric
was most sensitive.
An alternative approach might base metric selection on statistical correlations with water chemistry, land
use, or similar variables. The main problem with statistical correlation is that no single variable can summarize all the
human activities that degrade streams and provide a linear ranking of sites. Watersheds may be degraded by mining,
in which case water chemistry data are appropriate; or by channel modification by dams or roads, in which case
physical data are relevant; or by hatcheries and stocking, in which cases fish density may be an appropriate correlate
for human-induced degradation. In our study area, although logging was the primary land use, the percentage area of
the watershed logged did not differentiate between sites that had been clearcut to the river bank and those sites that
had left riparian areas intact Another limitation was thai the year of the cut and the frequency of cutting were also
unavailable.
We doubt that any single land-use or chemical variable can reliably rank sites, except in unusual
circumstances; indeed, that is a primary reason for evaluating river condition according to the resident biology. We
relied on our knowledge of freshwater systems to judge which human actions were most important and destructive.
We are confident that sites that appear to a human observer to be heavily used and visibly damaged really are more
degraded than sites that appear pristine, except, perhaps, in the case of chemical contamination. Along a continuum of
degradation, we used the extreme sites to test metrics. Reliable metrics were then combined into an index which can
be used to sort out the middle, or moderately degraded sites.
Principal eomnonCTits analysis. For the multivariate analysis, land-use data were not sufficient to perform a
canonical correspondence analysis; therefore, we chose principal components analysis of the macioinvertebrate data
and identified important site characteristics afterwards on the principal component plots. Principal components
analysis calculates the line ("the component") that extracts the maximum amount of statistical variance from a cloud
of points (Tabachnick and Fidell 1989). F*gh point in the cloud represents a stream site. The number of dimensions
through which the line passes is equal to the number of taxa collected. The typical application of this procedure to
monitoring data uses species lists and abundances to interpret differences between stream sites (Nonis and Georges
1993). PC A was based on 46 sites from the 1990 Umpqua National Forest data, the same data set used to test metrics.
45
-------
Multivariate analyses are founded on the assumption that data are distributed normally; abundance data (as
we used for this PCA) typically are not. Two problems in particular are, first, the presence of numerous zeros "M.
second, frequent high abundances, or right-skewed data. Abundance data are typically transformed by normalizing
(i.e., subtracting the mean and dividing by the standard deviation) or by taking the logarithm. Neither approach was
very satisfactory because neither transformation resulted in normally distributed variables. We performed PCA for
three versions of the same data: normalized, log transformation, and no transformation.
Results
Our first task was to select a subset of physically similar sites to test metrics. Next, we divided sites into
three groups according to the intensity of land usd in their watersheds. Careful attention to outliers in the land-use
^afa led to corrections and new hypotheses. We tested metrics by evaluating their response along a gradient of
the dominant human influence, with special emphasis on their ability to clearly distinguish the most and least
degraded sites. Next, we constructed B-IBI from component metrics and tested B-IBI with a different data set.
Finally, we tested the RBP index and analyzed the invertebrate data with PCA.
B-IBI development
riassifving sites and evaluating human influence. Our goal in classifying sites was to obtain a subset of sites
that were physically similar and did not have human influence confounded with other physical attributes. For
example, high elevation sites were logged less (Figure 2) and had fewer roads in their watershed; thus, high elevation
sites were biased toward more pristine conditions. We excluded high elevation sites from the analysis and chose sites
below 2500 ft because they formed a cohesive group and were spread more evenly across the range of human
influence. High elevation sites also had much higher taxa richness; but we could not determine if high taxa richness
was due to pristine conditions or elevation.
Percent area logged measured human influence at the watershed scale; instream condition measured human
influence at the riparian scale. We expected measurements at two different scales to roughly agree for these sites
because the primary human activity (logging) affected both the watershed and the instream channel (Figure 3). Given
the many sources of variability and error in land-use variables, plotting land-use variables against each other provided
a check for data consistency. Unexpected outliers pointed us toward data errors and new hypotheses to be tested. For
example, two sites with low harvested area but very poor instream condition made us aware of several watersheds for
which timber harvest was underestimated. Harvested area was underestimated because estimates were only for public
aPH did not inflate logging on private land within the watershed. We corrected these points by estimating that
the proportion of area harvested on private land was the same as on forest land, probably an underestimate. Sites with
the best instream condition had been logged, but had been left to recover (upper fortes of Cow Creek) or had been
logged from ridgeline roads rather than roads along the valley floor (Quarn Creek).
For 1990, we selected four sites (Calf Creek, South and East Forks of Cow Creek, and Quartz Creek) as the
least degraded and four sites (Brownie and Tom Creek, lower Cow Creek, upper and lower Elk) as the most degraded.
46
-------
The 1992 data set was smaller and three best sites (upper Cow Creek, N. Boulder Creek, and Quartz Creek) and two
worst sites (Elk Creek, lower Jackson Creek) were selected to determine if results were consistent with the 1990
metric analysis. Least degraded sites were not pristine: all watersheds have been logged; rather we chose sites that
were the best available. Best sites did not have roads paralleling the mainstem channel, had been logged least or
longest ago, and were not easily accessible to humans. The most degraded sites in this data set were affected by roads
along the bank of the stream, heavy logging, grazing and urbanization.
Decline in taxa richness is one Of the most reliable indicators of degradation (Ford 1989, Barbour et al.
1995) for diverse taxa: periphyton (Bahls 1993); phytoplankton (Schelske 1984); riverine fish (Kan 1981, Miller et
al. 1988, Ohio EPA 1988, Lyons 1992, Lyons etal. 1995); estuarine fish (Deegan et al. 1993); and invertebrates (Ohio
EPA 1988, Kerans and Karr 1994, DeShon 1995). Therefore, we used taxa richness to test whether road density,
presence/absence of roads in the channel, or percent area logged most accurately reflected the impact of human
activities on the biota. Taxa richness declined significantly (Spearman's rho, p < 0.05) as area logged increased
(Figure 4), but did not show a clear response to road density or presence. We were initially concerned that most of the
sites were similarly degraded and, therefore, might not provide a broad enough range of degradation to test metrics.
Significant correlation between area logged and taxa richness in spite of the inaccuracy of the land-use data gave us
confidence that the sites were sampled from a gradient of human-induced change, and that the best and worst sites
were truly different. Because our analytic was limited to existing data sets, our sites probably do not include the true
extremes.
Testing metrics. Our goal was to construct an index with metrics from each of four broad classes: taxa
richness and composition, tolerant and intolerant taxa, feeding ecology, and population attributes. We selected
metrics that best distinguished most and least degraded sites for 1990 and 1992 (Table 2). Only metrics that worked
for both 1990 and 1992 were included in B-IBI (Table 3). Because 1992 was a drought year, we had greater
confidence that metrics selected were minimally influenced by this source of natural variability.
Taxa richness and composition. Total taxa richness and richnesses of Ephemeroptera, Plecoptera, and
Tricfaoptera easily separated the best sites from the poor sites, and, in general, declined monotonically across the
range of degradation (See figure 5 for examples). Plecoptera taxa richness declined first in response to human-
induced changes, Ephemeroptera next, and Trichoptera last The presence of PtCTOHarevs indicated better stream
condition. Pteronarcvs. a large, long-lived plecopteran, seems susceptible to scour during high flows (R. Wisseman,
pers. observ). Neither dipteran nor chironomid taxa richnesses responded reliably to human-induced changes.
Tolerant and intolerant metrics. For tolerant and intolerant metrics, we identified the most tolerant (about 35
out of 200 total taxa) and extremely intolerant taxa (about 35 taxa). We also identified taxa that were specifically
tolerant or intolerant to fine sediment which included some of the taxa in the general tolerant/intolerant metrics but
others as well. Our goal is to eventually replace the general metrics with more specific metrics that identify what the
taxa are tolerant to or intolerant of, for example, toxic contamination, increased temperatures, or scour.
Typically, tolerant metrics are calculated as a percentage of the total number of individuals in a sample;
47
-------
intolerant metrics are calculated as taxa richness. Tolerant organisms occur at all sites but tolerant taxa tend to
dominate as conditions are degraded. Intolerant organisms, when present, are typically a small percentage of the total.
Accurate estimates of relative abundance for intolerant taxa are virtually impossible to obtain with a reasonable
sampling effort; yet, the presence of these taxa sends a strong signal about local environmental quality. We predicted,
therefore, that sediment tolerant organisms scored as a percentage would perform better than as a taxa richness metric
and, furthermore, that the opposite would be true for sediment intolerant organisms. Empirical results
theoretical expectations based on previous studies: tolerants performed better calculated as percentages of total
abundance and in tolerants performed better as taxa richness.
Feeding ecology. We were surprised to find that feeding ecology metrics (Merritt and Cummins 1984)
calculated as either percentage of individuals or taxa richness, failed to distinguish the most and least degraded sites
(See Figure 5 for examples). No doubt human disturbance affects the trophic composition of the resident biotic
assemblage; we suspect the inability to detect a change was caused by the plasticity of many organisms as they
develop from nymphs to adults, thus making it difficult to assign taxa to a single group. Percentage of scrapers in the
sample distinguished good from poor sites in both 1990 and 1992, but the relationship was reversed: in 1990 the best
sites had a smaller percentage of scrapers, but in 1992 the best sites had a higher percentage of scrapers. The year
1992 was unusually dry suggesting that temporal variation in environmental conditions may affect scrapers and
possibly the relative abundances of other trophic groups as well.
Population attributes. High abundance relative to other species was tested as percentage of the most
abundant taxa in the total sample for one through five most numerically abundant taxa. The best form for this metric
calculated dominance as the relative abundance (percentage) of the three most abundant taxa. Total abundance did not
satisfy our criteria for selection; nonetheless, we included it because we expect very low abundances only in the most
extreme situations, such as when high levels of urban or industrial contaminants are present (unpublished data from
urban streams, Karr 1981). Very high abundances may also indicate degradation, particularly in agricultural areas
where nutrient enrichment affects invertebrates. In general, abundance (or population size or density) is one of the
single most variable aspects of an assemblage; thus, all other metrics were calculated as a percentage of total
abundance (relative abundance) or as taxa richness.
Constructing B-IBI. Metrics differed in their range of values; therefore, we transformed them to a similar
scale before combining them into a summary index, B-IBI. Following the method developed for the fish IBI (Karr
1981, Karr et al. 1986), we divided the range of each metric into three parts and assigned a new score of 5 (indicates
similar to expected, or reference, condition), 3, or 1 (indicates strong deviation from expected condition) (Table 3,
Figure 6). Metrics were added to get a final B-IBI score (maximum = 55; minimum = 11). We determined the
potential range of metrics based on 82 sites: 34 from the Umpqua NF (1990), 30 from the Rosebug BLM (1991), and
18 from the Medford BLM (1991).
Simple, general rules for setting metric scoring criteria are difficult to define because they depend on the
original sampling design that generated the data if sites from a region are sampled such that one-third are located in
48
-------
pristine areas, a third in moderately degraded areas, and a third in very degraded areas, then one can start by dividing
the range of each metric at the 33rd and 67th percentiles. In our case, we suspectedrhaf our data sets included neither
very degraded sites, e.g., sites from urban areas, nor pristine sites with no human influence. Therefore, we tried to err
on the conservative side by broadening the middle category (3) to include more sites, thus making it more difficult for
a site to score a 5 or 1. Natural breakpoints in the distribution probably reflect relevant biology. Where these occurred,
we adjusted the scoring criteria to fall at these points. Setting scoring criteria is an iterative process and should be
revisited as more regional data become available across a broader range of human influences.
Part of the data (30 Roseburg sites) used to set scoring criteria were used to test the B-IBI. We do not expect
their inclusion to bias our test of the index because they were only used to assess the range of metrics. For most
metrics, very low and very high scores were found in the Umpqua NF, Roseburg, and Medford BLM areas. It was not
the case that Roseburg sites scored 1 for most metrics and Umpqua scored S; rather, average metric scores for the
Umpqua were higher than for Roseburg, which provided another check of the metrics.
Testing B-IBI
We tested the hypothesis that B-IBI scores for Umpqua stream sites would be higher, reflecting higher biodc
integrity, than for Roseburg sites. Although the range of B-IBI scores was similar for the two areas, the poorest sites
in Roseburg were more degraded than the poorest sites in the Umpqua. The Roseburg area includes BLM and private
land in a checkerboard pattern. Private land, in general, has been logged more heavily and has been degraded further
by urbanization, roads, mining and agriculture. A significant difference between index scores for the two areas (p <
0.05; Figure 7) indicated that B-IBI reflects the differing effects of human-induced degradation on instream biota in
the two study regions.
Testing RBP
The RBP index as modified by Oregon DEQ (Table 4) failed to detect a difference between sites in Roseburg
BLM and Umpqua NF (p > 0.05; Figure 7). RBP metrics represent a set of working hypotheses that must be tested
(like all metrics) with data from the region to be monitored. RBP metrics have not been tested in the Northwest (and
only minimally tested elsewhere). We tested the rnmpniwnt RBP metrics with the same criteria described above and
found that only four of the metrics could rfitHngnith between most and least degraded sites (Table 4). Two of those
metrics were included in B-IBL
Multivariate Analysis
PCA results were similar for the three versions of the data (normalized, log transformed, and no
transformation). Sites identified as most and least degraded, large river, and high elevation showed some tendency to
group together for all three data versions of PCA when plotted against the first and second principal components;
however, the groups defined by the sites were not exclusive and showed considerable overlap (Figure 8). Sites that
49
-------
were outlieis from their groups in one analysis were also outliers in tbe other two analyses. "Hie most degraded sites
grouped together better than the least degraded sites for all three analyses.
We suspect that zeros in the data matrix had a strong effect oo tbe solution of the PCA {Tabacbnick and
Fidell 1989) for two reasons. Hm, data transfonnauon had little effect on the PCA outcome, probably became zero
abundances were not changed by transfonnadon. Second, degraded sites pooped together best, prcbably because
with the fewest taxa, they contained the largest number of zeros in the data matrix.
We looked tor patterns in the loadings of the 198 taxa on die principal components. We found a slight
tendency for Ephemeroptera. Plecoptera, and Trichoptra taxa to load more ftequemly on the -n-| n, which
separated most and least degraded sites (PCD. W: did not detect patterns in the loadings for intolerant and tolerant
taxa. Other taxa grouped together on principal component axes provided little additional insight into the structure of
the assemblage. Rather than trying to work through the staustical analysis and detetmine how and why species were
grouped together, we found it easier to compare sites and search for patterns in tbe raw data by simply ^
taxa were present or absent in similar sites. By focusing on biological patterns associated with specific, testable
hypotheses, we were better able to describe and detect the environmental condition of sites than we could from
multivariate plots.
Discussion
In this paper, we emphasize the process of testing metrics and developing an index. Even with a rather
narrow range of disturbance, we were able to test metrics and develop an index that responded to human-induced
degradation. We selected metrics according to their response to levels of human disturbance in the watershed,
primarily logging and road building. One may ask, if the biological condition of sites can be determined so easily,
why bother with biological monitoring? Although sites near the ends of the spectrum are easy to judge, moderately
degraded sites are not. TTius we started with the extremes, tested our hypotheses, and applied the results to sort the
middle. Judging the condition of streams based on the instream biota allows us to look for ^3iw differences. Large
scale differences that a human observer notices, such as large woody debris, may not be as important to the resident
biology as other aspects of the river that humans can not see or measure.
Too often biological patterns in complex data sets are reported in terms of complicated, high level statistics,
such as principal components analysis. In contrast, we demonstrated how graphical analysis and simple statistics can
be used to interpret complex data and understand biological patterns. To many field biologists, "statistics" ^
"multivariate statistics." Monitoring reports have been using the same multivariate techniques the 1960's in
spite of a plethora of statistical techniques available (Potvin and Travis 1993). Multivariate analyses make
decisions based on statistical properties of the data, e.g., tbe structure of the covariance matrix, rather than the
biological judgement of the investigator. Because most multivariate analyses are based on taxa lists and abundances
they fail to incorporate our biological knowledge of the animals' natural history, and especially their responses to
human activities. Multivariate analyses are most appropriate for exploratory analysis, that is, developing hypotheses
about systems for which little is known. We know enough about freshwater systems to test hypotheses rather than
50
-------
continue generating them; we must move forward with testing hypotheses directly rather than simply reporting
complexity.
The object of a good index is not to include and measure every aspect of stream biology that responds to
human influence; biological systems are too complicated to capture completely. The effects of human-induced
degradation on rivers and streams that biologists are being asked to measure in a monitoring context are not subtle.
Multimetric indexes formalize what any good biologist, familiar with local biota, knows about the biological
condition of a stream. Narrative and numeric indexes are essential tools for communicating biological condition to
policy makers and concerned citizens so that they can make informed decisions about the management of aquatic
resources.
Metric development
Metrics were raimiarpri as taxa richness or percentages depending on the attribute they were measuring. A
good biomonitoring index should not be sensitive to slight differences in rainfall or location, therefore, we are less
interested in which particular species of stonefly is found at a site because the identity of the species may depend on
subde microhabitat differences or broader biogeographic patterns. Similarly, species abundance responds
dramatically to small changes in the environment On the other hand, taxa richness of stone flies, for example, is
robust to natural variability. We measure how many stonefly taxa are present, because most stoneflies require large
cobble, and fast, clean, cool water.
Trophic composition metrics (e.g., shredders, scrapers, and predators), tolerance metrics, and dominance
(relatively most abundant taxa) metrics are scored as percentages rather than taxa richness because they evaluate
processes at the level of the assemblage. Taxa in these groups may be present at all sites, but their relative proportion
changes with degradation. We also trophic composition metrics as taxa richness metrics when they failed to
discriminate sites as percentages; ™»»fhw approach worked. We suspect that trophic composition is affected by human
disturbance and that these metrics did not respond because the feeding ecology of many taxa change as they develop
into adults.
Although ratio metrics have been included in other monitoring protocols, e.g., ratio of scrapers to filterers,
we recommend against rt*«n for two reasons. First, the numerator and denominator vary together; therefore, a pair of
very large values for scrapers and filterers may yield the same ratio value as a pair of very small values. We expect
that large values for these two groups may have a different biological meaning than small values. In short, if two
attributes of an assemblage are potentially important, they should be evaluated individually. Second, when two
variables are combined as a ratio, the ratio tends to have higher variance than either variable alone (Sokal and Rohlf
1981). Barbour et al. (1992) found ratio metrics for invertebrates to be more variable than other metrics. On the other
hand, percentages, or relative abundances, are based on the binomial distribution and are much more reliable.
Our goal is to replace the general tolerant and intolerant metrics with more specific metrics that describe
what the organisms are tolerant to; for example, the sediment tolerant and intolerant metrics developed for these Ham
51
-------
As appropriate data sets become available, we suspect another more inclusive metric will include Ptemnar^
with other taxa, such as freshwater clams, that are sensitive to scour because of their large adult size and time needed
to complete their life cycle.
"Indexes"
To most biologists, the word "index" carries with it semantic baggage that is the legacy of diversity indexes.
Multimetric indexes are often condemned along with diversity indexes (Suter 1993, Nonis and Georges 1993)
although they differ fundamentally in what they measure and how they are constructed (Fausch et al. 1990).
Biomonitoring indexes can be divided into three types: (1) diversity and similarity indexes; (2) pollution tolerant
indexes, or "biotic indexes"; and (3) multimetric indexes.
Diversity indexes and most similarity indexes combine information about abundance (or density) and
species richness into a measure of "evenness." The concept of evenness was developed in the context of information
theory, but has not been meaningfully interpreted in terms of ecological theory (Hurlburt 1971). Diversity indexes
sometimes respond to degradation because they are strongly correlated with taxa richness (Camargo 1992), an
attribute of the assemblage that, on its own, is easy to interpret in terms of biology. The response of these indexes to
systematic changes in the assemblage are often erratic, inconsistent, dependent on initial conditions, and can give
misleading interpretations of biological data (Wolda 1981, Boyle et al. 1990).
Pollution tolerance indexes, or biotic indexes, assign a pollution tolerance value to every species and
calculate an index score for a site as a function of the number of individuals of each tolerance class (Chutter 1972
Winget and Mangum 1979, Hilsenhoff 1982, Lenat 1993). These indexes estimate an average pollution tolerance for
the assemblage. In western streams, chemical and organic pollution resulting from agriculture and urban sewage
represent only part of the problem. Grazing, timber harvest, and road building affect streams and their resident biota
differently than organic pollution. A potential problem with this index format is that by calculating the tolerance of
every organism, even those that are not particularly sensitive, a strong response by a few taxa can be mi«^ jf ^
assemblage is dominated by a large number of insensitive taxa; that is, the signal may be lost in the noise. We avoided
this problem with tolerant and intolerant metrics by concentrating on the responses of only the most and least tolerant
organisms.
Each component metric of a multimetric index measures an attribute of the assemblage that is the product of
evolutionary and biogeographic processes at a site (Karr 1995). Metrics are interpreted (and assigned a score of 5,3,
or 1) based on metric values observed in pristine, or best available, sites. The n metrics represent a measures of the
biotic integrity of a stream. Variance is reduced and precision increased by taking an average (or sum) of several
measurements because variance decreases as a function of sample size. The functional form of multimetric indexes
allows us to take advantage of properties of the mean. The Central Limit Theorem (CasseBa and Berger 1990) "*"><
that averages tend to be distributed normally; consequently, multimetric index scores can be tested with familiar
statistics, such as ANOVA or regression (Fore et al. 1994).
52
-------
MM, M aructured more lite familiar economic indexes suet as to consumer price index ot
the index of leading indicate* (Mitchell and Bun-s 1938). These indexes Summarize tbe cunent condition or state of
die economy and reduce tbe inherent natural variability by including tke prices of several items or die cunent levels
of several indicators. As for multimetric indexes, economic indicators are comtaed into an overall index by
tbem to a standari year, after convesion, index components are summed. H*se composite indexes have
survived 60 years of discussion and criticism and are still widely used to intense. economic trends by economists.
policy makers, and citizens (Auerbach 1982).
Making decisions
The collection of da* for biological monitoring is not a goal in itself; da* should be collected to answer
specific questions about the management of environmental resources (Yoder 1994). Simple statistical tests are
appropriate in some situations, such as, wben tbe goal is to regulate a polluter by assessing sites upstream and
dOTrasuvam of a point source. But wben a biologist or statistician repons a significant difference, an aim audience
„ „ . -—ifi-j-nce is dependent on sample size: given a large enough
will ask. How different? and. In what way? Detecting significance is oepenue
, ro»t*Tm*n 1990). In a monitoring context, we are less interested in
sample size a difference can always be detected (Petennan lywj.
i* evaluating differences (Stewart-Oaten et al. 1992), and especially
testing for statistical significance than we are in evaluating
. . . k. used for statistical comparisons, but they also provide a
their biological relevance. Mul time trie indexes can be ....
yardstick tbai can be used to rank sites, for example. to determine which sites are tbe best candidates for restoration,
fcdex scores for shnilar sites within a region p«md« » »««> for im^ scores and idendfying trends.
The complexity capcuredby multivariate analyse, make them poor communication tools for an audience
j . . „n (arlA have been) based on multivariate statistical analysis of
composed of non-scientists. Management decisions can ( ....
, 7_,ii i qq-i Wright et al. 1993); but the decision process is
biological data (DaviesetaL 1993, Reynoldson and Zarull 1993, wngnt
opaque to anyone less than an expert in these types of analyses. Complicated statistical decision rules are not easily
adapted to a simple task like ranking streams from most to least degraded. Multimetric indexes condense and
summarize infoimation and can be used to compare sites over a lstge geographic area. On a smaller scale, tbe
componemmetrics are avails to make mo™ site-spedfic assessment such as,. pinpoint source of degradation
(Yoder and Rankin 1995b) or identify what aspects of the assemblage a restoration affects. Although a single number,
the index, is used to evaluate the relative condition of sites within a region, detailed information about individual sites
is not tost, but is recorded in the individual metrics (Simon and Lyons 1995).
The temptation (and tbe pressure) exists to establish biological criteria as soon as possible; but if an index is
con^ of biolo,^ and relWl. ^ ^ ^ ^ ^ ^ ^
u wo..c/» the component metrics were never adequately tested.
failed to detect differences in southwestern Oregon because
Fine-tuning B-1B1
Some of the steps involved in developing a biomonitoring index may need to be revisited as data
accumulate, but we expect many of die conclusions &om this study to hold true across the Pacific Northwest region.
53
-------
In particular, the metrics that responded reliably to the effects of logging will likely respond to other types of
degradation such as mining or agriculture. Whether the metric increases or decreases as human inflnmr<» increases
will probably also be consistent. New metrics may be tested and included in the index, but we expect the majority of
the metrics to be retained. The ranges for the scoring criteria will probably shift as we evaluate data from more
severely degraded streams (e.g., Puget Sound Basin). Our goal is for B-IBI to be composed of metrics that are
generally applicable for large regions.
A good multimetric index includes a balance of metrics that respond across the range of degradation.
Because our data sets were from moderately degraded sites, we were unable to specifically test the sensitivity of
metrics under extremely degraded conditions. Abundance and percent sediment tolerants were sensitive at the most
degraded end of the spectrum while taxa richness of Plecoptera and sediment intolerants were more sensitive for less
degraded sites. Total taxa richness has a more monotonic response across the range of disturbance. As new data are
evaluated, we intend to assess the redundancy of metrics and the overall balance of B-IBI (Kerans and Karr 1994).
By carefully selecting component metrics we can minimize the effects of natural geographic variability and
weather so that changes in the assemblage that result from human actions are revealed in sharp relief. Complete
separation of the effects of natural variability from human-induced variability is not realistic because disturbance
exacerbates the effects of natural events (Schlosser 1990); for example, flood events are more extreme in damaged
watersheds. Multimetric index scores reflect this reality and consequently are more variable for heavily degraded
sites (Karr et al. 1987, Steedman 1988, Fore et al. 1994, Yoder and Rankin 1995a).
Statistical variability of the index determines how precise a measurement tool it is for comparing sites and
making management decisions. State and federal agencies have not yet agreed on the number of individuals to be
identified from an invertebrate sample or the number of replicates needed from a stream site. We intend to evaluate
the statistical power of B-IBI for the different sampling protocols in order to determine how many classes of integrity
B-IBI can distinguish, as was done for the fish EBI (Fore et al. 1994). If we can reduce the amount of effort involved
in identifying samples without sacrificing reasonable precision, the money saved for each site assessment translates
into more streams visited each year.
Acknowledgments
We thank M. Jones and J. Dose, Umpqua Ranger District, L. Lindell, Medford BLM, and E. Rumble and S.
Hofford, Roseburg BLM for providing information on land use. M. Lagerioef (EPA, Region 10), G. Hayslip (EPA,
Region 10), R. Hafele (Oregon Department of Environmental Quality), R. Plomikoff (Washington Department of
Ecology), and D. Zaroban (Idaho Department of Environmental Quality) participated in many discussions that guided
the development of this manuscript. P. Larsen, R. O'Connor, and R. Hughes reviewed an early version of this
manuscript The research described in this article has been funded wholly or in part by the U.S. Environmental
Protection Agency under Cooperative Agreement #3B1053NAEX to JRK.
54
-------
Literature Cited
Allan, J. D., and A. S. Flecker. 1993. Biodiversity conservation in running waters. Bioscience 43:32-43.
Auerbach, A. J. 1982. The index of leading indicators: "Measurement without theory", thirty-five years later. Review
of Economics and Statistics 64(4):589-595.
Bahls, L. L. 1993. Periphyton Bioassessment Methods for Montana Screams. Water Quality Bureau, Department of
Health and Environmental Sciences, Helena, MN.
Barbour, M. T., J. L. Plafkin, B. P. Bradley, C. G. Graves, R. W. Wisseman. 1992. Evaluation of EPA's rapid
bioassessment benthic metrics: metric redundancy and variability among reference stream sites.
Environmental Toxicology and Chemistry 11:437-449.
Barbour, M. T., J. B. Stribling, and J. R. Karr. 1995. Multimetric approach for establishing biocriteria and measuring
biological condition. Pages 63-77 in W. S. Davis and T. P. Simon (editors). Biological and
criteria: tools for water resource planning and decision making. Lewis Publishing, Boca Raton.
Boyle, T. P., G. M. Smillie, J. C. Anderson, D. R. Beeson. 1990. A sensitivity analysis of nine diversity and seven
similarity indices. Research Journal of the Water Pollution Control Federation 62:749-762.
Camargo, J. A. 1992. New diversity index for assessing structural alterations in aquatic communities. Bulletin of
Environmental and Contaminant Toxicology 48:428-434.
Casella, G., and R. Li Berger. 1990. Statistical inference. Wadsworth, Inc., Belmont, CA.
Chutter, F. M. 1972. An empirical biotic index of the quality of water in South African streams and rivers. Water
Research 6:19-30.
Cummins, K. W., M. A. Wilzbach, D. M. Gates, J. B. Perry, and W. B. Taliaferro. 1989. Shredders and riparian
vegetation. Bioscience 39:24-30.
Davies, S. P., L. Tsomides, D. L. Courtemanch, and F. Drummond. 1993. Maine Biological Monitoring and
Biocriteria Development Program. Maine Department of Environmental Protection, Bureau of Water
Quality Control, Division of Environmental Evaluation and Lake Studies, Augusta, Maine.
Davis, W. S., and T. P. Simon (editors). 1995. Biological assessment and criteria: tools for water resource planning
and decision making. Lewis Publishing, Boca Raton.
DeShon, J. E. 1995. Development and application of the invertebrate community index (ICI). Pages 217-244 in W S
Davis and T. P. Simon (editors). Biological assessment and criteria: tools for water resource planning and
decision making. Lewis Publishing, Boca Raton.
Deegan, L. A., J. T. Finn, S. G. Ayvasian, and C. Ryder. 1993. Feasibility and Application of the Ttyjw 0f Biotic
Integrity to Massachusetts estuaries (EBI). Final Project Report to Massachusetts Executive Office of
Environmental Affairs, Department of Environmental Protection, North Grafton, MA.
Dose, J. J., and B. B. Roper. 1994. Long-term changes in low-flow channel widths within the South Umpqua
Watershed, Oregon. Water Resources Bulletin 30:993-1000.
Fausch, K. D., J. R. Kan, and P. R. Yarn. 1984. Regional application of an index of biotic integrity ha^ on stream
fish communities. Transactions of die American Fisheries Society 113:39-55.
Fausch, K. D., J. Lyons, J. R. Karr, and P. L. Angermeier. 1990. Fish communities as indicators of environment
degradation. American Fisheries Society Symposium 8:123-144.
55
-------
Ford, J. 1989. The effects of diemical stress on aquatic species composition and community structure. Pages 99-144
in S. A. Levm, M. A. Harwell, J. R. Kelly, K. D. Kimball (editors) Ecotoxicology: problems and approaches
Springer Verlag, New York, NY.
Fore, L. S., J. R. Kairand L. L. Conquest 1994. Statistical properties of an index of biotic integrity used to evaluate
water resources. Canadian Journal of Fisheries and Aquatic Sciences 51:1077-1087
Frissell, C. A. 1993. Topology of extinction and endangerment of native fishes in the Pacific Northwest and
California (USA). Conservation Biology 7(2):342-354.
Hayslip, G. A. 1993. EPA Region 10 in-stream biological handbook (for wadable streams in the Pacific Northwest)
EPA-910-9-92-013. U. S. Environmental Protection Agency-Region 10, Environmental Services Division
Seattle, Washington.
Hilsenhoff, W. L. 1982. Using a biotic index to evaluate water quality in streams. Department of Natural Resources,
Madison, Wisconsin. Technical Bulletin Number 132.
Hughes, R. M., and J. R. Gammon. 1987. Longitudinal changes in fish assemblages and water quality in the
Willamette River, Oregon. Transactions of the American Fisheries Society 116:196-209.
Hughes, R. M., and R. F. Noss. 1992. Biological diversity and biological integrity: current concerns for lakes and
streams. Fisheries 17(3): 11-19
Hurlburt, S. H. 1971. The nonconcept of species diversity: a critique and alternative parameters Ecology 52(4V577
586.
Kan-, J. R. 1981. Assessment of biotic integrity using fish communities. Fisheries 6(6):21-27.
Karr, J. R. 1991. Biological integrity: a long-neglected aspect of water resource management. Ecological
Applications 1:66-84.
Karr. J. R. 1995. Ecological integrity and ecological health are not the same. In P. Schulze (editor). Engineering
within ecological constraints. National Academy of Engineering. National Academy Press, Washington D
Karr, J. R., K. D. Fausch, P. L. Angermeier, P. R. Yant, and 1.J. Schlosser. 1986. Assessment of biological integrity in
running water a method and its rationale. Illinois Natural History Survey Special Publication, Number 5
Champaign, Illinois.
Kair, J. R., and B. L. Kerans. 1991. Components of biological integrity: their definition and use in development of an
invertebrate IBI. Pages 1-16 in W. S. Davis and T. P. Simon (editors). Proceedings of the 1991 Midwest
pollution control biologists meeting. EPA 905-R-92-003. U. S. Environmental Protection Agency,
Environmental Sciences Division, Chicago, THinni<
Karr, J. R., L. A. Toth, and D. R. Dudley. 1985. Fish communities of Midwestern rivers: a history of degradation.
Bioscience 35:90-95.
Karr. J. R., P. R. Yant, and K. D. Fausch. 1987. Spatial and temporal variability of the index of biotic integrity in three
midwestem streams. Transactions of the American Fisheries Society. 116:1-11.
Kerans, B. L., and J. R. Karr. 1994. A benthic index of biotic integrity (B-IBI) for rivers of the Tennessee Valley.
Ecological Applications 4(4):768-785.
Kerans, B. L., J. R. Karr, and S. A. Ahlstedt. 1992. Aquatic invertebrate assemblages: spatial and temporal
differences among sampling protocols. Journal of the North American Benthological Society 11:377-390.
Lenat, D. R. 1993. A biotic index for the southeastern U. S.: derivation and list of tolerance values, with criteria for
assigning water-quality ratings. Journal of the North American Benthological Society 12(3):279-290.
56
-------
Lyons, J. 1992. Using the index of biotic integrity (IB I) to measure environmental quality in waimwater streams of
Wisconsin. US Forest Service, United States Dept. of Agriculture, St. Paul, MN. General Technical Report
NC-149.
Lyons, J. S. Navarro-Perez, P. A. Cochran, E. Santana C., and M. Guzman-Arroyo. 1995. Index of biotic integrity
based on fish assemblages for the conservation of streams and rivers in west-central Mexico. Conservation
Biology 9:569-584.
Meiritt R. W„ and K. W. Cummins (editors). 1984. An Introduction to the Aquatic Insects of North America. 2nd
edition. Kendall/Hunt Publishing Company, Dubuque, Iowa
Miller, K. L., and 13 other authors. 1988. Regional applications of an index of biotic integrity for use in water
resource management Fisheries 13(5): 12-20.
Mitchell, W. C., and A. F. Bums. 1938. Statistical Indicators of Cyclical Revivals. National Bureau of Economic
Research, New York, NY.
Moyle, P. Band J. E. Williams 1990. Biodiversity loss in the temperate zone: decline of the native fish fauna of
California. Conservation Biology 4:275-284.
Mulvey, M., L. Caton, and R. Hafele. 1992. Oregon nonpoint-source monitoring protocols and stream bioassessment
field manual for macroinvertebnues and habitat assessment Oregon Department of Environmental Quality
Laboratories, Portland, Oregon.
Nehlsen, W., J. E. Williams, and J. A. Lichatowich. 1991. Pacific salmon at the crossroads: stocks at risk from
California, Oregon, Idaho, and Washington. Fisheries 16(2):2-21.
Nonis, R. H., and A. Georges. 1993. Analysis and interpretation of benthic macroinvertebrate surveys. Pages 234-
285 in D. M. Rosenberg and V. H. Resh (editors). Freshwater biomonitoring and benthic macroinvenebrates.
Chapman and Hall, New York.
Ohio EPA. 1988. Biological criteria for the protection of aquatic life, volumes 1-3. Ecological Assessment Section,
Division of Water Quality Monitoring and Assessment, Ohio EPA, Columbus, Ohio.
Pacific Rivers Council. 1995. A call for a comprehensive watershed and wild fish conservation program in eastern
Oregon and Washington. 2nd edition. Pacific Rivers Council, Eugene, Oregon.
Petexman, R. M. 1990. Statistical power analysis can improve fisheries research and management Canadian Journal
of Fisheries and Aquatic Sciences. 47:2-15.
Plafkin, J. L., M. T. Barbour, K. D. Porter, S. K. Gross, and R. M. Hughes. 1989. Rapid bioassessment protocols for
use in streams and rivers: benthic macroinvenebrates and fish. Assessment and Water Protection Division,
U.S. Environmental Protection Agency, Report EPA/440/4-89-001. Washington: EPA.
Potvin, G, and J. Travis. 1993. Special feature. Statistical methods: an upgrade for biologists. Ecology 74:1614-
1676.
Reynoldson, T. B., and M. A. Zarull. 1993. An approach to the development of biological sediment guidelines. Pages
177-200 in S. Woodley, J. Kay, and G. Francis (editors). Ecological integrity and the management of
ecosystems. St Lucie Press, Delray Beach, Florida.
Schelske, C. L. 1984. In situ and natural phytoplankton assemblage bioassays. Pages 15-47 in L. E. Shubeit (editor).
Algae as ecological indicators. Academic Press, London, England.
Schlosser, I. J. 1990. Environmental variation, life history attributes, and community structure in stream fishes:
implications for environmental management and assessment. Environmental Management 14:621-628.
57
-------
Simon, T. P., and J. Lyons. 1995. Application of the index of biotic integrity to evaluate water resource integrity in
freshwater ecosystems. Pages 245-262 in W. S. Davis and T. P. Simon (editors). Biological assessment and
criteria: tools for water resource planning and decision making. Lewis Publishing Boca Raton.
Sokal, R. R- and F-J- Rohlf-1981. Biometry. 2nd edition. Freeman and Company, New York.
Steedman, R. J. 1988. Modification and assessment of an index of biotic integrity to quantify stream quality in
southern Ontario. Canadian Journal of Fisheries and Aquatic Sciences 45:492-501.
Stewart-Oaten, A., J. R. Bence, and C. W. Osenberg. 1992. Assessing effects of unreplicated perturbations: no simple
solutions. Ecology 73:1396-1404.
Suter, G. W., n. 1993. A critique of ecosystem health concepts and indexes. Environmental Toxicology and
Chemistry 12:1533-1539.
Tabachnick, B. G., and L. S. Fidell. 1989. Using multivariate statistics. 2nd edition. HarperCollins Publishers, Inc.,
New York.
US Forest Service (USFS). 1992. Wild and scenic river eligibility and determination. Umpqua National Forest,
Roseburg, Oregon.
The Wilderness Society. 1993. The living landscape. Volume 1. Wild salmon as natural capital; volume 2. Pacific
salmon and federal lands. The Wilderness Society, Washington, D.C.
Winget, R. N., and F. A. Mangum. 1979. Biotic condition index: integrated biological, physical and chemical stream
parameters for management. US Forest Service, Intermountain Region, Provo, Utah.
Wolda, H. 1981. Similarity indices, sample size, and diversity. Oecologia 50:296-302,
Wright, J. F., M. T. Furse, and P. D. Armitage. 1993. RIVPACS - a technique for evaluating the biological quality of
rivers in the U. K. European Water Pollution Control 3:15-25.
Yoder. C. 0.1994. Integrating basic and applied benthic science. Journal of the North American Benthological
Society 13(3):391-398.
Yoder, C. 0., and E. T. Rankin. 1995a. Biological criteria program development and implementation in Ohio. Pages
109.144 in W. S. Davis and T. P. Simon (editors). Biological assessment and criteria: tools for water
resource planning and decision making. Lewis Publishing, Boca Raton.
Yoder, C. O., and E. T. Rankin. 1995b. Biological response signatures and the area of degradation value: new tools
for interpreting multimetric data Pages 263-286 in W. S. Davis and T. P. Simon (editors). Biological
assessment and criteria: tools for water resource planning and decision making. Lewis Publishing, Boca
Raton.
58
-------
TABLE 1. For each data set, number of sites available, number that were appropriate for
B-IBI development (in parentheses), and analysis for which data were used (footnotes). High
elevation sites and large river sites were excluded for index development but were used in
PC A. After exclusion of these sites, some data sets bad too few sites remaining for analysis.
Year
Medford
Roseburg
Umpqua
1990
46* (34blC)
1991
22(18°)
30 (30°,cl)
30 (20*)
1992
21 (0)
12(0)
30 (24b)
aPCA
b Metric testing
c Determine ranges of metrics
d Test B-IBI and RBP
59
-------
TABLE 2. Metrics tested; their predicted response to degradation; and their ability to clearly
distinguish least degraded from most degraded sites.
Metric Predicted Metrics thet could
response distinguish sites
Taxa richness & composition
Total number of taxa a
Ephemeroptera taxa a
Plecoptera taxa a
Pteronarcys taxa a
Trichoptera taxa a
Dipteran taxa
Chironomid taxa
Tolerant / intolerant
Intolerant taxa a
Sediment intolerant taxa a
Sediment tolerant taxa
% Toleranta
% Sediment toleranta
% Sediment intolerant
% Oligochaetes
% Chironomids
Feeding ecology
% Scraper
% Predator
% Gatherer
% Shredder
% Filterer
% Omnivore
Scraper taxa
Shredder taxa
Omnivore taxa
Population attributes
Abundance a
% Dominance (3 taxa)a
Decrease
Decrease
Decrease
Decrease
Decrease
Decrease
Increase
Decrease
Decrease
Increase
Increase
Increase
Decrease
Increase
Increase
Decrease
Increase
Increase
Decrease
Increase
X
X
X
X
X
X
X
X
X
X
a Component metric included in benthic index of biotic integrity (B-ffil).
60
-------
TABLE 3. Metrics included in B-IBI for Oregon streams, their response to human influence, and scoring
criteria. A score of 5 indicates a metric score thai deviates little or none from expected condition, 3
indicates moderate deviation, and 1 strong deviation from expectation.
Metric
Total number of taxa
Ephemeroptera taxa
Plecoptera taxa
Trichoptera taxa
Pteronarcys taxa
Intolerant taxa
% Tolerant
Sediment intolerant taxa
% Sediment tolerant
% Dominance (3 taxa)
Abundance
Response to
degradation
Decrease
Decrease
Decrease
Decrease
Decrease
Decrease
Increase
Decrease
Increase
Increase
Decrease
1
<40
<8
<6
<6
0
<2
>0.4
0
>0.15
>0.55
<500
3
40^54
8-11
6-9
6-9
NA
2-5
0.2 - 0.4
1
0.05-0.15
0.40-0.55
500 -1500
5
>54"
>11
>9
>9
>0
>5
<02
>1
<0.05
<0.40
>1500
61
-------
TABLE 4. Metrics included in Oregon DEQ's RBP, their predicted response to human influence,
and whether they could distinguish least degraded from most degraded sites.
Metric
Total number of taxa
Predicted response
Could it distinguish?
EPTtaxa
% EPTtaxa
% Chironomids
% Dominance (1 taxon)
% Shredders
% Filterers
% Scrapers
Hilsenhoff biotic index
Diversity index, H'
Decrease
Decrease
Decrease
Increase
Increase
Increase
Increase
Decrease
Increase
Decrease
yes
yes
yes
no
no; 2 or 3 taxa
works better
no
no
no
no
yes; but strongly corre-
lated with taxa richness
62
-------
CM
O
*k_
¦4—•
Q
r= .03
Resource condition
FIGURE 1 Hypothetical relationships between resource condition and candidate biological metrics. Metric 1 is
is me more .ffccdvc m«nc in «. of ta poo, su^cal
correlation.
63
-------
>
LU
o
o
o
CM
o
o
Instream condition
Poor Good
FIGURE 2. Relationship between elevation and instream condition (based on channel morphology, bank
erosion and stability, and riparian zone features) in the Umpqua NF (n»=41). Sites above 2500 ft were excluded
from analysis because there were no degraded sites at high elevation; thus, elevation was confounded with
degradation.
64
-------
Poor
Instream condition
Good
FIGURE3. Relationship between watershed condition at a site (percent area nor logged) and instream
condition (riparian cover, bank stability, and other physical features). Dashed lines denote direction of expected
correlation. When outliers were re-evaluated we found that area logged was underestimated (l, 2,4); data were
tabulated incorrectly (3,5); and some togging had occurred long ago (6).
65
-------
co
X
CO
-------
lo
£0 t-
X
«
(0 O
i_ f-
£
Q.
o
O
CD
a.
in
+ +
• •
+ •
• • •
• ••
• • H •
• • • •
• •
T
A
CO
o
"5
"O
CD
c
CD
O
CD
Q.
o
CO
o
CM
O
O
o
o
o
* +
• •
• • • •
• "TT •
• * J** m
I
: $
0.2 0.4 0.6 0.8
Percent area logged
CO
o
CO
1—
©
-o
"O
o
CD
CM
d
j^
CO
•*—>
o
c
CD
©
O
0
o
CL
o
0.2 0.4 0.6 0.8
Percent area logged
FIGURE 5. Metrics that could clearly least degraded ("+") from most degraded stream sites
were used to construct B-IBI. Two taxa richness metrics (on left) distinguished the two groups of sites. Trophic
structure metrics, such as percent of predators and shredders (on right), failed to distinguish the groups of sites.
Arrows to the right of plot indicate the range of metrics scores for good and poor sites.
67
-------
1
TAX A
TPLE
10
I
80
?
14
TINT
14
%TOL
80%
0%
%DOM
80%
20%
ABUN
3500
T
B-IBI score 15
site L. Elk Cr.
47
EF Cow Cr.
FIGURE 6. Because metrics have different ranges they are scored as 5 (indicating little or no deviation from
expected, or reference, condition), 3, or 1 (indicating strong deviation from expected condition) in order to put
tbfTn on the same scale. B-IBI score equals the sum of eleven metric scores. Human-induced degradation in and
around L. Elk Creek was reflected in a low B-IBI score. TAXA=total taxa richness; TPLE=Plecoptera taxa
richness; TINT=intolerant taxa richness; %TOL= Percent tolerants; %DOM=Percent dominance; ABUN=total
abundance.
68
-------
LL
Q
O
in
a Roseburg BLM
o Umpqua NF
A A
A O
A °
A O
10
A O
A O
i °
A O
1 2
A O
' A 0
* n
A O
*\0
A O
A o
20 30 40
B-IBI score
50
60 80
RBP score
100
.. cmp> of B-IBI scores and RBP scores for sites in the
FIGURE 7. Hie cumulative distribution 20) The CDF is the percentage of sites with B-IBI scores
RosdHTC .tenet d . »> art tts Umpquidumc. (n -20). ton* of difference
lower than the B-IBI score on the x-axis. (one-sided t-test; p < 0.05) in the Roseburg reflected higher
69
-------
o
Q_
PC I
FIGURE 8. Sites from Umpqua NF sampled in 1990 plotted according to a principal components analysis of the
"normalized" data. Most degraded ( ~ ), high elevation ( ± ), and large river (L) sites showed a tendency to
group together. Least-degraded sites (+) and other remaining sites ( ^ ) did not clump together for these or
other component plots. PCA of log transformed and uncransfonned data gave similar plots.
70
-------
DEVELOPEDG BIOASSESSMENT PROTOCOLS
FOR MONTANA WETLANDS
R. Apfelbeck J; BMs, L. J; Shapley, M. 2; Gerritsen, J. 3;
Barbour, M. 3; Stribling, J. 3; Charles, D. 4; Acker, F. 4
Presented by Randy Apfelbeck
Montana Department of Environmental Quality
Cogswell Building, 1400 Broadway
Helena, MT 59601
NOTES:
Eighty wetlands were sampled throughout Montana from April through
September during 1993 and 1994 for developing wetland bioassessment protocols
using macroinvertebrate and periphyton. Wetlands were sampled for
rnacroinvertebrate, periphyton, water column chemistry and sediment trace
metal chemistry. Hydrologic, geologic and climatic data were collected from
maps and existing databases. Additional hydrologic field measurements were
taken to assist in developing a better understanding of the water quality for
sites that were highly evaporative or predominantly groundwater supported.
A wetland classification system is currently being developed that will group
reference wetlands with similar habitats and water quality in order to improve
detection of anthropogenic impairments. The classification system incorporates
ecoregions and geomorphic processes. Presently 10 wetland classes have been
delineated.
Macroinvertebrate metrics identified as being useful in developing an index to
assess wetland water quality include: 1) # taxa; 2) % dominant taxa (1, 2, 5); 3)
# of Plecoptera, Odonata, Ephemeroptera and Trichoptera taxa; 4) # of
individuals; 5) % and # of Chironomidae and % Orthocladiinae/total
Chironomidae; 6) % and # of Crustacea and Mollusca; 7) # of Hirudinea,
Spongillidae and Sphaeriidae taxa; and 8) Hilsenhoff Biotic Index.
Preliminary results indicate detection of impairment that appear to be caused
by heavy metals, nutrients, salinity, sediment and fluctuating water levels.
Scuds (Hyallela azteca) were identified as being a potential indicator of
ephemeral wetlands (absence or very low numbers), saline wetlands (absence)
and alkali wetlands (very abundant). Periphyton bioassessment protocols are
currently being developed by The Academy of Natural Sciences of
Philadelphia.
1 Montana Department of Environmental Quality
2Montana Natural Heritage Program
3Tetra Tech, Inc., Owings Mills, MD
4Academy of Natural Sciences, Philadelphia, PA
71
-------
PROJECT OBJECTIVES
1. Develop Rapid Bioassessment Protocols
1. Determine Natural Spatial Variability
3. Develop Classification System
4. Detect Wetland Water Quality Impairment
5r Simple And Cost Effective Approach
-------
Montana DEQ Wetland Characterization Sites
-------
SAMPLING METHODS
1. Location
2. Sampling Period
3. Macroinvertebrates
4. Periphyton
5. Water Column
6. Field Analysis
7. Sediment -
8. Photo Documentation
9. Field Observations
CLASSIFICATION INDICATORS
1. Vegetation
2. Alkalinity
3. Scuds
4. Wildlife
5. Evaporative Salts
Watershed / Wetland Area
7. Topographic Position
WETLAND CLASSIFICATION
1. Ecorcgion
A) Rocky Mountain Ecoregion
B) Intermmmtain Valley Ecoregion
C) Plains Ecoregion
2. Gtomorphic
A) Open Lakes
B) Closed Basin
C) Riparian
D) Ephemeral •
3. Chemical Delineation
:W0 Conductivity
III! B) PH •=;•
ji;:j!||| C> !!: Alkalinity
MACROINVERTEBRATE INDEX
DEVELOPMENT
1. Identify And Count Taxa
2. Develop And Test Potential Metrics
3. Select Core Metrics
4. Metric Scoring
5. Aggregation Of Metric Scores
6. Index Scores
7. Assessment Of Sites
-------
75
-------
i
METRICS SELECTED
stressor;
#
Taxa
Decrease
%
Dominance Taxa (1>2,5)
Increase
#
Plec / Odon / Eph / Trie Taxa
Decrease
#
Individuals
Decrease
%
Chironomidae
Increase
#
Chironomidae
Decrease
%
Orthocladiinae / Total Chir
Decrease
#
Crustacea & Mollusca Taxapypi;
Decrease
Crustacea / Molluscs Taxa
Increase
!ll#l
Leech, Sponge and Clunt Taxa
Decrease
HQsenhoff Biotic Index
Increase; ; ;;
ASSESSMENT
For Each Wetland Class:
1. Evaluate Reference Sites
2. Evaluate Sites With Know Impairment
3. Use Physitochemiciil Data To Assist In
Evaluation
4. Use Field Notes And Historical Information
& Evaluate Variability
6. Evaluate Outlier Sites
7. Determine Condition
INDEX SCORE
METRICS RANGE
Hilsenhoff Biotic Index
Leech / Sponge / Clam
Crustacea / Mollusca
Chironomidae (midge)
# Individuals
Plec / Odon / Eph / Trie
Inverse % Dominance
# Taxa
1-14
0-18
0-22
0-16
0-24
0-25 •
1-28 •'
iii Total Index Score
"•T-T-T-T-T:
111! 7-119
FUTURE OBJECTIVES
If
1. Refine The Classification System
2. Determine Temporal VariabiiiiyThrough
Replication
3. Evaluate Seasonality And Determine Index.
Period •
4. Refine The Habitat Assessment Approach To
p i Reduce Subjectivity
$. Evaluate Using the Wetland Reassessment
Approach For tow (Gradient Prairie Streams
-------
MONTANA
WETLAND MACROINVERTEBRATE WATER QUALITY INDICES
REFERENCE WETLANDS
LU
DC
o
o
CO
X
LLl
Q
<
AVERAGE INDEX SCORES
METRICS
jJJ - J v.™™ H hilsenhoff biotic index
& 60 ¦ leech/sponge/clam
CD (TO _ EPHEMERAL _ ¦
w 50 s CRUSTACEA/MOLLUSCA
g 40 - 1 CHIRONOMIDAE (MIDGE)
> 30 - ¦ # INDIVIDUALS
O 20 - SAUNE B PLEC/ODON/EPH/TRIC
O in _ ~ INVERSE % DOMINANCE
# TAX A
120
110
100
90
80
70
60
50
. 40
30
20
10 H
o
OPEN LAKES/MOUNTAIN & VALLEY
RIPARIAN/MOUNTAIN & VALLEY
CLOSED BASIN/NONALKALI
RIPARIAN/PLAINS
OPEN LAKES/PLAINS
HEADWATER
CLOSED BASIN/ALKALI
CLOSED BASIN/SW
EPHEMERAL
SALINE
WETLAND CLASSES
-------
WETLAND 1.1 iQFtC'fr I VERTEBRATE WATER QUALIT Y INDICE!
CLASS 3
GROUNDWATER BUHWTED CLOSED BASIN WETLANDS
&ITJWJj-«TER9>
fc » «i8
*fi3H=8E S DGMWECE
SJ
00
rEtt»C-:r.PDE H«OE3i
MD MaCROIUVERTEERaTE WATER OUALITY. INDICES
CLASS 1
MOUNTAIN HEADWATER AND CLOSED BASIN WETLANDS
TJUjrE WIO IHTIJRAU.V AfCC! • 1
= hlsbih.iff botc »rc>
::: LEEOH'SPTWE, CL»M
:;=:Ctsuaw.CEi:!»iittt«j3ffi«:.
: 3i « rfeiciuLri
:r:: rco'i.-rpsi Tnbhh
:•:= ritERS; Si raiij«Sui
. a » Uh*
1101 1TANA '
MONTANA
WETLAND MAC-ROB K'ERTEBRATE WATER QUALITY IHDIQE3
CLASS 2
WETLANDS ASSOCIATED WITH STIEAMS, OPfllNOS AND FENS
trfijrn.
:: 1^BX£P0M5E-TH.«iM
=: i^isfiiXfc'MCitJfeiSsK |
* CMR-J ltl;'iTr.E:!M.DGE
arAupwcteu
:i»eruAHD:CDDE:na:€3:
-------
^1
ririp
•»t ER*f l£?: 1CTE I W.-EI-
v^TLAT-b MACRO*4VEPiTE6F»fcTE WATER OUAUTY INDICES'
CLASS 5
GROUNDWATER 8UPP0RTED CLOSED BASIN WETLANDS
! WrU£ ¥HAW . t
= HLS0JHOPF POTC IWE*
LEE- H'SPOt K3E • C l>J.l
: r STOSW€%?l»l2tt.W«*:
¦ <:«hct Htstiia: st.tfcttfef :: i
is#- 1
:::: ^pti ixoiQ?H/TTO^::::
• ig #1«R3E. H b&M 1«MI£:.i:.
: MONTANA
a MLce-IHOFF 60TKJ WOE*
« 1£EW'0W3E-CIAM
:»SS«SSt*56S.;»aUiil#8K:
¦ •.•HHao.iciEij.fDjEt:::
a «:.cvrt.iK.s::.::: ::-
fLKir:ccc»ia»H--iPt:
.»Mhskse* cdw-w re:::
WETLAIID M AORC'lfMERtEERATE WATER QUALITY INDICES
CLASS 7
MONTANA /
CLOSED BASIN WETLANDS WITH LARGE WATERSHEDS
¦ SOURCE (.P VATEft PM.MLY &JHFMX CfMttoGC!
MONTANA.
ID MA;:fv"#IVEFiTEEFlATE WATER DUALITY IHLOES-
CLASS 8
SAUNE WETLANDS
ciwn»c} ccmxn /Lena weiuho mrae'
i i nt?Ti: fr i Tr f i H i i
::::l::Hii:ii:i:H!}?iEiUW®:i5OT6:iwi^Si:liiliii:i
: MONTANA;
METRICS
s hilcojh:.ff potio HOP-*
LEEvH SPONGE. CL AM
;; ~;
:: ¦ CMFKir.i.iC-^E: ft.1CiG£i:::
iiaiiiiwnKDisiigiiliiililHHll?
:::«:iiteEhSE V; CiCttiff WOIC:::
i;*:
ilWETUANO MAC Rl'INVERTEBRaTE WATER QUALITY INDICES
CLASS? 8
EPHEMERAL WETLANDS OF THE PLAINS ECE ie««j ||
-------
WETLAND MACRO* JVERTEBRATE WATER. QUALITY: INDICE!
CLASS 9
OPEN LAKE WETLANDS OF THE PLAINS EOOBEQBM
iwEtl-ijcc •wth LdftOE •#ateh3hs® fir- 'Xttflcwsi
«tw
s rtLSSJHOFF BCTC I'lDEr
" L£E'H TFCI-aE-CL-M
is: O^.C.T-^Ek.-WXLUSrJ'i':
moi-itama :
WETLAND MA'~ROIt I ERTEBR-TE WATER QUALIT Y IMDIOEt
REFERENCE WETLANDS ,
AVERAGE 1HDEX SCOPES
43 rv^«. tuoa- wm uu i —r
mvtfx%'njkteo ^ :xT: lUil
wm iraa/nxn :::: 25=
seumcm 5 S ^ iSi HT:
o.tmu htftH'ftlKHt
(LiiHll iftMfc'W zz:
METRKJ8
ML8EHHOT BCTVw W-CC-
LEECH-'P.POriGE-CLJW
0 RUSTAC £A- "KfXLUoCA
IMMiML
DC*i«^KE:
00
O
CFTBMAL
ii-itt'Lfti its i c'CfiBitifttss:
•-'ETLANC1 MACR'Jllf ERTEPRATE WATER QUALITY IHDIOES
CLASS 10
OPEN LAKE WETLANDS
MITSfcOUflMM W.Lf. Ato FKt7 MOUNTAIN ECOreOCi®
~ HLBB-W*F BCTC ItCE'.
:: LEEC" •RfO'fjE-C-LW.i
ii:=iH^tetS:.».,i.'
-------
AN INDEX OF BIOLOGICAL INTEGRITY
FOR THE RED RIVER ECOREGION: A BIOLOGICAL "DESERT"
Eric Pearson
North Dakota Department of Health
1200 Missouri Avenue
Bismarck, ND 58502
NOTES:
The Index of Biotic Integrity (IBI) is a numerical expression of the fish
community structure. This study, initiated with a grant from the U.S.
Environmental Protection Agency, involves a number of agencies including the
State Health Department, Division of Water Quality, the North Dakota
Game and Fish Department, Minnesota Pollution Control Agency, Minnesota
Department of Natural Resources, the U.S. Geological Survey, NAWQA
program and Regions 5 and 8 of the U.S. EPA.
We currently are proposing 17 metrics which are measures of species richness
and composition, trophic composition, and fish abundance and condition. These
IBI scores will allow the Health Department to assess the aquatic life impacts
along the Red River and at various sites throughout the basin. These impacts
may be anything from habitat degradation (e.g., isolation or snag removal) to
toxic impacts such as sewage outfalls or chemical spills. The Health
Department hopes to expand this program to other basins as an ongoing
component of if s ambient stream monitoring program.
81
-------
UPDATE ON THE USE OF MACROINVERTEBRATES
IN MONTANA STREAM BIOASSESSMENTS
Bob Bukantis
Montana Department of Environmental Quality
Water Quality Division
A206 .Cogswell Building
Heleha, MT 59620-0901
NOTEg:
We have been using EPA's Rapid Bioassessment Protocol III modified as
follows: sampling is done with a D-net (1 mm mesh) employing a 1 minute
traveling kick method (2 minutes in mountain streams). Samples are taken from
riffles during the summer and subsampled to 300 +/- 30 organisms in the lab.
Level of taxonomic determination is specified, usually to genus or species.
Analysis of data so far suggests that Montana can be lumped into Mountain,
Foothill, Spring Creek, and Plains Bioregions. Work done on methods
development/testing suggests that the sampling method produces less variable
data than Surber samplers, and shows good repeatability among nearby riffles
sampled. There is good repeatability between years in Mountain streams, but
not for Plains streams.
Investigator bias was not evident with 2 people taking samples on a foothills
stream. However, there appeared to be a significant effect of investigator bias
on a large Spring Creek.
Provisional, regional reference communities are currently being used in
evaluation of macroinvertebrate samples taken for stream assessments. These
reference communities have been modeled by describing community
characteristics for non-impaired, slight, moderate, and severe impairment
categories with expected ranges of metric values. Data from reference streams
were used to set limits of impairment; a subjective evaluation of existing data
was used to set levels of impairment.
Specific reference communities are developed on long-term projects by
compositing the best attributes (metric values) available from the pool of
samples taken from the project system. Use of thes&so-called "Internal
Reference Communities" increases the ability to discriminate impairment
between sites.
Analysis of sampling variation shows promise for use in selection of
appropriate batteries of metrics for each Bioregion, and for refinement of
setting impairment limits.
82
-------
SAM^Lb VARIATION IN MAGROINVbR l EbRATES
00
CO
HI
o
u
tr
LU
LL
— ' o
Q ,U
o
cc
LU
CL
HI
0
<
CC
LU
§
0.5
fi ^
h +•
'A f n n i j
* *
¦**
j.
i !
* i
I ln
-L ± i !
|i ( : s i i HI I ^ | j \ j!j
, T L! i I! »j !! ri n 7 ! I i? n!
1 til t ! i i • •' : T f i H I ' i i: u I
4-- *
r>
^ <2fO^ *
¦\V-V
Ji!i
¦ : r4
1,1^ jrviM?
°?
FOOTHILLS ECOREGiON
-------
SAMPLE VARIATION !N MACROINVERTEBRATES
FOOTHILLS ECOREGION
-------
SAMPLE VARIATION ih MACROINVERTEBRA i ES
00
cn
o'V<^ * °°vv^
FOOTHILLS ECOREGION
-------
SAMPLE VARA i ION IN N/1ACROINVER7EBRA
' I I I I I I : | | | | | | , | | | | |
PRING ECOREGION
-------
SAMPLE VARIATION IN \MCROINVERTEBRATES
1.5
LU
O
z:
LU
cr
lu-
ll
ll
Q
"..0
00
IIJ
o
DC
LU
Q_
Ui
0
<
CC
LU
5
0.5
0.0
h
T" i
SPRING ECOREGION
-------
SAMPLE VARIATION IN MACRCINVERTEBRATES
UJ
o
z
u
cc
ID
LL
U_
a
00
00
1.5
I I i
£ 1.0
<£-
UJ
o
tx
UJ
CL
LLi
<3
<
cc
LU
5
0.5
0.0
0
EL
t~t
c
,-te
[i 1 ::
in ~ © i I
I: U
1 X.
M :4>
&• o 0
SPRING. ECOREGiON
-------
0.22
0.2
0.18
0.16
0.14
0.12
0.1
0.08
0.06
0.04
0.02
0
-------
SAMPLING VARIATION
SURBER VS. KICKNET
' BEAR CREEK (DOWNSTREAM)
ggg SURBER I I KICKNET
DATA BY: INTER-FLUVEJNC., 1994
-------
Z
<
UJ
2
SAMPLING VARIATION
SURBER VS. KICKNET
BEAR CREEK {DOWNSTREAM)
SURBER I 1 KICKNET
DATA BY: INTER-FLUVE, INC., 1994
-------
ro
<
LLI
2
SAMPLING VARIATION
SURBER VS. KICKNET
BEAR CREEK(DOWNSTREAM)
SURBER I I KICKNET
DATA BY: INTER-FLUVE, INC., 1994
-------
TAXA RICH
EFT RICH
%DOM TAXA
BIOTiC INDEX
MTI
%CG + CF
%SCR+SHR
¦o %EPT IND
.2 QSI
a QSI-FFG
E DIC-5
& %HYDR OF TRICH
% CHIRON
8 %EPHEM
'£ %PLECOP
® %MULTIVOLT
2 %UNIVOLTINE
%SEMIVOLT
BCI-cQT
SHANNON H
# OF SCR TAXA
# OF PRED TAXA
# OF COLL-GATH
% BAETID OF EPHEM
FOOTHILLS STREAMS
0.2 1 0.4
0.3
Average Percent Difference (# of Sites Sampled)
-------
I
FOOTHILLS
JS.
a
£
a
-------
INVESTIGATOR BIA
c_n
0
u
c
0)
£
£
D
c
0)
o
5
Q.
0>
cn
£
<5
>
<
0.5
F*1 SPRING
WM BIG SPRING CREEK
-------
INVESTIGATOR BIA
V£5
cr»
O
O
c
£
c
0)
o
5
CL
0)
O)
2
%
>
<
. 4s
-------
REFERENCE
COMMUNITIES
•Upst r oam-Down stream
•Paired Wa "I e r<; h ed
•Internal
•Ecoregion
BIO ASSESSMENT
•Select Field Sites
•Col Iec t s amp Ies
•Subsample samples
Make Taxonomic
Deter minat ions
•Summar i ?e Data
•Interpret Data
97
-------
DATA SUMMARY
AND
INTERPRETATION
•COMPUTE me t r i cr,
•COMPARE metrics to
REFFRENCE
~assign metric scores
• SUM me trie scores
•CONVERT to PERCENT
SCQELLMi_CR ITFRIA
non s I i ght moflftratp SfiVftf e
me trie fi 4 2 Q
Taxo rich >28 28-24 24-19 <19
HPT rich >19 19 18 17-1& <16
Biotic <3 3-4 4-5 >5
&DOM taxon <25 25-35 35-4S >45
XCoilCG+F) <60 60-70 70-BD >80
5KScr + Shr >55 55-40 40-25
-------
PLAINS REFERENCE STREAMS
SmniMT! I-BITFPIA ..
non .5 I jgh-t moderate
me t r i c
fi
4
p n
Taxa richness
>24
?a-ia
1 B - 1 2
<12
EPT r i chness
B-6
5-3
<3
Id i ot i c index
<5
5-fi
6-7
>7
9SDOM t a x o n
<30
30-45
15-60
>60
SKCo 1 1 (G+F )
<60
GD-0D
00-95
>05
% EPT
>5U
50-30
3D- 1D
<10
'^ScrTShf '
# Predator taxa
S. Diversity
">"30"
>5
>3 . (J
<40
TD™i'5
4-5
3.0-2.4
4n-nn
3-4
2.4-1 B
fin.fln
<3
<3
<1 .1
?sn
(McGutre 199S)
FOOTHILL AND VALLEY
REFERENCE STREAMS
SCORING CRITERIA
n«n
c. I 1 ght
modprat p
sevorn
mo1 r i r
r
4
?
n
Taxa r i chness
>28
2B-21
21-14
<14
EPT r i chness
>14
14- 13
12-11
<11
Biot i c index
<4
4-5
5-G
>6
9GDOM tax on
<30
30-40
40-50
>50
99Coll[GtF)
<60
BD-75
75-90
>90
5KSc r +Sh r
>30
io pn
20- 10
<10
Hydropsych.
<7r>
7S-HfS
05-95
>95
CPT
..*60
GO 45
45-10
>in
(McGu i r* 19QS)
99
-------
THREEMILE CREEK
JJJ.iJWir, Ror
H/qfl ft/QT
a mi
Taxa richness
32
31
31
fcPT r i chriess
15
13
1fi
Biot i c i ndex
4 . 7
3 .5
3 . D
% Dom taxon
22
23
.in
% Ccl ! -FTG
80
60
52
5K iicr + Shr
19
2y
30
% EPT
28
11
5D
% Mli11 i volt i ne
BR
31 ..
31
(Menu ire 1995)
THREEMILE CREEK ~]
I I I i nr. i n Rnnr h Prl
a/QD ft/QI fl/qq
Taxa r i chness
4
4
4
EPT r i chness
D
0
0
B i ot i c i ndex
2
2
2
% Dom taxon
6
B
6
% Col 1-TFG
2
6
6
% 5criShr
2
4
b
% EPT
2
4
4
8j Mul t i voU i ne
n
4
4
Total score
1E
30
32
% of Reference
33%
6 3%
61%
C1 assif i cat i on
MOD
fiL f
SL 1
(McGutro 1995)
100
-------
THREEMILE CREEK REFERENCE
VA
SCORING CRITERIA
non
metr ic
sli ght
rmnri e r a t ^ sever?.
2.
Tax a richness.>33.6
ED~ r i c hnes s >22.5
Biotic Index >1.36
%COM taxon
%Co\ I[G+-)
%Sz r+5h r
%~PT
%MuIt i v oIt
>8 . 4
22 . 5 - 2C
1 96-1.5'
S.4-6.3
5.3-4.2
H
33 6-25.2 25.2-16.8 <15.8
20-17 ."5 <17.5
.6"- 1 .15 <'.15
<4 . 2
>43.6 40.8-35.7 35.7-30 6 <30.6
>31.2 31.2-23.4 23.4-15.6 <15 6
>51.75 51.75-34.5 34.5-17.25 <17.25
>4 . B
4.B-3.R
3.B-2.4 <2.4
(McGuire 1992J
-------
Relative Percent Difference Between
1990-94 In Reference Streams
PJ gins Mounts, ias.
Tsxa Ri chness
EPT Richness
S5 Dominant
% Co II ector <; ( g+f }
% Scrapur s
% Co I I -gath
% Scr+Shr
% Ch i r on
% FPT
B i ot i c index
S. Di versi ty
H Pr^rlatpr taxa
IS
b
19
7
3D
17
11
19
32
16
4H
21
54
1U
5b
54
IB
7
11
23
18'
G
1JL
(McGuire 1Q95)
102
-------
USE OF LAKE BENTHOS AS A BIOASSESSMENT TOOL
Malcolm Butler
Department of Zoology
North Dakota State University
Fargo, ND 58105
NQTES;
The profundal benthos has been used to classify lakes into types for over 70
years. Numerous classifications and indices have been generated based on
(a) observed patterns in the distribution of taxa, and (b) knowledge of the
ecological requirements or tolerances of specific benthic animals. Such
indicator-based systems have served well to monitor lakes for impacts whose
effects on the fauna are known. Multivariate approaches permit a broader
focus in the search for patterns within faunistic and environmental data.
We are surveying the profunda! benthos across a wide diversity of lakes in
Region VIII states, for which basic limnological data are available. We hope
to identify characteristic benthic communities and indicator taxa that reflect
environmental differences such as lake depth, size, region, and trophic state. I
will illustrate the general approach with an initial data set on 12 North
Dakota lakes that was analyzed by Dr. Trefor B. Reynoldson of the Canadian
National Water Research Institute. Three groups of lakes resulted from cluster
analysis of the benthic macroinvertebrate communities, and non-metric
multidimensional scaling identified potential indicator taxa. Discriminant
function analysis showed that lake depth, size, trophic state index, and
dissolved oxygen at Zmax did a fair job of distinguishing the three lake groups
independently defined by benthic macroinvertebrate communities.
To further illustrate the generality and potential of this approach, I present a
similar analysis of 26 Swedish lakes published by Dr. Richard K- Johnson
(Department of Environmental Assessment, Swedish University of Agricultural
Science). Johnson uses alternative multivariate methods to classify lakes
according to their biota (TWINSPAN) and to seek environmental correlates
with species abundance patterns (Canonical Correspondence Analysis) prior to
producing a predictive model of species occurrence with discriminant function
analysis.
The most obvious results from these analyses often reiterate the limnological
wisdom of seat-of-the-pants biologists from decades past. However,
multivariate approaches have the potential to identify subtle, confounded, or
unexpected patterns in extensive data sets.
103
-------
03/08/94 16:21:33.14 DENO Dakota Lakes Invertebrate Communities
0.1460
0.3228
0.4996
Bradd (
Wars (
1)
Fordv (
1)
Larimr (
1)_
Dd Colt (
D|
R. Willw(
2)
HcVill (
2)
L. Will (
2)
Kota Ry (
2)
Smishk (
2)
Silver 1(
3)
Skjerm (
3)
0.6764
I
0.8532
I
1.0300
I
0.1460
0.3228
0.4996
0.6764
0.8532
1.0300
-------
Dakota Lakes - SSH Ordination
1.5
1
0.5
CM
~ 0
o
0)
-0.5
-1
-1.5
-1
-0.5
0
Vector 1
0.5
1
Gp 1
¦
Gp 2
~
Gp 3
*
-------
Dakota Lakes - SSH Ordination
CO
o
o
1.5
1
0.5
0
-0.5
C¦ decorums ^
Chaoboruz
Gp 1
¦
Gp 2
~
Gp 3
-1
-1.5
-2
¦
-1
-0.5
0
Vector 1
C. ~f-ai ^
0.5
1
-------
OBS
1
2
3
4
5
6
7
8
9
10
11
12
GP
1
1
1
2
2
3
3
2
1
2
1
2
SAS
SITE
ZMAX
ZMN
AREA
DO
TSI
Bradd
4
1.8
28.1
5.0
70
Wars
6
3.4
21.6
1.1
61
Fordv
6
3.2
75.7
2.3
69
RWillw
6
3.1
12.1
2.5
55
LWillw
7
2.9
62.6
5.9
51
Silverl
3
2.6
36.8
4.1
70
Skjerm
5
3.0
16.2
2.9
47
Smishk
7
2.9
75.9
2.9
54
Larimr
9
4.5
24.1
0.6
59
McVill
6
3.6
13.1
0.6
66
DdColt
10
5.6
45.7
0.3
63
KotaRy
7
3.5
11.9
1.1
49
-------
Discriminant Analysis - Dakota Lakes
2
1.5
1
0.5
0
-0.5
-1
-1.5
H
{2
i
.r
-V
V
V)
i.
vs
-2
II
She\ HouJ
H.gh DO
-6
La r^e.
L.olO DO
-2 0
DF1 (93%)
Gp1
¦
Gp2
~
Gp3
-------
Prediction of Dakota lakes invertebrate assemblages
Discriminant Analysis
Cross-validation Summary using Linear Discriminant Function
Number of Observations and Percent Classified into Group:
1 2 3 Total
From Group
1
2
3
0
5
40%
60%
100%
2
2
2
1
5
40%
40%
20%
100%
3
0
0.
2
2
100%
100%
Total 4 5 3 12
33% 42% 25% 100%
Error Count Estimates for Group:
1 2 3 Total
Rate 60% 60% 0 40%
-------
Prediction of Dakota lakes invertebrate assemblages
Discriminant Analysis
Resubstitutlon Summary using Linear Discriminant Function
Number of Observations and Percent Classified into Group.
1 2
From Group
Total
1
4
80%
1
20%
0
5
100%
2
1
20%
4
80%
0
5
100%
3
0
0
2
100%
2
100%
Total
5
42%
5
42%
2
17%
12
100%
Error Count Estimates for Group:
1
Rate 20%
2
20%
3
0
Total
13%
-------
c o
3 r1
^ 33
a *
> C_
(Q O
IS. ZT
O 3
C CO
f s
- O
CO ©
Q/S
CD "
=3 O
o
P m
on <
o =;¦
x g
o 3
£ S3
P 2-
0)2.
>
o? w
- U)
a>
3
CD
3
O
-vl
c
"O
"O
0)
E.
0>
E. ro
a
*
CO
?
a>
a.
® «<
CO
a>
S
o'
p
3?
a>
CL
w"
ZJ"
Swedish reference lakes
(n = 26)
I
(n = 22)
C. flavicans (Meigen)
Ch. anthracinus Zett.
Z. zalutschicola Lipina
5. coracina (Zett.)
I
(n = 4)
Micropsectra spp.
H. subpilosus (Kieff.)
Paracladoplema spp.
S. ferox (Eisen)
I
(n = 5)
P. triquetra Tshern.
Dicrotendipes spp.
P. prasinatus (Staeg.)
Microtendipes spp.
A. aquaticus L.
M. angustata Curtis
G4
BD7
F4
G9
BD15
Group 2
I—
(n = 9)
(n = 17)
S. coracina
S. rosenschoeldi (Zett.)
C. flavicans
I
HI
AC 16
BD20
U3
Group 1
I
(n = 8)
C. flavicans Monodiamesa spp.
Ch. plumosus L. S. rosenschoeldi
Ch. tenuistylus Brundin Pisidium spp.
E6 04
F10 05
H8 P6
K2 R4
N4
Group 3
A4 W9
E5 X3
N6 XI1
S18 Y1
Group 4
-------
Figure 6 from: R.K. Johnson (1995)
Axis 2
>epth
^Micropsectra spp.
S. coraiina/zett.) S. /e™^(Eisen)_
Ch. tenuistylus Brundin •
C. flavicans (Meigen)
Ch, plumosus I
Ch. anthracinusZ*
Z. zalutscfuZoixL^LAm
tipma
.nosenst
rfodiamesa jgjAroa \
. ^f^atitude
A. aquitti{MS L.
p. triqiUHra Tshern.
02(min)
Microtendipes sppiffajicrotendipes spp
M. angustata Curtis* J
prasinatus (Staeg.) #1
Temperature
Axis 2
|H1i
E5 _
XJ3
• ##AC16
pHV#
BD20.
^Area-Axis 1
Latitude
O 2(min)
Te
mperature
-------
• •win. i i.i\, ouiiiidUM [ittifOJ
Table 3. The predicted probability of occurrence of selected taxa in limed reference lakes
compared with observed values (frequency of lakes where the taxon occurred).
Discriminant function analysis was performed with seven environmental variables and
TWINSPAN group classification of 26 Swedish reference lakes (see text). MP = median
probability of taxon occurrence and % O = percent of lakes where taxon was found
occurring.
TWINSPAN group 1 TWINSPAN group 4
MP
%0
MP
%0
Oligochaeta
99
100
Procladius spp.
75
100
Procladius spp.
75
100
S. coracina
75
100
Tanytarsus spp.
75
75
Tanytarsus spp.
75
100
Pisidium spp.
75
67
Oligochaeta
75
100
S. ferox
75
58
Pisidium spp.
75
0
H. apicalis
75
58
C. flavicans
66
100
Nematoda
75
50
Monodiamesa spp.
66
0
Protanypus spp.
75
42
Hydracarina
57
100
Micropsectra spp.
75
33
Z. zalutschicola
56
100
Turbellaria
75
25
S. rosenschoeldi
56
50
H. subpilosus
75
25
Ch. anthracinus
47
50
Monodiamesa spp.
75
17
H. apicalis
47
0
Paracladopelma spp.
75
8
Polypedilum spp.
38
0
S. coracina
50
92
Nematoda
38
0
S. rosenschoeldi
50
67
Micropsectra spp.
28
0
Hydracarina
50
58
Turbellaria
28
0
Psectrocladius spp.
50
25
T. tubifex
28
0
H. marcidus
50
25
S. ferox
19
0
Polypedilum spp.
50
17
H. marcidus
19
0
Gammarus spp.
50
0
Psectrocladius spp.
19
0
T. tubifex
50
0
Protanypus spp.
9
0
C. flavicans
25
58
Gammarus spp.
0
0
Z. zalutschicola
0
33
H. subpilosus
0
0
Ch. anthracinus
0
17
Paracladopelma spp.
0
0
-------
ASSESSING EFFECTS OF METALS ON BENTHIC MACROINVERTEBRATE
COMMUNITIES IN ROCKY MOUNTAIN STREAMS
Peter Kiffney
Department of Fishery and Wildlife Biology
Colorado Stale University
Fort Collins, CO 80523
NOTES:
Biological assessment to determine ecological integrity of aquatic systems has
become increasingly important with the realization that chemical monitoring
is not always protective of aquatic life. Our research group has used a number
of approaches in assessing the biological integrity of Rocky Mountain streams,
such as whole effluent toxicity tests, field biomonitoring, and toxicity tests
using indigenous organisms in experimental streams. I will discuss the merits of
these methods, and provide an example of each approach in evaluating the
effects of metals on stream benthic organisms. Because of the inherent natural
temporal and spatial variability of Rocky Mountain streams, I suggest that an
integrative approach that incorporates both experiments and field
observations be used when assessing the effects of anthropogenic inputs on
stream organisms.
The traditional approach to assessing the biological effects of contaminants on
aquatic systems has been the single-species toxicity test. These tests can be
reproducible, easy to perform, comparable across laboratories, and defensible in
court. However, single-species tests are also ecologically unrealistic for a
number of reasons. For example, test organisms are typically not found in
mountain streams (i.e., Ceriodaphnia dubia) and, thus, responses to an effluent
observed in a single species test may not be applicable to responses of natural
populations in the receiving stream.
An alternative method to single-species toxicity testing is the use of indigenous
organisms in experimental streams. In a series of experiments, I used artificial
substrates to collect invertebrates and transfer them to experimental streams
where they were exposed to metals. At the end of lOd, the effects of metal
treatment on population-, and community-level indices were examined using
ANOVA or regression techniques. With this experimental system, I examined
the influence of stream altitude on the response of benthic macroinvertebrate
communities to metals. Results from these experiments showed that the
response of invertebrates to metals was affected by stream altitude, as insects
from high-altitude streams were more sensitive to metals than similar groups
from low-altitude streams. Additionally, my research showed that the most
sensitive group to metals was mayflies, especially heptageniid mayflies.
Thus, this family may be a reliable and sensitive indicator of metal-pollution
in western streams.
Field biomonitoring using stream invertebrates has become increasingly
important in biological assessment of aquatic systems. Biomonitoring studies
have limitations, however, because of a number of statistical problems.
Nevertheless, the use of field biomonitoring can provide insights into processes
occurring in metal-polluted streams. In the Arkansas River, our group has
collected stream invertebrates to measure metal levels in tissue and for
community structure. Because this is a single system study, inference from these
104
-------
results are limited; therefore, we have also collected similar data from five
other metal-polluted streams in the central mountains of Colorado. Results
from these surveys have allowed us to develop metrics for metal-contaminated
streams. Some metrics, such as abundance of caddisflies, were confounded by
other abiotic factors (altitude), whereas abundance of mayflies or mayfly
species richness were consistently lower at metal-polluted sites than at
reference sites. Because this study included surveys of six streams, results have
a broader implication for a larger population of metal-polluted streams in
Colorado.
Since there are advantages and disadvantages associated with all methods of
biological assessment, I suggest that biologists responsible for evaluating the
ecological integrity of aquatic systems use an integrative approach. For
example, in our research group we have combined stream biomonitoring with
whole effluent toxicity tests, and/or toxicity experiments with indigenous
organisms to evaluate the ecological effects of metals on natural populations of
stream invertebrates in Rocky Mountain streams. This approach has provided
metrics that have been found to be consistently sensitive to metal pollution in
Colorado Rocky Mountain streams, and may be applicable to other regions of
the intermountain west.
105
------- |