Ontario Benthos Biomonitoring Network Participants’ Training Updated April 2006
Jan 13, 2016
Ontario Benthos Biomonitoring Network Participants’ Training
Updated April 2006
Clear Lake Inflow, 22-May-2005Longitude: -74.7° Latitude: 45.0°
Sampled by: Jones & Dmytrow
Test Site AtypicalHypothesis-test Results Index Contributions
D F Pnon-central
7.52 94.31 0.03
Index Contributions*
CA1(abundance) CA2(abundance) Richness % Chironomidae % EPT
D 6.52 3.38 1.53 5.04 4.22F 110.28 100.73 2.3 65.81 6.4
P 0.015 0.014 0.177 0.038 0.032
*values in bold typeface are beyond 2 st. dev. from the reference-site mean
Summary Statistics
Reference Sites (nref=15) Test Site*
mean St. Dev. meanCA1(abundance) -0.84 0.43 0.09
CA2(abundance) -0.34 0.41 2.74
Richness 13 4.16 13% Chironomidae 34 12.31 33% EPT 62 8.90 46
*values in bold typeface are beyond 2 st. dev. from the reference-site mean
Stream reference sites with test-site like collection method, gear type, mesh size, collection season, and flow permanence were selected based on similarity (Euclidean distance) to the following test-site habitat features: dominant substrate, elevation, latitude, longitude, and catchment area. Euclidean distances for reference sites ranged from 5 to 72. Total Euclidean distance for 15 reference sites and 5 attributes was 494
Standard Report (OBBN Vision)
InstructorsChris Jones, Ministry of Environment, Benthic Biomonitoring Scientist and
OBBN coordinator (Lead Instructor)
Nicole Dmytrow, Saugeen Conservation, OBBN Assistant Coordinator (Sampling, Benthos Identification)
Ron Reid, Ministry of Environment, Benthos Scientist (Sampling, Benthos Identification)
Michelle Bowman, University of Toronto (RCA bioassessment calculations: Test Site Analysis)
Desired OutcomeParticipants understand the purpose and
administration of the OBBN, and demonstrate competence with its methods.
This course is part of OBBN’s quality assurance plan: certification is one way of protecting the credibility of the
Network.
The OBBN is part of the Canada-wide Canadian Aquatic Biomonitoring Network (CABIN). We are working on
standard training and certification requirements for CABIN.
Participants’ Certification• 2 types of certificates (Participant, Trainer)• To be certified, participants must:
– Pass a general multiple-choice test– Correctly identify 40 of 44 benthos specimens to the coarse
OBBN 27-group level• In addition to above, trainers must:
– Assist with teaching the course– List at least 2 diagnostic characters for each specimen on the
benthos identification test (without consulting references)
Student Instructors
Rebecca Crockford, District of Muskoka
Lynette Dawson, Quinte Conservation
Gerry Sullivan, Otonabee Region Conservation Authority
Angela Wallace, Gartner Lee
Biomonitoring Knowledge vs. Degree of OBBN InvolvementDegree of Network involvement vs. OBBN Knowledgea
y = 1.3454x + 6.8247
R2 = 0.274
y = -1.2307x + 6.5219
R2 = 0.25570
2
4
6
8
10
12
14
16
1 2 3 4 5 6
Ordinal Degree of involvement
# o
f An
swe
rs
Correct
Don't Know
Equation R2 C = 6.82 + 1.35(ODI) 0.27 DK = 6.52 - 1.23(ODI) 0.26 C = 10.5 + 0.072(MI) 0.11
DK = 3.29 - 0.0716(MI) 0.12 Regression results (C = number of correct answers, DK = number of questions
answered ‘Don’t Know’, ODI = ordinal degree of OBBN involvement; all listed comparisons are significant at the =0.05 level)
Agenda: Day 1Welcome to the Course PurposeBackground (Need for Biomonitoring; Benthos as Indicators; Benthos facts; Complementarity of Stressor- and Effect-based Monitoring; OBBN Components, Principles, and Status Update)
Chris Jones, Gerry Sullivan
The Reference Condition Approach(RCA Overview; Definition of Reference Site; OBBN Reference Site Sampling Strategy; Criteria for minimally impacted; Spot the reference site; Example of RCA Bioassessment)
Chris Jones,Angela Wallace
Protocol(Standardization vs. Flexibility, Collection Procedures For Lakes, Streams, Wetlands, Processing Methods, Archiving,Habitat Characterization)
Chris Jones,Lisa Campbell
Sampling: Kennisis River and Lake of Bays(Student trainers as group leaders)
Nicole Dmytrow, Chris Jones, Ron Reid, Gerry Sullivan, Angela Wallace, Lisa Campbell, Lynette Dawson, Rebecca Crockford
Sieve Samples Nicole Dmytrow
Agenda: Day 2
Benthos Picking (random sub-sampling to obtain ~100-count sample)
Nicole Dmytrow, Chris Jones, Ron ReidGerry Sullivan, Angela Wallace, Lisa Campbell, Lynette Dawson, Rebecca Crockford
Benthos Identification (OBBN 27-group Level)- Diagnostic features of each group (slide show)- Examples from the DESC reference collection
(demonstration)- Practice using specimens collected yesterday (hands-on)
Chris Jones, Nicole Dmytrow,Rebecca Crockford, Lynette Dawson
Practice identification skills Chris Jones, Nicole Dmytrow
Students to identify specimens in front of class (microscope projection), highlighting diagnostic characters
Chris Jones
Agenda: Day 3
Assessment: Is Test Site Within Normal Range?-Summary Metrics-Hypothesis Testing (TSA)
Michelle Bowman, Chris Jones
Review Gerry Sullivan, Angela Wallace, Lisa Campbell, Lynette Dawson, Rebecca CrockfordChris Jones, Nicole Dmytrow
Certification Test (Optional) Chris Jones, Nicole Dmytrow
Take-up test, general discussion, and wrap-up Chris Jones
Biomonitoring Rationale • Legislation & policy stress protection of biota
– Biological definitions of impairment and adverse impact in Ontario
– “biological integrity” in U.S. Water Pollution Control Act
– The EU Water Framework Directive requires both “good ecological status” and “good chemical status” of surface water
• Management stresses protection/rehabilitation of biota:– Target setting– Performance evaluation
(Roux et al. 1999, Jones et al. 2005b, Jones 2006)
Biomonitoring Rationale II
“Biomonitoring is required … because the consequences of environmental stress can only be determined by an appraisal of the
biota”.
Wright (2000)
What are Benthos?
• Bottom-dwelling aquatic invertebrates
• Include animals like insects, worms, mollusks, crustaceans, and mites
Caddisfly of the family Helicopsychidae
Mayfly of the family Ephemerellidae.
Why Use Benthos As Bioindicators?
• Abundant and widespread • Nobody cares• Easily and inexpensively sampled• Sedentary (unlike fish)• Long lived (months to years)• Many species with different
tolerances • Respond to both water and
sediment chemistry• Readily archived for future
reference• Provide early-warning
Stream benthos collection in the Raisin River watershed
Benthos are excellent indicators of aquatic ecosystem health.
(Rosenberg & Resh 1993, 1996; Mackie 2001)
Complementarity of Stressor- and Effect-based Monitoring
Stressor-based Approach Effect-based Approach
Monitoring focus
Stressors causing environmental change, i.e., chemical and physical inputs
Effects (responses) of natural and/or anthropogenic disturbances, e.g., changes in the structure and function of biological communities
Management focus
Water quality regulation: controlling stressors through regulations
Aquatic ecosystem protection: managing ecological integrity
Primary indicators
Chemical and physical habitat variables, e.g., pH, dissolved oxygen, copper concentration
Structural and functional biological attributes (e.g., relative taxa abundances, frequency of deformities)
Assessment end points
Degree of compliance with a set criterion or discharge standard
Degree of deviation from a benchmark or desired biological condition
Adapted from Roux et al. (1999)
Stressor and Effect-based Approaches are Complementary
Phosphorus Data: 1997 - 2001
0
0.02
0.04
0.06
0.08
0.1
Pretty River @ hwy. 26, Collingwood
mg
/L
Zinc Data: 1997 - 2001
0
5
10
15
20
25
Pretty River @ Hwy. 26, Collingwood
ug
/l
Biology
Chemistry
Benthos data, Pretty River, October 1996; reference site data, 1997-2000
= Ontario Water Quality Objective
Pretty River, Highway 26,
Collingwood, Ontario
Pretty
Mad R.Noisy Nottawasaga
Pine 1
Pine 2
Sheldon
CA1
CA
2
95% confidence ellipse
Technical Issues
• Unlike water chemistry, no guidelines or “biocriteria”exist
• Complex; many confounding factors: biota respond to things other than stressor of interest
• No standard sampling protocol• Taxonomy requires special
expertise• Experts disagree on
hypothesis-testing procedures and interpretation
• Cost
The application of benthos biomonitoring has been limited by a number of technical issues.
OBBN Background
OBBN: a collaborative lake-, stream-, and wetland-bioassessment network
Leads: Ontario Ministry of Environment and Environment Canada (EMAN), but part of national CABIN program
1. Evaluate aquatic ecosystem condition
2. Measure effectiveness of programs
3. Provide biological complement to Provincial Water Quality Monitoring Program
4. Support development of biocriteria for aquatic ecosystem condition
Purposes
Aquatic mite
Barriers to Biomonitoring in Ontario
Standard Protocol
Data Sharing
Training
ImplementationStatus
Protocol
ResearchAnalytical Software
Database
Training
OBBN
• Train-the-trainer • Integration with North American Benthological Society Taxonomic Certification Program (NABS TCP)
• Query tool, data exporter, automated bioassessment-hypothesis test, reporting module
• spring 2006 release date
• Collaborative projects required to develop OBBN products
• Current focus is on understanding sources of variance and evaluating methods
• On-line• Printed
manual subject to Ministry approval
• National integration• Launched 31 Oct. 2005• ~30 organizations have
accounts
http://obbn.eman-rese.ca
OBBN Partners
OBBN Leads
• Ontario Ministry of Environment
• Environment Canada’s Ecological Monitoring
and Assessment Network
Technical Advisory Committee
• Universities • Conservation Authorities
• Ontario’s Ministries of Environment and Natural
Resources• Environment Canada
• Trout Unlimited• District of Muskoka
Certified Participants
• All Sectors
OBBN Partner Roles
Partners
• Sampling (for their own purposes and to collaborate on regional, provincial, and national reporting)
• Data-sharing• Research
OBBN Leads
• Coordinate 5 program components
• Provide technical advice and equipment
• Research
Technical Advisory Committee
• Technical guidance and review
• Research• Program Priorities• Problem Solving
Data-sharing Agreement
I understand and accept that as a partner in the Canadian Aquatic Biomonitoring Network, data entered into this system is freely shared among all Network participants.
I further understand and accept that CABIN and its partners put no restrictions on, and do not regulate, how data is used by network members.
Although I have made every attempt to ensure the quality of the data I enter into the database, I make no guarantee about the accuracy of that data, and assume no liability associated with its use.
OBBN Socio-economics and Demography
Vocational Sector (n=38)c
0%5%
10%15%20%25%30%35%
PS Gov CA Acad Ed NGO
Age (n=37)
0%
10%
20%
30%
40%
20-29 30-39 40-49 50-59 60-69
Per
cent
of R
espo
nses
Highest Level of Education Achieved
(n=38)a
0%
10%
20%
30%
40%
50%
CD UUG UGD
Employment Status (n=38)b
0%
20%
40%
60%
80%
100%
U R PT FT Other
Per
cent
of R
espo
nses
OBBN participants’ socio-economic status and demography (aCD = college diploma; UUG = university undergrad. degree; UGD = university grad.degree; bU = unemployed; R = retired; PT = part-time; FT = full-time; cPS = private sector; Gov = government; CA = conservation authority; Acad = academic; Ed = education; NGO = non-government or non-profit organization
Motives of Participation
0% 20% 40% 60% 80% 100%
TE (n=37)
AEC (n=39)
AMD (n=38)
GE (n=34)
GRR (n=37)
PE (n=38)
MO (n=37)
R (n=36)
Percent of Responses
VeryImportant
SomewhatImportant
NotImportant
Motives of OBBN participation (R = research; MO = meeting others with common interests; PE = performance evaluation (i.e., evaluating performance of water management programs; GRR = guiding rehabilitation or restoration; GE = Guiding enforcement; AMD = Assessing or managing biodiversity; AEC = Assessing/managing ecological condition; TE = Training/education)
*
Perspectives on Network
Implementation (I)
0% 10% 20% 30% 40% 50% 60% 70%
Choice of sampling sites (n=35)
Choice of data shared (n=34)
Developing and refining methods (n=32)
Analysis and interpretation (n=32)
Follow-up action (n=29)
Percent of Responses
Full control (5)
4
3
2
No Control (1)
Relationship Type (based on degree of participant control)a Control Partnership Collaboration Co-optation Who determines monitoring protocol? Participants Shared Shared Government Who selects sites to be monitored? Participants Participants Shared Government Who determines analytical methods, interpretation, and data distribution?
Participants Participants Shared Government
Who determines follow-up action? Participants Participants, then government
Shared Government
Types of government-participant relationships in monitoring programs (adapted from Savan et al. 2004).
Participants’ perceived control or influence over components of the OBBN
• 88% categorized participant-government relationship type as partnership or collaboration
**
Benthos: From Snot Globules to Jewelry
Caddisfly larva (Hydropsychidae)
Anterior view of water-boatman head (Corixidae)
Mayflies
True Flies
Black Flies
Caddisflies
Leeches
Dragonflies & Damselflies
Biocriteria and the Reference Condition Approach
Biocriteria“Healthy is Variable.”
–Dr. Robert Bailey, University of Western Ontario
(Kilgour et al. 1998, Bowman & Somers 2005)
StreamSampleDatePartnerHYDRACARINATrhypochthoniidae 2 1EPHEMEROPTERABaetidae 81 49Ephemerellidae 1 2PLECOPTERALeuctridae 1 1Capniidae 1 0Perlodidae 6 5Chloroperlidae 0 1TRICHOPTERARhyacophilidae 2 1Hydropsychidae 2 3COLEOPTERAElmidae 11 20DIPTERAChironomidae 20 29Ceratopogonidae 3 2Tipulidae 4 6Simulidae 0 2Empididae 1 0
Total: 135 122
• 2 equally healthy sites may have different biological assemblages
• Need to determine what normal is• Biomonitoring conundrum: Is an
observed difference greater than expected by chance? Is it biologically meaningful?
• Biocriteria are critical values for hypothesis tests
• The “normal range” is a pragmatic biocriterion
Baxter BaxterRiffle 1 Riffle 2
16-Aug-04 16-Aug-04ORCA ORCA
Biocriteria“Healthy is Variable.”
–Dr. Robert Bailey, University of Western Ontario
(Kilgour et al. 1998, Bowman & Somers 2005)
• 2 equally healthy sites may have different biological assemblages
• Need to determine what normal is• Biomonitoring conundrum: Is an
observed difference greater than expected by chance? Is it biologically meaningful?
• Biocriteria are critical values for hypothesis tests
• The “normal range” is a pragmatic biocriterion
StreamSampleDatePartnerHYDRACARINATrhypochthoniidae 2 1EPHEMEROPTERABaetidae 81 49Ephemerellidae 1 2PLECOPTERALeuctridae 1 1Capniidae 1 0Perlodidae 6 5Chloroperlidae 0 1TRICHOPTERARhyacophilidae 2 1Hydropsychidae 2 3COLEOPTERAElmidae 11 20DIPTERAChironomidae 20 29Ceratopogonidae 3 2Tipulidae 4 6Simulidae 0 2Empididae 1 0
Total: 135 122
Experimental Designs for Bioassessments
Monitoring for WhereNo
Monitoring for WhenYesNo
Temporal (Before-After)No
Optimal Impact Study (BACI)YesYesNo
Modern Analog ApproachNo
Reference Condition ApproachYesNo
Impact from Spatial PatternNo
Spatial Study (Control-Impact)YesYesYes
Experimental Design NameIs there a control area?
Is when and where known?
Has the impact occurred?
Monitoring for WhereNo
Monitoring for WhenYesNo
Temporal (Before-After)No
Optimal Impact Study (BACI)YesYesNo
Modern Analog ApproachNo
Reference Condition ApproachYesNo
Impact from Spatial PatternNo
Spatial Study (Control-Impact)YesYesYes
Experimental Design NameIs there a control area?
Is when and where known?
Has the impact occurred?
(Adapted from Green 1979 [Bowman and Somers 2005]; see also Underwood 1997)
History of the RCA
• A product of researchers working on the common challenge of studying an environment where an impact had (or was likely to have) occurred, but when and where the impact occurred were not known
• UK: RivPACS, Australia: AusRivAS, Canada: BEAST
• U.S.: Rapid-Bioassessment Procedures
(Wright et al. 2000, Bailey et al. 2004, Barbour et al. 1999, Bowman and Somers 2005)
Reference Condition Approach (RCA)
“Long-term monitoring programs…provide the measures of normal (reference data) against which the abnormal is judged. It is impossible to convince a court that something is wrong if ‘right’ is not defined.” – MOEE Biomonitoring Review Committee, 1994
Reference site
Test site
Multiple, minimally impacted control sites define the normal range of biological conditions to be expected at a test site
RCA StepsThe RCA has the following 5 steps (Bailey et al. 2004):
1. Minimally impacted reference sites are randomly selected and their biological communities and habitats are characterized.
2. Reference sites are grouped according to the similarity of their biological assemblages and/or habitats (depending on the approach used, a model that predicts a test site’s reference-state assemblage type, hence its reference-site group membership, may be built using a set of natural-habitat or physiographic attributes that are known to distinguish assemblage types).
3. A test site is sampled to characterize its biological community and habitat.
4. Appropriate reference sites are selected to define the normal or expected test-site condition.
5. Statistically test the bioassessment null hypothesis (i.e., that the test site is in reference condition).
Sample benthos and habitat at a variety ofrandomly selected, minimally impacted reference sites
Summarize the biological condition of reference sites. Group reference sites having similar biological communities.
Build a statistical model that predicts group membership based “niche variables” (physiographic variables that account for separation between groups)
Sample the biological community of a test site and characterize its niche attributes. Summarize biological condition using a set of metrics
Use physiographic model to predict test site to a reference group.
No
NoYes
Site likely unimpaired. Resample periodically and confirm reference
group selection
Site may be impaired. Confirm reference group selection and resample.
If same result, investigate for causes of
impairment
Establish normal range of biological condition for test site using appropriate reference site group ( ref±2SD)
Suitable reference site group available?
Yes
Biological condition of test site is within normal range?
RCA Steps
RCA Messiness
-Different definitions of minimal impact, reference site classification methods, summarization and hypothesis-testing procedures (e.g., Wright et al. 2000, Linke et al. 2005).
-Different researchers have different approaches to each step (Bowman and Somers 2005)
RCA Step-1 Challenges : Reference Sites and Minimal Impact
1. Minimally impacted reference sites are randomly selected and their biological communities and habitats are characterized.
• “Sites that are not disturbed by human activities are ideal reference sites; however, land-use practices and atmospheric pollution have so altered the landscape and quality of water resources … that truly undisturbed sites are rarely available (Barbour et al. 1996). ”
• Standard criteria for minimal impact don’t exist• It is particularly difficult to find reference sites for large waterbodies and for any
waterbodies in areas where climate and geography favour agriculture or urban development
• randomly selecting reference sites may be difficult because of their restricted and aggregated spatial distribution, and because of their remote location and difficult access (Hughes 1995).
Reference Site Criteria: Wyoming
(U.S. EPA 1996)
Different weights for different attributesDifferent thresholds for different eco-regions
OBBN: Qualitative Definition of Minimal Impact
CRITERIA FOR “MINIMALLY IMPACTED” Well downstream of significant point sources Minimal regulation of water level (minimal affect from dams and impoundments) Extensive naturally vegetated buffer Well forested catchment Minimal development or urban land use in catchment Minimal agricultural land use in catchment Minimal impervious cover and artificial drainage in catchment Minimal anthropogenic acidification (i.e. pH matches expectation based on local geology) Water chemistry better than regulatory guidelines, e.g. Ontario Ministry of Environment PWQO’s (REF PWQO)
(Jones et al. 2004)
RCA Step-1 Challenges: What is a reference site?
X
OBBN Approaches to RCA Step 1• Sample a wide range of sites (but also ensure relevance to test sites)
• Reserve at least 10% of annual sampling effort for reference site re-sampling (same sites each year, or different sites in different years, or a combination of the two strategies)
• Ideally, sample enough reference sites to adequately describe the normal ranges of different types of waterbodies (~30 sites per group; Bowman and Somers 2005)
• Where insufficient reference sites exist, estimate normal range using best available sites, modeling, and applying best professional judgment.
Remember: • We don’t know how many assemblage types there are
• Try to sample some unusual sites (e.g. large rivers, clay plain streams)
OBBN Approaches to RCA Step 1
• Standard methods required: location, taxa counts, habitat data
• OBBN Coordinators provide QC checks on reference-site samples; confirmed taxa enumerations and physiographic data returned to collector
• Depending on question, impacted sites may be used in bioassessments; however, minimally impacted sites are always useful for determining relative condition
Use of Impacted Control Sites
CA1
CA2
Urban mine-impacted test site
Urban control site
Minimally impacted reference site
(e.g., Reynoldson et al. 2005)
? ?
(Hypothetical Data)
Use of Impacted Control Sites
CA1
CA2
Urban mine-impacted test site
Urban control site
Minimally impacted reference site
(e.g., Reynoldson et al. 2005)
!
(Hypothetical Data)
Send Reference
Site Samples(But Not
Like This)
RCA Challenges, Steps 2-4 : Sampling Methods, Classification and Prediction
2. Reference sites are grouped according to the similarity of their biological assemblages and/or habitats (depending on the approach used, a model that predicts a test site’s reference-state assemblage type, hence its reference-site group membership, may be built using a set of natural-habitat or physiographic attributes that are known to distinguish assemblage types).
3. A test site is sampled to characterize its biological community and habitat.
4. Appropriate reference sites are selected to define the normal or expected test-site condition.
RCA Challenges, Steps 2-4
• No agreement on sampling methods (collection, sample processing, taxonomic resolution)
• No agreement on data summarization (multivariate, multi-metric, hybrid)
• Difficult to know a priori which habitat attributes (and scale) to measure
• Numerous questions about classification: – Method (a priori vs a posteriori, statistical methods)? – # of groups? – # of sites per group? – Habitat measures to match ref and test sites?
Grouping reduces residual variation among reference sites and increases power of assessment BUT:• It goes against our knowledge that communities change continuously across environmental gradients• How many groups are there? CA1
CA
2Reference Gp. 1
Reference Gp. 2
Reference Gp. 3
Test
(Gerritsen et al. 2000)
ReferenceSites
Test Site
Why Classify?
Grouping reduces residual variation among reference sites and increases power of assessment BUT:• It goes against our knowledge that communities change continuously across environmental gradients• How many groups are there? CA1
CA
2Reference Gp. 1
Reference Gp. 2
Reference Gp. 3
Test
(Gerritsen et al. 2000)
ReferenceSites
Test Site
Why Classify?
2 main ways to group sites: a priori and a posteriori
a priori a posterioriGrouping method
Groups based on assumptions about factors that determine community composition (e.g., ecoregion); May under- or over-estimate # of groups because assumptions about deterministic factors may be incorrect; within- and between-group variance may not be optimal
Biological community composition dictates group; # of groups tends to make more biological sense
Prediction Easy; if you know the habitat attributes you know the group
Can be tricky because not all between-group variation can be explained and because deterministic factors may not be adequately measured
Different Approaches to Classification
Messiness in Classification
Different reference-site classification methods will result in different models of reference condition (e.g., Wright et al. 2000, Bowman and Somers 2005)
A1A2A3A4
A5
A6
A7
A8
A9
B1
B2B3
B4
B5
B6
B7
B8
B9
C1
C2C3C4
C5C6C7
C8
C9
D1D2
D3
D4
D5D6
D7D8
D9
CA
A9
A8
A7
A6
A5
A4A3A2
A1
D8
D4C4
B9
B8
B7
B6
B5B3B2
C1B4
B1D2
D1
C9
C8
C7C6C5
C3C2
D3
D5D6
D7
D9
TWINSPATWINSPANC
A9
A8
A7
A6
A5
A4A3A2
A1
B9
B8
B7
B6
B5B3B2
B1
C9C7C6C5
C4C3C2
C1B4
D3
D9
D8 D7D6
D5
D4
D2D1
C8
UPGMAD
A9
A8
A7
A6
A5
A4A3A2
A1
C3
B9
B8
B7
B6
B5B3B2
B1
D3
C9C7C6C5
C4C2
C1B4
D9
D8 D7D6
D5
D4
D2D1
C8
Ward'sE
A8
A6
A5
A4A3A2
A1
B9
B8
B7
B6
B5
B4
B3B2
B1
C3D3
D7D6
C8
C7C6C5
C4C2
C1
D9
D8 D5
D4
D2D1
C9
A9
A7
K-meansF
A1
A2A3A4
A5
A6A7
A8A9
B1
B2B3
B4
B5
B6B7
B8
B9
C1
C2 C3C4
C5C6
C7C8C9
D1 D2
D3
D4
D5D6 D7D8
D9
NMDSBA
PCO1
PC
O2
PCO1
PC
O2
Further Messiness in Classification
A 2-axis Principle Coordinates Analysis ordination plot showing a seemingly appropriate set of 22 reference sites defining an assemblage type (left), and an alternate classification (right) of two groups of 20 sites that results from adding additional data for an assemblage type that was under-represented in the solution shown at left. Ellipses represent 90% confidence bounds for each assemblage type. Hypothetical data: Group 1 sites (diamond symbols) were simulated as randomly distributed variables (mean PCO1 = 1, mean PCO2 = 3.5); group 2 (squares) had mean PCO1 = 4 and PCO2 = 1. The standard deviations for PCO1 and PCO2 values was 1 for both groups.
OBBN Approach, Steps 2-4• Balance standardization with flexibility• Classification-free reference-site matching: Nearest Neighbour• Sampling is more than just collecting bugs: in data-driven approach, niche
variables used to select reference sites for test sites• Habitat characterized with site-, reach-, and catchment-scale measures• To summarize biotic composition, a variety of indices should be used,
because each summarizes and emphasizes different patterns in the assemblage. Further guidance may be given as we learn more about responses to stressors in different parts of the province
• Analytical software defaults will reflect current knowledge and recommendations
• Selecting reference sites will be automated by OBBN/CABIN database• Refining models is a research priority
Classification vs. Nearest Neighbour
Predictor 1
Pre
dic
tor
2
ClassificationApproach
Nearest-Neighbour orClassification-free Approach
Predictor 1
Pre
dic
tor
2
(Simulated Data)
RCA Challenges, Step 55. Statistically test the bioassessment null hypothesis
(i.e., that the test site is in reference condition).
• How much deviation from normal is ecologically significant? What level of confidence is required?
• Hypothesis-testing methods differ in the way they implicitly define “health” or biological integrity, in their assumptions, in their manner of quantifying biological condition and effects, in the format of their outputs, and in the predictability of their response to stress (Norris and Hawkins 2000)– U.K. and Australia: Ratio of expected-to-observed taxa richness, (e.g., Davies
2000 and Moss 2000)
– U.S.: Multi-metric scores, with biocriteria set using regional reference sites (e.g., Barbour and Yoder 2000);
– Canada: Ordination-axis-scores compared against confidence ellipses for reference sites (e.g., Reynoldson et al. 2000).
Ecologically Significant Effect
• When testing bioassessment hypotheses (H0: test site normal), critical effect size must be defined a priori
• Central test (H0: no difference) not biologically meaningful or management-relevant
• OBBN-recommended: 95% of reference site distribution …but need to consider Type I (false positive) & Type II (false negative) error rates and their consequences
(Bowman and Somers 2005, Jones et al. 2004)
Biocriteria Messiness: Error
Rate and Effect Size Considerations
Null Hypothesis
Decision True False
Reject H0Type I error
(false positive, α)
Correct decision
Accept H0Correct decision
Type II error
(false negative, ) (From Bailey et al. 2004)
Biocriteria: Summary of Key Points• Biocriteria: critical values for testing bioassessment null hypothesis (H0:
test site normal)• Confidence in bioassessment decision (i.e., pass or fail) depends on how
well we model normal range, and therefore how well we estimate probabilities of false positives and false negatives
• Setting biocriteria means trade-offs between Type-I and Type-II error rates: consider the consequences of these errors (management responses and costs)
• There is no magic α-level• Determining Type-II error rate requires a set of observations that are
known to deviate from normal by a specified effect size (this requires simulated data)
(Bailey et al. 2004, Jones et al. 2004, Bowman and Somers 2005)
OBBN Approach, RCA Step 5• Use data from the same season• Test Site Analysis (TSA; Bowman and Somers 2005,
2006a, and 2006b) is recommended method for testing bioassessment null hypothesis; Represents a convergence of multivariate and multi-metric methods:– Multiple indices are used to summarize composition – A non-central multivariate equivalence test (e.g., McBride 1993)
is calculated using all indices and considering redundancies among the summary indices (test statistics include D, F, and p)
– Why not just count-up individual passes and fails?– If the site fails, a discriminant analysis is done to describe the
effect size associated with each of the indices used in the equivalence test thereby characterizing the test-site’s response signature.
OBBN Approach, RCA Step 5
• OBBN recommends 95th percentile of reference-site distribution as biocriterion (but need to consider error rates and power appropriate for specific studies
• This step will ultimately be automated by OBBN database
Summary Index 1
Su
mm
ary
Ind
ex
2
Reference
Test
Centroid
(Simulated Data)
Does our Site Pass?
Cumulative Probability
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Bray-Curtis Distance
Per
cen
tile
(Simulated Data)
Walker's
Keast
Black
Sheldon 2SheldonPine
Nott2
Noisy
Nott
Centre
Silver
Misc. Diptera
Simulidae
Tipulidae
Ceratopogonidae
Chironomidae
Gastropoda
Coleoptera
Trichoptera
Megaloptera
PlecopteraAnisopteraEphemeroptera
Mites
Decapoda
Amphipoda
IsopodaHirudinea
OligochaetaTurbellaria
-8
-6
-4
-2
0
2
4
6
8
10
-4 -3 -2 -1 0 1 2 3 4
columns
rows
50% Ellipse
75% Ellipse
95% Ellipse
99.9 Ellipse
RCA Bioassessment Example
RCA Bioassessment Example
WillowPine Everett
Pine River (Mulmur)
mad River
Boyne River
Teesw ater RiverNorth Saugeen River
Penetangore North
Penetangore South
Misc. DipteraSimulidae
Tipulidae
Ceratopogonidae
Tabanidae
Chironomidae
Gastropoda
ColeopteraLepidoptera Trichoptera
Megaloptera
Hemiptera
PlecopteraZygoptera
Anisoptera
Ephemeroptera
Mites
Decapoda
Amphipoda
Pelecypoda
Isopoda Hirudinea
Oligochaeta
Turbellaria
-4
-3
-2
-1
0
1
2
3
-4 -3 -2 -1 0 1 2 3 4
columns
row s
50% Ellipse
75% Ellipse
95% Ellipse
Sampling Methods
Sampling Protocols
Standardization vs. FlexibilityBiomonitoring
Component Recommendation
Study Design Reference Condition Approach
Benthos Collection Method
Travelling-Kick-and-Sweep (where possible); replication in lakes and wetlands, sub-sampling in streams
Mesh Size 500 m Time of Year Any season; assessment comparisons use data from the same season
Picking In lab (preferred) or in field (optional); preserved (preferred) or live (optional), microscope (preferred) or visually unaided (optional); random sub-sampling using Marchant Box (preferred) or Bucket Method (optional) to provide a minimum 100-animal count per sample
Taxonomic Level Mix of 27 Phyla, Classes, Orders and Families (minimum); Family (preferred); Genus/Species (optional, recommended for reference sites)1
Analysis (Bioassessment Hypothesis Testing)
Test Site Analysis (TSA; see Appendix 9): Mahalanobis distance (e.g., Legendre and Legendre 1998) calculated across selected summary metrics; non-central significance test to determine if biological distance between test site and reference site group mean is larger than a specified effect size; if the null hypothesis (H0: │Dtest – Dreference mean │≤ critical effect size) is rejected, use discriminant function analysis to identify metrics contributing most to the separation between the test site and reference condition
Protocol Instruction Format
1. Sampling unit/inference
2. Replication
3. Benthos collection methods
General Comments:1. Some protocols require evaluation and may be updated2. There may be situations in which protocols will not work as
written. In this case, adapt as necessary3. If time or property access limit ability to apply techniques, collect
what you can. Some information is better than none4. Obtain landowner permission5. Avoid sensitive times (e.g., fish spawning) and sensitive habitats6. Adjust sampling effort if experience shows a habitat to have
exceptionally high or low benthos densities
(Excerpt from Protocol Manual)
Sub-sampling vs. Replication
• Sub-sampling: “In some experimental situations, several observations may be made within the experimental unit … such observations are made on sub-samples of sampling units. Differences among sub-samples within an experimental unit are observational differences rather than experimental unit differences”
• Replication: “When a treatment appears more than once in an experiment, it is said to be replicated.”
(Steel and Torrie 1980)
Lakes • Sampling Unit• Replication• Collection method
Replicate #1
Replicate #2
Replicate #3
Transect
1 m depth contour
Lake Segment (sampling unit) • Sampling unit is
“lake segment”
• 10 minute traveling kick and sweep along transects
• 3 replicates collected
Streams• Sampling unit• Alternate
definitions (pg. 21)
A
B
Cross Section A-B
A B
Top of both banks approximately same height from water surface
Channel Mid Line
Thalweg
Cross-over Point
Sampling Reach Boundary
Flow Direction
Streams • Replication & collection methods
Transect Traveling Kick and Sweep
Flow
Optional Transect
Sampling Location
Sampling ReachBoundary
• Samling unit encompasses 2 riffles and 1 pool (often meander sequence)
• 2 transect subsamples in riffles, one in pool
• ~ 3 minute, 10 m kick
Pool
Riffle or
cross-over
Riffle or
cross-over
Riffle or
cross-over
Pool
Applying Traveling Kick and Sweep in Large or Small Streams
Flow
TransectSampled portion of transect
Current Speed Distribution1 2 3 4 5
Stratum boundary
Flow
Transect
Supplementary Transect
Pool
Riffle
Riffle
Streams: Grab Sampling
Ekman, Ponar or other grab sample
Sampling ReachBoundary
Flow
OptionalTransect
• Sampling unit encompasses 2 riffles and one pool (meander sequence)
• 2 transects in riffles, 1 transect in pool• Each subsample is a composite of 3 (or more) grabs
Pool
Riffle or
cross-over
Riffle or
cross-over
Riffle or
cross-over
Pool
Traveling Kick Transect
Stovepipe Core Sample
Jab and Sweep Sample
1 m depth contour
2 m depth contour
Wetland Segment (replicate)
• Sampling Unit• Replication• Collection Methods
Wetlands
Wetlands: Selecting Collection Method
Water Depth
Substrate Type
Plant Density
Recommended Gear
Recommended Technique
0.15-1 m Stable (e.g., sand/gravel)
Low D-net Traveling kick and Sweep
0.05-1 m Soft (e.g., organic, muck)
moderate D-net Jab and Sweep
<0.05 m or saturated soils
Soft to moderately stable
Any Stovepipe Corer Core
Summary of Collection Methods
Collection Method Streams Lakes Wetlands
Traveling kick and sweep; standard method for wadeable habitats Grab samples (Ekman Dredge, Ponar Grab, or similar); option for deep water sites O OJab and Sweep; option for wadeable, sparsely vegetated, soft sediments OCoring; option for deep or very shallow water (especially in shallow wetland soils) O O
Artificial substrate; option for atypical habitats or special studies O O O
Sampling Groups
1 2 3Gerry Sullivan Angela Wallace Lisa Campbell
Christine Spedalieri Nancy Harrtrup Robin TapleyChris Brown Rebecca Scobie Valerie StevensonTrevor Middel John Haselmayer Scott Parker
Ben Jewiss Suzanne Partridge Liisa KearneyCassandra Borm Alana Nunn Julie Hordowick
4 5Lynette Dawson Rebecca Crockford
Beth Gilbert Rajesh BejankiwarMarnie Guindon Erin McGauley
Diana Tyner Sara KellyDebbie DePasquale Carolyn Paterson
Vince D'Elia Josh Hevenor
Sample Processing
• Sieve
• Sub-sample
– Marchant Box (preferred)
– Bucket method
• Sort carefully (Optional: microscope or magnifier)
• Identify and tally (taxonomic level matches training)
• 100 count (minimum)
• Preserve and archive sample
Sample Processing: Transporting to Lab
• Sieve in net in field
• Release non-benthos
• Keep live samples cool
• Label transport containers inside and out (date, location, sample number, etc.)
Sample Processing: Sieving
• Must be done to remove fines
• Preliminary done in field, thorough done in lab
• 0.5 mm mesh sieve
• Remove large pieces (rocks, wood)
Sample Processing: Sub-sampling & Picking
• Need random sub-samples• 100-count but sort entire last sub-sample• Consider suction device if using Marchant Box• If using Bucket Method, estimate portion picked
by weight or volume• A bit of soap will sink floaters• Screen for fast moving• Sort thoroughly
Benthos ID: 27 Group Level
Coelenterata(Hydras)
Turbellaria(Flatworms)
Nematoda(Roundworms)
Oligochaeta(Aquatic Earthworms)
Hirudinea(Leeches)
Isopoda(Sow Bugs)
Decapoda (Crayfish)
Trombidiformes-Hydracarina(Mites)
Ephemeroptera(Mayflies)
Anisoptera (Dragonflies)
Zygoptera(Damselflies)
Amphipoda(Scuds)
Plecoptera(Stoneflies)
Hemiptera (True Bugs)
Megaloptera (Fishflies, Alderflies)
Trichoptera (Caddisflies)
Lepidoptera(Aquatic Moths)
Coleoptera (Beetles)
Gastropoda (Snails, limpets)
Pelecypoda (Clams)
Chironomidae(Midges)
Tabanidae (Horse and Deer Flies)
Culicidae(Mosquitos)
Ceratopogonidae(No-see-ums)
Tipulidae (Crane Flies)
Simuliidae(Black Flies)
Misc. Diptera (Misc. True Flies)
Version 1.0, revised March 2004Ontario Benthos Biomonitoring Network
Water Body Name: _________________________ Site #: ____________ Replicate #: ______ Date (mm/dd/yyyy) and Time: _________________________
Organization: _____________________________ Department_______________________ Address:_____________________________________________
Contact: ________________ Phone: _________________ E-mail: _____________________________ % picked for 100-count ______ # of vials: _________
Circle Method: (Sub-sampling) Marchant Box / Teaspoon (Location) Field / Lab (Preservation) Live / Preserved (Magnification) Microscope / Unaided
Sample Processing: Preservation
• Formalin or Alcohol can be used
• Small volumes can be discharged to septic system or municipal sewage system
• Safe storage
• Avoid poisonous denatured alcohols
• Replace formalin with alcohol after a couple of days
Habitat Characterization
Done for 2 reasons:1. Niche Attributes
2. Diagnosis
diagnostic useful in determining cause (often of biological impairment)
niche variable a natural habitat (often physiographic) variable that accounts for a significant portion of the difference in biological condition between reference site groups
Habitat Characterization (Table 10, Pg. 37)
Measured at site Measured remotely (GIS)
Location (latitude & longitude)
Organic matter, areal coverage
Elevation Riparian vegetation
Water temperature Canopy cover (%)
Dissolved oxygen, pH, conductivity, alkalinity
Aquatic macrophytes and algae
Maximum Depth Bank full width (m)
Maximum hydraulic head
Instantaneous discharge (m3/s)
Wetted width Perennial or intermittent
(presence of standing water)
Dominant substrate classes
Drainage area
Base Flow Index
Basin relief
Mean annual lake evaporationLength of main channel Mean annual precipitationMean Annual Run-offMean Annual SnowfallMaximum Watershed ElevationMean ElevationMaximum Flow Distance
Minimum Watershed ElevationMean Slope of Watershed Catchment PerimeterShape factorSlope of main channelTributary densityCatchment land cover (areal proportions of 28 land cover types) OrderAspectAreaPerimeterFetch
TSA
Insert TSA Section: Michelle Bowman
General Discussion/Review
Certification Test• Test is optional• Passing grade for both multiple-choice and benthos identification tests is
90%• For benthos identification test:
– Participants can use references– Trainers are not permitted to use references, and a correct answer
includes both the taxonomic group and at least 2 diagnostic characters
• Students cannot be immediately certified without a passing grade, but arrangements can be made for a re-test (you do not have to redo the course to re-take the test)