summary

Ecological Flows Science Advisory Board (EFSAB)
Meeting Summary – June 19, 2012
Archdale Building, Raleigh NC
X APPROVED for distribution

Attendance
Members
Donnie Brewer, Environmental Mgt Commission
Mark Cantrell, US Fish & Wildlife Service
Bob Christian, East Carolina University
Tom Cuffney, U.S. Geological Survey
Linda Diebolt, Local Governments
Chris Goudreau, NC Wildlife Resources Commission
Jim Mead, NC Division of Water Resources
Sam Pearsall, Environmental Defense Fund
Judy Ratcliffe, NC Natural Heritage Program
Jaime Robinson, NCAWWA-WEA
Fritz Rhode, US National Marine Fisheries Service (online)
Jay Sauber, NC Division of Water Quality
Bill Swartley, NC Division of Forest Resources

Alternates
Vernon Cox, NC Department of Agriculture
Sarah McRae, US Fish & Wildlife Service
Steve Reed, NC Division of Water Resources
Van Stancil, NC Wildlife Resources Commission
Fred Tarver, NC Division of Water Resources

NC Division of Water Resources
Don Rayno
Sarah Young

Guests (onsite)
Alex Cohn, TNC
Leigh Habegger, APNEP
Jennifer Phelan, RTI
Ian McMillian, DWQ
Breda Munoz, RTI
Erin Thompson, APNEP
Jason Williamfeld, NC Forest Service

Guests (online)
Craig Brombley
Bill Crowell
Lisa Gordon
Kyle Hall
Haywood Pthisic, LNBA

NCSU Cooperative Extension Facilitation Team
Mary Lou Addor, Natural Resources Leadership Institute
(NRLI)
Patrick Beggs (WECO) Watershed Education for
Communities and Officials
Christy Perrin (WECO) Watershed Education for
Communities and Officials
Nancy Sharpless, Natural Resources Leadership Institute
(NRLI)

In memory of our friend, Steve Reed
This summary comes out following the passing of our friend and colleague, Steve Reed, NC DWR.
The facilitation team expresses our sincere condolences to his friends and colleagues. We will miss
him and his leadership in the EFSAB process, and will do our best to support DWR and the EFSAB in
meeting their goals as we move forward.

The purpose of the Ecological Flows Science Advisory Board:
The Ecological Flows Science Advisory Board will advise NC Department Environment and Natural
Resources (NCDENR) on an approach to characterize the aquatic ecology of different river basins
and methods to determine the flows needed to maintain ecological integrity.

Presentations, reports, and background information about the E-Flows SAB are available at:
www.ncwater.org/sab The presentations from the June meeting are available at
http://www.ncwater.org/Data_and_Modeling/eflows/sab/presentations/20120619/

The EF SAB will meet August 28 at the Stan Adams Education Center at 9:00am.
Address and map are on the last page.
EFSAB Meeting Summary - 6/19/2012 Page 2 of 31

I. Executive Summary
Comparison of 3 Current Studies (Jim Mead)
Jim provided a hand-out that clarified and compared three external studies that were originally
scheduled for presentations on the agenda. (Item 3 was postponed for the August meeting.)
1. Bio-Fidelity Test Hydrologic Stream Classification
2. RTI Internal Research & Development Project
3. The Nature Conservancy’s Four-Basin Environmental Flow Project

Update: Biological Fidelity Analysis of Stream Classes (Jennifer Phelan)
RTI is undertaking a study with the Environmental Defense Fund. An update was given. The
project objective is to adopt a stream classification system that represents the distribution of
aquatic biota in North Carolina, by evaluating the 7 EFS stream classes, comparing fidelities of
aquatic biota to different stream classification systems (EFS and McManamay et al., 2011), and
determining if the current EFS stream classes need to be modified to more accurately describe the
distribution of biota. RTI intends on including 578 NHD+ catchments (217 mountain, 231
Piedmont, 130 coastal plain) in the biofidelity analyses. These catchments were selected based on
“minimally altered” water quality and flow condition status and presence of at least one biological
monitoring station (NC Benthic Macroinvertebrate, NC Fish Community and/or Natural Heritage
Inventory programs). Aquatic bioata data from the 1981-2010(30 year) period of record will be
used in the fidelity analyses. Stream classes will be determined using 40 years (1967-2006)
WaterFALL hydrologic data and current instream flow alterations (withdrawals, discharges and
dams). Jennifer shared the results of a preliminary Random Forest statistical analyses testing the
fidelity of individual species to the 7 EFS stream classes. These preliminary analyses were
conducted on 106 of the 185 NHD+ catchments used to develop the EFS stream classes. The
fidelities of benthic macroinvertebrates (presence- absence) were assessed in the analyses.
Although some species showed some degree of fidelity to some of the stream classes, the results
were inconclusive due to the small and uneven number of records for each stream class, the
potential confounding influences of water quality and instream flow alteration (analyses were not
restricted to minimally altered NHD+ catchments), and the uneven and generally low number of
records for each species.
Macroinvertebrate Assemblage Correspondence to Hydrologic Classes & Discussion (Tom
Cuffney)
Tom discussed how he addressed some of the analysis issues: a) 1 class with only 1 site; b)rare
taxa, c) data collected from multiples seasons, there is not one index period, d) data is ordinal,
e)ambiguous taxa, f) what is lowest taxa level you want to represent in data.
After addressing those issues within the data, he conducted 2 analyses, using the same
macroinvertebrate dataset (106 NHD+ catchments) as RTI:
EFSAB Meeting Summary - 6/19/2012 Page 3 of 31

ANOSIM (Analysis of Similarity) hypothesis: there is no difference between the assemblages of
classes. Results: For the qualitative data set without Class E, it is not significant, no difference. For
the qualitative (presence/absence data), it is significant, but with low values.

He identified another question to consider: is there a hydrologic classification that optimizes the
correlation to the distribution of assemblages?
Based on this question he conducted additional analyses: Indicator value index and Phi Analysis for
Fidelity. Results: for both species and genus, quantitative and qualitative analyses did show
statistically significant association.

Fidelity analysis conclusions:
 We can extract taxa that are indicators of hydrologic clusters
 The key question is whether the clusters are optimal as indicators, if not then fidelity
analyses are not informative
 Identifying optimal clusters is really important.

The data used for these analyses included all benthic macro invertebrate sites, including impaired
sites, which introduces a lot of variability. These analyses could be repeated using the data
associated with the 578 minimally altered NHD+ catchments identified and provided by RTI.

Discussion about Fidelity testing
Some main discussion points included:
 the desire to overlay a category such as ecoregion to the hydrologic classification to help
develop subsets of classes.
 It’s a challenge to fit the biological systems with the hydrologic systems
 The work to link biology to stream classes is at the forefront, won’t be perfect but will
improve over time
RTI internal project: Flow Alteration-Biological Response Relationships to assist with the
determination of ecological flows & small group discussion
Jennifer Phelan provided an overview of RTI’s approach of using
 Unaltered/Historic and current flows determined with WaterFALL model and a
modification/subset of the TNC Indicators of Hydrologic Alteration (IHA) flow metrics
 Large state-wide NC fish community dataset
 Potential inclusion of other influential factors in the analyses (e.g., habitat, interspecific
interactions, preceding year climatic events, water quality).
to determine species-specific flow alteration-biological response relationships that are useful to
water managers. She presented the criteria used for selecting fish species to include, the criteria for
selecting flow metrics, and other influential factors (and sources) that could be included in a
multiple regression analysis. She provided handouts with the resources used for selecting fish
species, a table of some of the fish species, and the flow metrics that will be used in the analyses.

General suggestions from EFSAB included:
 Abandon the parameter of “presence on 303(d) list”, since not all listings are created for the
same reasons. Using some of the variables leading to 303(d) listing makes more sense.
 Consider grouping the species into guilds, and adopt guild metrics (species diversity, total
count, and.or diversity index (e.g., Shannon-Weaver) as the biological response variable.
 Consider including 7Q10 as a flow metric
EFSAB Meeting Summary - 6/19/2012 Page 4 of 31

 The EFSAB would like to see the guild species list
 Consider temperature as a water quality variable

The EFSAB broke into small groups to answer specific questions. See main section for responses.

Revisiting the DWR Trial Balloon Discussion
The EFSAB responded again to DWR’s proposals to:
Stop using >120% of habitat available under a flow regime in habitat modeling
There appeared to be general support from the group to keep the > 120% scenario in winter
analyses. The majority of EFSAB are okay with dropping the >120% analysis in the other seasons,
with a few members wanting to keep it in spring as well as winter, and with a strong preference
from a member to keep the >120% analysis in all seasons.
Reduce the number of flow scenarios analyzed, based on the table presented in April.
The EFSAB agreed to the table of proposed flow scenarios for DWR to analyze, with the exception
that 7Q10 should be analyzed later as it’s a commonly used flow standard employed by other states.
The EFSAB does not need to see it included in the results presented to them.

Table of Contents / June 19, 2012 Agenda (click to go to section)

I. Executive Summary .................................................................................................................................................. 2
II. Introduction ................................................................................................................................................................. 5
III. April 24, 2012 Meeting Summary ....................................................................................................................... 5
IV. Comparison of 3 current studies ......................................................................................................................... 5
V. Update - Biological Fidelity Analysis of Stream Classes ............................................................................ 7
VI. Macroinvertebrate Assemblage Correspondence to Hydrologic Classes ......................................... 13
VII. Discussion ................................................................................................................................................................... 18
VIII. RTI internal project: Flow Alteration-Biological Response Relationships to assist with the
determination of ecological flows ..................................................................................................................... 20
IX. Small Group Discussion ......................................................................................................................................... 26
X. Revisiting the DWR Trial Balloon Discussion .............................................................................................. 28
XI. Upcoming Agenda Discussion ............................................................................................................................ 30
XII. Directions to August Meeting ............................................................................................................................. 31

EFSAB Meeting Summary - 6/19/2012 Page 5 of 31

II. Introduction
May Lou Addor welcomed everyone to the 13th meeting of the NC Ecological Flows Science Advisory
Board. Attendees, in person and online, introduced themselves and their affiliation. The EFSAB
was reminded about the ground rules, the process for raising questions, and reminded to project
their voice loudly to for the room microphone to be able to pick it up. Attendees were reminded
that the session was being recorded.

III. April 24, 2012 Meeting Summary
The April 24, 2012 meeting summary needs a few editorial changes, such as the meeting location
was incorrect. The meeting summary for April 24, 2012 was approved upon correction of those
changes.
All approved Meeting Summaries of the EFSAB are located at: www.ncwater.org/sab

IV. Comparison of 3 current studies
Jim Mead summarized the 3 concurrent studies taking place that will aid the EFSAB.
We previously discussed two potential pathways to determine ecological flows:
1. habitat modeling, and
2. incorporating biological data into flow mechanisms.
At the April EFSAB meeting, we focused on habitat modeling approach again. This month we are
talking mainly about the biological approach. Studies are underway by RTI and The Nature
Conservancy that we will all benefit from. The Nature Conservancy project will be presented at our
August meeting. The following table summarizes the differences & similarities between the 3
projects:

EFSAB Meeting Summary - 6/19/2012 Page 6 of 31

Bio-Fidelity Test
Hydrologic Stream Classification
RTI Internal Research &
Development Project
The Nature Conservancy’s
Four-Basin Environmental Flow
Project
Wa
t
e
r
F
A
L
L

Hy
d
r
o
l
o
g
i
c

Mo
d
e
l
i
n
g

 30 to 40-year climate data
 2006 land cover
 Minimize upstream flow alteration
and water quality effects by site
selection

 30 to 40-year climate data
1. 2006 land cover, plus any flow
alterations
2. Potential natural vegetation (PNV)
or 1970 land cover, and no flow
alterations

 30 to 40-year climate data
1. 2006 land cover, plus any flow
alterations
2. 1970 land cover and no flow
alterations

Fl
o
w

An
a
l
y
s
i
s

 Stream Classification using EFS and
McManamay systems

 Compare #1 to #2 for degree of
flow alteration
 Focus more on mid- and low flow
metrics for water management

 Compare #1 to #2 for degree of flow
alteration
 Evaluate full range of flow metrics
Bi
o
l
o
g
i
c
a
l

D
a
t
a
 Benthos, fish & Natural Heritage
Program data
 Individual species (RTI) and
community-based analysis approaches
(USGS-Tom Cuffney)

 Fish: up to 20 species from a
“hydrology” guild
 Guilds based on Persinger et al.,
2010
 Fish metric based on count
 Fish – similar guilds to RTI
 Plus EPT, crayfish and mussels
 Fish metric based on count
Si
t
e
s

 Separate mountain, piedmont &
coastal
 Filter eliminates sites with upstream
flow alteration if total drainage area at
the monitoring site is less than twice
the total drainage area of the
upstream alteration.
 Within region, stratified by yes/no
upstream flow alteration
 For each region randomly select ~200
sites rated excellent to good-fair

 ~600 sites associated with fish
sampling data (NC Fish Community
and USGS NAQWA)
 Cape Fear, Tar, Roanoke and Little
Tennessee basins
 Focus on sites with multiple samples
over time
 Greater detail on fewer sites
Re
s
u
l
t
s

 How well do the stream classes
describe the spatial distribution of
aquatic biota (i.e., a higher probability
of a species or community being
present in one stream class over
another)
 Does the classification system need
revision?

 Ecological response curves:
x-axis =% flow alteration
y-axis = fish metric based on
species level count
 Uses space (multitude of sites with
varying amounts of flow
alteration) as surrogate for change
in flow in (same site)over time
 Ecological response curves:
x-axis = % flow alteration
y-axis = fish metric based on species
level count
 Uses flow changes over time from
multiple samples
 Also will include descriptive analysis
of basin conditions
Ti
m
e
l
i
n
e

 Complete by 9/30/12

 Complete by 9/30/12

 Complete by June 2013

EFSAB Meeting Summary - 6/19/2012 Page 7 of 31

V. Update - Biological Fidelity Analysis of Stream Classes

Jennifer Phelan of RTI International presented an update of their work on Biological Fidelity
Analysis of Stream Classes. The Environmental Defense Fund is funding this project. The slideshow
is available on the project website: http://www.ncwater.org

Outline of Presentation
 Background
 Current Status:
o Site selection
o Aquatic biota
o Stream classification
o Preliminary statistical analysis results
 Individual species
 Community assemblage
Project Objectives:
 To adopt a stream classification system that represents the distribution of aquatic biota in
North Carolina.
o Evaluate the 7 stream classes of the Environmental Flow Specialist (EFS) hydrological
stream classification system.
A. Coastal Streams
B. Small Stable Streams
F. Medium Stable Streams
C. Large Stable Streams
E. Large Piedmont Rivers
D. Small Flashy Streams
G. Small Seasonal
o Compare fidelities of aquatic biota to different stream classification systems
o If necessary, modify the EFS stream classes to more accurately describe the distribution
of biota
Study Methods:
 Aquatic Biota and Datasets:
o Fish (NCDENR DWQ Fish Community, USGS NAQWA, WRC Trout, WRC Diversity
(Gameland Surveys))
o Benthos (Aquatic species in Natural Heritage Program database)
o Aquatic species (NCDENR Natural Heritage Program database)
EFSAB Meeting Summary - 6/19/2012 Page 8 of 31

 Sites:
o 185 NHD+ catchments with USGS gages (used to develop EFS stream classes)
o Add additional sites for a final number of ~600 NHD+ catchments with stream classes
based on WaterFALL hydrologic data
 Aquatic biota data within each NHD+ catchment will be paired with the assigned stream
class
 Stream Classifications Systems (all three use flow metrics):
o EFS
o McManamay et al. (2011)
o Konrad (in review)
 Statistical Analyses:
o Random Forest (individual species)
o Similarity Analyses (communities)
Current Status:
1. Site Selection
2. Aquatic Biota
3. Stream Classifications
4. Preliminary Statistical Analyses

1. Site Selection
 185 NHD+ catchments with EFS stream classes. These sites paired with NC Benthic
Macroinvertebrate monitoring station data were used in th preliminary analyses that
will be presented today (a total of 106 of the 185 NHD+ catchments had monitoring
stations)Criteria for choosing the additional sites (for a total of ~ 600 sites) include:
o equal representation of 3 physiographic regions
o “minimally impaired” condition:
 Water quality: Excellent, Good or Good-Fair Benthos site condition
 Water quantity (instream flow alteration): total drainage area of
monitoring station is at least twice the size of the total drainage area of
the closest upstream flow alteration
o Within each physiographic region, proportional representation of streams with
and without upstream flow alterations (Example- if 20% of sites had upstream
flow alterations, then 20% of sites selected in a physiographic region also did)
o Random selection of qualifying sites
 Final number of sites (selection criteria applied to 185 NHD+ catchments + new sites):
o 578:
 217 mountains
 231 piedmont
 130 coastal plain
EFSAB Meeting Summary - 6/19/2012 Page 9 of 31

Question: looking at the map, is there a reason why the Albemarle Pamlico Sound plain is devoid of
sites.
Response: I imagine it is the availability of sites paired with the selection criteria.
Comment: There are very few sites out there that fit all the criteria.

3. Aquatic Biota
 ll monitoring sites within 578 NHD+ catchments:
o Fish, benthos, NHP aquatic species
o 30 years of data (1981-2010) assuming stream classes based on ~ 2010 condition
(the following table is not reflective of the date restriction- these will likely be
reduced.)
Aquatic Biota Number of Monitoring Sites
Benthos 636
Fish 238
Natural Heritage
Inventory
656

3. Stream Classification
 Stream classes based on WaterFALL hydrologic data:
o Generate for all 578 NHD+ catchments:
EFSAB Meeting Summary - 6/19/2012 Page 10 of 31

 2006 NLCD
 40 years of climate date (1967-2006)- 40 years is a length of time that
minimizes probability of stream class changes
 Instream flow alterations (dams, withdrawals and discharges – 1995-2012)
o Plan to generate these stream classes within the next couple of months
o Calibrate and QAQC stream classes with a comparison against classes determined by
USGS gage data
 Stream classification systems:
o EFS (7 classes), McManamay et al. (2011) (8 classes)
4. Preliminary Statistical Analyses
 Analyses:
o Random Forest (individual species)
 Presence/absence
 Similarity Analysis (communities)
 Presence/absence
 Abundance
 Dataset:
o Benthos (106 out of 185 NHD+ catchments with EFS stream classes)
o Taxonomic ambiguity resolved (by Tom Cuffney)
o Remove Parent – Keep Child (resolved to Genus level) dataset used in both sets of
analyses
 Number of NHD catchments for each stream class

Stream
Class
Number of NHD+
Catchments
A 14
B 41
C 6
D 30
E 1
F 8
G 6

EFSAB Meeting Summary - 6/19/2012 Page 11 of 31

Random Forest Approach
Random Forest is a decision-tree modeling and classification approach (Cutler et al., 2007)
A decision tree is a predictive model that uses a set of binary rules (yes/no) to split the data based
on the predictor variable

Example Table 1: Random Forest Analysis – An example of Biological fidelity to stream class

Example Table 2: Random Forest Analysis – An example of NO Biological fidelity to stream class

EFSAB Meeting Summary - 6/19/2012 Page 12 of 31

In the Example Table 1, for species 2 and 10 there is high fidelity to stream classes, in contrast to
Example Table 2, which shows that the class system is not working well.

Table 3: Preliminary Statistical Analyses

 Table 3 shows a subset of the initial results (13 of the 408 species included in the analysis,
shown because they illustrate some level of fidelity)
 Red shows a higher level probability of likelihood of occurrence
 Probability ranges from -1 to 1. The closer to 1, there is higher correlation to the species
being there. If the value is negative, there is no correlation.
 Only ~10% of taxa (41 out of 408) showed a higher fidelity to one stream class over another
 Some of the results may be an artifact of the low numbers of NHD+ catchments for each
stream class (1-41)
 If the number is small, it is positive correlation, but without much strength
 Values can be translated into probability values, though the values are low. Importance
value is a measure of occurrence of species in stream class and ability to predict stream
class by occurrence.
 Stream Class E has only 1 record (site), whereas stream class D has 30.
 Minimum number of NHD+ catchments for more robust analyses is ~ 40.
 36% of taxa had < 3 records
 Fidelities could be further confounded by including all benthos sites in the analysis
 The lowest level in this analysis is genus.

EFSAB Meeting Summary - 6/19/2012 Page 13 of 31

VI. Macroinvertebrate Assemblage Correspondence to Hydrologic Classes

This section summarizes Tom Cuffney’s (US Geological Service), presentation on his work. Rather
than looking at each genus or taxa, we’re looking at correspondence of assemblages in to the
hydrologic classes. We’ve talked about some of the analysis issues: a) 1 class with only 1 site;
b)rare taxa, c) data collected from multiples seasons, there is not one index period, d) data is
ordinal, e)ambiguous taxa, f) what is lowest taxa level you want to represent in data. When you talk
about things that are genus or species level, that is the lowest level that is represented in the data.
The data actually had data recorded for higher levels, so it’s not all genus or all species.
a) Regarding uneven distribution, you have several groups with multiple sites, but class E only has
one. We eliminated Class E because you can’t handle a class with one site when doing these
analyses. There is such a difference among classes, I did simulations where I randomly pulled
out 10 sites from ABD, and ended up data sets with all classes with all data classes. So data
used in analyses include all sites, all sites except E, and 10 random sites from A,B, and D.
b) Rare Taxa- along the y access is number of taxa in a group, X access is number of sites they
appear in. First one, 22% of taxa in data set appeared in only one site. 13% appeared in 2 data
sets. 55% of the taxa appear at 5 or fewer sites. There are a lot of rare taxa in here. What we
did here was to create additional data sets- a data set that had all the taxa in it, a data set that
had taxa that occurred in more than 5 sites, more than 15, and then I lost patience at 20.
c) Multiple seasons- most data is from July, Aug(~56% of it). I would up not doing anything with
this because I didn’t want to throw out the other 44% of the data. Bear in mind that these data
are from all months of the year, if you know about benthic you know some won’t be there all
times, or they may be there but may be too small to identify.
d) Ordinal Data - data produced by the state is ordinal data, its listed as absent, rare, common or
abundant. In quantitative analysis they are typically assigned 0,1,3, 10 when they are
calculating a benthic index, so that’s what we did.
e) Ambiguous parent- lets look at site 1, we’ve got data at order, genus, and 2 species within that
genus. We have some redundant information, if you have Baetas pluto (species), you know you
have Baetis (genus), Baetidae (Family) and Ephemeroptera (order). We want to eliminate as
much redundancy as we can. That’s what I mean when talking about resolving ambiguous taxa.
For this site 1, Ephemeropter and Baetidae are ambiguous parents.
Resolving ambiguous taxa and recoding data (tiny table). What we do is to remove ambiguous
parents > order level. We dropped it out.
For baetis where we had with 3 individuals, we divided them among 3 children in proportion to
their abundance. So in this example we added 1/11 to B. flavistriga to equal 1.3, and added
10/11 to B. pluto to equal 12.7. Once we had all that done, then we recoded it back to original
ordinal data (1,3,10).
We also did qualitative presence/absence recoding, so everything is just ones or zeros.
So now we have 2 more data sets- we have a lot of data sets.
f) Lowest Taxonomic level- genus and species level is lowest. When we look at the number of
ambiguous taxa, specifying your particular lowest taxa has an effect on that. If we specify the
species level we have 116 taxa that were ambiguous parents, that’s 19% of the total, 5,819 is
EFSAB Meeting Summary - 6/19/2012 Page 14 of 31

almost 18%. If we go to genus then it drops to 27 and only 6.6% that are ambiguous, and the
abundance drop to 2.6%. We’re really reducing amount of ambiguity in data sets.
So we added to the analyses one more analysis with species level, and one with genus level as
lowest.
Data preparation steps included:
1. Set lowest taxa level: Genus, Species
2. Took that data set and remove rare taxa: 0, 5, 10, 15, 20 sites
3. Remove ambiguous parents ≥ Order
4. Resolve ambiguous taxa:
Distribute abundance of ambiguous parents among children, proportionately
Resolve ambiguities separately for each sample- so it reduces the number of ambiguous
parents but does not remove them from the data set.
5. Ordinal and Qualitative (P/A) data sets
6. 10 site simulations for clusters A, B, and D

Analysis:
ANOSIM(Analysis of Similarity): which tests correspondence between hydrologic classes and
invertebrate classes
Indicator Value Analysis: after running ANOSIM I got interested in another analysis that identifying
indicator values that differentiate among hydrologic classes. Other methods for identifying
indicator values.
ANOSIM
Comes from a commercial package called Primer 6. In this case, hydrologic classes represent
treatments within the ANOVA. Analysis is based on resemblance (similarity or dissimilarity) of
assemblages between sites. It looks at ranks of those similarities between samples in the
underlying triangular similarity matrix. Looks at resemblances between sites, and looks at ranks of
those similarities in the underlying similarity matrix.
(Slide 12)Data Matrix (Site by species, with abundances in the table) is the upper one on the slide.
Can run similarity based on assemblages. For qualitative, used Kendal Rank Correlations on that,
since data was ordinal. For qualitative analysis (presence/absences) used a Sorenson Index. That
results in a site by site resemblance matrix so you get a sense of how similar those sites are in terms
of assemblages. That’s the data this ANOSIM works with. (Slide 13) It looks at the average within
and between classes of the assemblage matrix- that’s the colored areas represented within classes ,
(A vs A) then , like ANOVAs do, compares those with areas outside of that.
The Test statistic is R (slide 14) is actually the average rank similarity between classes, the diff
between that and the average rank similarity within classes divided by a constant based on the end
number of samples.
Hypothoses Ho: There is no differences between the assemblages of classes. R=0. This is
comparing permutations- is it different from a random permutation of the data? See slide 14 for
equations and variables.
IT does Global and Pair-wise tests:
Global test of significance: a significant result means there are differences somewhere that should
be examine further.
Pair-wise test of significance
EFSAB Meeting Summary - 6/19/2012 Page 15 of 31

 Extract pairs of classes
 Re-ranked
 Repeat test
ANOSIM results (slide 16): For quantitative data set, here is one with all sites and classes,
without Class E, ns means it’s all not-significant. On basis of quantitative analysis there is no
match, none randomly better. If you do it qualitatively, you do get significant results (Slide 17).
These are R values that vary from 0 to 1, most are pretty low (~ .20 for example). While better than
totally random, I wouldn’t classify them as very good.
All that effort put into doing those random 10 things with the other stuff, in final analysis was
wasted time.
Look at Pair-wise test for genus level, the yellow values are p values that are significant, 5% or or
less. What’s interesting is that there are 7 that are insignificant. 5 of the 7 are all associated with
class D. There is something problematic with class D.
Interpretation: Quantitative analysis shows no correspondence. It’s not statistically significant, it’s
random.
Qualitative: there is a statistically significant correspondence is relatively low.
Unanswered question: is there a hydrologic classification that optimizes the correlation to the
distribution of assemblages? This can’t be answered with what we are doing today. The
qualitative analysis shows it is better than random, but is it better than an alternative? We don’t
know.
So I took it further and did different kind of fidelity analysis to identify the taxa most strongly
associated with a-priori groups/hydrologic clusters. I used 2 different methods commonly used in
ecology.
Indicator value index (Dufrene and Legendre 1997): looks at indicator value for species I- max of
indicator value of species I for the different clusters. Each indicator value for a species within a
cluster is based on average abundance and on occurrence within clusters. Average abundance
refers to specificity, occurrence of species in different clusters is referred to as fidelity. For a
quantitative analysis it takes into account specificity and fidelity.
Phi Analysis for Fidelity (Tichy and Chytry 2006)- only works with occurrence data. Long
denominator (slide 24) they say helps this analysis work better when you have differences in
sample size among clusters, which is what we have here.
Fidelity analysis data sets (slide 25)- just used a couple data sets (the one where we distributed
ambiguous parents among the children, remove ambiguous parents ≥ Order, no rare taxa removed).
Did use species, genus, and looked at both quantitative and qualitative.
Global results were both for species and genus- quantitative and qualitative analyses did
show statistically significant association.
(Slide 27) Hydrologic Class A: These are the ones where the indicator values scores were
statistically significant under random permutation. Want you to notice the size of this. It’s nice to
look through here and see where we’ve got a species here, you find the same genus of that species
winds up being important in a lot of cases.
Hydro class B: Lots of candidates here.
Class C: Also a lot of candidates.
EFSAB Meeting Summary - 6/19/2012 Page 16 of 31

Class D: Only 1 taxa made that cut. There are definitely some problems with class D.
Class E was eliminated since it included only 1 site.
Class F: There were also a lot of candidates.
Class G: May also have some problems, relatively few candidates in there.
Fidelity Analysis Conclusions:
 We can extract taxa that are indicators of hydrologic clusters,
 the key question is not whether we can do that its whether the clusters are optimal as
indicators. If not, then fidelity analyses are not informative.
 Identifying optimal clusters is really important. In the fidelity analyses I did here, you can
identify the optimal clusters for individual taxa by looking at different cluster
methods…decide which is the best cluster for a particular species, then it gets interesting
because then you say, how many taxa show a good peak at x number of clusters, and start
getting the idea of how many clusters do you need to have to start pulling it apart. I
guarantee it will be tedious- 610 species in this data set. We can weed that down by getting
rid of rare taxa, then you have several hundred taxa…as you keep repeating same test over
and over your error rate goes up but there are ways of doing that.

Q: The tables near the end in the fidelity analysis, some of those taxa show up significant in
multiple classes, right? Or are they all unique to a class?
Response (R): They are all unique. It looks for the maximum value of the cluster.
Q: Is that for the genus level as well?
R: Yes, it is true for both genus and species, each one is a separate analysis but combined in one
table.
Q: Benthos- this is all sites included regardless of impairment?
R (RTI): The Data set included all benthos data that could be matched up with the 185 USGS sites.
A total of 106 NHD+ catchments (and associated Benthos data) were included in the analyses. .
There were no water quality of flow condition filters applied to this preliminary data set.
R (Cuffney): Then we may not be seeing correspondence because there are crappy sites in there.
C: Seems like presence/absence data it might be a better indicator.
R: Quantitative data you have probability that you will encounter the organism, and there’s also a
probably that there is a certain abundance level, and that joint probability will be a lot noisier. In
the work we’ve done for urban stuff, the qualitative stuff usually gives a clearer interpretation.
R(RTI): Of the 106 sites, approximately 60 of them meet criteria of the qualitative.
Q: from webinar: Would we gain any information (better resolution) by stratifying hydro classes
A-G by eco-region (mtn, piedmont, coast)? Example- Class B in mtns, Class B in piedmont, Class B at
coast?
R: Possibly, it’s a reasonable approach.
C: We had 185 gauged sites to develop EFS classes, at 106 we had a corresponding biological
monitoring site. That is the set Tom and RTI used. But if you apply the filters (any upstream flow
EFSAB Meeting Summary - 6/19/2012 Page 17 of 31

alteration need to be 2 times the drainage area, needs to be no lower biological classification than
good-fair/good-excellent) do you know how many sites would be left?
R: ~ 60 sites.
Jim: So 40 of those points that your preliminary and Tom’s analyses, some had either water quality
or water quantity alterations or both, which we are trying to remove as we go forward. The Class D,
the one with only 1 taxa associated, is the small flashy streams, Carolina Slate belt, has real mix of
some highly altered watersheds and some not. It could be Ellerbe creek for example.
C: Either most prone to impairment, or it would drive benthos to not be a good indicator of health.
R (Cuffney): To really evaluate either of these things, we need to work with sites with no
impairment. I have no confidence in this now. If there are really bad sites in here, that introduces a
lot of variability.
R (RTI): The idea for preliminary analysis was to use what data we had. Now that we have 578
sites that have gone through all the selection criteria…that would be the master data base, we can
do the same analysis on it and do a better analysis.
C: So good news is, it sounds like we have the techniques to do it, to remove the bad sites.
C: RTI can give you the data and you’ll have to go through all the computations again.
Q: This was fidelity of taxa to a-priori hydrologic classifications. You’re good at flipping it around
and looking at the community classes. Would they give us a different hydrologic break-down, to do
a cluster analysis of communities? If you looked at communities and see how they interpret the
hydrology?

R(Cuffney): yes, I’m a biologist and that’s how I’d go at it, someone who is a hydrologist would do it
the other way, the best would be to meet in middle, which we’re working at.

C: You can come up with a valid biological classification, but if you get a stream without any
biological data, how do you classify it without collecting more data?
C: Back to the canary in coal mine…if you could come up with some class for those streams…then
they could then be used as predictors of classification if you know the hydrology.
R: Once we have the 578 sites we can look at a number of alternative hydrologic classes and see
how those match up, can even do fidelity analysis, go through alternatives and see where you get
the max benefit, or indicator of Phi score in this case. Then look at it and say, this one works for this
set of species, etc., settle down to the classification that works best and subset of organisms that
works best.
Q: Regarding the WaterFALL model will this data be geospatially tagged? (facilitator didn’t hear
answer)
R(RTI): We will have a lot more data ready that can go through analysis (600 benthos site, 200 fish
sites, NHP sites).
C: We’re not just trying EFS class, but also McManamay class, to see if that had any greater
community association or species presence absence data.
Jen: if theEFS and McManamay classification systems are different, then it makes sense to test the
fidelities of biota to each set of stream classes. This repeated analysis is not be a problem since the
hydrology data will have already been generated for the sites.
Jim: I would not be surprised if there were similarity in classes.
EFSAB Meeting Summary - 6/19/2012 Page 18 of 31

Jen: EDF has already done analysis between EFS and McManamay systems to see if similar?
Sam: EFS and McManamay looked at each other’s systems and reported to us they were extremely
similar.

VII. Discussion of biofidelity analyses

Comment (C): It suggests to me - this is a very preliminary glimpse and it could really change when
you add a lot more data and filter out the streams that don't fit our criteria.
Response (R) (RTI): But, what if you added an overlay classification of mountain, Piedmont, and
coastal, or a different sorting category, like gradient or size. Usually there is a correlation between
location downstream and size of drainage area of a subwatershed. These different things may be
able to be used as overlays on top of the hydrologic classification. Plus, these things can be done
without site specific data.
R: I agree we have looked at such small samples that is hard to draw conclusions from them. I'm
optimistic that when we look at a lot more sites we may see some better correlations.
C: The working hypothesis for the RTI work is that we will not find a high degree of fidelity until we
subset our classes using some strategy that takes into account topoedaphic factors, such as the
shape of the land and the soils. So that remains as a future step, that is not yet funded.
C: It is important for us to remember that what we are trying to figure out is how to create a set of
classification for rivers and streams in NC based on hydrology. The issue is about altering
hydrology and determining how much alteration constitutes a breach of ecological integrity. We
need to come up with classes based on hydrology and then report how those classes effect
ecological integrity.
C: All the IFIM work and the fidelity work bears on that question. We are building up a pretty good
pile of ideas about what we should say about those classes in terms of how to define ecological
integrity or thresholds of ecological integrity or hydrologic regimes that insure ecologic integrity,
and we need to do that for every class. Currently we are still trying to figure out the classes and this
fidelity testing is to see if we have good classes and if not to see if we can come up with good
classes.
C: Based on our mandate, the classes need to be based on hydrology because that is what people are
altering.
C: Unfortunately these classes don't deal with alterations of water quality, such as a pollutant.
Question (Q): Is there some fidelity related to an ecoregion overlay of current EFS classes?
R: We've talked about various kinds of topoedaphic or other strategies to overlay classes to subset
them, and ecoregion definitions were one of them.
C: It seems like the fidelity we are searching for is actually going to be tied more closely to whatever
type of ecoregion we would overlay.
C: The hydrology alone is not trying to describe the distribution or the distinction of between
biology in different physiographic regions.
C: Mary Davis's work is about stream classification, not hydrologic or biologic classification.
EFSAB Meeting Summary - 6/19/2012 Page 19 of 31

C: RTI approach is unique by trying to test the biofidelity classes, trying to actually link the biology
to the stream classes.
R: All the other classifications systems out there are theoretical in nature, so testing this stuff puts
us in the forefront
C: This seems to be a philosophical discussion here. There are 2 hierarchical systems, one for
biology and one for hydrology, and at the top level we can say all streams have life and at bottom
we say only 1 stream has an individual of any particular type. We are trying to force-feed these two
hierarchies into one another. We are trying to get a specific species or guild to fit into a hydrology
hierarchy at a certain level. I think there is a balance of what biological hierarchical level
corresponds to a hydrological hierarchical level. This is a challenge.
C: In terms of terrestrial systems, we do this all the time and it is relatively easier because the
variables are easier to measure and we have better data about the species that occupy unique
climatic topoedaphic spots. We are struggling with how to do this with aquatic systems all across
the country. We will come up with something that is really good, but probably flawed and that will
get better over time.
Online comment: I think it is impossible to separate water quality conditions from ecological
integrity. Flow must play a role as well.
Q: At one point in the analysis do you start adding that level of detail in to determine if it makes a
difference ?
R: The process we developed is to test the fidelity at 600+ sites and assuming that we will not see a
great deal of revealed fidelity, subset the classes using multiple different strategies to see if one of
those results demonstrates higher fidelity.

Process flow chart:
Box 2 is the multi step process we are talking about, including all of the RTI work and Tom's work.
Box 3 and 4 begin the convergence of all the work
Box 5 is when we take all this and make decisions.

EFSAB Meeting Summary - 6/19/2012 Page 20 of 31

VIII. RTI internal project: Flow Alteration-Biological Response
Relationships to assist with the determination of ecological flows
Jennifer Phelan, Research Triangle Institute (RTI), provided an overview of their internal research
and development project. A summary follows: We thought our internal project would be of
interest to NC because we are developing flow alteration-biological response relationships in NC
using NC data. We don’t have results yet. We’ve spent time trying to fine-tune methods with input
from many of you. I’m going to try to keep this presentation as simple as possible.

Project Objectives:
To use a combined approach of:
 Unaltered/historic and current flows determined with WaterFALL model
 large, state-wide aquatic biota datasets
(monitoring programs)
to determine species-specific flow alteration – biological
response relationships that are useful to water managers
(some flow metrics are challenging for water resource
managers to use).

One of the outputs we’d like to produce is like this curve
(slide 3 and shown here), expressing species responses to
changes of flow. We are planning on doing a specific species
analyses and representing species response by changes in
abundance.

To explain our statistical approach, we wanted to use a space for time approach as opposed to a
time for time approach. You can go to one location and change the flow and see biological response.
In our case we’re trying to take a large set of data to try to represent those changes across the
landscape. (slide 4) Instead of evaluating changes in flow and species response in a single location,
we intend on estimating % change in flow at each location where monitoring data (and species
responses) are available and use a scatterplot of these points to produce the curves. There are
issues of other factors that influence responses, but we will try to address that as well.

Poff & Zimmerman did meta- analysis of data and studies concluded with this statement. I
highlighted words relevant to our analyses.

“Transferable quantitative relationships between flow alteration and ecological response” cannot
be easily developed, and “large databases, if analyzed with an eye toward degree of flow
alteration, carefully selected response metrics, stream typology, and multiple environmental
drivers, hold the potential to reveal important relationships” – Poff and Zimmerman (2010)

This type of approach has been attempted with variety of other studies (slide 6)
 Middle Potomac Environmental Flows (Mary Davis presented on this, a quantile
regression approach)
 Virginia DEQ Environmental Flows (in process)
 Carlisle et al. (2011) –a USGS effort across U.S. using NAQWA USGS gauges to show
biological responses to alterations. He showed with these large data sets that changes
in flow is a significant predictor.
EFSAB Meeting Summary - 6/19/2012 Page 21 of 31

Describing methods by focusing on each of the components of biological response/flow alteration
relationship curve:
1. What species will be used in the analyses to represent the biological response
2. Flow metrics that will be used to represent flow alteration
3. The statistical analyses that will be used to determine the significance and nature of the
relationship (ie. the response equation)
1. Methods: Species- Fish
 Top of the food chain integrators
 Relatively fast recovery time to climatic events that may cause local extirpation
 Valued by wide segments of society so may be easier to translate and communicate to
public

Datasets:
 NC DWQ Fish Community
 USGS National Water Quality Assessment Program (NAQWA)
Both datasets use comparable methods, comparable stream lengths, number of passes,
record species abundance and/or count values.

Species criteria:
 Look at individual species relationships AND
 Assign species to guilds indicative of flow conditions (4 different guilds identified by
Persinger in North Fork of Shenendoah River, VA(2010).
o Fast-generalist
o Riffle
o Pool-run
o Pool-cover
 Up to 20 species (to do a very good job on a fewer number of species rather than a so-so job
on larger number of species)
 Riffle guild- will apply our method of analysis to a single guild as a pilot Criteria used in
selecting species to include in the analyses:
o Well distributed (state or regionwide)
o Large number of records to make analysis easier and robust (> 50)
o Life history stages and requirements well known
o Single guild through-out life
This is a collaborative efforts- thanks to Chris Goudreau and Bryn Tracy. We’ve identified all the
species for NC and which guilds they’re associated with. Why a combined species/guild approach?
We want to identify canaries- those species most responsive to flow alterations-then by linking
them with guilds there might be an ability to extrapolate flow metrics to entire guild. There could
be value in linking individual species to guild. Also, DWR’s PHABSIM (habitat simulations) are built
upon guilds, we talked about possibility of assigning those different guilds to the 4 guilds identified
here to compare results of these methods in later stages of the block diagram discussed just before
lunch.
EFSAB Meeting Summary - 6/19/2012 Page 22 of 31

Although still in the planning phase, we may conduct an exploratory statistical analyses on all
species that fall under riffle guild and do show significant responses to flow alterations and try to
use that as a criteria to narrow down species for fuller suite of analyses.

2. Methods- Flow metrics
Criteria for flow metrics:
 Biologically relevant - can be related to biological response/supported by flow-biology
hypotheses
o Review of flow-biology hypotheses for fish:
o Increases and decreases in low, high and median/stable/base flows important to
fish
o Important events all months of the year
 Amenable to management - water managers can both understand and use them to
determine ecological flows
o Metrics represented the five components of flow, but removed 2
o Magnitude, timing, duration (I removed frequency and, rate of change since they
would be difficult for managers to use for water allocation/withdrawal policy)
o Also decided to focus on low flows since these are the challenges DWR is facing in
determining ecological flows
 Can be expressed as % change- some metrics are difficult to express as percentage changes
(e.g., flow fall and rise rate or time of occurrence)
 Can be effectively modeled within WaterFALL –some parameters are more challenging to
model (e.g., rise and fall rates or rate of change measures)

Flow metrics
 Reviewed TNC Indicators of Hydrologic Alteration (IHA) = 67 metrics, in light of criteria
 Additional metrics were added
Monthly Measures Annual Measures
Extreme low flow (10th percentile) 3-, 7-, 30-, and 90-day minimum
Low flow (25th percentile) 3-, 7-, 30-, and 90-day maximum
Median/Base flow (50th
percentile)
Extreme low flow (10th
percentile)
High flow (75th percentile) Extreme low flow count
Extreme low flow duration
(longest duration during year)

A final list of flow metrics were provided as a handout. In the table, bold= IHA metrics, the others
are added.
EFSAB Meeting Summary - 6/19/2012 Page 23 of 31

We combined both monthly measures, annual measures. For monthly measures, percentiles of
distributions for those months over a long period of record . For annual metrics, rolling averages of
3, 7,30, and 90 day minimums and maximums over the entire year. We’ll also account for extreme
low flow events. For low flows, together with low flow counts (number times below value), and of
the longest duration of low flow events during the year. Reason was for this came from meeting
with Jim, rather than an average, it was better to take the longest duration of the low flow because
these events are likely to have more impact on biology than the median or average extreme low
flows.

Flow metric calculations:
 WaterFALL hydrographs for unaltered/historic and current condition:
o Unaltered/Historic = Potential Natural Vegetation (PNV) or 1970s land-cover
o Current condition = 2006 NLCD + sources of instream flow alteration (dams,
withdrawals, discharges)
 30+ years of climate data (to reduce influences due to climatic variation and express %
change as from human alterations)
o Expressed as % change
 Focus analyses on reductions in flow (you can have positive changes in relationships, we
want to focus on the decreasing flows)

Q: Unaltered condition is related back to 1970s?
R: Either potential natural vegetation if there had not been land use changes, or 1970s data layer.
Reason is that in discussions with TNC and others, there is an argument to use 1970s for base layer.
C: I wouldn’t call that unaltered.
R: Yes, you’re right, the 1970’s layer is an earlier time (historic), while the other layer is unaltered.

Unaltered (historic) vs current layer expressed as % change.
-We’ll focus on the decreases in flows relationships.

3. Method: statistical analysis approach
The distinction between multivariate and univariate analysis (slide 22):
Multivariate = (equation shown on left of slide)multi-dimensional relationship often represented as an
equation. These X represent the influential factors (e.g., water quality, climate, habitat, those kind of
factors as predictors of biological response- arrows show those factors), and the flow metrics would be
one or several of these factors.
Univariate = (graph shown on right)relationship between single influence (e.g., flow metric) and
biological response
Benefits of using multiple predictors (multiple regression): gain a better understanding of the degree
to which species is influenced by altered flows versus other factors; and reduces the amount of error
(i.e., accounts for a larger proportion of the variance in the relationship).
We can tease out variation to say 10% is due to fragmentation, 20% is due to water quality, and it can
show what are the main drivers for certain species. It can help managers determine which species is
most sensitive to altered flows.
EFSAB Meeting Summary - 6/19/2012 Page 24 of 31

Negatives of multiple regression analyses- Application of relationship is restricted to locations where
have data for the other factors (or need to assume default or average values for factors); Complicates
the flow alteration – biological response relationship since you have to use defaults or average factors if
you don’t have the data, but allows us to have an understanding of other factors.

Co-variates including in the statistical analyses
Component Attribute/Dataset
Water Quality DO, pH, conductance (DWQ Habitat and USGS NAQWA
datasets)
303d listing (possibly)
Channel Morphology Sinuosity/Linear Length
Slope
Physiographic Region
Habitat NC Habitat Score (DWQ Habitat dataset)
Fragmentation- 2 ways (stream length between obstructions)
– same stream length and summed stream length of network
Climatic Events Preceding year total precipitation (to account for drought or
high water years)
Average of preceding 3 years of total precipitation
Date Sampling Month
Interspecific Interactions Influence of exotic and injurious introduced/ exotic fish
species

More on some of these components: Channel morphology- we think changes in flow will have
different effects based on morphology. Habitat- a variety of habitats exist in a stream length
(amount of cover, channelization, alterations). DWQ dataset has done good job of characterizing
these variables. We think fragmentation will be a significant factor- what habitat will species have
access to based on barriers to upstream, downstream movement (like dams). Fragmentation can
be done 2 ways- stream length between upstream and downstream barriers, for a single stream
length, versus a network analysis which takes into account all connected streams which are
bounded by barriers.

EFSAB Meeting Summary - 6/19/2012 Page 25 of 31

Univariate analysis- linear or quartile regressions (or comparable non-parametric analyses). We
will determine which approach based on distribution of data.
Benefit is only based on response to % change of flow metric, so application in water management
is more straight forward.
Negative is a potentially large amount of error which will reduce the strength of the relationships.

We hope the combination of analyses will provide:
o water managers with an understanding of the relative importance of flow alterations and
other influential variables for each species (some species are more influenced by flow
metrics than others).
o Flow-biology relationships with which to model/predict the impacts of flow alteration on
biological responses (will help to develop ecological flows).

The project is still in the planning phases. Any questions?

Q: Multivariate- you’ve shown multiple regressions, and simple regressions. Multivariate
regressions would have multiple x and multiple y.
R: It is a single y, with multiple variables.
C: The term multivariate analysis is incorrect. You’re conducting multiple simple regressions, both
are univariate procedures. I suggest clarifying that in because statisticians would notice.

Q: Is the parameter of 303 (d) list a presence-absence? Was it current list? Will you include the
parameter that caused it to be on the 303(d) list?
R: we’re still examining it , it will be categorical.
C: I suggest you abandon that because the 303(d) list combines policy decisions with water
quality decisions, doesn’t necessarily give you a ranking about water quality. You can use
some of the variables considered for the 303(d) list without the complexities of the policy decisions.
R: Good feedback. The problem would be the variables that are measured at water quality
monitoring stations, how far you can extrapolate those monitoring values?
C: That needs to be teased out, as an example, all streams are on 303(d) list because of the
mercury advisory, clearly you don’t want to include all those.

Q: Isn’t 303(d) those water bodies that are waiting for TMDLs?
C: No, it’s more complex. It is a list of impaired waters, but there are also policy decisions in that
not just those biologically challenged, but others get on the list too.

Online Q: Will you share guild species lists with group?
R: Good point. We think it will be useful for the committee so when its ready we’ll share it with
group.
C: Brynn and I (Chris Goudreau) are sharing first draft, Fritz Rhodes (EFSAB member representing
National Marine Fisheries) is an excellent fish taxonomist we can run it by him.
R: Great, more minds and more consensus reached it will be better.
(Fritz says he is glad to help)

Online Q: how were water quality variables chosen, were temp, DO, PH, conductance considered?
R: Temperature is one of the variables in dataset, and we did consider it. We could include it
in the analysis as exploratory, yes.

EFSAB Meeting Summary - 6/19/2012 Page 26 of 31

Q: Regarding the handout of flow metrics- for the list of annual metrics- 3, 7, 30, and 90 day
minimums are listed twice? Should that read “maximum”?
R: Yes that is a typo, my apologies.

IX. Small Group Discussion
The SAB broke into 4 groups for discussion. Planning team members composed 5 questions to help
move the discussion forward. The groups kept notes and reported out afterward. Since we only
had enough people for 4 groups, we set #3 aside and asked all the groups to discuss it if they had
time.
The questions were:
1. Flow metrics proposed for RTI IR&D – How do you think the proposed metrics could be
adopted to establish ecological flows and include in river basin models and planning? Are some
more suitable as triggers for red flags than others? Are there any not currently included in the
list outlined in the presentation that you think should be added?
2. Single species biological response proposed for RTI IR&D – What are your thoughts on the
proposed approach of evaluating individual species within a guild (probably the riffle guild
described by Persinger et al, 2010) to determine “canary” species for ecological flow
determinations? Will this identify the most flow sensitive organisms? Will this be explainable
and defensible for non-scientists?
3. Statistical analysis proposed for RTI IR&D – could both the univariate and multivariate analysis
results be used to determine ecological flows. If so, how?
4. Process diagram for stream classification through ecological flows – Do you agree with
proposed path for determining ecological flows, as outlined by the block diagram? Are there
any important steps missing?
5. Large piedmont rivers are not assessed by the state biological monitoring programs and the
coastal plain has a fewer number of monitoring stations. In addition, the link between flow and
habitat and species response to flow is less well understood for the coastal plain and possibly
less direct. Therefore, it may be difficult to determine ecological flows for these 2 stream types
using biological response or habitat alteration models. What do you propose would be an
appropriate pathway(s) to determine ecological flows for these rivers/region?

Report of the small group discussion follows.

1. Flow metrics proposed for RTI Internal Research & Development project. How do you
think the proposed metrics could be adopted to establish ecological flows and include in
river basin models and planning? Are some more suitable as triggers for red flags than
others? Are there any not currently included in the list outlined in the presentation that
you think should be added?
Small group discussion results:
Add the 7Q10, the 1-day low, and the 1-day high. These are metrics that are easily understandable
and widely used, allowing more people to understand and compare the work.
2. Single species biological response proposed for RTI IR&D – What are your thoughts on the
proposed approach of evaluating individual species within a guild (probably the riffle
guild described by Persinger et al, 2010) to determine “canary” species for ecological
EFSAB Meeting Summary - 6/19/2012 Page 27 of 31

flow determinations? Will this identify the most flow sensitive organisms? Will this be
explainable and defensible for non-scientists?
Small group discussion results:
 Why not use the count of all species within a guild rather than selecting a particular group
 Assign each species to one of four Persinger guilds.
 Determine count by total number of individuals for each species within each guild
 Allow richness (number of species) at each site, as well as abundance (count).
 We will have regional differences whether by guild or species.

Q: I’m concerned there will a guild or two missing from the analysis. Certainly mussel habitats are
being left out.
R: In this analysis RTI is currently focusing only on the fish, as a first step, but there will be other
studies, such at the TNC study which will be broader than just fish.
R: With respect to what guilds to include for fish, we compared the different guilds that are used by
Persinger and other researchers and some species don't fit into any of the four Persinger guilds.
This category is currently called other. it is a place holder for now. The TNC work may pick up
some of these.
We are comparing these species against the literature about species habitat in southern states.
3. Statistical analysis proposed for RTI IR&D – could both the univariate and multivariate
analysis results be used to determine ecological flows. If so, how?
From management perspective, the univariate approach lends itself to the model.
Multiple regression might be useful as foundation information for understanding what is going on
out there in addition to flow. What else helps explain the differences in species or guild count?
But it doesn't seem to plug into the final use of the model.
Also a propensity score analysis would allow the percentage contribution of each variable.

4. Process diagram for stream classification through ecological flows – Do you agree with
proposed path for determining ecological flows, as outlined by the block diagram? Are
there any important steps missing?
Small group discussion results:
 Incorporate feedback loops
 Add filters such as physiographic province/edaphic features into a feedback loop
 Define canaries as indicator species or drivers; will canaries be readily identifiable?
 Overall it is a good plan or process, and discussion throughout today has identified important
issues.
Online question: has DWQ clarified that tolerant species = canaries?
Response: As it relates to flow, no, this has not been determined. Current tolerant and intolerant
species lists were developed for water quality standards, not flow. (think, pollution tolerance.)

5. Large piedmont rivers are not assessed by the state biological monitoring programs and
the coastal plain has a fewer number of monitoring stations. In addition, the link
between flow and habitat and species response to flow is less well understood for the
EFSAB Meeting Summary - 6/19/2012 Page 28 of 31

coastal plain and possibly less direct. Therefore, it may be difficult to determine
ecological flows for these 2 stream types using biological response or habitat alteration
models. What do you propose would be an appropriate pathway(s) to determine
ecological flows for these rivers/region?

Small group discussion results:
We broke that into 2 topics.
1. How many streams don’t already have FERC licensing or something tied to them that already
contributes flow information.
2. Is there the same response relationship between small wadable streams and their biology and
the larger streams, because we don't have as much data. Is there another stream class that does
have enough data, that through this biological evaluation and looking at eco-regions, etc, maybe
one of these other stream classifications may apply to large Piedmont streams.
How many stream are we talking about here and will we find something comparable that we can
substitute.
C: Most Piedmont rivers have some kind of altered flow.
Coastal Plain:
We talked about the break between upper coastal plain and the lower area that has tidal influence,
where OASIS stops working. We may need to put a line on the map to define what will work
with our methodology and what will need a new method. Development is occurring in the
lower coastal plain and technology improves, they may be withdrawing brackish water.
We discussed what may need to be addressed. Is it just hydrograph, or do we include dissolved
oxygen and salinity.
Depending on what happens upstream, the tidal wedge line may creep upstream so that needs to be
considered.
Is WATERFALL a better use for the lower coastal plain? We are not sure.
There are also groundwater issues in the coastal plain that need to be incorporated.

X. Revisiting the DWR Trial Balloon Discussion

In April, NCDWR presented a “Trial Balloon”, which refers to a proposal to test reactions. The
entire proposal is detailed in the April Meeting summary. In brief summary, it was to discontinue
analyzing the flow scenarios of >120%, Annual and Monthly 7Q10, September Median in Spring and
Winter, and the Percentage of Mean Annual Flows of 10%, 20% and 30% in Spring and Winter. The
remainder flow scenarios were to be kept. Also, DWR suggested establishing criteria that err on the
side of slightly higher ecological flows so there would be lead time to meet increasing water
demands

Small groups met to discuss various habitat modeling scenarios, and the trial balloon proposal in
April (summarized in April Summary). Mary Lou Addor handed out copies of the summary, and
reviewed the results of these discussions. She explained a few minor items in the summary that
she would correct. Pages 9-10 include a summary of responses to the trial balloon. Question 3
EFSAB Meeting Summary - 6/19/2012 Page 29 of 31

(regarding the suggestion to err conservatively) was discussed for feedback, and did not need a
specific response from the group for moving forward. Questions 2 and 4 regarding the Trial
Balloon need to be revisited to ensure everyone’s interests are considered. These are:
Question 2 (in brief): We propose to focus on habitat results <80% of habitat available
under an unaltered flow regime, and not continue analyzing the >120% threshold
Question 4 (in brief): The 15 flow scenarios listed in the table should be used in analyses
moving forward (dropping the scenarios listed in the first paragraph above).

Mary Lou asked the group to take a few minutes to look at questions 2, 4, look at trial balloon, and
think if is there anything that needs to be added or shared with Jim and Fred as they move forward
with their analyses. Particularly if you were not present in April, this is an opportunity to share
with the group. John Crutchfield is not here, she’ll call him next week to get his feedback. We want
to make sure everybody is on same page moving forward.

Regarding Question 2: Are there any changes to this? It sounds like there wasn’t agreement for
various reasons.

C: I still feel strongly that >120 can be informative. If you have wide increases in certain types of
habitat it’s because other types of habitat are being reduced. I think there is info to be gained.

C: Some wanted to drop, some didn’t.
C: I think Jim picked up on point in winter when flows are high there is more reason to keep it,
because when reducing flows you are impacting in positive ways shallow guilds. In winter it may
come into play more so than other seasons.

C: When I looked at arranging raw data into different chart formats, it seemed clear that in the
>120 groups there was really significant increases in some types of habitat, which reflects decrease
in other types of habitat. If we didn’t continue to run the 120 scenario for all seasons, I think we’ll
lose info.

Facilitator’s note: There appears to be general support to keep the > 120% scenario in winter
analyses. The majority of EFSAB are okay with dropping the >120% analysis in the other seasons, with
a few members wanting to keep it in spring as well as winter, and with a strong preference from a
member to keep the >120% analysis in all seasons.

Question 4: The table below indicates which of the 15 flow scenarios analyzed so far, by season,
should continue to be evaluated. In April we heard: Group 1- keep all spring scenarios; Group 2-
keep all summer scenarios; Group 3- Agree with proposal for Fall scenarios (but analyze 7Q10
further into the analysis so people know it was analyzed); Group 4- agree with proposal for winter
scenarios. Seemed like there was more agreement here, regarding what scenarios to keep in based
on season. Are there any more comments?

Q: if you look on chart, if you keep 2 of the seasons but drop 2, it may be confusing to look at the
overall picture. Was there discussion about whether we should you keep it all for that reason?
Jim: I recall folks said they were agreeable not including 7Q10 in the charts but advising it would be
useful to run it to have in back pocket for final report because it is commonly used. I agree, it is not
hard to run it, we have a template, but not burden everyone by putting it in our graphs for our
meetings.

EFSAB Meeting Summary - 6/19/2012 Page 30 of 31

Q: Relative to the meeting discussion and today’s discussion about which efforts to add in, this is
regarding PHABSIM analyses. Today we talked about keeping in other metrics in the biological
analyses, would it be useful to have for comparison sake, the 7Q10 and low end (10% mean annual
flow)?

Jim: I think in terms of using flow metrics for RTI study, it picks up more of that -duration points
rather than % of average flow, it’s a frequency based number as opposed to % of an annual amount.
I think it may have more lending itself to a biological response curve- the duration or frequency
based approach. Don’t see a need to make them all line up. I agree with the 7Q10 suggestion.
Facilitator’s note: The EFSAB agreed to the table of proposed flow scenarios for DWR to analyze, with
the exception that 7Q10 should be analyzed as it’s a commonly used flow standard employed by other
states. The EFSAB does not need to see it included in the results presented to them.

XI. Upcoming Agenda Discussion
The remaining 2012 meeting dates and meeting locations are listed below and posted online at
www.ncwater.org/SAB.
August 28, 2012 - Stan Adams Training Center, Jordan Lake Educational State Forest
Sept 25, 2012 - Stan Adams Training Center, Jordan Lake Educational State Forest
October 23, 2012 - Stan Adams Training Center, Jordan Lake Educational State Forest
November 27, 2012 - Archdale Building

Over the next few meetings, the SAB will:
 Hear an update on the RTI biofidelity work,
 Have a presentation from The Nature Conservancy on their current 4 basin environmental
flow project
 Hear from Division of Water Resources on a plan for 2012-2013.
 Hear a presentation from Thomas Payne. DWR will request him to review EFSAB plan of
action and habitat modeling results so far. Thomas will be visiting Duke’s Nicolas School in
October.
Issues to be addressed:
 The use of Persinger’s 4 guilds and what species may be missing.
 How does using the Piedmont based guild set affect the EFSAB work in the mountains and
coast?
 What other guidelines for the coast are needed? What ecological concerns are present for
the EFSAB coastal work?

EFSAB Meeting Summary - 6/19/2012 Page 31 of 31

XII. Directions to August Meeting

On August April 28, 2012 we will meet at the:
Stanford M. Adams Training Facility at Jordan Lake Educational State Forest
2832 Big Woods Road, Chapel Hill, NC 27517
Map link: http://go.ncsu.edu/stanadams
From Rt 64 and Big Woods Road, it will be the first Forest Service sign on the right. Pass the office
building and continue on through the gate to the education center.