United States
Environmental Protection
Agency
Office of Research and
Development
Washington, DC
EPA/620/R-99/003
Julv 1999
&EPA
SHAT
Surface Waters
QUANTIFYING PHYSICAL HABITAT
IN WADEABLE STREAMS
Environmental Monitoring and
Assessment Program
-------
EPA/620/R-99/003
July 1999
QUANTIFYING PHYSICAL HABITAT
IN
WADEABLE STREAMS
by
Philip R. Kaufmann1, Paul Levine2, E. George Robison3,
Curt Seeliger2, and David V. Peck1
1 U.S. Environmental Protection Agency
Regional Ecology Branch
Western Ecology Division
National Health and Environmental Effects Research Laboratory
Corvallis, OR 97333
2OAO Corp.
c/o U.S. Environmental Protection Agency
200 SW 35th St.
Corvallis, OR. 97333
3Oregon Department of Forestry
2600 State St.
Salem, OR. 97310
ENVIRONMENTAL MONITORING AND ASSESSMENT PROGRAM
NATIONAL HEALTH AND ENVIRONMENTAL EFFECTS RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
RESEARCH TRIANGLE PARK, NC 27711
-------
NOTICE
This research described in this report has been funded wholly or in part by the U.S.
Environmental Protection Agency under Cooperative Agreements CR-818606 and CR-
824682 with Oregon State University, Contract 68-W5-0065 with OAO Corp., and Contract
68-C6-005 with Dynamac, Inc. This document has been prepared at the EPA National
Health and Environmental Effects Research Laboratory (Western Ecology Division,
Corvallis, Oregon).
This work is in support of the Environmental Monitoring and Assessment Program
(EMAP). It has been subjected to the Agency's peer and administrative review, and
approved for publication as an EPA document. Mention of trade names or commercial
products does not constitute endorsement or recommendation for use.
The correct citation for this document is:
Kaufmann, P.R, P. Levine, E.G. Robison, C. Seeliger, and D.V. Peck. 1999.
Quantifying Physical Habitat in Wadeable Streams. EPA/620/R-99/003. U.S.
Environmental Protection Agency, Washington, D.C.
-------
ABSTRACT
We describe concepts, rationale, and analytical procedures for characterizing
physical habitat in wadeable streams based on raw data generated from methods similar or
equal to those of Kaufmann and Robison (1998) that are used by the U.S. Environmental
Protection Agency (USEPA) in its Environmental Monitoring and Assessment Program
(EMAP). We provide guidance for calculating measures or indices of stream size and
gradient, sinuosity, substrate size and stability, habitat complexity and cover, woody debris
size and abundance, residual pool dimensions and frequency, riparian vegetation cover and
structure, anthropogenic disturbances, and channel-riparian interaction. The EMAP surveys
locate sample reaches using a randomized, systematic design (Stevens and Olsen 1999).
Within sample reaches, the EMAP field approach also employs a randomized, systematic
design to systematically locate and space habitat observations on stream reaches, each of
which have a length 40 times their lowflow wetted width. Two-person crews typically
complete EMAP habitat measurements in 1.5 to 3.5 hours of field time. While this time
commitment is greater than that required for more qualitative methods, these more
quantitative methods are more repeatable (more precise). For EMAP field crews in which
four people collect a variety of physical, chemical, and biological information, this level of
effort is about 25% to 33% of that spent on biological measures.
We evaluated sampling precision of field habitat survey methods employed by
EMAP in several hundred streams in Oregon and the Mid-Atlantic region, comparing
variance among streams ("signal") with variance between repeat stream visits
(measurement "noise"). Metrics with S/N <2.0 distort estimates of regional distributions
based on survey results, and severely limit analyses of associations by regression and
correlation; when metric S/N ratios are >10, these problems are relatively insignificant.
Quantitative channel morphology and riparian canopy densiometer measurements had
precise S/N ratios mostly between 6 and 20. Flow-sensitivity and ambiguity in features to
be measured limited precision of some physical measurements, but still resulted in metrics
of moderate to high precision (S/N 2-15). Semi-quantitative measurements (e.g. substrate
size) and visual presence-absence determinations (e.g., canopy presence) also had
moderate to high precision (S/N 2-16) The semi-quantitative metric group also included
several integrated metrics, such as mean substrate diameter, that were very precise (S/N
>20). Visual estimates of riparian canopy cover tended to have low to moderate precision
S/N <4.0, as did visual estimates of fish cover. Commonly used flow-sensitive measures
-------
(e.g. riffle/pool and width/depth ratios) and qualitative visual assessments (e.g., EPA's
Rapid Bioassessment Protocol habitat scores) tended to be imprecise (S/N <2). While
visual judgement methods are attractive because of their rapidity in the field and in data
reduction, their lack of precision limits their use in many applications. The final measure of
the utility of a habitat approach is whether it is useful for interpreting controls on biota or
impacts of human activity. We recommend that researchers examine a full suite of habitat
variables and consider the patterns of natural and anthropogenic controls and disturbances
in their region, the type of biota, and their own particular research objectives, in addition to
the precision of the habitat variables.
IV
-------
EXECUTIVE SUM MARY
We provide here a general description, rationale, and analytical guidance for
entering, verifying, summarizing, interpreting, and evaluating the precision of field data
generated by applying the physical habitat assessment methods similar or equal to those
described by Kaufmann and Robison (1994,1998). Those methods are currently used as
the standard method of stream habitat data collection by the U.S. Environmental Protection
Agency (USEPA) in its Environmental Monitoring and Assessment Program (EMAP). In the
broadest sense, stream habitat includes all the physical, chemical, and biological attributes
that influence or provide sustenance to organisms within the stream (Karr et al., 1986).
Physical habitat, for the purposes of EMAP field measurements, primarily concerns physical
elements, but includes some biological elements, such as aquatic macrophyte, riparian
vegetation, and large woody debris that are important in providing or controlling habitat
structure. Using the EMAP field physical habitat raw measurements as a starting point, we
provide guidance for calculating measures or indices of stream size and gradient, substrate
size and stability, habitat complexity and cover, riparian vegetation cover and structure,
anthropogenic disturbances, and channel-riparian interaction.
The EMAP surveys locate sample reaches using a randomized, systematic design
(Stevens and Olsen 1999) that results in a set of sample sites that is regionally
representative. Each sample reach is a length of stream channel 40 times as long as its
wetted channel width at the time of sampling. The EMAP field approach also employs a
randomized, systematic design to specify the location and spacing of habitat measurements
and observations within sample reaches (Kaufmann and Robison 1994,1998). This reach-
scale field sampling design at the makes calculating spatially representative stream reach
habitat characterizations straightforward.
This report will help researchers calculate and evaluate the utility of metrics that
summarize field data generated using EMAP methods and other similar quantitative habitat
survey methods. The derived reach-level metrics can then be used in analyses to interpret
regional patterns or temporal trends in habitat conditions, as well as associations with
biological or other data. The field-based physical habitat measurements from EMAP habitat
characterization are best used as a complement to other information (e.g., water chemistry,
temperature, and remote imagery of basin land use and land cover). The combined data
-------
analyses will more comprehensively describe additional habitat attributes and larger scales
of physical habitat or human disturbance than are evaluated by the field assessment alone.
The physical habitat data collected by the EMAP and Regional EMAP (REMAP) field
techniques described by Kaufmann and Robison (1994,1998) produce a large amount of
data that, for most uses, must be condensed to stream reach-level summaries that describe
particular aspects of physical habitat. In Appendix II, we describe a number of validation
and verification activities that are necessary before calculating numerical summaries,
metrics or descriptions of stream habitat from raw habitat survey data. These activities
include checking and reconciling data file structure, missing values, values out of range,
and illogical or unlikely combinations of variable values based on ecoregion, channel
morphology, or other internal relationships among variables. The stream reach metric
calculation approaches we recommend include simple statistical summaries, areal cover
estimates from areal cover class data, proximity-weighted disturbance indices, and
measures of woody debris abundance, residual pool dimensions and frequency, sinuosity,
and bed substrate stability. We describe procedures for calculating these reach-level
summary statistics, and have appended a CD (compact disk) containing documented SAS
(Statistical Applications Software) computer code to make these calculations.
Effective environmental policy decisions require stream habitat information that is
accurate, precise, and relevant. We evaluated sampling precision of field habitat survey
methods employed by the USEPA's EMAP in several hundred streams in Oregon and the
Mid-Atlantic region. We compared variance among streams ("signal") with variance
between repeat stream visits (measurement "noise"). Quantitative channel morphology and
riparian canopy densiometer measurements were precise (signal:noise (S/N) ratios mostly
6:1 to 20:1) when applied to features that are clearly defined and not excessively sensitive
to differences in flow stage. Flow-sensitivity (e.g. width, depth) and ambiguity in features to
be measured (e.g. incision height) limited precision, but still resulted in metrics generally
within the moderate to high precision range (S/N 2.0 to 15). Semi-quantitative
measurements (e.g. substrate size metrics) and presence-absence determinations (e.g.
visual estimates of canopy presence) also had moderate to high precision (S/N 2.0 to 16)
that was generally intermediate in precision between that of the two groups of quantitative
metrics, those that are flow-sensitive and those that are flow-independent. The semi-
quantitative metric group also included several integrated metrics, such as mean substrate
diameter, that were very precise (S/N >20). Visual estimates of riparian canopy cover
tended to have low to moderate precision S/N <4.0, as did visual estimates of the areal
cover of fish concealment features. Commonly used flow-sensitive measures (e.g.
riffle/pool and width/depth ratios) and qualitative visual assessments (e.g. EPA's Rapid
Bioassessment Protocol habitat scores) tended to be imprecise (S/N <2).
vi
-------
Based on our results, we make the following generalizations concerning the
precision of habitat measurement and assessment approaches:
Measurements are more precise than visual estimates, but
carefully-designed visual estimation procedures can be
nearly as precise as measurements. To enhance precision,
these visual observations are limited to measurable
characteristics (e.g. cover or presence), rather than
judgements of habitat quality, and they are made at multiple
locations within a reach.
Flow-sensitivity and complex definitions of habitat features
can degrade precision of quantitative measurements (e.g.
bankfull height and incision).
Flow-sensitivity and subjectivity in habitat-unit classifications
(e.g. %Pool) can seriously limit their usefulness in
contrasting stream habitat among streams or in tracking
changes in habitat through time.
The precision of multiple visual cover-class determinations
can be improved by re-interpreting this information as extent
of presence-absence of some defined feature (e.g. summed
vegetation cover in two layers reinterpreted as percent of
observations in which cover is >0% in both layers), but
perhaps at the expense of decreased sensitivity to stress.
The precision of separate metrics can be improved by
combining them into more integrated metrics, (e.g., the
precision of %Substrate <16mm diameter is more precise
than separate metrics of %Fine Gravel, %Sand, and %Fines;
the precision of %Pools+Glides is more precise than
%Pools), but perhaps at the expense of decreased sensitivity
to stress..
While visual judgement methods are attractive because of
their rapidity in the field and in data reduction, their lack of
precision limits their use in many applications.
VII
-------
At least 20 within-season pairs of repeat visits to 8 to 20 field
sites spread over several years are required for confident
assessment of within-season precision in physical habitat
metrics. These repeat samples are ideally drawn as a
random or stratified random sub-sample from a regional
probability sample of stream reaches.
Metrics with S/N < 2.0 distort estimates of regional
distributions based on survey results, and severely limit
analyses of associations by regression and correlation.
When metric S/N variance ratios are >10, field measurement
variance and short-term temporal fluctuations cause
relatively insignificant error and distortion in estimates of
regional population distribution functions and offer relatively
insignificant obstacles to analyses of association using
regression and correlation.
In EMAP field surveys, two people typically complete the specified channel, riparian,
and discharge measurements required in the quantitative approach within 1.5 to 3.5 hours
of field time. In addition to Physical Habitat data, a 4-person EMAP field crew collects
chemical water samples and data on fish assemblages, macroinvertebrates, periphyton,
and frequently benthic metabolism. The time commitment for collecting physical habitat
information is between 25% and 33% of the effort expended on biological measures, a
balance we feel is appropriate. While the time commitment for collecting data using the
EMAP habitat methods is greater than that required for more qualitative methods, the
greater expenditure results in more repeatable (more precise) characterizations and
assessments. The quantitative methods also provide greater flexibility in interpretation and
re-interpretation than do the qualitative habitat scoring approaches, because interpretations
of habitat quality are made during data analysis, rather than during field data collection.
The final measure of the utility of a habitat characterization approach is whether it
contains useful information for interpreting controls on the biota or impacts of human
activity. In regional surveys, or in temporal series, this measure of performance is
demonstrated through analysis of associations among variables. As with precision, this
aspect of habitat metric utility is also region-specific, and dependent on the type of
biological assemblage of interest and the type of human disturbances present. We offer
guidance to researchers in limiting the suite of habitat variables to be considered in
analysis, based on our own research and that of others who have used EMAP habitat data
viii
-------
in a variety of multivariate and other types of investigations associating habitat with fish,
macroinvertebrates, periphyton, benthic metabolism, and landscape disturbances. The
"short list" of 18 habitat variables includes representatives that we consider generally the
most important from each of the 7 aspects of habitat presented in the introduction to this
report. However, we recommend that researchers examine the full suite of variables
available and take into consideration the patterns of natural and anthropogenic controls and
disturbances in their region, the type of biota, and their own particular research objectives,
in addition to the precision of the habitat variables.
IX
-------
TABLE OF CONTENTS
Section Page
NOTICE ii
ABSTRACT iii
EXECUTIVE SUMMARY v
FIGURES xiii
TABLES xiv
ACKNOWLEDGMENTS xvi
ACRONYMS, ABBREVIATIONS, AND MEASUREMENT UNITS xvii
1 INTRODUCTION 1
1.1 PURPOSE OF THIS REPORT 1
1.2 PHYSICAL HABITAT COMPONENTS 1
1.3 SAMPLING CONSIDERATIONS 5
1.3.1 Sampling Season 5
1.3.2 Sample Reach Length 6
1.3.3 Intermittent and Ephemeral Streams 6
1.3.4 Habitat Characterization, Interpretation, and Replicability 6
2 SYNOPSIS OF EMAP PHYSICAL HABITAT FIELD METHODS 8
2.1 DESIGN AND RATIONALE 8
2.2 HABITAT DATA COLLECTION 11
2.2.1 Longitudinal Profile 11
2.2.2 Large Woody Debris Tally 12
2.2.3 Slope and Sinuosity Measurements 12
2.2.4 Substrate and Channel Dimension Cross-sections 14
2.3.5 Bank Morphology 16
2.3.6 Canopy Cover (Densiometer) 16
2.3.7 Riparian Vegetation Structure 16
2.3.8 Fish Cover, Algae, and Aquatic Macrophytes 17
2.3.9 Human Influence 19
-------
TABLE OF CONTENTS (Continued)
Section Page
3 CALCULATING REACH SUMMARY METRICS 20
3.1 INTRODUCTION 20
3.1.1 Conceptual Approaches 20
3.1.2 Data Files, Verification, and Metric Calculation Codes 20
3.2 METRIC CALCULATION PROCEDURES AND RATIONALE 27
3.2.1 Channel Morphology Statistical Summaries 27
3.2.2 Channel Cross-Section and Bank Morphology 27
3.2.3 Sinuosity 33
3.2.4 Slope 35
3.2.5 Residual Pool Analysis 35
3.2.6 Substrate Size and Composition 42
3.2.7 Bed Substrate Stability 43
3.2.8 Fish Cover 51
3.2.9 Large Woody Debris 52
3.2.10 Riparian Canopy Cover (Densiometer) 53
3.2.11 Riparian Vegetation Structure 54
3.2.12 Riparian Human Disturbances 55
3.2.13 Metric Variable Labeling and File Merging 56
4 PRECISION OF HABITAT CHARACTERIZATION 57
4.1 THEORETICAL CONSIDERATIONS 57
4.2 HABITAT METRIC PRECISION RESULTS 59
4.2.1 Channel Morphology and Habitat Classifications 60
4.2.2 Substrate 64
4.2.3 Fish Cover and Large Woody Debris 66
4.2.4 Riparian Vegetation 69
4.2.5 Riparian Human Activities and Disturbances 72
4.2.6 EPA's Rapid Bioassessment Protocol (RBP) Habitat Quality Scores ... 73
4.3 SUMMARY OF HABITAT METRIC PRECISION 77
4.4 IMPLICATIONS OF HABITAT MEASUREMENT PRECISION 82
4.4.1 Effects on Estimates of Regional Population Distributions 82
4.4.2 Effects on Associations Between Variables 84
4.5 GENERALIZATIONS AND RECOMMENDATIONS CONCERNING METRIC
PRECISION 86
4.6 OTHER CONSIDERATIONS IN SELECTING PHYSICAL HABITAT
METRICS 88
5 LITERATURE CITED 91
XI
-------
TABLE OF CONTENTS (Continued)
Section Page
APPENDIX A. COMPLETED EXAMPLES OF FIELD DATA FORMS FOR PHYSICAL
HABITAT CHARACTERIZATION A-1
APPENDIX B. DATA ENTRY, VERIFICATION, AND DATABASE STRUCTURE B-1
APPENDIX C. COMPUTER CODE FOR DATA VERIFICATION AND VALIDATION . . . C-1
APPENDIX D. COMPUTER CODE FOR PHYSICAL HABITAT METRIC
CALCULATION D-1
APPENDIX E. ILLUSTRATION OF RAW DATA AND COMPLETED METRIC
CALCULATIONS FOR SEVERAL STREAM REACHES E-1
XII
-------
FIGURES
Figure Page
Figure 1. Sample reach layout (plan view) 10
Figure 2. Substrate sampling cross-section 15
Figure 3. Boundaries for visual estimation of riparian vegetation, instream fish
cover, and human influences (plan view) 18
Figure 4. Residual pool profile. A) based on channel thalweg elevation data;
B) based on thalweg depth and reach slope data 37
Figure 5. Frequency distribution of signal to noise ratios for physical habitat
variables, grouped according to the types of habitat attributes assessed. . 79
Figure 6. Frequency distribution of signal to noise ratios for physical habitat
metrics, grouped according to the measurement approach 80
XIII
-------
TABLES
Table Page
TABLE 1. COMPONENTS OF EMAP-SURFACE WATERS PHYSICAL HABITAT
PROTOCOL 9
TABLE 2. HABITAT CLASSIFICATION AT CHANNEL UNIT SCALE 13
TABLE 3. PHYSICAL HABITAT MEASUREMENT DATA FILES AND THEIR
CONTENTS 22
TABLE 4. VARIABLES IN THE PHYSICAL HABITAT MEASUREMENT DATA
FILES 23
TABLE 5. COMPUTER PROGRAMS AND DATA FILES FOR CALCULATING
REACH-LEVEL PHYSICAL HABITAT METRICS 26
TABLE 6. PHYSICAL HABITAT METRIC VARIABLE NAMES AND LABELS 28
TABLE 7. PRECISION OF PHYSICAL HABITAT METRICS FOR QUANTITATIVE
STREAM CHANNEL MORPHOLOGY IN THE MID-ATLANTIC REGION
AND OREGON 61
TABLE 8. PRECISION OF PHYSICAL HABITAT METRICS FOR STREAM
CHANNEL HABITAT CLASSIFICATION IN THE MID-ATLANTIC
REGION AND OREGON 62
TABLE 9. COMPARISON OF PRECISION IN METRICS DESCRIBING STREAM
REACH POOL HABITAT IN SURVEYS OF THE MID-ATLANTIC
REGION AND OREGON 64
TABLE 10. PRECISION OF PHYSICAL HABITAT METRICS FOR STREAM
REACH SUBSTRATE IN THE MID-ATLANTIC REGION AND
OREGON 65
TABLE 11. PRECISION OF PHYSICAL HABITAT METRICS FOR INSTREAM FISH
COVER AND LARGE WOODY DEBRIS (WITHIN BANKFULL
CHANNEL) IN THE MID-ATLANTIC REGION AND OREGON 69
TABLE 12. PRECISION OF PHYSICAL HABITAT METRICS FOR CANOPY
DENSITY, COVER, AND PRESENCE IN MULTIPLE LAYERS OF
RIPARIAN VEGETATION ALONG STREAMS OF THE MID-ATLANTIC
REGION AND OREGON 70
TABLE 13. PRECISION OF PHYSICAL HABITAT METRICS FOR COVER AND
PRESENCE WITHIN SINGLE LAYERS OF RIPARIAN VEGETATION
IN STREAMS OF THE MID-ATLANTIC REGION AND OREGON 71
TABLE 14. COMPARISON OF PRECISION OF FOUR STREAMSIDE RIPARIAN
CANOPY COVER METRICS IN THE MID-ATLANTIC REGION AND
OREGON 73
TABLE 15. PRECISION OF PHYSICAL HABITAT METRICS FOR STREAMSIDE
RIPARIAN HUMAN ACTIVITIES AND DISTURBANCES IN THE MID-
ATLANTIC REGION AND OREGON 74
XIV
-------
TABLES (Continued)
Table Page
TABLE 16. PRECISION OF RAPID BIOASSESSMENT PROTOCOL (RBP)
HABITAT QUALITY METRICS (BARBOUR AND STRIBLING,
(1991) WITHIN SAMPLE SEASON IN THE MID-ATLANTIC
REGION AND OREGON 75
TABLE 17. CONTRASTING PRECISION OF SDDEPTH IN SEPARATE AND
COMBINED SURVEYS OF STREAMS IN THE MID-ATLANTIC REGION
AND THE WILLAMETTE BASIN IN OREGON 82
TABLE 18. THEORETICAL MAXIMUM OBSERVED CORRELATION COEFFICIENTS
(0 BETWEEN TWO METRICS OF VARYING PRECISION, AS
MEASURED BY S/N (CJ2st(yr)/a2rep) 86
TABLE 19. PHYSICAL HABITAT VARIABLES MOST FREQUENTLY USED 89
xv
-------
ACKNOWLEDGMENTS
We acknowledge the research leadership of Steve Paulsen, John Stoddard, and Phil
Larsen (U.S. EPA ORD, Corvallis, Oregon), without which this work would not have been
possible. We also thank the following reviewers for thoughtful insights, and helpful, critical,
and sometimes painful comments on earlier drafts: Dan Averill (Dynamac, Inc., Corvallis,
OR), Mike Barbour (TetraTech, Inc.), Steve Bauer (Pocket Water, Inc., Boise, ID), Ken
Bazata (Nebraska DEQ, Lincoln, NE), Bob Beschta (Oregon State University, Corvallis,
OR), John Buffington (U.S.Geological Survey, Boulder, CO), Lyle Cowles (EPA Region VII,
Kansas City, KS), Doug Drake (Oregon DEQ, Portland, OR), Jim Green (EPA Region III,
Wheeling, WV), Mike Griffith (Oak Ridge Institute for Science and Education, Oak Ridge,
TN), Tom Kincaid (Dynamac, Inc., Corvallis, OR), Phil Larsen, Harry Leland (U.S.
Geological Survey, Boulder, CO), Cindy Lin (EPA Region IX, San Francisco, CA), Craig
McFarlane (EPA ORD, Corvallis, OR), Glenn Merritt (Washington Dept. of Ecology,
Olympia, WA), Dave Montgomery (University of Washington, Seattle, WA), Mike Mulvey
(Oregon DEQ, Portland, OR), Joel Pederson (EPA Region IX, San Francisco, CA), LeRoy
Poff (Colorado State University, Fort Collins, CO), Geoff Poole (EPA Region X, Seattle,
WA), Charlie Rabini (U.S. Geological Survey, Columbia, MO), Steve Ralph (EPA Region X,
Seattle, WA), Sam Stribling (Tetra-Tech, Inc.), Billy Schweiger (EPA Region VII, Kansas
City, KS), Scott Urquhart (Oregon State University, Corvallis, OR), and Ian Waite (USGS-
BRD, Portland, OR). We also thank Marlys Cappaert (OAO Corp., Corvallis, OR)for
assistance in providing data and computer code on CD-ROM and the WED Web site.
Marge Hails-Avery (National Asian Pacific Center on Aging, Senior Environmental
Employment Program, Corvallis, OR) assisted with preparing the figures. Last, but certainly
not least, we thank the many fine field crew members from EPA, State, University, and
contractor organizations in the Pacific Coastal regions (EPA Regions IX and X), the Rocky
Mountains (EPA Region VIII), the central U.S. (EPA Region VII), and the Mid-Atlantic
Region (EPA Region III), who collected habitat data and have helped us to develop these
procedures for interpreting this data.
XVI
-------
ACRONYMS, ABBREVIATIONS, AND MEASUREMENT UNITS
Acronyms and Abbreviations
ANOVA Analysis of variance
CV Coefficient of variation
DBH Diameter at breast height
EMAP Environmental Monitoring and Assessment Program
EMAP-SW Environmental Monitoring and Assessment Program-Surface Waters
Resource Group
EPA (U.S.) Environmental Protection Agency
GPS Global positioning system
LWD Large woody debris
MAHA Mid-Atlantic Highlands Assessment
MAIA Mid-Atlantic Integrated Assessment
NAWQA National Water-Quality Assessment Program
NHEERL National Health and Environmental Effects Research Laboratory
ORD Office of Research and Development
OSU Oregon State University
PHab physical habitat
QA quality assurance
QC quality control
RBP (EPA) Rapid Bioassessment Protocol
REMAP Regional Environmental Monitoring and Assessment Program
RMSE Root mean square error
SD Standard deviation
USEPA U.S. Environmental Protection Agency
USGS United States Geological Survey
WED Western Ecology Division (Corvallis, OR)
XVII
-------
ACRONYMS, ABBREVIATIONS, AND MEASUREMENT UNITS
(CONTINUED)
Measurement Units
cm centimeter
ft feet
in inches
hr hours
kg kilogram
m meter
m2 square meters
m3 cubic meters
mm millimeter
N Newton
s second
XVIII
-------
1 INTRODUCTION
1.1 PURPOSE OF THIS REPORT
We provide here a general description, rationale, and analytical guidance for verifying
and summarizing field data generated by methods equal or similar to the quantitative and
semi-quantitative habitat characterizations described by Kaufmann and Robison (1994,
1998). We also present results concerning the precision of these habitat methods, based
upon field surveys in a variety of settings in the United States. The field approach
described by Kaufmann and Robison (1994, 1998) has been used as the standard method
of stream habitat data collection by the EPA's Environmental Monitoring and Assessment
Program (EMAP), by various Regional Environmental Monitoring and Assessment
Programs (REMAPs) in many states and EPA Regions, by several National Parks, and by
private industries. Thousands of streams have been sampled using EMAP methods
throughout the Mid-Atlantic region, the central U.S., Colorado, California, and the Pacific
Northwest between 1993 and 1998. This guide will help researchers to calculate and
evaluate the utility of metrics that summarize field data generated by these EMAP methods
and similar approaches. The reach metrics derived from these field data can then be
analyzed to interpret regional patterns or temporal trends in habitat conditions, and
associations with biological or other data. Field-based physical habitat measurements from
EMAP habitat characterizations are best used as a complement to other information (e.g.,
water chemistry, temperature, and remote imagery of basin land use and land cover). The
combined data analyses will more comprehensively describe additional habitat attributes
and larger scales of physical habitat or human disturbance than are evaluated by the field
assessment alone. This entire report, including appendices, may also be available in the
future on the EMAP website (http://www.epa.gov/emap).
1.2 PHYSICAL HABITAT COMPONENTS
In its broadest sense, habitat in streams includes all physical, chemical, and biological
attributes that influence or provide sustenance to organisms within the stream (Karr et al.,
1986). Physical habitat is an operational definition often used by stream ecologists to refer
to the structural attributes of habitat, and typically excludes water chemistry and physical
attributes such as water clarity, temperature, and light intensity. On the other hand, aquatic
1
-------
macrophyte, riparian vegetation, and large woody debris measurements are commonly
included in physical habitat assessments (e.g., Platts et al., 1983; Hankin and Reeves,
1988; Kaufmann and Robison, 1994, 1998; Overton et al., 1997; Stanfield et al., 1997)
because of their role in modifying habitat structure and light inputs, although they are
actually biological measures. Summarizing the habitat results of a workshop conducted by
EMAP on stream monitoring design, Kaufmann (1993) identified seven general physical
habitat attributes important in influencing stream ecology:
Stream Size -- Channel Dimensions
Channel Gradient
Channel Substrate Size and Type
Habitat Complexity and Cover
Riparian Vegetation Cover and Structure
Anthropogenic Alterations
Channel-Riparian Interaction
Like biological characteristics, all these habitat attributes vary naturally - including
the type and intensity of anthropogenic disturbance. Even in the absence of anthropogenic
disturbances, expected values of the other habitat attributes vary according to their
ecological setting. Within a given physiographic-climatic region, stream drainage area and
overall stream gradient are likely to be strong natural determinants of many aspects of
stream habitat, because of their influence on discharge, flood stage, and stream power (the
product of discharge times gradient). In addition, all these attributes may be directly or
indirectly altered by anthropogenic activities.
Stream Size is the primary determinant of the quantity of lotic habitat. The general
size class of a stream, based on its drainage area, stream order or annual runoff, is
relatively immutable. However, anthropogenic activities frequently alter channel
dimensions, floods, and low flow discharges, therefore altering the quantity and quality of
aquatic habitat. Kaufmann (1993) recommended that monitoring programs make field
measurements of thalweg depth, depth cross-sections, wetted and bankfull width, and
discharge as indicators of stream size. Field measurement of "baseflow" using a current
meter at one channel cross-section is typical of most habitat monitoring procedures, and
supplements an approximation of mean annual discharge calculated from watershed area
and generalized runoff data (e.g., runoff maps by Bishop et al., 1998).
Channel Gradient is a very important determinant of the potential energy in a
stream that can be converted into water velocity (Leopold et al., 1964). If discharge,
channel cross-section area, channel shape, and hydraulic roughness are all held constant,
-------
then the velocity of water in a stream is determined by the water surface gradient (Chow,
1959). The water surface gradient of a reach is essential for calculating stream power, QS,
and bed shear stress, pgRS, where Q=discharge, S=water surface slope, R=hydraulic
radius (channel cross-section area •*• wetted perimeter), p=mass density of water, and
g=gravitational acceleration (Dingman, 1984). These hydraulic characteristics can, in turn,
be used to estimate the bedload transport capacity and bed particle size that a stream can
move under various flow conditions. By comparing observed and mobile particle sizes
(Dingman, 1984), we can evaluate the stability of the stream bed and infer whether the
sediment supply to the stream may be augmented from enhanced upslope erosion resulting
from anthropogenic and other disturbances.
Channel Substrate: Bottom characteristics, including aquatic macrovegetation, are
often cited as major controls on the species composition of macroinvertebrate, periphyton,
and fish assemblages in streams (e.g., Hynes, 1972; Cummins, 1974; Platts et al., 1983).
Along with bedform (e.g., riffles and pools), substrate size influences the hydraulic
roughness and consequently the range of water velocities in a stream channel. It also
influences the size range of interstices that provide living space and cover for
macroinvertebrates, salamanders, sculpins, and darters. Substrate characteristics are often
sensitive indicators of the effects of human activities on streams (MacDonald et al., 1991).
Decreases in the mean substrate size and increases in the percentage of fine sediments,
for example, may destabilize channels and indicate changes in the rates of upland erosion
and sediment supply (Dietrich et al., 1989). Consequently, changes in substrate size
distributions are often indicative of catchment and streamside disturbances that alter
hillslope erosion or mobilize sediment. Accumulations of fine substrate particles also fill the
interstices of coarser bed materials, reducing habitat space and its availability for benthic
fish and macroinvertebrates (Platts et al., 1983; Hawkins et al., 1983; Rinne, 1988). In
addition, circulation of well-oxygenated water is impeded when fine particles embed coarser,
more permeable substrates. Most practitioners (e.g., Platts et al., 1983; Bauer and Burton,
1993) recommend a systematic "pebble count," as described by Wolman (1954), to quantify
the substrate size distribution, with visual assessments of substrate embeddedness as
described by Platts et al. (1983). The EPA stream monitoring design workshop (Kaufmann,
1993) recommended also including estimates of aquatic macrophyte and filamentous algal
cover because of their role as substrates and because their presence may be a useful
indication of water velocities and trophic status.
Habitat Complexity and Cover for Aquatic Fauna: When other needs are met,
complex habitat with abundant cover should generally support greater biodiversity than
simple habitats that lack cover (Gorman and Karr, 1978; Benson and Magnuson, 1992).
Habitat complexity is, however, difficult to quantify. The EPA's stream monitoring workshop
-------
participants agreed that the following components of complexity should be assessed
(Kaufmann, 1993):
Habitat Type and Distribution (e.g., Bisson et al., 1982; Frissell et al., 1986;
Hankin and Reeves ,1988; Hawkins et al., 1993; Montgomery and
Buffington, 1993, 1997, 1998).
Large Woody Debris count and size (e.g., Harmon et al., 1986; Robison and
Beschta, 1990).
In-Channel Cover: Percentage of areal cover of various types of features
that could provide fish concealment, including undercut banks, overhanging
vegetation, large woody debris, boulders (e.g., Hankin and Reeves, 1988;
Kaufmann and Whittier, 1997; Kaufmann et al., in review).
Residual pools, channel complexity, hydraulic roughness (e.g.,. Lisle 1982,
1987; Kaufmann, 1987a, 1987b; Robison and Kaufmann, 1994).
Width variance and bank sinuosity (Moore and Gregory, 1988).
Estimates of residual pool (Lisle 1982, 1987) frequency and size distribution, and
reach-scale indices of slackwater volume, channel morphometric complexity, and hydraulic
roughness can be quantitatively estimated from simple and rapid systematic profiles of
width and depth along stream reaches (O'Neill and Abrahams, 1984; Kaufmann, 1987a,
1987b; Robison and Beschta, 1990; Stack, 1989; Kaufmann and Robison, 1994, 1998;
Robison and Kaufmann, 1994, 1998). Indices of morphometric and hydraulic complexity
may be correlated with nutrient retentivity and may also be indicators of high flow velocity
cover in a stream reach (Kaufmann, 1987a, 1987b). Residual pool depths and volumes
also give an indication of habitat space during extremely low flows (Lisle, 1986, 1987).
Riparian Vegetation: The importance of riparian vegetation to channel structure,
cover, shading, nutrient inputs, large woody debris, wildlife corridors, and as a buffer
against anthropogenic perturbations is well recognized (Naiman et al., 1988; Gregory et al.,
1991). Riparian canopy cover over a stream is important not only for its role in moderating
stream temperatures through shading, but also as an indicator of conditions that control
bank stability and the potential for inputs of coarse and fine particulate organic material
(MacDonald et al., 1991). Organic inputs from riparian vegetation become food for stream
organisms and provide structure that creates and maintains complex channel habitat. The
EPA stream monitoring workshop participants recommended evaluating channel shading
(using canopy densiometer measurements) and riparian vegetation structure [by visual
estimates of the areal cover and type of vegetation in three layers (canopy, mid-layer, and
-------
ground cover)], distinguishing evergreen from deciduous vegetation, and woody trees and
shrubs from herbaceous vegetation.
Anthropogenic Alterations and Disturbances: Land use, buildings, and other
evidence of human activities in the stream channel and its riparian zone may, in
themselves, serve as habitat quality indicators; they may also serve as diagnostic indicators
of anthropogenic stress. The EPA's stream monitoring workshop recommended field
assessment of the frequency and extent of both in-channel and near-channel human
activities and disturbances. In-channel disturbances include channel revetment, pipes,
straightening, bridges, culverts, and trash (e.g., car bodies, grocery carts, pavement blocks,
etc.). Near-channel riparian disturbances include buildings, lawns, roads, pastures,
orchards, and row crops.
Channel/Riparian Interaction: Anthropogenic activities including grazing, farming,
flood control, channel revetment, and urbanization can result in the separation of streams
from their floodplains and riparian zones. The secondary effects on channel structure,
riparian vegetation, and ephemeral aquatic habitats can markedly affect biotic integrity of
stream ecosystems. Expectations for the potential magnitude and extent of interaction of
streams with the terrestrial environment differ for streams according to their channel type
and degree of valley constraint (Rosgen, 1985, 1994; Gregory et al., 1991; Stanford and
Ward, 1993). Possible metrics that might contribute to an index of channel/riparian
interaction include channel sinuosity, channel incision, and channel morphometric
complexity (based on the spatial pattern and variability in channel width and depth profile
data).
1.3 SAMPLING CONSIDERATIONS
1.3.1 Sampling Season
The EMAP stream indicator development workshop participants concluded that,
although physical habitat could be evaluated during any season, it would be most effective if
habitat evaluations were concurrent with biological sampling (Hughes, 1993). Generally the
most advantageous time for biological sampling in regional scale monitoring programs was
identified as a low flow season after leaf out and not closely following major flood events.
For most of the United States, this is the summer season, although some regional
differences are likely and should be examined. For example, late summer (August) might
be appropriate for snowmelt systems in the Rocky Mountains, while spring might be more
appropriate in parts of the arid southwest.
-------
1.3.2 Sample Reach Length
Local habitat shows repeating patterns of variation that are associated with riffle-pool
structure and meander bend morphology. If field sampling reaches are not long enough to
incorporate these patterns, the choice of the exact reach location will unduly influence
habitat characteristics observed in the field. For example, comparing a pool reach in a
disturbed basin with a riffle reach in a pristine basin will yield misleading results, no matter
how well the basins are matched in size, topography, climate, and lithology. Recognizing
the advantages of standardized reach lengths that are long enough to incorporate local
habitat-scale variation, large-scale monitoring and assessment programs in the U.S. and
Canada sample reach lengths that increase in proportion to stream size, typically measured
as multiples of wetted or bankfull width. Based on fish assemblage and habitat sampling
requirements, the U.S. EPA's EMAP program specifies sample reaches that are 40 times
their low flow wetted width (Klemm and Lazorchak, 1994; Lazorchak et al., 1998), the U.S.
Geological Survey's NAWQA program specifies reach lengths of 20 times wetted width
(Fitzpatrick et al., 1998), Simonson et al. (1994) specifies 30 to 35 times wetted width for
Upper Midwest streams, and Ontario Ministry of Environment specifies reaches 20 times
bankfull width (Stanfield et al., 1997).
1.3.3 Intermittent and Ephemeral Streams
The EMAP stream indicator development workshop participants felt that most
perennial stream habitat measures would be appropriate for intermittent and ephemeral
streams (Kaufmann, 1993). It is important to make such measurements on dry and near-
dry channels in order to characterize available aquatic habitat space and quantify changes
over time that might result from such influences as climate change and irrigation withdrawal.
Understandably, one would obtain values of zero for measures such as discharge, pool
depths, and wetted width in a dry stream.
1.3.4 Habitat Characterization, Interpretation, and Replicability
There are two conceptually different approaches to assessing habitat characteristics
or interpreting habitat quality. In one approach, exemplified by the EPA's Rapid
Bioassessment Protocols (Plafkin etal., 1989; Barbourand Stribling, 1991; Barbouret al.,
1997), habitat quality is interpreted directly in the field by biologists while sampling the
stream reach. This approach takes about 15 to 20 minutes of field time and quickly yields a
habitat quality assessment. However, the quality of that assessment depends upon the
knowledge and experience of the field observer to make proper and consistent
-------
interpretations of both the natural expectations (potentials) and the biological consequences
(quality) that can be attributed to the observed physical attributes.
The second conceptual approach confines observations to habitat characteristics
themselves (whether they are quantitative or qualitative). The regional patterns and trends
in these habitat characteristics themselves may well be of interest, as might be their
association with stream biota or watershed land use. Furthermore, a habitat quality index
may be derived by ascribing quality scoring to the habitat measurements as part of the data
analysis process. Typically, this second type of habitat assessment approach employs
more quantitative data collection, as exemplified by field methods for EMAP (Kaufmann and
Robison 1994, 1998), a habitat characterization framework proposed by Simonson et al.
(1994), the field methods of the U.S. Geological Survey's National Water Quality
Assessment (NAWQA) program (Meador et al., 1993; Fitzpatrick et al., 1998) and others
cited by Gurtz and Muir (1994). These field approaches typically define a reach length
proportional to stream width and employ transect measurements that are systematically
spaced (Kaufmann and Robison, 1994, 1998; Simonson et al., 1994; Fitzpatrick et al.,
1998), or spaced by judgement to be representative (Meador et al., 1993). They usually
include measurement of substrate, channel and bank dimensions, riparian canopy cover,
discharge, gradient, sinuosity, in-channel cover features, and counts of large woody debris
and riparian human disturbances. They may employ systematic visual estimates of
substrate embeddedness, fish cover features, habitat types, and riparian vegetation
structure. The field time requirement for these more quantitative habitat assessment
methods is usually 1.5 to 3.5 hours with a crew of two people. Because of the greater
amount of data collected, they also require more time for data summarization, analysis, and
interpretation. On the other hand, the more quantitative methods and less ambiguous field
measurements result in considerably greater precision when compared with qualitative
approaches.
-------
2 SYNOPSIS OF EMAP PHYSICAL HABITAT FIELD METHODS
2.1 DESIGN AND RATIONALE
This is not a field manual. We provide here only a synopsis of the EMAP field
methods for physical habitat in wadeable streams, which are described in a detailed field
training manual by Kaufmann and Robison (1994, 1998). Our purpose in the present
document is to describe the data collection methods generally so that readers can fully
understand the data reduction procedures and their rationale. Table 1 lists the components
of the EMAP field methods for physical habitat. These methods are most efficiently applied
during low flow conditions and after leafout of terrestrial vegetation, but may be applied
during other seasons and at higher flows, except as limited by considerations of safety. It is
designed for monitoring applications where robust, quantitative descriptions of reach-scale
habitat are needed, but time is limited.
The midpoint locations of sample reaches in the EMAP surveys are specified using a
randomized, systematic design (Stevens and Olsen, 1999). The EMAP field protocol
defines the length of each sampling reach proportional to wetted stream width at the time of
sampling, and then systematically places measurements to represent the entire reach
statistically. Field crews measure upstream and downstream distances of 20 times the
wetted channel width from the predetermined midpoints to center each 40 channel-width
field sampling reach (Lazorchak et al., 1998). A minimum reach length is set at 150 m.
Within each sample reach, the approach described by Kaufmann and Robison (1994, 1998)
employs a systematic sampling design to locate the actual habitat measurements and
observations. Once the downstream end of the reach is located, 11 transect positions are
set at 1/1 Oth of the sample reach length, or four times the mean wetted channel width apart.
Thalweg depth measurements are spaced at very tight intervals 1/100th the sample reach
length apart (1/150th in streams < 2.5 m wide); whereas channel wetted width,
cross-section profiles, substrate, bank characteristics and riparian vegetation structure are
measured at larger spacings (Figure 1). Woody debris is tallied along the full length of the
stream reach, and discharge is measured at one location (Table 1). A set of completed
EMAP field forms for habitat data collection is presented in Appendix A.
The randomized, systematic sampling design minimizes bias in the placement and
positioning of measurements in EMAP field sampling and facilitates the calculation of
8
-------
TABLE 1. COMPONENTS OF EMAP-SURFACE WATERS PHYSICAL HABITAT PROTOCOL
Adapted from Kaufmann and Robison (1994, 1998)
Component Description
1. Longitudinal Profile: Measure maximum (thalweg) depth, classify aquatic habitat, determine
presence of soft/small sediment at 10-15 equally spaced intervals
between each of 11 channel cross-sections (100-150 along entire
reach). Measure wetted width at 11 channel cross-sections and midway
between cross-sections (21 measurements).
2. Large Woody Debris: Between each of the channel cross sections, tally large woody debris
numbers within and above the bankfull channel according to size
classes.
3. Channel and Riparian Cross-Sections: at 11 cross-section stations placed at equal intervals
along reach length:
Measure: channel cross section dimensions, bank height, undercut, angle (with rod and
clinometer); gradient (clinometer), sinuosity (compass backsite), riparian canopy cover
(densiometer).
Visually Estimate3: substrate size class and embeddedness; areal cover class and type
(e.g., woody) of riparian vegetation in Canopy, Mid-Layer and Ground Cover; areal cover
class offish concealment features, aquatic macrophytes and filamentous algae.
Observe & Record3: human disturbances and their proximity to the channel.
4. Discharge: In medium and large streams (defined later) measure water depth and velocity (at
0.6 depth with electromagnetic or impeller-type flow meter) at 15 to 20 equally
spaced intervals across one carefully chosen channel cross-section. In very small
streams, measure discharge with a portable weir or time the filling of a bucket.
3 Substrate size class and embeddedness are estimated, and depth is measured for five particles taken at five equally-spaced
points (2 marginal, 3 mid-channel) on each cross-section. The cross-section is defined by laying the surveyor's rod or tape to
span the wetted channel. Woody debris is tallied over the distance between each cross-section and the next cross-section
upstream. Riparian vegetation and human disturbances are observed 5m upstream and 5m downstream from the cross
section station. They extend shoreward 10m from left and right banks. Fish cover types, aquatic macrophytes, and algae are
observed within channel 5m upstream and 5m downstream from the cross section stations. These boundaries for visual
observations are estimated by eye.
representative reach characteristics from raw data. Measures are taken over defined
channel areas, and these sampling areas or points are located systematically at spacings
that are proportional to low flow channel width (at the time of sampling). This systematic
sampling design scales the sampling reach length and resolution in proportion to stream
size. It also allows statistical and spatial series analyses of the data that are not possible
with other designs. The authors of the EMAP methods strove to make the approach
objective and repeatable by using easily learned, repeatable measures of physical habitat in
-------
Instream fish cover
Riparian Vegetation &
Human Disturbance
Substrate and Channel
Measurements
Upstream end of
sampling reach
Channel/Riparian
Cross section
Transect
Thalweg profile
intervals
Woody Debris Tally
(between transects)
Downstream end of
sampling reach \_/
Figure 1. Sample reach layout (plan view). From Kaufmann and Robison, 1998.
10
-------
place of estimation techniques wherever possible. Where estimation is employed, they
direct the sampling crew to estimate attributes that are otherwise measurable, rather than to
interpret the quality or importance of the attribute to biota or its importance as an indicator
of disturbance directly. However, they included traditional visual classification of channel
unit scale habitat types because they have been useful in past studies and enhance
comparability with other work.
The field time commitment to gain repeatability and precision is greater than that
required for more qualitative methods. In our field surveys, two people typically complete
the specified channel, riparian, and discharge measurements in about 3 hours of field time.
However, the time required can vary considerably with channel characteristics. On streams
up to about 4 m wide with sparse woody debris, measurements can be completed in less
than 2 hours. Crews may require up to 3.5 hours in large (> 10 m wide), complex streams
with abundant woody debris and deep water. Modifications from earlier EMAP methods
reduced the number of width measurements from 100 to 21 on sample reaches, reducing
the time required to < 3.5 hours even on most large, complex wadeable streams.
2.2 HABITAT DATA COLLECTION
2.2.1 Longitudinal Profile
Data from the longitudinal profile allows calculation of indices of stream size,
residual pool dimensions and spacing, pool sedimentation, channel complexity, and the
relative proportions of habitat types such as riffles and pools. The longitudinal profile is a
survey of thalweg (maximum) depth, wetted width, habitat class (sensu Bisson et al., 1982),
and presence of loose substrate with diameter < 16 mm. Measurements are collected at
100 or 150 equally-spaced points along the centerline of the channel between the two ends
of the stream reach. The "thalweg" is simply the deepest portion of the channel or cross-
section. The longitudinal profile of EMAP habitat protocol proceeds upstream in the middle
of the channel, rather than along the thalweg itself (though each thalweg depth
measurement is taken at the deepest point at each cross-section or incremental distance
upstream between cross-sections). The longitudinal profile measurements are spaced
evenly a distance of Vb to 1/4 the channel width's distance apart over the entire sample reach
length. This close spacing insures that they do not "miss" deep areas and habitat units that
are about as long as the channel is wide (Robison, 1998). A minimum sample reach length
of 150 m is set in the EMAP stream sampling methods in order to adequately sample fish;
the necessity for close depth sampling intervals results in a total of 150 thalweg sampling
increments in streams with wetted widths < 2.5 m. With the exception of backwater pools,
11
-------
channel unit scale habitat classifications pertain to the main channel. In addition to the
qualitative criteria listed in Table 2, these channel-unit scale habitat units should be at least
as long as the channel is wide. Wetted width is measured perpendicular to the mid-channel
line at 21 locations [at each of the 11 transects and midway between transects (Figure 1)].
2.2.2 Large Woody Debris Tally
The large woody debris (LWD) component of the EMAP Physical Habitat protocol
allows quantitative estimates of the number, size, total volume and distribution of wood
within the stream reach. The EMAP methods for LWD are a simplified adaptation of those
described by Robison and Beschta (1990). LWD is defined here as woody material with
small end diameter of at least 10 cm (4 inches), and length of at least 1.5 m (5 ft). For each
LWD piece, field surveyors first visually estimate length and both end diameters in order to
place it in one of twelve diameter and length categories. The diameter classes are 0.1 m to
< 0.3 m, 0.3 m to < 0.6 m, 0.6 m to < 0.8 m, and > 0.8 m, based on the large end diameter.
The length classes are 1.5m to < 5.0 m, 5 m to < 15 m, and > 15 m, based on the portion of
the LWD piece that is > 10 cm diameter. EMAP field crews separately tally all pieces of
LWD that are at least partially in the channel up to bankfull height, then those that span
above, but not into, the bankfull channel.
2.2.3 Slope and Sinuosity Measurements
The slope, or gradient, of the stream reach is useful in three different ways. First,
the overall stream gradient is one of the major stream classification variables, giving an
indication of potential water velocities and stream power, which are important controls on
aquatic habitat and sediment transport within the reach. Second, the spatial variability of
stream gradient is a measure of habitat complexity, as reflected in the diversity of water
velocities and sediment sizes within the stream reach. Lastly, using methods described by
Stack (1989) and Robison and Kaufmann (1994), the water surface slope will allow us to
compute residual pool depths and volumes from the multiple depth and width
measurements taken in the longitudinal profile. Compass bearings between cross section
stations, along with the distances between stations, allow us to estimate the sinuosity of the
channel (ratio of the length of the reach divided by the straight line distance between the
two reach ends). Slope and bearing are measured by "backsiting" with a clinometer and
compass downstream between cross-section transects B and A, C and B, D and C, etc., up
to the 11th cross section K (Figure 1).
12
-------
TABLE 2. HABITAT CLASSIFICATION AT CHANNEL UNIT SCALE3
Adapted from Kaufmann and Robison, (1998)
Class
Code Description
Pools: Still water, low velocity, smooth, glassy surface, usually deep compared to other parts of
the channel:
Plunge Pool
Trench Pool
Lateral Scour Pool
Backwater Pool
Dam Pool
Pool
Glide
Riffle
Rapid
Cascade
Falls
Dry Channel
PP - Pool at base of plunging cascade or falls.
PT - Pool like trench in stream center.
PL - Pool scoured along bank.
PB - Pool separated from main flow off side of channel.
PD - Pool formed by impoundment above dam or constriction.
P - Pool (unspecified type)
GL Water moving slowly, with smooth, unbroken surface - low
turbulence
Rl Water moving, with small ripples, waves and eddies - waves
not breaking, surface tension not broken, sound: "babbling",
"gurgling".
RA Water movement rapid and turbulent, surface with
intermittent Whitewater with breaking waves - sound:
Continuous rushing, but not as loud as cascade.
CA Water movement rapid and very turbulent over steep
channel bottom. Most of the water surface broken in short
irregular plunges, mostly Whitewater- sound: "Roaring."
FA Free falling water over vertical or near vertical drop into
plunge, water turbulent and white over high falls, sound:
from "splash" to "roar", depending upon discharge.
DR No water in channel
Code
Pool-Forming Element Category
N
W
R
B
F
WR, RW, RBW
OT
Not Applicable, Habitat Unit is not a pool
Large Woody Debris.
Rootwad
Boulder or Bedrock
Unknown cause (unseen fluvial processes)
Combinations
Other - note in comments
Note that in order for a channel habitat unit to be distinguished, it must be at least as wide or long as the channel is wide.
13
-------
2.2.4 Substrate and Channel Dimension Cross-sections
Measurements of substrate and channel dimensions at channel transects contribute
directly to assessments of habitat volume, channel and bed stability, and habitat quality for
benthic macroinvertebrates, periphyton, and fish. In the EMAP field methods (Kaufmann
and Robison, 1994, 1998), substrate size and embeddedness are evaluated at each of the
11 detailed cross-sections using a combination of procedures adapted from those described
by Wolman (1954), Bain et al. (1985), Platts et al. (1983), and Plafkin et al. (1989). The
basis of the procedure is a systematic selection of 5 substrate particles from each of the 11
channel cross sections (Figure 2). In the process of measuring sediment sizes at each
channel cross section, surveyors also measure the wetted width of the channel and the
water depths at each sediment sample point. If the wetted channel is split by a mid-channel
bar, the five substrate points are centered between the wetted width boundaries regardless
of the bar in between. Consequently, sediment particles selected in some cross-sections
may be "high and dry". For dry channels, field crews make cross-section measurements
across the unvegetated portion of the channel. The crews visually estimate the size of
particles according to the following classes:
RS Bedrock (Smooth).... > 4000 mm
RR Bedrock (Rough) > 4000 mm
HP Hardpan > 4000 mm
BL Boulders > 250 to 4000 mm
CB Cobbles > 64 to 250 mm
GC Gravel (Coarse) > 16 to 64 mm
GF Gravel (Fine) > 2 to 16 mm
SA Sand > 0.06 to 2 mm
FN Silt, clay, muck < 0.06 mm
WD Wood Regardless of Size
OT Other Regardless of Size
Field crews visually examine surface stains, markings, and algae to aid their
estimation of the average percentage embeddedness of particles that are larger than sand
within the 10-cm circle around the measuring rod. Embeddedness is the fraction of a
particle's surface that is surrounded by (embedded in) fine sediments on the stream bottom.
Sand and finer substrates are defined as 100% embedded.
14
-------
Right
Bank
Figure 2. Substrate sampling cross-section. From Kaufmann and Robison (1998).
15
-------
2.3.5 Bank Morphology
Bank morphology measurements contribute to assessments of channel stability
during flood flows, long-term channel down-cutting, and fish concealment features such as
undercut banks. Bank angle and undercut distances are measured on the left and right
banks at each of the 11 cross sections. To measure bank angle, field surveyors lay a meter
stick down against the bank with one end at the water's edge and measure the angle with a
clinometer (e.g., vertical bank = 90 degrees), an adaptation of procedures described by
Platts et al. (1983). If the bank is undercut, they measure the horizontal distance of the
undercutting. They also measure and record the wetted width of the channel, the width of
exposed mid-channel gravel or sand bars, the estimated incision height (depth, measured
from the water surface to the first valley terrace above bankfull height), and the estimated
height (depth) and width of the channel at bankfull stage.
2.3.6 Canopy Cover (Densiometer)
Canopy densiometer measurements are a relatively precise, objective means for
quantifying riparian vegetation cover, though they tell little about the type or structure of this
vegetation. Vegetative cover over the stream is measured at each of the 11 detailed cross
section stations using a Convex Spherical Densiometer, model B (Lemmon, 1957), modified
as described by Mulvey et al. (1992) to measure cover in 4 quadrats. For each of the 11
stations, densiometer measures are taken separately in four directions positioned at the
center of the stream. These 44 observations are used to estimate canopy cover over the
channel. Surveyors also measure canopy density facing the banks at the wetted channel
margins at both sides of each of the 11 cross-sections. These 22 bank densiometer
readings complement the visual estimates of vegetation structure and cover within the
riparian zone itself, and are particularly important in wide streams, where riparian canopy
may not be recorded when using the densiometer while positioned mid-stream.
2.3.7 Riparian Vegetation Structure
Visual estimation procedures are used to characterize the type and amount of
various types of riparian vegetation. This semi-quantitative assessment is used to evaluate
the condition and level of disturbance of the stream corridor. It also indicates the present
and future potential for various types of organic inputs and shading.
Observations to assess riparian vegetation apply to the riparian area within 5 m
upstream and downstream of each of the 11 cross-section stations (Figure 1). They include
the visible area from the stream back a distance of 10 m (30 ft) shoreward from both the left
16
-------
and right banks, creating a10mx I0m riparian plot on each side of the stream, centered
on the transect (Figure 3). The riparian plot dimensions are estimated, not measured. On
steeply sloping channel margins, the 10 m x 10 m plot boundaries are defined as if they
were projected down from an aerial view. If the wetted channel is split by a mid-channel
bar, the bank and riparian measurements are made on each side of the channel, not the
bar.
In the EMAP field methods, the riparian vegetation is conceptually divided into three
layers: a CANOPY LAYER (> 5 m high), an UNDERSTORY (0.5 to 5 m high), and a
GROUND COVER layer (< 0.5 m high). Large and small diameter trees are distinguished in
the canopy layer, as are herbaceous and woody vegetation in the understory. Note that
several vegetation types (e.g., grasses or woody shrubs) can potentially occur in more than
one layer. Similarly note that some attributes other than vegetation are possible entries for
the "Ground Cover" layer (e.g., barren ground). The type of vegetation (Deciduous,
Coniferous, broadleaf Evergreen, Mixed, or l\[one) in each of the two taller layers (Canopy
and Understory) are recorded. A layer is considered "Mixed" if more than 10% of its areal
coverage is made up of an alternate vegetation type. Areal cover is estimated separately in
each of the three vegetation layers. Areal cover can be thought of as the amount of shadow
cast by a particular layer alone when the sun is directly overhead. The maximum cover in
each layer is 100%, so the sum of the areal covers for the combined three layers could add
up to 300%. The four entry choices for areal cover within each of the three vegetation
layers are "0" (absent: zero cover), "1" (sparse: < 10%), "2" (moderate: 10 to 40%), "3"
(heavy: 40 to 75%), and "4" (very heavy: > 75%).
2.3.8 Fish Cover, Algae, and Aquatic Macrophytes
The EMAP habitat characterization includes semi-quantitative visual estimates of the
areal cover of a number of channel features that are important (alone or in combination with
other measures) for assessing habitat complexity, fish cover, macroinvertebrate habitat, and
channel disturbance. These include filamentous algae, aquatic macrophytes, woody debris
>0.3 m diameter, brush and small woody debris, overhanging vegetation < 1 m above the
water surface, undercut banks, boulders, and artificial structures. Filamentous algae are
long, streaming filaments of microscopic algal cells that often occur in slow moving, nutrient
rich waters with little riparian shading. Aquatic macrophytes are floating, submerged, or
emergent water loving plants, including mosses and wetland grasses that could provide
cover for fish or macroinvertebrates. Woody debris comprise the larger pieces of wood that
can influence cover and stream morphology. Brush/small woody debris pertains to the
smaller wood that primarily affects cover but not morphology. Overhanging vegetation
within one meter of the surface is the amount of brush, twigs, small debris, etc. that is not
17
-------
10m
Flow
10m
10m
RIPARIAN
PLOT
(Left Bank)
Cross-sectibn Transect
5 m
5 m
Instream Fish
Cover Plot
RIPARIAN
PLOT
(Right Bank)
10m
Figure 3. Boundaries for visual estimation of riparian vegetation, instream fish cover, and
human influences (plan view). From Kaufmann and Robison (1998).
18
-------
in the water but is close to the stream and provides cover. Boulders are typically basketball
to car sized particles. In-channel artificial structures may include those designed for fish
habitat enhancement and also those discarded (e.g., cars or tires) or purposefully placed for
diversion, impoundment, channel stabilization, or other purposes. Observations to assess
the areal cover of these in-channel features apply to the channel area within 5 m up- and
downstream from each of the 11 cross-section stations (Figure 3). The four entry choices
for areal cover of fish concealment and other features are "0" (absent: zero cover), "1"
(sparse: < 10%), "2" (moderate: 10 to 40%), "3" (heavy: 40 to 75%), and "4" (very heavy: >
75%), the same as those used for riparian vegetation cover.
2.3.9 Human Influence
The field evaluations of the presence and proximity of various important types of
human land use activities in the stream riparian area are used in combination with mapped
watershed land use information to assess the potential degree of disturbance of the sample
stream reaches.
At each of the 11 detailed Channel and Riparian Cross-Sections, crews evaluate the
presence/absence and the proximity of 11 categories of human influences:
Walls/ dikes/ revetments
Buildings
Pavement
Roads/railroads
Pipes
Landfills/trash
Parks/lawns
Row crops
Pasture/range/hay fields
Logging operations
Mining activities.
Observations are confined to the stream and riparian area within 5 m upstream and 5 m
downstream from each cross-section transect (Figure 3). The presence of each of these
human activities and their proximity to the stream channel are evaluated and recorded
separately for the left and right sides of the channel and banks. Proximity is distinguished
according to whether the activity is within the channel or its margin, within the 10 m x 10 m
riparian plot, or farther than 10m from the bank.
19
-------
3 CALCULATING REACH SUMMARY METRICS
3.1 INTRODUCTION
3.1.1 Conceptual Approaches
The physical habitat (PHab) measurements made using the EMAP and REMAP field
techniques described by Kaufmann and Robison (1994, 1998) produce a large amount of
data that, for most uses, must be condensed to stream reach summaries that describe
particular aspects of physical habitat. For example, the 55 separate observations of
substrate particle size class that are made at each stream reach are condensed into
statistics that summarize the substrate characteristics of each reach, not individual particles
or transects within a reach. For substrate, reach summaries include the geometric mean
substrate particle diameter and the percentages of the reach substrate composed of sand,
fines, or other size classes. EMAP's physical habitat sampling design within each reach is
systematic; this feature makes calculating representative summaries of the habitat
characteristics of stream reaches straightforward. These averages, percentiles and other
data summaries are correctly interpreted as spatially representative estimates of the habitat
characteristics measured. The metric calculation approaches we recommend include
simple statistical summaries, areal cover estimates from cover class data, proximity-
weighted estimates, and specific approaches for calculating woody debris abundance,
residual pool characteristics, sinuosity, and bed stability of stream reaches. In this section,
we discuss the conceptual basis and operational details for calculating summary metrics for
all of the types of EMAP habitat data; the SAS computer code files are included as
Appendix D on the compact disk. A set of raw data and an example calculation of all
metrics for several sample reaches is included as Appendix E on the compact disk. This
report and its appendices may be available in the future on the EMAP website
(http://www. epa.go v/emap).
3.1.2 Data Files, Verification, and Metric Calculation Codes
We use the following conventions to identify data files, variables, and computer code
throughout this document:
20
-------
Data file names are written in lowercase, bold italics, (e.g.,
sub_bank, thalweg).
PHab measurement and metric variable names, as well as
location and identification variable names and their
values are written in uppercase letters, (e.g., WT_WID,
TRANSDIR, STRMJD, PCT_POOL, ORC38, ORST97-
047).
When referring to a piece of Statistical Applications Software
(SAS) computer code we have included, we will specify
"SAS code", and name the code in UPPERCASE
ITALICS, or identify the code by description (e.g., ". . .
SAS code MH_RP applied to the sub_bank data file.").
Before calculating physical habitat metrics, it is necessary to insure that the raw
physical habitat data are accurate and that the raw data files match the intended structure
that accommodates metric-calculating algorithms like the SAS code we include in this
report. There are six general procedures for data verification and validation. They address
(1) data file structure, (2) missing values, (3) allowable ranges, (4) unusual values, (5)
plausible channel morphology, and (6) other evaluations of internal logic and consistency.
These are described in detail in Appendix B, with computer code to aid this process
(Appendix C on compact disk). After completing verification and validation of PHab data
sets, there will be a comment file plus six "final" data files containing verified, validated raw
physical habitat measurements (Table 3). The raw habitat variables contained in these files
are listed and defined in Table 4; their file structure is indicated by "location" variables that
include STRMJD, YEAR, VISIT_NO, TRANSECT, TRANSDIR, STA_NUM, and
INCREMNT. File structure and variable formats are illustrated in Appendix E. These final
data files comprise the source data for computing all reach-level metrics.
We have included computer code that calculates stream reach level summary
metrics as Appendix D, located on the compact disk. By "reach level", we mean that these
are single value summaries of numerous attributes of each stream reach. These metric
values are each calculated from numerous measurements or observations on each stream
reach that are contained in one of the six detailed data files. There may be several pieces
of code which calculate different metrics from the same data file, as summarized in Table 5.
This code has been designed to run on the final PHab data files without prior alterations.
Unless otherwise stated, missing and invalid values are excluded from these calculations.
In order to use the code we have included, first run the verification/validation programs,
starting with AARD\7X\RK.SX\Swhich, besides checking data structure and values, will
21
-------
TABLE 3. PHYSICAL HABITAT MEASUREMENT DATA FILES AND THEIR CONTENTS
Data File3
thalweg
sub_bank
fishcov
Igwoody
canpycov
riparian
phabcom
Contents
Thalweq Profile Data: Depth, Wetted-Width, Habitat Class, Pool-Forming
Code; Presence of Mid-Channel Bars, Side-Channels, and soft/small
sediment in thalweg.
Slope and Bearing - Primary and Supplemental Backsites: These data are
initially in the file slopebrg, then merged into thalweg during
verification/ validation process
Channel Cross-Section data: Wetted and Bankfull Widths, Mid-channel Bar
Width, Bankfull Depth, Incision Height, Bank Angle, and Undercut
Distance. Depth, Embeddedness, and Substrate Size Class at cross-
section verticals.
Areal cover class of Filamentous algae, Aquatic Macrophytes, Large Woody
Debris, Brush/small Woody Debris, Overhanging Vegetation, Undercut
Banks, Rock Ledge and Boulder cover, Artificial Structures.
Number of wood pieces in each of 12 diameter-size classes both for wood at
least partially within bankfull channel and wood above Bankfull
channel.
Canopy Density: Lemmon Model B Canopy Densiometer measurements.
Visual Riparian Estimates:
Canopy: Vegetation Type, Large-diameter Tree Cover, Small-diameter
Tree Cover.
Understory: Vegetation Type, Woody Cover, Non-woody Vegetation
Cover.
Ground Cover: Woody, Herbaceous, Barren or duff.
Human Influences - Presence and proximity of:
Walls/dikes/revetments
Buildings
Roads/railroads
Pavement
Influent or effluent pipes
Landfill or trash
Parks or lawns
Row crop agriculture
Pasture/range/hayfields
Logging operations
Mining activity
Narrative field and data entry comments
Data file names may vary somewhat from those we use in this report, but their basic content and structure will be as
presented, and their file names should be similar to those presented in this table.
22
-------
TABLE 4. VARIABLES IN THE PHYSICAL HABITAT MEASUREMENT DATA FILES
(These are prior to metric calculation)
Variable Name
Description
thalweg variables:
STRMJD
YEAR
VISIT_NO
DATE_COL
TEAMJD
TRANSECT
TRANSP
STA_NUM
INCREMNT
INDI
BARWID
BARYES
BEARING
CHANUNIT
DEPTH
POOLFORM
REACHLEN
SEDIMENT
SIDECHAN
SLOPE
BEAR2, BEARS
SLOPE2, SLOPES
PROPORTN, PROPORT2
PROPORT3,
WT_WID
COMMENT
COM FLDF
Individual site identification
Year of sampling
Visit number-within year
Date of site visit (MM/DD/YYYY)
Field sampling team identification
Sampling transect number or letter (A-K)
Transect spacing (m)
Thalweg depth sampling station number (0-9, or 0-14)
Reach length / number of stations per reach
Number of stations per transect
Mid-channel bar width (m)
Mid-channel bar presence (X = Yes)
Backsite bearing-transect to transect (degrees)
Habitat unit code
Thalweg water depth (cm)
Pool forming agent code
Length of sample reach (m)
Presence of fine sediments < 16mm (Y or N)
Presence of side channel (SC or Y for Yes)
Backsighted reach gradient (%)
Supplemental backsight bearing (0-359 degrees) - named
SUPBEAR in older data sets
Supplemental slope reading (%) -named SUPSLOPE in
older data sets
Proportion of intra-transect segment represented by
SLOPE, SLOPE2, SLOPES; and BEARING, BEAR2,
and BEARS, respectively
Wetted width (m)
Comments
Comment flag
Igwoody variables:
STRMJD
YEAR
VISIT_NO
DATE_COL
TEAMJD
TRANSECT
PIECES
PIECJDIA
PIECJ.EN
PI EC TYP
Individual site identification
Year of sampling
Visit number-within year
Date of site visit (MM/DD/YYYY)
Field sampling team identification
Sampling transect number or letter
Number of woody debris pieces tallied within diameter-length
class
Diameter Class (S, M, L, X) - S = 0.1to 0.3 m; M= 0.3 to 0.6
m; L = 06 to 0.8 m; X = > 0.8m
Length Class (S, M, L) - S=1.5-5 m; M=5-15 m; L= > 15 m
Woody debris WET/DRY at Bankfull flow
(continued)
23
-------
TABLE 4 (Continued)
Variable Name
Description
sub bank variables:
STRMJD
YEAR
VISIT_NO
DATE_COL
TEAMJD
TRANSDIR
TRANSECT
ANGLE
BANKHT
BANKWID
BARWID
DEPTH
DIST_LB
EMBED
INCISED
SIZE_CLS
UNDERCUT
WT WID
individual site identification
Year of sampling
Visit number-within year
Date of site visit (MM/DD/YYYY)
Field sampling team identification
Direction or position of measurement at transect
Sampling transect number or letter
Bank angle (degrees)
Bankfull height (m)
Bankfull width (m)
Mid-channel bar width (m)
Water depth (cm)
Distance from left bank at each TRANSDIR (m)
Substrate embeddedness (%)
Stream incision, height of first terrace above bankfull (m)
Substrate size class
Lateral distance of bank undercut (m)
Wetted width (m)
Comment flag pertaining to variable ******
fishcov variables:
STRMJD
YEAR
VISIT_NO
DATE_COL
TEAMJD
TRANSECT
ALGAE
BOULDR
BRUSH
MACPHY
OVRHNG
STRUCT
UNDCUT
WOODY
Individual site identification
Year of sampling
Visit number-within year
Date of site visit (MM/DD/YYYY)
Field sampling team identification
Sampling transect number or letter
Fish cover- areal plot coverage for filamentous algae
Fish cover- areal plot coverage for boulders or rock ledges
Fish cover - areal plot coverage for brush or small woody
debris
Fish cover- areal plot coverage for aquatic macrophytes
Fish cover- areal plot coverage for overhanging vegetation
Fish cover - areal plot coverage for artificial structures
Fish cover - areal plot coverage for undercut banks
Fish cover - areal plot coverage for large woody debris
Comment flag pertaining to variable ******
(continued)
24
-------
TABLE 4 (Continued)
Variable Name
Description
canpycov variables:
STRMJD
YEAR
VISIT_NO
DATE_COL
TEAMJD
TRANSECT
TRANSDIR
DENSIOM
Individual site identification
Year of sampling
Visit number-within year
Date of site visit (MM/DD/YYYY)
Field sampling team identification
Sampling transect number or letter
Direction and position of measurement at transect
Canopy cover densiometer reading (0-17)
riparian variables:
Vegetation:
STRMJD
YEAR
VISIT_NO
DATE_COL
TEAMJD
TRANSECT
TRANSDIR
BTRE
CANV
GCB
GCNW
GCW
NONW
STRE
UNDV
WOOD
Human Disturbance:
BLDG
CROP
LDFL
LOG
MINACT
PARK
PIPE
PSTR
PVMT
ROAD
WALL
******
Individual site identification
Year of sampling
Visit number-within year
Date of site visit (MM/DD/YYYY)
Field sampling team identification
Sampling transect number or letter
Direction of observation at transect
Riparian Canopy - large tree cover (> 0.3 m diameter)
Riparian Canopy - Dominant vegetation type
Riparian Ground cover - Barren or Duff
Riparian Ground cover - Non-woody vegetation
Riparian Ground cover - Woody vegetation
Riparian Understory - Non-woody vegetation cover
Riparian Canopy - Small tree cover (< 0.3 m diameter)
Riparian Understory - Dominant vegetation type
Riparian Understory - Woody vegetation cover
Human Influence-buildings
Human Influence-row crops
Human Influence-landfill/trash
Human Influence-logging operations
Human Influence-mining activity
Human Influence-park/lawn
Human lnfluence-pipes(inlet/outlet)
Human Influence-pasture/range/hay field
Human Influence-pavement
Human Influence-roads
Human Influence-wall/dike/riprap
Comment flag pertaining to variable ******
25
-------
TABLE 5. COMPUTER PROGRAMS AND DATA FILES FOR CALCULATING REACH-LEVEL
PHYSICAL HABITAT METRICS
Computer Code
MH_DWC1.sas
MH_HBCL.sas
MH_RP.sas
MH_SLP.sas
MH_SIN.sas
MH_SUBS.sas
MH_EMB.sas
MH_LMET.sas
(depends on previous
metric calculations, so
run last)
MH_BKF.sas
MH_ANGCUT.sas
MH_FCV.sas
MH_WOOD.sas
MH_DEN.sas
MH_VGPC.sas
MH_CMTYP.sas
MH_HUM.sas
MHLABELS.sas
PHABMET.sas
PHABBEST.sas
Data file
thalweg
thalweg
thalweg
thalweg
thalweg
sub_bank
sub_bank
mhsubs
mhjslp
mhwdcm
mtha
mhbkf
mhresp
sub_bank
sub_bank
fishcov
Igwoody
canpycov
riparian
riparian
riparian
phabmet or
final metric
data files
Final metric
data files
phabmet
Reach-Level Metrics Calculated
Mean and standard deviation (SD) of depth, wetted width, Width:Depth
ratio, and Width-Depth product.
% of reach in each habitat class.
Reach aggregate and individual residual pool metrics.
Mean and SD for reach slope.
Reach sinuosity from backsighted bearings.
Substrate size [% by class, mean and SD of size; median, lower and
upper quartiles (Q., Q3),and interquartile range of size class].
Log10(geometric mean diameter).
Reach mean and standard deviation of embeddedness (channel center
measurements only and channel center + stream margin
measurements)
Log10 of geometric mean substrate diameter in mm (LSUB_DMM), and
model estimate of Log10 of maximum diameter of mobile substrate in
mm (LTEST). "Relative bed stability" (LRBS_TST) = Log10(observed
mean substrate diameter/ mobile substrate diameter).
Mean and SD of bankfull width, bankfull height, and incision height
Mean, SD, Q., and Q3, and interquartile range of bank angle and
undercut distance
Areal cover and proportional presence offish concealment features
Counts and volumes of large woody debris (LWD) size classes
Mean and SD of canopy densiometer values (calculated separately for
mid-channel and bank measurements)
Riparian vegetation cover and presence metrics
Riparian vegetation type (proportion of reach with each type)
Proximity-weighted presence of human influences
Labels all final metric variables
Merges all final metric files to create phabmet
Extracts a reduced set of commonly-used variables from phabmet
26
-------
merge and assemble the appropriate sets of data for metric calculations. Edit the name of
the required final PHab data file in the code where requested (file renaming may be
necessary). The resulting data file should be saved and named appropriately for future
merging with other EMAP metric data (e.g., fish or macroinvertebrate assemblage data
files). The variable names and definitions of most of the reach physical habitat metrics
calculated by these computer codes are listed in Table 6.
3.2 METRIC CALCULATION PROCEDURES AND RATIONALE
3.2.1 Channel Morphology Statistical Summaries
Wetted channel dimension measurements (e.g. width, depth, bank angle) can be
directly reduced to whole-reach habitat characterizations by calculating their means and
standard deviations, or by calculating percentages of observations within stated bounds.
Because the data are systematically spaced, these averages and percentiles are estimates
of the spatial distributions of the habitat characteristics measured. For example, the mean
of the 100 thalweg depth measurements is an estimate of the mean thalweg depth in the
stream reach.
When code MH_HBCL is applied to data file thalweg, it uses frequencies of
observations in each habitat class to calculate reach level percentages of each class (e.g.,
PCT_PP, the percentage of the reach length classified as plunge pool habitat). In addition,
it combines classes into percentages of the following broader habitat classifications:
PCT_FAST = % (falls + cascades + rapids + riffles).
PCT_SLOW = % (all pool types + glides).
PCT_POOL = % all pool types (including impoundment, backwater, plunge,
lateral scour, and trench).
PCT_DRS = % (dry channel + subsurface flow).
When code MH_DW1C is applied to data file thalweg, it calculates simple reach
level means and standard deviations for wetted width and thalweg depth. It also calculates
mean width-depth products and width:depth ratios (Table 6), as well as the number of non-
missing values of width, depth, width x depth, and width/depth.
3.2.2 Channel Cross-Section and Bank Morphology
When code MH_BKF is applied to data file sub_bank, it calculates simple reach
level means and standard deviations for bankfull width and height, incision height, and also
27
-------
TABLE 6. PHYSICAL HABITAT METRIC VARIABLE NAMES AND LABELS
Variable Description
Identification Variables:
STRMJD
VISIT_NO
YEAR
Stream reach identification code
Site visit number
Year of site visit
Channel Morphology Metrics
REACHLEN
XDEPTH
SDDEPTH
XWIDTH
SDWIDTH
XWXD
SDWXD
XWD_RAT
SDWD_RAT
PCT_FA
PCT_CA
PCT_RA
PCT_RI
PCT_GL
PCT_PD
PCT_PP
PCT_PL
PCT_PT
PCT_PB
PCT_P
PCT_DR
PCT_SB
PCT_DRS
PCT_FAST
PCT_SLOW
PCT POOL
Length of sample reach (m)
Mean thalweg depth (cm)
Standard deviation of thalweg depth (cm)
Mean wetted width (m)
Standard deviation of wetted width (m)
Mean wetted width x depth (m2)
Standard deviation of wetted width x thalweg depth (m2)
Mean wetted width / depth (m/m)
Standard deviation of wetted width /thalweg depth (m/m)
Percent falls
Percent cascade
Percent rapids
Percent riffle
Percent glide
Percent impoundment pool
Percent plunge pool
Percent lateral scour pool
Percent trench pool
Percent backwater pool
Percent pool (unspecified type)
Percent dry channel
Percent subsurface flow
Percent dry or subsurface flow (PCT_DR + PCT_SB)
Percent falls + cascades + rapids + riffles
Percent glides + all pool types
Percent all pool types
Channel Cross-section and Bank Morphology Metrics:
XBKA
XUN
MEDBKUN
XBKF_W
XBKF_H
XING H
Mean bank angle (degrees)
Mean bank undercut distance (m)
Median bank undercut distance (m)
Mean bank full width (m)
Mean bank full height (m)
Mean incision height (m)
(continued)
28
-------
TABLE 6 (Continued)
Variable
Description
Channel Sinuosity and Slope Metrics:
SINU
XSLOPE
XBEARING
VSLOPE
Channel Sinuosity =Reach length / Straight line distance between reach ends)
Water surface gradient over reach (%)
Mean direction of reach flow (degrees)
Standard deviation of water surface gradient (%)
Residual Pool Metrics:
AREASUM
RP100
RPV100
TOTPVOL
RPGT50
RPGT75
RPGT100
RPXLEN
RPXAREA
RPMDEP
RPMLEN
RPMAREA
PCTRSED
PCTPSED
Residual pool total vertical profile area (m2/reach)
Mean residual depth [cm; equivalent to residual pool vertical profile area
(m2/100 m of reach)]
Residual volume per 100 m of reach (m3/100 m)
Residual volume for the entire reach (m3 / reach)
Residual pools with residual depth > 50 cm (number/reach)
Residual pools with residual depth > 75 cm (number/reach)
Residual pools with residual depth > 100 cm (number/reach)
Mean length of residual pools in reach (m/pool)
Mean residual pool area (mVpool)
Maximum residual depth of deepest residual pool in reach (cm)
Length of longest residual pool in reach (m)
Area (vertical profile) of largest residual pool in reach (rrWpoor)
Presence of thalweg small sediments (% of reach length)
Presence of thalweg small sediments (% of residual pool length)
Substrate Size and Composition Metrics:
SUB_X
SUB_V
SUB_Q3
SUB_MED
SUB_Q1
SUBJQR
LSUB_DMM
XEMBED
VEMBED
XCEMBED
VCEMBED
PCT_RS
PCT_RR
PCT_BDRK
PCT_BL
PCT_CB
PCT_GC
PCT BIGR
Substrate mean size class (see text)
Standard deviation of substrate size class (see text)
75th percentile of substrate size class
Substrate median size class
25th percentile of substrate size class
Interquartile range of substrate size class
Log10[estimated geometric mean substrate diameter (mm)]
Substrate mean embeddedness - channel + margin (%)
Standard deviation of embeddedness - channel + margin <°'
Substrate mean embeddedness - mid channel (%)
Standard deviation of embeddedness - mid-channel (%)
Substrate % smooth bedrock (>4000mm)
Substrate % rough bedrock (>4000mm)
Substrate % bedrock
Substrate % boulder(250-4000mm)
Substrate % cobble (64-250mm)
Substrate % coarse gravel (16-64mm)
Substrate % coarse gravel and larger (>16mm)
(continued)
29
-------
TABLE 6 (Continued)
Variable
Description
Substrate Size and Composition Metrics:
PCT_GF
PCT_SFGF
PCT_SA
PCT_FN
PCT_SAFN
PCTJDM
PCT_WD
PCT_ORG
PCT_RC
PCT_HP
PCT OT
Substrate % fine gravel (2-16mm)
Substrate % fine gravel and smaller (<=16mm)
Substrate % sand (0.6-2mm)
Substrate % fine (silt/clay; < 0.6 mm)
Substrate % sand + fines (<2mm)
Substrate % organic detritus
Substrate % wood
Substrate % wood or detritus
Substrate % concrete
Substrate % hard pan
Substrate % miscellaneous other types
Bed Substrate Stability Metrics:
LTEST
LDMB BW4
LRBS 1ST
LRBS BW4
Log10 [Erodible substrate diam (mm) ] - Quick Estimate:
LTEST=log10[13.7x(0.5xXDEPTHx10)(XSLOPE/100)]
Log10 [Erodible substrate diam (mm) ] - Estimate 2:
LDMB_BW4=Log10[13.7(Rbf- Rw- Rp)xS)
where:
Rbf = 0.5[(XDEPTH x 10)+(XBKF_H x 1000)],
Rw =(V1W_MSQx 1000),
Rp = (0.5xRP100 x 10),
S = (XSLOPE-MOO),
and if Rw > (Rbf - Rp) then (Rbf - Rw - Rp) = 0.1 (Rbf - Rp)
Log10 [Relative Bed Stability] = (observed mean substrate diameter)
(erodible substrate diameter) - Quick Estimate:
LRBS_TST=LSUB_DMM - LTEST
Log10 [Relative Bed Stability] = (observed mean substrate diameter)
(erodible substrate diameter) - Estimate 2:
LRBS BW4 = LSUB DMM - LDMB BW4
Fish Cover Metrics:
XFC_ALG
XFC_AQM
XFC_LWD
XFC_BRS
XFCJDHV
XFC_RCK
XFC UCB
Filamentous algae areal cover
Aquatic macrophyte areal cover
Large woody debris areal cover
Brush and small woody debris areal cover
Overhanging vegetation areal cover
Boulder and rock ledge areal cover
Undercut bank areal cover
(continued)
30
-------
TABLE 6 (Continued)
Variable Description
Fish Cover Metrics (continued):
XFCJHUM Artificial structure areal cover
XFC_ALL Sum of areal cover from all fish concealment types except algae and aquatic
macrophytes.
XFC_BIG Sum of cover from large wood, boulders, over-hanging banks and human
structures
XFC_NAT Sum of cover from large wood, brush, overhanging vegetation, boulders and
undercut banks
PFC_xxx: Proportion of reach with named cover types present, regardless of amount of
cover, (xxx represent the last 3 letters of the cover type variables
above)
Large Woody Debris (LWD) Metrics:
LWD Size Definitions:
LWD is tallied in a matrix with three length and four large end diameter classes:
Length
Diameter S (1.5 m -5 m) M (> 5 m -15 m) L (> 15 m)
S(0.1m-0.3m) T S M
M (> 0.3m-0.6m) S M L
L(>0.6m-0.8m) S L L
X(>0.8m) ML X
The codes T, S, M, L, and X in this table are progressively larger piece sizes from very small to very
large. A nominal mean volume is calculated for each piece of LWD according to its diameter-length
class membership as described by Robison (1998):
Ti[1.33(Class minimum Diameter-^2)2]x[l.33(Class minimum Length)]
Total numbers and volumes of LWD in each diameter-length class are regrouped and assigned to
one of five cumulative wood size classes:
Class 1 - T, S, M, L, X Very small to Very Large
Class 2 - S, M, L, X Small to Very Large
Class 3 - M, L, X Medium to Very Large
Class 4 - L, X Large to Very Large
Class 5 - X Very Large
(continued)
31
-------
TABLE 6 (Continued)
Variable
Description
Large Woody Debris (LWD) Metrics:
LWD Metric Variables:
C1W...C5W
V1W... V5W
C1WM100 ...C5WM100
V1WM100... V5WM100
C1W_MSQ...C5W_MSQ
V1W_MSQ ... V5W_MSQ
C1T...C5T
V1T... VST
C1TM100 ...C5TM100
V1TM100 ... V5TM100
LWD
LWD
LWD
LWD
LWD
LWD
LWD
LWD
1...5
LWD
LWD
1...5
in active channel (pieces/reach) - size classes 1...5
volume in active channel (mVreach) - size classes 1...5
in active channel (pieces/100m) - size classes 1 ...5
volume in active channel (m3/100m) size classes 1 ...5
in active channel (pieces/m2) - size classes 1...5
volume in active channel (mVm2) - size classes 1...5
in and above active channel (pieces/reach) - size classes 1...5
volume in and above active channel (mVreach) - size classes
in and above active channel (pieces/100m) - size classes 1 ...5
volume in and above active channel (m3/100m) - size classes
Riparian Cover (Densiometer) Metrics:
XCDENBK
XCDENMID
Mean % canopy density at bank
Mean % canopy density mid-stream
Riparian Vegetation Cover and Structure Metrics:
XCL Riparian canopy (> 5 m high) cover - trees > 0.3m DBH
XCS Riparian canopy (> 5 m high) cover- trees < 0.3m DBH
XMW Riparian mid-layer (0.5 to 5 m high) woody cover
XMH Riparian mid-layer (0.5 to 5 m high) herbaceous cover
XGW Riparian ground-layer (< 0.5 m high) woody cover
XGH Riparian ground-layer (< 0.5 m high) herbaceous cover
XGB Riparian ground-layer (< 0.5 m high) bare ground cover
XC Riparian canopy cover (XCL+XCS)
XM Riparian mid-layer cover (XMW + XMH)
XG Riparian ground-layer vegetation cover (XGW + XGH)
XCM Riparian canopy + mid-layer cover (XC + XM)
XCMW Riparian canopy + mid-layer woody cover (XC + XMW)
XCMG Riparian cover, sum of 3 layers (XC + XM + XG)
XCMGW Riparian woody cover, sum of 3 layers (XC + XMW + XGW)
XPCAN Riparian canopy presence (proportion of reach)
XPMID Riparian mid-layer presence (proportion of reach)
XPGVEG Riparian ground cover presence (proportion of reach)
XPCM Riparian canopy and mid-layer presence (proportion of reach)
XPCMG 3-layer riparian vegetation presence (proportion of reach)
(continued)
32
-------
TABLE 6 (Continued)
Variable
Description
Riparian Vegetation Cover and Structure Metrics (continued):
PCAN_C Coniferous riparian canopy presence (proportion of reach)
PCAN_D Deciduous riparian canopy presence (proportion of reach)
PCAN_E Broadleaf evergreen riparian canopy presence (proportion of reach)
PCAN_M Mixed riparian canopy type presence (proportion of reach)
PMID_C Coniferous riparian mid-layer presence (proportion of reach)
PMID_D Deciduous riparian mid-layer presence (proportion of reach)
PMID_E Broadleaf evergreen riparian mid-layer presence (proportion of reach)
PMID_M Mixed riparian mid-layer type presence (proportion of reach)
Human Disturbance Metric Variables:
W1H BLDG
W1H WALL
W1H PVMT
W1H ROAD
W1H_PIPE
W1H LDFL
W1H PARK
W1H CROP
W1H PSTR
W1H LOG
W1H MINE
W1 HALL
W1JHNOAG
W1_HAG
Riparian
Riparian
Riparian
Riparian
Riparian
index)
Riparian
Riparian
Riparian
Riparian
Riparian
Riparian
Riparian
Riparian
sum)
Riparian
human
human
human
human
human
human
human
human
human
human
human
human
human
human
disturbance
disturbance
disturbance
disturbance
disturbance
disturbance
disturbance
disturbance
disturbance
disturbance
disturbance
disturbance
disturbance
disturbance
- Buildings (proximity-weighted index)
- Channel revetment (proximity-weighted index)
- Pavement (proximity-weighted index)
- Roads (proximity-weighted index)
- Pipes, influent and effluent (proximity-weighted
- Trash and Landfill (proximity-weighted index)
- Parks and Lawns (proximity-weighted index)
- Row Crop Agriculture (proximity-weighted index)
~ Pasture and Grass fields (proximity-weighted index)
- Logging (proximity-weighted index)
- Mining (proximity-weighted index)
index - All types (proximity-weighted sum)
index- Non-agricultural types (proximity-weighted
index - Agricultural types (proximity-weighted sum)
lists the number of non-missing values used in these calculations as a quality assurance
check (Table 6).
When code MH_ANGCUT is applied to data file sub_bank, it calculates simple
reach level means and standard deviations for bank angle and undercut distance. In
addition, it calculates the median, upper and lower quartiles, and the inter-quartile range of
these variables within each reach (Table 6). It also lists the number of non-missing values
used in these calculations.
3.2.3 Sinuosity
Sinuosity (Schumm, 1963), or thalweg sinuosity (Leopold and Wolman 1957) is a
mathematical expression of the degree of tortuosity or twisting of a stream channel as
33
-------
observed from above. Sinuosity is measured as the distance along the channel "as the fish
swims" divided by the direct line-of-site distance "as the crow flies" up the valley between
the two ends of the reach. Kaufmann and Robison (1998) prescribe compass bearing
backsites between each of the 11 channel cross-section transects in the EMAP field
methods, yielding 10 bearings and distances from which one can crudely map the course
of the channel within the limits of the sample reach. It is extremely important to define the
resolution and scale over which the measurements for sinuosity are made. For consistency
in EMAP, we define the scale as a reach 40 times as long as its wetted width at low flow
and the resolution of measurements as one measurement of length and bearing per 4
channel-widths' distance. To calculate sinuosity, the following calculations are made:
Sinuosity = (Reach length) •*• ("Crows" distance)
= [Z(DT)] * [(Z"Northing")2 + (I"Easting")2f2;
(1)
where:
"Northing" and "Easting" are, respectively, the northern and
eastern vector components of the distance from the
downstream starting point.
DT = Distance along channel between transects,
I = summation over transects,
0 = compass bearing in radians= 2n-(degrees of bearing/360 degrees).
When code MH_SIN is applied to the data file thalweg, the first part of this code
creates a data file containing backsight bearings, the distance between transects, and
weights for each bearing backsite based on the proportion of between-transect distance
over which the bearings were measured. Next, the code creates variables equaling the
cumulative x ("Easting") and y ("Northing") coordinates for each transect, based on an
arbitrary downstream starting point and the bearings and distances from each transect to its
adjacent downstream transect. The straight line distance from the furthest upstream
transect K back to transect A is then calculated as the denominator of Equation 1. As
shown in the previous paragraph, sinuosity (Table 6) is calculated as the ratio of the total
(tortuous) distance along the course of the channel divided by the straight line distance
between the two ends of the reach. Code MH_SIN prepares a data file with cumulative
34
-------
x and y coordinates of each transect, then optionally plots a simple planform map of the
course of the reach. These plots may be used to compare stream reaches between visits
or years, or to validate backsighted bearing data. Calculations of sinuosity and reach
planform maps for streams with missing inter-transect bearings should be considered
suspect if the field data has missing inter-transect bearings (i.e. if TRAN_N <10). The
metric file created, mhjsin, includes the crow's distance and along-stream distance
between reach ends, reach sinuosity, distance between transects, and the number of
transects.
3.2.4 Slope
Code MH_SLP calculates the channel centerline length-weighted mean slope for
each sample reach. When the code is applied to data file thalweg, it creates a new data
set from the longitudinal profile data, from which mean and standard deviation of slope are
calculated for each reach. The code bases these calculations on backsighted slope
measurements between transects, weighting each slope measurement by the proportion of
the inter-transect distance to which it applies. Field crews enter one or more slope
backsites between transects, estimating these proportions. If supplemental measurements
are entered without specifying inter-transect proportions, we assume the measurements
were evenly-spaced between transects. For example, if two slopes are entered without
proportions, we assume that these measurements were halfway between transects, and
give weights of 50% to both the main and supplemental backsights. The code calculates
length-weighted means for each inter-transect distance. Because the inter-transect
distances within a reach are equal, the code calculates a length-weighted mean as the
arithmetic average of all the inter-transect slope values (Table 6). Missing inter-transect
slopes are not used in the calculation. The output data file becomes the permanent metric
file mhjslp**, where ** is set to the last two digits of the year of field measurements.
3.2.5 Residual Pool Analysis
Mean residual depth, reach-aggregates of residual pool area, and other residual
pool statistics are flow-independent measures of channel morphology that have distinct
advantages over highly flow-dependent descriptions, such as visual pool classifications.
Refining a concept first introduced in general terms by Bathurst (1981) in a discussion of
factors controlling hydraulic resistance due to gravel bars, Lisle (1982, 1986, 1987) defined
a "residual pool" as an area in a stream that would contain water even at zero discharge
due to the damming effect of its downstream riffle crest. Kaufmann (1987a,b) concurrently
applied this idea at long reach scales (30 to 40 times their baseflow wetted width). He
defined individual pool and aggregate stream reach residual pool longitudinal profile areas
35
-------
on the basis of tightly-spaced, systematic thalweg bed elevation measurements. Residual
pool depths are calculated from these longitudinal bed elevation profiles by projecting a
level horizontal line upstream from the downstream control point (i.e., riffle crest) of each
residual pool until it meets an elevation greater than or equal to the elevation of the control
point (Figure 4A). Subsequent residual pool characteristics are then based upon the
elevation difference between the residual surface and the stream bed. Stack (1989), Stack
and Beschta (1989), and Robison and Kaufmann (1994) subsequently developed rapid field
techniques for approximating residual pool characteristics from much less time-consuming
thalweg depth profile and reach water surface slope data. Robison and Kaufmann (1994)
and Robison (1998) describe the rapid approach in detail, reporting close agreement in
reach aggregate residual pool statistics compared with values determined using the more
rigorous (but more time-consuming) thalweg bed elevation technique that requires surveyed
bed elevation data.
In the rapid residual pool approximation, the residual surface is graphically
represented on a longitudinal depth profile by a line extending upstream from the
downstream riffle crest to a thalweg depth shallow enough for the bed to intercept this line.
Because this residual surface is, by definition, perpendicular to the gravity vector, the slope
of the water surface must be accounted for and the depth measurements corrected
accordingly. This correction appears as a downward "tilt" of the residual surface in the
upstream direction when it is plotted on a longitudinal profile of water depth (Figure 4B).
The relation between the angle of correction ("tilt") and the reach mean slope was
determined empirically by Stack (1989), to best match residual pool areas based on surveys
of bed elevation:
Correction angle of residual surface = 0.12 + 0.25(slope) (2)
Once a residual pool longitudinal profile of a reach is calculated, individual pool
dimensions and reach aggregate volumes or summary statistics can be calculated from
these profiles. Kaufmann (1987a) found reach aggregate residual pool vertical profile
areas to be closely related to the amount of large woody debris and transient hydraulic
storage volume in stream reaches in the Oregon Coast Range; the mean values for
individual pool volume and maximum residual depth varied with pool type and pool-forming
agents (e.g., debris jam, rootwad, single log, boulder). We describe below an approach for
calculating reach aggregate, and individual measures of the dimensions of residual pools
from longitudinal thalweg profiles measured using the EMAP survey methods of Kaufmann
and Robison (1994,1998).
36
-------
1082
E 1078
Residual Pool Profiles
Lookout Creek
1066
0
200 400 600
Distance (m)
800
Figure 4. Residual pool profile. A) based on channel thalweg elevation data; B) based on
thalweg depth and reach slope data. Adapted from Robison and Kaufmann, 1994.
Lisle and Hilton (1992) proposed an index of sediment supply for gravel-bed
streams, based on the volume of fine sediment in pools. Using field measurements of
residual pool depths and fine sediment layer depths at many points within pools (determined
by pounding a metal bar down through fine sediment to a hard gravel-cobble pool bottom),
this approach evaluates the proportion of the pool volume filled with fine sediments. The
fraction of pool volume filled with fines (V*) was calculated as the fines volume divided by
the sum of fines volume plus residual pool volume:
V* =
vf _ vf
V,
(3)
sp
where Vf, Vr, and Vsp are, respectively, the volume of fines in the pool, the residual pool
volume, and the scoured pool volume (i.e., Vf + Vr).
37
-------
Lisle and Hilton (1992) found this index of pool sedimentation to be positively related
to sediment production rates, and to the area and intensity of logging and road building in
Northern California stream drainage basins. For routine monitoring in surveys such as
EMAP, Kaufmann and Robison (1998) found it impractical to obtain all the residual depth
and fines depth information necessary to formally calculate V*. However, by collecting
information on the presence of "soft, small" substrate (< 16mm diameter) at 100 to 150
points along the thalweg, a related surrogate of V* can be calculated from EMAP survey
data as the percentage of the total length of residual pools in each stream reach that have
loose, relatively mobile fine sediments deposited along the thalweg (PCTPSED). We
expect that this index would be less sensitive to pool filling from upslope sediment
production than V*. However, we hypothesize that, as pool volume is filled by fine
sediments, the pool bottom surface area covered by fines will increase, and this trend
should be detected in EMAP surveys as an increase in the presence of fine sediments
within the residual pool portion of the thalweg bottom profile (as well as a decrease in the
depth and volume of the residual pools themselves).
The SAS code MH_RP prepares longitudinal thalweg profile data from the file
thalweg and then calculates individual and aggregate residual pool statistics. Using the
code, the downstream riffle crest (control point) of each residual pool in the longitudinal
thalweg depth profile is detected iteratively, by scanning the sequence of thalweg depths for
a value that exceeds the previous (downstream) depth by an amount determined using
Stack's correction angle equation, as follows:
depth0 > depth(M) + {[0.12 + 0.25(slope)] x interval} (4)
where:
depthw is the measured thalweg depth at the ith point in the upstream
direction along the sampled reach. Depthm, used for the first
measured depth(/ = 1), is taken to be the thalweg depth mean
minus the thalweg depth standard deviation, to allow detection of
the first pool in those cases where a sampled reach begins in a
pool;
7'is the thalweg depth measurement index, starting at 1 and continuing
to the upstream end of the reach, generally either 100 or 150;
slope is the mean water surface slope of the reach, and
interval is the distance between consecutive measurement points.
38
-------
Once this riffle crest is detected, the residual pool continues upstream until it's
calculated residual surface intersects the thalweg bed surface. This point is detected by
scanning the sequence of thalweg depths until the following relation is true:
depth(e) < depth(b) + {[0.12 + 0.25(slope)] x [interval x (e - Jb)]}; (5)
where:
depthw is the water depth at the estimated downstream riffle crest defining this
particular residual pool;
depthre; is the measured thalweg depth at the estimated upstream end of the
residual pool. It represents the point at which the residual surface intersects
the streambed. This point may or may not be the downstream control point
of the next upstream residual pool;
slope is the mean water surface slope of the reach;
interval is the distance between consecutive measurement points;
b is the value of the index / at which the downstream starting point of the pool
was detected using Equation 4;
e is the value of the index / at which this relation holds, marking the detected
upstream end of the residual pool.
The last term in Equation 5, [interval x (e - Jb)], is the length of the detected residual
pool. The residual depth is defined as the difference between the measured thalweg depth
and the calculated residual surface depth, as shown in Equation 6:
resd(/) = depth(/) - {depth(b) + [0.12 + 0.25(slope)][interval x (e - Jb)]} (6)
where:
resd(/) is the residual depth at the ith measurement point along the sample reach.
The "sagittal" area (i.e., the longitudinal profile area) of the residual pool is the
summation over the pool of the product of the thalweg depth at a point and the interval
between it and the measurement preceding it. This is given in Equation 7:
area
)
-------
Wetted widths are obtained at each transect and halfway between transects. We
calculate residual widths corresponding to each measured thalweg depth by linear
interpolation between measured width values. Assuming a triangular channel cross section,
the residual width is calculated from wetted width, thalweg depth, and residual depth using
the geometric relationship in Equation 8, as described by Robison and Kaufmann (1994)
and Robison (1998). Again, assuming a triangular channel cross-section, the volume of the
residual pool can then be calculated from both residual depths and widths using Equation 9:
(8)
volume = ^[(0.5xresdf/)¥reswf/j¥intervalY| (9)
where:
resw(/) is the residual width at measurement point /';
wt_wid(/) is the measured wetted width at measurement point /'.
The SAS code MH_RP creates a temporary data set containing a downstream
starting control point depth equal to the mean thalweg depth in the reach minus its standard
deviation. This surrogate control point depth is prepended to the sequence of measured
depths and used as an initial depth for detecting residual pools. This allows the detection
and inclusion of a portion of a residual pool which may be located at the downstream
starting point of thalweg measurements. The temporary data set also contains the mean
slope of the reach, the reach length, the distance (increment) between inter-transect
measurement locations (STA_NUMs), and a value equal to the number of measurements
between transects minus one, as well as variables used for identification and other
accounting processes.
The metric calculation code (MH_RP) then performs the following operations on the
temporary data set to calculate aggregate residual pool metrics and statistics:
1) The relevant data (depth, sediment presence/absence, wetted width,
TRANSECT, STA_NUM, and a flag denoting if the data are in the main
or a side channel) are transposed so that values for an entire reach are
in a single record.
40
-------
2) The beginning and end of each 'segment' is determined, where a segment is
a section of the data sequence bounded by either the beginning of data
collection, the end of data collection, or by a gap in the data lasting
longer than 10 intervals (or stations, if you prefer). A reach consists of
a segment for the main channel, and one or more for the side channels
(if present and sampled).
3) Determine the 'head' of each defined segment. These are points which put a
segment in its context - they may be a point in the main channel (where
a side channel empties into it) or the surrogate starting control point (if
the segment is either the main channel or a side channel which
empties into the main channel below the study reach).
4) The pool calculations are done for each segment to find incremental lengths,
sedimented lengths, depths, widths, areas and volumes of the residual
pools.
5) Side channel pools occurring at channel confluences are joined to the main
channel pool at the confluence if present.
These first five steps are done for each reach in the study. At the completion of all reaches,
and prior to the next step, it is possible to plot out the depths, residual pool surfaces, and
sediment locations for each reach.
6) The incremental values are then transposed to place values for each
measurement point on separate records. They are then summed over
the length of the individual pools to obtain residual dimensions and
derivative statistics, such as means and quartiles. These values are
saved in a file mhrpind**, where ** is set to the last two digits of the
year that the study took place.
7) Reach summary metrics are calculated from the data for individual pool data
within each reach (Table 6). These are stored in the file mhresp**,
where ** is set to the last two digits of the year that the study took
place.
41
-------
3.2.6 Substrate Size and Composition
Systematic channel pebble counts can be directly reduced to whole-reach substrate
characterizations by calculating percentages of observations within stated size classes.
Because the data are systematically spaced, these averages and percentiles are interpreted
as unbiased representations of the substrate characteristics measured. For example, the
percent sand substrate obtained from the 55 particles in the EMAP pebble count (Kaufmann
and Robison, 1998) is an estimate of the percent of the surface area of the stream reach
that has surficial substrate composed of sand.
In an approach adapted from that of Bain et al. (1985), we calculate a mean
substrate diameter after assigning numeric values to the substrate size classes that are
proportional to the logarithm of the midpoint diameter of each size class. We calculate
substrate mean diameter class (SUB_X) and its standard deviation (SUB_V) as the
arithmetic mean of the numerically-transformed size classes (shown in the following
paragraph). The logarithmic nature of the substrate size classes specified in the EMAP
methods makes these mean size class values proportional to the geometric mean substrate
diameter. Based on assigning geometric midpoint diameters to each particle diameter
class, we derived the following relationship to transform mean diameter class values into
estimates of the Iog10 of mean substrate diameter in millimeters (LSUB_DMM):
If SUB_X < 2.5 then LSUB_DMM = -4.61 + (2.16 x SUB_X) (10a)
If SUB_X > 2.5 then LSUB_DMM = -1.78 + (0.960 x SUB_X) (10b)
When code MH_SUBS is applied to data file sub_bank, it uses the frequency of particles in
each substrate size class to calculate reach level percentages in each size class (from
Table 6) and in the following combined size classes:
PCT_SAFN = % substrate in size classes smaller than sand (< 2mm)
PCT_SFGF = % substrate in size classes < fine gravel (< 16 mm)
PCT_BIGR = % substrate in size classes larger than fine gravel (> 16mm)
PCT_BDRK = % substrate as smooth or rough bedrock.
PCTJDRG = % substrate composed of wood or organic detritus
This code assigns numeric values to the various substrate size classes. Substrate
classes RR, RS, HP, and RC (concrete) are assigned a value of 6 (see Section 2.2.4 for
code definitions). Class BL is assigned a value of 5, CB a value of 4, GC a value of 3.5, GF
42
-------
a value of 2.5, SA a value of 2, and FN a value of 1. The program then calculates simple
reach level means, standard deviations, and percentiles for substrate size classes and lists
the number of non-missing values used in the calculations. Using Equations 10a and 10b, it
calculates the Iog10 of geometric mean substrate diameter (mm) from mean substrate size
class values (Table 6).
When computer code MH_EMB is applied to data file sub_bank, it calculates a
reach mean and standard deviation of substrate embeddedness for the channel alone and
for the channel plus channel margin together (Table 6), as well as listing the number of non-
missing values used in the calculations. Note that observations of sand and fine substrates
are assigned embeddedness values of 100%, and bedrock (smooth or rough), hardpan, and
concrete are assigned embeddedness values of 0%.
3.2.7 Bed Substrate Stability
Many human activities directly or indirectly alter the size and composition of stream
substrates. Consequently, excessive erosion, transport and deposition of sediment in
streams and rivers is a major problem in surface waters throughout the United States. In
fact, the 1996 National Water Quality Inventory [Section 305(b) Report to Congress] ranked
sediments as a leading cause of water quality impairment in assessed surface waters.
Accumulations of fine substrate particles fill the interstices of coarser bed materials,
reducing habitat space and its availability for benthic fish and macroinvertebrates (Platts et
al., 1983; Hawkins et al., 1983; Rinne, 1988). However, substrate sizes vary naturally in
streams of different sizes and slopes. In order to evaluate impairment from sedimentation,
it is essential to have some measure of how much the substrate size (e.g., % fines) in a
stream deviates from that expected in the absence of human activities. The bed substrate
stability index we derive herein is a measure of stream bed textural "fining" that occurs as a
response to increases in the rate of upland erosion, and the increased mobility or instability
of the bed substrate that accompanies such inputs of fine textured substrates.
Among streams flowing at the same slope, larger streams naturally tend to have
coarser substrates than lower gradient streams, because their generally deeper flows exert
more shear stress on their beds and tend to quickly transport fine substrates downstream
(Leopold et al., 1964; Morisawa, 1968). The size composition of a streambed depends on
the balance between the rates of supply of various sediment sizes to the stream and the
rate at which the flow takes them downstream, i.e., the stream's sediment transport capacity
(Mackin, 1948, Schumm, 1971). The sediment supply rate and the type and size of
particles delivered to a stream by upslope erosion and mass transport is controlled and
influenced by basin characteristics, including lithology, topography, climate, vegetative
43
-------
cover, runoff characteristics, and land disturbances. On the other hand, the potential
erosive, or transport, capability of a stream is largely dependent on its slope, watershed
area and runoff regime, characteristics which determine the velocity and depth of water
flow. This potential transport capability can be lessened by channel features that impart
hydraulic roughness and dissipate energy.
Good quality in-channel habitat is generally neither excessively stable (substrate
coarse relative to transport capability), or unstable (substrate fine relative to transport
capability). Some movement of the streambed is beneficial and essential to maintaining
habitat quality, because it allows flows to scour and rework substrates to maintain complex
pool habitat and to clean gravels that are important for fish spawning and production of
aquatic invertebrates. Although natural rates of sediment input vary among and within
regions, human activities can alter these inputs. Excessive watershed erosion from these
activities can transport large amounts of fine sediments into streams, leading to frequent
bed mobility and poor instream habitat. Conversely, some human alterations like dredging,
channelization or upstream impoundments, may lead to a lack of fine sediments in some
parts of the channel, but an excess in other places. Clearing vegetation from banks and
riparian areas may increase siltation and reduce large woody debris in streams. Logging or
farming up to the stream banks, building roads across or along streams, dredging and
straightening the stream channel, and building dams or other diversion structures in the
stream channel may destabilize stream banks and change bottom substrate size and
composition.
Wilcock (1998) argues that, for gravel-bed rivers, a threshold of sand proportion
greater than 20% to 40% initiates bed instability, as the bed transitions from a "framework-
supported" bed of gravel and cobble to a "matrix-supported" bed in which these larger
particles are supported by mobile sand and finer materials. However, these substrate
stability guidelines are limited to streams that naturally have gravel or coarser substrates.
By comparing the size range of streambed sediments with a stream's erosive capability (i.e.,
bed shear stress) during typical flood conditions (e.g., Dingman, 1984: Dietrich et al., 1989:
Buffington and Montgomery, 1992: Buffington, 1995: Montgomery et al., 1999), we can
assess bed stability over a wider range of stream slopes, drainage areas, and substrates. If
most of the streambed sediments are finer than the size the stream is capable of moving,
those sediments move frequently, and are therefore relatively unstable. Following is a
derivation of the relative bed stability index (Dingman, 1984), adapted to the type of data
collected in a regional survey such as EMAP.
Sediment transport theory (e.g., Simons and Senturk, 1977) allows an estimate of
the average streambed shear stress or erosive tractive force (rbf) on the bed during bankfull
44
-------
flow. Stream channels can be very complex, exhibiting a wide range in local bed shear
stress due to small-scale spatial variation in slope, depth, and roughness within a channel
reach. The following estimate of rbf is an expression of the average shear stress over the
bed of the entire reach over which the measurements of channel morphology are averaged:
(11)
where:
rbf = Bankfull channel bed shear stress (Force per unit area = kg-m-s~2= N-m~2)
pH20 = mass density of water [kg-m~3]
g = gravitational acceleration [m-s~2]
Rbf = Bankfull channel hydraulic radius = Cross-sectional area •*• wetted perimeter
[m]
S = Channel water surface slope (Dimensionless: m/m)
Assumptions: This estimate of Tbf pertains to flow conditions that are uniform (spatially), and
steady (relatively unchanging in time as water flows through the reach). These conditions
are approximated by making measurements over long reaches within which discharge can be
assumed to change only very slowly. Paola and Mohrig (1996) argue that quasi-steady flow
can be assumed if discharge fluctuations occur on time scales » u/gS, while quasi-uniform
flow can be assumed if reach lengths are » Fi/S; where u is mean velocity, g is gravitational
acceleration, R is hydraulic radius (approximately flow depth), and S is slope. Buffington
(1998) used Paola and Mohrig's criteria to evaluate the suitability of applying Equation 11 in a
study partitioning shear stress in complex pool-riffle channel reaches. We followed the same
logic in examining the characteristics of 150 EMAP sample reaches in the Mid-Atlantic region
survey to determine how well they fit these criteria for approximating steady, uniform flow.
Assuming bankfull mean velocities of~1m~2, the median value ofu/gS was 0.2 minutes with
a maximum of 3 minutes in these sample reaches. The quasi-steady flow assumption
appears reasonable: even during storm events, discharges could easily be assumed to
remain approximately constant for much longer periods of time. The quasi-uniform flow
assumption is somewhat strained, but still reasonable, as more than half the 150 reaches in
the same survey had reach lengths 6 times R/S. Less than 25% had reach lengths shorter
than 2. 5 times R/S, and these were reaches with slopes < 0. 5%. Therefore, the calculation
of reach-average shear stress as TM = pH2o9^b^ IS probably a reasonable approximation
overall, but it is better for moderate to high gradient than for low gradient streams (slope <
1%).
Bed particle movement by erosion is dependent upon particle size, water depth and
velocity, the difference between the mass densities of fluid and particles, and the shape and
45
-------
arrangement of particles. The shear stress necessary to move a particle, the critical shear
stress (rc), can be defined as a function of particle diameter as follows:
(12)
where:
8 = "Shields parameter", expressing the erodibility of particles as related to their
shape, arrangement, and the presence and size of other particles. It is
experimentally determined by relating shear stress to particle diameter.
Yalin and Karahan (1979) report a value of 0.044 for relating D50 of non-
cohesive, spherical particles > 0.1 mm diameter to shear stress in "fully
rough" turbulent flows (i.e., Reynolds number > 70 and values of RbfS > 10~4
m). The great majority of natural streams are "fully rough" (e.g., > 99% of
150 stream reaches sampled by EMAP in the Mid-Atlantic region passed
these criteria).
Psed ~ PH o = The difference between the mass densities of sediment particles
and water [kg-m~3]
D = the substrate particle diameter [in m, if rc is in N-nr2].
By setting critical shear stress equal to bankfull shear stress,
Tc = Tbf (13a)
(13b)
we calculate the critical (or maximum) diameter of particles that can be transported at
bankfull flow (Dcbf). Because it is calculated from a representation of the average or median
bankfull shear stress within a long reach of channel, Dcbf is an estimate of the average or
median substrate critical diameter within the reach:
46
-------
2
°cbf =7-7 Hi (14a)
# Psed -PH20 9
Substituting g=9.81 m-s2, B = 0.044, psed = 2,650 kg-nr3 (average density for silicate
mineral substrate particles), and p^ o = 998 kg-m~3 for freshwater at 20 °C, Equation 14a
simplifies to the following expression, where Dcbf and Rbf are in the same units (e.g, if Rbf is
in mm, then the estimate of Dcbf is mm):
Dcbf=13.7RbfS (14b)
A first-cut of these estimates makes the simplifying assumption of a wide
(Width/Depth > 20), longitudinally uniform channel of triangular cross-section, lacking large
woody debris and pool-riffle structure. Under these wide channel assumptions, hydraulic
radius can be approximated by mean channel depth, because wetted perimeter is then
approximately equal to wetted width. Then assuming approximately triangular flow cross-
sections (see Robison and Beschta, 1989), mean depth of a channel cross-section can be
calculated as one-half maximum (thalweg) depth. Rbf, the mean bankfull hydraulic radius
(m), is calculated from EMAP data by adding mean bankfull height (XBKF_H ) plus mean
thalweg depth [(0.5 x XDEPTH)/10], and then dividing the sum by two (dividing XDEPTH by
10 converts thalweg depth to m). This first estimate (pgRbfS) gives the potential bed shear
stress, or total bankfull shear stress based on a hydraulically simple, low-roughness channel
of a size and slope that would be determined mostly by topographic and climatic
considerations (i.e., slope and discharge). The effective shear stress experienced by
particles on the bed of actual stream channels is typically reduced from this potential value
by sources of roughness that differ among streams, including bank irregularities, bars, and
wood (Einstein and Banks, 1950, Nelson and Smith, 1989, Buffington, 1998).
In an approach analogous to those of other researchers (e.g., Buffington and
Montgomery, 1992; Buffington, 1995), we incorporate terms to accommodate the reduction
in shear stress and consequent bed textural "fining" that results from the presence of large
woody debris and channel cross section irregularities. In our approach, we adjust Dcbf by
reducing Rbf in Equation 14b by the approximate roughness height of large woody debris
and pool-riffle scale channel irregularities. We calculate the effective hydraulic radius, R*bf,
by subtracting from Rbf the large woody debris "mean depth" (Rw = m3 wood volume per m2
47
-------
channel surface area = m wood "depth" = V1W_MSQ), and the cross-section mean residual
depth (Rp = one-half the thalweg mean residual depth = [0.5 x RP100]/10), multiplying
RP100 by 0.5 to approximate mean depth in a triangular cross-section, and dividing by 10 to
convert cm to m. Note that in the calculation R*bf = (Rbf - Rw - Rp), we make the simplifying
assumptions that woody debris volume per channel bottom area is uniform over the wetted
and bankfull channel, and that residual depth is a reasonable surrogate for the large scale
roughness due to bars and irregularities in channel cross-section. There is good
justification for using residual depth as a surrogate for large scale roughness, as residual
depth has been shown to be related to transient hydraulic storage volume and, if discharge
is taken into account, also to the Darcy-Weisbach friction factor (Kaufmann, 1987a; 1987b).
However, the assumption of uniform large woody debris spacing likely underestimates the
reduction in shear stress due to LWD. It might be possible to apply an adjustment to the
equations to augment Rw to account for LWD "dumpiness," or to alter the approach for
encorporating the roughness due to pools, bars, bends, and other channel irregularities.
The current calculations make the working assumption that the mean roughness "heights"
of LWD and pool-riffle structure translate directly to reductions in effective hydraulic radius.
Other approaches to more realistically incorporate the influence of roughness elements
typically require more intensive data collection to describe their size, shape, spacing and
orientation in relation to stream flow (e.g., Shields and Gippel, 1995; Buffington, 1998). The
EMAP field data lack the detailed information that would allow direct adjustments to account
for these complexities.
Finally, we calculate RBS, a measure of Relative Bed Stability (Dingman, 1984) as
the ratio (D50/D*cbf) of the observed substrate median diameter (approximated by the
geometric mean diameter) divided by the average critical diameter at bankfull flow (the
reach average for the largest particle that is mobile during bankfull flow):
PDC _ "50 _ "50
KDO — j — —. r- / -I C\
Dcbf (l3.7R*bfS) (15)
Similar comparisons of observed substrate size with that predicted from shear stress
have been used to evaluate the effects on substrate from changes in sediment supply (e.g.,
Montgomery et al., 1999) or large scale roughness elements such as LWD (Buffington,
1995; 1998). The RBS ratio is also conceptually similar to the "Riffle Stability Index" of
Kappesser (1993), and is also similar to the ratio discussed by Dietrich et al. (1989) of
median diameters of the substrate armor layer divided by that of the substrate beneath that
layer, which is taken to be the bedload.
48
-------
In calculating RBS using EMAP physical habitat data, we approximate the D50 by the
bed surface substrate geometric mean diameter, SUB_DMM, and the reach average critical
diameter (D*cbf) by DMB_BW4, both expressed in mm (see Table 6). Note that R*bf must
also be expressed in mm to match DMB_BW4. For convenience and to normalize their
variances, we find it useful to express RBS values as logarithms. An RBS value of 1
corresponds to an LRBS (i.e., Log10 of RBS) value of 0. A value of LRBS equal to 0 results
when a stream has D50 = D*cbf, indicating that the average bed shear stress during bankfull
flows is sufficient to move the median (or in this case, the geometric mean) particle size.
There is some uncertainty, when D50 = D*cbf, whether, in a particular stream, particles
smaller than D50 will be selectively mobilized at lower flows, with different sizes of particles
moving at progressively higher flows; or if the entire bed will become mobile in threshold
fashion (see discussion by Montgomery et al., 1999). Parker et al. (1982) postulated that, in
relatively coarse stream beds with a wide range of particle sizes, small grains "hidden" by
larger armor-layer particles become mobile only when the larger particles move. At that
threshold, the whole bed becomes mobile. A further complexity advanced by Wilcock
(1997) is that partial mobility may occur in some channels, where only some substrate
particles of a given size are mobile at a particular flow. Montgomery et al. (1999) observe
that, at present, there is no scientific consensus regarding the particular conditions under
which threshold (equal) mobility, selective transport, or partial mobility of stream bed
substrates will occur. These authors suggest, however, that selective transport may occur
in highly sediment-laden rivers, but when sediment supply rates decline, this process over
time results in the winnowing away of fines and the development of an armor layer. They
further suggest that". . . as selective transport represents high or continuous sediment
loading (Sutherland, 1987), . . . threshold mobility (Parker et al., 1982) characterizes
channels with relatively low or intermittent sediment supply
On the basis of the previous discussion, we may assume that if D*cbf is accurately
estimated, and that the RBS ratio D50/D*cbf is < 1, then at least half the bed substrate
particles become mobile during bankfull flows that typically occur every year or two. A high
positive value of RBS (e.g., 1000, an LRBS value of 3.0) indicates an extremely stable,
immovable stream substrate like that in an armored canal, a "tailwater" reach below a dam,
or other situations where the sediment supply is low, relative to the hydraulic competence of
the stream to transport bedload sediments downstream (Dietrich et al., 1989). In contrast,
very small values (e.g., 0.003, an LRBS value of -2.5) indicate a channel composed of
substrates that are frequently moved by even relatively small floods. Note that LRBS values
are logarithms of ratios, so a value of -2.5 denotes a stream in which bankfull flows have
sufficient mean tractive force to move particles with diameter 300 times larger than the
geometric mean particle size in the stream.
49
-------
The RBS ratio D50/D*cbf is not only a measure of the mobility of stream bed
substrates, but also gives an indication of the supply of sediment to the stream channel. An
increase in the percentage of fine substrate particles ("textural fining") in a stream bed often
occurs when sediment supply is augmented due to land use activities that increase hillslope
erosion (Lisle, 1982; Dietrich et al., 1989; Lisle and Hilton, 1992), suggesting an augmented
sediment supply in relation to the stream's downstream bedload transport capability
(Dietrich et al., 1989). Buffington (1998) hypothesized that for a given bed shear stress,
stream bed substrate size (D50) should be inversely related to sediment supply, because
sediment supplies that overwhelm the local sediment transport capacity would reduce the
bed-surface D50 through deposition of fine-grained particles that are typically in transport.
The RBS ratio D50/D*cbf is therefore an indicator of sediment supply, but the relationship
between RBS and sediment supply should be quantitatively proportional only if the channel
is at equilibrium, i.e., if the rate of sediment transport through the channel is equal to the
rate of sediment supply (Buffington, 1998). Even in streams draining "pristine" watersheds
that are at equilibrium between sediment supply and transport, one might expect different
characteristic values of RBS that are dependent upon the "natural" rates of erosion
(Buffington 1998). In the absence of human activities, these natural erosion rates would
depend upon climate, basin geology, geomorphology, channel position within the
watershed, and related features such as glaciers and natural landslide frequency. However,
even if a stream channel is not yet in equilibrium with a recently augmented sediment
supply or a pulse of sediment influx (e.g., a landslide), a decrease in its RBS ratio will
indicate the increased sediment supply, but the ratio cannot then be used to estimate the
magnitude of the sediment supply.
We hypothesize that, given a natural disturbance regime, sediment supply in
watersheds not altered by human disturbances may be roughly in long-term equilibrium with
transport. RBS values for streams draining watersheds relatively undisturbed by humans
should tend toward a characteristic value typical to the region. Dietrich et al. (1989) argue
that, in most relatively undisturbed watersheds and streams, low hillslope erosion rates will
allow some surface coarsening or armoring of streambeds. In these situations LRBS
should be near or slightly above 0, and this is what we generally observe in the EMAP
stream populations sampled in the mid-Atlantic region and in Western Oregon (P.R.
Kaufmann, unpublished data). The least-disturbed of EMAP streams sampled in the
Midwest Cornbelt/Great Plains generally tend to have LRBS values between -0.5 and +0.5.
In all of these regions, progressive intensity of human land uses is generally associated with
progressive sediment "fining", indicated by declining values of LRBS (P.R. Kaufmann,
unpublished data). Highly disturbed basins typically had LRBS < -1.0 in the mid-Atlantic
region, and < -2.0 in Western Oregon and the Midwest Cornbelt/Great Plains, except where
50
-------
streams were extensively channelized, revetted, or dredged. In such cases, we often
observed extremely high LRBS values (e.g., > 2.0).
Based on these considerations, we have set condition thresholds after workshop
discussions with regional stream ecologists. They are expressed as LRBS, the Log10 of
RBS, and are set in the spirit of hypotheses to be refined as we learn more about the
reference levels for RBS in undisturbed streams in various regions, associations between
RBS and disturbance, and the relationships between RBS and biota:
Mid-Atlantic Highlands Cornbelt/Great Plains
"Good Condition" >0.2 to 1.0 >-0.5 to+0.5
"Impaired" >-1.0 to + 0.2 and >1.0 to 2.0 >-2.5 to -0.5 and >0.5 to 2.5
"Highly Impaired" <-1.0 and >2.0 <-2.5 and >2.5
Instructions for calculating these bed substrate stability metrics from EMAP physical
habitat data are contained in Table 6 and in Appendix D (on the compact disk). These
instructions include algorithms for calculating LDMB_BW4=Log10(D*cbf), the estimated critical
median diameter (in mm) for bed substrate under bankfull conditions, with adjustments for
Large Woody Debris and channel complexity. We also give instructions for calculating the
estimated relative bed stability under bankfull conditions (LRBS_BW4 = Log10[D50/D*cbf]).
The calculations must be made from a dataset that merges channel, substrate, residual
pool, woody debris, and fish cover variables, as these are not contained in the single PHab
data files. In the EMAP data management system, data sets named phabmet or phabbest
contain the required suite of variables for these calculations. The estimates LTEST and
LRBS_TST are also calculated using SAS code in Appendix D (compact disk) and are
contained in the EMAP data files. They are empirical approximations of the more refined
estimates LDMB_BW4 and LRBS_BW4, but do not explicitly involve bankfull depth, residual
pool depth, and large woody debris volume in calculations. They are rough empirical
approximations using one-half the mean thalweg depth as a surrogate for effective bankfull
hydraulic radius: R*bf ~ (0.5 x [XDEPTH/10]), a relationship that may sometimes be
inaccurate, but can be derived as a quick "ballpark" estimate requiring very little field data.
3.2.8 Fish Cover
Field data estimating the presence and cover of fish concealment features according
to field procedures described by Kaufmann and Robison (1994, 1998) consists of visual
estimates of the cover class category of eight specific types of features in 11 observation
plots distributed along each stream sample reach. To calculate reach-level metrics of fish
51
-------
cover, apply code MH_FCVto the data file fishcov. The metric summaries calculated by
this code are whole-reach averages, based on cover or presence estimates at 11 stations.
For each fish concealment type, field crews estimated areal cover in four classes: absent
(0), sparse (0 to 10%), moderate (10 to 40%), heavy (40 to 75%) and very heavy (> 75%).
Based on cover estimation techniques described by Daubenmire (1968), reach fish cover
metrics are calculated by assigning cover class midpoint values (i.e., 0%, 5%, 25%, 57.5%,
and 87.5%) to each plot's observations and then averaging those cover values across all 11
stations. These calculations yield metrics of single types of fish concealment features listed
in Table 6 (e.g., XFC_BRS -- the proportional areal cover of brush and small woody debris).
In addition, the code calculates cover estimates of the following combined cover types,
whose summed cover may exceed 100% (i.e., proportional areal cover > 1.0) :
XFC_ALL = sum of proportional areal cover from all types of fish "cover"
excluding algae and aquatic macrophytes.
XFC_NAT = sum of proportional areal cover from natural concealment features
(rocks and boulders, overhanging vegetation, brush, LWD, and undercut
banks)
XFC_BIG = sum of proportional areal cover from large features (Rocks and
boulders, LWD, undercut banks, and artificial structures).
Further summarization of fish cover information is accomplished by calculating reach
fish cover presence metrics as the fraction of the 11 in-channel plots that had cover values
> 0 for each category of fish concealment feature (e.g., PFC_BRS) or for defined
combinations of features (e.g., PFC_BIG in Table 6).
3.2.9 Large Woody Debris
Large woody debris (LWD) pieces observed between each transect and the next
upstream transect in each stream reach are tallied on the EMAP habitat field forms
according to 12 diameter and length size classes, with indication whether the piece is in or
outside the bankfull channel. To calculate LWD metrics, code MH_WOOD is applied to
data file Igwoody. A nominal size-class volume is assigned to each piece, based on
empirical pilot studies in which the dimensions of every piece of woody were measured
(Robison, 1998). Reach summary metrics are then calculated as the total number and
estimated volume of wood in various size classes.
Tallies and wood volume metrics are calculated for the original 12 length-diameter
classes; separately for wood in the active channel, wood spanning but not in the active
channel, and the combined total wood in the reach:
52
-------
Length
Diameter
S (0.1 m to 0.3m)
M (> 0.3 m to 0.6 m)
L (> 0.6 m to 0.8 m)
X (> 0.8 m)
S (1.5m to 5 m)
T
S
S
M
M ( > 5 m to 1 5 m)
S
M
L
L
L(> 15m)
M
L
L
X
The codes T, S, M, L, and X in this table are progressively larger piece sizes from
very small to very large. A nominal mean volume is calculated for each piece of LWD
according to its diameter-length class membership as described and tested by Robison
(1998):
Volume=7r
.33[
Class Minimum Diameter^
pl.33(Class Minimum Length)] (16)
Total numbers and volumes of LWD in each diameter-length class are regrouped
and assigned to one of five cumulative wood size classes:
Class 1 - T, S, M, L, X (Very small to Very Large)
Class 2 - S, M, L, X (Small to Very Large)
Class 3 - M, L, X (Medium to Very Large)
Class 4 - L, X (Large to Very Large)
Class 5 - X (Very Large)
As done for the separate diameter-length classes, MH_WOOD calculates separate
summaries for the two locational classes plus a combined total, expressing them on per
reach, per 100 m, and per m2 basis. Unless noted in the data set, missing values in the
wood tallies are assumed to be zero.
3.2.10 Riparian Canopy Cover (Densiometer)
The multiple canopy densiometer measurements collected in the EMAP field
methods (Kaufmann and Robison, 1994, 1998) can be directly reduced to whole-reach
canopy density characterizations by calculating their means and standard deviations.
Because the data are systematically spaced, these averages and percentiles are spatially
representative estimates of canopy density on and along the stream. When code MH_DEN
53
-------
is applied to data file canpycov, it calculates separate reach level summary statistics for
mid-channel and bankside canopy cover. Mean densiometer readings, their standard
deviation, and the number of non-missing observations used in these calculations are
generated separately for the 44 instream measurements (4 observations x 11 stations), and
the 22 bank measurements (2 observations x 11 stations). These metrics are converted to
percent canopy density by dividing the mean and standard deviation of densiometer reading
values by 17, the highest possible canopy densiometer value, and multiplying the result by
100.
3.2.11 Riparian Vegetation Structure
Riparian vegetation cover field data collected by EMAP methods consist of visual
cover class estimates for a number of specific features or types of vegetation in multiple
observation plots distributed along each stream sample reach. The desired metric
summaries are whole-reach averages, rather than multiple separate cover or presence
estimates for each transect or riparian plot.
Riparian vegetation type and areal cover are visually estimated in three layers at
each of the 22 riparian vegetation plots located at the left and right sides of 11 transects.
The vegetation layers are: canopy (> 5 m high), mid-layer (0.5 to 5 m high) and ground
cover (< 0.5 m high). Coniferous, deciduous, and broadleaf evergreen vegetation types are
distinguished in the canopy and mid-layer. Canopy layer tree cover is divided into large
diameter (> 0.3 m) and small diameter (< 0.3 m) trees; woody and herbaceous vegetation
cover are distinguished in the mid-layer; and woody, herbaceous, and barren ground are
distinguished in the ground cover layer. For each of the vegetation layer categories, areal
cover is estimated in four classes: absent (0), sparse (0 to10%), moderate (10 to 40%),
heavy (40% to 75%) and very heavy (> 75%).
When code MH_VGPC is applied to the data file riparian, it produces two
permanent metric data files, mh_canrar\d mh_canp. Reach riparian cover metrics
(mh_canr) are calculated by assigning the cover class mid-point value, as described for fish
cover, to each riparian plot's observations and then averaging those cover values across all
22 stations. These calculations yield the single vegetation type metrics listed in Table 6
(e.g., XCL, the areal cover proportion of large diameter trees). Further summarization of
riparian vegetation information is accomplished by summing the areal cover or tallying the
presence of defined combinations of riparian vegetation layers or vegetation types. The
riparian vegetation metrics that sum more than one layer may have areal cover proportions
greater than 1.0 (i.e., cover > 100%):
54
-------
XC = total canopy level cover (XCL + XCS)
XM = total mid layer veg. cover (XMW + XMH)
XG = total ground cover (XGW + XGH)
XCM = total canopy and mid veg. layer cover (XC + XM)
XCMG = total veg cover from canopy, mid and ground layers (XC + XM + XG)
XCMW = total canopy plus mid layer woody cover (XC + XMW)
XCMGW = total woody veg. cover in all three cover layers (XC + XMW + XGW).
In the second metric data file (mh_canp) calculated from riparian, the code
MH_VGPC assigns an indicator of presence to each type of vegetative cover. Reach
riparian cover presence metrics are calculated as the fraction of the 22 riparian plots with
non-zero cover values for any given vegetation-layer category (e.g., XPCL, the proportion of
the reach that is bordered by large diameter trees). As for riparian cover metrics,
combinations of cover presence are calculated (Table 6).
Finally, estimated percentages of the reach riparian area with canopy and midlayer
comprised of deciduous, coniferous, broadleaf evergreen, or mixed vegetation types (Table
6) are calculated by applying code MH_CMTYP to the data file riparian.
3.2.12 Riparian Human Disturbances
In the EMAP field methods, crews record the presence and proximity of 11
predefined types of human land use or disturbance based on 22 separate visual
observations at both the left and right sides of the channel at 11 transect locations.
Observations are specified in three proximity categories: "B" within the channel or on the
stream bank, "C" within the 10 m x 10 m riparian sample plots, and "P" behind or adjacent
to the plots. For each of the 11 disturbance categories, we calculate the proportion of
riparian stations where the disturbance was in each proximity category (e.g., for roads,
these are calculated within the MH_HUM computer code as the variables BXPROAD,
CXPROAD, and PXPROAD). These metrics are calculated by applying code MH_HUMio
data file riparian. In addition, we calculate proximity-weighted disturbance indices by
tallying the number of riparian stations at which a particular type of disturbance was
observed, weighting each observation according to its proximity to the stream, and then
averaging over the 22 riparian stations on the reach (e.g., W1H_ROAD in Table 6).
Weightings were 1.5 for disturbance observations within the channel or on the stream bank
("B"), 1.0 for observations within the 10 m x 10 m riparian sample plots ("C"), and 0.667 for
those behind or adjacent to the plots ("P"). Aggregated metrics were calculated by
combining simple metrics of the various types of human disturbance observations (e.g.,
W1JHALL in Table 6).
55
-------
3.2.13 Metric Variable Labeling and File Merging
The code PHABMET merges all the separate types of physical habitat metric files
into the new file phabmet. If one desires a reduced set of variables, code PHABBEST may
be applied to phabmet to create phabbest, a file containing only the most commonly used
habitat variables. Although the separate metric calculation codes label all the variables they
create, we also provide the code MHLABELS to attach labels to every variable in the
merged metric data file phabmet or the separate metric files. To use the labeling code,
enter the permanent data file name of each metric file (or phabmet at the appropriate place
in MHLABELS.
56
-------
4 PRECISION OF HABITAT CHARACTERIZATION
4.1 THEORETICAL CONSIDERATIONS
Effective environmental policy decisions require stream habitat information that is
accurate, precise, and ecologically relevant (i.e., it contains useful information for
interpreting controls on the biota or impacts of human activity). We evaluated sampling
precision of field habitat survey methods employed by the USEPA's EMAP in probability
samples of several hundred streams in Oregon (Herlihy et al., 1997) and the Mid-Atlantic
region (Paulsen et al., 1991) between 1993 and 1996. We compared the within-year
variance among streams ("signal") with the variance between repeat stream visits within the
same year (measurement "noise"), combining our estimates of within-year variance over
four years of study. We employed this statistical model in our evaluation of the precision of
habitat metrics:
ijk
(17)
where / indexes years (/' = 1993 to 1996), j indexes stream reaches (j := 1 , . . . , n,), k indexes
visits to a particular stream reach (k = 1 , . . . , rm), Yijk is the measured metric value for visit
k to stream j during year /', /j is the grand mean among stream reaches, T,- is the mean
difference of metric values during year /from the grand mean, S^ is the mean difference of
stream j from the mean within year /', and Eijk is the residual variation that we have termed
"noise." We assume that T,- ~ (0, o2yr), S^
variance is as follows:
(0, o2st(yr)) and Eijk
(0, o2rep). The analysis of
Source
df
Mean Square
E(Mean Square)
Years
Streams(Years)
Residual
y-1
<«,-•*)
MSYear
MSResidual
a2rep
, + c2ayr
57
-------
The term a2st(yr) is the within-year variation among streams in a region that is not attributable
to measurement uncertainty or interannual variability, and these differences among streams
are frequently the "signal" of interest in a regional survey. Residual variation, or "noise"
variance (o2rep) in our model was estimated by pooling the variances of repeat visits to a
random subset of the probability sample of streams visited during the summer sampling
periods (in Oregon, the subset was a stratified random sample). We have attempted to
accurately represent the variances that would be encountered in a large-scale survey, not a
highly controlled research project in which all measurements can be made by the same
researchers at precisely the time that the data are needed. Therefore, in addition to
measurement variation, we recognize that the noise variance includes the combined effects
of within-season habitat variation, differences in estimates obtained by separate field crews,
and uncertainty in the precise relocation of the unmarked sample reaches (relocated on
subsequent visits using global positioning system (GPS) receivers, map, compass,
landmarks, and field notes). Because our purpose here is to examine the precision with
which habitat attributes can be measured during a period in which habitat quality is believed
not to have changed, we do not include variation between years, which may more likely
include actual changes in these attributes. Rather, we block our ANOVA by year,
examining the within-year variation in repeat measurements. We define the precision of
physical habitat metrics using three measures: orep, CV, and the signal:noise variance ratio,
' / 2
st(yr/O" rep-
S/N, calculated as o2st(vr)/o2
The first expression of metric precision, orep, is the root mean square error (RMSE)
from our variance model, and is equivalent to the pooled standard deviation of repeat
measurements (SDrep) of a habitat metric. The lower the value of orep for a given habitat
metric, the more precise the measurement. The units of this expression of precision are the
actual units of the habitat measurement. Consequently, orep is particularly useful in cases
where an investigator is familiar with the habitat attribute, its expected response to
disturbances, and the relationship between biota and changes in the numeric value of the
habitat metric. Precision comparisons based on orep are more difficult to interpret for
"indexes" that have less tangible meanings (e.g., W1_HALL, a proximity-weighted index of
riparian human disturbances). It is also difficult to compare values of orep among metrics
that are expressed in different units or have different potential ranges.
The second measure of precision is the coefficient of variation (CV =100 orep/Mean).
Many researchers use the CV as their primary expression of precision. Here, as is typical in
regional applications, the CV is calculated as the pooled SD of replicates (in this case
repeat visits) divided by the grand mean across sites. This CV can be misleading because
of its dependence on the regional mean, which may differ substantially among field
applications. For example, it would be reasonable to conclude that measurement precision
58
-------
is equal in two regional surveys in which canopy cover was measured with orep=0.1, even if
regional mean riparian canopy covers were 0.1 and 0.9. However, the respective CVs for
the two surveys would be [100 x (0.1/0.1)] = 100% and [100 x (0.1/0.9)] = 11%, indicating
vastly different "precision." The measure orep avoids this problem by expressing the
precision as equal in the two cases, and we feel it is more appropriate for evaluating and
comparing the precision of techniques applied across many streams.
In the third expression of precision, the S/N ratio (o2st(yr)/o2rep), we compare the
variance of the habitat metric observed across a regional sampling of streams ("signal") with
the "noise" variance resulting from field measurement within the sampling season. This
variance ratio is related to "intraclass correlation" or "heritability" (Snedecor and Cochran,
1980). Heritability is a similar ratio defined as o2st(yr)/(o2st(yr) + o2rep). The higher the value of
S/N, the more precise the metric is relative to the context of its regional variation. One
advantage of this measure of precision is that it facilitates comparisons among different
metrics. When the regional stream sample set and the subset of repeat streams are both
random samples, S/N is related to the F-ratio commonly used in analysis of variance for
evaluating the ability of a metric to discern differences among streams over the "noise" of
measurement variation. This F statistic is calculated as {MSstream(year)}/(MSResiduai), and is an
estimate of (o2rep + C1o2st(yr))/o2rep, where a, is a constant varying between 1 and r, the
number of times the repeat-sample streams are visited. If all the sample streams were
visited the same number of times, a, would be equal to r(see Neter and Wasserman,
1974). Our signal:noise ratio is related to the F statistic as follows:
S/N = Signal: Noise ratio = st(yr) =i - '- (18)
The higher the S/N ratio is for a habitat metric surveyed within a region, the more
that metric is able to discern differences among streams. If anthropogenic changes in
habitat are similar in type and magnitude to the differences observed among streams
across the region, then S/N is also a useful predictor of the metric's potential for discerning
trends or changes in habitat in single or multiple sites.
4.2 HABITAT METRIC PRECISION RESULTS
Three measures of precision, orep, CV, and the S/N ratio (o2st(yr)/o2rep), are presented
in a series of tables for a large selection of EMAP physical habitat metrics. They are
grouped in tables according to the type of metric as follows: Channel Morphology
measurements, Channel Habitat Classifications, Substrate, Fish Cover, Riparian
59
-------
Vegetation, and Riparian Human Activities and Disturbances. In cases where metric values
are tangibly understandable or have well-constrained ranges (e.g., substrate size
percentages and cover proportions), we compare precision within a group of similar metrics
primarily on the basis of orep in the following discussions, with consideration of S/N to
compare their precision with other types of metrics and to assess the likely utility of metrics
in a regional survey. Where the range of a variable is not rigidly constrained (e.g. width,
slope, depth), we rely mainly on values of S/N to express and compare precision. However,
orep gives a tangible measure of the precision of these metrics expressed in their units of
measurement (for example, it is inherently useful to know that mean Residual Depth was
measured with a precision of + 1.6 cm in the Mid-Atlantic region survey, regardless of the
magnitude of regional variation in that metric). We avoid focusing our assessment of metric
precision on CVs, but present these commonly used numbers and show how they can
frequently be misleading.
4.2.1 Channel Morphology and Habitat Classifications
Quantitative measurements of relatively flow-independent channel morphology (e.g.,
XSLOPE, SDDEPTH) and Residual Pool metrics (e.g., RP100, RPGT75, RPXAREA) were
quite precise, with S/N ratios ranging from 6 to 33 (Table 7). Even though mean depth,
mean width, and mean width-depth product (XDEPTH, XWIDTH, XWXD) vary somewhat
with flow stage, they were also reasonably precise, with S/N values ranging from 6.9 to 15.
Presumably because of their greater dependence upon flow stage, mean width-depth ratio
(XWD_RAT), standard deviation of width-depth ratio (SDWD_RAT), mean bank angle
(XBKA), and mean undercut distance (XUN) were imprecise (S/N values mostly < 2.9).
Channel percentages in various habitat classifications (e.g., % Pool Habitat) are
similarly dependent on flow stage, but are also considerably dependent on personal
judgement; as expected, these were relatively imprecise, with S/N ratios mostly < 2 (Table
8). Exceptions that were alternately quite precise (S/N = 7.5 to 21) or very imprecise (S/N
< 2) in one or the other regional survey included %Falls, %Cascades, and the aggregated
metrics %Fast Water and %Slow Water Habitats. Compared with other channel habitat
features, these classifications are less variable within the summer baseflow sampling period
and more reliably recognized in the field. Stream habitat unit classifications lack
repeatability both because of their subjectivity and their flow-dependency (Platts et al.,
1983). One commonly reported channel description, %Pool, had orep= 11 to 16%, CV = 48
to 88%, and measured values varied almost as much between visits as among streams
(S/N = 1.2 and 2.1 in the two surveys). These findings, in general agreement with Ralph et
al. (1994), Roper and Scarnecchia (1995), Wood-Smith and Buffington (1996), and Poole et
al. (1997), are of concern, because many stream monitoring efforts collect this type of data
60
-------
TABLE 7. PRECISION OF PHYSICAL HABITAT METRICS FOR QUANTITATIVE STREAM
CHANNEL MORPHOLOGY IN THE MID-ATLANTIC REGION AND OREGON
(for the Mid-Atlantic Region, n=169 with 50 replicates; for Oregon, n=44 with 22 replicates)
Variable Name — Description
CHANNEL MORPHOLOGY METRICS3
XDEPTH - Thalweg mean depth (cm)
SDDEPTH - Thalweg Std. Deviation of depth (cm)
XWIDTH - Mean Wetted Width (m)
SDWIDTH - Std. Deviation of Wetted Width (m)
XWXD - Mean Width-Depth Product (m2)
SDWXD - Standard Deviation of Width-Depth
Product (m2)
WD_RAT - Mean ratio of Wetted Width to
Thalweg Depth
SDWD_RAT - Standard Delation of Width-Depth
Ratio
AREASUM - Residual Pool Vertical Profile Area
(m2/reach)
RP100 - Mean Residual Depth (m2/100 m = cm)
RPGT75 - Number of Residual Pools with Depth
> 73 cm (number/reach)
RPXAREA - Mean Residual Pool Vertical Profile
Area (m2/pool)
RPMDEP - Maximum Residual Depth of Deepest
Pool in Reach (cm)
XINC_H - Mean Incision Height (m)
XUN - Mean Bank Lateral Undercut Distance (m)
XBF_H - Mean Bankfull Height (m)
XBF_W - Mean Bankfull Width (m)
XBKA — Mean bank angle (degrees)
XSLOPE - Mean Channel Gradient (%)
VSLOPE - Std. Deviation of Channel Gradient (%)
SINU - Channel Sinuosity
RMSE=arep
(in units of metric)
Mid-
Atlantic
6.4
1.7
0.93
0.58
0.79
0.32
6.8
6.5
4.6
1.6
0.60
0.69
14
0.38
—
0.33
1.7
8.1
0.80
0.40
0.10
Oregon
6.2
3.4
0.89
0.60
0.80
0.75
2.6
3.4
7.6
2.2
0.98
1.0
34
0.76
0.025
0.13
1.1
8.4
0.87
0.66
0.25
cv=arep/- •(%)
Mid-
Atlantic
22
13
18
38
39
32
32
52
19
17
47
30
19
26
—
71
24
18
42
43
8.3
Oregon
17
23
17
35
33
61
16
36
25
19
52
41
37
53
70
22
12
22
25
46
20
s/N =a2st(yr)/a2rep
Mid-
Atlantic
7.3
16
15
6.4
8.2
15
0.9
0.8
29
16
9.5
33
7.6
7.6
—
0.2
5.2
2.4
18
14
4.1
Oregon
6.9
6.0
14
5.1
8.1
2.9
6.5
2.9
17
9.0
8.2
6.8
1.5
0.8
2.1
3.5
24
2.0
24
4.7
1.1
' Variable names in bold are aggregate metric variables.
61
-------
TABLE 8. PRECISION OF PHYSICAL HABITAT METRICS FOR STREAM CHANNEL HABITAT
CLASSIFICATION IN THE MID-ATLANTIC REGION AND OREGON
(for the Mid-Atlantic Region, n=169 with 50 replicates; for Oregon, n=44 with 22 replicates)
Variable Name — Description
CHANNEL HABITAT CLASSIFICATION
METRICS3
PCT_FA - % Falls
PCT_CA - % Cascades
PCT_RA - % Rapids
PCT_RI - % Riffles
PCT_GL - % Glides
PCT_PP - % Plunge Pools
PCT_PD - % Impoundment Pools
PCT_PT - % Trench Pools
PCT_PL - % Lateral Scour Pools
PCT_PB - % Backwater Pools
PCT_POOL % Pools / Reach Length
PCT_SLOW % Pools+Glides / Reach
Length
PCT_FAST % Fast Water Habitat / Reach
Length
PCT_DRS % Dry or Submerged Flow/
Reach Length
RMSE=arep
(in units of metric)
Mid-
Atlantic
0.4
1.7
12
18
14
3.3
5.2
9.0
8.4
1.1
11
16
16
9.9
Oregon
1.0
3.7
3.0
14
15
5.0
12
12
9.5
1.2
16
12
12
1.2
cv=arep/- •(%)
Mid-
Atlantic
204
105
229
39
42
215
199
250
196
310
88
35
31
397
Oregon
382
47
80
59
79
219
140
100
211
411
48
23
25
586
S/N = C72st(yr)/a2rep
Mid-
Atlantic
19
10
~0
0.7
1.9
1.1
3.0
0.1
~0
0.4
1.2
1.7
1.6
0.7
Oregon
~0
21
1.6
1.6
2.1
0.1
1.4
2.5
0.2
~0
2.1
7.5
7.6
0.9
1 Variable names in bold are aggregate metric variables.
and intend to use it for trend assessment. Although flow-dependent data are necessary
and useful as covariates to aid understanding biological data, their flow-dependency
severely limits their use in assessing channel habitat changes in response to human
activities. To increase the precision of stream habitat classifications, Wood-Smith and
Buffington (1996) recommend determining or envisioning habitat types at a characteristic
stage (e.g., bankfull); Roper and Scarnecchia (1995) recommended simplifying or
aggregating complex habitat classifications and replacing them when possible with objective
measurements. To the same end, Ralph et al. (1994) recommended using direct, objective
measurements over visual and subjective judgements.
62
-------
To illustrate the effects that aggregating classifications, measuring rather than
estimating, and decreasing flow stage dependency have on metric precision, it is instructive
to compare metrics that quantify pool habitat in stream reaches (Table 9). Individual
metrics quantifying the percentage of specific pool types visually classified by field crews
were the least precise of these metrics. Though orep values increased when habitat classes
were combined into PCT_POOL (5 pool classes) and PCT_SLOW (5 pool classes + glides),
this effect does not indicate a loss of precision, as it results simply because these summed
metric values (and their variances) are greater than their subcomponent values. Note, for
example, that CV's of these combined classes are substantially lower than those for glide or
pool habitat metrics taken individually. The increase in precision of aggregating these pool
classifications is indicated by the slight to moderate increases in the S/N ratios of the
PCT_POOL and PCT_SLOW over the general S/N values of separate pool classes. Even if
field crews made perfect, consistent classifications of pool habitat along the thalweg, the
characteristics upon which these classifications depend are themselves dependent upon the
flow stage. An area that is clearly a pool during low flows may become a glide, riffle, or
even a rapid at higher flows. The lower S/N ratios of PCT_POOL and PCT_SLOW in the
Mid-Atlantic region (April-June), compared with the Oregon survey (July-Sept) may have
resulted because that survey was conducted during more rapidly changing springtime flow
conditions.
Mean thalweg depth (XDEPTH) is dependent upon stream basin size, flow stage
and the bed profile of the stream reach. While it appears quite precise, XDEPTH likely
derives much of its regional variability, and thus its high S/N ratio, from the broad size range
of streams in the surveys. Mean residual depth (RP100) and its surrogate SDDEPTH, are
flow-independent indices of bottom complexity, pool vertical profile area, and pool volume
(Kaufmann, 1987a; Robison and Kaufmann, 1994). Though it is not meaningful to compare
directly the orep values of RP100 and SDDEPTH with those for PCT_POOL and
PCT_SLOW, their S/N ratios were 3 to 8 times higher, indicating a much higher level of
precision for the quantitative, flow-independent metrics. Like XDEPTH, part of the regional
variance of stream residual pool area depends upon the stream basin size, which, for a
given runoff, influences the mean annual and flood stage discharges. Consequently S/N
ratios for RP100 and SDDEPTH are enhanced somewhat by regional differences in stream
size.
A flow-dependent analogue of PCT_POOL, percent residual depth (P_RESD), is
derived by dividing mean thalweg residual depth by mean thalweg total depth. In absolute
terms (orep), P_RESD was somewhat more precise than PCT_POOL (Table 9). Relative to
among-stream variation (i.e., considering S/N), P_RESD had about the same very low
precision as PCT_POOL in the Mid-Atlantic survey (S/N=1.1), but was moderately precise in
63
-------
TABLE 9. COMPARISON OF PRECISION IN METRICS DESCRIBING STREAM REACH POOL
HABITAT IN SURVEYS OF THE MID-ATLANTIC REGION AND OREGON
(forthe Mid-Atlantic Region, n=169 with 50 replicates; for Oregon, n = 44 with 22 replicates)
Variable Name - Description
POOL HABITAT METRICS3
PCT_P* - % of individual pool types - 5
metrics
PCT_POOL - % Pools / Reach Length
PCT_SLOW - (% Pools+Glides) / Reach
Length
XDEPTH - Mean Thalweg Depth (cm)
SDDEPTH - Standard Deviation of Thalweg
depth (cm)
RP100 - Mean Residual Depth (m2/100 m =
cm)
P RESD - % Residual Depth = 100 x
(RP100/XDEPTH)
RMSE=arep
(in units of metric)
Mid-
Atlantic
Oregon
1.1 to 12
11
16
6.4
1.7
1.6
9.8
16
12
6.2
3.4
2.2
5.4
cv=arep/- •(%)
Mid-
Atlantic
Oregon
100 to 411
88
35
22
13
17
29
48
23
17
23
19
18
S/N = C72st(yr)/a2rep
Mid-
Atlantic
Oregon
0 to 2.5
1.2
1.7
7.3
16
16
1.1
2.1
7.5
6.9
6.0
9.0
4.9
1 Variable names in bold are aggregate metric variables.
the Oregon survey (S/N=4.9), where flows were relatively stable within each year's field
season.
4.2.2 Substrate
Substrate metrics were reasonably precise, with 10 of 14 substrate percent
composition metrics having orep < 7% in one or both surveys (Table 10). However, the
following information suggests that the precision of substrate metrics could be improved by
increasing the number of "pebbles" in the systematic pebble count from 55 up to 100
particles, for example, even if size classifications were still determined by eye. Assuming
binomial sampling probabilities (Snedecor and Cochran, 1980), where the standard
deviation = {[p(1-p)]/n}%, pebble counts of 55 particles taken from substrates with
compositions in the range of 0 to 50% of a designated size class should have orep between
4 and 7%, assuming no error in size classifications. The observed orep values are typically
in this range, suggesting that the precision of these field methods is limited more by the
number of substrate particles in the sample than by uncertainties in the size classifications,
which are judged visually. Percent Sand + Fines, for example, had orep = 7.7% in the Mid-
Atlantic region survey. With a regional sample mean PCT_SAFN = 32%, we would predict
64
-------
TABLE 10. PRECISION OF PHYSICAL HABITAT METRICS FOR STREAM REACH SUBSTRATE
IN THE MID-ATLANTIC REGION AND OREGON
(for the Mid-Atlantic Region, n=169 with 50 replicates; for Oregon, n=44 with 22 replicates)
Variable Name — Description
SUBSTRATE METRICS3
PCT_RR - % Substrate - Rough Bedrock
PCT_RS - % Substrate - Smooth Bedrock
PCT_BL - % Substrate - Boulders
PCT_CB - % Substrate - Cobbles
PCT_GC - % Substrate - Large Gravel
PCT_GF - % Substrate - Fine Gravel
PCT_SA - % Substrate - Sand
PCT FN - % Substrate - Fines (Silt.Clay, and
Muck)
PCT_HP - % Substrate - Hardpan
PCT_SAFN - % Substrate - Sand + Fines
PCT_SFGF - % Substrate < 16mm diameter
PCT_BIGR - % Substrate > 16 mm diameter
PCT_BDRK - % Substrate - Bedrock
PCT_WD - % Substrate - Organic Debris
PCT_OT - % Substrate - Miscellaneous
XEMBED - % Substrate Embedded - mid-
channel + margin
XCEMBED - % Substrate Embedded - mid-
channel only
SUB_X - Mean Substrate Size Class (0 to 6)
SUB_V - Std. Deviation of Substrate Size
Class (0 to 6)
LSUB_DMM - Log10[Estimated Geometric
Mean Substrate Diameter (mm)]
LTEST - Log10[Mobile Substrate Diameter
(mm)]
LRBS_TST - Log10(Relative Bed Stability)
RMSE=arep
(in units of metric)
Mid-
Atlantic
9.2
5.6
5.1
4.9
7.4
5.3
9.8
11
0.2
7.7
7.5
6.2
9.1
3.6
3.6
15
17
0.20
0.20
0.26
0.27
0.35
Oregon
2.6
4.6
6.2
6.2
7.6
8.7
7.9
7.4
3.0
11
12
8.1
4.0
3.8
4.7
9.5
13
0.24
0.18
0.32
0.27
0.44
cv=arep/- •(%)
Mid-
Atlantic
255
210
58
26
45
47
68
60
310
24
17
12
144
146
480
27
36
6.4
18
n.a.
n.a.
n.a.
Oregon
100
176
42
42
45
87
118
32
142
36
30
16
76
102
217
18
27
7.8
17
n.a
n.a
n.a
S/N = C72st(yr)/a2rep
Mid-
Atlantic
~0
0.7
3.9
8.0
1.4
2.2
1.4
2.8
11
10
11
19
0.7
0.2
~0
1.9
1.5
22
2.4
20
2.6
9.0
Oregon
3.1
0.7
6.0
2.9
1.4
~0
0.1
15
12
7.1
5.0
16
3.9
0.4
0.6
7.7
4.1
23
3.9
24
7.4
6.8
1 Variable names in bold are aggregate metric variables.
65
-------
that arep would be approximately 6.3% in repeated samples of 55 particles, if there were no
error in the classification of particle size for each particle. The observed orep value of 7.7%
for PCT_SAFN means that if the observed mean value for a stream reach was 25% Sand +
Fines, the true reach mean + 1SD is estimated to be 25% + 7.7%. The Log10 of the
geometric mean substrate diameter (LSUB_DMM) had orep = 0.26 in the same survey,
giving a + 1 SD range of 5.5 to 18mm for a mean diameter of 10 mm. The S/N ratios of
substrate metrics were mostly high, with half of the 22 metrics having S/N > 5 and 7 metrics
with S/N > 10. Aggregated or averaged metrics (e.g., PCT_SAFN, PCT_BIGR,
LSUB_DMM, and LRBS_TST) tended to have much higher S/N ratios than percentages of
single classes of substrate (e.g., PCT_RR, PCT_SA).
4.2.3 Fish Cover and Large Woody Debris
All metrics for single fish cover types (e.g., XFC_UCB, XFC_RCK) are expressed as
areal cover proportions that range from 0 to 1. Because fish concealment features can be
layered upon each other, the sums of several types of cover types (e.g., XFC_NAT) can
theoretically have a total equal to the number of cover types in the sum. Even though all of
the fish cover metrics are visual estimations, the "plots" over which they are determined are
well-defined, and their classifications are tightly constrained. As a result, almost all were
estimated with relatively low standard errors (orep < 0.1) by field crews (Table 11). In fact,
half of the fish cover metrics were quite precise, with orep < 0.05 in one or both surveys.
Because the variances the sums of variables are additive, summed cover metrics such as
XFC_NAT, with orep= 0.18, had relatively higher orep than typical for their subcomponents
(which are single fish cover types). Signal:noise ratios of the metrics for single and
aggregate fish cover categories ranged from low to moderate (0 to 6.2), the majority having
S/N values between 2 and 4.
Fish cover presence metrics (e.g., PFC_LWD) differ from fish cover metrics (e.g.,
XFC_LWD) by being estimates of the percent of the reach length with any of the cover type
present, in contrast to being an estimate of the average areal cover of that type in the
reach. Like single cover-type metrics, metrics expressing the portion of the stream reach in
which one or more cover types were present have ranges constrained from 0 to 1. Because
many fish cover elements were typically present (though not necessarily abundant) at all
transects in most streams of the MAHA and Oregon surveys, regional variation in the values
of cover-presence metrics was not generally as great as that for metrics expressing the
amount of cover of fish concealment features. Consequently, S/N ratios for aggregated
cover-presence metrics were relatively low in the MAHA survey (S/N < 2) and barely
moderate in the Oregon survey (S/N 2.8 to 2.9), driven down not as much by imprecise
measurement, as by low variance among streams.
66
-------
TABLE 11. PRECISION OF PHYSICAL HABITAT METRICS FOR INSTREAM FISH COVER AND
LARGE WOODY DEBRIS (WITHIN BANKFULL CHANNEL) IN THE MID-ATLANTIC REGION
AND OREGON (for the Mid-Atlantic Region, n=169 with 50 replicates; for Oregon, n=44 with 22
replicates)
Variable Name — Description
FISH COVER AND LARGE WOODY
DEBRIS (LWD) TALLY METRICS3
RMSE=arep
(in units of metric)
Mid-
Atlantic
Oregon
cv=arep/- •(%)
Mid-
Atlantic
Oregon
S/N = C72st(yr)/a2rep
Mid-
Atlantic
Oregon
Fish Cover, Algae and Macrophyte Metrics:
XFC_ALG - Filamentous Algae -Areal
Cover Proportion
XFC_AQM - Aquatic Macrophytes - Areal
Cover Proportion
XFC_LWD - Large Woody Debris - Areal
Cover Proportion
XFC_BRS - Brush - Areal Cover Proportion
XFCJDHV - Overhanging Vegetation - Areal
Cover Proportion
XFCJJCB - Undercut Bank - Areal Cover
Proportion
XFC_RCK - Boulder, Rock Ledge - Areal
Cover Proportion
XFC_HUM - Artificial Structures - Areal
Cover Proportion
XFC_NAT - Sum of Natural Fish Cover
Types - Areal Cover Proportion
XFC_ALL - Sum of All Fish Cover Types -
Areal Cover Proportion
XFC_BIG - Sum of LWD, Undercut Bank,
and Rock Cover - Areal Cover Proportion
PFC_ANY - Portion of Reach with Any Type
of Cover
PFC_NAT - Portion of Reach with Natural
Cover
PFC_BIG - Portion of Reach with LWD,
Undercut Bank, or Rock Cover
0.067
0.031
0.040
0.037
0.11
0.040
0.095
0.13
0.18
0.22
0.18
0.15
0.12
0.22
0.089
0.068
0.036
0.065
0.069
0.040
0.14
0.006
0.18
0.18
0.14
0.07
0.07
0.12
224
102
142
59
87
64
55
679
39
46
64
16
13
26
197
117
53
63
36
56
64
203
28
28
40
7.1
7.1
14
0.8
4.7
0.2
1.2
0.6
2.1
3.5
~0
1.7
0.8
0.7
~0
1
0.4
0.9
2.8
3.9
1.0
5.1
6.2
2.1
3.6
2.8
2.8
2.9
~0
~0
2.9
' Variable names in bold are aggregate metric variables.
(continued)
67
-------
TABLE 11 (continued)
Variable Name — Description
FISH COVER AND LARGE WOODY
DEBRIS (LWD) TALLY METRICS3
RMSE=arep
(in units of metric)
Mid-
Atlantic
Oregon
cv=arep/« -(%)
Mid-
Atlantic
Oregon
S/N = C72st(yr)/a2rep
Mid-
Atlantic
Oregon
Selected Large Woody Debris Tally Metrics:
Log10(C1WM100) - LWD, all sizes
(Pieces/100 m)
Log10(C4WM100) - LWD, Large +Extra
Large sizes (Pieces/100 m)
Log10(V1WM100) - LWD Volume, all sizes
(m3/100m)
Log10(V4WM100) - LWD Volume, Large +
Extra Large sizes (m3/100 m)
0.50
0.93
0.53
1.17
0.40
0.63
0.34
0.82
n.a.
n.a.
n.a.
n.a.
n.a.
n.a.
n.a.
n.a.
3.9
~0
2.5
~0
7.0
2.4
12
2.5
1 Variable names in bold are aggregate metric variables.
In an absolute sense, LWD tallies produced reach estimates that were quite
imprecise, but when viewed relative to the magnitude of metric values or the range of
variation within the region, the precision of LWD tally metrics was moderate to high,
depending on the size class of LWD. The range of LWD abundance in streams within both
regions was quite extreme, with distributions and repeat variances that were decidedly
skewed. Repeat visit variance was proportional to the number or volume of woody debris in
the tally. For this reason, we conducted the ANOVA on log-transformed data. Values of
orep in a selection of LWD count and volume tally metrics ranged from 0.50 to 1.15 in the
Mid-Atlantic survey and 0.34 to 0.82 in Oregon (Table 11). Because these are log-
transformed values, they may be interpreted as proportional orep values of 3 to 15 times the
measured value for Mid-Atlantic streams and 2 to 6 times the measured value for Oregon
Streams. The total amount of LWD and the variation among streams in the Oregon survey
was substantially higher than in the Mid-Atlantic survey, leading to higher S/N values. In
Oregon, the median count for all sizes of LWD in the 35 sample stream reaches was 12
pieces (median volume =5.4 m3) per 100 m, compared with 5 pieces (volume =1.8 m3) per
100 m in the Mid-Atlantic reaches. The greater abundance of LWD in Oregon streams is
accentuated when one considers only the "Large" and "Extra Large" length and diameter
classes (see Table 6 for size class definitions). In Oregon, the medians for the sample
reaches were 0.5 pieces per 100 m and 1.5 m3 per 100 m, compared with medians of zero
in the Mid-Atlantic survey. As a result primarily of the greater quantity and variation in
Oregon streams, metric precision in terms of S/N ratios was high for LWD (all sizes
combined) in Oregon streams (7.0 to 12), compared with only moderate precision in the
68
-------
Mid-Atlantic survey (S/N = 2.5 to 3.9). For large and extra-large diameter and length
classes, S/N precision was barely moderate in Oregon (S/N = 2.4 to 2.5), but very low in
the Mid-Atlantic (S/N ~0), because wood of that size is relatively rare in that region.
4.2.4 Riparian Vegetation
Riparian canopy cover measured with a canopy densiometer was determined with
virtually the same precision by field crews in the Mid-Atlantic region and Oregon surveys
(orep = 5.7 and 5.8 % for XCDENMID). Corresponding values of S/N were 19 and 15 for
XCDENMID in the two surveys (Table 12). The 2 sets of crews had similar backgrounds of
education and experience, used the same field manual, and were trained identically.
However, the Oregon survey and its repeat site visits were done in midsummer (July 1 to
Sept. 15), whereas the Mid-Atlantic region streams were sampled between April and June
(after snowmelt but before full leafout), to optimize detecting acidic deposition effects on
chemistry and biota. In the Mid-Atlantic region, crews were instructed to "imagine" tree
cover under leafout conditions based on the extent of bare branches and newly budding
leaves they frequently encountered in the riparian canopy. Surprisingly, actual changes in
canopy cover and ambiguity in reading densiometer values during the spring sampling
period did not appear to erode the precision of measurements in that region. Canopy
density measurements taken at the stream banks (XCDENBK) were less precise than mid-
channel measurements in the Mid-Atlantic region; the reverse was true in Oregon. Lower
precision in Mid-Atlantic region streams may have resulted because that survey was
conducted during more rapidly changing springtime flow conditions in which stream width
was declining, changing the distance between the wetted bank and the edge of riparian
vegetation. However, the smaller number of bank densiometer measurements (22),
compared with mid-channel measurements (44) is theoretically sufficient to account for the
lower precision in XCDENBK. Based on bionomial sampling theory, this difference in the
number of individual densiometer observations is sufficient to result in a orep value 1.41
times higher using 22, rather than 44 measurements (and assuming equal mean canopy
densities for the bank and mid-channel).
Besides the canopy densiometer metrics, the other riparian vegetation metrics
showed considerable range in precision (Tables 12 and 13). With the exception of XC in
both regions, and XCL in Oregon, and the ground cover metrics XGH and XGB in the Mid-
Atlantic region, virtually all single vegetation type cover magnitude (not presence) metrics
were relatively imprecise, with orep mostly between 10% and 20% and S/N < 2.5. Cover
magnitude estimates summing two or more layers of vegetation were similarly imprecise or
moderately precise. In contrast, virtually all single and combined cover presence metrics
(i.e., those with names beginning with XP... or P...) were moderately precise, with orep < 8%
69
-------
TABLE 12. PRECISION OF PHYSICAL HABITAT METRICS FOR CANOPY DENSITY, COVER,
AND PRESENCE IN MULTIPLE LAYERS OF RIPARIAN VEGETATION ALONG STREAMS OF
THE MID-ATLANTIC REGION AND OREGON
(for the Mid-Atlantic Region, n=169 with 50 replicates; for Oregon, n=44 with 22 replicates)
Variable Name — Description
RIPARIAN VEGETATION METRICS -
MULTIPLE LAYER3
RMSE=arep
(units of metric)
Mid-
Atlantic
Oregon
cv=arep/- •(%)
Mid-
Atlantic
Oregon
S/N = C72st(yr)/a2rep
Mid-
Atlantic
Oregon
Canopy Densiometer Metrics:
XCDENMID - Canopy Cover Midstream -
Densiometer (%)
VCDENMID - Std. Deviation of Canopy Cover
Midstream - Densiometer (%)
XCDENBK - Canopy Cover at Bank -
Densiometer (%)
VCDENBK - Std. Deviation of Canopy Cover
at Bank - Densiometer (%)
5.7
3.7
8.0
5.6
5.8
3.9
3.9
5.7
7.5
19
10
31
8.1
21
4.4
42
19
9.3
7.3
5.3
15
4.3
17
2.2
Visual Cover Estimation Metrics:
XCM - Sum of Canopy + Mid-Layer Cover
(Proportion of Riparian)
XPCM — Both Canopy and Mid-Layer Present
(Proportion of Riparian)
XCMW - Sum of Woody Canopy + Mid-Layer
(Proportion of Riparian)
XCMG - Sum of Canopy + Mid-Layer +
Ground Cover (Proportion of Riparian)
XPCMG - 3-Layers of Vegetation Present
(Proportion of Riparian)
XCMGW - Sum of Woody Vegetation Cover
in 3 Layers (Proportion of Riparian)
0.33
0.09
0.22
0.41
0.10
0.25
0.27
0.08
0.22
0.40
0.08
0.36
40
11
28
29
13
28
34
9.8
33
28
9.8
40
0.6
7.1
2.3
0.3
5.8
2.3
0.8
7.9
1.4
0.1
8.0
0.7
1 Variable names in bold are aggregate metric variables.
and S/N > 5. The cover presence metrics XPCAN and (in the Oregon surveys) PMID_C had
high precision (orep <5 and S/N > 10), rivaling that of canopy densiometer metrics.
It is instructive to examine the precision that can be gained by applying quantitative
methods or by changing the interpretation of visual cover observations from estimates of
cover magnitude to estimates of cover presence. Table 14 compares the precision of
densiometer measurements (XCDENMID, XCDENBK) with that of purely visual estimates of
canopy cover (XC, XPCAN) that have similar or identical conceptual meaning. It is clear
70
-------
TABLE 13. PRECISION OF PHYSICAL HABITAT METRICS FOR COVER AND PRESENCE
WITHIN SINGLE LAYERS OF RIPARIAN VEGETATION IN STREAMS OF THE MID-ATLANTIC
REGION AND OREGON (for the Mid-Atlantic Region, n=169 with 50 replicates; for Oregon, n=44
with 122 replicates)
Variable Name - Description
RIPARIAN VEGETATION METRICS -
SINGLE LAYER3
XCL - Large Diameter Tree Canopy Cover
(Proportion of Riparian)
XCS - Small Diameter Tree Canopy Cover
(Proportion of Riparian)
XC - Tree Canopy Cover (Proportion of
Riparian)
XPCAN -Tree Canopy Presence (Proportion
of Riparian)
XMW - Mid-Layer Woody Vegetation Cover
(Proportion of Riparian)
XMH - Mid-Layer Herbaceous Vegetation
Cover (Proportion of Riparian)
XM - Mid-Layer Vegetation Cover (Proportion
of Riparian)
XPMID - Mid-Layer Vegetation Presence
(Proportion of Riparian)
XGW - Ground Layer Woody Vegetation
Cover (Proportion of Riparian)
XGH - Ground Layer Herbaceous Vegetation
Cover (Proportion of Riparian)
XGB - Ground Layer Barren or Duff Cover
(Proportion of Riparian)
XG - Ground Layer Vegetation Cover
(Proportion of Riparian)
PCAN_C - Conifer Riparian Canopy
(Proportion of Riparian)
PCAN_D - Broadleaf Deciduous Riparian
Canopy (Proportion of Riparian)
PCAN_M - Mixed Conifer-Broadleaf Canopy
(Proportion of Riparian)
PMID_C - Conifer Riparian Mid-Layer
(Proportion of Riparian)
RMSE=arep
(units of metric)
Mid-
Atlantic
0.097
0.11
0.14
0.07
0.12
0.19
0.22
0.09
0.07
0.11
0.085
0.14
0.03
0.14
0.10
0.04
Oregon
0.057
0.12
0.12
0.08
0.12
0.13
0.19
0.03
0.17
0.16
0.07
0.22
0.11
0.13
0.16
0.02
cv=arep/- •(%)
Mid-
Atlantic
51
41
32
8.4
37
272
57
10
55
27
29
26
169
22
49
136
Oregon
38
55
33
8.7
41
100
44
3.5
77
40
47
36
58
31
65
55
S/N = C72st(yr)/a2rep
Mid-
Atlantic
0.9
1.5
2.3
10
1.6
~0
0.1
3.6
1.4
3.6
6.5
1.9
4.3
4.9
7.2
6.5
Oregon
4.6
1.4
2.4
10
0.9
0.9
0.6
2.1
0.1
1.1
2.0
~0
8.5
7.4
2.9
37
' Variable names in bold are aggregate metric variables.
(continued)
71
-------
TABLE 13 (Continued)
Variable Name — Description
RIPARIAN VEGETATION METRICS -
SINGLE LAYER3
PMID_D - Broadleaf Deciduous Riparian Mid-
Layer (Proportion of Riparian)
PMID_M - Mixed Conifer-Broadleaf Canopy
(Proportion of Riparian)
RMSE=arep
(units of metric)
Mid-
Atlantic
0.15
0.13
Oregon
0.33
0.32
cv=arep/« -(%)
Mid-
Atlantic
23
61
Oregon
58
87
s/N = a2st(yr)/a2rep
Mid-
Atlantic
3.8
4.4
Oregon
0.7
0.6
1 Variable names in bold are aggregate metric variables.
that the least precise of the four canopy cover metrics is XC, which depends upon visual
judgement to estimate both the canopy cover in a set of riparian plots and the dimensions of
those 22 plots. Precision is substantially increased by reinterpreting the same visual data to
calculate percentage cover presence along the stream reach (XPCAN), rather than cover
magnitude (orep decreases from 12-14% to 7-8% and S/N increases from 2.3-2.4 up to 10).
Not surprisingly, field crews using purely visual judgement were able to estimate canopy
presence more reliably than canopy cover. Analogous patterns can be seen by comparing
the precision of XM with XPMID, XCM with XPCM, and XCMG with XPCMG (Table 12).
Depending on the region, however, these gains in precision were often accompanied by a
reduction in the range of variability among sites. Choosing metrics of vegetation presence
over vegetation cover may also result in a loss of ecological information and decreased
sensitivity to stress in some regions. As mentioned in a previous paragraph, XCDENBK
and XCDENMID are expressions of mean canopy cover based on, respectively, 22 bank
and 44 mid-channel canopy densiometer observations along the sample reach. Both the
quantitative densiometer approaches, XCDENBK and particularly XCDENMID, are more
precise than the purely visual estimation procedures (Table 14), but have the disadvantage
of lacking ecological specificity. Unlike XC, for example, the densiometer measurements
make no distinction between shrub and tree cover. For these reasons it is advantageous to
retain both semi-quantitative visual riparian estimates and quantitative canopy densiometer
measurements in characterizing riparian vegetation cover and structure.
4.2.5 Riparian Human Activities and Disturbances
Riparian human disturbance metrics ranged from low to high precision, but most
were in the low to moderate range; slightly more than a third had S/N ratios >4, but half had
S/N <2 in one or the other regional survey (Table 15). The precision of individual metrics
varied greatly between the two surveys. In the Mid-Atlantic region, the most precise
72
-------
TABLE 14. COMPARISON OF PRECISION OF FOUR STREAMSIDE RIPARIAN CANOPY
COVER METRICS IN THE MID-ATLANTIC REGION AND OREGON
(for the Mid-Atlantic Region, n=169 with 50 replicates; for Oregon, n=44 with 22 replicates)
Variable Name - Description
RIPARIAN CANOPY COVER METRICS
Visual % Tree Canopy Cover (XC * 100)a
Visual % Tree Canopy Presence (XPCAN
x 100)a
XCDENBK - Canopy %Cover at Bank -
Densiometer
XCDENMID - Canopy %Cover Midstream
- Densiometer
RMSE=arep
(% Cover)
Mid-
Atlantic
14
7.1
8.0
5.7
Oregon
12
8.0
3.9
5.8
cv=arep/- •(%)
Mid-
Atlantic
32
8.4
10
7.5
Oregon
33
8.7
4.4
8.1
S/N = C72st(yr)/a2rep
Mid-
Atlantic
2.3
10
7.3
19
Oregon
2.4
10
17
15
' Values expressed as % for comparison purposes.
individual disturbance metrics (arep < 0.05 or S/N >7.0) were those assessing revetments,
influent/effluent pipes, pastures, row crops, and logging activities. Precision was generally
lower in the Oregon survey, and those metrics assessing lawns/parks, buildings, pastures,
influent/effluent pipes, and pavement were most precise (orep < 0.06 or S/N > 4.9).
Seemingly a rather straightforward observation, the road disturbance metric W1H_ROAD
was the only metric determined with low precision in both surveys (orep > 0.15 and S/N <
1.4). The poor performance of this metric may have resulted from inconsistent inclusion of
paths, railroads, and pavement in the tally, and the inconsistent tallying of roads that are
not directly observed, but are heard (traffic) or known to be beyond the riparian plot.
Aggregating (summing) human disturbance metrics into variables such as W1_HALL
resulted in generally higher orep values than the subcomponents, which is not surprising,
because the repeat visit variances (o2rep) of the summed subcomponents are additive. For
the same reason, the aggregated human disturbance metrics generally had S/N ratios
approximately midway within the range exhibited by their subcomponents. The summed
agricultural disturbance metric W1_HAG was determined with greater precision in both
surveys (S/N 6.9 and 8.8) than the non-agricultural disturbance sum (S/N 3.4 and 0.9).
4.2.6 EPA's Rapid Bioassessment Protocol (RBP) Habitat Quality Scores
In tandem with the more intensive EMAP habitat characterization procedures that
are the primary focus of this report, the EPA and OSU field crews in the Mid-Atlantic and
Oregon surveys employed EPA's Rapid Bioassessment Protocol (RBP) habitat quality
assessment field procedures as described by Barbour and Stribling (1991) and Klemm and
73
-------
TABLE 15. PRECISION OF PHYSICAL HABITAT METRICS FOR STREAMSIDE RIPARIAN
HUMAN ACTIVITIES AND DISTURBANCES IN THE MID-ATLANTIC REGION AND OREGON
(for the Mid-Atlantic Region, n=169 with 50 replicates; for Oregon, n=44 with 22 replicates)
Variable Name — Description
RIPARIAN HUMAN DISTURBANCE
METRICS3
W1H_BLDG - Riparian Human Disturbance -
Buildings (Proximity-weighted index)
W1H_CROP - Riparian Human Disturbance -
Row Crop Agriculture (Proximity-weighted
index)
W1H_LDFL - Riparian Human Disturbance -
Trash and Landfill (Proximity-weighted index)
W1H_LOG — Riparian Human Disturbance -
Logging (Proximity-weighted index)
W1H_MINE - Riparian Human Disturbance -
Mining (Proximity-weighted index)
W1H_P ARK- Riparian Human Disturbance -
Parks and Lawns (Proximity-weighted index)
W1H_PIPE - Riparian Human Disturbance -
Pipes, Influent or Effluent (Proximity-weighted
index)
W1H_PSTR - Riparian Human Disturbance -
Pasture, Grass or Hay Field (Proximity-
weighted index)
W1 H_PVMT - Riparian Human Disturbance -
Pavement) (Proximity-weighted index)
W1H_ROAD - Riparian Human Disturbance -
Roads (Proximity-weighted index)
W1H_WALL- Riparian Human Disturbance -
Channel Revetment (Proximity-weighted
index)
W1JHALL - Riparian Human Disturbance
Index (Proximity-weighted)
W1_HAG — Riparian Human Disturbance
Index - Agricultural Types (Proximity-
weighted)
W1JHNOAG - Riparian Human Disturbance
Index - Non-agricultural Types (Proximity-
weighted)
RMSE=arep
(in units of metric)
Mid-
Atlantic
0.13
0.05
0.13
0.05
0.07
0.13
0.03
0.15
0.16
0.15
0.02
0.51
0.17
0.45
Oregon
0.09
0.09
0.32
0.36
—
0.03
0.04
0.14
0.06
0.16
0.17
0.78
0.12
0.76
cv=arep/- •(%)
Mid-
Atlantic
82
118
178
98
364
100
162
50
125
56
20
41
49
50
Oregon
74
168
245
167
—
70
170
70
73
63
331
66
47
81
S/N = C72st(yr)/a2rep
Mid-
Atlantic
1.3
7.1
1.9
16
1.2
1.8
3.4
7.9
0.6
1.4
185
3.3
6.9
3.4
Oregon
5.3
2.7
~0
0.3
—
18
0.1
4.9
11
1.2
~0
0.9
8.8
0.9
' Variable names in bold are aggregate metric variables.
74
-------
Lazorchak (1994). RBP precision was assessed using the same ANOVA procedures to
compare among-stream variance to within-year repeat sampling variance (see Section 4.1).
For RBP habitat procedures, precision estimates in the Mid-Atlantic region were based on
sampling 459 streams with 36 within-season repeat visits over two field seasons (1993 and
1994). In the Oregon survey, RBP precision was assessed from a sample of 34 streams,
with 16 within-season repeat visits over the period from 1993-1996.
The RBP habitat quality assessment consists of 12 subcomponent habitat
assessment metrics that are summed to yield the RBP Habitat Quality Score (Barbour and
Stribling 1991). Separately for twelve aspects of channel and riparian habitat (Table 16),
field surveyors use their observations and judgement to rate habitat condition from poor
(score=0) to excellent (score=20). These subcomponent scores sum to a potential range
from 0 to 240 for the RBP habitat quality total score. Subcomponent metric orep values
ranged from 2.0 to 4.3 points and CV's ranged from 12% to 32%, somewhat higher (less
precise) than those reported by Barbour and Stribling (1994), who measured CVs of 5% to
20% for the RBP sub-metrics in replication by 17 investigators rating habitat quality during
one day on a single, good quality mountain stream in New Mexico. The range of S/N ratios
for the RBP sub-metrics in our two surveys ranged from 0 to 7.4. The "Riffle Frequency"
metric in the Oregon survey had the highest S/N ratio (7.4); all others had S/N < 4.2. The
"Channel Alteration", "Sediment Deposition", "Grazing and other Disruptive Pressure", and
"Riparian Zone Width" metrics had moderate precision, with S/N between 2 and 4.2 in at
least one of the surveys. Field determinations of the remaining seven RBP habitat
subcomponent metrics were rather imprecise relative to among-stream variation, with S/N
between 0 and 1.8 in both surveys.
The RBP Habitat Quality Assessment total score had orep values of 23 and 20 points,
respectively, for the Mid-Atlantic and Oregon surveys. Our CVs of 14% and 12% were
relatively low and similar to those reported for RBP habitat assessment in other studies
(e.g., Barbour and Stribling, 1994; Hannaford et al., 1997). On the basis of these orep and
CV values, the repeat visit variance of the RBP habitat score is relatively small compared
with its potential range of variation and its mean value. If the total RBP score faithfully
represents habitat quality over its potential range of 0 to 240 points, these orep values would
indicate a good potential for discerning among-stream variation and changes in habitat
quality over time. However, at least in the two regions considered, and in agreement with
findings by Hannaford and Resh (1995) in several California streams, we did not observe
great variation in the total score among streams relative to variation between visits to the
same site. The RBP total score usually revealed sites with very severe habitat degradation
and those with very high quality, in agreement with Hannaford et al. (1997), who observed
75
-------
TABLE 16. PRECISION OF RAPID BIOASSESSMENT PROTOCOL (RBP) HABITAT QUALITY
METRICS (BARBOUR AND STRIBLING, 1991) WITHIN SAMPLE SEASON IN THE MID-
ATLANTIC REGION AND OREGON (for the Mid-Atlantic Region, n=459 streams with 36 repeats
over 2 years;
for Oregon, n=34 streams with 16 repeats over 3 years)
RAPID BIOASSESSMENT PROTOCOL
HABITAT METRICS3
Instream Cover for Fish - Score (0-20)
Epifaunal Substrate - Score (0-20)
Embeddedness (or Pool Substrate*) - Score
(0-20)
Velocity/Depth Regime (or Pool Variability^)
- Score (0-20)
Channel Alteration - Score (0-20)
Sediment Deposition — Score (0-20)
Riffle Frequency (or Channel Sinuosity") -
Score (0-20)
Channel Flow Status - Score (0-20)
Bank Condition - Score (0-20)
Bank Vegetative Protection - Score (0-20)
Grazing or Other Disruptive Pressure —
Score (0-20)
Riparian Vegetation Zone Width — Score (0-
20)
RBP Habitat Quality Total Score - (0 to
240)
RMSE=arep
(in units of metric)
Mid-
Atlantic
3.7
4.3
3.6
3.2
2.0
2.5
2.8
3.2
2.5
3.7
2.3
2.9
23
Oregon
3.4
3.6
2.8
3.8
2.9
2.9
2.0
4.1
2.3
3.8
2.9
3.0
20
cv=arep/« -(%)
Mid-
Atlantic
28
30
28
25
12
19
18
22
18
25
15
24
14
Oregon
24
27
20
32
19
23
15
31
17
25
17
27
12
S/N = C72st(yr)/a2rep
Mid-
Atlantic
0.7
~0
0.6
0.9
2.0
2.5
1.1
0.8
1.8
0.4
3.3
4.2
1.6
Oregon
1.0
0.6
1.1
1.4
1.1
1.9
7.4
0.1
1.3
0.2
1.0
3.1
3.3
a Variable names in bold are aggregate metric variables.
b Mid-Atlantic Region survey data did not include repeat visits to measure these low gradient stream habitat assessment
features; Oregon survey included repeat visits to both low gradient and high gradient streams.
greater consistency among observers in RBP habitat scores for obviously pristine or
obviously degraded streams than for moderately impaired streams.
The great majority of our Mid-Atlantic region sites had mid-value total scores within a
range that did not greatly exceed the range observed between measurements made on
different visits to single streams. Barbour and Stribling (1994) observed a 19 point range in
total RBP habitat scores for determinations of habitat quality at a single good quality site in
New Mexico. However, in a comparison of 10 sites, they reported that, in general, within-
76
-------
site variability was smaller than variability among sites. In our surveys, S/N ratios for the
total score were quite low, indicating either a true lack of variation in habitat quality among
streams, or lack of RBP habitat metric responsiveness to actual habitat quality variation.
This was particularly true in the Mid-Atlantic region, where measurement variance was
almost as great as the variance among streams in the region (S/N = 1.6). In Oregon, the
S/N ratio was somewhat higher (3.3), perhaps consistent with the smaller number of field
crews involved and the more localized regional extent compared with the Mid-Atlantic region
survey (more similar to conditions evaluated by Barbour and Stribling [1994]). In the smaller
region, crews could be trained under conditions more closely resembling the range of
conditions they later sampled, a factor reported by Hannaford et al. (1997) to be important
in RBP habitat training. However, the S/N value of the RBP total score in Oregon was also
heavily influenced by the one high S/N value of one of its subcomponent metrics, "Riffle
Frequency" ("Channel Sinuosity" in low gradient streams), which had S/N=7.4. In addition,
combining the lowland Willamette Valley with the upland Cascade Mountain ecoregion in
the Oregon survey enhanced regional variability of the RBP score in the Oregon survey.
The "Riffle Frequency", "Sediment Deposition, and several other RBP metrics tend to score
high gradient, coarse-bedded, forested mountain streams higher than low gradient streams.
When the ANOVA for Oregon streams is calculated separately to factor out ecoregional
differences, sample sizes are rather small to make firm conclusions, but S/N values of 0.6
and 1.4 (for the Cascade Mts. and Willamette Valley) suggest the same result as found for
the Mid-Atlantic region: either the streams lack habitat quality variation within the two
Oregon ecoregions or the RBP habitat score is unable to discern actual habitat quality
differences above the "noise" of measurement uncertainty.
After applying four separate qualitative habitat quality assessment approaches in a
survey of northern prairie streams in the U.S., Stauffer and Goldstein (1997) recommended
that the variability of index scores could be improved by basing them on counts (or
measurements) rather than judgements in the field. Hannaford and Resh (1995) attributed
inconsistent RBP habitat assessments to a combination of "viewer" error and differences in
the precise location where multiple habitat observations were made in their study. They
reasoned that increased training or substitution of measurements for judgement might
reduce "viewer" error. To reduce variability due to small-scale differences in site location,
they recommended spreading a series of observations over a longer stream segment than
required by Plafkin et al. (1989) or Barbour and Stribling (1991).
4.3 SUMMARY OF HABITAT METRIC PRECISION
Our results from testing physical habitat field sampling procedures and analytical
approaches exhibit consistent patterns of metric precision in each of the two regional
77
-------
surveys. The most precise general classes of metrics were those describing channel
morphology and substrate. Most of the metrics with S/N > 6 were in these two categories
(Figure 5). Riparian vegetation and human disturbance metrics had intermediate precision;
although these two categories had many imprecise metrics (S/N < 2) most of their metrics
had S/N values > 2 and many > 6. Fish cover metrics had intermediate S/N ratios (most
between 2.8 and 6.0) in the Oregon survey, but lower values (most from 0.8 to 2.1) in the
Mid-Atlantic region survey. Since the absolute precision of these Mid-Atlantic metrics was
actually better than that in the Oregon survey (lower orep values in Table 11), the results
indicate low precision only with respect to a low variability of fish cover among streams in
the Mid-Atlantic region. Two classes of metrics were generally imprecise: those involving
visual channel habitat unit classification and RBP habitat quality assessments. Most
metrics in these classes had S/N < 2.0.
Most of the metric classes discussed in the previous paragraph contain a mix of
metrics determined using various approaches that involve greater or lesser degrees of
quantification, judgement, and sensitivity to variation in stream flow stage. We regrouped
the various metrics according to their measurement approach to more clearly compare the
precision of various approaches (Figure 6). The categories include:
QMS: Quantitative measurement of stable and easily defined features (e.g.,
slope, residual depth, canopy density),
F_QM: Quantitative measurement of flow-dependent or difficult-to-define
features (e.g. thalweg depth, incision height),
SQM: Semi-quantitative measurements or determinations of presence-
absence (e.g. substrate size, canopy presence, LWD tally metrics),
VSC: Visual estimates of areal cover (e.g., visual estimates of areal cover of
riparian vegetation and fish concealment features),
F_HB: Visual determinations of flow-sensitive channel unit class (e.g. %Riffle,
%Pool), and
JUD: Visual assessments requiring field judgements of habitat quality (e.g.
RBP Habitat Quality Score).
In both the Mid-Atlantic and Oregon surveys, measurement group QMS was clearly
the most precise group of metrics; most of these metrics had S/N between 6 and 18 (Figure
6). Flow-sensitivity and ambiguity in measured features certainly degraded the precision of
group F_QM relative to the first group, but group F_QM metrics were still generally within
the moderate precision range (S/N 2.0 to 6.0). Semi-quantitative measurements and
presence-absence determinations (group SQM) were intermediate in precision between the
78
-------
HI
w
o
o
w
2O
10
o
—f— i —-r—-——-j————T——
CHMR HBCL SUB FCV RVG
CLASS OF MEASUREMENTS
HDIS
—r—
RBPH
A) Mid-Atlantic Region
20
O
1-
Hl
CO
i 10
li
z
0
CO
0
i
^— " ^_ 1 ' 1
: | 1^^ |"|
r^i ' I , "T11111' """"H Sf1111
_, ( p. , --,,_ _f_ _ ( 1 — ..
CHMR HBCL SUB FCV RVG HDIS RBPH
CLASS OF MEASUREMENTS
B) Oregon
Figure 5. Frequency distribution of signal to noise ratios for physical habitat variables,
grouped according to the types of habitat attributes assessed. Measurement class codes:
CHMR - channel morphology, HBCL - channel habitat unit classification, SUB - substrate, FCV -
fish cover, HDIS - human disturbance, and RBPH - Rapid Bioassessment Protocol habitat scores.
Heavy bar = median; box = 25th to 75th percentiles, whiskers = 10th and 90th percentiles.
79
-------
(D
V)
3»
2CM
ia
O4
n
QMS F_QM SQM VSC F_HB
CLASS OF MEASUREMENTS
JUD
A) Mid-Atlantic Region
a:
LU
O
z
li
<
3O-
2O
1O-
• r
QMS F_QM SQM VSC F_HB
CLASS OF MEASUREMENTS
JUD
B) Oregon
Figure 6. Frequency distribution of signal to noise ratios for physical habitat metrics,
grouped according to the measurement approach. Measurement classes: QMS - Quantitative
measurement of stable and easily defined features, F_QM - Quantitative measurement of flow-
dependent or difficult-to-define features, SQM - Semi-quantitative measurements or determinations
of presence-absence, VSC - Visual estimates of areal cover, F_HB - Visual determinations of flow-
sensitive channel unit class, and JUD - Visual assessments requiring field judgements of habitat
quality. Heavy bar = median; box = 25th to 75th percentiles, whiskers = 10th and 90th percentiles.
80
-------
two quantitative measurement groups. Precision relative to regional variation (S/N)
generally declined from moderate to low in the three remaining measurement groups:
visual estimations of areal cover (VSC), flow-sensitive habitat classifications (F_HB), and
field judgements of habitat quality (JUD).
Our findings are in general agreement with those of Wang et al. (1996), who
evaluated among-observer precision based on replicate observations by six observers in
three Wisconsin streams. They found that". . . stream width and water depth were
estimated most precisely; these were followed by substrate composition, cover for fish, and
bank susceptibility to erosion. Estimates of bank vegetation or land use and gravel
embeddedness were the least precise. . .". Platts (1981) and Hogle et al. (1993) found that
precision and accuracy of habitat measurements were related to the clarity and detail of
habitat definitions and measurement procedures. Platts (1981), Hogle et al. (1993), Ralph
et al. (1994), Poole et al. (1997), and Wang et al. (1996) generally advocated
measurements over visual observations or judgements for most habitat attributes.
However, Wang et al. (1996) found, as we did, that visual determinations of some
attributes, such as substrate cover, can be reasonably precise if these observations are
repeated over a length of reach and the spatial boundaries of these observations are tightly
controlled. Ralph et al. (1991, 1994) and Poole et al. (1997) caution that habitat unit
classifications, even when carried out under the same flow conditions, are not sufficiently
repeatable to be used in trend monitoring applications, due to the subjectivity of these types
of observations. Poole et al. (1997) argue that, while they can be used in regional trend and
status estimates, their lack of precision unnecessarily complicates data analysis and
interpretation.
An unexpected result from our analysis was that a substantial number of repeat field
samples (20 to 70) were required to reliably determine values of orep and S/N, though
statisticians advocate sample sizes of 40 to 50 for variance estimates (Scott Urquhart,
personal communication). Values of these measures of precision determined on the basis
of < 20 pairs of within-season field revisits were surprisingly variable (Table 17). As a
typical example, estimates of orep for the metric SDDEPTH ranged from 1.0 cm to 1.7 cm
with 7 to 15 pairs of within-season repeat visits each year in the Mid-Atlantic region, and
from 1.0 cm to 5.4 cm with 6 to 8 pairs in Oregon. Similarly, with identical methods and
training, S/N ratios ranged from 11 to 47 in the various years of surveying Mid-Atlantic
streams, and from 4 to 73 in Oregon. Both orep and S/N can be substantially overestimated
or underestimated when based on 6 to 15 pairs of repeat samples. We are led to the
conclusion that an adequate evaluation of precision requires 20 to 50 pairs of within-season
repeat samples, and that these pairs should be spread across several years. In addition,
81
-------
TABLE 17. CONTRASTING PRECISION OF SDDEPTH IN SEPARATE AND COMBINED
SURVEYS OF STREAMS IN THE MID-ATLANTIC REGION AND THE WILLAMETTE BASIN IN
OREGON
Source of Variation
Degrees of freedom
Repeats
Model
Total
RMSE
Kepi
(cm)
C /M
O/IM —
°" st(yr/°" rep
Mid-Atlantic Region
1993 Stream(year)
1994 Stream(year)
1995 Stream(year)
1996 Stream(year)
1993-96 Stream(year)
7
13
15
15
49
80
86
15
14
199
87
99
30
29
248
1.0
2.2
1.7
1.5
1.7
47
11
20
18
16
Willamette Basin, Oregon
1993 Stream(year)
1994 Stream(year)
1995 Stream(year)
1996 Stream(year)
1993-96 Stream(year)
6
8
8
0
22
43
7
7
33
33
49
15
15
33
115
1.0
5.4
1.4
—
3.4
73
4
53
—
6
some of these stream visits should be to the same streams over several years, if the
interannual component of variance is to be determined.
4.4 IMPLICATIONS OF HABITAT MEASUREMENT PRECISION
4.4.1 Effects on Estimates of Regional Population Distributions
The population variance observed in a regional survey in any given year (o2obs) is the
sum of the "true", or "signal" variance among streams during that year (o2st(yr)), plus within-
season replicate "noise" variance (o2rep) that results from short-term temporal variation or
measurement variation occurring within the sampling "window". For that year, o2obs = o:
The effect of o2 is to distort the observed frequency distribution of the habitat
st(yr)
+ O'
rep-
metric across streams in the region. For example, this effect can be envisioned graphically
as an increase in the outward spread of a normal bell-shaped curve from its mean (Overton
,1989: Paulsen et al., 1991: Larsen and Urquhart, 1993). As long as the habitat metric
noise variance is normal and homogeneous (and the measurements themselves are
82
-------
unbiased), the mean and median habitat metric values of the regional stream population are
estimated without bias, no matter how imprecise the measurements. Estimates of other
stream population percentiles (e.g., the 10th, 25th, and 75th percentiles) have insignificant
absolute bias when the signal-to-noise ratio is >10 (Overton, 1989). However, percentiles
other than the median acquire progressively greater biases with declines in the ratio of
signal to noise (Overton, 1989; Paulsen et al., 1991). Paulsen et al. (1991) show, for
example, that when noise variance of a metric is equal to signal variance (S/N = o2st(yr)/o2rep
= 1.0), the apparent value of SDst(yr) for that metric is inflated by a factor of 1.29. In a
population with an underlying normal distribution, a survey using that metric would
overestimate the percentage of streams with habitat metric values greater than the mean +
1(SDst(yr)) at 24%—considerably greater than the true value of 16% (these percentages are
taken from normal distribution tables). If this metric value were the threshold of
"acceptability," then the survey would overestimate the number of streams with acceptable
habitat quality by a factor of 1.5.
In general, if metric measurement variance is greater than about 50% of the
variance among streams (i.e., S/N = o2st(yr)/o2rep < 2.0), and we believe that the regional
sample of streams spans an ecologically meaningful range of the metric, then the metric
may be too imprecise to answer with confidence certain kinds of questions posed
concerning the proportions of the stream population within stated ranges of the metric
(Paulsen et al., 1991). In such cases, we must either refrain from making high-resolution
statements, seek a more precise metric, or achieve greater precision in the original metric.
It might also be possible to correct the population estimates after-the-fact by "deconvolving"
population distribution functions (Stefanski and Bay, 1996), assuming the components of
variance are quantified. Even though one can increase the precision of stream
measurements by revisiting all streams two or more times and averaging the results, this is
usually an expensive alternative. As long as temporal variation within the summer season is
not the major source of uncertainty, the precision of physical habitat characterization of
stream reaches could be increased either by increasing the number of within-reach habitat
observation points, or by reducing the uncertainty of observations at each observation point.
If most of the variance of a metric is the result of substantial uncertainty in the
classifications or measurements themselves at a given position on a reach (rather than
short-term temporal variation in the true value of the habitat parameter being measured),
then little is gained by increasing the number of measurements. In this case one is better
advised to increase the precision of the measurements themselves, or to develop more
precise metrics by changing calculation or aggregation procedures (e.g., see Section 4.2.4
and Hughes etal., 1998).
83
-------
4.4.2 Effects on Associations Between Variables
In addition to its effects on survey estimates of the regional population distributions,
measurement imprecision can have predictable, but often unappreciated effects on our
ability to detect correlation between variables and our ability to assess the amount of
variation that can be attributed to a potential causal factor in using regression analysis. To
illustrate, consider a biological measurement ("Variable 1") and ancillary habitat
measurement ("Variable 2"). The Pearson product-moment correlation coefficient, r,
measures the proportion of the total variation of two variables that is shared between the
two variables (Snedecor and Cochran 1980):
°1 2
r = —'— (19)
where
o1j2 is the covariance between the two variables,
a1 is the standard deviation of Variable 1,
o2 is the standard deviation of Variable 2.
We apply assumptions made by Allen et al. (1999) to evaluate the effect of
measurement error on correlations. Assume for a moment that in reality, the underlying
characteristics that we attempt to measure with Variables 1 and 2 are perfectly correlated,
but the variables themselves are subject to random, unbiased measurement errors that are
uncorrelated between the two variables. If those assumptions hold, then:
°1.2,st(yr)
~ (20)
( £2 772 v f~2 772 ^
I \°1.st(yr) +01,rep Xy^stfyr) +°2,rep I
The effect of measurement imprecision on statistical correlation is clear from this
expression. The greater the proportion of measurement error, o2rep, the smaller will be the
observed r value, and the smaller will be the proportion of variance (r2) explained by a
regression predicting Variable 1 from Variable 2. Standardizing variances and retaining the
assumptions above, a rearrangement of Equation 20 yields a useful expression to relate the
magnitude of o2st(yr)/o2rep to the maximum observable values of r and r2 that could be
obtained in correlations and regressions between two variables:
84
-------
rmax -
5
°i,
st(yrj
°i2
rep
^.stfyrj
°2,rep
(21 a)
rmax —
(21 b)
'"max ~~
(21 c)
Under the stated assumptions, if two variables have S/N values of 3, for example,
we would not expect to observe a correlation between them higher than r^ = [3/(1 + 3)]% x
[3/(3 + 1)]1/z =0.75, even if in truth, the attributes measured by these variables are more
closely associated. When linear regression is used to predict Variable 1 from Variable 2,
we would expect a regression model to explain no more than 56% of the variance in
variable 2 (rzmax = 0.56). For convenience in interpreting the implications of differing S/N
values of metrics discussed in this report, Table 18 compares the predicted values of rmax for
various combinations of measurement precision. If the S/N values of two variables are both
1, a true underlying correlation of r=1.00 will be observed as a correlation of r=0.50;
similarly, only when two variables have S/N > 10 (or one has 5 and the other > 25) does
the observed correlation exceed r=0.90. Furthermore, if the less precise of two variables in
an association analysis has S/N < 2, the highest correlation coefficient (r) expected with the
other variable is 0.81, regardless of how precise the other variable is. Similarly, the highest
coefficient of determination (r2) expected in this case is 0.67. These observations have
85
-------
TABLE 18. THEORETICAL MAXIMUM OBSERVED CORRELATION COEFFICIENTS (r)
BETWEEN TWO METRICS OF VARYING PRECISION, AS MEASURED BY S/N ( a2st,vr/a2reD)a
S/N =
S/N =
1
2
3
5
10
25
50
100
1
0.50
0.58
0.61
0.65
0.67
0.69
0.70
0.70
2
-
0.67
0.70
0.75
0.78
0.80
0.81
0.81
3
-
-
0.75
0.79
0.83
0.85
0.86
0.86
5
-
-
-
0.83
0.87
0.90
0.90
0.91
10 25 50 100
-
-
-
-
0.91
0.93 0.96
0.94 0.97 0.98
0.95 0.98 0.99 0.99
Assuming the underlying correlation between attributes measured by the two metrics is 1.
important implications for correlation and regression analyses. The "strength" of the
underlying relationship suggested by observed association should be based on the
maximum explainable variation ("discarding" measurement variance, or "noise"). However,
this should not be interpreted as a rationale for embracing field approaches that have
inherently high measurement variability.
4.5 GENERALIZATIONS AND RECOMMENDATIONS CONCERNING METRIC
PRECISION
Based on our results, we make the following generalizations concerning the
precision of habitat measurement and assessment approaches:
1) Measurements are more precise than visual estimates, but carefully-
designed visual estimation procedures can be nearly as precise as
measurements. To enhance precision, these visual observations are limited
to measurable characteristics (e.g., cover or presence), rather than
judgements of habitat quality, and they are made at multiple locations within
a reach.
2) Flow-sensitivity and complex definitions of habitat features can degrade
precision of quantitative measurements (e.g., bankfull height and incision).
86
-------
3) Flow-sensitivity and subjectivity in habitat-unit classifications (e.g., %Pool)
can seriously limit their usefulness in contrasting stream habitat among
streams or in tracking changes in habitat through time.
4) The precision of multiple visual cover-class determinations can be improved
by re-interpreting this information as extent of presence-absence of some
defined feature (e.g., summed vegetation cover in two layers reinterpreted
as percent of observations in which cover is > 0% in both layers), but
perhaps at the expense of decreased sensitivity to stress.
5) The precision of separate metrics can be improved by combining them into
more integrated metrics. For example, the precision of %Substrate <16mm
diameter is more precise than the separate metrics of %Fine Gravel,
%Sand, or %Fines; the precision of %(Pools + Glides) is more precise than
%Pools. However, the gain in precision may be at the expense of
decreased sensitivity to stress.
6) While visual judgement methods are attractive because of their rapidity in
the field and in data reduction, their lack of precision limits their use in many
applications.
7) At least 20 within-season pairs of repeat visits to 8 to 20 field sites spread
over several years are required for confident assessment of within-season
precision in physical habitat metrics. These repeat samples are ideally
drawn as a random or stratified random sub-sample from a regional
probability sample of stream reaches.
8) Metrics with S/N < 2.0 distort estimates of regional distributions based on
survey results, and severely limit analyses of associations by regression and
correlation.
9) When metric S/N variance ratios are >10, field measurement variance and
short-term temporal fluctuations cause relatively insignificant error and
distortion in estimates of regional population distribution functions, and offer
relatively insignificant obstacles to analyses of association using regression
and correlation.
87
-------
4.6 OTHER CONSIDERATIONS IN SELECTING PHYSICAL HABITAT METRICS
At the beginning of this chapter, we emphasized the importance of accuracy,
precision, and ecological relevance in characterizing physical habitat. When initially
selecting a physical habitat measurement approach, the additional dimensions of
practicality, effort and cost come in to play. If the variables have already been measured
and the data is in-hand, one can choose the best set of variables to use in a particular
analysis, regardless of their cost. When selecting variables, it is obviously important that
those variables are accurate, faithfully depicting the attribute of habitat that we intend or
understand them to depict. Secondly, it is important that they be precise, sufficiently
repeatable so that measurement variation does not eclipse differences we want to be able
to detect. As we discuss earlier in this chapter, lack of precision can severely limit the utility
of a variable, but precision must be viewed relative to the magnitude of difference (or
change) one wants to detect. The S/N ratios we calculated compare measurement
precision with the variation of a metric across a region, adopting the observed range as a
surrogate for the range of "important" variation. Expected ranges of condition or
magnitudes of temporal change in any particular variable may differ substantially among
regions, or in the same region viewed at different scales. For this reason, S/N ratios should
be viewed only as predictions of the ability of a variable to discern "important" differences
within the same region from which they were calculated. Furthermore, S/N ratios calculated
from a particular survey (e.g. MAHA or Oregon) should be viewed only as approximations of
what the precision of measurements would likely be in a different region. For this reason,
we chose not to exclude or drop variables from consideration in a national monitoring
program solely on the basis of poor precision in a given region.
The final measure of the utility of a habitat characterization approach is whether it
contains useful information for interpreting controls on biota or impacts of human activity. In
regional surveys, or in temporal series, this measure of performance is demonstrated
through analysis of associations among variables. As with precision, this aspect of habitat
metric utility is also region-specific, and dependent on the type of biological assemblage and
the type of human disturbances present. However, we can offer some guidance based on
our own research and that of others who have used EMAP habitat data. Table 19 lists the
variables used most often in variety of multivariate and other types of analyses associating
habitat with fish assemblages (Herlihy et al., 1997; Hughes et al., 1998; McCormick et al., in
review; Howlin et al., in preparation), macroinvertebrate assemblages (Bryce et al., 1999; Li
et al., in review; Griffiths et al., in review), periphyton assemblages (Hill et al., in review; Pan
et al., 1999; Griffiths et al., in review), benthic metabolism (Hill et al., 1998), and landscape
disturbance (Bryce et al., 1999). It is evident that the list includes representatives from each
88
-------
TABLE 19. PHYSICAL HABITAT VARIABLES MOST FREQUENTLY USED
VARIABLE3
DESCRIPTION:
Channel Morphology:
XDEPTH Mean thalweg depth (cm)
SDDEPTH Standard deviation of thalweg depth (cm)
XWIDTH Mean wetted width (m)
XWXD Mean wetted width x depth (m2)
RP100 Mean residual depth (m2/100 m reach length) =cm
XBKF_W Mean bankfull width (m)
XBKFJH Mean bankfull height (m)
XINCJH Mean incision height (m)
SINU Channel Sinuosity
XSLOPE Water surface gradient over reach
Substrate:
LSUB_DMM
XEMBED
PCT_FN
PCT_SA
PCT_RC
PCTJHP
PCT_SAFN
PCT_SFGF
PCT_BIGR
PCT_BDRK
LTEST
LRBS_TST
LDMB_BW4
LRBS BW4
Substrate %
Substrate %
Substrate %
Log10[estimated geometric mean substrate diameter (mm)]
Substrate mean embeddedness - channel + margin (%)
Substrate % fine (silt/clay)
Substrate % sand (0.6 to 2mm)
concrete
hard pan
sand + fines (< 2 mm)
Substrate % fine gravel and smaller (< 16mm)
Substrate % coarse gravel and larger (> 16mm)
Substrate % bedrock
Log10 [Erodible substrate diameter (mm) ] - Estimate 1 (see text)
Log10 [Relative Bed Stability] - Estimate 1 (see text)
Log10 [Erodible substrate diameter (mm) ] - Estimate 2 (see text)
Log10 [Relative Bed Stability] - Estimate 2 (see text)
Fish Cover and Woody Debris:
XFC_ALG Filamentous algae areal cover
XFC_AQM Aquatic macrophyte areal cover
XFC_LWD Large woody debris areal cover
XFC_BRS Brush and small woody debris areal cover
XFC_OHV Overhanging vegetation areal cover
XFC_BIG Sum of cover from large wood, boulders, over-hanging banks and human
structures
XFC_NAT Sum of cover from large wood, brush, overhanging vegetation, boulders and
undercut banks
V1W_MSQ LWD volume in active channel (m3/m2) - size classes 1 to 5)
V1TM100 LWD volume in and above active channel (m3/100m) - size classes 1 to 5)
Variable names in bold are considered to be most important.
(continued)
89
-------
TABLE 19 (Continued)
VARIABLE3
DESCRIPTION:
Riparian Vegetation Cover and Structure:
XCDENBK Mean % canopy density at bank
XCDENMID Mean % canopy density midstream
XCL Riparian canopy (> 5 m high) cover - trees > 0.3 m DBH (diameter at breast height)
XGB Riparian ground-layer (< 0.5 m high) bare ground cover
XC Riparian canopy cover (XCL+XCS)
XCM Riparian canopy + mid-layer cover (XC + XM)
XCMGW Riparian woody cover, sum of 3 layers (XC + XMW + XGW)
XPCAN Riparian canopy presence (proportion of reach)
XPCM Riparian canopy and mid-layer presence (proportion of reach)
XPCMG 3-layer riparian vegetation presence (proportion of reach)
PCAN_C Coniferous riparian canopy presence (proportion of reach)
Human Disturbances:
W1 H_WALL Riparian Human disturbance - Channel revetment (proximity-weighted index)
W1H_LOG Riparian Human disturbance -Logging (proximity-weighted index)
W1JHALL Riparian Human Disturbance Index (proximity-weighted sum)
W1JHNOAG Riparian Human Disturbance Index- Non-agricultural types (proximity-weighted
sum)
W1JHAG Riparian Human Disturbance Index - Agricultural types (proximity-weighted sum)
a Variable names in bold are considered to be most important.
of the seven aspects of habitat presented in the Introduction to this report. Of the 49
variables, we highlight a balanced set of 18 variables that we consider generally the most
important, but recommend that researchers examine the suite of variables in Table 6 and
take into consideration their precision (Tables 7 through 15), the region, type of biota, and
their own particular research objectives.
90
-------
5 LITERATURE CITED
Allen, A.P., T.R. Whittier, P.R. Kaufmann, DP. Larsen, R.J. O'Conner, R.M. Hughes, R.S.
Stemberger, S.S. Dixit, R.O. Brinkhurst, and AT. Herlihy. 1999. Concordance of
taxonomic richness patterns across five assemblages in lakes of the northeastern U.S.
Can. J. Fish. Aquat. Sci. 56:739-747.
Bain, M.B., J.T. Finn, and H.E. Booke. 1985. Quantifying stream substrate for habitat
analysis studies. North Amer. J. Fish. Mgmt. 5:499-500.
Barbour, M.T., J. Gerritsen, B.D. Snyder, and J.B. Stribling. 1997. DRAFT Revision to
Rapid Bioassessment Protocols for Use in Streams and Rivers. 841-D-97-002. U.S.
Environmental Protection Agency, Office of Water, Washington, DC.
Barbour, M.T. and J.B. Stribling. 1991. Use of habitat assessment in evaluating the
biological integrity of stream communities, in: G. Gibson ( ed.), Biological criteria:
research and regulation, proceedings of a symposium, 12-13 December 1990,
Arlington, Virginia. EPA-440/5-91/005. U.S. Environmental Protection Agency, Office
of Water, Washington, D.C.
Barbour, M.T. and J.B. Stribling. 1994. A technique for assessing stream habitat structure.
pp 156-178 jri: Riparian Ecosystems in the Humid U.S.: Function, Values and
Management. Conference Proceedings, National Association of Conservation
Districts, Washington D.C. March 15-18, 1993, Atlanta, Georgia.
Bathurst, J.C. 1981. Bar resistance of gravel-bed streams. ASCE J. Hydraulics Div.
106:1276-1278.
Bauer, S.B., and T.A. Burton. 1993. Monitoring Protocols to Evaluate Water Quality Effects
of Grazing Management on Western Range/and Streams. EPA 910/9-91-001. U.S.
Environmental Protection Agency, Region X, Seattle, WA. 166 p.
91
-------
Benson, B.J. and J.J. Magnuson. 1992. Spatial heterogeneity of littoral fish assemblages
in lakes: relation to species diversity and habitat structure. Can. J. Fish. Aquat. Sci.
49:1493-1500.
Bishop, G.D., M.R. Church, and C. Daly. 1998. Effects of improved precipitation estimates
on automated runoff mapping: Eastern United States. J. Am. Wtr. Resour. X\ssoc.
34(1):159-166.
Bisson, P.A., J.L. Nielsen, R.A. Palmason, and L.E. Grove. 1982. A system of naming
habitat types in small streams, with examples of habitat utilization by salmonids during
low stream flow. pp. 62-73 in: N.B. Armantrout (ed.), Acquisition and utilization of
aquatic habitat inventory information. Symposium Proceedings, October 28-30, 1981,
Portland, OR. The Hague Publishing, Billings, MT.
Bryce, S., D.P. Larsen, R.M. Hughes, and P.R. Kaufmann. (1999). Assessing the relative
risks to aquatic ecosystems in the Mid-Appalachian region of the United States. J. Am.
Wtr. Resource Assoc. 35(1):23-36.
Buffington, J.M. 1995. Effects of Hydraulic Roughness and Sediment Supply on Surface
Textures of Gravel-bedded Rivers. M.S.Thesis, University of Washington, Seattle,
WA. 184p.
Buffington, J.M. 1998. The Use of Streambed Texture to Interpret Physical and Biological
Conditions at Watershed, Reach, and Sub-reach Scales. Ph.D. dissertation, University
of Washington, Seattle, WA. 148p.
Buffington, J.M. and D.R. Montgomery. 1992. Effects of hydraulic roughness and sediment
supply on bed surface textures in gravel-bed streams. EOS, Trans AGU 73:231.
Chow, V.T. 1959. Open-channel Hydraulics. McGraw-Hill Book Co., New York, 680 p.
Cummins, K.W. 1974. Structure and function of stream ecosystems. Bioscience 24:631-
641
Daubenmire, R. 1969. Plant Communities: A Textbook of Plant Synecology. Harper &
Row, New York. 300 p.
Dietrich, W.E., J.W. Kirchner, H. Ikeda, and F. Iseya. 1989. Sediment supply and the
development of the coarse surface layer in gravel bed rivers. Nature 340(20):215-217.
92
-------
Dingman, S.L 1984. Fluvial Hydrology. W.H. Freeman, New York. 383 p.
Einstein, H.A. and R.B. Banks. 1950. Fluid resistance of composite roughness. EOS, Trans.
AGU 117:1121-1146.
Fitzpatrick, F.A., I.R. Waite, P.J. D'Arconte, M.R. Meador, M.A. Maupin, and M.E. Gurtz.
1998. Revised Methods for Characterizing Stream Habitat in the National Water-
Quality Assessment Program. Water-Resources Investigations Report 98-4052, U.S.
Geological Survey. Raleigh, N.C. 67 p.
Frissell, C.A., W.J. Liss, C.E. Warren, and M.D. Hurley. 1986. A hierarchical framework for
stream habitat classification: viewing streams in a watershed contest. Environ. Mgmt.
10(2): 199-214.
Gorman, O.T. and J.R. Karr. 1978. Habitat structure and stream fish communities. Ecology
59(3):507-515.
Gregory, S.V., F.J. Swanson, W.A. McKee, and K.W. Cummins. 1991. An ecosystem
perspective of riparian zones. Bioscience 41(8):540-551.
Griffith, M.B., B.H. Hill, AT. Herlihy, and P.R. Kaufmann. (in review). Multivariate analysis
of periphyton assemblages in relation to environmental gradients in Colorado Rocky
Mountain streams J. North Amer. Benthol. Soc.
Griffith, M.B., P.R. Kaufmann, AT. Herlihy, and B.H. Hill, (in review). Analysis of
macroinvertebrate assemblages in relation to environmental gradients in Rocky
Mountain streams. J. North Amer. Benthol. Soc.
Gurtz, M.E. and T.A. Muir (eds). 1994. Report of the I nteragency Biological Methods
Workshop. Open-File Rep. 94-490, U.S. Geological Survey, Raleigh, N.C. 85 p.
Hankin, D.G., and G.H. Reeves. 1988. Estimating total fish abundance and total habitat
area in small streams based on visual estimation methods. Can. J. Fish. Aquat. Sci.
45:834-844.
Hannaford, M.J., M.T. Barbour, and V.H. Resh. 1997. Training reduces observer variability
in visual-based assessments of stream habitat. J. North Amer. Benthol. Soc.
16(4): 853-860.
93
-------
Hannaford, M.J. and V.H. Resh. 1995. Variability in macroinvertebrate rapid-
bioassessment surveys and habitat assessments in a northern California stream. J.
North Amer. Benthol. Soc. 14(3):430-439.
Harmon, M.E., J.F. Franklin, F.J. Swanson, P. Sollins, S.V. Gregory, J.D. Lattin, N.H.
Anderson, S.P. Cline, N.G. Aumen, J.R. Sedell, G.W. Lienkaemper, K. Cromack, Jr.,
and K.W. Cummins. 1986. Ecology of coarse woody debris in temperate ecosystems.
/Advances in Ecological Research 15:133-302.
Hawkins, C.P., J.L. Kershner, P.A. Bisson, M.D. Bryant, LM. Decker, S.V. Gregory, D.A.
McCullough, C.K. Overton, G.H. Reeves, R.J. Steedman, and M.K. Young. 1993. A
hierarchical approach to classifying stream habitat features. Fisheries 18:3-12.
Hawkins, C.P., M.L. Murphy, and N.J. Anderson. 1983. Density offish and salamanders in
relation to riparian canopy and physical habitat in streams of the northwestern United
States. Can. J. Fish. Aquat. Sci. 40(8): 1173-1186.
Herlihy, A., P.Kaufmann, L.Reynolds, J.Li, and E.G.Robison. 1997. Developing indicators
of Ecological Condition in the Willamette Basin, pp 275-282 In: A. Laenen and D.
Dunnette (eds.). River Quality: Dynamics and Restoration. Lewis Publishers, CRC
Press, Boca Raton, FL.
Hill, B. H., A. T. Herlihy, P. R. Kaufmann, and R. L. Sinsabaugh. 1998. Sediment microbial
respiration in a synoptic survey of mid-Atlantic region streams. Freshwater Biology.
39:493-501.
Hill, B.H., AT. Herlihy, P.R. Kaufmann, R.J. Stevenson, and F.H. McCormick. (in review).
The use of periphyton assemblage data as an index of biotic integrity. J. North Amer.
Benthol. Soc.
Hogle, J.S., T.A. Wesche, and W.A. Hubert. 1993. A test of the precision of the habitat
quality index model II. North Am. J. Fish. Mgmt. 13:640-643.
Howlin, S., R.M. Hughes, and P.R. Kaufmann. (in prep.) Development of an Index of
Biointegrity for coldwater streams.
Hughes, R.M. (ed.). 1993. Stream Indicator and Design Workshop. EPA/600/R-93/138.
U.S. Environmental Protection Agency, Office of Research and Development,
Corvallis, Oregon.
94
-------
Hughes, R. M., T. M. Kincaid, P. R. Kaufmann, A. T. Herlihy, L. Reynolds, and D. P. Larsen.
1998. A process for developing and evaluating indices of fish assemblage integrity.
Can. J. Fish. Aquat. Sci. 55:1618-1631.
Hynes, H.B.N. 1972. Ecology of Running Waters. Univ. of Toronto Press, Canada. 555p.
Kappesser, G. 1993. Riffle Stability Index. In-house report for Idaho Panhandle National
Forests. USDA Forest Service Report, Coeur d'Alene, ID.
Karr, J.R., K.D. Fausch, P.L. Angermeier, P.R. Yant, and I.J. Schlosser. 1986. Assessing
Biological Integrity in Running Waters: A Method and its Rationale. Illinois Nat. Hist.
Survey Spec. Publ. 5. Champaign, IL.
Kaufmann, P.R. 1987a. Channel Morphology and Hydraulic Characteristics of Torrent-
Impacted Forest Streams in the Oregon Coast Range, U.S.A. Ph.D. Thesis.
Department of Forest Engineering/Hydrology, Oregon State University, Corvallis, OR.
225 pp.
Kaufmann, P.R. 1987b. Slackwater Habitat in Torrent-Impacted Streams. Pages 407-408.
in: R.L. Beschta, T. Blinn, G.E. Grant, F.J. Swanson, and G.E. Ice (eds.). Erosion and
Sedimentation in the Pacific Rim. International Association of Hydrologic Science.
Publication No. 165. Proceedings of an International Symposium, August 3-7, 1986.
Oregon State University, Corvallis, OR.
Kaufmann, P.R. (ed.). 1993. Physical Habitat, pp. 59-69 in: R.M. Hughes (ed.). Stream
Indicator and Design Workshop. EPA/600/R-93/138. U.S. Environmental Protection
Agency, Office of Research and Development, Corvallis, Oregon.
Kaufmann, P.R. 1998. Stream Discharge, pp 67-76 Iri: J.M. Lazorchak, D.J. Klemm and
D.V. Peck (eds.). Environmental Monitoring and Assessment Program - Surface
Waters: Field Operations and Methods for Measuring the Ecological Condition of
Wadeable Streams. EPA/620/R-94/004F. U.S. Environmental Protection Agency,
Office of Research and Development, Washington, D.C.
Kaufmann, P.R., R.M. Hughes, T.R. Whittier, S.A. Thiele, and D.P. Larsen. (in review).
Lake Shore and Littoral Habitat Structure: Precision and Biological Relevance of a
Field Survey Method. Can. J. Fish. Aquat. Sci.
95
-------
Kaufmann, P.R. and E.G. Robison. 1994. Physical Habitat Assessment. pp6-1 to 6-38 ]n:
D.J. Klemm and J.M. Lazorchak (eds.). Environmental Monitoring and Assessment
Program 1994 Pilot Field Operations Manual for Streams. EPA/620/R-94/004. U.S.
Environmental Protection Agency, Office of Research and Development, Cincinnati,
Ohio.
Kaufmann, P.R. and E.G. Robison. 1998. Physical Habitat Characterizaton. pp 77-118 In:
J.M. Lazorchak, D.J. Klemm and D.V. Peck (eds.). Environmental Monitoring and
Assessment Program - Surface Waters: Field Operations and Methods for Measuring
the Ecological Condition of Wadeable Streams. EPA/620/R-94/004F. U.S.
Environmental Protection Agency, Office of Research and Development, Washington,
D.C.
Kaufmann, P.R. and T.R. Whittier. 1997. Habitat Assessment. Pages 5-1 to 5-26 in:
J.R. Baker, D.V. Peck, and D.W. Sutton (Eds.). Environmental Monitoring and
Assessment Program -Surface Waters: Field Operations Manual for Lakes.
EPA/620/R-97/001. U.S. Environmental Protection Agency, Washington, D.C.
Klemm, D.J. and J.M. Lazorchak (eds.). 1994. Environmental Monitoring and Assessment
Program 1994 Pilot Field Operations Manual for Streams. EPA/620/R-94/004. U.S.
Environmental Protection Agency, Office of Research and Development, Cincinnati,
Ohio.
Larsen, D.P., and N.S. Urquhart. 1993. A framework for evaluating the sensitivity of the
EMAP design, pp 119-155 in D.P. Larsen and S.J. Christie (eds.), EMAP-Surface
Waters 1991 Pilot Report. EPA/620/R-93/003. U.S. Environmental Protection Agency,
Office of Research and Development, Washington, D.C.
Lazorchak, J.M., D.J. Klemm and D.V. Peck (eds.). 1998. Environmental Monitoring and
Assessment Program - Surface Waters: Field Operations and Methods for Measuring
the Ecological Condition of Wadeable Streams. EPA/620/R-94/004F. U.S.
Environmental Protection Agency, Office of Research and Development, Washington,
D.C. 211pp plus Appendices.
Lemmon, P.E. 1957. A new instrument for measuring forest overstory density. J. Forestry.
55(9):667-669.
Leopold, L.B. and M.G. Wolman. 1957. River Channel Patterns - Braided, Meandering,
Straight. U.S. Geol. Surv. Prof. Paper 282B.
96
-------
Leopold, , L.B. M.G. Wolman, and J.P. Miller. 1964. Fluvial Processes in Geomorphology.
W.H. Freeman and Co., San. Fran. CA, USA. 522 p.
Li, J., AT. Herlihy, W. Gerth, P.R. Kaufmann, S.V. Gregory, and D.P. Larsen. (in review).
Quantifying variation in stream macroinvertebrate assemblages at multiple spatial
scales: the relative influence of sample size and spatial distribution. Freshwater
Biology.
Lisle, I.E. 1982. Effects of aggradation and degradation on riffle-pool morphology in natural
gravel channels, northwestern California. Water. Resour. Res. 18(6):1643-1651.
Lisle, I.E. 1986. Effects of woody debris on anadromous salmonid habitat, Prince of Wales
Island, Southeast Alaska. North Amer. J. Fish. Mgmt. 6:538-550.
Lisle, T. E. 1987. Using" Residual Depths" to Monitor Pool Depths Independently of
Discharge. USDA Forest Service Pacific. SW Forest and Range Exper. Sta. Research
Note PSW-394. 4pp.
Lisle, I.E. and S. Hilton. 1992. The volume of fine sediment in pools: an index of sediment
supply in gravel-bed streams. Water Res. Bull. 28(2):371-383.
MacDonald, L.H., A.W. Smart, and R.C. Wissmar. 1991. Monitoring Guidelines to Evaluate
Effects of Forestry Activities on Streams in the Pacific Northwest and Alaska. WPA
910/9-91-001. U.S. Environmental Protection Agency, Region X, Seattle, Washington.
166 p.
Mackin, J.H. 1948. Concept of the graded river. Geol. Soc. Am. Bull. 59:463-512.
McCormick, F.H., R.M. Hughes, P.R. Kaufmann, AT. Herlihy, and D.V. Peck, (in review).
Development of an index of biotic integrity for the Mid-Atlantic Highlands region.
Trans. Amer. Fish. Soc.
Meador, M.R., C.R. Hupp, T.F. Cuffney, and M.E. Gurtz. 1993. Methods for characterizing
stream habitat as part of the National Water-Quality Assessment Program. U.S.
Geological Survey Open-File Report 93-408. U.S. Geological Survey, Raleigh, N.C. 48
pp.
97
-------
Montgomery, D.R. and J.M. Buffington. 1993. Channel Classification, Prediction of
Channel Response, and Assessment of Channel Condition. Washington State
Timber/Fish/Wildlife Agreement, Report TFW-SH10-93-002, Dept. of Natural
Resources, Olympia, WA.
Montgomery, D.R. and J.M. Buffington. 1997. Channel-reach morphology in mountain
drainage basins, Geol. Soc. Am. Bull. 109:596-611.
Montgomery, D.R. and J.M. Buffington. 1998. Channel processes, classification, and
response, pp 13-42 Iri: R. Naiman and R. Bilby (eds.). River Ecology and
Management. Springer-Verlag, New York.
Montgomery, D.R., M.S. Panfil, and S.K. Hayes. 1999. Channel-bed mobility response to
extreme sediment loading at Mount Pinatubo. Geology 27(3):271-274.
Moore, K.M.S. and S.V. Gregory. 1988. Summer habitat utilization and ecology of cutthroat
trout fry (Salmo clarki) in Cascade mountain streams. Can. Jour. Fish. Aquat. Sci.
45:1921-1930.
Morisawa, M. 1968. Streams, Their Dynamics and Morphology. McGraw-Hill, New York.
175 p.
Mulvey, M., L. Caton, and R. Hafele. 1992. Oregon Nonpoint Source Monitoring Protocols:
Stream Bioassessment Field Manual: for Macroinvertebrates and Habitat Assessment.
Oregon Department of Environmental Quality, Laboratory Biomonitoring Section. 1712
S.W. 11th Ave. Portland, Oregon, 97201. 40 p.
Naiman, R.J., H. Decamps, J. Pastor, and C.A. Johnston. 1988. The potential importance
of boundaries to fluvial ecosystems. J. North Amer. Benthol. Soc. 7(4):289-306.
Nelson, J.M. and J.D. Smith. 1989. Flow in meandering channels with natural topography.
pp. 69-102 in: S. Ikeda and G. Parker (eds.). River Meandering. American
Geophysical Union, Washington, D.C.
Neter, J. and W. Wasserman. 1974. Applied Linear Statistical Models. Richard D. Irwin,
Inc., Homewood, IL. 60430. 842 p.
O'Neill, M.P. and A.D. Abrahams. 1984. Objective identification of pools and riffles. Water
Resour. Res. 20(7):921-926.
98
-------
Overton, C.K., S.P. Wollrab, B.C. Roberts, and M.A. Radko. 1997. R1/R4
(Northern/lntermountain Regions) Fish and Fish Habitat Standard Inventory
Procedures Handbook. Gen. Tech. Rep. INT-GTR-346. Ogden, UT: USDA Forest
Service Intermountain Research Station. 73 p.
Overton, W.S. 1989. Effects of Measurement and Other Extraneous Errors on Estimated
Distribution Functions in the National Surface Water Surveys. Technical Report 129.
Department of Statistics, Oregon State University, Corvallis, OR.
Pan, Y., R.J. Stevenson, B.H. Hill, P.R. Kaufmann, and AT. Herlihy. 1999. Spatial patterns
and ecological determinants of benthic algal assemblages in Mid-Atlantic streams,
U.S.A. J. Phycology 35:460-468.
Paola, C. and D. Mohrig. 1996. Paleohydraulics revisited: palaeoslope estimation in coarse-
grained braided rivers. Basin Research 8:43-254.
Parker, G., P.C. Klingeman, and D.G. McLean. 1982. Bedload and size distribution in paved
gravel-bed streams. J. Hydraul. Div., ASCE 108:544-571.
Paulsen, S.G., D.P. Larsen, P.R. Kaufmann, T.R. Whittier, J.R. Baker, D.V. Peck, J.
McGue, D. Stevens, J. Stoddard, R.M. Hughes, D. McMullen, J. Lazorchak, and W.
Kinney. 1991. Environmental Monitoring and Assessment Program (EMAP) Surface
Waters Monitoring and Research Strategy - Fiscal Year 1991. EPA 600/3-91/022.
U.S. Environmental Protection Agency, Washington, DC. 184 p.
Plafkin, J.L, M.T. Barbour, K.D. Porter, S.K. Gross, R.M. Hughes. 1989. Rapid
Bioassessment Protocols for Use in Streams and Rivers: Benthic Macroinvertebrates
and Fish. EPA/440/4-89/001. U.S. Environmental Protection Agency, Assessment and
Watershed Protection Division, Washington, DC.
Platts, W.S. 1981. Stream inventory garbage in - reliable analysis out: only in fairy tales.
pp. 75-84. In: N.B. Armantrout (ed.). Acquisition and Utilization of Aquatic Habitat
Inventory Information. American Fisheries Society, Western Division, Bethesda,
Maryland.
Platts, W.S., W.F. Megahan, and G.W. Minshall. 1983. Methods for evaluating stream,
riparian and biotic conditions. Gen. Tech. Rep. INT-138. U.S. Forest Service,
Intermountain Forest and Range Experiment Station, Ogden, UT. 70 p.
99
-------
Poole, G.C., C.A. Frissell, and S.C. Ralph. 1997. In-stream habitat unit classification:
inadequacies for monitoring and some consequences for management. J. Amer.
Water Resour. Assoc. 33(4):879-896.
Ralph, S.C., T. Cardoso, G.C. Poole, L.L Conquest, and R.J. Naiman. 1991. Status and
Trends of Instream Habitat in Forested Lands of Washington: The Timber-Fish-Wildlife
Ambient Monitoring Project. Biennial progress report, Center for Streamside Studies,
University of Washington, Seattle, Washington.
Ralph, S.C., G.C. Poole, L.L. Conquest, and R.J. Naiman. 1994. Stream channel
morphology and woody debris in logged and unlogged basins of western Washington.
Can. J. Fish. Aquat. Sci. 51:37-51.
Rinne, J. 1988. Effects of livestock grazing exclosure on aquatic macroinvertebrates in a
montane stream, New Mexico. Great Basin Naturalist 48(2): 146-153.
Robison, E.G. 1998. Reach Scale Sampling Metrics and Longitudinal Pattern Adjustments
of Small Streams. Ph.D. Dissertation. Dept. of Forest Engineering and Hydrology.
Oregon State University, Corvallis, OR. 254 p.
Robison, E.G. and P.R. Kaufmann. 1994. Evaluating two objective techniques to define
pools in small streams, pp. 659-668, |n: R.A. Marston and V.A. Hasfurther (eds.).
Effects of Human Induced Changes on Hydrologic Systems. Summer Symposium
proceedings, American Water Resources Association,. June 26-29, 1994, Jackson
Hole, Wyoming. 1182 pp.
Robison, E.G. and R.L. Beschta. 1989. Estimating stream cross-sectional area from
wetted width and thalweg depth. Physical Geography 10(2): 190-198.
Robison, E.G. and R.L. Beschta. 1990. Characteristics of coarse woody debris for several
coastal streams of southeast Alaska, USA. Can. J. Fish. Aquat. Sci. 47(9): 1684-1693.
Roper, B.R. and D.L. Scarnecchia. 1995. Observer variability in classifying habitat types in
stream surveys. North Amer. J. Fish. Mgmt. 15:49-53.
Rosgen, D.L. 1985. A stream classification system, pp 91-95 jn: R.R. Johnson, C.D.
Zeibell, D.R. Patton, P.P. Ffolliott, and R.H. Hamre (eds.). Riparian Ecosystems and
Their Management: Reconciling Conflicting Uses. USDA For. Serv. Gen. Tech. Rep.
RM-120. Fort Collins, CO.
100
-------
Rosgen, D.L 1994. A classification of natural rivers. Catena 22:169-199.
Schumm, S.A. 1963. Sinuosity of alluvial rivers of the Great Plains: Geol. Soc. Am. Bull.
74:1089-1100.
Schumm, S.A. 1971. Fluvial geomorphology: Channel adjustment and river
metamorphosis, pp. 5-1 to 5-22 jn: H.W. Shen (ed.). River mechanics, Vol. I. H.W.
Shen, Fort Collins, CO.
Shields, F.D. and C.J. Gippel. 1995. Prediction of effects of woody debris removal on flow
resistance. J. Hydraul. Engr., ASCE 121:341-354.
Simons, D.B. and F. Senturk. 1977. Sediment Transport Technology. Water Resources
Publications, Fort Collins, CO. 80522, USA. 807 p.
Simonson, T.D., J. Lyons, and P.O. Kanehl. 1994. Quantifying fish habitat in streams:
transect spacing, sample size, and a proposed framework. North Amer. J. Fish. Mgmt.
14:607-615.
Snedecor, G.W. and W.G. Cochran. 1980. Statistical Methods. Seventh edition, Iowa
State University Press, Ames, I A, USA. 507 p.
Stack, W.R. 1989. Factors Influencing Pool Morphology in Oregon Coastal Streams. M.S.
Thesis, Oregon State University. 109 p.
Stack, W.R. and R.L. Beschta. 1989. Factors influencing pool morphology in Oregon
coastal streams, pp. 401-411 |n: W.W. Woessnerand D.F. Potts (eds.). Headwaters
Hydrology Symposium. American Water Resources Association. 708 p.
Stanfield, L, M. Jones, M. Stoneman, B Kilgour, J. Parish, and G. Wichert. 1997. Stream
Assessment Protocol for Ontario. Ontario Ministry of Natural Resources, Fish and
Wildlife Branch. Peterborough, Ontario, Canada, K9J 8M5.
Stanford, J.A. and J.V. Ward. 1993. An ecosystem perspective of alluvial rivers:
connectivity and the hyporheic corridor. J. North Amer. Benthol. Soc. 12(1):48-60.
Stauffer, J.C. and R.M. Goldstein. 1997. Comparison of three qualitative habitat indices
and their applicability to Prairie streams. North Amer. J. Fish. Mgmt. 17:348-361.
101
-------
Stefanski, L. A. and J. M. Bay. 1996. Simulation extrapolation deconvolution of finite
population cumulative distribution function estimatiors. Biometrika 83(2): 496-417.
Stevens, D. L., Jr. and A. R. Olsen. (in press). Spatially restricted surveys overtime for
aquatic resources. J. Agric., Biol, and Environ. Statistics.
Sutherland, A.J. 1987. Static armour layers by selective erosion, pp. 243-267 |n: C.R.
Thorne, J.C. Bathurst, and R.D. Hey (eds.). Sediment Transport in Gravel-bed Rivers.
John Wiley & Sons, Chichester, UK.
Wang, L., T.D. Simonson, and J. Lyons. 1996. Accuracy and precision of selected stream
habitat estimates. North Amer J. Fish. Mgmt. 16:340-347.
Wilcock, P.R. 1997. The components of fractional transport rate. Water Resour. Res.
33(1):247-258.
Wilcock, P.R. 1998. Two-fraction model of initial sediment motion in gravel-bed rivers.
Science 280:410-412.
Wolman, M.G. 1954. A method of sampling coarse river-bed material. Trans. Am.
Geophys. Union 35(6):951-956.
Wood-Smith, R.D. and J.M. Buffington. 1996. Multivariate geomorphic analysis of forest
streams: Implications for assessment of land use impact on channel condition. Earth
Surf. Proc. and Landforms 21:377-393.
Yalin, M.S. and E. Karahan. 1979. Inception of sediment transport. Proc. ASCE J.
Hydraulics Div. 105(HY-11):1433-1443.
102
-------
APPENDIX A. COMPLETED EXAMPLES OF FIELD DATA FORMS FOR PHYSICAL
HABITAT CHARACTERIZATION
A completed set of EMAP field data forms for physical habitat characterization is
included here. These forms are the same as those presented in the EMAP-Surface Waters
field operations manual for wadeable streams (Lazorchak et al., 1998). Figure A-1
illustrates the form used to record instream and riparian measurement data at each cross-
section transect. Figure A-2 shows the data form used to record measurement data
collected along the longitudinal thalweg profile. Figure A-3 is an example of the form used
to record backsighted slope and bearing measurements. Physical habitat data from one
sample reach would be contained on 12 sheets. The cross-section and longitudinal thalweg
profile data occupy the front and back sides of 11 sheets, one for each transect. The slope
and bearing data for all transects of a reach are contained on one sheet.
A-1
-------
a
ui
0-
<5 <
r to
_l 00
i!
^i
J> u
(1 o
[40
(
g
i l&
Eiitft
<«Ei>
U. > 1.
• o £
= CJ O
SIZE C
COD
o|
Ol
X
GING
SURF
7o
IIII
0.0)
II
« -2g
-------
Reviewed by (initial):
PHab: THALWEG PROFILE & WOODY DEBRIS FORM - STREAMS
SITE NAME:
/Vf,tt
DATE: 7 1 if 1 97 VISIT: 81 D2 _
SITE ID: MAIA97
TEAM ID (X): 01 D2 D3 D4 D5 D6 D7 D8
TRANSECT(X): DA-B HB-C DC-D DD-E DE-F DF-G DG-H DH-I DW Dj-K
THALWEG PROFILE
Increment (m) -
/.J-
STA-
TION
THALWEG
DEPTH
(cm)
(XXX)
WETTED
WIDTH (ra)
(XX.X)
BAR WIDTH1
SOFT/SMALL
SEDIMENT
(X FOR YES)
CHANNEL
UNIT
CODE
POOL
FORM
CODE
SIDE
CHANNEL
X FOR YES)
COMMENTS
IH
Rf
13
ftj
N
N
ft
HO
3f
f.H
/.o
pr
31
PT
PT
?T
LARGE WOODY DEBRIS (2 10 cm SMALL END DIAMETER.; i 1.5 m LENGTH)
- TALLY EACH PIECE -
DIAMETER
LARGE END
0.3 -0.6m
PIECES ALL/PART IN BANKFULL CHANNEL
01
o
0
/(//
nr
o
JZ
0
PIECES BRIDGE ABOVE BANKFULL CHANNEL
(7
\Q_
CHANNEL UNIT CODES
RA
CA
Pool, Plunge
Pool, Trench
Pool, Lateral Scour
Pool, Backwater
Pool, Impoundment
Glide
Riffle
Rapid
Cascade
Falls
Dry Channel
POOL FORM CODES
Not» pool
Large Woody Debris
Rootwad
Boulder or bedrock
Unknown, fluvial
Other (note in comments)
FLAG
COMMENTS
Fl
FlagCodes: K=nomeasurementmade; U=suspect measurement; F1,F2,etc.=misc.flagsassignedbyeachfieldcrew. Explainallflags in comments. 1BMeasure
Bar Width at Station 0 and Mid-Station (5 or 7), X small column if bar present at the rest of the stations.
Rev. 06/02/97 (st_phct.97)
PHab: CHANNEL/RIPARIAN CROSS-SECTION & THALWEG PROFILE FORM - STREAMS - 2
Figure A-2. Thalweg Profile and Woody Debris Form. From Kaufmann and Robison (1998).
A-3
-------
.5
'c
01
DC
CO
0>
§
LU
cc
O z
a o
g^
a.
is
m
CO
UJ
3
CO
3
CO
t
UJ
s?
S
vs
ViJ
s
m
ol
8?
O
o
a-
Oo
o
m
s?
0?
o
H
H
H
s?
Ol
fx
H
5?
S?
ol
O
H
PO
tfi
I
cc
EC
O
LL
LU
OH
UJ
O.
3
CO
J2
o
Figure A-3. Slope and Bearing data form. From Kaufmann and Robison (1998).
A-4
-------
APPENDIX B. DATA ENTRY, VERIFICATION, AND DATABASE STRUCTURE
B.1 DATA STRUCTURE B-2
B.1.1 Introduction and Typographical Conventions B-2
B.1.2 Structure of Data Collection B-2
B.1.3 Data File Structure B-4
B.2 DATA VERIFICATION AND VALIDATION B-5
B.2.1 Getting Started B-5
B.2.2 Preliminary Structure and Variable Attribute Check B-6
B.2.3 Verifying and Validating Individual Data Files B-7
B.2.3.1 Data File Structure B-7
B.2.3.2 Missing Values B-8
B.2.3.3 Values Outside Allowable Range B-8
B.2.3.4 Unusual Values B-8
B.2.3.5 Channel Morphology B-9
B.2.3.6 Internal Logic B-9
B.3 COMPUTER CODE FOR VER/VAL OF INDIVIDUAL DATA FILES B-10
B.3.1 File canpycov Checks B-10
B.3.2 File fishcov Checks B-10
B.3.3 File riparian Checks B-11
B.3.4 File sub_bank Checks B-13
B.3.5 File thalweg Checks B-16
B.3.6 File Igwoody Checks B-19
B.3.7 Computer Code Involving More Than One Data File B-20
B.4 FINAL DATA VALIDATION CONSIDERATIONS B-21
B-1
-------
B.1 DATA STRUCTURE
B.1.1 Introduction and Typographical Conventions
Before beginning verification and validation (VerVal) of Physical Habitat (PHab) data
it is advisable to be familiar with the sampling techniques and protocol used to collect the
data. Of particular interest is the layout and structure of the sampling reach, as this directly
determines the structure of the individual PHab data files. The general sampling scheme is
discussed in Section 2 of the main body of this report (Synopsis ofEMAP Physical Habitat
Field Methods) and also the Physical Habitat chapter of EMAP's field operations and
training manual (Kaufmann and Robison, 1998).
We use the following typographical conventions for identifying data file names,
variable names, and VerVal computer code throughout this document:
Data file names are written in lower case, bold italics, (e.g. sub_bank,
thalweg).
PHab measurement and metric variable names are written in upper
case letters, as are the names and values of location and identification
variables (e.g. WT_WID, TRANSDIR, STRMJD, PCT_SA, ORC38,
ORST97-047).
When referring to a particular part of this document we will use the
term 'section', and italicize the name of the section we are referring to,
(e.g., 'in the structure check section of this document').
When referring to a piece of Statistical Applications Software (SAS)
code we have included we will specify 'SAS code,' and name the code
in upper case italics, or identify the code by description (e.g.'. . .in SAS
code MH_RP, or 'Run the SAS code that examines the sub_bank data
file.').
B.1.2 Structure of Data Collection
Every stream sampling reach is designated by a unique stream identification
number, and delineated with eleven equally spaced sampling cross sections called
TRANSECTS, which are labeled A through K. Habitat characteristics are measured
according to one of four sampling location layouts relative to the reach and its 11 transects
(Figure B-1):
B-2
-------
Cross-section Sampling
locations (TRANSDIRs)
LF LC CT RC
Transect C
Channel/Riparian
Cross section
Transect
(TRANSECTS)
Upstream end of
sampling reach /
Thalweg profile
intervals
(STA_NUMs)
Downstream end of
sampling reach
Figure B-1. Layout of sampling reach, showing sampling points across transects and thalweg
measurement points.
B-3
-------
Substrate and depth data (sub_bankf\\e) are collected at points called
TRANSDIRs, arranged along the channel cross-section at each transect, as
shown on the inset of Figure B-1.
Canopy densiometer measurements (canpycovfile) are made at TRANSDIR
positions LF and RT, and then in four directions from one mid-channel point,
designated CU, CD, CL, and CR (denoting the left and right bank
measurements and the four mid-channel measurements towards the
upstream, downstream, left hand, and right hand directions.
Fish cover is evaluated at 11 in-channel plots, each centered on a transect cross-
section (fishcovf\\e).
Visual riparian cover estimates and human disturbance tallies are evaluated on
streamside plots on the left and right banks, centered on each of the 11
transect cross-sections (riparian file).
Thalweg profile data are collected along the mid-channel length of the sampling
reach at points called STA_NUMs (station numbers), of which there are 10
or 15, equally spaced between each of the 11 transect cross sections
(thalweg f\\e).
B.1.3 Data File Structure
Data files are arranged as matrices, with observations running in rows, and variables
running in columns. Every observation is made up of individual cell entries containing data
points / values for each variable.
In addition to the field comments file, phabcom, there are six standard PHab data files
containing measurements and observations entered from the PHab field forms: canpycov,
fishcov, Igwoody, riparian, sub_bank and thalweg. The types of variables within these
files are listed in Table 4 of the main body of this report. The data files are organized by
individual stream sampling events. Each sampling event is identified by a unique set of
identification variables consisting of STRMJD, YEAR, VISIT_NO, DATE_COL and
TEAMJD. The number of observations within each sampling event is determined by what
measurements were made in the field, and at what physical location they were made within
the sampling reach. Variables in the PHab data files named TRANSECT, TRANSDIR and
STA_NUM are called sampling location variables, and designate the specific location where
a particular PHab measurement was made within the reach.
Every stream should have eleven TRANSECTS, labeled A through K. Within each
TRANSECT there are varying numbers of observations based on which measurements
were made in the field, and at what locations (TRANSDIRs or STA_NUMs) on the sampling
reach they were made.
B-4
-------
B.2 DATA VERIFICATION AND VALIDATION
B.2.1 Getting Started
We provide SAS Code files (Appendix C on the attached compact disk) to automate
the VerVal process as much as possible. However, the code we have included is designed
only as a structural guide for Verification and Validation of PHab data files. Do not rely
solely on the SAS code provided for assuring the quality of PHab data files. All
possibilities can not reasonably be expected to be captured by any fixed VerVal procedure.
Therefore keen observational skill and professional judgement will always be required to
determine if particular values are possible, or logical within each individual context. It is also
important to keep in mind that the code may need to be altered to accommodate differences
in data entry, variable formatting, file names, and sometimes the field sampling protocols
themselves. It is advisable to fix file format irregularities prior to using the metrics
calculation codes. These calculations may be quite complex, making the code difficult to
modify without introducing errors. File format irregularities can be fixed by modifying the
VerVal code or by writing new programs.
For each region and year:
1) Obtain all requisite data files for a particular region and year (canpycov, fishcov,
Igwoody, riparian, sub_bank, thalweg, phabcom).
2) Create a folder (directory) to hold the raw data for that region and year, and
place these seven data files in it. A copy of the verified data files should be
retained locally. One way to do this is by adding a "0" suffix to the file
names. Alternatively, these files could be kept in a separate directory.
3) Create working copies of each data file. Name them by adding the suffix 1, 2, 3,
etc., to the original file names, to designate successive versions created
during VerVal. File names usually must be limited to 8 digits; for example
we shortened names like riparian and canpycov to riparand canpy. To
indicate the year of study, we typically alter the seven file names by
shortening them to 6 digits, then adding the last two numbers of the study
year to make 8-digit file names. Place all successive working files in the
folder containing all data for the given region and year. It is important to
record in a log book or file all changes made when creating new or updated
data files.
4) Before beginning VerVal it is important to make sure that all PHab data files are
present for each region and year, that these files have the correct structure,
and that they contain the necessary identifying variables with the correct
attributes.
B-5
-------
B.2.2 Preliminary Structure and Variable Attribute Check
With all the requisite data files available, run the VerVal overview computer code
AARDVARK.sas, so named in order that it shows up first in an alphabetized directory listing,
and also because it performs a miscellaneous variety of tasks. The computer code is
thoroughly documented, so that a data manager can apply it and be clearly informed about
what the code is doing to the data set. This section of code examines the entire set of data
files checking for inconsistencies in variable naming and formatting, improper variable
names, improper variable types (numeric or character). These problems should be fixed
before proceeding. The code will add missing variables and set the variables values to their
expected length. However, note that text variable values may be shortened, possibly
truncating text data. Improper variable types are a condition which requires the individual
attention of the user, and thus are not automatically changed in the code. Variables which
are unexpected may exist in a data set either as a result of non-standard naming (e.g.,
STREAM instead of STRMJD), or may be supplemental data specific to an individual study.
In either case these conditions also require the attention of the user and are not deleted by
the program. AARDVARK.sas also checks sampling date (DATE_COL) values within data
files, noting irregularities in the output. It also creates a table to assist in checking for
inconsistencies among data files and for determining values for important, but often
unrecorded variables, such as REACHLEN and INCREMNT.
Among the formatting requirements of the metric calculation computer codes are the
following:
1) All STRMJD's should be for the correct region (e.g.. OR , MA , etc.), and in
the correct form. In most cases STR _ID should be a multi-character string
which uniquely identifies that a stream reach ('MA003S', 'ORC34', 'ORST97-038',
'00028S'). If STRMJD's are not in the right form, create a new version of data
file making the necessary corrections.
2) All dates should be for the correct year and in the correct form (MM/DD/YYYY).
3) Variables appearing in more than one data file should have the same attributes.
Variable name, type, and length should be the same in all data files for a given
region. All data files should have appropriate ID and sampling location variables.
ID variables may include STRMJD, VISIT_NO, YEAR, DATE_COL and
TEAMJD. Sampling location variables are TRANSECT and TRANSDIR or
TRANSECT and STA_NUM. (The thalweg data file has a sampling location
variable called STA_NUM, while the sub_bank data file has a corresponding
variable called TRANSDIR.). If variables are incorrectly named or missing, or if
B-6
-------
variable attributes are inconsistent among data files, create a new version of the
data file making the necessary corrections.
In addition to the general file formatting issues, there are segments of AARDVARK.sas that
merge data (e.g. bearing and slope data are merged from slopebrg into thalweg), and
segments of code that put specific files into the format expected by the metric calculation
codes that will later be applied to them. These segments may be run or modified as
needed. Once these changes have been made, the data overview program should be rerun
to check the changes made.
B.2.3 Verifying and Validating Individual Data Files
Many of the VerVal code files we provide are tailored to each of the PHab data files
(canpycov, fishcov, Igwoody, riparian, sub_bank, thalweg), and each bears the name of
the data file to which it pertains. To apply the code, the name of the data file to be checked
should be typed where indicated near the beginning of the program text. The code should
be run in sequence, one test section at a time. Before running each successive check,
correct deficiencies indicated by the previous section. In order to confirm that corrections
have been made, each section of code should be re-run on each data file after it is updated.
Each time changes are made to a PHab data file and an updated file version is
created, it is important to reopen the data file (making the appropriate changes in the file
name, if necessary), and re-run the previous DATA steps with the newly modified data
before continuing to the next section of that VerVal SAS code. Otherwise changes to the
data file will not be recognized by the code.
There are six general steps in verifying and validating PHab data files. It is vital that
these be done in the prescribed order, and that each step be completed before proceeding
to the next. The SAS code files included for VerVal of each PHab data file contain
numbered sections that correspond to these six sequential steps, but not all steps apply to
every data file. These six steps of verification/validation check the following aspects of the
data: data file structure, missing values, values outside of an allowable range, unusual
values, channel morphology visualizations, and internal logic among variables. These steps
are generally described in Sections B.2.3.1 through B.2.3.6. Because some cross-checks
are based on related variables among the PHAB data sets, we recommend validating the
sub_bank data first, followed by the thalweg data.
B.2.3.1 Data File Structure-
The first section of each VerVal code aids validating and correcting data file structure,
insuring that every stream visited has the correct ID number, the correct number of visits,
B-7
-------
transects, and observations within transects. This is a critical portion of VerVal, because
each final data file must have a standardized structure that reflects the field sampling design
in order for the habitat metric calculation code to run properly. Each data point or cell entry
in a data file should reflect an individual measurement or observation recorded in the field.
Each subsection of "Computer Code for VerVal of Individual Data Files" begins with a
brief description of the sampling design that gives rise to the structure of the data file.
Subsections then list the identification variables and briefly explain how they relate to the
data file structure. The identification variables (STRMJD, YEAR, VISIT_NO, DATE_COL,
TRANSECT, TRANSDIR, STA_NUM) should have exactly the number of cell entries listed
in the "Structure Check' subsections within the subsequent section entitled "Computer Code
for VerVal of Individual Data Files". Inequalities may indicate that data for a transect,
station, etc, were not recorded in the field, not entered during data entry, assigned an
incorrect identifying value, or that additional side channel observations were included in the
survey. Discrepancies in the number of identification variable values present should be
validated and corrected at this stage. The number of entries for the PHAB variables
themselves (e.g. DENSIOM, UNDERCUT, etc.) may vary somewhat due to missing values
in the data file. The number of non-missing values for these variables should equal, or be
slightly lower than the number listed number in the individual data file "Structure Check'
subsections. However, at the structure check step, it is not yet necessary to validate every
entry for each PHAB variable, but rather to ensure that the overall structure and number of
observations are correct.
B.2.3.2 Missing Values-
This section of code aids in identifying missing values. These should be verified by
cross-checking with the data forms. Values which are present on the data forms, but
missing from the data file, should be documented in the separate VerVal log book or file,
and added to the data file. Values which are missing from the data files, but are found to be
0 on the field forms should likewise be corrected. Large woody debris tally data are an
exception, as missing values indicate zero counts for a particular size/location class.
B.2.3.3 Values Outside Allowable Range-
This section of code lists observations in which variable values are outside of an
allowable range, (example: riparian vegetation cover class: 0 - 4, DENSIOM: 0 - 17). Illegal
values should be compared against the data forms. Values that cannot be reconciled
should be documented, and changed to "missing" in the revised data file.
B.2.3.4 Unusual Values-
This section of code uses summary statistics to examine the range of continuous
numerical data (example: DEPTH, WT_WID values). The unusual values identified by this
B-8
-------
examination should be verified by checking the field forms to ensure that they are not
transcription errors. Remaining unusual values should be evaluated using professional
judgement. They may be changed and documented, or left to stand accordingly. Impossible
values (e.g., DEPTH=700m) that cannot be reconciled should be documented, and changed
to missing values in a revised data file.
B.2.3.5 Channel Morphology-
This section of some of the VerVal codes graphs channel morphology values as
transect or stream profiles so that one can visually determine if some values stand out as
outliers, even though they may not be unusual relative to the whole data set. Values of
many channel morphology measures need to be evaluated in a reach-specific and cross
section-specific context. For example, a wetted width of 30 m may not stand out as unusual
in a large data set. But it would be very unusual in a stream with all other widths < 3 m.
Outliers should be verified and documented, and an attempt to reconcile them should be
made using best professional judgement. If this is not possible, a decision must be made to
remove outliers, or to leave them stand. Regardless which choice is made, proper
documentation is essential.
B. 2.3.6 Internal Logic-
This section of VerVal codes lists various instances where the data contains logical
inconsistencies. Some examples are:
-The sum of cover class values within a vegetation layer should not convert to a
cover percentage > 100% (e.g., class 3 + class 4 = 57.5% + 87.5% = 145%);
-Canopy type or densiometer cover indicate vegetation presence, but visual cover
estimates are zeros (e.g. CANV = D,C,E, or M; but BTRE and STRE both
equal 0 or missing).
Apparently illogical values that are identified and listed by the code should be checked
against field forms, documented, and where possible, professional judgement should be
used in reconciling values to specific guidelines listed in individual 'logic check sections of
this document. When apparently illogical values cannot be reconciled, decide whether to
replace the questionable values with missing values, or let them stand; then fully document
these changes.
B-9
-------
B.3 COMPUTER CODE FOR VER/VAL OF INDIVIDUAL DATA FILES
B.3.1 File canpycov Checks
Structure check:
Canopy cover is measured by taking a densiometer reading at six places,
(TRANSDIRS) spaced along, and around each of eleven transects (see figure 2-1),
resulting in 66 observations per stream visit.
Expected Frequencies of Variables:
STRMJD , VISIT_NO, DATE_COL, TRANSECT - 66 for each stream visit.
TRANSDIR - 6 per TRANSECT, one for each value CD, CL, CR, CU, LF and RT.
DENSIOM - Canopy cover measurement should have a maximum of 66 values per
stream visit.
Missing value check:
For DENSIOM - As described in Verifying and Validating Individual Data Files'
overview section.
Allowable range check:
DENSIOM should only have whole number values from 0 to 17.
B.3.2 File fishcovChecks
Structure check:
Areal cover available for fish from several different sources is estimated over a stream
plot centered on each transect, resulting in eleven observations per stream visit.
B-10
-------
Expected Frequencies of Variables:
STRMJD , VISIT_NO, DATE_COL, TRANSECT - 11 for each stream visit.
Fish cover estimates for the PHAB variables ALGAE, BOULDR, BRUSH, MACPHY,
STRUCT, OVRHNG, UNDCUT and WOODY each should have a maximum of 11
values per stream visit.
Missing value check:
For PHAB variables mentioned above: As described in 'Verifying and Validating
Individual Data files' overview section.
Allowable range check:
All fish cover variables should only have whole number values from 0 to 4.
B.3.3 File riparian Checks
Structure check:
At each transect, vegetation type and cover in three layers are estimated, and human
influences are recorded within a riparian plot on each stream bank, resulting in 22
observations per stream visit.
Expected Frequencies of Variables:
STRMJD, YEAR, VISIT_NO, DATE_COL, TRANSECT -- 22 for each stream visit.
TRANSDIR - Two entries for each transect, one each of value LF and RT.
Canopy and understory vegetation type (CANV, UNDV) and cover (BTRE, STRE,
WOOD, NONW, GCW, GCNW, and GCB), and human influences (WALL, BLDG,
PVMT, ROAD, PIPE, LDFL, PARK, CROP, PSTR, LOG and MINACT) should have a
maximum of 22 entries per stream visit.
Missing value check:
For PHAB variables mentioned above - As described in 'Verifying and Validating
Individual Data Files" overview section.
B-11
-------
Allowable range check:
CANV and UNDV should only have values D, C, E, M, or N. BTRE, STRE, WOOD,
NONW, GCW, GCNW and GCB should only have whole number values from 0 to 4.
All human influence variables should only have values of 0, B, C or P.
Logic checks for Cover Class Variables:
Within each of the three layers of riparian vegetation cover, values should not add to
more than the equivalent of 100%, assuming the mid-point % cover for each cover
class entry. For each of the vegetation-layer categories, areal cover was estimated in
four classes: absent (0), sparse (0-10%), moderate (10-40%), heavy (40%-75%) and
very heavy (>75%). The mid-points of these cover classes are, respectively: 0%, 5%,
25%, 53%, and 87.5%. Therefore, the following combinations are violations of this
logic:
1) If BTRE=4, then STRE cannot be 3 or 4, and vice-versa.
2) If WOOD=4, then NONW can not be 3 or 4, and vice-versa.
3) If any of GCW, GCNW or GCB=4, then none of the others can be 3 or 4.
Running section 6 of the SAS code for VerVal of PHAB data files lists violations of this
rule of logic. In these cases each value should be reduced by one. For example, if
BTRE and STRE both=4, then reduce both values by one. This will bring the equivalent
total below 100%. If BTRE=4 and STRE=3, or vice-versa, then reduce both values by
one. This will preserve the relationship of dominant and sub-dominant vegetation cover
values, while also reducing the equivalent vegetation layer total to below 100%
Logic Checks for Cover Type Variables:
If CANV indicates there is no cover in the canopy vegetation layer (i.e. CANV = N), and
there is a non-zero value for BTRE or STRE, then CANV should be changed to
missing.
If UNDV indicates there is no cover in the understory vegetation layer (i.e. UNDV= N),
and there is a non-zero value for WOOD or NONW, then UNDV should be changed to
missing.
If BTRE and STRE both have values of zero, and CANV is not equal to N, then CANV
should be changed to N.
B-12
-------
If WOOD and NONW both have values of zero, and UNDV is not equal to N, then
UNDV should be changed to N.
Comments Cross Check:
Occasionally, survey crews will list human influence factors in comments, but neglect
to include them in the riparian data section of the field form. The comments cross-
check section of code lists comments as well as all the human influence variables. If
any human influences are listed in the comments, but have no values in the PHab
variables, then document them, and add them as 'P' to a revised data file.
B.3.4 File sub_bank Checks
Structure Check:
Substrate information is measured at each of five points (TRANSDIRs) across each
TRANSECT, resulting in 55 observations per stream visit. Bank characteristics are
measured at both stream banks, at each TRANSECT, resulting in 22 values per
stream visit. Width and bankfull characteristics are recorded once at each transect,
resulting in eleven values for each stream visit.
Expected Frequencies of Variables:
STRMJD, YEAR, VISIT_NO, DATE_COL, TRANSECT - 55 for each stream visit.
TRANSDIR - There should be five entries for each TRANSECT, one each of value LF,
CL, CT, CR and RT.
Distance to the left bank (DIST_LB), water depth (DEPTH), substrate size class
(SIZE_CLS) and embeddedness (EMBED) should have one entry per TRANSDIR for
a total of 5 per transect and 55 per stream visit.
Bank angle (ANGLE) and undercut (UNDERCUT) should only have entries at
TRANSDIRs LF and RT for each TRANSECT, totaling a maximum of 22 entries per
stream visit. Wetted width (WT_WID), bankfull width (BANKWID) and height
(BANKHT), stream incision (INCISED) and bar width (BARWID) should have one entry
per TRANSECT, totaling a maximum of 11 entries per stream visit. These values
should be listed atTRANSDIR=RT.
B-13
-------
Missing Value Check:
UNDERCUT- If an UNDERCUT value is missing, it should be compared with the
UNDCUT value in the fishcovdata file. If the UNDCUT value is 0, then the missing
UNDERCUT (sub_bank value) should be changed to 0. If the UNDCUT value in
fishcov\s not 0, then the missing sub_bank value must be left as missing.
BARWID - If a BARWID value is missing, it should be compared with the BARWID
value in the thalweg data file for the same TRANSECT at STA_NUM=0. If there is a
BARWID value at this location in the thalweg data, then the missing BARWID value in
the sub_bankf\\e may be replaced with that value.
For other PHAB variables: As described in 'Verifying and Validating Individual Data
Files' overview section.
Allowable Range Check:
SIZE_CLS can have values of RS, RR, RC, HP, BL, CB, GC, GF, SA, WD, FN, OT, or
missing.
EMBED can have values of 0 to 100, inclusive.
ANGLE can have values from 0 to 180.
Unusual Value Check:
UNDERCUT values greater than 1 meter should be considered very unusual.
INCISION values greater than BANKHT + 5m should be considered very unusual.
BARWID values should be generally small (0 to 6, but always less than WT_WID).
Channel Morphology Checks:
This section graphs channel cross section and longitudinal profile plots for DIST_LB,
DEPTH, WT_WID, BANKWID and BANKHT by stream visit. These plots should be
examined for outliers, values that do not fall within a reasonable range for each individual
stream.
B-14
-------
Logic Checks:
Distance From Left Bank - DIST_LB values should be 0 (or very close to 0) when
TRANSDIR=LF, then it should increase successively by 25% of the wetted width as
measurement locations progress across the channel (TRANSDIR positions CL, CT,
CR), until at TRANSDIR=RT, the value of DIST_LB=WET_WID. The first part of
section 6 of the SAS code for VerVal of sub_bank data lists TRANSECTS where
DIST_LB is not 0 at TRANSDIR=LF and/or where DIST_LB does not equal WT_WID
at TRANSDIR=RT. The second part of this section of code identifies transects with
irregular TRANSDIR spacings.
Incision, Bankfull Height and Wetted Width, Bankfull Width - BANKWID should be >
WT_WID. INCISED should be > BANKHT. The VerVal code lists data that are
counter to these rules. Occasionally field crews incorrectly record values of INCISED
that are measurements of incision above bankfull height rather than from the water
surface. In this case INCISED should be changed to the value of BANKHT plus
INCISED. INCISED values are also sometimes mistakenly switched with BANKHT
values.
Wetted Width, Depth - The VerVal code lists TRANSECTS where WT_WID =0 (most
probably signifying a dry channel), but where a DEPTH value other than 0 is entered.
Validation can be performed by cross checking with thalweg data. If CHAN UN IT =
DR at STA_NUM=0 (thalweg data file), then WT_WID and DEPTH values (sub_bank)
should both = 0. If thalweg data indicates that there was water flow at STA_NUM=0
for that TRANSECT then appropriate adjustments to sub_bank data must be made.
Bar Width, Wetted Width - In the EMAP field methods, a mid-channel bar is an in-
channel feature. Therefore, BARWID cannot equal or exceed WT_WID; the VerVal
code lists violations of this logic.
Comments Check- This section of VerVal code lists comments if SIZE_CL =OT,
signifying "other". Field surveyors are instructed to elaborate on atypical substrates in
comments. If 'hardpan' or 'concrete' are listed, then SIZE_CL should be changed to
'HP' or'RC', respectively.
By definition, bedrock and hardpan substrate size classes are not embedded (if
SIZE_CLS=RR, RS, HP or RC then EMBED=0). Conversely, sand or smaller
substrate size classes are by definition completely embedded (if SIZE_CLS=SA or FN
then EMBED=100). These corrections should be made universally by running the last
B-15
-------
section of the VerVal code for sub_bank after the rest of the VerVal code for this data
file has been completed.
B.3.5 File thalweg Checks
Structure Check:
Thalweg characteristics are measured at either ten or fifteen stations (depending on
the width of the stream) equally spaced as STA_NUM=0 through 9 or 14 between each
TRANSECT, A through K. There is only one station at the location of transect K.
Stream width is recorded once at each transect (STA_NUM=0) A through K, and at the
middle position (either STA_NUM=5 or 7) of each transect A through J. Stream slope
and compass bearing backsights are recorded once at each TRANSECT from B to K.
Slope and bearing data are originally entered in a file called slopebrg, which is
merged with thalweg data by the VerVal code AARDVARK.sas.
Expected Frequencies of Variables:
STRMJD, YEAR, VISIT_NO, DATE_COL, TRANSECT -- either 101 or 151 for each
stream visit. Data for STA_NUM '0' of TRANSECT 'K' are often not collected, so
thalweg depth values typically number 100 or 150 per stream visit.
STA_NUM - 10 or 15 for each of the ten TRANSECTS A through J, and one for
TRANSECT=K.
Thalweg depth (DEPTH), presence of fine sediments (SEDIMENT), habitat unit
(CHANUNIT), and pool forming agent (POOLFORM) should normally have either 10 or
15 values, one per STA_NUM, for TRANSECTS A through J, and one or zero for
TRANSECT K, totalling a maximum of 100, 101, 150, or 151 entries per stream visit.
Wetted width (WT_WID), and bar width (BARWID) should have a maximum of 20 or
21 entries per stream visit, with 2 entries (at station numbers 1 and either 5 or 7) at
each transect A through J, and 1 or zero at transect K. The primary backsighted
values for stream gradient (SLOPE), and bearing (BEARING) should have 10 entries
per stream visit. These values should be listed at STA_NUM=0 for each transect B
through K. Supplemental slope and bearing measurements may have 10 or fewer
entries per stream visit. These supplemental measurement variables include SLOPE2,
SLOPES, BEAR2, BEARS, PROPORT2, PROPORT3, and in older data sets:
SUPSLOPE and SUPBEAR.
B-16
-------
Missing Value Check:
In order to make this stage of data checking easier, the SAS code we break this
section into several parts to handle groups of variables differently:
1) WT_WID, SLOPE, BEARING, PROPORTN, INCREMNT, REACHLEN, BARWID,
SIDECHAN, CHANUNIT and POOLFORM: If WT_WID is missing and
STA_NUM is 0, then replace WT_WID with the value of WT_WID from sub_bank
data for the same transect. For other PHab variables, procedures are as
described in the "Verifying and Validating Individual Data Files" overview section.
2) There is a separate section for the variable DEPTH. This variable is used in
calculating residual pool statistics, which are degraded by missing values. (Field
crews should be encouraged to make their best estimate for points where depth
cannot be measured directly, and to flag these entries accordingly.) If a large
number of depth values are missing, or there are several gaps in one reach, then
it may be impossible to calculate residual pool statistics. For this reason extra
care should be taken in verification, including the following suggestions:
If one depth value is missing, and there are values for
depth at the adjacent station numbers, then the
missing value may be replaced by interpolating
between the two adjacent values.
If there are several depth values missing in a row it may be
possible to evaluate residual pools based on an abbreviated
reach segment. This will entail adjusting reach lengths in the
residual pool statistic calculations.
3) This section of VerVal code lists missing fine substrate presence (SEDIMENT)
values for those thalweg stations at the first STA_NUM of a transect separately
from those between transects. This is because missing SEDIMENT values at the
transect cross-section location (STA_NUM=0) can be inferred from SIZE_CL
substrate data in the sub_bank data file:
If the observation at the first station of a transect is
missing, then check the sub_bank data file
value for SIZE_CL at the deepest TRANSDIR
from the same TRANSECT. If this value is
GF, SA or FN, then the missing value should
be replaced with 'Y'. Otherwise the missing
value should be replaced with 'N'.
B-17
-------
Missing SEDIMENT values between transects are
checked as described in Section B.2.3 of this
Appendix: "Verifying and Validating Individual
Data Files".
Allowable Range and Unusual Value Checks:
SEDIMENT should only have values of Y or N; but field crews frequently indicate fine
sediment presence with Y and absence by a missing value.
CHANUNIT should only have values of PP, PT, PL, PB, PD, GL, Rl, RA, CA, FA or
DR.
POOLFORM should have values of N, W, R, B, F, or O, or a combination of these.
SIDECHAN can have values of either Y or N. If there is a missing value, it should be
changed to N.
SLOPE, or any supplemental slope values greater than 20% are rare. Values between
0 and 0.1% should be examined to see that they were not recorded as 1/100th of the
true slope.
BEARING, or any supplemental bearings should only have values from 0 to 359.
PROPORTN, PROPORT2, AND PROPORT3 should have values from 0 to 100%, and
should sum to 100%.
BARWID values should be generally small (0 to 6, but always less than WT_WID).
Channel Morphology Check:
This section of code produces, for each stream visit, a longitudinal profile of wetted
width for the reach, and depth profiles for each of the reach segments between
transects. These plots should be examined for obvious outliers that fall outside a
reasonable range of variation within each individual stream and visit.
Logic Checks:
WT_WID, DEPTH, CHANUNIT - This section lists transects where WT_WID, DEPTH
or CHANUNIT suggest a dry channel, but one or more of the other variables do not
agree, (e.g., where CHANUNIT = DR, but DEPTH or WT_WID has a non-zero value).
Validation can be aided by cross-checking with sub_bank data.
B-18
-------
Bar Width, Wetted Width - The EMAP field methods define mid-channel bars as
in-channel features. Therefore BARWID cannot be > WT_WID. This section lists
violations of this relationship.
Slope and Bearing - If no SLOPE, BEARING and/or PROPORTN values are listed, but
there are (supplemental) values for SLOPE2, SLOPES, BEAR2, BEARS, PROPORT2,
or PROPORT3, the missing primary value should be replaced with the supplementary
value, and the supplementary value removed.
Increment and Reach Length - If each transect has 10 stations, then INCREMNT
should be equal to, or approximately equal to REACHLEN •*• 100. If each TRANSECT
has 15 stations, then INCREMNT should be equal to, or approximately equal to
REACHLEN * 150.
Shift in Slope and Bearing Recordings -- If TRANSECT A has a value listed for SLOPE
or BEARING, but TRANSECT K does not, then check if the slope and/or bearing
recordings are all shifted by one TRANSECT (i.e., backsight at B was recorded at A,
etc.). There should be one value listed for each TRANSECT from B to K, all located at
STA_NUM=0.
Width to Depth Ratio - This section lists transects where (WT_WID/DEPTH) is > 50, or
< 1. These values are suspect, and should be verified with sub_bank channel cross-
section data.
B.3.6 F\\e Igwoody Checks
Structure Check:
Pieces of wood greater than 0.1 m in diameter and 1.5m in length are counted and
recorded between every two transects, separated into 12 size classes each for pieces
at least partially wet during bankful flows (PIEC_TYP='WET') and those within bankful
width, but above bankful flows (PIEC_TYP='DRY'). The Igwoody data file may not
contain observations for size classes in which LWD was not recorded. In this case,
check field forms to determine when it is safe to assume that missing observations in
any given size class are meant to be zeros. In these cases there may be significantly
less than 240 observations per reach, but the metric calculation SAS code will run
correctly regardless of missing observations.
B-19
-------
Expected Frequencies of Variables:
STRMJD, VISIT_NO, DATE_COL, TRANSECT - 240 per stream visit. There may be
less than this number if zeros were omitted when no LWD was observed.
PIEC_LEN, PIEC_DIA and PIECES should all have 24 entries per transect totaling 240
entries per stream visit.
PIEC_TYP should have 12 values of 'WET' and 12 values of 'DRY' for each of the
eleven TRANSECTS, totalling 240 entries for each stream visit
Missing Value Check:
For PHab variables mentioned above - As described in "Verifying and Validating
Individual Data Files" overview section.
Unusual Value Check:
Extremely large numbers, or numbers which are disproportionately large for an
individual stream reach should be checked against the field forms.
Data Restructuring:
This part of the VerVal code restructures the Igwoody data file to facilitate calculating
reach level metrics. The resulting data file should be saved as the final Igwood* data
file.
B.3.7 Computer Code Involving More Than One Data File
There are several instances where values resulting from the same or similar
measurements appear as the same or different variables, but in two different PHab data
files. The VerVal code we provide will compare these values and generate a list of
observations that do not agree. Values listed should be verified by examining the original
data forms. Instances where values are found to be inconsistent should be reconciled and
documented.
FISHSUB - This code compares undercut measurements in the sub_bank data file
(UNDERCUT), with estimates of the areal cover of undercut from fishcov data
(UNDCUT). It is logically possible for fishcov to have a positive value for UNDCUT,
with sub_bank having a value of 0 for UNDERCUT. However if there is a positive
B-20
-------
value for UNDERCUT (sub_bank), and UNDCUT (fishcov) has a value of 0, then
UNDCUT should be changed to 1, indicating that there is at least some bank undercut
within the fishcov plot.
THALSUB - The first section of this code lists observations where thalweg values for
WT_WID and/or BARWID do not match those for sub_bank. The second part of this
code checks for situations where either thalweg or sub_bank data suggest a dry
stream channel, but there are depth and/or wetted width values, and/or the CHANUNIT
value is not DR.
LABELS - When all PHab data files have been thoroughly validated, verified and
documented, variables should be labeled. At this point the sequential number of each
final PHab data file should be added to the macro data lines at the end of the included
LABELS SAS code. Running this code will automatically label all variables in the
canpycov, fishcov, riparian, sub_bank, thalweg, and Igwoody data files.
B.4 FINAL DATA VALIDATION CONSIDERATIONS
A final level of quality assurance is accomplished by examining reach level metrics
after they are calculated from the raw habitat measurement files (see Section 3 of main
body of this report). Some errors within the various PHab data sets can be seen quite
obviously from this overall vantage point. These types of checks include the following:
Check metric variables that indicate how many values were used to calculate
metrics (e.g. NBNK, N_BA, NSLP). Values deviating from the expected numbers
may indicate data structure problems; if there are excessive numbers of missing
values, a reach level metric value based on the remaining observations may be
suspect.
Particular ecoregions have predictable expectations for some variables,
particularly for canopy and mid-layer cover types (for example, one would not
expect high cover values for coniferous vegetation in the Great Plains region).
Cases where data is contrary to expectations may indicate incorrect values in the
physical habitat data file riparian. Similarly, metric values that are an order of
magnitude higher or lower than the values from all other stream reaches in an
ecoregion, or that look incongruent when compared to others, should be double-
checked.
If a metric value for a reach is equal to the minimum or maximum possible for the
associated raw measurement variable, (e.g., if XCDENMID=0, or
B-21
-------
XFC_ALG=1.00) then every observation for that variable in the raw physical
habitat measurement data file would have to be at the minimum or maximum
value (e.g., DENSIOM=0 and ALGAE=4) for the stream in question.
If a metric variable indicating variance (e.g., VSLOPE, SDDEPTH ) is equal to 0 in
a particular stream reach, then all raw data values must be equal to the reach
mean value (e.g., XSLOPE, XDEPTH) for all observations of the associated
measurement variable in that stream (e.g., SLOPE and DEPTH in file thalweg).
To check sinuosity metrics, compare REACH and REACHLEN. If TRAN_N=10 (all
bearing values were present and used in calculating sinuosity), then REACH should equal
REACHLEN. If this is not the case, verify the number of bearing values used to calculate
sinuosity for the stream visit in question.
Finally, errors are inevitably revealed during data analysis, as inconsistencies may
be discovered in relationships such as those between slope and substrate size, or between
depth variance and residual pool area. We hope that this guide helps to identify most errors
and inconsistencies.
B-22
-------
APPENDIX C. COMPUTER CODE FOR DATA VERIFICATION AND VALIDATION
Programs and thorough program documentation to assist with data verification and
validation of EMAP Physical Habitat data (described in Appendix B) are included on the
enclosed compact disk. They were developed for use with the Statistical Analysis System
(SAS; Version 6.12 for Windows) software. These programs are formatted as plain ASCII
text files, and so can be read and printed from any word processor. In SAS, a file is
retrieved into the Program Editor window, modified as necessary, and submitted to run it.
These programs and their updates may also be available in the future from the EMAP
website (http://www.epa.gov/emap)
C-1
-------
APPENDIX D. COMPUTER CODE FOR PHYSICAL HABITAT METRIC CALCULATION
Programs and thorough program documentation to calculate the various metric
variables from the individual measurement EMAP Physical Habitat variables (described in
Section 4 of the main report) are included on the enclosed compact disk. They were
developed for use with the Statistical Analysis System (SAS; Version 6.12 for Windows)
software. These programs are formatted as plain ASCII text files, and so can be read and
printed from any word processor. In SAS, a file is retrieved into the Program Editor window,
modified as necessary, and submitted to run it. These programs and their updates may
also be available in the future from the EMAP website (http://www.epa.gov/emap).
D-1
-------
APPENDIX E. ILLUSTRATION OF RAW DATA AND COMPLETED METRIC
CALCULATIONS FOR SEVERAL STREAM REACHES
Included on the enclosed compact disk are files that provide an example raw data
collected from several stream reaches, and the metric variables calculated using procedures
described in this report. These data files are formatted as comma-delimited ASCII files and
as SAS export files; we have included a program to convert to local SAS format. These
files may also be available in the future from the EMAP website (http://www.epa.gov/emap)
E-1
------- |