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

-------