Data supplement to EPA 840-B-21008
Development and Evaluation of the
Beta Streamflow Duration
Assessment Method (SDAM) for the
Western Mountains (WM)
May 2022
Report EPA 840-R-22002
-------
Development and Evaluation of the Beta
Streamflow Duration Assessment Method
for the Western Mountains
Data supplement
Prepared by Raphael D. Mazor. Southern California Coastal Water Research Project. Costa
Mesa, CA 92626
In collaboration with the U.S. Environmental Protection Agency's Streamflow Duration
Assessment Method Project Delivery Team:
Ken Fritz
Office of Research and Development
Cincinnati, OH 45268
Tracie-Lynn Nadeau
Office of Wetlands, Oceans, and Watersheds
Portland, OR 97205
Brain Topping
Office of Wetlands, Oceans, and Watersheds
Washington, DC 20004
Julie Kelso, ORISE Fellow
Office of Wetlands, Oceans, and Watersheds
Washington, DC 20004
This document has been reviewed in accordance with U.S. Environmental Protection Agency
policy and approved for publication. Any mention of trade names, manufacturers or products
does not imply an endorsement by the United States Government or the U.S. Environmental
Protection Agency. EPA and its employees do not endorse any commercial products, services,
or enterprises. Funding was provided under contract EP-C-17-001 for data management and
analysis and EP-C-16-006 for data collection. The views expressed in this report are those of the
authors and do not necessarily represent the views or policies of the U.S. Environmental
Protection Agency.
Suggested citation:
Mazor, R.D., Fritz, K.M., Topping, B., Nadeau, T.-L., and Kelso, J. 2022. Development and
Evaluation of the Beta Streamflow Duration Assessment Method for the Western Mountains.
Document No. EPA 840-R-22002.
2
-------
Introduction
Streamflow duration assessment methods (SDAMs) are rapid, field-based methods to determine flow
duration class at the reach scale. The conceptual framework and process steps presented by Fritz and
others (2020) were followed to integrate the three key components of an SDAM development study
(hydrological data, indicators, and study reaches) and develop a beta SDAM for the Western Mountains
(WM; Mazoretal. 2021c).
This supplemental document describes the data collection, data analysis, and evaluation steps
that resulted in the beta SDAM WM. The SDAM Project Delivery Team is making this document
available to inform public review and comment on the beta method. For a complete description
of the beta SDAM WM protocol, please see the User Manual (Mazor et al. 2021c). For more
information on the collaborative effort between the U.S. Environmental Protection Agency
(EPA) and the U.S. Army Corps of Engineers (Corps) to develop regional SDAMs for nationwide
coverage, please see here.
Streamflow duration classes
Streamflow duration governs important ecosystem functions (such as support for aquatic life,
sediment transport, and biogeochemical processing rates) and streamflow duration classes are
often used to guide watershed management decisions, including assessing the applicability of
water quality standards. Our definitions of streamflow duration classes followed those used by
Nadeau (2015):
Ephemeral reaches flow only in direct response to precipitation. Water typically flows
only during and/or shortly after large precipitation events, the streambed is always
above the water table, and stormwater runoff is the primary water source.
Intermittent reaches contain sustained flowing water for only part of the year, typically
during the wet season, where the streambed may be below the water table or where
the snowmelt from surrounding uplands provides sustained flow. The flow may vary
greatly with stormwater runoff.
Perennial reaches contain flowing water continuously during a year of normal rainfall,
often with the streambed located below the water table for most of the year.
Groundwater typically supplies the baseflow for perennial reaches, but the baseflow
may also be supplemented by stormwater runoff or snowmelt.
For these definitions, a reach is a section of stream or river along which similar hydrologic
conditions exist (e.g., discharge, depth, velocity, or sediment transport dynamics) and
consistent drivers of hydrology are evident (e.g., slope, substrate, geomorphology, or
confinement). A channel is an area that is confined by banks and a bed and contains flowing
water (continuously or not).
3
-------
Overview of the beta method for the Western Mountains
The beta SDAM for the WM uses a small number of indicators to predict the streamflow
duration class of stream reaches in the WM. Some indicators are measured through desktop
analysis, while others are quantified during a single field visit. The beta SDAM WM results in
one of four possible classifications: ephemeral, intermittent, perennial, and at least
intermittent. The at least intermittent category occurs when an intermittent or perennial
classification cannot be made with high confidence, but an ephemeral classification can be
ruled out.
The tool uses a machine learning model known as random forest. Random forest models are
increasingly common in the environmental sciences because of their superior performance in
handling complex relationships among indicators used to predict classifications. We previously
used this approach to develop regional SDAMs for the Arid West (AW; Mazor et al. 2021a,
2021b) and Pacific Northwest (PNW; Nadeau et al. 2015, Nadeau 2015). Because the beta
method for the WM includes continuous indicators, the random forest model was not able to
be simplified into a decision tree or table, as was done with the beta SDAM AW (Mazor et al.
2021b) and SDAM PNW (Nadeau et al. 2015). Consequently, the random forest model for the
beta SDAM WM requires specialized software to run, so we developed an online open-access,
user-friendly web application to facilitate efficient and consistent use of the beta SDAM WM
protocol for those that do not have access to specialized software.
The degree of snow influence at an assessment reach was used to stratify the WM region
(snow-influenced and non-snow influenced areas) because persistent snow can be an
important water source affecting flow duration in streams. Snow influence is measured as
the mean snow persistence within a 10-km radius of the assessment reach (Hammond et al.
2017). Snow persistence is the fraction of time that snow is present on the ground between
January 1 and July 3; for the beta SDAM WM, snow persistence is calculated as the average of
the years between 2000 and 2020. Assessment reaches where the mean snow persistence is
greater than 25% are classified as snow-influenced, as this threshold differentiates areas where
snow is minimal from areas where snow is intermittent, transitional, or persistent (Hammond
et al. 2018). Although climate change and annual variation may change the degree of snow
influence affecting a reach in any given year, the stratification for this beta method is based on
a fixed 21-year time period that should be robust to short-term changes in climate. Snow-
influenced areas are prevalent in the Rocky Mountains, as well as at higher elevations in
Arizona and the Sierra Nevada of California. Non-snow influenced areas are prevalent in the
coastal mountains and valleys of northern California, the Sierra Nevada Foothills, and the
mountains of southern New Mexico, but they are also found throughout other regions of the
WM (Figure 1).
4
-------
Snow persistence
<25 (minimal snow)
25-50 (intermittent snow) - *
50-75 (transitional snow)
>75 (persistent snow)
Figure 1. Average snow persistence in the western United States. Data accessed from Hammond et al. (2017). Snow-influenced
areas are defined as those with mean snow persistence greater than 25 (i.e., on average, snow is on the ground more than 25%
of the time between January 1 and July 3). Portions of the west outside the WM region are presented with a gray overlay.
Methods and Results
Study area
The WM encompasses nearly 1 million km2 in the western United States, covering portions of
twelve western states. The region is defined by a combination of variables related to climatic,
landcover, vegetation, and soil conditions; for purposes of the current study, portions of the
WM region that overlap with the states of Washington, Oregon, and Idaho were excluded
(Figure 2; U.S. Army Corps of Engineers 2010). The WM includes low-elevation temperate
rainforests along the coast that rarely freeze, although much of the region is characterized by
high-elevation snow-dominated mountain ranges, including the Sierra Nevada, Rocky
Mountains, and Cascades. Typical vegetation is coniferous forests, although higher elevations
are characterized by grassland and tundra. Total annual rainfall typically exceeds 20 inches.
Ephemeral and intermittent reaches may be found at any position within a watershed but are
more common in smaller headwaters, where flow accumulation is insufficient to sustain longer-
duration flows.
5
-------
Although few large cities are found within the WM, several growing metropolitan areas are
found in bordering portions of the AW and Great Plains, such as Denver, Reno, and Salt Lake
City. Thus, the need for an SDAM in permitting and management programs is high in this
region. Within the WM, at least two SDAMs are currently in use but are applicable to only
specific geographic areas: the Pacific Northwest (PNW) method (Nadeau 2015), and the New
Mexico (NM) method (New Mexico Environment Department (NMED) 2011). However, prior to
the current study, the rest of the region lacked any tool to classify streamflow duration. Our
effort focused on the portion of the WM outside the PNW (Figure 2).
i-region
California and Nevada
Northern Rockies
Central Rockies
Southern Rockies
Figure 2. Sub-regions of the WM.
This method applies to WM region of the United States as defined in the National Wetland
Plant list (U.S. Army Corps of Engineers 2010, Lichvar et al. 2016), excluding the WM region that
overlaps with the states of Washington, Oregon, and Idaho. For reaches near regional borders
or for reaches in atypical (e.g., arid) conditions within the WM, consult the Western Mountains
regional supplement (U.S. Army Corps of Engineers 2010) to determine whether this method is
appropriate.
Development of the Beta SDAM WM
To develop this method, the steps described in Fritz et al. (2020) were followed, as detailed
below.
Preparation
At the outset of the project, we assembled a regional steering committee (RSC) consisting of
technical staff at Corps Districts and EPA Regional Offices in the WM region that manage
programs where streamflow duration information is often needed (e.g., Clean Water Act
programs, including permits and enforcement). RSC members were selected based on their
expertise in both scientific and programmatic elements relevant to streamflow duration
classification needs. The RSC served several functions in the development process, such as
reviewing technical products, facilitating connections with local experts, and identifying
resources such as sources of hydrologic data.
Sut
6
-------
We identified candidate indicators that were supported by the scientific literature (reviewed in
(Mazor and McCune 2021) or used in existing SDAMs developed for portions of the WM;
specifically, the New Mexico SDAM (NM method; (NMED 2011), and the SDAM PNW (PNW
method; (Nadeau 2015). Following input from the RSC, these candidate indicators were then
screened using the criteria described by Fritz and others (2020), including:
Consistency: Does the indicator consistently discriminate among flow duration classes
(e.g., demonstrated in multiple studies)?
Repeatability: Can different practitioners take similar measurements, given sufficient
training and standardization?
Defensibility: Does the indicator have a rational mechanistic relationship with flow
duration, as either a response or a driver?
Rapidness: Can the indicator be measured during a one-day reach-visit (even if
subsequent lab analyses are required)?
Objectivity: Does the indicator rely on objective (often quantitative) measures, as
opposed to subjective judgments of practitioners?
Robustness: Does human activity complicate indicator measurement or interpretation
(e.g., poor water quality may affect the expression of some biological indicators)?
Practicality: Can practitioners realistically sample the indicator with typical capacity,
skills, and resources?
Candidate indicators were included in the study (Table 1) if they met all of the above criteria or
were included in the NM or PNW SDAMs to facilitate comparison across the methods (McCune
and Mazor 2019).
Identify candidate reaches
We had two objectives in selecting candidate reaches for the WM region covered by this study:
first, to include a sufficient number of reaches in each streamflow duration class to characterize
variability in indicator measurements; and second, to select reaches representing the range of
key natural and disturbance gradients within the region to aid applicability of the method in
anticipated conditions across the WM region. To support our goal of geographic
representativeness, we established four sub-regional strata in the WM (Figure 2): one stratum
for California and Nevada (comprising both the cold Sierra Nevada mountains, and the warmer
North Coast of California) and one each for the Southern, Central, and Northern Rocky
Mountains. We aimed to select 150 publicly accessible stream-reaches (one assessed location
per reach) with equal representation of perennial, intermittent, and ephemeral flow duration
classes among and within the four WM sub-regions.
7
-------
Table 1. Candidate indicators evaluated in the present study. Indicators with "NM" in the Origin column were measured
following the NM method protocol (NMED 2011) and indicators marked with "PNW" were measured following the PNW
protocol (Nadeau 2015); other indicators (OTH) were measured with protocols developed for this study (available here) and
come from sources reviewed in a study by Mazor and McCune (2021) or recommendations from the BSC. Asterisks (*) indicate
hydrologic indicators that are considered direct measures of water presence.
Candidate indicator
Geomorphic indicators
Sinuosity
Bankfull width
Floodplain channel
dimensions
Particle size/stream
substrate sorting
In-channel
structure/riffle pool
sequence
Sediment
deposition on
plants and debris
Hydrologic indicators
Surface and
subsurface flow*
Isolated pools*
Water in channel*
Seeps and springs*
Hydric soils
Soil moisture and
texture*
Woody jams
Biological indicators
Live and dead algal
cover
Filamentous algal
abundance
Stream shading
Description Origin
Visual estimate of the curviness of the stream NM
channel
Width of the channel at bankfull height PNW
Visual estimate of the extent of channel NM
entrenchment and connectivity to the
floodplain
Visual estimate of the extent of evidence of NM
substrate sorting within the channel
Visual estimate of the diversity and NM
distinctiveness of riffles, pools, and other flow-
based microhabitats
Visual estimate of the extent of evidence of NM
sediment deposition on plants and on debris
within the floodplain
Estimate of the percent of the reach-length PNW
with surface and subsurface flow
Number of pools in the channel without any PNW
connection to flowing surface water
Visual estimate of the extent of surface flow in NM
the channel
Presence/absence of springs or seeps within NM
one-half channel width of the channel
Presence/absence of hydric soils within the NM
channel, measured at up to three locations
Extent of soil saturation and texture measured OTH
at three locations in the channel
Number of woody jams within the channel OTH
Visual estimate of the percent of streambed OTH
covered by live or dead algal growth
Estimate of the overall abundance of NM
filamentous algae within the channel
Percent shade-providing cover above the OTH
streambed measured with a densiometer at
three locations
8
-------
Candidate indicator
Hydrophytic plant
species
Fish
Aquatic
invertebrates
Aquatic
invertebrates
Amphibians
Mosses and
liverworts
Differences in
vegetation (riparian
corridor)
Absence of upland
rooted plants in the
streambed
Presence of iron-
oxidizing fungi or
bacteria
Presence of aquatic
or semi-aquatic
snakes
Geospatial indicators
Location and
watershed
characteristics
Long-term normal
precipitation and
temperature
Long-term mean
snow persistence
between January 1
and July 3
Description Origin
Number of obligate (OBL) or facultative wet PNW
(FACW)-rated plants (as listed in Lichvar et al.
2016) growing within the channel or a half-
channel width from the channel
Estimate of the overall abundance offish (other NM
than non-native mosquitofish) in the channel
Abundance and richness of aquatic invertebrate PNW
families collected from the channel
Estimate of the overall abundance of aquatic NM
invertebrates within the channel
Estimate of the overall abundance of NM
amphibians within the channel
Visual estimate of the percent of streambed OTH
and banks covered by live or dead bryophytes
or liverworts
Visual estimate of the distinctiveness of NM
vegetation in the riparian corridor compared to
surrounding upland vegetation
Visual estimate of the extent of upland rooted NM
plants growing within the streambed
Presence of oily sheens indicative of iron- NM
oxidizing fungi or bacteria within the
assessment reach
Presence of aquatic or semi-aquatic snakes PNW
(e.g., most garter snake species) in the channel
Latitude, longitude, and elevation OTH
30-y normal mean annual and monthly OTH
precipitation and 30-y normal mean, maximum,
and minimum annual temperature (PRISM
climate data; Hart and Bell 2015).
Snow persistence (Hammond et al. 2017) OTH
9
-------
Yes
Insufficient
record
r i
Zyear< 37
No
r
No
r
No
J
r
Unclassified
Yes
Perennial
Yes
Ephemeral
Yes
Intermittent
Figure 3. Flowchart used to classify reaches based on continuous measures of water presence (e. g., USGS stream gages). DOR:
days of record. Zyear: Average number of dry days per year. Myear: Average length of longest continuous wet period per year, in
days. For USGS gages, at least 20 years of data were analyzed whenever possible.
To screen reaches for use in method development, we first compiled a list of 1166 candidate
study reaches based on existing hydrologic data records (e.g., U.S. Geological Survey (USGS)
stream gages, water presence logger, wildlife cameras, field photos), published studies, and
interviews with local experts familiar with the specific reach's hydrology. Most of these reaches
(858) were derived from the database of gages operated by the USGS and nearly all of them
were perennial (as determined by applying the flowchart in Figure 3). Consequently, other
sources were required to identify candidate ephemeral and intermittent reaches. Hydrologic
data collected for other purposes (e.g., gages maintained by local flood control agencies, or
local natural resource managers) provided another 239 reaches. Published studies and public
land management plans yielded 49 candidate reaches and consultation with local experts
provided another 30. Whenever possible, multiple sources of hydrologic information were used
to confirm classifications. In the resulting set of reaches, 9.6% were determined to be
ephemeral, 15.6% were intermittent, and 74.7% were perennial.
Classified reaches were prioritized for study inclusion based on the number and type of data
sources available to determine actual streamflow duration classification. Reaches where flow
duration could be determined based on multiple data sources (e.g., water presence loggers and
expert knowledge) were categorized as "preferred" for study inclusion. Reaches classified
based solely on interpretation of USGS stream gage data without consultation of a local expert
were categorized as "USGS gage" reaches. Reaches classified through local expertise alone
10
-------
were categorized as "acceptable" and included in the study to fill gaps in study sub-regions
where an insufficient number of "preferred" and "USGS gage" reaches classified as intermittent
or ephemeral could be identified.
Of these 1166 reaches, 149 reaches were sampled (31 ephemeral, 66 intermittent, and 52
perennial reaches) in a sampling campaign that ran from July 2019 to October 2020. Post-
sampling site classifications were reviewed in light of the data collected, including the Stream
Temperature, Intermittence, and Conductance (STIC; Chapin et al. 2014) logger data collected
at 48 "baseline" sites that were revisited multiple times over a year (baseline sites are
described under Data collection below). If sampling events produced direct observations of
stream hydrology inconsistent with the initial classification (e.g., ephemeral reaches flowing
during site visits without antecedent precipitation), then field notes and field photos were used
to determine reach flow duration. Each of these cases triggered case-by-case review of all
available materials by the project delivery team and the RSC to determine if the original
classification should remain the same, be updated, or excluded from analysis.
In the final data set of 149 sampled reaches, streamflow duration class was directly determined
from USGS stream gage records at 48% of reaches (41 perennial and 30 intermittent reaches,
but no ephemeral reaches; Error! Reference source not found., Figure 4Error! Reference
source not found.)- Other sources of hydrologic data used to directly classify study reaches
include continuous data loggers (48 reaches), trail cameras, published studies, and consultation
with local experts. Multiple sources of hydrologic data were used to classify 47 of the ungaged
assessment reaches and a single source was used at 33 ungaged study reaches. In general,
more hydrologic data were available at perennial reaches than at intermittent or ephemeral
reaches.
Figure 4. Locations of 31 ephemeral, 66 intermittent, and 52 perennial study stream reaches used to develop the beta SDAM
WM.
11
-------
Table 2. Distribution of sites used to develop the beta SDAM WM. Baseline sites were visited three times throughout the study
and had water presence loggers installed and validation sites were visited once throughout the study and did not have loggers
installed.
Validation Baseline
Class
Gaged
Preferred
Acceptable
Gaged
Preferred
Total
Ephemeral
0
5
22
0
4
31
-California and
Nevada
0
0
8
0
2
10
-Central Rockies
0
2
4
0
1
7
-Northern Rockies
0
0
6
0
0
6
-Southern Rockies
0
3
4
0
1
8
Intermittent
16
10
10
12
18
66
-California and
Nevada
5
2
1
5
5
18
-Central Rockies
2
4
3
0
8
17
-Northern Rockies
6
0
6
2
4
18
-Southern Rockies
3
4
0
5
1
13
Perennial
31
6
1
10
4
52
-California and
Nevada
9
0
0
4
0
13
-Central Rockies
4
5
1
0
2
12
-Northern Rockies
9
1
0
3
1
14
-Southern Rockies
9
0
0
3
1
13
Data collection
Reaches were sampled following the development protocol (available here and in the
supplementary material of Mazor et al. 2021c), which covers measurement of indicators
identified in Mazor and McCune (2021), as well as "Level 1" indicators of the NM method
(NMED 2011), and all indicators of the PNW method (Nadeau 2015). STIC loggers (Chapin et al.
2014) were deployed at 48 "baseline" reaches and were revisited a total of three times each
over a year; "validation" sites were visited once and did not have loggers. For further details on
STIC data loggers and their verification/calibration, deployment, and data retrieval, see
Schumacher and Fritz (2019). The sampling protocol used in this study was identical to that
used to develop the beta SDAM AW. Mazor et al. (2021a) provides a summary of these data
collection protocols. Sampled study sites are shown in Figure 4. Forty-two of these study sites
were noted as disturbed by human activity (e.g., channelization, discharges, diversions) by field
crews.
Data analysis
Metric calculation
Candidate indicator data were used to calculate 72 candidate metrics: 37 biological metrics, 7
geomorphological metrics, 8 hydrologic metrics (7 of which were direct measures of water
presence), and 20 geospatial metrics (Table 3).
12
-------
Table 3. Metrics evaluated for the development of the beta SDAM WM. PctDom: Percent of observations with the most common value (typically zero). PvlvE: F-statistic from a
comparison of mean values at perennial, intermittent, and ephemeral reaches. Absolute t-statistic from a comparison of mean values at ephemeral and at least intermittent
reaches (EvAUj, at perennial and non-perennial reaches (PvNP), at flowing intermittent and perennial reaches (Pvlwet), and at non-flowing intermittent and ephemeral reaches
(Evldry). rf_MDA: Variable importance from a random forest model, measured as mean decrease in accuracy. Screen: Indicates if the metric passed or failed screening criteria in
Table 4. Ord: Ordinal metrics. Bin: Binary metrics. Con: Continuous metrics. Asterisks (*) indicate hydro logic metrics that directly measure the presence of water. NM: Metrics
derived from candidate indicators used in the SDAM NM. OBL and FACW: Obligate and facultative-wet wetland indicator plants, respectively (Lichvar et al. 2016). EPT:
Ephemeroptera, Plecoptera, and Trichoptera insect orders. GOLD: Gastropoda, Oligochaeta, and Diptera invertebrate groups. OCH: Odonata, Coleoptera, and Heteroptera insect
orders.
Metric Description Form PctDom Range PvlvE EvALI PvNP Pvlwet Evldry rf_MDA Screen
Biological
fishabund_score2
Abundance of fish, excluding
Ord
73%
3
8.91
7.09
3.01
0.51
1.44
0.0004
Pass
mosquitofish (NM)
DifferenceslnVegetation_score
Differences in vegetation between the
Ord
34%
3
31.10
6.28
6.35
1.47
1.93
0.0027
Pass
riparian corridor and adjacent uplands
score (NM)
UplandRootedPlants_score
Absence of upland rooted plants in the
Ord
47%
3
6.04
2.60
3.29
0.20
0.53
-0.0003
Pass
streambed score (NM)
iofb_score
Presence of iron-oxidizing bacteria and
Bin
85%
1.5
6.30
5.18
2.74
0.96
1.00
0.0003
Pass
fungi score (NM)
mayfly_abundance
Abundance of mayflies
Con
47%
66
52.47
10.78
8.41
4.18
1.16
0.0111
Pass
perennial_abundance
Abundance of perennial indicator taxa
Con
58%
90
16.06
6.48
4.72
2.23
1.56
0.0062
Pass
perennial_taxa
Number of perennial indicator taxa
Con
58%
14
16.27
7.21
4.75
1.67
1.02
0.0007
Pass
perennial_live_abundance
Abundance of perennial indicator taxa
Con
58%
90
15.90
6.44
4.71
2.19
1.17
0.0063
Pass
(living specimens only)
snake_score
Presence of aquatic snakes
Bin
97%
1
0.31
0.27
0.72
0.10
1.00
0.0000
Fail
vert_score
Presence of aquatic vertebrates
Bin
86%
1
1.66
0.88
1.67
0.32
1.79
0.0001
Fail
vert_sumscore
Number of aquatic vertebrate types
Ord
92%
2
0.48
0.16
0.80
0.15
1.36
0.0001
Fail
present (fish, amphibians, snakes,
turtles)
hydrophytes_present
Number of OBL and FACW plant species
Ord
20%
13
15.71
5.41
4.65
0.76
0.24
0.0012
Pass
present in the channel or within a half-
channel width of the channel
13
-------
Metric
Description
Form
hydrophytes_present_noflag
alglivedead_cover_score
moss_cover_score
liverwort_cover_score
PctShading
TotalAbundance
Richness
EPT_abundance
EPT_taxa
EPT_relabd
EPT_reltaxa
GOLD_abundance
GOLD_taxa
OCH_abundance
OCH_taxa
GOLD_relabd
GOLD_reltaxa
OCH_relabd
OCH_reltaxa
GOLDOCH_relabd
GOLDOCH_reltaxa
Noninsect abundance
Number of OBL and FACW plant species Ord
present in the channel or within a half-
channel width of the channel (excluding
those with a flagged unusual
distribution)
Cover of live or dead algae on the Ord
streambed
Moss cover on the streambed Ord
Liverwort cover on the streambed Ord
Percent shading on the streambed Con
Total abundance of aquatic Con
invertebrates
Total richness of aquatic invertebrate Con
families
Abundance ofEPT Con
Number of EPT families Con
Relative abundance of EPT families Con
Relative richness of EPT families Con
Abundance of GOLD Con
Number of GOLD families Con
Abundance of OCH Con
Numer of OCH families Con
Relative abundance of GOLD taxa Con
Relative richness of GOLD taxa Con
Relative abundance of OCH taxa Con
Relative richness of OCH taxa Con
Relative abundance of GOLD and OCH Con
taxa
Relative richness of GOLD and OCH taxa Con
Abundance of non-insect taxa Con
PctDom
Range
PvlvE
EvALI
PvNP
Pvlwet
Evldry
rfJVIDA
Scree
20%
13
14.79
5.37
4.44
0.58
0.09
0.0001
Pass
34%
4
45.84
10.23
8.03
2.06
2.51
0.0049
Pass
63%
3
0.21
0.26
0.65
0.85
0.35
0.0000
Fail
88%
3
1.38
2.46
1.20
0.57
1.06
-0.0001
Pass
8%
1
2.54
0.66
2.25
1.90
0.13
0.0001
Pass
21%
287
35.93
9.72
7.09
2.73
0.11
0.0077
Pass
21%
36
45.87
10.24
8.53
2.87
0.13
0.0067
Pass
34%
150
37.86
9.62
7.27
3.16
0.84
0.0107
Pass
34%
27
37.14
9.86
7.38
2.82
0.90
0.0095
Pass
34%
1
13.78
3.86
5.36
2.27
0.86
0.0021
Pass
34%
2
16.42
5.83
5.42
1.94
1.39
0.0013
Pass
33%
91
31.93
9.39
6.62
2.17
0.30
0.0025
Pass
33%
14
32.04
10.28
6.59
1.70
0.27
0.0012
Pass
44%
74
10.19
5.26
3.86
1.19
1.03
-0.0002
Pass
44%
11
9.61
4.39
3.82
0.62
1.03
-0.0010
Pass
33%
1
4.11
2.40
2.12
1.56
0.18
0.0030
Pass
33%
1
6.66
3.44
2.48
1.66
0.06
0.0022
Pass
44%
1
0.06
0.04
0.36
0.02
0.32
-0.0001
Fail
44%
1
0.03
0.18
0.06
0.76
0.21
0.0002
Fail
27%
1
2.93
2.00
1.58
1.51
0.06
0.0011
Pass
27%
1.4
4.75
2.74
2.01
1.94
0.09
0.0008
Pass
50%
87
6.76
5.02
2.87
0.40
0.37
0.0001
Pass
-------
Metric
Description
Form PctDom Range PvlvE EvALI PvNP Pvlwet Evldry rf_MDA Screen
Noninsect_taxa
Richness of non-insect taxa
Con
50%
11
7.34
5.68
2.81
0.15
0.05
0.0001
Pass
Noninsect_relabund
Relative abundance of non-insect taxa
Con
50%
1
0.37
0.30
0.69
1.21
0.21
0.0003
Fail
Noninsect_reltaxa
Relative richness of non-insect taxa
Con
50%
1
0.54
0.62
0.53
1.37
0.22
-0.0009
Fail
Geomorphological
Sinuosity_score
Channel sinuosity score (NM)
Ord
33%
3
4.30
1.52
2.76
1.42
0.68
0.00
Pass
ChannelDimensions_score
Channel dimensions score (NM)
Ord
37%
3
0.52
0.97
0.31
0.57
0.28
0.00
Fail
RifflePoolSeq_score
Riffle-pool sequence score (NM)
Ord
31%
3
11.92
2.66
5.07
2.48
0.09
0.00
Pass
SubstrateSorting_score
Substrate sorting score (NM)
Ord
33%
3
8.64
2.78
4.14
2.11
0.56
0.00
Pass
SedimentOnPlantsDebris_score
Sediment on plants and debris score
Ord
91%
1.5
0.43
0.70
0.97
0.42
0.38
0.00
Fail
(NM)
BankWidthMean
Mean bank-width
Ord
2%
48
8.54
5.29
3.44
1.67
1.32
0.01
Pass
Slope
Valley slope
Ord
15%
26
1.24
1.48
0.61
0.69
0.43
0.00
Fail
Hydrologic
WaterlnChannel_score
* Water in channel score (NM)
Ord
48%
6
110.02
17.82
13.98
3.88
1.85
0.03
Pass
HydricSoils_score
Presence of hydric soils in the channel
Bin
76%
3
8.20
3.77
3.42
1.89
0.97
0.00
Pass
score (NM)
springs_score
* Presence of springs or seeps in the
Bin
98%
3
0.63
1.00
1.00
1.00
0.00
0.00
Fail
channel score (NM)
SurfaceFlow_pct
* Percent of reach with flowing surface
Ord
50%
100
102.77
19.20
13.77
3.53
1.00
0.03
Pass
water
SurfaceSubsurfaceFlow_pct
* Percent of reach with flowing surface or
Ord
86%
100
6.66
4.20
2.04
3.14
1.97
0.00
Pass
subsurface water
lsolatedPools_number
* Number of isolated pools (no
Ord
89%
9
3.49
0.78
2.92
1.53
1.84
0.00
Pass
connection to flowing surface water)
WoodyJams_number
Number of woody jams in the reach
Ord
79%
10
0.98
0.22
1.42
0.49
1.10
0.00
Fail
SoilMoist_MaxScore
* Maximum soil moisture score in the
Ord
72%
2
55.23
8.53
8.44
0.00
1.94
0.01
Pass
reach
Geospatial
Elev_m
Elevation
Con
3%
3250
2.11
1.97
1.49
0.41
0.89
0.00
Pass
tmean
Mean annual temperature
Con
3%
17
1.66
1.95
0.92
0.34
0.67
0.00
Fail
tmax
Maximum annual temperature
Con
2%
18
1.97
2.17
0.39
1.11
0.76
0.00
Pass
tmin
Minimum annual temperature
Con
2%
17
1.56
1.63
1.40
0.43
0.54
0.00
Fail
-------
Metric
Description
Form
PctDom
Range
PvlvE
EvALI
PvNP
Pvlwet
Evldry
rfJVIDA
Scree
MeanSnowPersistence_
.10
Mean snow persistence within a 10-km
radius of the reach
Con
1%
82
2.69
2.52
1.45
0.00
0.91
0.00
Pass
MeanSnowPersistence_
.05
Mean snow persistence within a 5-km
radius of the reach
Con
1%
86
2.97
2.67
1.35
0.17
1.11
0.00
Pass
MeanSnowPersistence_
.01
Mean snow persistence within a 1-km
radius of the reach
Con
1%
84
2.53
2.51
1.18
0.25
1.02
0.00
Pass
ppt
Mean annual precipitation
Con
2%
1603
0.80
0.23
1.38
0.76
0.54
0.00
Fail
ppt.mOl
Mean January precipitation
Con
2%
337
0.90
0.80
1.44
0.38
0.33
0.00
Fail
ppt.m02
Mean February precipitation
Con
2%
293
0.50
0.35
1.09
0.33
0.53
0.00
Fail
ppt.m03
Mean March precipitation
Con
2%
254
0.49
0.41
1.06
0.42
0.37
0.00
Fail
ppt.m04
Mean April precipitation
Con
2%
143
1.08
0.82
0.99
1.33
0.66
0.00
Fail
ppt.m05
Mean May precipitation
Con
2%
107
1.93
1.29
1.07
2.28
0.36
0.00
Pass
ppt.m06
Mean June precipitation
Con
3%
129
2.20
1.51
0.99
2.15
0.53
0.00
Pass
ppt.m07
Mean July precipitation
Con
2%
102
0.37
0.86
0.57
0.27
0.53
0.00
Fail
ppt.m08
Mean August precipitation
Con
2%
131
0.05
0.06
0.31
0.50
0.17
0.00
Fail
ppt.m09
Mean September precipitation
Con
2%
80
0.49
0.66
0.93
0.57
0.17
0.00
Fail
ppt.mlO
Mean October precipitation
Con
2%
102
0.08
0.33
0.20
0.16
0.38
0.00
Pass
ppt.mll
Mean November precipitation
Con
2%
247
0.80
0.44
1.44
0.73
0.35
0.00
Fail
ppt.ml2
Mean December precipitation
Con
2%
367
0.74
0.68
1.34
0.39
0.32
0.00
Fail
16
-------
Metric screening
As an initial data exploration step, we visualized the relationships between streamflow duration
class (hereafter "flow class") and indicators by ordinating all 72 metrics for all samples in the
data set in a nonmetric multidimensional scaling using Gowers' distance. Convex hulls were
drawn around each streamflow duration class to help visualize their distributions in ordination
space. The 2-axis ordination was computed using the metaMDS function in the vegan R package
(Oksanen et al. 2019). Correlation coefficients (Spearman's rho) were calculated between
ordination axes and metric values. Wet and dry reaches were plotted separately to evaluate the
role of flow conditions at the time of the visit on flow duration indicators; streams with scores 4
and higher for the "Water in channel" indicator (WaterlnChannel_score) from the NM SDAM
were considered wet and scores 3 or lower were considered dry.
The ordination showed that perennial and ephemeral reaches were quite distinct, but
intermittent reaches overlapped considerably with the other classes (Figure 5). In general,
intermittent reaches that were dry on collection dates were similar to ephemeral reaches and
17
-------
0.1 -
cm 0.0
CO
~
-0.1 -
A A A A
* A
A A i 1* * A
\ A A A A A A A^* *
* .
* *
t A * A
* A A a/ *
A A* '
* ** |
* A A A «
* a * a & aj :
ml ##
A / a- **
. :
a
.
^
ft
A A A A A \\
. w
A A
A
A ^ A
A #
A
A
-0.2
0.0
0.2
MDS1
Dry
Flowing
Eph
Int
Per
o
sz
Cd
0.25-
0.00
-0.25-
-0.50-
-0.75 -
Geomorphological
Geospatial
Hydrological
h+-^7"
-0.75 -0.50 -0.25
0.00 0.25 -0.75
Rho with MDS1
-0.50 -0.25 0.00 0.25
Figure 5. A two-axis nonmetric multidimensional scaling of metrics based on biological, geomorphic, geospatial, and hydrologic
indicators. Panel A shows individual reaches. MDS: Multidimensional scaling axis 1 or 2. Eph: Ephemeral reaches. Int:
Intermittent reaches. Per: Perennial reaches. Circle: Reaches were dry during the site visit. Triangle: reaches were flowing during
the site visit. Panel B shows correlations (Spearman's rho) betvseen selected metrics and ordination axis scores; metrics with rho2
>0.5 are highlighted in blue (no geomorphological or geospatial metrics had rho2 > 0.5, nor did any metric have rho2> 0.5 with
18
-------
the second axis). Selected metrics are labeled: Biological metrics: A: Total aquatic invertebrate abundance. B: GOLD abundance.
C: EFT abundance. D: Perennial indicator taxa abundance. E: GOLDOCH relative richness. Geomorphological metrics: F: Bank
width. G: Slope. Geospatial metrics: H: Mean snow persistence within 10 km. I: Mean annual maximum temperature. Hydro logic
metrics. J: Percent of reach with surf ace flow. K: Soil moisture. L: Number of isolated pools.
intermittent reaches that had surface flow on collection dates were similar to perennial
reaches. Hydrologic and biological metrics were among the most strongly correlated with
ordination axes and no geomorphological or geospatial metric correlated with an ordination
axis with a rho2 greater than 0.5.
Metrics were evaluated using several criteria for inclusion in the beta SDAM (Table 4). We
developed criteria following approaches for screening metrics in bioassessment indices (e.g.,
Stoddard et al. 2008) and applied them to data from initial reach-visits (i.e., data from revisits
were withheld from analysis). One criterion was a distribution statistic, calculated as percent
dominance of the most common value (which was typically zero); all metrics had to meet this
criterion. The remaining criteria measured responsiveness of metrics (i.e., ability to discriminate
across flow classes). Most of these measures were based on statistical comparisons of mean
values at different subsets of reaches (e.g., t-statistic from a comparison of metric values at
perennial and non-perennial reaches), as has been used in other studies (Hawkins et al. 2010,
Cao and Hawkins 2011, Mazor et al. 2016). Another responsiveness statistic was based on
variable importance (specifically, mean decrease in accuracy) from a random forest model to
predict streamflow duration class from all candidate metrics; the model was calibrated using
the default option from the randomForest function in the randomForest package in R (Liaw and
Wiener 2002). Metrics had to meet at least one responsiveness criterion to be considered in
further analyses. A total of 47 of the 72 candidate metrics met these criteria and were
considered as screened metrics.
Table 4. Metric screening criteria. Metrics had to meet the distribution criterion and at least one responsiveness criterion to be
considered screen ed for further analysis.
Criterion
Distribution criterion
% dominance of <95%
most common value
Responsiveness criteria
PvlvE F>2
EvALI t>2
PvNP t>2
Pvlwet t>2
Evldry t>2
Definition
Frequency of most common value (typically, zero) in the
development data set
F-statistic in a comparison of values at perennial versus
intermittent versus ephemeral reaches
t-statistic in a comparison of values at ephemeral versus at
least intermittent reaches
t-statistic in a comparison of values at perennial versus
non-perennial reaches
t-statistic in a comparison of values at perennial versus
flowing intermittent reaches
t-statistic in a comparison of values at ephemeral versus
dry intermittent reaches
19
-------
rf_MDA Top Mean decrease accuracy (MDA) in a random forest model
quartile to predict perennial, intermittent, or ephemeral
streamflow duration class
Metric selection
The screened metrics were reduced to a final set of metrics for the beta SDAM based on their
importance in random forest models using the recursive feature elimination (rfe) function in
the R caret package (Kuhn 2020). Briefly, rfe is a form of stepwise selection where complex
models (i.e., those based on many metrics) are calibrated and simpler models are considered
iteratively by eliminating the least important metrics. We considered the most complex model
(i.e., 47 candidate metrics included) then iteratively eliminating 5 variables at a time in each
step based on low variable importance until a 20-variable model was identified; after this point,
only one variable was eliminated in each step. The best performing model (i.e., highest
accuracy in predicting streamflow duration class) was identified. Then, the simplest model (i.e.,
the one with the fewest variables) with accuracy within 1% of the best was selected to identify
the final set of metrics. If the best-performing model selected by this approach had more than
20 variables, the 20-variable model was selected. For this analysis, accuracy was measured with
Cohen's Kappa statistic a measure of accuracy that accounts for uneven distribution among
the three streamflow duration classes.
We applied this modeling process to different subsets of the dataset, including:
the full region-wide dataset;
datasets stratified by sub-regions shown in Figure 2 (4 total); and
datasets stratified into snow-influenced and non-snow influenced sites, based on mean
snow persistence greater than 25% calculated for a 1-km, 5-km, and 10-km buffer from
the sampling reach (2 strata for each of 3 buffers).
For each subset, the modeling process was implemented:
with or without considering geospatial metrics; and
with or without considering metrics based on direct measures of water presence.
There are advantages and disadvantages to including these metrics in an SDAM and thus we
evaluated options with and without them. Geospatial metrics may improve SDAM performance
but would require GIS analysis to use the resulting method. Direct measures of water presence
can also greatly increase performance, but this introduces circularity (because water presence
was used to confirm and update streamflow duration classes in the development data set) and
may degrade the ability of the SDAM to work during atypical conditions, such as drought. See
(Mazor et al. (2021b) for a discussion of the implications of including geospatial metrics and
direct measures of water presence in SDAMs.
20
-------
To explore all these options, we developed 20 sets of models for different subsets of reaches
and combinations of predictors, with sets including between 1 and 5 models (44 models total;
Table 5). Analyses were conducted on data from the initial reach visits alone. For each of the 20
models, data were split into 80% training and 20% testing data sets, stratified by the 4 sub-
regions and 3 streamflow duration classes. Model design characteristics and optimal number of
metrics selected by rfe are shown in Table 5 and the selected metrics for each model are shown
in Figure 6.
Table 5. Design characteristics of the 44 models. H20: included direct measures of water presence. GIS: included geospatial
metrics, n sites: number of sites used in model training, testing, and evaluated for repeatability (revisit), rfe accuracy: accuracy
of best mode! produced by recursive feature elimination (rfe), measured as Cohen's Kappa or as out-of-bag (OOB) accuracy.
n reaches rfe accuracy
Model set
Stratum
H20
GIS
training
testing
revisit
# metrics
Kappa
OOB
Unstratified models
Unstrat
None
117
32
84
19
0.41
0.45
Unstrat GIS
None
Yes
117
32
84
15
0.41
0.38
Unstrat H20
None
Yes
117
32
84
16
0.47
0.38
Unstrat H20 GIS
None
Yes
Yes
117
32
84
3
0.47
0.28
Models stratified by region
Strat
California & Nevada
32
9
25
20
0.36
0.47
Strat
Central Rockies
27
9
21
18
0.52
0.26
Strat
Northern Rockies
29
9
19
19
0.27
0.41
Strat
Southern Rockies
26
8
19
3
0.51
0.31
Strat GIS
California & Nevada
Yes
32
9
25
16
0.38
0.28
Strat GIS
Central Rockies
Yes
27
9
21
20
0.52
0.41
Strat GIS
Northern Rockies
Yes
29
9
19
16
0.42
0.41
Strat GIS
Southern Rockies
Yes
26
8
19
3
0.45
0.38
Strat H20
California & Nevada
Yes
32
9
25
3
0.59
0.28
Strat H20
Central Rockies
Yes
27
9
21
14
0.54
0.26
Strat H20
Northern Rockies
Yes
29
9
19
10
0.39
0.45
Strat H20
Southern Rockies
Yes
26
8
19
3
0.52
0.31
Strat H20 GIS
California & Nevada
Yes
Yes
32
9
25
3
0.51
0.25
Strat H20 GIS
Central Rockies
Yes
Yes
27
9
21
20
0.51
0.3
Strat H20 GIS
Northern Rockies
Yes
Yes
29
9
19
20
0.25
0.31
Strat H20 GIS
Southern Rockies
Yes
Yes
26
8
19
8
0.49
0.31
Models stratified by snow influence
Snow influence within 1 km
SnowOl
Not snow-dominated
46
13
36
20
0.42
0.43
SnowOl
Snow-dominated
71
19
48
8
0.35
0.35
SnowOl GIS
Not snow-dominated
Yes
46
13
36
20
0.37
0.41
SnowOl GIS
Snow-dominated
Yes
71
19
48
18
0.35
0.32
SnowOl H20
Not snow-dominated
Yes
46
13
36
3
0.57
0.3
SnowOl H20
Snow-dominated
Yes
71
19
48
20
0.33
0.38
21
-------
n reaches rfe accuracy
Model set
Stratum
H20
GIS
training
testing
revisit
# metrics
Kappa
OOB
SnowOl H20 GIS
Not snow-dominated
Yes
Yes
46
13
36
6
0.6
0.28
SnowOl H20 GIS
Snow-dominated
Yes
Yes
71
19
48
10
0.48
0.35
Snow influence within 5 km
Snow05
Not snow-dominated
40
11
28
20
0.38
0.43
Snow05
Snow-dominated
77
21
56
15
0.48
0.36
Snow05 GIS
Not snow-dominated
Yes
40
11
28
14
0.33
0.4
Snow05 GIS
Snow-dominated
Yes
77
21
56
13
0.43
0.29
Snow05 H20
Not snow-dominated
Yes
40
11
28
13
0.48
0.25
Snow05 H20
Snow-dominated
Yes
77
21
56
20
0.43
0.3
Snow05 H20 GIS
Not snow-dominated
Yes
Yes
40
11
28
4
0.54
0.3
Snow05 H20 GIS
Snow-dominated
Yes
Yes
77
21
56
17
0.48
0.3
Snow influence within 10 km
SnowlO
Not snow-dominated
39
11
31
13
0.24
0.49
SnowlO
Snow-dominated
78
21
53
6
0.43
0.45
SnowlO GIS
Not snow-dominated
Yes
39
11
31
17
0.22
0.31
SnowlO GIS
Snow-dominated
Yes
78
21
53
11
0.52
0.33
SnowlO H20
Not snow-dominated
Yes
39
11
31
5
0.54
0.31
SnowlO H20
Snow-dominated
Yes
78
21
53
8
0.41
0.33
SnowlO H20 GIS
Not snow-dominated
Yes
Yes
39
11
31
4
0.42
0.31
SnowlO H20 GIS
Snow-dominated
Yes
Yes
78
21
53
13
0.52
0.24
Biological metrics (particularly those based on aquatic invertebrates) were among the most
widely selected metrics across model sets. Among non-biological metrics, mean bankfull width
was the only frequently selected geomorphological metric. Direct measures of water presence
were selected every time these measures were eligible for selection. Among geospatial metrics,
October precipitation was the most frequently selected metric (Figure 6).
22
-------
RifflePoolSeq_score
ppt.m05
Noninsect_taxa
MeanSnowPersistence_10
MeanSnowPersistence_05
3t.m06
_taxa
HydricSoils_score
GOLDOCH reltaxa
GOLDOChQelabd
tmax
SubstrateSorting_score
Noninsect_abundance
OCH_abundance
I i ve rwo rt_co ve r_sco re
iofb_score
SurfaceSubsurfaceFlow_pct
fishabund score2
PctShading
GOLD_relabd
SoilMoist_MaxScore
Sinuosity score
EPT_relabd
UplandRootedPlants_score
hydrophytes present
EPTj"eltaxa
ppt.mlO
perennial_taxa
hydrophytes_present noflag
GOLD taxa
GOLD_reltaxa
DifferenceslnVegetation_score
WaterlnChannel_score
SurfaceFlow_pct
GOLD_abundance
alglivedead_cover_score
TotalAbundance
EPT taxa
perennial_abundance
Richness
EPT_abundance
perennial live abundance
BanKWidthMean
mayfly_abundance
ro
E
^ ro
3 5>
¦<- lo
o o
5 5
o o
c c
co co
OC0C0C0C0C000000
iooooo
-------
Preliminary model calibration and performance assessment
Random forest models were then fit for each of the 20 options using the randomForest
function in the randomForest package in R (Liaw and Wiener 2002) using default parameters,
except that the number of trees was set to 1500 instead of the default 500. Only the initial visit
for reaches in the calibration data set was used for model fitting.
Model performance evaluation focused on two aspects: accuracy and repeatability. Accuracy
was assessed by calculating the same comparisons used to evaluate metric responsiveness
during the metric screening phase (e.g., ephemeral versus at least intermittent reaches,
perennial versus wet intermittent reaches, etc.; Table 4). Accuracy was measured using the
initial reach-visit in both the calibration training and testing data sets independently. We
compared training and testing measures to see if models validated poorly, suggesting that they
may be overfit.
Repeatability was assessed using data from the 48 reaches that were revisited (i.e., Baseline
sites; Error! Reference source not found.) and was calculated as the percent of reaches where
model classifications from visits were the same (regardless of classification accuracy). Due to
the limited amount of data, repeatability was only assessed on a region-wide basis and not
within each subregion; it was not analyzed separately for calibration and validation reaches.
Performance of the beta SDAM AW, SDAM PNW, and SDAM NM was also evaluated within the
training data set.
SDAM models newly developed through the current effort had better performance than
previously developed SDAMs (especially the beta SDAM AW), but among the new models,
performance was similar and there was no clear best model set (Table 6, Figure 7 and Figure 8).
Stratified model sets performed slightly better than the unstratified models and there were
modest improvements in accuracy achieved by including geospatial metrics, as well as direct
measures based on water presence. The RSC recommended the model set stratified by snow
influence calculated within a 10-km radius; furthermore, the RSC opted for the models that
included geospatial metrics (i.e., model set Snow 10 GIS) but did not recommend including
direct measures of water presence due to the potential introduction of circularity (water
presence during field visits was sometimes used to inform or verify the direct flow classification
of stream reaches), as described above.
24
-------
Table 6. Performance of the 20 model sets evaluated. PvlvE: Percent of reaches classified correctly as perennial, intermittent, or
ephemeral. EvAU: Percent of reaches classified correctly as ephemeral or at least intermittent. PvNP: Percent of reaches
classified correctly as perennial or non-perennial. Pvlwet: Percent of flowing reaches classified correctly as perennial or
intermittent. IvEdry: Percent of dry reaches correctly classified as intermittent or ephemeral. Train: Result for training data. Test:
Result for testing data. Model sets are described in Table 5. AW: Results for the beta SDAM AW. NM: Results for the SDAM NM.
PNW: Results for the SDAM PNW.
Accuracy
PvlvE EvALI PvNP Pvlwet IvEdry
Model set
Train
Test
Train
Test
Train
Test
Train
Test
Train
Test
Precision
AW
0.39
0.79
0.45
0.48
0.25
0.67
NM
0.58
0.8
0.72
0.66
0.46
0.87
PNW
0.57
0.79
0.78
0.64
0.46
0.82
SnowlO H20
GIS
0.74
0.59
0.88
0.81
0.85
0.78
0.76
0.63
0.7
0.54
0.81
Snow05 H20
GIS
0.7
0.75
0.86
0.88
0.84
0.88
0.72
0.81
0.67
0.64
0.83
SnowOl H20
GIS
0.68
0.78
0.88
0.88
0.79
0.91
0.66
0.84
0.7
0.69
0.82
Stratum H20
GIS
0.71
0.6
0.82
0.83
0.89
0.77
0.79
0.65
0.6
0.5
0.84
Unstrat H20
GIS
0.72
0.69
0.87
0.84
0.85
0.84
0.75
0.74
0.67
0.62
0.8
SnowlO H20
0.68
0.66
0.86
0.78
0.81
0.88
0.69
0.78
0.64
0.5
0.8
Snow05 H20
0.72
0.5
0.9
0.81
0.82
0.69
0.7
0.5
0.74
0.5
0.83
SnowOl H20
0.65
0.69
0.85
0.81
0.79
0.88
0.66
0.79
0.63
0.54
0.82
Stratum H20
0.68
0.6
0.83
0.77
0.84
0.83
0.76
0.7
0.55
0.47
0.83
Unstrat H20
0.62
0.75
0.85
0.88
0.77
0.88
0.62
0.79
0.63
0.69
0.8
SnowlO GIS
0.68
0.63
0.89
0.81
0.78
0.81
0.66
0.7
0.7
0.5
0.83
Snow05 GIS
0.68
0.69
0.85
0.88
0.83
0.81
0.72
0.68
0.61
0.69
0.84
SnowOl GIS
0.64
0.59
0.85
0.81
0.79
0.78
0.65
0.67
0.62
0.5
0.84
Stratum GIS
0.63
0.57
0.82
0.8
0.81
0.77
0.69
0.65
0.55
0.42
0.8
Unstrat GIS
0.62
0.53
0.81
0.81
0.8
0.72
0.68
0.47
0.5
0.6
0.84
SnowlO
0.54
0.69
0.79
0.84
0.74
0.84
0.59
0.63
0.44
0.75
0.73
Snow05
0.62
0.69
0.85
0.88
0.75
0.81
0.59
0.68
0.65
0.69
0.77
SnowOl
0.62
0.69
0.84
0.88
0.78
0.81
0.63
0.65
0.6
0.75
0.78
Stratum
0.63
0.46
0.82
0.8
0.8
0.66
0.66
0.42
0.58
0.5
0.83
Unstrat
0.55
0.63
0.79
0.81
0.74
0.81
0.58
0.67
0.49
0.57
0.79
25
-------
AW-
NM-
PNW -
SnowlO H20 GIS-
Snow05 H20 GIS -
SnowOI H20 GIS -
Stratum H20 GIS -
Unstrat H20 GIS -
SnowlO H20-
Snow05 H20 -
SnowOI H20-
Stratum H20 -
Unstrat H20 -
SnowlO GIS -
Snow05 GIS -
SnowOI GIS-
Stratum GIS-
Unstrat GIS -
SnowlO-
Snow05-
SnowOI-
Stratum -
Unstrat-
O.OO.20.5O.75.a!DOO.2e.5O.75.a]ilOO.2e.50.75 CGDOO.20.5O.75.amOO.20.5O.75.aEOO.20.5O.75.OO
Performance
Set Testing ¦ Training
Figure 7. Performance of the 20 model sets evaluated. Blue dots indicate the highest-performing model sets and red dots
indicate the next-best performing model sets. PvlvE: Percent of reaches classified correctly as perennial, intermittent, or
ephemeral. EvAU: Percent of reaches classified correctly as ephemeral or at least intermittent. PvNP: Percent of reaches
classified correctly as perennial or non-perennial. Pvlwet: Percent of flowing reaches classified correctly as perennial or
intermittent. IvEdry: Percent of dry reaches correctly classified as intermittent or ephemeral. Unstrat: Unstratified models.
Stratum: Models stratified by subregion. SnowlO: Models stratified by snow persistence. Model sets are described in Table 5.
AW: Results for the beta SDAM AW (Mazor et al. 2021a). NM: Results for the SDAM NM. PNW: Results for the SDAM PNW.
Accuracy
PvlvE
0
h 1 1 r-
Accuracy
EvALI
t 1 1 1 r
Accuracy
PvNP
*
»
»
t 1 1 r
Accuracy
Pvlwet
1
m
41
t 1 1 1 r
Accuracy
IvEdry
^
«
#
i i i
i
Precision
t 1 1 1 r
26
-------
SP
snowO
M
Jnsfra'
m
l>?,v
UWUJ
nstrat -L
ijp
M
W3IH
now,
M
: a:up"
w
snowQ
fm
ffat
Snowl
r"OWp
iai
il
AV
Snowl
anowQ
Ix
:: ..
SCIOWO
Stratur.,
unsfraT
Accuracy
PvlvE
Accuracy
EvALI
Accuracy
PvNP
Accuracy
Pvlwet
Accuracy
IvEdry
Precision
m
m *
U)
«
M*
O*
CO
CO
m mm m
mm m
m mm m
O
3
m a
mmm
m m
0.25 0.50 0.75
0.25 0.50 0.75 0.25 0.50 0.75 0.25 0.50 0.75
Performance
0.250.50 0.75 0
25 0.50 0.75
Evaluated
Stratum
CA-NV
Southern Rockies
Central Rockies
Northern Rockies
Snow-dominated
Not snow-dominated
Figure 8. Performance of the 20 model sets evaluated within strata defined by sub-region or sno w influence. The y-axis labels on
the left indicate the stratifications used to develop the models (if any) and the panel labels on the right indicate the
stratifications used to assess performance. PvlvE: Percent of reaches classified correctly as perennial, intermittent, or ephemeral.
EvALI: Percent of reaches classified correctly as ephemeral or at least intermittent. PvNP: Percent of reaches classified correctly
as perennial or non-perennial. Pvlwet: Percent of flowing reaches classified correctly as perennial or intermittent. IvEdry: Percent
of dry reaches correctly classified as intermittent or ephemeral. Model sets are described in Table 5. AW: Results for the beta
SDAM A W (Mazor et al. 2021a). NM: Results for the SDAMNM. PNW: Results for the SDAM PNW.
Simplification of the selected model set
Upon selection of the final model set (i.e., models that included geospatiai metrics and were
stratified by snow influence calculated within a 10-km radius), we attempted to simplify the
selected model set in three steps to make the SDAM easier to implement in the field while
improving (or at least not sacrificing) performance. Simplification occurred in three steps;
27
-------
1. Refinement of metrics
2. Increased confidence required for classifications
3. Addition of single indicators of at least intermittent flow
Refinement of metrics
The metric selection process described above identified an optimal set of metrics to use in the
SDAM, but it did so without considering difficulties in measuring each metric or effort required
to measure all of the metrics. For example, rfe may have selected a metric based on the total
number of aquatic invertebrates, even if there was little new information provided once 20
individuals were recorded. That is, SDAM users might be able to cease counting aquatic
invertebrates once 20 individuals were recorded. Simplifying metrics was intended to reduce
the burden on SDAM users and facilitate method use (e.g., avoid reliance on access to statistical
software). Some metrics were eliminated because they were closely related to another metric
in the selected model set (i.e., they described similar stream characteristics, such as mayfly
abundance and EPT abundance). Metrics that were more time-consuming to measure were
rejected if a simpler alternative was available and continuous metrics were converted to binary
or ordinal metrics based on visual interpretation of random forest partial dependence curves
(binary and ordinal metrics are typically more rapid to measure and easier to standardize than
continuous metrics). Accuracy and repeatability measures were re-evaluated to ensure that
overall model performance was not substantially diminished by the modifications.
The snow-influenced and non-snow influenced models were refined in parallel steps. At each
step, metrics were either eliminated, classified into categorical bins, or otherwise modified. The
impact on performance was assessed and the highest performing modification was selected for
further refinement. Performance was assessed in terms of three accuracy measures: PvlvE (i.e.,
proportion of reaches classified corrected as perennial, intermittent, or ephemeral), EvALI (% of
reaches classified correctly as ephemeral or at least intermittent), and Cohen's Kappa. The
metric refinement steps are described below. Asterisks (*) indicate the selected refinement at
each step; if no asterisk is shown, none of the refinements considered at that step were
selected and the selected option from the previous step was used for further analysis.
Snow-influenced model:
1. Select two aquatic invertebrate metrics:
a. Total abundance and richness
b. Total abundance and perennial indicator abundance*
c. Total abundance and richness of perennial indicator taxa
d. Total abundance and EPT abundance
e. Total abundance and richness of EPT taxa
f. Total abundance and GOLD abundance
g. Total abundance and richness of GOLD taxa
2. Add a third aquatic invertebrate metric
a. Richness of EPT taxa
28
-------
b. Richness of perennial indicator taxa *
c. Total richness
d. Richness of GOLD taxa
3. Bin richness of perennial indicator taxa metric
a. Two categories* (0 to 3, >4)
b. Three categories (0,1 to 3, >4)
4. Bin total and perennial indicator abundance
a. Three categories for total abundance (0,1 to 19, 20+) and perennial indicator
abundance (0, 1 to 5, >6)*
5. Bin mean bankfull width
a. Three categories (<2, 2 to 6, >6)*
6. Bin streambed algal cover
a. Two categories (<10%, >10%)*
7. Bin or drop geospatial metrics (NONE SELECTED)
a. Bin October precipitation at quartiles
b. Bin October precipitation at quintiles
c. Drop October precipitation
Refinements to the snow-influenced model improved model performance at most steps (Figure
9). These refinements included eliminating several variables and binning those that remained
into two or three categories. Unfortunately, no satisfactory way to bin the single geospatial
metric in this model (October precipitation) was identified, so it was retained as a continuous
variable for the beta SDAM WM.
29
-------
Snow-influenced model
Indicator refinement
1.00-
0.75-
(D
O
c
ro
£
!
o
t
a)
CL
0.50-
0.25-
Accuracy
measure
PvlvE
EvALI
Kappa
Selected
FALSE
TRUE
0.00.
I I I I I I I I
0 1 2 3 4 5 6 7
Step
Figure9. Impact of indicator refinement on the accuracy of the snow-influenced model. Solid lines show the performance of the
best model from each step. Dotted lines show the performance of model selected at each step. Dashed lines show performance
of the original model.
Non-snow influenced model
1. Select 2 aquatic invertebrate metrics
a. Total abundance and richness
b. Total abundance and abundance of perennial indicators
c. Total abundance and richness of perennial indicator taxa
d. Total abundance and EPT abundance
e. Total abundance and richness of EPT taxa
f. Total abundance and mayfly abundance
g. Total abundance and GOLD abundance
h. Total abundance and richness of GOLD taxa
i. Abundance and richness of EPT taxa
j. Abundance and richness of perennial indicator taxa
k. Mayfly abundance and total richness
I. Mayfly abundance and richness of perennial indicator taxa*
2. Add a third aquatic invertebrate metric (NONE SELECTED)
a. Total abundance
3. Remove an additional metric (NONE SELECTED)
a. Sinuosity
b. Mean bankfull width
30
-------
c. Fish abundance
4. Bin mayfly abundance
a. Five categories (0, 1 to 5, 6 to 10,11 to 15, >16)*
5. Bin richness of perennial indicator taxa
a. Four categories (0, 1, 2, >3)*
6. Bin mean bankfull width (NONE SELECTED)
a. Three categories (<2, 2 to 6, >6)
b. Bin at quartiles
7. Bin geospatial metrics (NONE SELECTED)
a. Bin May precipitation at three categories (<45, 45 to 50, 50+)
b. Bin May precipitation at quartiles
c. Bin maximum temperature at quartiles
d. Bin maximum temperature in two categories (<18, >18)
e. Bin maximum temperature and May precipitation based on quartiles
Refinements to the non-snow influenced model rarely improved model performance and most
refinements were rejected (Figure 10). The only refinement to substantially improve
performance was the binning of the mayfly abundance metric (step 4). Thus, the non-snow
influenced model retained more metrics in continuous forms than the snow-influenced model.
Non-Snow influenced model
Indicator refinement
1.00-
0.75-
<1>
O
c
ro
£
!
o
t
a)
CL
0.50-
0.25-
0.00 ¦
~ . . . .
-
4
6
Accuracy
measure
PvlvE
EvALI
Kappa
Selected
FALSE
TRUE
Step
Figure 10. Impact of indicator refinement on the accuracy of the non-snow influenced model. Solid lines show the performance
of the best model from each step. Dotted lines show the performance of model selected at each step. Dashed lines show
performance of the original model.
31
-------
Increased confidence required for classifications
Random forest models, when used in classification mode, traditionally make assignments based
on the class that receives the highest number of votes by each "tree" in the forest. Thus, in a 3-
way decision, the class with the most votes could receive much less than a majority of all
votesas low as 34%. The RSC believed such low-confidence classifications may not provide
sufficient defensibility for some management decisions, instead the RSC recommended
exploring approaches to distinguish between high- and low-confidence classifications.
Based on this input from the RSC, we explored increasing the minimum number of votes
required to make a confident classification from 30% to 100% by increments of 1%. When the
final model was applied to a novel test reach and a single class received a sufficient percent of
votes, then the reach was classified accordingly. If none met the minimum, but the combined
percent of votes for intermittent and perennial classes exceeded the minimum, then the reach
was classified as at least intermittent. In all other cases, the reach was classified as need more
information. This decision framework reflects the opinion of the RSC that distinguishing
between ephemeral and at least intermittent reaches is a high priority use of the SDAM, more
so than distinguishing between perennial and nonperennial (ephemeral and intermittent)
reaches. The percent of reaches under each of the five possible classifications with increasing
minimum vote agreement thresholds was calculated. The snow-influenced and non-snow
influenced models were analyzed together to evaluate the overall impact of this modification to
the entire WM.
At a minimum required proportion of votes of 0.5, only 5% of reaches were classified as at least
intermittent and none were classified need more information (Figure 11). Classifications of at
least intermittent first appear with a minimum proportion of 0.38 (0.45 in the testing data set),
whereas classifications of need more information appear at 0.51 (in both the training and
testing data sets). Although they cannot be ruled out, it appears unlikely that the beta SDAM
WM will result in classifications of need more information. Based on these results, the RSC
recommended a minimum proportion threshold of 0.5 for flow classification.
32
-------
Classification
NMI
ALI
'
¦ i
0.5 0.6 0.7 0.8
Minimum proportion of votes
Figure 11. Influence of the minimum proportion of votes required to make a classification on n (the number of reaches in each
class). NMI: Need more information. ALI: At least intermittent. P: Perennial. I: Intermittent. E: Ephemeral. The vertical black line
represents a minimum proportion of required votes of 0.5, reflecting the final recommendation of the RSC. The two red lines
represent the proportion of votes that first result in classification of ALI (the lower line) or NMI (the upper line). Only results from
the training data set are shown.
Addition of single indicators of at least intermittent flow
Single indicators can supersede model classifications of ephemeral to at least intermittent.
Single indicators provide technical benefits (i.e., improved accuracy), as well as non-technical
benefits, such as greater acceptance of the SDAM, given public understanding of the role of
streamflow duration in supporting wildlife and rapidity of determining a flow classification,
which is why they are used in most other SDAMs (e.g., NMED 2011, Nadeau et al. 2015, Dorney
and Russell 2018, Mazor et al. 2021a). The following potential single indicators, based on
recommendations from the RSC were evaluated:
Presence of aquatic invertebrates
Presence of EPT individuals, or at least 5 EPT individuals
Presence of hydrophytes, or at least 2 or 3 hydrophytic plant species
Algal cover > 10%
Presence offish
The number of instances where inclusion of the single indicator would correct a
misclassification (i.e., the reach was truly intermittent or perennial) and the number of times it
would introduce a misclassification (i.e., the reach was truly ephemeral) were quantified.
33
-------
Several single indicators had minimal impact on performance or introduced more errors than
they corrected (Figure 12). Based on these results, the RSC recommended using only the
presence of fish (apart from mosquitofish) as single indicators in the beta SDAM WM.
Aquatic vertebrates (incl. frog calls -
Aquatic vertebrates -
Aquatic snakes -
SDAM PNW single indicators -
SDAM NM single indicators -
Iron-oxidizing bacteria and fungi-
Hydrophytes (3+ species) -
Hydrophytes (2+ species) -
Hydrophytes (any)-
Hydric soils-
Fish or hydric soil or algae >10% -
Fish -
EPT (5+)-
EPT (any)-
Aquatic invertebrates -
Amphibians (incl. frog calls) -
Aquatic amphibians-
Algal cover >10% -
02468 10 02468 10
Number of sites changed
Net change I Worsen | No net change Improve
Set t Testing f Training
Figure 12. Influence of single indicators on performance of snow-influenced and non-snow influenced models
Performance of the beta SDAM WM
Performance of the final, simplified model for the beta SDAM WM is summarized in Table 7.
The overall accuracy was 74% in the training dataset (and 53% in the testing dataset), but this
accuracy increased to 93% in the training dataset (and 88% in the testing data set) when only
ephemera I versus at least intermittent classifications were considered (i.e., both blue and green
cells in Table 7 were treated as correct). Among 42 reaches marked as disturbed by human
activity, accuracy among all classes was 79% and 95% when only ephemeral versus at least
intermittent classifications were considered.
Snow-influenced
Non-snow influenced
. . .
34
-------
Table 7. Classifications of the final version of the beta SDAM WM on training and testing datasets. Blue cells indicate correct
classifications of perennial, intermittent, at least intermittent, and ephemeral reaches, whereas green cells indicate correct
classifications as ephemeral versus at least intermittent.
True streamflow duration class
Intermittent
Beta SDAM WM
Ephemeral
Dry
Flowing
Perennial
Classification
Train
Test
Train
Test
Train
Test
Train
Test
Ephemeral
20
4
4
1
0
0
0
0
Intermittent
3
3
17
3
16
5
8
7
At least
intermittent
1
0
2
0
2
1
3
0
Perennial
0
0
0
1
11
3
30
4
Data and code availability
All data used to develop the method and R code used in analysis are available here.
Next steps
Continued data collection within the WM is underway and will provide greater representation
of the diversity of stream conditions found within the region. Data from this effort will be used
to develop a final method (expected after 2023) to replace the beta method.
Acknowledgements
The development of this method and supporting materials was guided by a RSC consisting of
representatives of federal regulatory agencies in the Western U.S.: James T. Robb (U.S. Army
Corps of Engineers [USACE]South Pacific Division, Sacramento District), Robert Leidy (U.S.
Environmental Protection Agency [USEPA] Region 9), Aaron Allen (USACESouth Pacific
Division, Los Angeles District), Gabrielle C. L. David (USACEEngineer Research and
Development Center, Cold Regions Research and Engineering Laboratory), Loribeth Tanner
(USEPARegion 6), Rachel Harrington (USEPA - Region 8), Joe Morgan (USEPARegion 9),
Matt Wilson (USACEHeadquarters), Tunis McElwain (USACEHeadquarters), Silvia Gazzera
(USACE - Headquarters), Kevin Little, (USACE - Northwestern Division, Omaha District), Jess
Jordan (USACE - Northwestern Division, Seattle District), and Rose Kwok (USEPA
Headquarters).
We thank Abel Santana, Robert Butler, Duy Nguyen, Kristine Gesulga, and Anne Holt for
assistance with data management and Jeff Brown, Liesl Tiefenthaler, Mason London, John
Olson, Matthew Robinson, Emma Haines, Jess Turner, Katharina Zimmerman, Kelsey Trammel,
Marcus Beck, Savannah Pena, Abigail Rivera, and Andrew Caudillo for assistance with data
collection. Rob Coulombe provided training.
35
-------
Numerous researchers and land managers with local expertise assisted with the selection of
study reaches to calibrate the method: Patricia Spindler, Eric Stein, Andrew C. Rehn, Peter R.
Ode, Nathan Mack, Shawn McBride, Stephanie Kampf, Lindsey Reynolds, Kris Barrios, Marcia
Radke, Keith Bouma-Gregson, Kira Puntenney-Desmond, Andy Brummond, Don Lee, Ed Schenk,
Eric Hargett, Gabe Rossi, Mark Ockey, Sean Tevlin, Sean Lovill, Josh Smith, and Michael Bogan.
We thank the California Department of Fish and Wildlife's Aquatic Bioassessment Lab and
Daniel Pickard for use of imagery from the macroinvertebrate digital reference collection.
Cited literature
Cao, Y., and C. P. Hawkins. 2011. The comparability of bioassessments: a review of conceptual
and methodological issues. Journal of the North American Benthological Society 30:680-
701.
Chapin, T. P., A. S. Todd, and M. P. Zeigler. 2014. Robust, low-cost data loggers for stream
temperature, flow intermittency, and relative conductivity monitoring. Water Resources
Research 50:6542-6548.
Dorney, J., and P. Russell. 2018. North Carolina Division of Water Quality methodology for
identification of intermittent and perennial streams and their origins. Pages 273-279 in
J. Dorney, R. Savage, R. W. Tiner, and P. Adamus (eds.), Wetland and Stream Rapid
Assessments. Elsevier, San Diego, CA.
Fritz, K. M., T.-L. Nadeau, J. E. Kelso, W. S. Beck, R. D. Mazor, R. A. Harrington, and B. J. Topping.
2020. Classifying Streamflow Duration: The Scientific Basis and an Operational
Framework for Method Development. Water 12:2545.
Hammond, J. C., F. A. Saavedra, and S. K. Kampf. 2017. MODIS MOD10A2 derived snow
persistence and no data index for the western U.S. Available online:
https://www.hyd roshare.org/resource/lc62269aa802467688d25540caf2467e/
Hart, E., and K. Bell. 2015. Prism: Access Data From The Oregon State Prism Climate Project.
Hawkins, C. P., Y. Cao, and B. Roper. 2010. Method of predicting reference condition biota
affects the performance and interpretation of ecological indices. Freshwater Biology
55:1066-1085.
Kuhn, M. 2020. caret: Classification and Regression Training.
Liaw, A., and M. Wiener. 2002. Classification and regression by randomForest. R News 2:18-22.
Lichvar, R. W., D. L. Banks, W. N. Kirchner, and N. C. Melvin. 2016. The national wetland plant
list: 2016 wetland ratings. Phytoneutron 30:1-17.
36
-------
Mazor, R. D., and K. S. McCune. 2021. Review of flow duration methods and indicators of flow
duration in the scientific literature: Western Mountains. Pages 55. Southern California
Coastal Water Research Project, Costa Mesa, CA.
Mazor, R. D., A. C. Rehn, P. R. Ode, M. Engeln, K. C. Schiff, E. D. Stein, D. J. Gillett, D. B. Herbst,
and C. P. Hawkins. 2016. Bioassessment in complex environments: designing an index
for consistent meaning in different settings. Freshwater Science 35:249-271.
Mazor, R. D., B. J. Topping, T.-L. Nadeau, K. M. Fritz, J. E. Kelso, R. A. Harrington, W. S. Beck, K.
McCune, H. Lowman, A. Aaron, R. Leidy, J. T. Robb, and G. C. L. David. 2021a. User
Manual for a Beta Streamflow Duration Assessment Method for the Arid West of the
United States. Version 1.0. Pages 83. Document No. EPA-800-K-21001, U.S.
Environmental Protection Agency, Washington, D.C. Available online:
https://www.epa.gov/sites/production/files/2021-
03/documents/user_manual_beta_sdam_aw.pdf.
Mazor, R. D., B. J. Topping, T.-L. Nadeau, K. M. Fritz, J. E. Kelso, R. A. Harrington, W. S. Beck, K. S.
McCune, A. 0. Allen, R. Leidy, J. T. Robb, and G. C. L. David. 2021b. Implementing an
operational framework to develop a streamflow duration assessment method: A case
study from the Arid West United States. Water 13:3310.
Mazor, R. D., B. J. Topping, T.-L. Nadeau, K. M. Fritz, J. E. Kelso, R. A. Harrington, W. S. Beck, K. S.
McCune, A. 0. Allen, R. Leidy, J. T. Robb, G. C. L. David, and L. Tanner. 2021c. User
Manual for a Beta Streamflow Duratoin Assessment Method for the Western Mountains
of the United States. Version 1.0. Pages 116. Document No. EPA 840-B-21008, U.S.
Environmental Protection Agency, Washington, D.C. Available online:
https://www.epa.gov/system/files/documents/2021-12/beta-sdam-for-the-wm-user-
manual.pdf
McCune, K., and R. D. Mazor. 2019. Review of flow duration methods and indicators of flow
duration in the scientific literature: Arid Southwest. Pages 90; Available online:
https://ftp.sccwrp.org/pub/download/DOCUMENTS/TechnicalReports/1063_FlowMeth
odsReview.pdf. Southern California Coastal Water Research Project, Costa Mesa, CA.
Nadeau, T.-L. 2015. Streamflow Duration Assessment Method for the Pacific Northwest. Pages
36. Document No. EPA 910-K-14-001, U.S. Environmental Protection Agency, Region 10,
Seattle, WA. Available online: https://www.epa.gov/system/files/documents/2022-
03/sda m-pnw_nov-2015-final.pdf.
Nadeau, T.-L., S. G. Leibowitz, P. J. Wigington, J. L. Ebersole, K. M. Fritz, R. A. Coulombe, R. L.
Comeleo, and K. A. Blocksom. 2015. Validation of rapid assessment methods to
determine streamflow duration classes in the Pacific Northwest, USA. Environmental
Management 56:34-53.
37
-------
New Mexico Environment Department (NMED). 2011. Hydrology protocol for the
determination of uses supported by ephemeral, intermittent, and perennial waters.
Page 35. Surface Water Quality Bureau, New Mexico Environment Department,
Albuquerque, NM.
Oksanen, J., F. G. Blanchet, M. Friendly, R. Kindt, P. Legendre, D. McGlinn, P. R. Minchin, R. B.
O'Hara, G. L. Simpson, P. Solymos, M. H. M. Stevens, E. Szoecs, and H. Wagner. 2019.
vegan: Community Ecology Package.
Schumacher, C., and K. M. Fritz. 2019. Standard Operating Procedure: Verifying/Calibrating,
Deploying, Retrieving Stream Temperature, Intermittency, and Conductivity (STIC) Data
Loggers, and Downloading and Converting Data. EPA Report D-WQD-ECB-024-SOP-02.
Environmental Protection Agency, Washington, D.C.
Stoddard, J. L., A. T. Herlihy, D. V. Peck, R. M. Hughes, T. R. Whittier, and E. Tarquinio. 2008. A
process for creating multimetric indices for large-scale aquatic surveys. Journal of the
North American Benthological Society 27:878-891.
U.S. Army Corps of Engineers. 2010. Regional Supplement to the Corps of Engineers Wetland
Delineation Manual: Western Mountains, Valleys, and Coast Region (Version 2.0). Page
153. U.S. Army Engineer Research and Development Center, Vicksburg, MS: U.S. Army
Engineer Research and Development Center.
Links
Beta Streamflow Duration Assessment Method for the Western Mountains user manual:
https://www.epa.gov/streamflow~duration~assessment/beta~streamflow~duration~assessment~
method-western-mountains
Reginal Streamflow Duration Assessment Methods website: https://www.epa.eov/streamflow-
duration-assessment
Web application for the beta SDAM for WM: https://sccwrp.shinyapps.io/beta sdam win/
Western Mountain beta SDAM data and R code: https://doi.ore/10.2 >066
38
------- |