Author’s Accepted Manuscript
TheValley Bottom Extraction Tool (V-BET): a GIS
tool for delineating valley bottoms across entire
Jordan T. Gilbert, William W. Macfarlane, Joseph
To appear in: Computers and Geosciences
Received date: 13 May 2016
Revised date: 21 July 2016
Accepted date: 22 July 2016
Cite this article as: Jordan T. Gilbert, William W. Macfarlane and Joseph M.
Wheaton, TheValley Bottom Extraction Tool (V-BET): a GIS tool for delineating
valley bottoms across entire drainage networks, Computers and Geosciences,
This is a PDF file of an unedited manuscript that has been accepted for
publication. As a service to our customers we are providing this early version of
the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting galley proof before it is published in its final citable form.
Please note that during the production process errors may be discovered which
could affect the content, and all legal disclaimers that apply to the journal pertain.
The Valley Bottom Extraction Tool (V-BET): a GIS tool for delineating valley bottoms across entire
Jordan T. Gilbert1*, William W. Macfarlane1, Joseph M. Wheaton1
Department of Watershed Sciences, Utah State University, 5210 Old Main Hill, Logan, UT
*Corresponding author. email@example.com
The shape, size and extent of a valley bottom dictates the form and function of the associated
river or stream. Consequently, accurate, watershed-wide delineation of valley bottoms is
increasingly recognized as a necessary component of watershed management. While many
valley bottom delineation approaches exist, methods that can be effectively applied across
entire drainage networks to produce reasonably accurate results are lacking. Most existing
tools are generally designed to work using high resolution topography data (i.e. > 2m resolution
Digital Elevation Model (DEM)) and can only be applied over relatively short reach lengths due
to computational or data availability limitations. When these precise mapping approaches are
applied throughout drainage networks (i.e. 102 to 104 km), the computational techniques often
either do not scale, or the algorithms perform inconsistently. Other tools that produce outputs
at broader scale extents generally utilize coarser input topographic data to produce more
poorly resolved valley bottom approximations. To fill this methodology gap and produce
relatively accurate valley bottoms over large areas, we developed an algorithm that accepts
terrain data from one to 10 m with slope and valley width parameters that scale based on
drainage area, allowing for watershed-scale valley bottom delineation. We packaged this
algorithm in the Valley Bottom Extraction Tool (V-BET) as an open-source ArcGIS toolbox for
ease of use. To illustrate V-BET’s scalability and test the tool’s robustness across different
physiographic settings, we delineated valley bottoms for the entire perennial drainage network
of Utah as well as twelve watersheds across the interior Columbia River Basin (totaling 55 400
km) using 10 m DEMs. We found that even when driven with relatively coarse data (10 m
DEMs), V-BET produced a relatively accurate approximation of valley bottoms across the entire
watersheds of these diverse physiographic regions.
Keywords: Floodplain mapping, riverscape, valley setting, drainage network
analysis, riparian extent
The ability to identify and accurately delineate distinct fluvial landforms that make up
riverscapes is a crucial part of effective watershed management because these landforms can
dictate the form and function of these ecosystems riverscapes (Tarolli and Dalla Fontana,
2009). Valley bottoms are one of the most important fluvial landforms of critical geomorphic,
hydrologic and ecological significance is the valley bottom (Bendix and Hupp, 2000; Gallant and
Dowling, 2003; Hynes, 1975; Nardi et al., 2006). By definition, a valley bottom is comprised of
the active stream or river channel and the associated low-lying, contemporary floodplain (Fryirs
et al., 2015; Wheaton et al., 2015) (Figure 1). Valley bottoms can be bound by bedrock
hillslopes, or by other landforms along the margins of valleys such as alluvial fans, terraces
(abandoned floodplains), moraines or piedmonts (Fryirs et al., 2015). The shape, size and
extent of a valley bottom in relation to the associated stream channel’s width and position on
the valley floor are reflected in the degree of river confinement, which in turn dictates the
types of rivers that can form (Brierley and Fryirs, 2013) as well as their character, behavior and
capacity to adjust (Phillips, 2008; Phillips, 2010). As such, correctly delineating the valley
bottom and interpreting the valley-setting is critical to correctly interpreting river character and
behavior (Fryirs et al., 2015). Valley bottoms also represent the maximum possible extent of
riparian (Ilhardt et al., 2000; Macfarlane et al., In Revision). Riparian areas provide important
ecosystem functions (e.g., sediment sinks, hyporheic exchange) (Stanford, 2006; Stanford and
Ward, 1993; Wissmar, 2004) and support a disproportionate and diverse array of terrestrial and
aquatic organisms (Pollock et al., 1998). Consequently, accurate watershed-wide delineation of
valley bottoms is increasingly recognized as a necessary component of watershed management
(Bocco et al., 2005; Chowdary et al., 2009).
Many Geographic Information Systems (GIS)-based tools have been developed that either
derive valley bottoms from Digital Elevation Models (DEMs) (e.g. Nagel et al., 2014; Roux et al.,
2015; Straumann and Purves, 2008; Williams et al., 2000) or produce outputs from which a
valley bottom can be extracted (e.g. Dilts et al., 2010; Gallant and Dowling, 2003). Some
methods use coarse DEMs (10-30 m) to generally coarsely delineate valleys over large areas
(e.g. Gallant and Dowling, 2003; Nagel et al., 2014; Straumann and Purves, 2008). As high
resolution (e.g. > 2 m resolution) topographic data (e.g. LiDAR) have become increasingly
available (Passalacqua et al., 2015), new tools have emerged for use at the reach scale to
delineate valley bottoms and other landforms associated with fluvial settings (e.g. terraces and
fans) with a high degree of accuracy (e.g. Belmont, 2011; Dilts et al., 2010; McKean et al., 2009;
Stout and Belmont, 2014). Finally, perhaps the most robust, but also the most data intensive
and computationally expensive method is to use 1.5D (Brunner, 2010; Norman et al., 2003), 2D
(e.g. Bates et al., 1992) or 3D (e.g. MacWilliams, 2004) hydraulic models and map inundation
patterns, which can be interpreted as floodways or valley bottoms, associated with specific
discharges or floods of particular recurrence intervals (Merwade et al., 2008).
Prior to developing the Valley Bottom Extraction Tool (V-BET), we utilized various valley bottom
delineation tools and workflows in an attempt to determine their efficacy for delineating valley
bottoms across watershed (103 to 104 km2) or regional (104 to 106 km2) scale extents (Table 1).
Several tools we tested required digital terrain data of a higher resolution than was available
for entire watersheds or regions, and showed a high degree of sensitivity to the quality of that
topographic data (Cook and Merwade, 2009). Conversely, other tools that accommodate
coarser input data (e.g., 30 m DEMs) were designed to produce valley bottoms over large areas,
but their outputs were determined to be too coarse to be meaningful at reach scales, or
neglected headwater portions of the drainage network.
Most valley bottom delineation tools use either a slope algorithm designed to identify flat, low
lying features (e.g. Straumann and Purves, 2008), or a “flooding” algorithm in which a specified
depth is added to a relative (detrended) DEM (e.g. Kastens, 2008; Roux et al., 2015). A major
limitation to applying tools that use the “flooding” approach across entire watersheds is that
the method uses a single fill depth and therefore does not scale properly throughout the entire
drainage network. As such, an appropriate fill depth in the lower, mainstem portions of the
network results in exaggeration of the valley bottom extent toward the headwaters of the
watershed. Conversely, an appropriate fill depth in the headwaters of the network results in
underestimation of valley bottom extent in the mainstems. Approaches that use slope analyses
have similar scaling issues. Slope thresholds that are appropriate for delineating large, alluvial
valley bottoms will neglect confined, high-slope headwater valley bottoms. As such, standard
slope-based tools are often only applicable to broader valley bottoms of large rivers, and do not
produce outputs for entire drainage networks across the full range of stream orders.
The objective of this paper is to present and demonstrate a slope-based valley bottom
delineation approach that operates as a function of drainage area and scales outputs based on
their location within a watershed. This relatively modest improvement on the slope approach
suggested by Straumann and Purves (2008) represents a computationally elegant technique to
scale a straight-forward geoprocessing task in a pragmatic way. The need for such an approach
is underscored by the reality that resource managers need valley bottom mapping across large
regional (103 to 105 km2) extents. We illustrate V-BET’s utility by delineating valley bottoms at a
large range of scales (e.g. watersheds of ≤500 km2 to regions ≥ 20 000 km2) using coarse (10 m),
nationally available DEMs.
V-BET Algorithm and Tool
Although stream characteristics vary greatly depending on the biophysical settings of different
regions, generalizations can be made within biophysical settings as to how stream character
changes longitudinally (Montgomery and Buffington, 1997; Vannote et al., 1980). As such, V-
BET was based on the assumptions that: (1) valley bottom width is a function of upstream
drainage area (DA), with wider valley bottoms corresponding, crudely, to larger upstream DA
(Montgomery, 2002; Nardi et al., 2006), (2) the average slope of a valley bottom is related to
upstream DA; the larger the DA, the flatter the valley bottom (McNamara et al., 2006;
Montgomery, 2001; Schorghofer and Rothman, 2002; Tucker and Bras, 1998; Willgoose et al.,
1991), and (3) valley bottoms are relatively flat areas with margins often defined by abrupt
changes in slope (Gallant and Dowling, 2003) (see Figure 2).
The V-BET Algorithm
The V-BET algorithm proceeds according to seven steps (Figure 3). Step 1: Subset drainage
network using flow accumulation raster. V-BET requires a flow accumulation raster in which
the value for each cell represents the upstream DA (in km2) similar to Jenson and Domingue
(1988), which is used to segment the supplied drainage network by varying DA thresholds
(Figure 3.1). DA-based drainage network segmentation values are a user-defined parameter
and can be adjusted appropriately for the region of interest. In our application areas, streams
with DA of less than 25 km2 were generally confined headwater streams, those streams with DA
in the 25 to 250 km2 range were in a transition zone where valleys widened and slopes
decreased and those with DAs greater than 250 km2 were generally larger rivers or tributaries
in alluvial valleys. Therefore, when running V-BET, portions of the network for which the
upstream DA was greater than 250 km2 were classified as “large”, portions between 25 and 250
km2 were classified as “medium”, and locations with upstream DA of less than 25 km2 were
considered the “small” portion of the network. The three resulting portions of the input
network will have different maximum valley width and slope parameters associated with them
(Figure 3.2). In principle, this scale-dependent buffering could be implemented as a continuous
function of drainage area, but we sought to determine if a computationally simpler solution
Step 2: Buffer drainage network by varying maximum valley bottom widths. The next
procedure that V-BET performs is to buffer the drainage network by user specified values that
represent the maximum valley bottom widths of the large, medium and small portions of the
network (Figure 3.3). These width values can be obtained for the watershed of interest from
empirically sampling the maximum valley widths using a “measure” tool in any GIS based
program. The resulting buffers serve as a bounding extent for the subsequent valley bottom
delineations. In principle, this scale-dependent buffering could be implemented as a
continuous function of drainage area, but we sought to determine if a computationally simpler
solution would suffice.
Step 3 and 4: Derive slope raster and subset slope raster by varying DA thresholds. Next, the
tool performs a slope analysis on the input DEM in which an output raster is produced with
slope (in degrees) calculated for each pixel (Figure 3.4). Within each of the three (small,
medium and large) buffers created in the previous step, values below specified slope thresholds
are retained, and the remainder of the cells are removed (Figure 3.5). To determine the
appropriate slope thresholds to associate with each DA threshold that we used, we reviewed
published studies investigating the link between slope and DA (McNamara et al., 2006;
Montgomery, 2001; Schorghofer and Rothman, 2002; Tucker and Bras, 1998; Willgoose et al.,
1991). To further assess appropriate values for our application areas, we selected random
points to plot the relationship between DA and slope within the valley bottoms of two
watersheds within our study areas: the Weber watershed of Northern Utah and the Salmon
watershed of Central Idaho (5000 and 7500 points respectively). The Weber River, in general,
is a partly confined to unconfined river passing through alluvial valleys between mountains, and
the valley bottom slopes are relatively low. Conversely, the Salmon River is a mostly confined
river with bedrock controlled valleys and steep gradients. The best fit regression line for the DA
and slope data in both cases was a logarithmic equation. For the Weber this equation was:
For the Salmon it was:
where S is slope in degrees and DA is upstream DA in square kilometers. Comparing this
information from these two contrasting watersheds with the previously conducted analyses led
us to select slope thresholds for our applications of 12 degrees for the “small” portion of the
network, seven degrees for the “medium” portion of the network, and five degrees for the
“large” portion of the network. These values are not universally applicable, but rather
represent a potential upper limit relevant to the interior western U.S.
Step 5, 6 and 7: Convert to polygons, merge and clean. After the slope rasters have been
subset by the threshold values, they are converted into polygons (Figure 3.6). At this step,
there are often small polygons outside of the actual valley bottom in addition to the polygon
that represents the valley bottom. This is a result of relatively low slope areas that occur
outside of the valley bottom (beyond the extent of whatever geographic feature is confining
the valley bottom) but are still within the buffer width specified while running the tool. These
polygons, which are not associated with the valley bottom are removed. This results in an
initial valley bottom polygon which then undergoes cleaning steps. Polygons within a specified
distance of one another are aggregated together. Individual polygons smaller than a threshold
size that the user specifies are removed, and holes smaller than a selected threshold size are
filled. The resulting polygons are then smoothed, and the final output produced (Figure 3.7).
V-BET was written as a Python (version 2.7) script to leverage ArcGIS 10.2 and 10.3 functionality
through the ArcPy python module. The script was incorporated into a standard ArcMap tool
GUI for ease of use, with dependency on the “spatial analyst” extension of ArcGIS. For more
detailed instructions on running V-BET in ArcMap, and access to the source-code, see
Inputs and Preprocessing
V-BET requires two inputs: a DEM and a polyline drainage network (Table 2). The tool was
designed and tested to provide accurate outputs using either relatively coarse, nationally
available data (e.g. 10 meter DEMs from the USGS National Elevation Dataset (NED)) or high
resolution topographic data (e.g. 1 m airborne LiDAR). V-BET accepts as an input either
cartographically or hydrographically derived drainage networks. Cartographically derived
networks are traced using imagery at a set spatial scale to accurately portray the location of the
channel. An example of a cartographically derived drainage network is the USGS’s National
Hydrography Dataset (NHD) 1:24 000 dataset, which was derived using the USGS topographic
7.5-minute quadrangles. Hydrographically derived drainage networks are generated using a
DEM, and the greater the resolution of the DEM, the more accurate the resulting drainage
network will be. An example of this type of network is the NHDPlus Version 2 1:100 000 scale
network. Before applying V-BET, the drainage network input should be reduced to the portions
of the network for which the user is interested in deriving a valley bottom. For example, canals
and pipelines should generally not be included as such features do not form valley bottoms. In
our testing, we reduced our input network to the perennial network, removing canals,
pipelines, and intermittent and ephemeral streams.
The third, optional input to V-BET is a flow accumulation raster (in units of km2). Because V-
BET’s algorithm utilizes DA values, a flow accumulation raster is automatically derived from the
DEM if this optional input is left blank. The automatically derived flow accumulation raster is
saved to enable its use as an optional input in subsequent runs of the tool if the user wishes to
alter parameters. For cases in which the entire upstream watershed is not included in the area
of interest (i.e. the input DEM), using the input DEM to derive a flow accumulation will result in
false DA values. In such cases, a flow accumulation for the entire watershed is necessary to
derive accurate DA values, and must be provided by the user (meaning that the input is no
Evaluating, Editing and Validation
For most applications, we evaluate, manually edit and validate valley bottom polygons at a
scale of 1:10 000 across the entire associated drainage network. We recommend using a 3D
viewer, like Google Earth to evaluate the valley bottom outputs. KMZ files of V-BET outputs can
be imported into Google Earth, where the underlying DEM shows the terrain in 3D with the
valley bottom polygons overlain. We visually assess the overall accuracy of the valley bottom
polygons by determining if the valley bottom boundaries matched the breaks in slope where
valley bottoms transition into hillslopes or other confining features. The DEMs used in Google
Earth are fairly coarse (30 meters or greater depending on location), so this method provides a
preliminary evaluation, primarily informed by cues from aerial imagery context, which serve to
identify gross errors in the valley bottom polygons (Figure 4A). For manual editing, we typically
use high resolution orthoimagery from the National Agriculture Imagery Program (NAIP) (Figure
4B) as well as hillshades (Figure 4C) derived from the 10 m resolution input DEM to identify
landforms that bound valley bottoms. These features identified in the imagery and hillshade
are used to adjust the line work in the V-BET generated valley bottom polygons as necessary.
After editing the V-BET valley bottoms generated by V-BET, we validate the accuracy of the final
products using several GIS layers in ArcMap. These layers include: Soil Survey Geographical
(SSURGO) hydric soils, Federal Emergency Management Agency (FEMA) floodways, National
Wetland Inventory (NWI) riverine wetlands and digitized riparian vegetation (Utah Water
Related Land Use dataset). These layers represent features that are contained within valley
bottom extents and should, therefore, fall within the valley bottom polygon produced from V-
BET. Additionally, slope rasters maps are generated and are used to identify floodplains and hill
slopes and to analyze slope distributions within the valley bottom extents (Figure 5). Lastly, the
V-BET output is also validated using USGS 1:50 000 geologic maps. In alluvial valleys, geologic
maps depict an alluvial fill which, by definition, is sediment that has been eroded, reshaped,
and deposited by water (Monroe and Wicander, 2011). The presence of alluvium, therefore,
serves as strong evidence that the area has been influence by a flowing stream or river at some
point in time (Figure 6).
Illustrative Applications and Interpretations
Our illustration of valley bottom delineation using V-BET focused on the perennial drainage
networks of the state of Utah (25 700 km of streams), as well as twelve basins of fisheries
management concern within the interior Columbia River Basin (CRB), including the John Day
and Grand Ronde rivers in Oregon, the Tucannon, Entiat, and Wenatchee rivers and Asotin
Creek in Washington, and the upper Salmon, Yankee Fork, Lemhi and Clearwater river in basins,
Idaho (totaling 29 690 km of streams). The CRB effort was part of CHaMP (Columbia Habitat
Monitoring Program, http://champmonitoring.org) which tracks the status and trend of
anadromous salmonid habitat throughout the interior Columbia River Basin CRB (Bouwes et al.,
Utah is a physiographically diverse landscape covering 219 808 km2 that range from alpine
meadows to desert canyons that support a wide range of valley types (bedrock canyons, alluvial
valleys, etc.). Utah includes three primary physiographic regions, each with unique topographic,
geologic, and geomorphic characteristics: the Colorado Plateau, the Basin and Range, and the
Middle Rocky Mountains (USGS, 2003). Elevations in Utah range from 664 m at Beaver Dam
Wash in the southwestern corner of the state to 4,123 m at King’s Peak in the Uinta Mountains
(Utah State, 2016). Utah provides an ideal range of landscapes across which the robustness of a
semi-automated valley bottom delineation approach can be tested. Similarly, the CRB is
comprised of the Columbia Plateau Physiographic Province (USGS, 2003), which includes a
diverse range of landscapes, including broad plateaus and, deep canyons, steep mountains and
the rolling hills and deep soils of Washington and Oregon’s Palouse region (Figure 7).
V-BET was able to effectively delineate valley bottoms for the entire perennial drainage
network of Utah (25 700 km of streams), as well as the John Day, Grand Ronde, Tucannon,
Asotin, Entiat, Wenatchee, Upper Salmon, Yankee Fork, Lemhi and Clearwater basins within the
interior CRB (totaling 29 690 km of streams) (Figure 8, Table 3). Within our analysis
watersheds, summary metrics associated with valley bottoms varied significantly. The percent
of watersheds area that valley bottoms comprised ranged from 2.1% for Utah to 10.4% in the
upper Grand Ronde (Table 3). Although Utah had the lowest proportion of valley bottoms, it
had the largest effective valley bottom width (185 m), suggesting that the low proportion of
valley bottoms is a result of the low (0.12) perennial drainage density. The high drainage
density (0.8) and large valley bottom width (129 m) of the upper Grand Ronde result in the
largest proportion of valley bottom within any of the watersheds. Conversely, although the
Asotin has very high drainage density (0.85), its narrow effective valley bottom width (32 m)
results in a low proportion of valley bottom (2.7%) within the watershed. The variation in these
metrics illustrates the diversity of watersheds that V-BET delineated, and are useful for
characterizing and comparing different watersheds.
During V-BET’s development, while applying other valley bottom delineation methods, we
found the Fluvial Corridor Tool (FCT) (Roux et al., 2015) and the Valley Confinement Algorithm
(VCA) (Nagel et al., 2014) to be the most capable of the existing methods for watershed scale
delineation. However, FCT uses the “flooding” method and therefore is subject to the
associated scaling limitation, and VCA does not produce outputs for confined valley settings. To
overcome FCT’s limitation, we ran the tool twice for each test watershed, once using settings
for the lower portions of the network and once for the headwater portions of the network.
These two outputs were manually merged in convenient locations within the transition zone
where valley width decreased and slope increased. Merging the polygons in these locations
minimized the abruptness of the transitions between the two outputs, however they the
polygons still required manual smoothing, which was a time consuming editing process that
rendered the method inefficient at regional scales that required editing numerous watersheds.
V-BET was able to produce valley bottom polygons throughout the entire drainage network for
all of the study watersheds in which the tool was applied, and was able to delineate valley
bottoms in unconfined, partly confined and confined valley settings (Figure 8). Figure 9
compares the outputs of a single run of V-BET with two runs of FCT. Figure 9A shows the
headwater exaggeration that occurred using settings in FCT that appropriately delineated
broader alluvial valley bottoms. Figure 9B demonstrates the underestimation of these broader
alluvial valley bottoms that occurred using settings in FCT that appropriately delineated
headwater valley bottoms. A single run of V-BET produced an output that scaled appropriately
throughout the drainage network (Figure 9C).
Validation and Editing Examples
As there is no true measure of “valley bottomness” (Gallant and Dowling, 2003) the accuracy of
the outputs is difficult to validate statistically. However, the distinct characteristics of valley
bottoms were validated using various GIS datasets (Figure 4, Figure 5, Figure 6). Using slope
rasters, we compared slope distributions within the entire watershed to the distributions within
the valley bottom (Figure 5). As expected, the distribution of slopes within the valley bottom is
much narrower than throughout the remainder of the watershed and is skewed toward zero,
with occurrence drastically dropping as slope increases. Using geologic maps (Figure 6), we
compared the area of alluvial units to other units for both the entire map area and the valley
bottom. Within the portion of the watershed that the map covered, alluvium comprised only
3.6% of the overall area, whereas within the valley bottom it made up 67.5%. Because
landforms exist that are fluvial in nature but are not part of the valley bottom (e.g., terraces),
some alluvium is found outside of valley bottoms. Additionally, as mentioned in Section 2.3,
the V-BET outputs were compared against GIS datasets representing features found within
valley bottoms. This validation method suggests that V-BET’s output is capturing landscape
features that should be expected to fall within valley bottom extents such as hydric soils,
riparian vegetation, alluvial deposits and areas that are flat relative to the rest of the
When using other tools for watershed scale valley bottom delineation, errors in the output
required manual editing throughout the entire network. In contrast, the V-BET output required
only minor correction in headwater and low order streams, and errors in the lower portions of
the network were easily identifiable. Table 4 summarizes the effort required to edit V-BET
outputs to achieve good spatial precision at a scale of 1:10 000 for a portion of our study area.
On average, we were able to perform valley bottom edits for 419 km of streams per hour.
Because output errors requiring correction are systematic and easily recognized, manual editing
to correct them took less time than it did for any of the other tools we used. For example, in
the complex topography of the Blacks Fork and Smiths Fork drainages in the Uinta Mountains of
Utah (622 km2), where there is also a high concentration of headwater streams, manual editing
at a 1:10 000 scale with previous approaches took up to 12 hours, while editing the V-BET
output took only four hours.
Our motivations for developing valley bottom mapping across broad spatial extents with
adequate reach-scale resolution and accuracy were fundamentally scientific, and reinforced by
pressing resource management concerns. From an earth-sciences and geomorphic perspective,
the valley bottom is one of the most fundamental building blocks of the landscape and the
defining extent of riverscapes (Fryirs et al., 2015; Hynes, 1975). From a resource management
perspective, valley bottoms are a focal point for conflicting land uses. As the flattest parts of
the landscape, they are easily farmed and developed, creating conflict between human land
uses and existing ecosystems (Burby and French, 1981; Mount, 1995). V-BET allowed us to map
the extent of valley bottoms across entire drainage networks at an unprecedented scale,
providing basic mapping information for large regions in which no such information previously
existed. V-BET delineated valley bottoms for watersheds as large as 20 000 km2 in a single run,
which we could not previously achieve with any of the 11 existing tools we utilized.
For this application we used coarse, nationally available input datasets (NHD 1:24 000 scale
hydrography and NED 10 meter DEMs) with which we achieved watershed scale valley bottom
delineations that were relatively accurate, requiring relatively modest manual editing despite
the coarseness of the input data (Table 4). Nevertheless, V-BET can also be run with higher
resolution inputs to achieve more precise and more refined valley bottom mapping. While
higher resolution inputs can produce more precise valley bottom outputs, this does not equate
to significant differences in the valley bottom summary metrics (Figure 10). Highly accurate
representations of valley bottoms (i.e. spatially reflecting the true valley bottom at fine scales)
are only attainable by manually editing the output polygons at a desired scale for the specific
application. Finer scale editing results in more precise valley bottoms, but is more time
consuming, and its effects on summary metrics are also minimal. For example, in the Entiat
watershed the valley bottom comprised 2.6% of the watershed regardless of whether or not it
was edited, and the effective valley width only changed from 47 to 48 m (Table 3). The average
difference in area between edited and unedited outputs was 4.3 km2 (Table 4). Therefore, the
scale at which the output polygons are edited should be dictated by the purpose that the valley
bottom delineation will serve, as well as the time and resources available to produce these
Limitations of rRunning V-BET with 10m DEMs
As with any semi-automated process, Running V-BET run with 10 m data does have limitations.
In broad valleys with gradual slopes and minimal topographic complexity, very limited portions
of the clipped slope raster are excluded before being converted to polygons. Therefore, the
output is roughly the same as the buffer used to clip the slope raster. Similarly, 10 m DEMs
generally do not have the spatial resolution to adequately resolve valleys bottoms in headwater
streams (e.g. a 10 m resolution pixel cannot resolve a 5 m wide valley bottom). Consequently,
V-BET outputs in these settings are rough estimates of valley bottoms and resemble fixed width
drainage network buffers. Another limitation occasionally occurs with 10 m DEM inputs in
areas where there are flat terraces elevated above either or both sides of the valley bottom. If
the coarse 10 m DEMs does not resolve terrace risers (i.e., the slope between the valley bottom
and the terrace) sufficiently, erroneous connections form between the valley bottoms and the
terraces, and the terraces are not removed from the output. In these cases, a higher resolution
input DEM may improve both precision and accuracy.
V-BET and Other Alluvial Landforms
V-BET is fundamentally focused on delineating valley bottoms however, the valley can include
other landforms of alluvial origin (e.g. alluvial fans, or terraces). While the valley bottom of a
river does not include tributary alluvial fans formed by tributaries, such the fans may comprise
part of the tributary valley bottom of the tributary, in which case it would be included in the
output. Valleys also include landforms of glacial origin (e.g. lateral and terminal moraines),
hillslope origin (e.g. talus and colluvial fans) and even lacustrine origin (e.g. dry lake beds).
Anthropogenic margins (e.g. levees, rip-rapped channel margins, flood control walls, retaining
walls), anthropogenic landforms (e.g. any grading or development) and anthropogenic
structural elements (e.g. bridge piers, etc.) also make up and occupy the valley bottom. The
extent to which V-BET excludes other natural landforms and captures anthropogenic landforms,
margins and structural elements depends on their magnitude relative to the resolution of the
Valley bottom delineations are an important management tool as they define the boundaries
within which streams and rivers can exist and adjust. In order to be useful for broader scale
landscape management, valley bottom delineations must cover broad areas such as entire
watersheds or multiple congruent watersheds over large regions. However, high resolution
terrain data used to drive the majority of valley bottom delineation approaches is limited in
both scale and availability. Additionally, because most valley bottom delineation tools are
designed for use at the reach scale, their outputs fail to scale appropriately throughout
drainage networks. V-BET, in addition to accepting high resolution terrain data, can utilize
widely available 10 m DEMs to perform valley bottom delineations. With V-BET, parameters
also scale with DA, meaning that valley bottoms can be delineated for entire watersheds in a
single run of the tool., This allowed us to efficiently delineate thus allowing for efficient
delineation of valley bottoms over large and diverse physiographic regions, providing important
scientific and resource management information that has not existed until now.
Table 1 A Partial list of existing, published and/or available tools and/or algorithms for
deriving valley bottoms that were tested as part of the development of V-BET.
Method or Tool
Data Requirements & Key Parameters
Stream centerline, stream banks, flow paths,
Williams et al. (2000) Method
Watershed boundary, drainage network,
(Williams et al., 2000)
DEM, stream gauge information
Floodplain Mapping Tool (FMT)
DEM, average valley width
(Stout and Belmont, 2014)
DEM, raster drainage network
(Dilts et al., 2010)
DEM, raster drainage network
(Dilts et al., 2010)
River Bathymetry Toolkit
(McKean et al., 2009)
(Gallant and Dowling, 2003)
Object-based DEM classification
(Straumann and Purves, 2008)
DEM, NHD+ streamlines and waterbodies,
precipitation (PRISM), watershed boundaries
(Nagel et al., 2014)
Fluvial Corridor Toolbox
DEM, drainage network polyline
(Roux et al., 2015)
Table 2 The nationally available, free input data used to run V-BET across Utah and twelve
watersheds in the interior Columbia River Basin (CRB).
The streams and/or rivers along which valley bottoms
should be are derived
USGS National Hydrography Dataset Cartographic
1:24 000 scale http://nhd.usgs.gov/
Topography that adequately resolves presence of valley
bottom and can differentiate it from other valley feature
(e.g. terraces, fans and moraines)
USGS National Elevation Dataset 10 m Digital
Table 3 - Summary results and statistics of Utah and twelve watersheds in the interior
Columbia River Basin (CRB) for which V-BET was tested and demonstrated.
% Area in
km/km2 = km-1
State of Utah
Table 4 - Summary of the editing process required to produce V-BET outputs at a spatial
precision at a of 1:10 000 scale using V-BET outputs for a portion for different watersheds of
the study area.
Figure 1 - Conceptual diagram showing an example of a valley bottom and the confining
margins within a mountain setting of the interior western U.S.
Figure 2 - Conceptual diagram of headwaters, transition zone and larger alluvial rivers and
their associated valley bottom widths and slopes. Examples taken from the East Canyon
watershed within the Weber watershed, Utah (12N 450 605m E 4 523 443m N).
Figure 3 - Conceptual diagram of the V-BET workflow showing the input data required and the
seven processing steps.
Figure 4 - Examples of topographic and aerial photography context used to visually
interrogate and edit valley bottom outputs during the manual editing process. A. Google
Earth aerial imagery and topographic context. B. NAIP high resolution imagery. C. Hillshade
derived from 10 m DEM. Examples taken from a portion of the John Day watershed, Oregon
(11N 358 220m E 4 950 087m N).
Figure 5 - Slope analysis used to validate the V-BET outputs, showing slope distributions for
the entire watershed, and the valley bottom extent of the watershed. Example taken from a
portion of the Weber watershed, Utah (12N 426 347m E 4 579 152m N).
Figure 6 - USGS 1:50 000 geologic map used to validate the V-BET outputs, showing the
occurrence of alluvium versus other geologic units for the entire map and for only the
portions within the valley bottom. Example taken from a portion of the Weber watershed,
Utah (12N 441 041m E 4 544 468m N).
Figure 7 - Location map and physiographic context (using Level III Ecoregions) for areas in
which the V-BET is illustrated in this paper including the 12 example watersheds (51 423 km2)
in the interior Columbia River Basin (CRB) and the entire state of Utah (219 890 km2).
Figure 8 – Valley bottom delineations of the perennial drainage networks of Utah, and twelve
watershed throughout the interior Columbia River Basin (CRB).
Figure 9 - Illustration of the contrast in V-BET simulation versus FCT simulations with wide
valley (A) and narrow valley (B) settings. Example from a portion of the John Day watershed,
Oregon (11N 358 220m E 4 950 087m N).
Figure 10 – Shows the influence that DEM spatial resolution has on V-BET outputs. The top
panel (A) shows both outputs overlaid on a 10 m input DEM, whereas the bottom panel (B)
shows both outputs overlaid on a 1 m LiDAR input DEM. Example from a portion of the
Tucannon watershed, Washington (11N 419 713m E 5 150 675m N).
Documentation, instructions for running the tool and a downloadable toolbox can be found at
This work was supported by the U.S. Department of the Interior Bureau of Land Management
(USU Award No. 151010), the Utah Department of Natural Resources Endangered Species
Mitigation Fund (USU Award No. 140600), Utah Division of Wildlife Resources Pittman and
Robertson Fund (USU Award No. 150736), the Snake River Salmon Recovery Board through Eco
Logical Research (USU Award No. 200239) and the Bonneville Power Administration (BPA
project numbers: CHaMP 2011-006 and ISEMP 2013-017), as part of the Columbia Habitat
Monitoring Program (http://champmonitoring.org) through a sub-award from Eco Logical
Research (USU Award No. 150737). We thank Peter McHugh and Nate Hough-Snee (USU) for
their input in the development of V-BET and assistance with the manuscript. Martha Jensen,
Chris Brown, Joshua Gilbert and Chalese Hafen provided GIS support. We thank two anonymous
reviewers for their thoughtful reviews of this paper.
Bates, P.D., Anderson, M.G., Baird, L., Walling, D.E., Simm, D., 1992. Modelling floodplain flows
using a two-dimensional finite element model. Earth Surf. Process. Landf., 17, 575-588.
Belmont, P., 2011. Floodplain width adjustments in response to rapid base level fall and
knickpoint migration. Geomorphology, 128(1), 92-102.
Bendix, J., Hupp, C.R., 2000. Hydrological and geomorphological impacts on riparian plant
communities. Hydrological processes, 14(16), 2977-2990.
Bocco, G., Velázquez, A., Siebe, C., 2005. Using geomorphologic mapping to strengthen natural
resource management in developing countries. The case of rural indigenous
communities in Michoacan, Mexico. Catena, 60(3), 239-253.
Bouwes, N., J. Moberg, Weber, N., Bouwes, B., Bennett, S., Beasley, C., Jordan, C.E., Nelle, P.,
Polino, M., Rentmeester, S., Semmens, B., Volk, C., Ward, M.B., White, J., 2011.
Scientific protocol for salmonid habitat surveys within the Columbia Habitat Monitoring
Program, Integrated Status and Effectiveness Monitoring Program, Wauconda, WA.
Brierley, G.J., Fryirs, K.A., 2013. Geomorphology and river management: applications of the
river styles framework. John Wiley & Sons.
Brunner, G.W., 1995. HEC-RAS River Analysis System. Hydraulic Reference Manual. Version 1.0.
US Army Corps of Engineers, Davis, CA.
Brunner, G.W., 2010. HEC-RAS river Analysis System User's Manual, Version 4.1, US Army Corps
of Engineers, Hydrologic Engineering Center, Davis, CA.
Burby, R.J., French, S.P., 1981. Coping with floods: the land use management paradox. Journal
of the American Planning Association, 47(3), 289-300.
Chowdary, V., Ramakrishnan, D., Srivastava, Y., Chandran, V., Jeyaram, A., 2009. Integrated
water resource development plan for sustainable management of Mayurakshi
watershed, India using remote sensing and GIS. Water resources management, 23(8),
Cook, A., Merwade, V., 2009. Effect of topographic data, geometric configuration and modeling
approach on flood inundation mapping. Journal of Hydrology, 377(1), 131-142.
Dilts, T.E., Yang, J., Weisberge, P.J., 2010. Mapping Riparian Vegetation with Lidar Data,
ArcUser, pp. 18-21.
Fryirs, K.A., Wheaton, J.M., Brierley, G.J., 2015. An approach for measuring confinement and
assessing the influence of valley setting on river forms and processes. Earth Surface
Processes and Landforms, 41(5), 701-710.
Gallant, J.C., Dowling, T.I., 2003. A multiresolution index of valley bottom flatness for mapping
depositional areas. Water Resources Research, 39(12).
Hynes, H.B.N., 1975. Edgardo Baldi memorial lecture. The stream and its valley. Verhandlungen
der Internationalen Vereinigung fur theoretische und angewandte Limnologie, 19, 1-15.
Ilhardt, B.L., Verry, E.S., Palik, B.J., 2000. Defining riparian areas, Forestry and the riparian zone,
Orono, Maine, pp. 7-14.
Jenson, S.K., Domingue, J.O., 1988. Extracting topographic structure from digital elevation data
for geographic information system analysis. Photogrammetric engineering and remote
sensing, 54(11), 1593-1600.
Kastens, J.H., 2008. Some New Developments on Two Separate Topics: Statistical Cross
Validation and Floodplain Mapping. ProQuest.
Macfarlane, W.W., Gilbert, J.T., Jensen, M.L., Gilbert, J.D., Hough-Snee, N., McHugh, P.A.,
Wheaton, J.M., Bennett, S.N., In Revision. Riparian vegetation as an indicator of riparian
condition: detecting departures from historic condition across the North American
West. Journal of Environmental Management.
MacWilliams, M.L., 2004. Three-dimensional hydrodynamic simulation of river channels and
floodplains. PhD, Stanford University.
McKean, J., Nagel, D., Tonina, D., Bailey, P., Wright, C.W., Bohn, C., Nayegandhi, A., 2009.
Remote sensing of channels and riparian zones with a narrow-beam aquatic-terrestrial
LIDAR. Remote Sensing, 1(4), 1065-1096.
McNamara, J.P., Ziegler, A.D., Wood, S.H., Vogler, J.B., 2006. Channel head locations with
respect to geomorphologic thresholds derived from a digital elevation model: A case
study in northern Thailand. Forest Ecology and Management, 224(1), 147-156.
Merwade, V., Cook, A., Coonrod, J., 2008. GIS techniques for creating river terrain models for
hydrodynamic modeling and flood inundation mapping. Environmental Modelling &
Software, 23(10-11), 1300-1311.
Monroe, J., Wicander, R., 2011. The changing earth: exploring geology and evolution. Cengage
Learning, Belmont, CA.
Montgomery, D.R., 2001. Slope distributions, threshold hillslopes, and steady-state topography.
American Journal of Science, 301(4-5), 432-454.
Montgomery, D.R., 2002. Valley formation by fluvial and glacial erosion. Geology, 30(11), 1047-
Montgomery, D.R., Buffington, J.M., 1997. Channel-reach morphology in mountain drainage
basins. Geological Society of America Bulletin, 109(5), 596-611.
Mount, J.F., 1995. California rivers and streams: the conflict between fluvial process and land
use. Univ of California Press.
Nagel, D.E., Buffington, J.M., Parkes, S.L., Wenger, S., Goode, J.R., 2014. A landscape scale valley
confinement algorithm: delineating unconfined valley bottoms for geomorphic, aquatic,
and riparian applications, U.S. Dept of Agriculture, Forest Service, Rocky Mountain
Research Station, Fort Collins, CO.
Nardi, F., Vivoni, E.R., Grimaldi, S., 2006. Investigating a floodplain scaling relation using a
hydrogeomorphic delineation method. Water Resources Research, 42(9).
Norman, N.S., Nelson, J.E., Zundel, A.K., 2003. Improved Process for Floodplain Delineation
from Digital Terrain Models. J. Water Resour. Plan. Manage.-ASCE, 129(5).
Passalacqua, P., Belmont, P., Staley, D., Simley, J., Arrowsmith, J.R., Bodee, C., Crosby, C.,
DeLong, S.B., Glenn, N.F., Kelly, S., Lague, D., Sangireddy, H., Schaffrath, K., Tarboton, D.,
Wasklewicz, T., Wheaton, J., 2015. Analyzing high resolution topography for advancing
the understanding of mass and energy transfer through landscapes: A review. Earth
Science Reviews, 148, 174-193.
Phillips, J.D., 2008. Geomorphic controls and transition zones in the lower Sabine River.
Hydrological Processes, 22(14), 2424-2437.
Phillips, J.D., 2010. The job of the river. Earth Surface Processes and Landforms, 35(3), 305-313.
Pollock, M.M., Naiman, R.J., Hanley, T.A., 1998. Plant species richness in riparian wetlands-a
test of biodiversity theory. Ecology, 79(1), 94-105.
Roux, C., Alber, A., Bertrand, M., Vaudor, L., Piégay, H., 2015. “FluvialCorridor”: A new ArcGIS
toolbox package for multiscale riverscape exploration. Geomorphology, 242, 29-37.
Schorghofer, N., Rothman, D.H., 2002. Acausal relations between topographic slope and
drainage area. Geophysical Research Letters, 29(13).
Stanford, J., 2006. Landscapes and riverscapes, Methods in stream ecology. Academic Press,
New York, pp. 3-22.
Stanford, J.A., Ward, J., 1993. An ecosystem perspective of alluvial rivers: connectivity and the
hyporheic corridor. Journal of the North American Benthological Society, 12(1), 48-60.
Stout, J.C., Belmont, P., 2014. TerEx Toolbox for semiautomated selection of fluvial terrace and
floodplain features from lidar. Earth Surface Processes and Landforms, 39(5), 569-580.
Straumann, R.K., Purves, R.S., 2008. Delineation of valleys and valley floors, 5th International
Conference, GIScience, Park City, UT, pp. 320-336.
Tarolli, P., Dalla Fontana, G., 2009. Hillslope-to-valley transition morphology: new opportunities
from high resolution DTMs. Geomorphology, 113(1), 47-56.
Tucker, G.E., Bras, R.L., 1998. Hillslope processes, drainage density, and landscape morphology.
Water Resources Research, 34(10), 2751-2764.
USGS, 2003. A Tapestry of Time and Terrain: The Union of Two Maps - Geology and
Topography, Physiographic Regions.
Utah State, U.E., 2016. Physiography of Utah. Utah State University, Logan, UT.
Vannote, R.L., Minshall, G.W., Cummins, K.W., Sedell, J.R., Cushing, C.E., 1980. The river
continuum concept. Canadian journal of fisheries and aquatic sciences, 37(1), 130-137.
Wheaton, J.M., Fryirs, K.A., Brierley, G., Bangen, S.G., Bouwes, N., O'Brien, G., 2015.
Geomorphic mapping and taxonomy of fluvial landforms. Geomorphology, 248, 273-
Willgoose, G., Bras, R.L., RodriguezIturbe, I., 1991. A physical explanation of an observed link
areaslope relationship. Water Resources Research, 27(7), 1697-1702.
Williams, W.A., Jensen, M.E., Winne, J.C., Redmond, R.L., 2000. An automated technique for
delineating and characterizing valley-bottom settings. Environmental monitoring and
assessment, 64(1), 105-114.
Wissmar, R.C., 2004. Riparian corridors of Eastern Oregon and Washington: functions and
sustainability along lowland-arid to mountain gradients. Aquatic Sciences, 66(4), 373-
Valley bottom delineation is crucial to riverine ecosystem management
Existing delineation tools require LiDAR data to produce reach scale outputs
We developed V-BET to accept 10 meter data to produce watershed scale outputs
We demonstrate its utility by delineating valley bottoms over large, diverse regions