See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/251529648

Channel Pattern and River Floodplain Dynamics in Forested Mountain River

Systems

Article  in  Geomorphology · August 2006

DOI: 10.1016/j.geomorph.2006.01.030

CITATIONS READS

146 539

5 authors, including:

Timothy Beechie Michael M. Pollock

National Oceanic and Atmospheric Administration National Oceanic and Atmospheric Administration
117 PUBLICATIONS   4,673 CITATIONS    58 PUBLICATIONS   3,429 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Recovery of Pacific salmonids View project

Elwha River Dam Removal Monitoring View project

All content following this page was uploaded by Michael M. Pollock on 15 February 2019.

The user has requested enhancement of the downloaded file.

 Geomorphology 78 (2006) 124 – 141
www.elsevier.com/locate/geomorph

Channel pattern and river-floodplain dynamics in forested

mountain river systems
Timothy J. Beechie ⁎, Martin Liermann 1 , Michael M. Pollock 1 ,
Sarah Baker 1 , Jeremy Davies 1
Watershed Program, Environmental Conservation Division, Northwest Fisheries Science Center,
NOAA National Marine Fisheries Service, Seattle, WA, USA
Received 25 April 2005; received in revised form 30 September 2005; accepted 13 January 2006
Available online 28 February 2006

Abstract

Channel pattern effectively stratifies the dynamics of rivers and floodplains in forested mountain river systems of the Pacific
Northwest, USA. Straight channels are least dynamic, with relatively slow floodplain turnover and floodplains dominated by old
surfaces. Braided channels are most dynamic, with floodplain turnover as low as 25 years and predominantly young floodplain
surfaces. Island-braided and meandering channels have intermediate dynamics, with moderately frequent disturbances (erosion of
floodplain patches) maintaining a mix of old and young surfaces. Return intervals for the erosion of floodplains increase in the
order of braided, island-braided, meandering, and straight (8, 33, 60, and 89 years, respectively). A threshold for the lateral
migration of a channel occurs at a bankfull width of 15–20 m. The most likely mechanism underlying this threshold is that larger
channels are deep enough to erode below the rooting zone of bank vegetation. Above this threshold, channels not confined between
valley walls exhibit channel patterns distinguishable by slope and discharge, and slope–discharge domains can be used to predict
channel patterns. Meandering and braided patterns are most consistently identified by the model, and prediction errors are largely
associated with indistinct transitions among channel patterns that are adjacent in plots of slope against discharge. Locations of
straight channels are difficult to identify accurately with the current model. The predicted spatial distribution of channel patterns
reflects a downstream decline in channel slope, which is likely correlated with a declining ratio of bed load to suspended load.
Ecological theory suggests that biological diversity should be highest where the intermediate disturbance regime of island-braided
channels sustains high diversity of habitat and successional states through time.
© 2006 Elsevier B.V. All rights reserved.

Keywords: Channel–floodplain dynamics; Channel pattern; Bank erosion; Bed load transport

1. Introduction

Geomorphologists, biologists, and ecologists have

for decades used the classification of channel pattern as
⁎ Corresponding author. Fax: +1 206 860 3335. a common, cross-disciplinary language for describing
E-mail addresses: Tim.Beechie@noaa.gov (T.J. Beechie),
the general character of reaches of alluvial rivers
Martin.Liermann@noaa.gov (M. Liermann),
Michael.Pollock@noaa.gov (M.M. Pollock), Sarah.Baker@noaa.gov (Leopold and Wolman, 1957; Schumm, 1985; Ward et
(S. Baker), Jeremy.Davies@noaa.gov (J. Davies). al., 2002). No one has explicitly linked these patterns to
1
Fax: +1 206 860 3335. the biological and ecological processes that govern
0169-555X/$ - see front matter © 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.geomorph.2006.01.030
 T.J. Beechie et al. / Geomorphology 78 (2006) 124–141
biodiversity in river–floodplain ecosystems (Lewis et
al., 2000; Richards et al., 2002). Making this link first
requires an understanding of the geomorphological
template upon which river–floodplain ecosystems
develop, including the processes that control the
formation and persistence of floodplain surfaces. The
geomorphological processes that we refer to here are
mainly bank erosion and sediment deposition, which at
the reach scale are seen as lateral movements of the
channel across a floodplain. Processes of lateral
movement include lateral migration, avulsion, meander
cutoffs, and channel switching (Nanson and Beach,
1977; O'Connor et al., 2003). With each mechanism,
the river erodes some patches each year while other
patches accrete sediment and gradually rise in
elevation above the river bed (Hickin and Nanson,
1975; Nanson and Beach, 1977; Salo et al., 1986;
Hughes, 1997). Channel movements, thus, create a
shifting mosaic of patches of habitat of different ages
within the river corridor (Ward and Stanford, 1983).
Spatial scale at which the mosaic can be observed
varies with river and floodplain size, but typically
ranges from 101 to 104 ha (Nanson and Beach, 1977;
Salo et al., 1986).
This paper describes channel–floodplain dynamics in
forested mountain river basins, using channel pattern as
a basis for stratifying and predicting reach-level
dynamics throughout channel networks. We have three
Fig. 1. Study area and site locations for analysis of patch dynamics by channel pattern.
125
main objectives. The first objective is to identify the
threshold size of channel needed for lateral migration,
which identifies where in the channel network a stream
becomes large enough to erode forested floodplain
surfaces and form new floodplain surfaces. The second
objective is to illustrate that alluvial channel pattern
predicts channel–floodplain dynamics in reaches large
enough for lateral migration to occur. We describe these
dynamics in terms of frequency of disturbance and
formation of floodplain surfaces, as well as age structure
of floodplain surfaces. The third objective is to predict
the spatial distribution of channel patterns in mountain
river networks. We develop a predictive model for
channel patterns in the river network, map the
distribution in several large watersheds (drainage areas
∼2000 to 8000 km2), and evaluate the accuracy of the
predictive model. Finally, we discuss the implications of
network-scale patterns for residence time of floodplain-
stored sediment in mountain drainage basins, and for
regulating ecological diversity.
2. Study area
This study was focused in a mountainous region of
the Pacific Coastal Forest in North America (Fig. 1) that
encompasses the Puget trough and adjacent mountain
ranges. The Puget trough is oriented along a north–
south axis, with the Cascade Mountains to the east and
126 T.J. Beechie et al. / Geomorphology 78 (2006) 124–141
Olympic Mountains and Coast Range to the west
(Fig. 1) (Black and Silkey, 1998). Volcanic peaks in the
Cascade Mountains typically exceed 3000 m in ele-
vation, and uplifted ranges in the Olympic and Cascade
Mountains commonly exceed 1800 m. Mean annual
precipitation varies from less than 50 cm year− 1 in the
rain shadow of the Olympic Mountains to more than
600 cm year− 1 at higher elevations. Much of the pre-
cipitation at higher elevations falls as snow and melts
later in the spring. Areas at lower elevations receive
most precipitation as rain, and most runoff occurs in fall
and winter.
The Cascade Range is composed primarily of pre-
Tertiary formations (N 65 Ma old) in the north and
Tertiary formations (2–65 Ma old) in the south. Volcanic
peaks of Quaternary age (b 2 Ma old) occur in both
regions, forming the highest peaks in the range (Brown
et al., 1987). Floodplains tend to be relatively narrow in
the pre-Tertiary and Tertiary core of the Cascades, where
high-grade metamorphic rocks, volcanic rocks, and
local granitic intrusions form steep valley walls. Flood-
plains are wider to the west of this core, where low-
grade metamorphic rocks (phyllite and shale) and
continental sedimentary rocks form the Cascade foot-
hills. The widest floodplains are located in valleys
bounded by terraces comprised of unconsolidated
lacustrine clays, glacial till, and outwash gravels rising
to an elevation of at least 600 m (Heller, 1979; Brown et
al., 1987). Floodplains are also narrow in the Tertiary
Coast Range and Olympic Mountains, which consist
mainly of uplifted marine basalts and sedimentary rocks.
Headwater streams are typically steep (channel slope
N0.2) and relatively small (bankfull width b 5 m),
originating on mountain slopes underlain by bedrock.
Channel slopes decrease dramatically as streams
traverse terraces carved into valley-filling glacial
deposits (slopes typically between 0.01 and 0.08), and
channel slopes are typically b 0.01 on contemporary
floodplains (Beechie et al., 2001). Floodplain rivers in
the study area typically have gravel beds and gravelly to
sandy floodplains and banks.
A limited number of tree species comprise vegetation
on floodplain in the study area, which is part of the
Pacific Coastal Forest extending from northern Califor-
nia to Alaska. Dominant species include red alder (Alnus
rubra), black cottonwood (Populus trichocarpa), Sitka
spruce (Picea sitchensis), western hemlock (Tsuga
heterophylla), western red cedar (Thuja plicata), and
big leaf maple (Acer macrophyllum) (Franklin and
Dyrness, 1973). The general successional pattern is
from hardwood to conifer, with young patches occupied
by colonizing species such as alder and cottonwood and
old patches occupied by climax species such as Sitka
spruce, western hemlock, and western red cedar
(Crocker and Major, 1955; Fonda, 1974; Henderson et
al., 1989).
3. Methods
For the first two objectives (identifying a threshold of
channel migration and quantifying river–floodplain
dynamics), we used a simple remote-sensing approach
to estimate the distribution of patch ages on floodplains
and infer regimes of channel–floodplain dynamics. We
did not rely on measurements of the rate of channel
migration or transition matrices (e.g., O'Connor et al.,
2003), mainly because the time required to orthorectify
sequential aerial photography for many years and a large
number of sites is cost-prohibitive. The approach to the
third objective (predicting channel pattern) was to
develop a simple model based on slope–discharge do-
mains (Leopold and Wolman, 1957). This is a first-order
approximation, which ignores additional factors influ-
encing channel pattern, such as sediment supply,
sediment caliber, and influence of wood debris (Fether-
ston et al., 1995; Church, 2002; Abbe and Montgomery,
2002). Nevertheless, it has the advantage of well-
documented thresholds with which to develop a pre-
dictive model, and the added advantage of being widely
recognized across scientific disciplines.
For the aerial photograph analyses two important
assumptions were made. First, vegetation patterns can
be used as an indicator of floodplain dynamics because
channel movements of more than a few meters are
sufficient to modify the age distributions of vegetation
on the floodplain. In the Pacific Coastal Forest, some
floodplain patches are eroded annually, whereas other
patches persist for decades or centuries (Abbe and
Montgomery, 1996), hardwood forests establish rapid-
ly on new surfaces (Fetherston et al., 1995), and
succession to conifer species occurs within the first
century (Crocker and Major, 1955; Fonda, 1974). The
relative similarity in time scales of floodplain turnover
and successional processes in this region makes forest
stands a good indicator of surface age because new
surfaces are vegetated quickly and succession is rapid
enough that the progression of surface ages is readily
apparent.
Second, most of the variation in stand type and age
on a floodplain results from erosion and formation of
floodplain surfaces through channel movements.
Multiple geomorphic processes may drive erosion of
floodplain surfaces, including lateral channel migra-
tion and channel switching (O'Connor et al., 2003),
 T.J. Beechie et al. / Geomorphology 78 (2006) 124–141
island formation (Osterkamp, 1998; Gurnell et al.,
2001), crevasse splay formation (Nanson and Croke,
1992), development of side-valley tributary fans
(Florsheim, 2004), or floodplain aggradation or scour
(Nanson, 1986). The methodology is not specific to any
one of these processes, although the primary erosion
processes driving the development of channel patterns
are channel migration, avulsion, and channel switching.
Other factors such as local variation in topography or the
presence of nurse logs may affect stand type and age at
very small spatial scales (b101 m2, O'Connor et al.,
2003). To limit the influence of local topography or
other small-scale factors on the analysis, we did not
distinguish patches smaller than about 1/3 ha in the
measurements (equivalent to a circle approximately
20 m diameter).
3.1. Threshold of channel migration
Because lateral migration of the channel is arguably
the dominant floodplain erosion process in our study
rivers, we focused first on identifying a threshold of a
stream size or power below which a channel does not
migrate across its floodplain. Such a threshold is
determined by the relative magnitudes of forces acting
to erode the bank and forces resisting erosion (Nanson
and Hickin, 1986; Millar and Quick, 1993; Micheli and
Kirchner, 2002). Erosive forces are essentially a
function of channel size and slope (Nanson and Hickin,
1986), and resisting forces are a function of bank
material and vegetation characteristics (Millar and
Quick, 1993; Micheli and Kirchner, 2002). Variations
in bank material and vegetation characteristics are
relatively small within the study area (i.e., we focus
on gravel bed rivers and forested floodplains), mini-
mizing the influence of variation in resisting forces on
the threshold. Thus, the study focused on variation in
eroding forces (channel size and slope) in determining a
threshold of channel migration.
To determine how large a channel must be before it
has sufficient stream power to erode its banks and
migrate laterally across its floodplain, we compared
channel sizes of migrating and non-migrating channels
within the study area. We chose to use the width of the
bankfull channel rather than discharge or stream power
as the predictor variable because previous studies have
found that bankfull width was the best single predictor
of the rate of lateral migration (e.g., Nanson and Hickin,
1986). We selected sites with bankfull widths ranging
from less than 10 m to more than 50 m, and classified
each site as either migrating or non-migrating based on
examination of aerial orthophotographs. We classified a
127
site as migrating if one or more signs of significant
channel movement occurred, which we defined as
migration rapid enough to modify the structure of
floodplain forests. Signs of significant channel move-
ment included relict channels, scroll bars, meander
cutoffs, and multiple vegetation ages on the floodplain.
Channels with none of these indicators of channel
movement (i.e., the floodplain forest was of uniform age
and there were no signs of relict channels) were
classified as non-migrating. At each site, we measured
the width of the bankfull channel and width of the
floodplain in the field, except for large channels where
we could measure the width of the bankfull channel
from the photograph. The threshold of channel migra-
tion was determined by examining the distribution of the
Table 1
Summary of selected terms describing alluvial channel patterns,
illustrating sometimes overlapping and conflicting terms
Channel Description
pattern
Straight Single channel, sinuosity b1.5, alternating bars
(Leopold and Wolman, 1957).
Meandering Single channel, sinuosity N1.5, usually dominated by
suspended sediment load (Leopold and Wolman,
1957).
Braided Multiple flow paths separated by transient bars,
relatively stable bars, or islands (Leopold and
Wolman, 1957; Rust, 1978; Ferguson, 1987),
typically characterized by high bed load supply
(Ferguson, 1987).
Island-braided Multiple flow paths separated by forested islands
(Ward et al., 2002), usually in gravel bed rivers. May
be viewed as a subset of the braided channel definition
of Leopold and Wolman (1957), and of anabranching
channels (Schumm, 1985).
Anastomosing Multiple channels separated by bars or islands,
described both as identical to anabranching channels
(Knighton and Nanson, 1993) and distinct from
anabranching channels (Schumm, 1985). Also
described as synonymous with braided (Leopold et
al., 1964), and distinct from braided (Rust, 1978).
Generally accepted as having low bed load supply
(Ferguson, 1987) and rare lateral migration (Knighton
and Nanson, 1993).
Anabranching Multiple channels divided by islands N3 times wider
than water width at average discharge (Schumm,
1985). Overlaps definitions of braided (Leopold and
Wolman, 1957), anastomosing (Schumm, 1985), and
island-braided channels.
Wandering Multiple channels, generally in gravel bed rivers.
Can be either migrating (Type I) or avulsing (Type
II) (Carson, 1984). May be viewed as a subset of
Leopold and Wolman's (1957) braided channel, and
partly overlaps definitions of anastomosing and
anabranching channels (Knighton and Nanson,
1993). Similar to island-braided.
128 T.J. Beechie et al. / Geomorphology 78 (2006) 124–141
widths of channels between migrating and non-migrat-
ing channels.
3.2. Classification of channel pattern
Over the past several decades researchers have
developed a variety of classification systems for patterns
of alluvial river channels (e.g., Leopold and Wolman,
1957; Kellerhals et al., 1976; Schumm, 1985; Knighton
and Nanson, 1993; Ward et al., 2001), and have made
many attempts to identify channel attributes that might
predict those patterns (e.g., Leopold and Wolman, 1957;
Parker, 1976; Ferguson, 1987; Nanson and Croke, 1992;
Knighton and Nanson, 1993). No definitive classifica-
tion system has emerged, however, and decades of
attempts at refinement have produced an overlapping
and sometimes conflicting terminology (Table 1).
Moreover, the understanding of why such patterns
exist remains incomplete (Church, 2002). Nevertheless,
classification of channel pattern has facilitated commu-
nication within and across scientific disciplines, and
remains a useful approach to describing broad classes of
channel–floodplain systems.
The purpose in this paper is not to refine the
classification of channel patterns, but to describe a
geomorphological template for interpreting the dynam-
ics of river–floodplain ecosystems. We use a simple
classification of channel patterns to illustrate this ap-
proach: (1) straight, (2) meandering, (3) island-braided,
and (4) braided channels (Fig. 2). This classification is
based primarily on the early classification of braided,
Fig. 2. Illustration of the four channel patterns, indicating relative rates of lateral migration of the channel (i.e., migration that erodes floodplain
surfaces).
Table 2
Summary of channel pattern definitions used in this study
Channel pattern Definition
Straight Primarily single thread channel, sinuosity b1.5
Meandering Primarily single thread channel, sinuosity N1.5
Island-braided Multiple channels, mainly separated by
vegetated islands
Braided Multiple channels, mainly separated by
unvegetated gravel bars
meandering and straight channels of Leopold and
Wolman (1957), but also includes island-braided (or
wandering) channels as an intermediate type between
meandering and braided channels (Desloges and Church,
1989; Church, 2002; Ward et al., 2002). This classifica-
tion encompasses the range of major patterns of channels
identified in the literature, and also reflects the range of
rates of channel movement or rates of floodplain
turnover found in natural channels (Schumm, 1985).
Definitions of these types of channels are summarized in
Table 2.
3.3. Quantifying the dynamics of channels and
floodplains
To quantify the dynamics of channels and floodplains
within each of the identified channel patterns, we
selected 42 reaches encompassing the four types of
channels and analyzed the age distribution of channel
and floodplain surfaces in each reach. Classification of
channel pattern for each site was based primarily on the
 T.J. Beechie et al. / Geomorphology 78 (2006) 124–141 129

Table 3 over each aerial photograph and tallied the number of

Types of cover used in orthophotograph analysis of the distribution of grid points in each of five cover classes. The five cover
patch ages
classes were open channel, newly established vegetation
Symbol Description Age Mid- (crowns not distinct in 1 : 12,000 scale photo), small
range point age
(year) (year)
crown forest (crowns b5 m diameter), medium crown
forest (crowns 5–10 m diameter), and large crown forest
C Channel, no vegetation 0 0
(N10 m diameter) (Table 3). Spacing between points
V1 Newly established vegetation b5 3
(crowns not distinguishable), varied with the dimensions of the site so that the number
usually dominated by hardwood of sample points for each site was approximately 200,
species except in a few small channels where a point spacing of
V2 Forest with crown diameter b5 m, 5–25 15 20 m (the minimum patch size) limited the number of
usually dominated by hardwood
points to less than 100.
species
V3 Forest with crown diameter 5–10 m, 25–75 50 We estimated approximate range of ages of forest
usually mixed hardwood and patches by relating the diameter of the crown to time
conifer species since stand initiation at the intensive study site (an
V4 Forest with crown diameter N10 m, 75–250 a 150 island-braided reach of the Sauk River). We first
usually dominated by conifer
orthorectified sequential aerial photographs (1939–
species
2000), then bracketed the age of 45 points in nine
See text for description of hardwood and conifer species found in the
vegetation patches. For each point, we examined the
study area.
a
The maximum age is arbitrarily set at 250 years to facilitate sequence of aerial photographs to determine the earliest
estimation of return interval of river erosion into floodplain surfaces photo year in which the present-day stand did not yet
(described in text). Some floodplain surfaces may exceed 250 years exist (i.e., the last year in which the point was a
in age. channel). We also determined the photo year of first
appearance of the present-day stand, thereby bracketing
visual appearance of the channel in aerial photography, the year in which the stand was initiated. We then
although sinuosity was frequently used to help distin- assigned each of the 45 points an age mid-way
guish straight channels from meandering channels. All between the bracketing photo years. In a Geographic
reaches had a floodplain wide enough to allow Information System (GIS), we measured the diameter
significant channel migration across the floodplain. of five distinct crowns nearest the plot center, and
For straight, meandering, and island-braided channels, calculated mean diameter of the crowns for each point.
the floodplain was typically much more than four times Stand age was strongly related to mean diameter of
the width of the main channel. Braided channel flood- crowns (age = 2.52D1.45 ; r2 = 0.93) (Fig. 3). Based on
plains were often narrower relative to channel width,
mainly because braided channels are extremely wide
and often occupy a significant portion of the valley
floor. To the extent possible, we limited the analyses of
aerial photographs to sites where floodplain vegetation
had not been altered by human activity to capture the
distribution of stand ages and geomorphic surfaces
created by natural movements of the channel. For most
large rivers (i.e., greater than 50 m wide), floodplain
boundaries could be identified from aerial photographs
and topographic maps. Hence, site selection often
required no prior knowledge of the site. Narrower
floodplains were sometimes impossible to delineate
from aerial photographs and maps alone. Thus, for small
floodplains, we used visual observations and field
measurements of channel and floodplain dimensions
to locate potential sites, and then made final site se-
Fig. 3. Relationship of stand age to crown diameter. For patches that
lections based on the suitability of aerial photographs. were formed prior to the first photo year of 1939, we set a maximum
To characterize the distribution of ages of floodplain age of 100 years for the stand because the floodplain was logged ca.
surfaces at each site, we laid a grid of sampling points 1900.
130 T.J. Beechie et al. / Geomorphology 78 (2006) 124–141
Fig. 4. Slope–discharge thresholds separating meandering from island-
braided channels (lower line), and island-braided from braided
channels (upper line). Straight channels do not plot into a distinct
domain, but overlap meandering and island-braided channels.
the regression of age of the stand to diameter of the
crowns, we assigned an approximate range of ages to
each class of crown diameter. Definitions and range of
ages of stand classes are summarized in Table 3.
We characterized channel–floodplain dynamics at
each site using metrics based on the distribution of
surface ages. The primary metrics were the average age
of surfaces in a reach and age diversity of floodplain
surfaces in a reach. For each reach, we calculated the
average age of floodplain surfaces (agemean).
3CV1 þ 15CV2 þ 50CV3 þ 150CV4
agemean ¼ ð1Þ
CTotal −CC
where CV1–4 are number of sample points in classes V1
through V4, CC is the number of sample points in
channel, and CTotal is the total number of sample points.
Values of 3, 15, 50, and 150 are approximate midpoint
ages of vegetation classes from Table 3. Based on the
distribution of ages for each channel pattern, we
estimated mean return intervals between disturbances
of a point on the floodplain. We term this the ‘erosion
return interval’, which represents the average length of
time that a point spends as a floodplain before being
reoccupied by the channel.
We estimated the mean erosion return interval based
on the equation of Booth (1991)
X Island-braided
ft ¼ 1−e−pt ð2Þ
where ft is the cumulative proportion of floodplain
surfaces less than age t, p is the estimated probability of
erosion by the channel in any one year, and the average
time between disturbances is 1/p. We calculated
evenness (J′) of the five cover classes
P
HV − pi logðpi Þ
J V¼ ¼ ð3Þ
H Vmax logðsÞ
where H′ is the Shannon–Weiner diversity index, Hmax ′
is the maximum value of the diversity index, pi is the
proportion of the ith cover class, and s is the number of
cover classes. Evenness is an index ranging from zero to
one, where an index of zero indicates complete coverage
by a single cover class and an index of one indicates
equal proportions of all cover classes.
3.4. Prediction of channel–floodplain dynamics
Based on the confinement of a channel and the
slope–discharge domains of each channel pattern, we
constructed a simple stepwise model to predict channel
pattern in six river basins (drainage areas between 2000
and 8000 km2). The first step was to distinguish
between confined channels (floodplain width to channel
width ratio b4) and unconfined channels (floodplain
width to channel width ratio N 4). Channels exceeding
this confinement threshold have sufficient space to
develop a sinuosity of 1.5 and to achieve the meandering
channel pattern. Although confined channels may be
alluvial (Montgomery et al., 1996), the constraint
between valley walls generally limits the morphology
to straight channels with bed morphologies not
considered in this paper (e.g., step-pool or cascade,
Montgomery and Buffington, 1997). These channels are
mapped simply as “confined”. Within the unconfined
channels, channel patterns were distinguished based
on a slope–discharge plot of 47 reaches distributed
among the four patterns of channels (Fig. 4). Channel
slope was measured from USGS topographic maps
(elevation change over the reach divided by channel
length), and the 2-year flood discharge (approximating
Table 4
Thresholds of channel pattern used in a simple predictive model
Confinement Slope–discharge domain
ratio
Confined b4 NA
Straight N4 Q b 15 m2 s− 1, and S b 0.1(Q− 0.42),
Meandering N4 Q N 15 m2 s− 1, and S b 0.05(Q− 0.61)
N4 Q N 15 m2 s− 1,
0.05(Q− 0.61) b S b 0.1(Q− 0.42)
Braided N4 S N 0.1(Q− 0.42)
The confinement ratio is valley bottom width (Wv) divided by bankfull
width (Wbf). Channel slope (S) is in m/m− 1, and the predicted 2-year
flood discharge (Q) is in m3 s− 1.
 T.J. Beechie et al. / Geomorphology 78 (2006) 124–141 131

bankfull discharge) was estimated based on drainage
area and average precipitation upstream from each reach
(Sumioka et al., 1998). Drainage area and average
precipitation were calculated in GIS using a USGS 30-m
digital elevation model (DEM) and Natural Resource
Conservation Service digital data for mean annual
precipitation. Thresholds of slope–discharge between
the four channel patterns were delineated similarly to
previously published thresholds (Leopold and Wolman,
1957; Desloges and Church, 1989), but shifted slightly
toward steeper slopes. These shifts probably reflect the
relatively coarse bed material and dense bank vegetation
in the study area (Ferguson, 1987; Millar and Quick,
1993; Millar, 2000; Church, 2002). Within the uncon-
fined channels, most braided channels were above the
island-braided to braided threshold noted by Church
(2002), for which S is proportional to Q− 0.5 (Fig. 4 and
Table 4). Meandering channels tended to plot below a
line similar to that of Leopold and Wolman (1957), as
long as Q N 15 m3 s− 1. Island-braided channels were
between these two thresholds. Straight channels exhib-
ited the most overlap with other channel patterns,
precluding the definition of a slope–discharge domain
that would distinguish them from other patterns when

Fig. 5. Sample site locations for evaluating the accuracy of prediction of channel type in six test watersheds within the study area.
132 T.J. Beechie et al. / Geomorphology 78 (2006) 124–141
discharge exceeded 15 m3 s− 1. A grouping of straight
channels at low discharges, however, led us to define
straight channels as those with Q b 15 m3 s− 1.
Using GIS, we then predicted channel patterns
throughout the channel networks of several mountain
river basins that drain the western slopes of the
Cascade Mountains. Digital channel networks in the
test basins were based on existing hydrography data,
and attributed with slope, discharge, channel width,
and valley width in a GIS (J. Davies, unpublished
data). Slope for each reach was calculated based on
elevation change through the reach estimated from the
DEM divided by length of the reach in the digital
hydrography. Discharge was based on USGS regional
equations for estimating the 2-year flood as described
above (Sumioka et al., 1998). Mean width of the
valley was calculated for each reach in the GIS by
projecting multiple transects across the DEM-generat-
ed valley floor, and taking the average of individual
valley width measurements. Once attributes for all
reaches in all basins had been determined, we
predicted the channel pattern for each reach in all
Puget Sound watersheds (the region for which
attributed channel data were available) that originate
in high elevation mountains and have significant
lengths of lowland reaches. This ensured that water-
sheds were mountainous and of sufficient size to
develop large, low-gradient channels in the down-
stream reaches. Other watersheds in the analysis area
are either not mountainous, interrupted by large lakes,
or drop steeply into Puget Sound. We evaluated the
accuracy of predictions by randomly selecting 100 of
the predicted channel patterns in the six test basins
(Fig. 5), and comparing actual channel pattern
(determined from aerial photography) to predicted
channel pattern.
4. Results
4.1. An empirical threshold of channel migration
Examination of channel width and slope for
migrating and non-migrating channels reveals signif-
icant overlap in channel slopes among migrating and
non-migrating channels, but virtually no overlap of the
widths of bankfull channels (Fig. 6). Thus, there is no
indication of a threshold slope, but there is a clear
indication that some aspect of stream size determines
the potential for channel movement in a forested
mountain environment. Our data suggest that the
threshold channel size for migration across forested
floodplains is between 15 and 20 m bankfull width.
Fig. 6. Threshold of channel migration. Channels smaller than 15 m
bankfull width are non-migrating (i.e., the channel does not erode
floodplain surfaces), whereas channels wider than 20 m are migrating.
Channel slope is not a significant determinant of channel migration.
4.2. Channel pattern and river–floodplain dynamics
Across 42 floodplain sites in the Pacific Coastal
Forest of North America, patch ages were predominant-
ly old in straight channels and predominantly young in
braided channels, indicating low and high disturbance
frequencies, respectively (Fig. 7). In straight channels,
the median age of floodplain surfaces is 65 years,
whereas in braided channels, the median age is only
16 years (Table 5). Patch ages were more evenly
distributed in meandering and island-braided channels,
with median surface ages of 48 and 23 years,
respectively. The proportion of the valley floor occupied
by the channel is highest in braided channels (72%). Less
than one third of the valley floor is occupied by channels
in straight, meandering, and island-braided channels,
with the lowest proportion in meandering channels
(12%).
In straight channels, a high proportion of patches
persist well beyond 75 years, and many stands appear to
have reached the old growth stage (N250 years old) (Fig.
8). The mean erosion return interval for a point on the
floodplain of a straight channel is 89 years (Table 5).
Floodplains of meandering channels have a more even
distribution of surface ages than straight channels (Figs.
7 and 8), with a mean erosion return interval of 60 years.
Island-braided channels have an intermediate frequency
of disturbance, indicated by a broader distribution of
ages of floodplain surfaces and a mean erosion return
interval of 33 years. In braided channels, most patches
 T.J. Beechie et al. / Geomorphology 78 (2006) 124–141 133

channel by vegetated islands, with the remainder being

secondary flow paths separated from the main channel
by gravel bars. The Ls/Lm ratio in braided channels also
increases with increasing drainage area (Fig. 10). Most
channels (93%), however, are secondary flow paths
separated from the main flow by gravel bars, with few
side channels separated by vegetated islands.

4.3. Predicting channel patterns in a mountain river

network

Predicted channel patterns were generally distributed

as expected, with braided channels high in the network,
island-braided channels in mid-network, and meander-
ing channels low in the network (Fig. 11). Confined
channels are predicted in steeper portions of the network
where narrow valley floors are bounded by steep hill
slopes underlain by bedrock. Low-order tributaries in
the low-elevation (western) portion of the test basins cut
through terraces of deep unconsolidated sediments,
and, therefore, have valleys wide enough to be
classified as unconfined channels. These channels are
small and are, therefore, predicted to be straight chan-
nels. The remaining channel types are predicted only in
larger tributaries and rivers (Q2yr N 15 m3 s− 1). Braided
channels are predicted high in the network where
channels are relatively steep and small (but still have
Q2yr N 15 m3 s− 1), and meandering channels are lowest
Fig. 7. Distribution of the average age class of floodplain patches by
in the network where channel slope is low and rivers
channel type, indicating the predominance of young surfaces in
braided channels, old surfaces in straight channels, and diverse surface are large. Island-braided channels are predicted where
ages in braided and island-braided channels. Ages of vegetation slope and channel size are intermediate.
classes are V1: 0–5 years, V2: 5–25 years, V3: 25–75 years; V4: N75 The accuracy test for classification shows that
years. overall percentage of correct classification is only 45%
(Table 6). Most of the errors in classification result from
are less than 25 years old and the mean erosion return
interval is only 8 years. Channel patterns with Table 5
intermediate mean surface ages have the highest Measures of patch dynamics and patch attributes by channel pattern
diversity of patch ages (Fig. 9). Straight Meandering Island- Braided
In straight and meandering channels, the ratio of braided
side channel length to main channel length (Ls/Lm) is Mean percent channel 22% 12% 30% 72%
typically less than one (i.e., total side channel length is Mean percent of 3% 2% 18% 48%
floodplain
less than total main channel length in the reach) (Fig.
b5 years old
10). As these channels are predominantly single-thread Mean percent of 42% 24% 12% b1%
channels by definition, this is an expected result. floodplain
Island-braided channels have the greatest length of side N75 years old
channels among the channel types, and the Ls/Lm ratio Mean age of floodplain 85 63 41 12
surfaces (years)
increases with increasing drainage area (Fig. 10). Side
Median age of 65 48 23 6
channel length is approximately equal to main channel floodplain surfaces
length at a drainage area of 100 km2, but may be as (years)
much as five times greater than main channel length at a Probability of erosion 0.011 0.017 0.031 0.125
drainage area of 1000 km2. Most side channels (87%) in at a point in any year
Return interval (years) 89 60 33 8
the island-braided pattern are separated from the main
134 T.J. Beechie et al. / Geomorphology 78 (2006) 124–141

Fig. 8. Illustration of the distributions of the age of patches in each of the four channel patterns. Left column is the original orthophotograph with
floodplain boundaries delineated with heavy white lines; dashed white lines delineate the vegetation analysis area where land uses had modified
portions of a floodplain. Right column shows digitized ages of patches to illustrate proportions of the four age classes within each channel pattern.

the difficulty of predicting the straight channel pattern.
Percent omission error for straight channels is 77%,
indicating that many straight channels are incorrectly
predicted to be meandering, braided, or island-braided.
Percent omission error for straight channels is only 23%,
indicating that most channels predicted to be straight
channels are indeed straight. Braided and meandering
channels have the lowest omission errors (18% and
15%, respectively), and are correctly classified by the
model more than 80% of the time. Moreover, meander-
ing channels were never misclassified as braided
channels or vice versa.
5. Discussion
Channel patterns predict important aspects of
channel–floodplain dynamics in forested mountain
river systems, including return intervals of floodplain
erosion, extent and type of side-channel development,
and age diversity of floodplain and aquatic patches.
 T.J. Beechie et al. / Geomorphology 78 (2006) 124–141
Fig. 9. Age diversity of floodplain surfaces at first increases with
increasing mean age of floodplain surfaces, then decreases. Highest
patch age diversity is in meandering and island-braided channels,
which have intermediate disturbance frequencies.
Each of these attributes has important geomorphological
and ecological implications, which we can interpret
from the combination of results from our study, results
from other studies, and ecological theory. At the outset
of this study, we observed that channel patterns can be
arranged to represent a continuum of channel–flood-
plain dynamics, with rates of channel movement
increasing in the order of straight, meandering, island-
braided and braided channels. These lateral dynamics
strongly influence the rate of sediment exchange
between channels and floodplains, as well as the
geomorphic template upon which channel–floodplain
ecosystems develop. We also observed that channel
patterns are to some degree predictable based on simple
variables of channel confinement, channel slope, and
channel width. By employing such a model to predict
channel patterns throughout a forested mountain river
system, we illustrate the potential role of classification
of channel pattern in understanding sediment transfer
through mountain river networks and the likely
influence of channel–floodplain dynamics on ecological
diversity.
5.1. Channel pattern as a predictor of channel–
floodplain dynamics and ecological diversity
The most important outcome of this study is the
quantification of channel–floodplain dynamics within
each of four well-known channel patterns. Our purpose
was to describe channel–floodplain dynamics for each
pattern, and to quantify them in a way that has broad
implications for geomorphology and ecology. Hence,
our metrics focus on attributes of river–floodplain
135
systems that describe rates of floodplain turnover (e.g.,
erosion return interval, mean age of floodplain surfaces),
as well as complexity of the environment (e.g., side
channel length, age diversity of floodplain surfaces).
Whereas this study has the constraint that it was
conducted within a humid, forested region, the general
conclusions are transferable to river systems in many
ecological settings because relative rates of channel
movement among these channel patterns remain the
same. Some patterns may be rare or non-existent,
however, in some environments.
Erosion return intervals and mean ages of floodplain
surfaces indicate that rates of lateral channel movement
increase in the order of straight, meandering, island-
braided, and braided. Hence, persistence of floodplain
surfaces decreases in that same order, suggesting that
residence time of floodplain-stored sediment decreases
in the order of straight to braided channels. We can infer
that residence time of floodplain sediment in straight
channels is at least 101–102 years based on the average
erosion return interval for floodplains of straight
channels (89 years), but not more than 103 years
based on ages of contemporary floodplains of rivers in
the study area (Benda et al., 1992; Beechie et al.,
2001). By contrast, residence time of sediment in
braided river floodplains can only rarely exceed 101
years, as more than 85% of the valley floor is less than
5 years old and less than 1% of floodplain patches
persist beyond 75 years. Residence times of floodplain
sediments in meandering and island-braided channels
are intermediate in length. In island-braided channels,
Fig. 10. Ratio of side channel length to main channel length increases
as a function of drainage area in island-braided and braided channels.
Regression for braided channels (solid line) is Ls/Lm = 0.72A0.20,
r2 = 0.52. Regression for island-braided channels (dashed line) is Ls/
Lm = 0.37A0.24, r2 = 0.13. Ratios for meandering and straight channels
are b1, with no significant relationship to drainage area.
136 T.J. Beechie et al. / Geomorphology 78 (2006) 124–141

Fig. 11. Predicted channel patterns in the Puyallup watershed (A) indicate that most confined channels occur within the steep bedrock terrain of the
Cascade Mountains (eastern half of each watershed), and most straight channels are predicted to occur in the gentler terrain of terraced glacial sediment
(western half of each watershed). Insets illustrate braided reaches nearest the glacial headwaters on Mount Rainier and the downstream transition to
island-braided channels (B), and the transition from island-braided to meandering reaches in the lower river (C). Patterns were similar in the other five
test watersheds (location map), but line density is too high to effectively illustrate the locations of channel patterns at the small map scale.

the return interval of floodplain erosion averages 33 meandering channels is probably slightly longer (but
years but some patches persist well beyond 75 years, still on the order of 101–102 years), as indicated by the
suggesting a residence time on the order of 101–102 longer mean erosion return interval of 61 years and a
years. Residence time of floodplain sediments in higher proportion of stands persisting beyond 75 years.

Table 6
Error analysis for the prediction of channel pattern
Predicted Actual channel pattern Row total %Correct %Commission
channel pattern
Braided Island-braided Meandering Straight
Braided 9 9 0 11 29 31% 69%
Island-braided 1 15 1 13 30 50% 50%
Meandering 0 7 11 10 28 39% 61%
Straight 1 1 1 10 13 77% 23%
Column total 11 32 13 44
Percent correct 82% 47% 85% 23%
Percent omission 18% 53% 15% 77%
Overall percent correct 45%
 T.J. Beechie et al. / Geomorphology 78 (2006) 124–141
Metrics of environmental complexity indicate high-
est age diversity of floodplain surfaces at intermediate
disturbance intervals and lowest diversity at low and
high disturbance intervals. The age diversity of flood-
plain surfaces is low in straight channels, which have
long erosion return intervals, and most floodplain
surfaces are uniformly old. Age diversity is also low
in braided channels, which have the shortest erosion
return intervals and floodplain surfaces that are
predominantly young. Age diversity is highest in
island-braided and meandering channels, which have
intermediate erosion return intervals. The intermediate
disturbance hypothesis predicts that environments with
an intermediate frequency of patchy disturbances should
sustain the highest biological diversity (Connell, 1978;
Roxburgh et al., 2004). That is, an intermediate level of
disturbance maintains a mix of young and old patches
through time, which allows coexistence of colonizing
and climax species over long time periods. Our metrics
of channel–floodplain dynamics suggest that such
disturbance-mediated diversity is likely to be lowest in
straight and braided channels, where river–floodplain
landscapes are dominated either by old patches with
climax species (straight channels) or young patches with
colonizing species (braided channels). The highest
diversity should be found in island-braided channels,
which maintain a mix of young and old patches in which
colonizing and climax species coexist.
Another aspect of environmental complexity in
river–floodplain systems is the number and character
of side channels within a reach. Island-braided and
braided channels have the greatest length of side
channels among the four channel patterns. Island-
braided channels tend, however, to have higher diversity
of aquatic habitat than braided channels, mainly because
side channels of island-braided reaches vary in size and
habitat structure whereas side channels of braided
reaches are relatively homogenous (Ward et al., 2001).
Greater environmental complexity in island-braided
channels is in part a function of the high density of
complex boundaries between environments (ecotones)
(Ward et al., 1999), which is maintained by interactions
between river channels and floodplain forests as well as
by the high retention of wood debris at forested islands
(Gurnell et al., 2002). Meandering and straight channels
have relatively low side-channel lengths, and, thus, a
low frequency of ecotones compared to island-braided
channels. Coupling this result with the hypothesis that
highest biological diversity should occur at an interme-
diate ecotone frequency (Naiman et al., 1988; Ward et
al., 1999), we infer that highest biological diversity
should occur in island-braided channels.
137
5.2. Processes influencing the distribution of channel
patterns in forested mountain river systems
The development of channel patterns is controlled by
a number of factors that can be considered in a stepwise
fashion. The first consideration is whether or not a
channel is unconfined by valley walls and has room to
develop a floodplain. Where channels are confined, they
are prevented from developing alluvial channel patterns
simply because they do not have room to develop
meander bends or migrate across the floodplain. Where
channels are unconfined, the second consideration is
whether a channel is large enough to erode a forested
floodplain and initiate the development of one of several
possible channel patterns. Small channels do not have
sufficient strength of flow to erode forested floodplains,
and, therefore, cannot develop meandering, braided, or
island-braided patterns. Where channels are large
enough to migrate laterally through a forested flood-
plain, a third suite of factors determines which of several
channel patterns will develop. These factors (e.g., slope,
discharge, flow strength, sediment supply and caliber,
riparian vegetation) are often intertwined and to some
degree correlated with each other.
Employing this approach in six forested mountain
river systems yields an expected spatial arrangement of
channel patterns. Confined channels are predicted in
steep bedrock-bounded valley floors characteristic of
high elevation headwater reaches, whereas straight
channels are predicted in lower elevation headwater
streams with terrace-bounded valley floors. Both of
these channel classes exhibit relatively low lateral
channel–floodplain dynamics, either because they are
confined between bedrock hill slopes or because
channels are too small to erode the banks and migrate
laterally across the floodplains. Episodic sediment
supply to confined channels, however, may cause
infrequent aggradation or incision of the channel bed
and floodplain (Nanson and Croke, 1992; Benda and
Dunne, 1997).
Where channels are unconfined, channels with a
bankfull width smaller than 15–20 m cannot erode the
banks and are, therefore, straight. This threshold is
consistent with the discharge threshold in our model
(15 m 3 s − 1 ), as the two thresholds are roughly
equivalent based on typical relationships among channel
slope, width, depth, and discharge in the study area
(Beechie and Sibley, 1997). At least two physical
processes might explain this threshold. First, erosive
forces on the bank might exceed resistance of the bank
(including the influence of bank vegetation) when
bankfull width exceeds 15–20 m. Resisting forces are
138 T.J. Beechie et al. / Geomorphology 78 (2006) 124–141
Table 7
Parameter ranges used to calculate shear stress at the base of the bank
and critical shear stress for bank erosion
Parameter Value range Source
D50 0.01 to 0.1 m Beechie and Sibley, 1997
Bank angle 30° to 40° Millar, 2000
ϕ′ for densely 75° Millar, 2000
vegetated banks
Mean bankfull depth 0.23 to 0.92 m Beechie and Sibley, 1997
Channel slope 0.002 to 0.048 Beechie and Sibley, 1997
The ranges of values are from channels typical of the study area. Data
sources are listed in the right column.
primarily a function of grain size and reinforcement of
the bank by roots (Nanson and Hickin, 1986; Millar and
Quick, 1993; Millar, 2000). A simple calculation of
bank stability can be made based on the equation of
Millar (2000), which includes modifying the angle of
repose of bank sediment (ϕ) to account for bank
vegetation:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
sin2 h
ss ¼ 0:048ðgÞðS−1ÞðD50 Þtan/ V 1− 2
sin / V
where τ is the shear stress required to move bank
sediment, γ is the unit weight of water (N m− 3), S is
specific gravity of bank sediment, D50 is the median
diameter of sediment (m), ϕ′ is angle of repose modified
to account for bank roots, and θ is the bank angle. Based
on the range of typical values for channels spanning the
threshold width (Table 7), critical shear stress for bank
erosion ranges from 15 to 190 N m− 2, and shear stress at
the base of the bank ranges from 10 to 140 N m− 2. The
ratio of shear stress to critical shear stress, however,
does not increase with channel width, indicating that
channels at or above the threshold width are no more
likely to initiate erosion than those below the threshold
width.
The second possible explanation is that erosive forces
are insufficient to overcome resisting forces of bank
roots, but channel depth exceeds the rooting depth at the
threshold channel width. This would allow the channel
to erode its banks beneath the roots (Nanson and Hickin,
1986). The depths of roots for conifer species in riparian
zones of the study area are typically less than 0.5 m,
whereas the depths of roots for hardwood species are
usually less than 0.75 m. Local data for reaches spanning
the bankfull width threshold (Beechie and Sibley, 1997;
T. Beechie unpublished data; G. Pess, unpublished data)
show that bankfull depth is less than the rooting depth in
channels smaller than the threshold width (Fig. 12), and
bankfull depth is greater than rooting depth in channels
larger than the threshold width. Thus, it is plausible that
the threshold width is related to the channel depth at
which erosion can occur below the rooting zone.
Among unconfined channels with predicted dis-
charge N 15 m3 s− 1, the model predicts braided channels
highest in the network, island-braided channels in mid-
network, and meandering channels low in the network.
This longitudinal sequence of channel patterns conforms
to the idealized sequence described in previous work
(e.g., Nanson and Croke, 1992; Church, 2002), and
reflects downstream declines in flow strength and bed
load supply. Braided channels typically have a high
supply of bed load (Schumm, 1985; Desloges and
Church, 1989; Knighton and Nanson, 1993), and in the
study area braided channels are most commonly located
immediately downstream from glaciers or alpine terrain.
The braided pattern is also reinforced by low bank
resistance and limited occurrence of wood debris that
might force channel switching (Knighton and Nanson,
1993). Bank resistance and forced channel switching are
low in braided channels of the study area because of
smaller trees (less root strength) and the lack of sufficient
wood debris to create stable jams. Less root strength in
the banks reinforces the braided pattern (Millar and
Quick, 1993; Millar, 2000; Eaton et al., 2004), and less
wood debris inhibits the formation of stable hard points
that contribute to island-braiding. Nevertheless, both of
these factors may be a result of high lateral instability in
braided channels, rather than a cause of it. High lateral
instability may prevent development of old, stable
forests, so high root strength and abundant large wood
would not have the chance to develop. Hence, the
primary factor influencing formation of braided channels
Fig. 12. Width and depth dimensions of streams within the study area.
Depths of most channels are below the threshold of channel migration,
are shallower than typical rooting depths, and depths of most channels
above the threshold are deeper than typical rooting depths. Regression
equation for the relationship is h = 0.21 + 0.026Wbf, r2 = 0.87, P b 0.001.
 T.J. Beechie et al. / Geomorphology 78 (2006) 124–141
ultimately may be a high supply of large material in the
bed load (Desloges and Church, 1989; Dade and Friend,
1998; Church, 2002).
The downstream transition from braided to island-
braided channels mainly reflects decreasing channel
slope and decreasing bed load discharge (Desloges and
Church, 1989; Nanson and Croke, 1992; Church, 2002),
as well as the influence of increased erosion resistance of
the banks and increased role of wood debris (Knighton
and Nanson, 1993; Harwood and Brown, 1993; Abbe
and Montgomery, 1996). The formation of stable islands
is favored by reduced supply of sediment and by
persistent debris jams that establish erosion-resistant
points (Keller and Swanson, 1979; Fetherston et al.,
1995; Abbe and Montgomery, 2002). The movement of
abundant wood debris, however, also creates debris jams
that force channel switching and push flows onto the
floodplain (Harwood and Brown, 1993; Abbe and
Montgomery, 1996). The combination of these factors
produces a channel pattern that maximizes sediment
transport capacity by concentrating work done into
several narrower and deeper channels (Harwood and
Brown, 1993), and creates the most diverse age structure
of floodplain surfaces among the four channel patterns
addressed here.
The downstream transition from island-braided to
meandering channels reflects a continuing downstream
decline in channel slope and bed load discharge.
Sediment load in meandering channels tends to be
dominated by suspended load (Schumm, 1985), whereas
the upstream channel patterns are dominated by bed
load or mixed loads (Schumm, 1985; Knighton and
Nanson, 1993). Channels with a low supply of bed load
tend to have higher bank cohesion (because floodplain
deposits are typically finer grained), low channel slope,
and low width–depth ratio (Eaton et al., 2004), all of
which are typically associated with the meandering
channel pattern (Parker, 1976). The degree of meander-
ing is a function of grain resistance, bed form resistance,
and reach-scale form resistance (i.e., sinuosity), which
are maximized in meandering channels when sinuosity
is high (Eaton et al., 2004). That is, where relative grain
roughness (D/d) and bed form roughness cannot by
themselves maximize resistance in the fluvial system
(because grain size is small and bed load supply is low),
the channel achieves its equilibrium channel form by
increasing sinuosity and, thereby, reducing its slope
(Eaton et al., 2004).
Although the longitudinal sequence of channel
patterns was expected, our error analysis indicated that
the accuracy of predicting channel pattern with a simple
model is modest (overall accuracy of 45%). Meandering
139
and braided channels were clearly distinguished from
each other (neither was ever misclassified as the other),
and 77% of predicted straight channels were indeed
straight. A high proportion of meandering and braided
channels were misclassified, however, as island-braided
channels, reflecting the indistinct transitions among
adjacent channel patterns in the slope–discharge plot.
Moreover, many straight channels were misclassified as
one of the other three patterns. This occurs in large part
because our model cannot identify straight channels that
lie in the slope–discharge domains of meandering or
island-braided channels. Other factors that influence the
formation of straight channels include sediment supply
and resistance of banks to erosion (Knighton, 1998), but
these are difficult to account for in a GIS-based model
because of the lack of available data. In addition, many
channels predicted to be meandering, braided, or island-
braided appear to occupy confined valley floors, and,
therefore, have little room to develop any channel
pattern other than straight. This suggests that the classi-
fication of confinement may be inaccurate, and future
models should attempt to improve the prediction of
valley width relative to channel width.
6. Conclusions
Channel pattern effectively stratifies dynamics of
rivers and floodplains in forested mountain river
systems. Straight channels are least dynamic, with
relatively slow floodplain turnover and floodplains
dominated by old surfaces. Braided channels are most
dynamic, with floodplain turnover as low as 25 years
and predominantly young floodplain surfaces. Island-
braided and meandering channels have intermediate
dynamics, with moderately frequent disturbances main-
taining a mix of old and young surfaces. The primary
ecological implication of this finding is that the highest
biological diversity should be located in island-braided
channels. Transfer of this ecological prediction to other
environments should consider return interval of erosion
and persistence of floodplain patches, as well as
successional pathways of floodplain or aquatic species.
Where successional time frames of biota roughly match
the distribution of patch ages, this hypothesis should
apply. If successional time frames are much shorter than
typical erosion return intervals of braided channels,
however, then biota associated with all four channel
patterns may be characterized by climax species.
Conversely, where successional time frames are longer
than the life span of persistent patches in straight
channels, then most channel patterns may be character-
ized by colonizing species.
140 T.J. Beechie et al. / Geomorphology 78 (2006) 124–141
We have also shown that locations of each channel
pattern in mountain river networks are generally
predictable, with the exception that straight channels
are difficult to identify accurately with our current
model. Meandering and braided patterns are clearly
distinguished, and prediction errors are largely associ-
ated with indistinct transitions among channel patterns
that are adjacent in slope–discharge plots. The predicted
spatial distribution of channel patterns in forested
mountain river networks of the Pacific Northwest is
consistent with observed patterns elsewhere (i.e.,
braided in the headwaters, island-braided in mid
network, and meandering in the lower network).
Because this general sequence reflects downstream
declines in channel slope and ratio of bed load to
suspended load, it is likely transferable to many
mountain river systems. Transfer of our results to
other systems should consider the influence of sediment
caliber and bank vegetation on the expression of channel
patterns, both of which can shift thresholds at which
each channel pattern is manifest. Exceptions will occur
where headwater streams have relatively low sediment
supply and do not develop the braided pattern, or where
bed-load supply remains high throughout the network
and lower reaches do not develop the single-thread
meandering pattern. In non-forested systems, island-
braided reaches may be rare because of the lack of wood
debris jams that stabilize islands. Nevertheless, the
general prediction that disturbance-mediated biological
diversity is highest in mid-network should be widely
applicable because the fundamental geomorphological
and ecological processes are consistent across climatic
settings.
Acknowledgments
We thank George Pess for additional channel
geometry data used in the evaluation of the bankfull
width threshold for channel migration. We also thank
Peter Kiffney, George Pess, Joan Florsheim, and one
anonymous reviewer for helpful reviews of the
manuscript.
References
Abbe, T.B., Montgomery, D.R., 1996. Large woody debris jams,
channel hydraulics and habitat formation in large rivers. Regulated
Rivers : Research and Management 12, 201–221.
Abbe, T.B., Montgomery, D.R., 2002. Patterns and processes of wood
debris accumulation in the Queets River basin, Washington.
Geomorphology 51, 81–107.
Beechie, T.J., Sibley, T.H., 1997. Relationships between channel
characteristics, woody debris, and fish habitat in northwestern
Washington streams. Transactions of the American Fisheries
Society 126, 217–229.
Beechie, T.J., Collins, B.D., Pess, G.R., 2001. Holocene and recent
geomorphic processes, land use and salmonid habitat in two north
Puget Sound river basins. In: Dorava, J.B., Montgomery, D.R.,
Fitzpatrick, F., Palcsak, B. (Eds.), Geomorphic Processes and
Riverine Habitat, Water Science and Application, vol. 4. American
Geophysical Union, Washington D.C., pp. 37–54.
Benda, L., Dunne, T., 1997. Stochastic forcing of sediment routing and
storage in channel networks. Water Resources Research 33 (12),
2849–2863.
Benda, L., Beechie, T.J., Wissmar, R.C., Johnson, A., 1992.
Morphology and evolution of salmonid habitats in a recently
deglaciated river basin, Washington State, USA. Canadian Journal
of Fisheries and Aquatic Sciences 49, 1246–1256.
Black, R.W., Silkey, M., 1998. Water-quality assessment of the Puget
Sound Basin, Washington. Summary of stream biological data
through 1995. U.S. Geological Survey Water-Resources Investiga-
tions Report 97–4164.
Booth, D.E., 1991. Estimating prelogging old-growth in the Pacific
Northwest. Journal of Forestry 89, 25–29.
Brown, E.H., Blackwell, D.L., Christenson, B.W., Frasse, F.I.,
Haugerud, R.A., Jones, J.T., Leiggi, P.A., Morrision, M.L., Rady,
P.M., Reller, G.J., Sevigny, J.H., Silverberg, D.S., Smith, M.T.,
Sondergaard, J.N., Ziegler, C.B., 1987. Geologic Map of the
Northwest Cascades, Washington. Geological Society of America.
Map and Chart Series MC-61.
Carson, M.A., 1984. The meandering-braided river threshold: a
reappraisal. Journal of Hydrology 73, 315–334.
Church, M., 2002. Geomorphic thresholds in riverine landscapes.
Freshwater Biology 47, 541–557.
Connell, J.H., 1978. Diversity in tropical rain forests and coral reefs.
Science 199, 1302–1310.
Crocker, R.L., Major, J., 1955. Soil development in relation to
vegetation and surface age at Glacier Bay, Alaska. Journal of
Ecology 43, 427–448.
Dade, W.B., Friend, P.F., 1998. Grain-size, sediment-transport regime,
and channel slope in alluvial rivers. Journal of Geology 106,
661–675.
Desloges, J.R., Church, M.A., 1989. Wandering gravel-bed rivers. The
Canadian Geographer 33, 360–364.
Eaton, B.C., Church, M., Millar, R.G., 2004. Rational regime model of
alluvial channel morphology and response. Earth Surface Process-
es and Landforms 29, 511–529.
Ferguson, R., 1987. Hydraulic and sedimentary controls on channel
pattern. In: Richards, K. (Ed.), River Channels: Environment and
Process. Basil Blackwell Ltd, Oxford, UK, pp. 129–158.
Fetherston, K.L., Naiman, R.J., Bilby, R.E., 1995. Large woody
debris, physical process, and riparian forest development in
montane river networks of the Pacific Northwest. Geomorphology
13, 133–144.
Florsheim, J.L., 2004. Side-valley tributary fans in high-energy
floodplain environments, sediment sources and depositional
processes, Navarro Basin, California. Geological Society of
America Bulletin 116, 923–937.
Fonda, R.W., 1974. Forest succession in relation to river terrace
development in Olympic National Park, Washington. Ecology 55,
927–942.
Franklin, J.F., Dyrness, C.T., 1973. Natural vegetation of Oregon and
Washington. USDA Forest Service General Technical Report
PNW-8. Pacific Northwest Forest and Range Experiment Station,
Portland, Oregon. 417 pp.
 T.J. Beechie et al. / Geomorphology 78 (2006) 124–141
Gurnell, A.M., Petts, G.E., Hannah, D.M., Smith, B.P.G., Edwards, P.
J., Kollman, J., Ward, J.V., Tockner, K., 2001. Island formation
along the gravel-bed Fiume Tagliamento, Italy. Earth Surface
Processes and Landforms 26, 31–62.
Gurnell, A.M., Piegay, H., Swanson, F.J., Gregory, S.V., 2002. Large
wood and fluvial processes. Freshwater Biology 47, 601–619.
Harwood, K., Brown, A.G., 1993. Fluvial processes in a forested
anastomosing river: flood partitioning and changing flow patterns.
Earth Surface Processes and Landforms 18, 741–748.
Heller, P.L., 1979. Map Showing Surficial Geology of Parts of the
Lower Skagit and Baker Valleys, North Cascades, Washington.
U.S. Geological Survey, pp. 79–964. Open-file.
Henderson, J.A., Peter, D.H., Lesher, R.D., Shaw, D.C., 1989.
Forested Plant Associations of the Olympic National Forest. R6
Ecology Technical Paper 001-88. U.S. Forest Service, Pacific
Northwest Region, Portland, Oregon.
Hickin, E.J., Nanson, G.C., 1975. The character of channel migration
on the Beatton River, northeast British Columbia, Canada.
Geological Society of America Bulletin 86, 487–494.
Hughes, F.M.R., 1997. Floodplain biogeomorphology. Progress in
Physical Geography 21, 510–529.
Keller, E.A., Swanson, F.J., 1979. Effects of large organic material on
channel form and fluvial processes. Earth Surface Processes and
Landforms 4, 361–380.
Kellerhals, R., Church, M., Bray, D.I., 1976. Classification and
analysis of river processes. Journal of the Hydraulics Division,
ASCE 102 (HY7), 813–829.
Knighton, D., 1998. Fluvial Forms and Processes: A New Perspective.
Arnold, London.
Knighton, A.D., Nanson, G.C., 1993. Anastomosis and the continuum
of channel pattern. Earth Surface Processes and Landforms 18,
613–625.
Leopold, L.B., Wolman, M.G., 1957. River channel patterns: braided,
meandering and straight. In: Physiographic and hydraulic studies
of rivers. USGS Professional Paper, vol. 282-B. U.S. Government
Printing Office, Washington, D.C, pp. 39–85.
Leopold, L.B., Wolman, M.G., Miller, J.P., 1964. Fluvial Processes in
Geomorphology. W. H. Freeman and Co., San Francisco.
Lewis Jr., W.M., Hamilton, S.K., Lasi, M.A., Rodriguez, M., Saunders
III, J.F., 2000. Ecological determinism in the Orinoco floodplain.
BioScience 50, 681–692.
Micheli, E.R., Kirchner, J.W., 2002. Effects of wet meadow riparian
vegetation on stream bank erosion: 1. Remote sensing measure-
ments of streambank migration and erodibility. Earth Surface
Processes and Landforms 27, 627–639.
Millar, R.G., 2000. Effect of bank vegetation on alluvial channel
patterns. Water Resources Research 36, 1109–1118.
Millar, R.G., Quick, M.C., 1993. Effect of bank stability on geometry
of gravel rivers. Journal of Hydraulic Engineering 119,
1343–1363.
Montgomery, D.R., Buffington, J.M., 1997. Channel–reach morphol-
ogy in mountain drainage basins. Geological Society of America
Bulletin 109, 596–611.
Montgomery, D.R., Abbe, T.B., Buffington, J.M., Peterson, N.P.,
Schmidt, K.M., Stock, J.D., 1996. Distribution of bedrock and
View publication stats
141
alluvial channels in forested mountain drainage basins. Nature 381,
587–589.
Naiman, R.J., Decamps, H., Pastor, J., Johnston, C.A., 1988. The
potential importance of boundaries to fluvial ecosystems. Journal
of the North American Benthological Society 7, 289–306.
Nanson, G.C., 1986. Episodes of vertical accretion and catastrophic
stripping: a model of disequilibrium floodplain development.
Geological Society of America Bulletin 97, 1467–1475.
Nanson, G.C., Beach, H.F., 1977. Forest Succession and sedimentation
on a meandering–river floodplain, northeast British Columbia,
Canada. Journal of Biogeography 4, 229–251.
Nanson, G.C., Croke, J.C., 1992. A genetic classification of
floodplains. Geomorphology 4, 459–486.
Nanson, G.C., Hickin, E.J., 1986. A statistical analysis of bank erosion
and channel migration in western Canada. Geological Society of
American Bulletin 97, 497–504.
O'Connor, J.E., Jones, M.A., Haluska, T.L., 2003. Floodplain and
channel dynamics of the Quinault and Queets Rivers, Washington,
USA. Geomorphology 51, 31–59.
Osterkamp, W.R., 1998. Processes of fluvial island formation, with
examples from Plum Creek, Colorado and Snake River, Idaho.
Wetlands 18, 530–545.
Parker, G., 1976. On the cause and characteristic scales of meandering
and braiding in rivers. Journal of Fluid Mechanics 76, 457–480.
Richards, K., Brasington, J., Hughes, F., 2002. Geomorphic dynamics
of floodplains: ecological implications and potential modelling
strategy. Freshwater Biology 47, 559–579.
Roxburgh, S.H., Shea, K., Wilson, J.B., 2004. The intermediate
disturbance hypothesis: patch dynamics and mechanisms of
species coexistence. Ecology 85, 359–371.
Rust, B.R., 1978. A classification of alluvial channel systems. In:
Miall, A.D. (Ed.), Fluvial Sedimentology. Canadian Society of
Petroleum Geologists, Calgary, Alberta, pp. 187–198.
Salo, J., Kalliola, R., Hakkinen, I., Neimela, P., Puhakka, M., Coley, P.
D., 1986. River dynamics and the diversity of Amazon lowland
forest. Nature 322, 254–258.
Schumm, S.A., 1985. Patterns of alluvial rivers. Annual Review of
Earth and Planetary Sciences 13, 5–27.
Sumioka, S.S., Kresch, D.L., Kasnick, K.D., 1998. Magnitude and
frequency of floods in Washington. U.S. Geological Survey Water-
Resources Investigations 97–4277.
Ward, J.V., Stanford, J.A., 1983. The serial discontinuity concept of
lotic ecosystems. In: FontaineIII III, T.D., S.M., Bartell (Eds.),
Dynamics of Lotic Ecosystems. Ann Arbor Science Publishers,
Ann Arbor, pp. 29–42.
Ward, J.V., Tockner, K., Schiemer, F., 1999. Biodiversity of floodplain
river ecosystems: ecotones and connectivity. Regulated Rivers:
Research and Management 15, 125–139.
Ward, J.V., Tockner, K., Uehlinger, U., Malard, F., 2001. Understand-
ing natural patterns and processes in river corridors as the basis for
effective river restoration. Regulated Rivers: Research and
Management 17, 311–323.
Ward, J.V., Tockner, K., Arscott, D.B., Claret, C., 2002. Riverine
landscape diversity. Freshwater Biology 47, 517–539.