Long-term changes in river–floodplain dynamics: implications for salmonid habitat in the Interior Columbia Basin, USA

Ecological Applications, 21(5), 2011, pp. 1643–1658� 2011 by the Ecological Society of America

Long-term changes in river–floodplain dynamics: implications forsalmonid habitat in the Interior Columbia Basin, USA

MATTHEW J. TOMLINSON,1,3 SARAH E. GERGEL,1,4 TIMOTHY J. BEECHIE,2 AND MICHELLE M. MCCLURE2

1Centre for Applied Conservation Research, 2424 Main Mall, University of British Columbia, Vancouver,British Columbia V6T1Z4 Canada

2Northwest Fisheries Science Center, National Marine Fisheries Service, National Oceanic and Atmospheric Administration,2725 Montlake Boulevard East, Seattle, Washington 98112 USA

Abstract. Rivers and their associated floodplains are among the world’s most highlyaltered ecosystems, resulting in billions of dollars in restoration expenditures. Successfulrestoration of these systems requires information at multiple spatial scales (from localizedreaches to broader-scale watersheds), as well as information spanning long time frames. Here,we develop a suite of historical landscape indicators of riverine status, primarily from theperspective of salmonid management, using a case study in the Interior Columbia Basin,Washington, USA. We use a combination of historical and modern aerial photography toquantify changes in land cover and reach type, as well as potential fish habitat within channeland off-channel floodplain areas. As of 1949, ;55% of the Wenatchee River floodplain hadbeen converted to agriculture. By 2006, 62% had been modified by anthropogenicdevelopment, of which 20% was due to urban expansion. The historical percentage ofagricultural land in the watershed and the contemporary percentage of urban area surpassthresholds in land cover associated with deleterious impacts on river systems. In addition, theabundance of reach types associated with the highest quality salmonid habitat (island braidedand meandering reaches) has declined due to conversion to straight reach types. The areaoccupied by fish habitats associated with channel migration (slow/stagnant channels and drychannels) has declined approximately 25–30%. Along highly modified rivers, these habitatshave also become increasingly fragmented. Caveats related to visual quality and seasonaltiming of historical photographs were important considerations in the interpretation ofchanges witnessed for headwater island braided systems, as well as for floodplain ponds.Development of rigorous, long-term, multi-scale monitoring techniques is necessary to guidethe management and restoration of river–floodplain systems for the diversity of ecosystemservices they provide.

Key words: change detection; channel change; fish habitat; historical aerial photographs; InteriorColumbia Basin, Washington (USA); island braided; land cover; landscape indicators; meandering; reachclassification; salmonids.

INTRODUCTION

Rivers and their associated floodplains are among the

world’s most highly altered ecosystems (Tockner and

Stanford 2002). In the United States, billions of dollars

have been spent on restoration of our nation’s rivers

(Bernhardt et al. 2007), often with differing success

(Thompson 2006, Stewart et al. 2009, Whiteway et al.

2010). Recent calls for ‘‘process-based’’ approaches to

river restoration (Beechie et al. 2010, Naiman et al.

2010) have been recommended in order to avoid some of

the pitfalls seen in river restoration (Bernhardt et al.

2005). A process-based approach aims to restore

normative rates and magnitudes of biophysical processes

important for maintaining river–floodplain heterogene-

ity, processes such as channel migration, succession of

riparian vegetation, and large wood input (Naiman et al.

2010). Implementing such an approach not only requires

metrics quantifying changes in these processes over time

(Ward et al. 2001, Naiman et al. 2010) but also requires

information at the multiple spatial scales over which

riverine processes operate (e.g., from localized reaches to

broader-scale watersheds; Beechie et al. 2010). The other

key component of this approach is acquiring informa-

tion over long temporal scales, in order to bound the

normative conditions of a system (Beechie et al. 2010).

Understanding how riverine processes differ in more

‘‘natural’’ vs. more highly modified rivers is part of this

challenge for improving river restoration (Ward et al.

2001).

Maintenance of many stream functions and processes

is intricately linked to conditions in the riparian zones

and watersheds surrounding rivers (Jorgensen et al.

2009, Johnson and Host 2010). Because of these tight

Manuscript received 24 June 2010; revised 27 October 2010;accepted 2 November 2010; final version received 7 December2010. Corresponding Editor: M. J. Vander Zanden.

3 Present address: National Capital Commission, 202-40Elgin Street, Ottawa, Ontario K1P 1C7 Canada.

4 Corresponding author. E-mail: [email protected]

1643

linkages to both physical and biological attributes of

rivers, landscape indicators (i.e., measurements of land

use and land cover) are routinely used to infer the

potential condition of aquatic ecosystems (Innis et al.

2000, Gergel et al. 2002b, Johnson and Host 2010). For

example, changes in land cover have been directly linked

to changes in stream discharge, flood magnitude, and

channel geomorphology (White and Greer 2006). The

percentage of a riparian zone that is forested can affect

in-stream macroinvertebrate densities and water quality

(Sponseller et al. 2001). Interestingly, less work has been

conducted developing historical landscape indicators for

riverine systems (Fullerton et al. 2006, Theobald et al.

2009), despite the fact that historical landscape pattern

affects a diversity of ecological processes (Gergel et al.

2002b, Poudevigne and Baudry 2003, Lindborg and

Eriksson 2004). For example, Harding et al. (1998)

found in areas that were currently forested, historical

land cover (from ;50 years prior) was a better predictor

of stream biodiversity than contemporary measurements

of land cover.

Aerial photography possesses several features useful

in addressing long-term dynamics of ecosystems, land-

scapes, and riverine systems (Zanoni et al. 2008, Zomeni

et al. 2008, Morgan et al. 2010, Naiman et al. 2010).

Aerial photography (often available as early as the 1930s

or 1950s throughout the United States), can provide

information in a historical dimension preceding satellite

sources (e.g., Landsat imagery) by several decades

(Morgan et al. 2010). Furthermore, until the recent

availability of high spatial resolution imagery (Wulder

et al. 2004), aerial photography also provided informa-

tion at a finer spatial resolution, exceeding the 25-m or

30-m pixel sizes associated with moderate-resolution

imagery. Thus, historical landscape indicators derived

from aerial photography have potential to provide

multi-scale, long-term information needed for fostering

process-based monitoring of river systems (Zanoni et al.

2008, Naiman et al. 2010).

In the United States, river restoration is often driven

by conservation needs of species listed under the

Endangered Species Act. Anadromous salmonids (salm-

on and trout) are threatened, endangered, or at-risk

throughout much of their historical range in western

North America (Nehlsen et al. 1991, NRC 1996, Slaney

et al. 1996, Myers et al. 1998, Schindler et al. 2003). The

Columbia basin, which drains portions of British

Columbia and several U.S. states, represents some of

the most important salmonid habitat in the U.S. Pacific

Northwest; however, 84% of the salmon populations

therein are not currently viable (McClure et al. 2003).

Here, we develop a suite of historical landscape

indicators of riverine status, primarily from the perspec-

tive of salmonid management, using a case study in the

Wenatchee sub-watershed of the Interior Columbia

Basin. We use a combination of historical and modern

aerial photography to quantify changes in land cover,

reach type, as well as potential fish habitat within

channel and off-channel areas. The oldest available

contiguous set of historical aerial photographs for the

Wenatchee River (circa 1949) was contrasted with

equivalent modern (2006) imagery to address three

primary questions:

1. What were the historical distribution and relative

abundance of land cover in the floodplain and riparian

zone, and how has this changed through time?—Forest

cover reduces surface runoff, maintains stream temper-

ature, and provides large wood to streams (which in turn

provides food and habitat for many in-stream organ-

isms), whereas extensive conversion to agriculture can

reduce the provisioning of these services (Wang et al.

1997). Riparian vegetation is particularly important in

affecting physical processes in streams (Meehan et al.

1977, Bjornn and Reisser 1991, Gregory et al. 1991,

Pollock 1998, Naiman et al. 2000), and thus can directly

affect the survival of salmonids (Meehan et al. 1977,

Bjornn and Reisser 1991, Murphy and Meehan 1991,

Fausch and Northcote 1992). Riparian vegetation

influences stream temperature by blocking direct solar

radiation, and because the temperature requirements of

salmonids are quite specific (Bjornn and Reisser 1991),

changes in bankside vegetation can have significant

consequences on fish populations (Murphy and Meehan

1991). Fragmentation of floodplain and riparian areas

by roads and railroads can also disrupt river–floodplain

connectivity with significant impacts seemingly dispro-

portionate to their limited areal extent (Blanton and

Marcus 2009). Therefore, we implemented an historical

and modern land cover classification (agriculture, forest,

urban, and road) throughout the floodplain of the

Wenatchee System to provide an indication of which

areas have undergone the greatest changes in land cover

over time.

2. What were the historical relative abundances of

stream reach types, and have they changed over time?—

Delineation of stream reach types is useful in monitoring

the status of rivers and is related to potential availability

of suitable habitat for salmonids (Lunetta et al. 1997,

Burnett et al. 2007). Specific habitat requirements vary

greatly among salmonid species (Bjornn and Reisser

1991, Hicks et al. 1991) and among different life stages

of each species. For example, the preferred rearing

habitat for coho salmon (Onchorhynchus kisutch)

includes stream pools, whereas steelhead (O. mykiss)

prefer riffle habitat (Bugert et al. 1991, Young 2001).

Channel confinement and gradient greatly influence

spawning and rearing habitat potential for salmonids;

as channel confinement and gradient increase, spawning

habitat decreases (ICTRT 2007). By definition, channel

movement in unconfined reaches is less restricted,

thereby increasing the creation of off-channel floodplain

fish habitat (i.e., floodplain channels, ponds, and

wetlands). In contrast, channel confinement limits

lateral migration of meandering streams and the

creation of such habitats (Hall et al. 2007). Island

braided reaches are thought to maintain higher quality

MATTHEW J. TOMLINSON ET AL.1644 Ecological ApplicationsVol. 21, No. 5

salmonid habitat than less complex straight channels

(Beechie et al. 2006, Zanoni et al. 2008). Rivers naturallytransition among reach types (Beechie et al. 2006) with

changes in sediment supply or riparian forest condition.However, the predominance of certain transitions

beyond normative rates (i.e., conversion of highlycomplex channels to straight), could be problematicfor salmonid management if higher quality habitats are

not simultaneously regenerated. Here, we trackedchanges in stream reach types over time to determine

whether transitions among reach types were associatedwith land cover changes or related to the initial

conditions of reaches.3. What were the historical distribution and relative

abundance of floodplain and channel habitat, and how hasthis changed through time?—The complex floodplain and

channel features associated with unconfined streamreaches are especially indicative of high-quality salmo-

nid habitat. Floodplain channels provide critical habitatfor juveniles for both foraging and refuge (Brown and

Hartman 1988, Jeffres et al. 2008). Salmonids reared infloodplain channels can have higher rates of growth and

survivorship to adult than those in the adjacent mainstem (Sommer et al. 2001, Jeffres et al. 2008). Floodplain

areas also provide other important services relevant tosalmonids (e.g., nutrient and large wood input, filteringof runoff, and temperature regulation). Vegetated

islands (important features of island braided systems)typically form in channels downstream of large wood

jams and are indicative of high biological diversity(Beechie et al. 2006). A significant reduction in

floodplain habitat and connectivity due to road con-struction, water withdrawal, and channelization in the

Wenatchee System (Andonaegui 2001, Honea et al.2009) has engendered concern about the quality of

remaining fish habitat. Therefore, we mapped floodplainand channel habitat features (dry and slow/stagnant side

channels, ponds/wetlands, and vegetated islands) usinghistorical and modern imagery and quantified changes

in their amount and fragmentation throughout thefloodplain.

METHODS

Study site

The study area consists of the Wenatchee River and

four main chinook-bearing tributaries (Chiwawa, LittleWenatchee, and White rivers, and Nason Creek). These

five streams, located in Chelan County, Washington,USA (Fig. 1), are referred to collectively as the

Wenatchee System. The Wenatchee River watershed is;3550 km2 (;850 000 acres) with its mouth ;750 km

upstream of the confluence of the Columbia and thePacific Ocean. Western slopes contain glaciers, moist

alpine meadows, and open forests, with dense coniferousforest occupying lower altitudes. Drier eastern slopes arecovered by glaciers and alpine grasslands transitioning

to drought-tolerant coniferous forest (Jorgensen et al.2009). Primary land uses are related to forestry,

wilderness, agriculture, rangeland, residential, and

recreation and just over 75% of the land is managed

by the U.S. Forest Service (USFS).

The Wenatchee System was historically a highly

productive salmonid spawning area (Chapman et al.

1995, Ford et al. 2001, ICTRT 2003); whereas current

returns, specifically Upper Columbia spring chinook,

have been dramatically reduced from historical levels

(McClure et al. 2003). The basin now contains salmon

or steelhead populations belonging to four evolution-

arily significant units (ESUs), two of which are currently

endangered: Upper Columbia River spring-run chinook

salmon (O. tshawytscha), Endangered; Upper Columbia

River summer/fall-run chinook salmon (O. tshawyt-

scha); Upper Columbia River steelhead (O. mykiss),

Endangered; Lake Wenatchee sockeye salmon (O.

nerka); coho salmon (O. kisutch), Extirpated.

Aerial photographs

Historical landscape attributes were reconstructed

using the oldest, spatially contiguous photo coverage

of the Wenatchee System: USFS circa 1949 at a 1:20 000

scale. Just over 300 contact prints were created from the

original negatives by the National Archives and Records

Administration. The prints were scanned at high

resolution (1200 dpi) on an Epson Expression 1640

XL (Epson America, Long Beach, California, USA).

Attributes of the modern landscape were determined

from the USDA 2006 compressed county mosaic 1-m

FIG. 1. The Wenatchee River Watershed, Washington,USA.

July 2011 1645HISTORICAL FLOODPLAIN RECONSTRUCTION

color orthophoto (captured at 1:40 000) for Chelan

County (USDA 2006). Due to seasonal differences in

imagery acquisition, modern imagery had more visible

standing water thus exhibiting a greater total wetted

area than the historical imagery. Modern imagery was

primarily captured in July 2006 (monthly median

streamflow of 2800 ft3/s [79 m3/s]); whereas historical

imagery was captured in September 1949 (monthly

median streamflow of 942 ft3/s [27 m3/s]; Appendix A).

In comparing these median discharges with the entire

record (78 years) of monthly median stream discharges,

July 2006 flows fell in the 72nd and September 1949

flows in the 17th percentiles.

Creation of orthomosaic

Orthorectification is the process of removing geomet-

ric error, radial displacement, and tilt and relief

distortions from aerial photographs so that orthophotos

are of consistent scale and derived measurements are

planimetrically accurate. The major steps in the ortho-

mosaicing process were performed primarily with Alta

Photogrammetric Suite (APS) version 7.x (Groupe Alta,

Quebec City, Quebec, Canada). First, each scanned

image underwent interior and exterior orientation to

determine the three-dimensional space of each photo

and account for aircraft movement, respectively.

Second, ground control points (GCPs) and a USGS

30-m digital elevation model (DEM) were used to relate

the uncorrected (raw) image to the modern orthoima-

gery. Least-squares regression then transformed all

remaining (unreferenced) pixels to map coordinates.

RMSE values (the root mean-squared error) was ,1

pixel resulting in an error of 6 0.25 m. Third, bilinear

interpolation (which created new digital numbers for

each new pixel) and a transformation equation were

used to warp raw images into orthophotos, which were

then mosaiced.

Image classification

Four different classification schemes (Table 1;

Appendix B) were designed to distinguish features at

different spatial resolution and were implemented via

manual interpretation and digitization. The specifics and

ecological relevance of these classifications are explained

next.

Historical extent of the 100-year floodplain.—The 100-

year floodplain encompasses the majority of channel

and floodplain morphological change. The original

extent of the historical floodplain was not identifiable

from modern flood maps (e.g., Federal Emergency and

Management Agency) due to contemporary floodplain

modifications (levees, dams). Therefore, the historical

floodplain extent was digitized manually based on visual

interpretation of terrain features using USGS 10-m

DEM, 1:24 000 topoquads and historical orthorectified

photos (where appropriate), as abrupt boundaries

between floodplain and mountainous terrain were

clearly identifiable with these data layers. Subsequent

land cover and floodplain habitat photo-interpretation

occurred only within the boundaries of the historical

floodplain.

Generalized land cover scheme (question 1).—

Historical and modern land cover (agriculture, urban,

and forest/shrub) and road networks were digitized on-

screen from orthoimagery in ArcGIS 9.3 (Environmental

Systems Research Institute, Redlands, California, USA;

Table 1) to determine where floodplain features have

subsequently been fragmented or converted to other

uses. Roads were digitized as lines with a 5-m buffer for

areal measurements. A minimum mapping unit of 4 ha

was used, such that only patches of agriculture or urban

land �4 ha were delineated (Lunetta et al. 1997).

Infrastructure such as buildings and roads located within

a delineated agricultural or urban polygon was not

distinguished separately. However, in order to ensure

accurate representation of riparian vegetation, any

bankside forested vegetation (regardless of patch size)

nested within a larger agriculture or urban polygon was

delineated. This conservative approach helped ensure

that areal estimates of forest loss were not artificially

inflated. Because of its widespread use as a landscape

indicator, the percentage of the floodplain subjected to

anthropogenic modifications (either agriculture, urban,

or roads) was determined for each reach. Rivers were

then grouped according to their level of anthropogenic

modification based on the type, percentage, and extent of

these modifications.

Finally, the modern land cover classification was

ground-truthed using field data collected with a Trimble

GPS Pathfinder ProXT receiver (Trimble Navigation,

Sunnyvale, California, USA). The accuracy assessment

of the modern land cover interpretation was useful to

plausibly bound the ‘‘best case’’ accuracy for historical

land cover, impossible to ground-truth directly. We

quantified both producer’s and user’s accuracy. User’s

accuracy represents the probability to a map user that a

location on the map is actually that class on the ground.

It is inversely related to errors of commission, which

represent the probability of overestimating the amount

TABLE 1. Classification schemes delineated on historical (1949)and modern (2006) aerial photographs for the WenatcheeRiver System near Leavenworth, Washington, USA.

Classificationscheme Categories

Land cover agriculture; urban; forest/shrub; roads

Stream reach types

Confined cascade; step pool; plane bed; pool/riffleUnconfined straight; meandering; island-braided;

braided

Habitat features

Floodplain slow/stagnant channel; dry channel;pond/wetland

In-channel vegetated islands

Note: Detailed descriptions and mapping procedures can befound in Appendix C: Table C1.


of a given class on a map. In contrast, producer’s

accuracy is inversely related to errors of omission (or the

probability of missing, or under-mapping, a given class;

Foody 2002).

Stream reach type scheme (question 2).—Streams in

1949 and 2006 were divided into reach segments and

then classified by reach type based on confinement,

gradient, number of channels, and sinuosity (Appendix

C). Several steps were used to identify individual

reaches. First, stream centerlines and banks were

manually digitized from the orthoimagery. Reach

segments approximately 10–20 stream widths in length

(Montgomery and Buffington 1997) were delineated

beginning at the mouth of each stream and iteratively

moving upstream, guided by the 10-m DEM and stream

banks/centerlines, to identify discrete reach types. The

demarcated upper and lower boundary of each reach

segment remained at a fixed, identical location on the

orthoimagery in each time period. Each reach segment

was classified as either confined or unconfined using a

channel confinement ratio (CCR), a measure of flood-

plain width/channel width. Floodplain width was

calculated by measuring a perpendicular line from the

stream bank edge to the floodplain boundary; channel

width was calculated by measuring a perpendicular line

from opposite stream banks. CCR � 3.8 were classified

as confined, and CCR . 3.8 indicated an unconfined

reach (Hall et al. 2007). Third, unconfined channels were

further classified into four reach types (straight,

meandering, island braided, and braided) based on

number of threads per channel, sinuosity, using relative

tonal differences among vegetation, water, exposed

substrates, and channels visible in photographs

(Beechie et al. 2006). Confined channels were further

classified according to slope (using the DEM), and in

descending gradient included: cascade, step pool, plane

bed, and pool–riffle (Montgomery and Buffington

1997). Reach gradient was calculated by dividing the

elevation of the upper boundary of the reach segment by

the elevation at the lower boundary of the reach.

Additional details are provided in Appendix C.

Floodplain and channel habitat scheme (question 3).—

Floodplains and channels with an abundance of habitat

features such as connected side channels or vegetated

islands indicate potential areas of productive salmonid

habitat. Therefore, floodplain and channel habitat

features (Table 1) were manually digitized from histor-

ical and modern orthoimagery. Total area and number

of habitat features were summarized by reach.

Floodplain (off-channel) habitats included slow/stag-

nant channels, dry channels, and ponds/wetlands; in-

channel habitat included vegetated islands.

Change detection and statistical analyses

Land cover change (question 1) was quantified in a

transition matrix. Within each cover class in 1949, 1000

random points at least 100 m apart were located and

compared to land cover in 2006. To examine changes in

relative abundances of land cover types, the total area of

land cover classes was compared over time for

floodplain and riparian areas. The percentage of the

floodplain and riparian area modified by anthropogenic

alterations (urban, agriculture, and roads) was also

determined for each stream.

Reach type transitions through time (question 2) were

summarized using a transition matrix for all reaches (N

¼ 424). A subset of the reaches was examined further to

provide detailed information about the most common

reach type transitions, where reasonable replication

existed (n . 10). The emphasis was on (a) examining

changes in characteristics within a given reach over time,

as well as (b) comparing characteristics of reaches that

transitioned to reaches that did not transition. First, for

this subset, changes in reach characteristics over time

(i.e., sinuosity, length, anthropogenic modification) were

compared using a paired t test of Wilcoxon signed-

ranked scores (due to non-normality). Second, one-way

Kruskal-Wallis ANOVA (to account for non-normality)

was used to compare transitioning and non-transition-

ing reaches to determine if initial (1949) and ending

(2006) levels of sinuosity, length, or modification levels

differed. For these same reaches, the total amount of

change observed over time was also compared.

Changes in total area and number of habitat types

(question 3) were totaled across the study site.

Floodplain and channel habitat changes over time were

also summarized for each reach, and then changes in

mean area and mean number of features over time were

compared using paired t tests of their Wilcoxon signed-

rank scores. Because floodplain area differed widely

among reach segments, floodplain habitats were stan-

dardized by floodplain area (e.g., area of dry channel/

hectare floodplain). Because the length of stream in a

given reach varied among reach segments, and through

time, vegetated islands were standardized by reach

length (e.g., number of islands/kilometer reach length).

Statistical analyses were completed using SAS 9.1

(SAS Institute, Cary, North Carolina, USA) using a ¼0.05. Losses and declines in measured attributes are

represented with negative values. Because each analysis

was an independent test of discrete values (i.e., not

multiple pairwise comparisons), Bonferroni corrections

of alpha values were unnecessary.

RESULTS

1. What was the historical distribution and relative

abundance of land cover in the floodplain and riparian

zone, and how has this changed through time?—In 1949,

floodplain and riparian areas were dominated by forest/

shrub, and to a lesser extent, agriculture; urban land

cover was the least abundant (Fig. 2). Urban areas

showed the greatest increase in total area over time in

floodplain and riparian areas (602% and 372%, respec-

tively), whereas agriculture showed the largest decline

(25% and 19%, respectively; Fig. 2). Forest/shrub land

cover decreased by only 4% in total area in floodplain


and riparian areas (Fig. 2). The primary land cover

transition observed was to urban (Table 2). Both forest/

shrub and agriculture overwhelmingly transitioned to

urban (with probabilities of change of 0.17 and 0.41,

respectively). Urban cover was highly persistent, with a

0.98 probability of remaining so till 2006. Forest/shrub

and agriculture classes were the less persistent (0.78 and

0.58 probabilities of remaining) over time.

When considering all anthropogenic floodplain mod-

ifications collectively (agriculture and urban areas and

roads), the most highly modified river as of 2006 was the

Wenatchee River (63%; Fig. 3). Despite its limited areal

extent of modification (7%), Nason Creek was identified

as highly modified, as the extensive road network

alongside, and often within, the river corridor could

potentially result in disproportionate effects on riverine

condition. In contrast, the lowest levels of modification

in 2006 were observed in the Little Wenatchee (1%),

White (3%), and Chiwawa (4%) river floodplains (Fig.

3). Modifications to the Chiwawa River floodplain were

largely a result of a single, isolated, rural housing

development at the river’s confluence with the

Wenatchee River.

Overall accuracy of the modern land cover classifica-

tion (question 1) was 88%, with a kappa coefficient of

0.80 (indicating the accuracy of the classification was

80% better than chance; Table 3). Producer’s accuracies

for individual classes ranged from 98% to 69%; user’s

accuracies ranged from 94% to 82%. Agriculture and

urban classes had high producer’s accuracies (92% and98%, respectively) indicating that these classes had low

errors of omission (i.e., were unlikely to be missed).Forest/shrub had a lower producer’s accuracy (69%) and

was primarily misclassified as agriculture, thus underes-timating the forest/shrub class. Agriculture and forest/shrub had higher user’s accuracies (89% and 94%,

respectively) indicating low errors of commission andthat areas mapped as such were unlikely to be highly

overestimated. Urban had a lower user’s accuracy (82%)indicating higher error of commission. This was due to

agriculture being misclassified as urban and therebyslightly overestimating the urban class. Furthermore, if

modern errors of omission and commission for individ-ual classes bear any relevance for the historical

classification, some areas of historical forest/shrub mayhave been misclassified as agriculture, thus under-

representing historical abundance of forest/shrub bypotentially ;22%. Conversely, the historical abundanceof urban areas may have been overestimated if historical

agriculture was misclassified as urban (by 11%, as wascontemporary agriculture).

2. What were the original abundances of historicalstream reach types, and how have they changed through

time?—Changes in the relative abundance of differentreach types between historical and modern periods are

shown for all 424 reaches in the Wenatchee System (Fig.4). Reach type classifications for 68 reaches changed

between 1949 and 2006 (Table 4). The number ofstraight, island braided, and plane bed reaches increased

by 2%, 67%, and 11%, respectively, while the number ofmeandering and pool/riffle reaches decreased by 28%and 41%, respectively (Fig. 4). Reaches that weremeandering or island braided in 1949 overwhelmingly

transitioned to straight channels, with probabilities ofchange of 0.24 and 0.42, respectively (Table 4).Concomitantly, the probabilities of meandering and

island braided channels persisting over time were 0.66and 0.54, respectively (Table 4). In contrast, straight

reaches in 1949 were highly persistent, with a 0.91probability of remaining straight in 2006 (Table 4).

FIG. 2. General land cover types in (a) floodplain and (b)riparian areas of the Wenatchee System (Wenatchee, Chiwawa,White, and Little Wenatchee rivers and Nason Creek), withpercentage change from 1949 to 2006 given above bars.

TABLE 2. Transition matrices showing land cover change from1949 to 2006 at 3000 randomly generated points.

Land cover in 1949

Land cover in 2006

Agriculture Urban Forest/shrub

a) Raw tally matrix of instances of change

Agriculture 932 664 20Urban 2 130 0Forest/shrub 66 206 964

b) Probability matrix (0.0–1.0)

Agriculture 0.58 0.41 0.01Urban 0.02 0.98 0.00Forest/shrub 0.05 0.17 0.78

Note: The values in panel (a) are the numbers of points thatchanged between 1949 and 2006, and the values in panel (b) arethe probabilities of change between 1949 and 2006.


Specific reach type transitions of particular signifi-

cance (noted in bold in Table 4) were examined in more

detail, where specific characteristics of the channel,

floodplain, and riparian zone were compared to those

same characteristics in reaches that did not transition

(Appendix D). Eighteen meandering reaches transi-

tioned to straight (Table 4) and showed corresponding

decreases in mean sinuosity (1.6 to 1.3, P¼ 0.0003) and

FIG. 3. Percentage of (a) floodplain and (b) riparian area modified by anthropogenic development in 1949 vs. 2006 (i.e., roads,urban, and agriculture) in the Wenatchee System. Data from 1949 are displayed in the left column and data from 2006 in the rightcolumn for each pairing.

TABLE 3. Accuracy assessment of the general land cover mapped on modern orthoimagery.

Classified land type

Ground truth of land type

Agriculture Urban Forest/shrubUser’s

accuracy (%)

Agriculture 205 1 25 88.7Urban 13 97 9 81.5Forest/shrub 4 1 77 93.9Producer’s accuracy (%) 92.3 98.0 69.4 87.7�

Notes: Raw tally confusion matrix of land cover and calculated accuracies of the classificationscheme are provided. Agreement is shown along the diagonal (in boldface), and discrepancies areshown in off-diagonals. Overall accuracy is denoted with a dagger (�); kappa coefficient ¼ 0.8.User’s accuracy is calculated as a percentage of the row totals for each class, whereas producer’saccuracy is calculated as a percentage of column totals.


mean reach length (834 to 690 m, P ¼ ,0.0001;

Appendix D). Furthermore, the percentage of the

floodplain and riparian areas subjected to anthropo-

genic modifications increased by 3% (P ¼ 0.0009) and

2% (P ¼ 0.0327), respectively, in these reaches

(Appendix D). In contrast, the 49 meandering reaches

in 1949, which remained so in 2006, exhibited no

changes in sinuosity or reach length (Table 4) despite

the fact that both floodplain and riparian anthropo-

genic modification significantly increased by ;2%over time (P ¼ 0.0084 and 0.0305, respectively).

Furthermore, the mean change over time in modifica-

tion to floodplain and riparian areas seen in transition-

ing and non-transitioning reaches was statistically

similar (Appendix D). Meandering channels that

transitioned to straight, however, had lower initial (P

¼ 0.0003) and ending (P ¼ ,0.0001) mean sinuosity

than channels that remained meandering as of 2006

(Appendix D). Additionally, meandering channels that

transitioned to straight exhibited greater decreases in

mean sinuosity (�0.26 vs.�0.12; P¼ 0.0012) and mean

reach length (�144.3 vs. �47.9 m, P ¼ ,0.0001;

Appendix D) over time than non-transitioning chan-

nels. While initially similar in 1949, the meandering

channels that eventually transitioned to straight (by

2006) exhibited shorter reach lengths (689 m) than

those that remained meandering (796 m, P ¼ 0.0363;

Appendix D).

Just under one-half of island braided reaches (n¼ 10)

transitioned to straight (Table 4) and exhibited a

decrease in reach length (921 to 911 m, P ¼ 0.0481)

and an increase in floodplain (2%, P ¼ 0.0078) and

riparian (3%, P ¼ 0.0234) modification (Appendix D).

Furthermore, the area of vegetated islands decreased

FIG. 4. Number of different reach types for 424 reaches of the Wenatchee System in 1949 vs. 2006, and percentage change(above bars) over that period.

TABLE 4. Changes in reach type for all reaches (N¼ 424) in the Wenatchee System from 1949 to2006.

Reach type in 1949

Reach type in 2006

Straight Meandering Island braided Pool/riffle Plane bed

a) Raw tally matrix of instances of change

Straight 222 3 20 0 0Meandering 18 49 7 0 0Island braided 10 1 13 0 0Pool/riffle 0 0 0 9 8Plane bed 0 0 0 1 63

b) Probability matrix (0.0–1.0)

Straight 0.91 0.01 0.08 0 0Meandering 0.24 0.66 0.09 0 0Island braided 0.42 0.04 0.54 0 0Pool/riffle 0 0 0 0.53 0.47Plane bed 0 0 0 0.02 0.98

Notes:Dynamics of transitioning reaches (in boldface) are examined further in Appendix D. Thevalues in panel (a) are the numbers of points that changed between 1949 and 2006, and the values inpanel (b) are the probabilities of change between 1949 and 2006.


from 4.9 to 1.8 m2/reach (P ¼ 0.0273) within these

reaches (Appendix D). In contrast, the 13 island braided

reaches that remained island braided (Table 4) showed

no significant change in sinuosity, reach length, anthro-

pogenic modification, or area of vegetated islands over

time (Appendix D). When examining the characteristics

of island braided channels over time, significant

differences were seen between reaches that transitioned

to straight vs. those remaining island braided. In

channels that transitioned, initial (1949) and ending

(2006) reach length was shorter than in those remaining

island braided (P ¼ 0.0029 and 0.0043, respectively;

Appendix D). However, reaches that remained island

braided had greater initial and ending anthropogenic

modification of floodplain (P ¼ 0.0217 and 0.0255,

respectively) and riparian (P ¼ 0.0349 and 0.0407,

respectively) areas than those transitioning to straight

(Appendix D).

Based on aerial photograph interpretation, 20 straight

reaches transitioned to island braided (Table 4). These

reaches showed no significant change over time in

sinuosity or reach length. However, floodplain and

riparian modification increased 3% (P¼ 0.0038) and 2%(P ¼ 0.0131), respectively (Appendix D). Detection of

vegetated islands in headwater channels was extremely

difficult as severe glare (or overexposure) within the

channel on historical imagery was more prevalent in

narrow headwater streams than in the Wenatchee main

stem. In fact, 18 of the 20 reaches designated as ‘‘straight

to island braided’’ were located in narrow headwater

streams where glare was an issue. Historical straight

reaches that transitioned to island braided were com-

pared to non-transitioning island braided channels in

order to determine if some historical island braided

channels had been inadvertently classified as straight.

Initial and ending sinuosity did not differ among these

two groups (Appendix D). However, initial and ending

floodplain (P¼,0.0001 and ,0.0001) and riparian (P¼0.0002 and 0.0003) modification was higher in non-

transitioning reaches (Appendix D).

3. What was the historical distribution and relative

abundance of floodplain and channel habitat, and how has

this changed through time?—Across the entire study site,

the area of habitat features associated with channel

migration (slow/stagnant and dry channels) has de-

creased over time. While the total area of slow/stagnant

channels decreased 26% (Fig. 5a), the number of

features did not change (Fig. 5b). Dry channels

decreased both in total area (33%) and in number

(30%; Fig. 5a, b). However, when highly modified rivers

were distinguished from less highly modified rivers (Fig.

6), reductions in area were accompanied with an

increase in total number of slow/stagnant and dry

channels, indicating fragmentation of these habitat types

in the most highly modified rivers.

In contrast, vegetated islands and pond/wetlands

appeared to have increased in area and number from

1949 to 2006 (Fig. 5). The total area of floodplain ponds

and wetlands increased 4% and in-channel vegetated

islands increased by 8% (Fig. 5a). Furthermore, the

number of ponds/wetlands and vegetated islands in-

creased 46% and 83%, respectively (Fig. 5a). Modern

imagery was captured primarily in July during higher

stream discharges (2800 ft3/s [79 m3/s]), in contrast to

the historical imagery that was captured during lower

September discharges (942 ft3/s [27 m3/s]). The higher

flows and water levels associated with the modern

imagery would likely inflate areal measurements of

contemporary ponds and wetlands. Furthermore, the

increase in vegetated islands is likely a by-product of the

same mapping limitation (excessive glare) that limited

the identification of island braided reaches in nearly all

photos of headwater streams.

DISCUSSION

This work documents the condition of the Wenatchee

System in the 1940s, as well as the important landscape

and reach-level changes that have occurred since. These

changes have serious ecological repercussions for the

area. Historically abundant agricultural lands of the

1940s (likely forest and shrub-steppe originally) have

since been converted to urban uses. The minimal losses

of forest/shrub cover witnessed from 1949 to 2006 were

because the majority of urban expansion occurred

within areas already in agriculture as of 1949.

Dynamic reach types associated with the highest quality

salmonid habitat (island braided and meandering

reaches) have converted to straight channel forms with

lower fish habitat value (Beechie et al. 2006, Zanoni et

al. 2008). Across the study site, the area of habitats

associated with channel migration (slow/stagnant chan-

nels and dry channels) has declined by roughly 25–30%.

In the most highly modified rivers, these habitats have

also become increasingly fragmented. While the total

area of floodplain ponds/wetlands has increased, these

increases should be viewed within the context of higher

water levels in 2006 (likely resulting in more standing

water across the floodplain), inflating the extent of

contemporary ponds and wetlands visible on the

modern photographs. The historical abundance of

vegetated islands and island braided reaches was likely

underestimated due to severe glare on some historical

photographs; thus, underestimating losses of island

braided systems and associated habitat. This combina-

tion of changes suggests that substantive river–flood-

plain degradation has already occurred, and further may

be imminent, in the absence of active conservation,

restoration, or management. Next, we explain the

implications of these changes in more detail.

Implications of changes in land cover

When considering all anthropogenic land cover

modifications collectively, the most highly modified

systems in 2006 were the Wenatchee River and Nason

Creek; however, the patterns and types of anthropogenic

modification differed greatly among subbasins.


Floodplain modification along the Wenatchee main

stem was primarily due to agriculture and urban

development. In contrast, road development dominated

Nason Creek, confining the channel and likely contrib-

uting to the reductions in floodplain connectivity

(Blanton and Marcus 2009) and sinuosity (Andonaegui

2001). Higher road densities were observed in the

riparian corridor along Nason Creek, and roads directly

crossed the channel in several instances (Fig. 3b;

Andonaegui 2001). These changes are important as

road development in riparian zones can have an effect

on sediment loading, large wood input, and water

temperature fluctuations (Hicks et al. 1991) dispropor-

tionately large relative to the limited areal extent of

roads. In contrast, the Chiwawa River floodplain

remained largely undeveloped, reflected by a single

housing complex at the river’s mouth.

Current levels of floodplain modification along the

Wenatchee main stem (primarily urban development),

as well as the percentage of agricultural land in the

floodplain historically, both surpass threshold levels

associated with deleterious impacts to streams. In 1949,

;55% of the Wenatchee River floodplain was in

agriculture. By 2006, 62% had been modified by

anthropogenic land cover, of which 20% was due to

urban development. As little as 10–20% urban land

cover in a watershed can severely impact stream biotic

integrity and habitat quality in general (Wang et al.

1997), and urban areas and roads also negatively impact

salmonid habitat specifically (Hicks et al. 1991, Beechie

et al. 1994, Roth et al. 1996). Wang et al. (1997)

demonstrated that agricultural land covering .50% of a

watershed can negatively impact stream biotic integrity

(an index of fish assemblages; Lyons et al. 1996) and

stream habitat quality (a combined measure of stream,

channel, and riparian conditions; Simonson et al. 1993).

Furthermore, historical levels of agriculture can reduce

the diversity of contemporary fish and invertebrate

assemblages (Harding et al. 1998). Thus, historical

agricultural expansion, followed by urban conversion,

has likely contributed to declines in salmonid habitat

along the Wenatchee main stem for many decades.

FIG. 5. (a) Total area and (b) number of floodplain and channel habitat features observed in the Wenatchee System in 1949 vs.2006, with percentage change over time displayed above each set of paired bars.


Furthermore, the overall amounts of landscape

change may have been underestimated in two important

ways. Assessing the classification accuracy of historical

maps is extremely problematic (Schulte and Mladenoff

2001, Schulte et al. 2002), as appropriate reference data

may be lacking, impossible to obtain, or at a scale too

coarse for a particular purpose (Manies and Mladenoff

2000). Here, the accuracy of the modern land cover

classification was high (88%) and helped bound the best

case for the historical classification, as we assumed that

the accuracy of the historical land cover did not exceed

this. If the limited errors in class confusion for the

modern imagery were applied to the historical image,

two patterns emerge: the amount of forest/shrub cover

FIG. 6. Impact of floodplain land cover modification (i.e., agriculture, urban, and roads) on (a) total area and (b) number offloodplain and channel habitat features in 1949 vs. 2006. Wenatchee River and Nason Creek represent rivers with highly modifiedfloodplains, whereas Chiwawa, White and Little Wenatchee rivers represent rivers with less modified floodplains. Higher and lowermodification were determined from Fig. 3.


may have been underestimated historically and second,

the amount of urban areas was overestimated histori-

cally. Taken together, the reported losses of floodplain

and riparian vegetation are likely to be conservative

(minimum) estimates, whereas urban expansion may

have been even greater than reported.

Loss of meandering reaches and associated habitats

The Wenatchee System lost many meandering reaches

between 1949 and 2006 as they converted to straight

channels. Interestingly, meandering streams in 1949 that

transitioned were under similar anthropogenic pressures

as meandering streams that did not transition (Appendix

D). The primary difference, however, was the initial

condition of reaches. Transitioning reaches had lower

levels of mean sinuosity in 1949. Thus, meandering

reaches closer to the sinuosity threshold between

meandering and straight (sinuosity ¼ 1.5) had a greater

probability of crossing the threshold to the straighter

channel forms, placing them in a category at greater risk

of losing habitats created and maintained by river

meandering.

This general trend of channel straightening would be

expected to correspond to losses in off-channel habitats,

and our results confirm that the area was historically

richer in floodplain habitats associated with lateral

channel migration. Between 1949 and 2006, streams in

the Wenatchee System lost a substantial amount of dry

channel habitats, both in area and number. Slow/

stagnant channels decreased considerably in area. In

the highly modified rivers, dry channels and slow/

stagnant channels were reduced in area, yet increased

in number. Thus, the spatial pattern of these habitats

changed, becoming more fragmented over time in the

highly modified rivers. These losses indicate a reduction

in, and fragmentation of, two important habitat types

previously available to salmonids.

Loss of meandering streams and their associated

habitat has important and direct implications for

salmonid populations. As streams migrate less across a

floodplain, fewer new off-channel features are created

and existing ones become disconnected from the main

channel. Channel migration associated with meandering

reaches is also responsible for large wood input and

increased in-stream habitat variability due to bank

cutting and pool/riffle formation (Montgomery 1999,

Naiman et al. 2000). In the Puget Lowlands,

Washington, USA, stream straightening led to funda-

mental changes in stream morphology, greatly reducing

in-stream salmonid habitat (Collins et al. 2002).

Hydromodification (diking, ditching, dredging) associ-

ated with agriculture and urban land uses of the Skagit

River (Washington, USA) greatly reduced floodplain

channel habitat, which was linked to decreases in coho

smolt production (Beechie et al. 1994). A historical

reconstruction of the Danube River (Austria) demon-

strated that channel straightening led to decreases in off-

channel habitat, reductions in nutrient exchange, and

restricted organismal movement between floodplain

channels and the main stem (Hohensinner et al. 2004).

Furthermore, fish population recovery has shown to

benefit from restoration of floodplain habitat (Beechie et

al. 1994, Sommer et al. 2001, Jeffres et al. 2008).

Dynamics of island braided and straight reaches:

some caveats

Nearly one-half of the historical island braided

reaches transitioned to straight channels. These tran-

sitioning reaches had been subjected to greater increases

in floodplain modification over time and decreases in

reach length than reaches which remained island

braided. Decreases in reach length are generally

associated with decreases in sinuosity and lateral

channel migration which fosters large wood input to

streams. Island formation is dependent, in part, on large

wood input (Abbe and Montgomery 2003, Gurnell et al.

2005), and anthropogenic modification of floodplain

and riparian areas also reduces availability of large,

bankside wood. Thus, it is likely that the reduced stream

lengths and increased anthropogenic modifications

contributed to a reduction of available large wood

(both locally and from upstream), contributing to the

subsequent loss of island braided reaches.

Island braided reaches typically have the highest

biological diversity (Naiman et al. 2010) and produce

and maintain higher quality salmonid habitat than other

reach types examined in this study (Beechie et al. 2006,

Zanoni et al. 2008). Therefore, island braided reaches

should also be priorities for conservation or restoration.

Possible restoration efforts that promote formation of

in-channel islands could include riparian management

fostering the growth of bankside vegetation as potential

large wood input. Managers could also add large wood

to the channel in key locations forming jams that would

aid in-channel island formation. However, such efforts

have not always been successful in the long term,

especially in larger systems (Thompson 2006, Stewart et

al. 2009).

Although transitions among reach types can be the

result of natural geomorphic processes due to changes in

sediment delivery and large wood (Beechie et al. 2006),

the predominance of reach conversions to ‘‘straight’’

types can reduce habitat quality for salmonids: straight

channels have lower biological and habitat diversity

(Beechie et al. 2006). Between 1949 and 2006, many

meandering and island braided reaches transitioned to

straight reaches, yet very few made the reverse

transition. However, the observation that some histor-

ically straight channels transitioned to island braided is

likely a result of glare and poor image contrast, as loss of

stream power or increasing sediment supply are much

less likely explanations. In the narrow headwaters,

where nearly all such transitioning reaches were located,

photos for the majority of reaches had severe glare

issues. Correcting this misclassification would decrease

the historical abundance of straight channels, further


magnifying the increased dominance of straight reaches

witnessed over time. Better identification of historical

islands would also influence the perceived increase of in-

channel islands over time. Correcting and/or further

confirming this misclassification would entail targeted

acquisition of additional historical photos for such areas

(from different years), and is a recommended strategy

for such work.

Recommendations

Reestablishment of channel migration (where possi-

ble) can help create a self-sustaining mechanism for river

restoration and lead to long-term improvements in

stream habitat (Ward et al. 2002). Today, there are

few remaining meandering and island braided reaches in

the Wenatchee System, as many of these reaches have

converted to straight channels. The increasing predom-

inance and persistence of land cover and reach types

associated with lower quality salmonid habitat (agricul-

ture, urbanization, channel straightening) is of concern,

as creation of high-quality habitats is unlikely to be

fostered by such landscape and reach-level trends. As

such, the remaining meandering and island braided

reaches should be the focus of restoration and protec-

tion due to their importance to salmon populations.

Furthermore, as a result of different anthropogenic

modifications and their spatial distribution, mitigation

efforts should vary among subbasins. The Wenatchee

River main stem would potentially benefit from resto-

ration efforts designed to counter the effects of

agriculture and urban expansion. Nason Creek would

benefit from restoration efforts designed to mitigate

road impacts. The remaining three rivers (the Chiwawa,

White, and Little Wenatchee) are relatively undeveloped

and would benefit from efforts intended to maintain the

natural fluvial-geomorphic and ecological processes still

occurring (e.g., channel migration, annual flooding,

large wood input, and other efforts), and could

potentially serve as useful benchmarks for normative

rates of such processes. Possible restoration efforts for

sustaining existing, active meandering channels and

associated habitats could include reducing development

within the river’s floodplain or use of setback levees

(Gergel et al. 2002a), which permit limited overbank

flooding and lateral channel migration.

Conclusions

Even limited levels of urbanization, and the associated

increase in impervious surfaces, is of tremendous

management concern throughout the United States

(Blanton and Marcus 2009, Theobald et al. 2009).

Such changes have been linked to declines in water

quality and biotic integrity in a variety of regions, yet

more research using historical data sources over time is

needed to capture the complexity of land cover change

in floodplain systems (Freeman et al. 2003). Advantages

of air photos include the ability to quantify patterns

linked to processes at fine scales (patches and reaches) as

well as to the broader floodplain landscape. However, it

is important to consider their limitations and biases

when estimating normative rates of change. For

example, this approach may have underestimated some

landscape (urban expansion) and reach-level changes

(loss of island braided systems). Nonetheless, the general

approach is applicable to river floodplains worldwide

and can improve the quantification of components

needed to undertake process-based river restoration, in

a way that is cost-effective, over a time frame otherwise

unobtainable. Especially when used in combination with

other spatial data sources and/or field techniques (Poole

et al. 2002, Densmore and Karle 2009), links between

the past and the future can be explored (Lathrop et al.

2007).

Risks to anadromous salmonids remain pervasive

throughout much of their historical range in western

North America (Nehlsen et al. 1991, NRC 1996, Slaney

et al. 1996, Myers et al. 1998, Schindler et al. 2003). At

present, a lack of historical information on land cover,

riparian and channel condition make it difficult to

disentangle the various potential causes of population

declines (e.g., habitat degradation, overfishing, and

other causes; Nehlsen et al. 1991, NRC 1996). When

combined with the future spectre of changes in

temperature and hydrology likely under future climate

scenarios (Battin et al. 2007), the need to accurately

quantify and understand the long-term changes in river

floodplains becomes only more critical. New concepts,

combined with rigorous techniques, can help guide

management and restoration of river–floodplain systems

for the diversity of ecosystem services they provide.

ACKNOWLEDGMENTS

We thank Collin Ankerson, Nicholas Coops, and ScottHinch, as well as Jessica Morgan, Trevor Lantz, ShanleyThompson, and Chris Bater for their time and assistance withthis work. Insightful comments from two anonymous reviewersgreatly improved the manuscript. Primary funding for M. J.Tomlinson was provided by NOAA Fisheries with assistancefrom the Nature Trust of British Columbia Bert HoffmeisterScholarship and the Donald McPhee Fellowship. S. E. Gergelwas supported by the NSERC Discovery Grants Program. Theviews expressed in this paper are those of the authors and donot necessarily reflect those of NOAA or its constituentagencies.

LITERATURE CITED

Abbe, T. B., and D. R. Montgomery. 2003. Patterns andprocesses of wood debris accumulation in the Queets Riverbasin, Washington. Geomorphology 51:81–107.

Andonaegui, C. 2001. Salmon, steelhead, and bull trout habitatlimiting factors for the Wenatchee Subbasin (Water ResourceInventory Area 45) and portions of WRIA 40 within ChelanCounty (Squilchuck, Stemilt and Colockum drainages).Washington State Conservation Commission, Olympia,Washington, USA.

Battin, J., M. W. Wiley, M. H. Ruckelshaus, R. N. Palmer, E.Korb, K. K. Bartz, and H. Imaki. 2007. Projected impacts ofclimate change on salmon habitat restoration. Proceedings ofthe National Academy of Sciences USA 104:6720–6725.

Beechie, T. J., E. Beamer, and L. Wasserman. 1994. Estimatingcoho salmon rearing habitat and smolt production losses in a


large river basin, and implications for habitat restoration.North American Journal of Fisheries Management 14:797–811.

Beechie, T. J., M. Liermann, M. M. Pollock, S. Baker, and J.Davies. 2006. Channel pattern and river-floodplain dynamicsin forested mountain river systems. Geomorphology 78:124–141.

Beechie, T. J., D. A. Sear, J. D. Olden, G. R. Pess, J. M.Buffington, H. Moir, P. Roni, and M. M. Pollock. 2010.Process-based principles for restoring river ecosystems.BioScience 60:209–222.

Bernhardt, E. S., et al. 2005. Synthesizing U.S. river restorationefforts. Science 308:636–637.

Bernhardt, E. S., E. B. Sudduth, M. A. Palmer, J. D. Allan,J. L. Meyer, G. Alexander, J. Follastad-Shah, B. Hassett, R.Jenkinson, R. Lave, J. Rumps, and L. Pagano. 2007.Restoring rivers one reach at a time: results from a surveyof U.S. river restoration practitioners. Restoration Ecology15:482–493.

Bjornn, T., and D. Reisser. 1991. Habitat requirements ofsalmonids in streams. Pages 83–138 in W. Meehan, editor.Influences of forest and rangeland management on salmonidfisheries and their habitats. American Fisheries SocietySpecial Publication, Bethesda, Maryland, USA.

Blanton, P., and W. A. Marcus. 2009. Railroads, roads andlateral disconnection in the river landscapes of the continen-tal United States. Geomorphology 112:212–227.

Brown, T. G., and G. F. Hartman. 1988. Contribution ofseasonally flooded lands and minor tributaries to theproduction of coho salmon in Carnation Creek, BritishColumbia. Transactions of the American Fisheries Society117:546–551.

Bugert, R. M., T. C. Bjornn, and W. R. Meehan. 1991. Summerhabitat use by young salmonids and their responses to coverand predators in a small Southeast Alaska stream. Transac-tions of the American Fisheries Society 120:474–485.

Burnett, K. M., G. H. Reeves, D. J. Miller, S. Clarke, K.Vance-Borland, and K. Christiansen. 2007. Distribution ofsalmon-habitat potential relative to landscape characteristicsand implications for conservation. Ecological Applications17:66–80.

Chapman, D., C. Peven, A. Giorgi, T. Hillman, and F. Utter.1995. Status of spring chinook salmon in the mid-ColumbiaRiver. Don Chapman Consultants, Boise, Idaho, USA.

Collins, B. D., D. R. Montgomery, and A. D. Haas. 2002.Historical changes in the distribution and functions of largewood in Puget Lowland rivers. Canadian Journal of Fisheriesand Aquatic Sciences 59:66–76.

Densmore, R. V., and K. F. Karle. 2009. Flood effects on anAlaskan stream restoration project: the value of long-termmonitoring. Journal of the American Water ResourcesAssociation 45:1424–1433.

Fausch, K. D., and T. G. Northcote. 1992. Large woody debrisand salmonid habitat in a small coastal British Columbiastream. Canadian Journal of Fisheries and Aquatic Sciences49:682–693.

Foody, G. M. 2002. Status of land cover classification accuracyassessment. Remote Sensing of Environment 80:185–201.

Ford, M., et al. 2001. Upper Columbia River steelhead andspring chinook salmon population structure and biologicalrequirements. Upper Columbia River Steelhead and SpringChinook Salmon Biological Requirements Committee, Seat-tle, Washington, USA.

Freeman, R. E., E. H. Stanley, and M. G. Turner. 2003.Analysis and conservation implications of landscape changein the Wisconsin River floodplain, USA. Ecological Appli-cations 13:416–431.

Fullerton, A. H., T. J. Beechie, S. E. Baker, J. E. Hall, andK. A. Barnas. 2006. Regional patterns of riparian character-istics in the interior Columbia River basin, Northwestern

USA: applications for restoration planning. LandscapeEcology 21:1347–1360.

Gergel, S. E., M. D. Dixon, and M. G. Turner. 2002a.Consequences of human-altered floods: levees, floods, andfloodplain forests along the Wisconsin River. EcologicalApplications 12:1755–1770.

Gergel, S. E., M. G. Turner, J. R. Miller, J. M. Melack, andE. H. Stanley. 2002b. Landscape indicators of humanimpacts to riverine systems. Aquatic Sciences 64:118–128.

Gregory, S. V., F. J. Swanson, W. A. Mckee, and K. W.Cummins. 1991. An ecosystem perspective of riparian zones.BioScience 41:540–551.

Gurnell, A., K. Tockner, P. Edwards, and G. Petts. 2005.Effects of deposited wood on biocomplexity of rivercorridors. Frontiers in Ecology and the Environment3:377–382.

Hall, J. E., D. M. Holzer, and T. J. Beechie. 2007. Predictingriver floodplain and lateral channel migration for salmonhabitat conservation. Journal of the American WaterResources Association 43:786–797.

Harding, J. S., E. F. Benfield, P. V. Bolstad, G. S. Helfman, andE. B. D. Jones. 1998. Stream biodiversity: the ghost of landuse past. Proceedings of the National Academy of SciencesUSA 95:14843–14847.

Hicks, B. J., J. D. Hall, P. A. Bisson, and J. R. Sedell. 1991.Responses of salmonids to habitat changes. Pages 483–518 inW. Meehan, editor. Influences of forest and rangelandmanagement on salmonid fisheries and their habitats.American Fisheries Society Special Publication, Bethesda,Maryland, USA.

Hohensinner, S., H. Habersack, M. Jungwirth, and G. Zauner.2004. Reconstruction of the characteristics of a naturalalluvial river-floodplain system and hydromorphologicalchanges following human modifications: the Danube River(1812–1991). River Research and Applications 20:25–41.

Honea, J. M., J. C. Jorgensen, M. M. McClure, T. D. Cooney,K. Engie, D. M. Holzer, and R. Hilborn. 2009. Evaluatinghabitat effects on population status: influence of habitatrestoration on spring-run chinook salmon. FreshwaterBiology 54:1576–1592.

ICTRT. 2003. Independent populations of chinook, andsteelhead and sockeye for listed evolutionarily significantunits within the interior Columbia River Basin domain.National Oceanic and Atmospheric Administration, Nation-al Marine Fisheries Service, Northwest Fisheries ScienceCenter, Seattle, Washington, USA.

ICTRT. 2007. Viability criteria for application to InteriorColumbia Basin salmonid ESUs. National Oceanic andAtmospheric Administration, National Marine FisheriesService, Northwest Fisheries Science Center, Seattle, Wash-ington, USA.

Innis, S. A., R. J. Naiman, and S. R. Elliott. 2000. Indicatorsand assessment methods for measuring the ecologicalintegrity of semi-aquatic terrestrial environments. Hydro-biologia 422:111–131.

Jeffres, C. A., J. J. Opperman, and P. B. Moyle. 2008.Ephemeral floodplain habitats provide best growth condi-tions for juvenile chinook salmon in a California river.Environmental Biology of Fishes 83:449–458.

Johnson, L. B., and G. E. Host. 2010. Recent developments inlandscape approaches for the study of aquatic ecosystems.Journal of the North American Benthological Society 29:41–66.

Jorgensen, J. C., J. M. Honea, M. M. McClure, T. D. Cooney,K. Engie, and D. M. Holzer. 2009. Linking landscape-levelchange to habitat quality: an evaluation of restorationactions on the freshwater habitat of spring-run chinooksalmon. Freshwater Biology 54:1560–1575.

Lathrop, R. G., D. L. Tulloch, and C. Hatfield. 2007.Consequences of land use change in the New York–New


Jersey Highlands, USA: landscape indicators of forest andwatershed integrity. Landscape and Urban Planning 79:150–159.

Lindborg, R., and O. Eriksson. 2004. Historical landscapeconnectivity affects present plant species diversity. Ecology85:1840–1845.

Lunetta, R. S., B. L. Cosentino, D. R. Montgomery, E. M.Beamer, and T. J. Beechie. 1997. GIS-based evaluation ofsalmon habitat in the Pacific Northwest. PhotogrammetricEngineering and Remote Sensing 63:1219–1229.

Lyons, J., L. Wang, and T. D. Simonson. 1996. Developmentand validation of an index of biotic integrity for coldwaterstreams in Wisconsin. North American Journal of FisheriesManagement 16:241–256.

Manies, K. L., and D. J. Mladenoff. 2000. Testing methods toproduce landscape-scale presettlement vegetation maps fromthe U.S. public land survey records. Landscape Ecology15:741–754.

McClure, M. M., E. E. Holmes, B. L. Sanderson, and C. E.Jordan. 2003. A large-scale, multispecies status, assessment:anadromous salmonids in the Columbia River Basin.Ecological Applications 13:964–989.

Meehan, W. R., F. J. Swanson, and J. R. Sedell. 1977.Influences of riparian vegetation on aquatic ecosystems withparticular reference to salmonid fishes and their food supply.USDA Forest Service General Technical Report RM-43,Tuscon, Arizona, USA.

Montgomery, D. R. 1999. Process domains and the rivercontinuum. Journal of the American Water ResourcesAssociation 35:397–410.

Montgomery, D. R., and J. M. Buffington. 1997. Channel-reach morphology in mountain drainage basins. GeologicalSociety of America Bulletin 109:596–611.

Morgan, J. L., S. E. Gergel, and N. C. Coops. 2010. Aerialphotography: a rapidly evolving tool for ecological manage-ment. BioScience 60:47–59.

Murphy, M. L., and W. Meehan. 1991. Stream ecosystems.Pages 17–46 in W. Meehan, editor. Influences of forest andrangeland management on salmonid fisheries and theirhabitats. American Fisheries Society Special Publication,Bethesda, Maryland, USA.

Myers, J. M., R. G. Kope, G. J. Bryant, D. Teel, L. J.Lierheimer, T. C. Wainwright, W. S. Grant, F. W. Waknitz,K. Neely, S. T. Lindley, and R. S. Waples. 1998. Statusreview of chinook salmon from Washington, Idaho, Oregon,and California. U.S. Department of Commerce, NOAATechnical Memorandum NMFS-NWFSC-35.

Naiman, R. J., J. S. Bechtold, T. J. Beechie, J. J. Latterell, andR. Van Pelt. 2010. A process-based view of floodplain forestpatterns in coastal river valleys of the Pacific Northwest.Ecosystems 13:1–31.

Naiman, R. J., R. E. Bilby, and P. A. Bisson. 2000. Riparianecology and management in the Pacific coastal rain forest.BioScience 50:996–1011.

Nehlsen, W., J. E. Williams, and J. A. Lichatowich. 1991.Pacific salmon at the crossroads: stocks at risk fromCalifornia, Oregon, Idaho, and Washington. Fisheries16:4–21.

NRC [National Research Council]. 1996. Upstream: salmonand society in the Pacific Northwest. National AcademyPress, Washington, D.C., USA.

Pollock, M. M. 1998. Biodiversity. Pages 430–452 in R. J.Naiman and R. E. Bilby, editors. River ecology andmanagement: lessons from the Pacific coastal ecoregion.Springer, New York, New York, USA.

Poole, G. C., J. A. Stanford, C. A. Frissell, and S. W. Running.2002. Three-dimensional mapping of geomorphic controls onflood-plain hydrology and connectivity from aerial photos.Geomorphology 48:329–347.

Poudevigne, I., and J. Baudry. 2003. The implication of pastand present landscape patterns for biodiversity research:introduction and overview. Landscape Ecology 18:223–225.

Roth, N. E., J. D. Allan, and D. L. Erickson. 1996. Landscapeinfluences on stream biotic integrity assessed at multiplespatial scales. Landscape Ecology 11:141–156.

Schindler, D. E., M. D. Scheuerell, J. W. Moore, S. M. Gende,T. B. Francis, and W. J. Palen. 2003. Pacific salmon and theecology of coastal ecosystems. Frontiers in Ecology and theEnvironment 1:31–37.

Schulte, L. A., and D. J. Mladenoff. 2001. The original U.S.public land survey records: their use and limitations inreconstructing presettlement vegetation. Journal of Forestry99:5–10.

Schulte, L. A., D. J. Mladenoff, and E. V. Nordheim. 2002.Quantitative classification of a historic northern Wisconsin(USA) landscape: mapping forests at regional scales.Canadian Journal of Forest Research 32:1616–1638.

Simonson, T. D., J. Lyons, and P. D. Kanehl. 1993. Guidelinesfor evaluating fish habitat in Wisconsin streams. USDAForest Service General Technical Report NC-164. NorthCentral Forest Experiment Station, St. Paul, Minnesota,USA.

Slaney, T. L., K. D. Hyatt, T. G. Northcote, and R. J. Fielden.1996. Status of anadromous salmon and trout in BritishColumbia and Yukon. Fisheries 21:20–35.

Sommer, T. R., M. L. Nobriga, W. C. Harrell, W. Batham, andW. J. Kimmerer. 2001. Floodplain rearing of juvenile chinooksalmon: evidence of enhanced growth and survival. CanadianJournal of Fisheries and Aquatic Sciences 58:325–333.

Sponseller, R. A., E. F. Benfield, and H. M. Valett. 2001.Relationships between land use, spatial scale and streammacroinvertebrate communities. Freshwater Biology46:1409–1424.

Stewart, G. B., H. R. Bayliss, D. A. Showler, W. J. Sutherland,and A. S. Pullin. 2009. Effectiveness of engineered in-streamstructure mitigation measures to increase salmonid abun-dance: a systematic review. Ecological Applications 19:931–941.

Theobald, D. M., S. J. Goetz, J. B. Norman, and P. Jantz. 2009.Watersheds at risk to increased impervious surface cover inthe conterminous United States. Journal of HydrologicEngineering 14:362–368.

Thompson, D. M. 2006. Did the pre-1980 use of in-streamstructures improve streams? A reanalysis of historical data.Ecological Applications 16:784–796.

Tockner, K., and J. A. Stanford. 2002. Riverine flood plains:present state and future trends. Environmental Conservation29:308–330.

USDA. 2006. Compressed county mosaic 1-m color orthophotofor Chelan County. USDA-FSA Aerial Photography FieldOffice, FSA National Agriculture Imagery Program (NAIP),Salt Lake City, Utah, USA.

Wang, L. Z., J. Lyons, P. Kanehl, and R. Gatti. 1997.Influences of watershed land use on habitat quality andbiotic integrity in Wisconsin streams. Fisheries 22:6–12.

Ward, J. V., K. Tockner, D. B. Arscott, and C. Claret. 2002.Riverine landscape diversity. Freshwater Biology 47:517–539.

Ward, J. V., K. Tockner, U. Uehlinger, and F. Malard. 2001.Understanding natural patterns and processes in rivercorridors as the basis for effective river restoration.Regulated Rivers: Research and Management 17:311–323.

White, M. D., and K. A. Greer. 2006. The effects of watershedurbanization on the stream hydrology and riparian vegeta-tion of Los Penasquitos Creek, California. Landscape andUrban Planning 74:125–138.

Whiteway, S. L., P. M. Biron, A. Zimmermann, O. Venter, andJ. W. A. Grant. 2010. Do in-stream restoration structuresenhance salmonid abundance? A meta-analysis. CanadianJournal of Fisheries and Aquatic Sciences 67:831–841.


Wulder, M. A., R. J. Hall, N. C. Coops, and S. E. Franklin.2004. High spatial resolution remotely sensed data forecosystem characterization. BioScience 54:511–521.

Young, K. A. 2001. Habitat diversity and species diversity:testing the competition hypothesis with juvenile salmonids.Oikos 95:87–93.

Zanoni, L., A. Gurnell, N. Drake, and N. Surian. 2008. Islanddynamics in a braided river from analysis of historical maps

and air photographs. River Research and Applications

24:1141–1159.

Zomeni, M., J. Tzanopoulosa, and J. D. Pantis. 2008.

Historical analysis of landscape change using remote sensing

techniques: an explanatory tool for agricultural transforma-

tion in Greek rural areas. Landscape and Urban Planning

86:38–46.

APPENDIX A

Monthly median historical (1949) and modern (2006) stream discharge, as well as median monthly discharge and range(maximum and minimum discharge) spanning 1929–2007 for the Wenatchee River below Leavenworth, Washington, USA(Ecological Archives A021-074-A1).

APPENDIX B

A comparison of coarse- and fine-scale landscape views using aerial photography (Ecological Archives A021-074-A2).

APPENDIX C

A description of classification schemes mapped in this research and identifying characteristics and scales used in digitization(Ecological Archives A021-074-A3).

APPENDIX D

A comparison of characteristics between transitioning reaches (identified in Table 4) and the reaches that did not transition to adifferent reach type from 1949 to 2006 (Ecological Archives A021-074-A4).


Documents

Long-term changes in river–floodplain dynamics: implications for salmonid habitat in the Interior Columbia Basin, USA