urban impacts on physical stream condition: effects...

ABSTRACT: An assessment of physical conditions in urbanstreams of the Puget Sound region, coupled with spatially explicitwatershed characterizations, demonstrates the importance of spa-tial scale, drainage network connectivity, and longitudinal down-stream trends when considering the effects of urbanization onstreams. A rapid stream assessment technique and a multimetricindex were used to describe the physical conditions of multiplereaches in four watersheds. Watersheds were characterized usinggeographic information system (GIS) derived landscape metricsthat represent the magnitude of urbanization at three spatialscales and the connectivity of urban land. Physical conditions, asmeasured by the physical stream conditions index (PSCI), werebest explained for the watersheds by two landscape metrics: quanti-ty of intense and grassy urban land in the subwatershed and quan-tity of intense and grassy urban land within 500 m of the site (R2 =0.52, p < 0.0005). A multiple regression of PSCI with these metricsand an additional connectivity metric (proximity of a road crossing)provided the best model for the three urban watersheds (R2 = 0.41,p < 0.0005). Analyses of longitudinal trends in PSCI within thethree urban watersheds showed that conditions improved when astream flowed through an intact riparian buffer with forest or wet-land vegetation and without road crossings. Results demonstratethat information on spatial scale and patterns of urbanization isessential to understanding and successfully managing urbanstreams.(KEY TERMS: urbanization; rivers/streams; geomorphology; landuse/land cover; spatial scale; habitat.)

McBride, Maeve and Derek B. Booth, 2005. Urban Impacts on Physical StreamCondition: Effects of Spatial Scale, Connectivity, and Longitudinal Trends. Jour-nal of the American Water Resources Association (JAWRA) 41(3):565-580.

INTRODUCTION

Urban development, coupled with human popula-tion growth, threatens local and global ecosystems

(Zipperer et al., 2000). Urbanization of the PugetSound region has dramatically altered the naturalstreamflow regime and the physical and geomorphicconditions within stream systems (Booth, 1990; Mayet al., 1997). As a result of development, once forestedland has been replaced with buildings, roads, andlawns. These land cover changes, as well as theextensive changes to the soil profile and the nativevegetation community, have altered conditions andprocesses in lowland streams, which in turn haveimpaired stream health (Booth, 1991).

The altered physical and geomorphic conditions inurban streams are diverse and complex (Hammer,1972; Neller, 1988; Booth, 1990; Booth and Jackson,1997; May et al., 1997; Caraco, 2000; Pizzuto et al.,2000; Hession et al., 2003). In general, urban streamstend to have enlarged cross-sectional dimensions(Hammer, 1972; Caraco, 2000; Pizzuto et al., 2000;Booth and Henshaw, 2001; Hession et al., 2003), accel-erated bed and bank erosion (Neller, 1988; Roesnerand Bledsoe, 2003), decreased amounts of large woodydebris (LWD) and other roughness elements (May etal., 1997; Finkenbine et al., 2000), and simplified mor-phology (Pizzuto et al., 2000). The grain size distribu-tion commonly shifts to smaller sizes in urbanstreams (Booth and Jackson, 1997); conversely, small-er grain sizes may be selectively removed in highlyurbanized systems where transport capacity greatlyexceeds sediment supply (Pizzuto et al., 2000; Finken-bine et al., 2000).

Assessments of the complex physical or biologicalconditions of urban streams are often attempted byusing multimetric indices, measures that integrate

1Paper No. 04032 of the Journal of the American Water Resources Association (JAWRA) (Copyright © 2005). Discussions are open untilDecember 1, 2005.

2Respectively, Research Assistant, Department of Civil and Environmental Engineering, University of Vermont, 213 Votey Building,Burlington, Vermont 05405; and Research Associate Professor, Department of Civil and Environmental Engineering, University of Washing-ton, Box 352700, Seattle, Washington 98195 (E-Mail/McBride: [email protected]).

JOURNAL OF THE AMERICAN WATER RESOURCES ASSOCIATION 565 JAWRA

JOURNAL OF THE AMERICAN WATER RESOURCES ASSOCIATIONJUNE AMERICAN WATER RESOURCES ASSOCIATION 2005

URBAN IMPACTS ON PHYSICAL STREAM CONDITION: EFFECTS OFSPATIAL SCALE, CONNECTIVITY, AND LONGITUDINAL TRENDS1

Maeve McBride and Derek B. Booth2

multiple components to indicate an overall condition(Plafkin et al., 1989; Rankin, 1995; Raven et al., 1998;Barbour et al., 1999; Karr and Chu, 1999). This inte-grative approach to measuring conditions can helpdiagnose causes of degradation in complex ecologicalsystems (Karr and Chu, 1999). Another benefit ofmultimetric indices is their statistical versatility.Because multimetric indices are continuous and canbe normally distributed, familiar tests can be appliedto identify significant differences in index values(Karr and Chu, 1999).

There is a need for dependable, statistically soundtools to evaluate the amount, location, and distribu-tion of urban land in watersheds. Quantitative meth-ods that link landscape patterns and ecologicalprocesses are considered critical to basic ecologicalresearch (Turner and Gardner, 1991). Measures ofurbanization that go beyond single watershed scalenumbers will help to understand and predict theseverity and extent of urban effects on stream sys-tems. With better information on the interaction ofland cover change and stream ecosystems, it shouldbe possible to improve policies and managementstrategies for protecting stream integrity in develop-ing areas (Wear et al., 1998). Based on these assump-tions, this study had three main objectives: (1) toassess instream physical and geomorphic conditionsand their variability within individual urban streams;(2) to measure urbanization using a range of alterna-tive landscape metrics; and (3) to identify relation-ships between physical stream conditions and variousspatial scales and degrees of urbanization.

METHODS

Study Streams

Multiple stream reaches were studied within fourwatersheds in the Puget Sound Lowland region withsimilarities in watershed size, surface geology, andrelief ratio (Figure 1). In total, 70 sites were sampled:7 in Juanita Creek, 28 in Swamp Creek, 22 in LittleBear Creek, and 13 in Thorndyke Creek. The water-sheds range from approximately 17 to 60 km2 and arepredominantly underlain by glacial till (Table 1). Therelief ratios, defined as the difference in elevationbetween the highest and lowest points of the water-shed divided by the length of the watershed (Dunneand Leopold, 1978), range from 11 to 23 m/km.

The study watersheds were selected to span arange of urban land cover (Table 1). Thorndyke Creek,on the western side of the remote Olympic Peninsula(Figure 1), served as a reference stream. Thorndyke

Creek’s watershed has very little development and ispredominantly forested, although some logging hasoccurred in the watershed. Approximately 20 percentof the upland areas of Thorndyke Creek’s watershedwere logged at the time of this study. The watershedof Juanita Creek, which flows into the northwest sideof Lake Washington, is highly urbanized. Little BearCreek and Swamp Creek, also tributaries in the LakeWashington watershed system, both have moderatelevels of urbanization. Forested areas in all water-sheds are predominantly second-growth or third-growth forests.

Field Methods

Physical conditions in the study streams were sam-pled using a rapid assessment technique during thesummer of 2000. The assessments were based onaverage conditions within 100 m reaches. Assessmentreaches were randomly located approximately every300 to 500 m along the mainstem channel, exceptwhere access was prohibited, in wetlands, or in nonal-luvial reaches (e.g., reaches constrained by bankarmoring). The location of the downstream end ofeach sample reach was located using a Garmin 12XLglobal positioning system (GPS) unit. These pointlocations are hereinafter referred to as sites.

Quantitative and qualitative measures were takento describe channel morphology, estimate channeldimensions, and characterize bed substrate. Bed mor-phology was classified as cascade, step-pool, planebed, pool-riffle, or dune-ripple (Montgomery and Buff-ington, 1998). The presence of sediment storage barswas recorded (Knighton, 1998). Channel planformwas classified as straight, meandering, or braided(Leopold and Wolman, 1957). Gradient was measuredat each site using a clinometer and stadia rod. Bank-full width and average bankfull depth were measuredat one representative riffle and pool feature for eachsite, using a tape and stadia rod. An estimate of bank-full cross-sectional area was derived from the productof average bankfull width and depth. An enlargementratio was then calculated as the ratio of the measuredchannel size to an expected channel size determinedfrom a regional regression of bankfull cross-sectionalarea to watershed size for nonurban streams (Booth,1990). Streambank stability was visually evaluatedand ranked as stable, slightly unstable, moderatelyunstable, or unstable (Henshaw and Booth, 2000).Channel spanning pools with a residual depth greaterthan one-fourth of the bankfull depth were tallied(Montgomery et al., 1995). Large woody debris pieceswere tallied within the active bankfull channel ifLWD was at least 25 cm in diameter and 3 m in

JAWRA 566 JOURNAL OF THE AMERICAN WATER RESOURCES ASSOCIATION

MCBRIDE AND BOOTH

length (Scholz and Booth, 2001). The structural com-plexity of the stream was visually assessed andranked in four classes from excellent/complex topoor/simple. The structural complexity rank wasbased on the sites’ diversity in channel geometry,planform, types of pool and riffle features, and overallstructure (McBride, 2001). Substrate size of active rif-fles or bar features was determined using the pebblecount method, where 100 clasts were selected ran-domly from the riffle or bar surface (Wolman, 1954).Both substrate embeddedness (Barbour et al., 1999)and substrate cementation of riffle features (McBride,2001) were each ranked in four visual classes (poor,fair, good, and excellent). The substrate embedded-ness rank was based on an assessment of the embed-dedness of approximately 10 randomly selected

individual clasts. All measurements were made by thesame observer and under similar base flow conditionsbetween July and September 2000.

Spatial Methods

A GIS based spatial analysis was used to charac-terize the landscape contributing to each sampledsite. Several spatial data sources were employed to characterize the study watersheds, including land cover (30 m, Landsat; Center for Water andWatershed Studies, 1998; Hill et al., 2003); elevation(10 m, 1:24,000 digital elevation model; UniversityLibraries, 1999); wetlands (1:24,000, National


URBAN IMPACTS ON PHYSICAL STREAM CONDITIONS: EFFECTS OF SPATIAL SCALE, CONNECTIVITY, AND LONGITUDINAL TRENDS

Figure 1. Locator Map With Study Watersheds and 1998 Land Cover (Center for Water andWatershed Studies, 1998) Using a Classification Simplified From Hill et al., 2003.

Wetlands Inventory; U.S. Fish and Wildlife Service,1987-1989); and roads (1:24,000; Puget Sound Region-al Council, 1997, unpublished, data). The land coverclassification is a 30 m grid that distinguished a totalof seven categories, three of which were “urban” cate-gories – intense urban land, grassy urban land, andforested urban land. Intense urban lands are areaswith the highest amounts of pavement, and totalimpervious area (TIA) is approximately 92 percent inthis category (Hill et al., 2003). Grassy urban landsareas distinguish areas with high amounts of pave-ment and moderate amounts of grassy or shrub vege-tation, and TIA is approximately 74 percent (Hill etal., 2003). Forested urban lands are areas with highpercentages of pavement and moderate amounts offorest vegetation, and TIA is approximately 34 per-cent (Hill et al., 2003).

Three landscape zones were delineated for eachsampled site to characterize the magnitude andpotential hydraulic connectivity of urban land at dif-ferent spatial scales. Often, the primary zone of inter-est is the watershed, the total contributing area of thelandscape. Subwatersheds were delineated for eachsampled site using GIS. A second delineated zone wasthe “buffer,” which was defined as the total riparianarea upstream from the site location (Figure 2). Twobuffer zones of different widths, 100 m and 200 m,were created. The third zone of interest was the“local” zone, defined as that portion of the total water-shed uphill from the site location and within a speci-fied distance (Figure 2). Two local zones of differentsizes, with boundaries 500 m and 1,000 m from thesampling site, were created. Both buffer and localzone boundaries were determined along topographicflow paths. The areas of the buffer and local zoneswere not extracted from the subwatershed zones, amethod preferred by some researchers (Fitzpatrick etal., 2001; Wang and Kanehl, 2003). The methodolo-gies used in this study for buffer and local zones aresimilar to other spatial analyses (Roth et al., 1996;Allan et al., 1997; Schuft et al., 1999), particularlythose of Morley and Karr (2002).

Following the delineation of the three spatial zones(subwatershed, buffer, and local), landscape metricsthat characterize both the magnitude of urban devel-opment and the connectivity of urban land weredefined. Magnitude metrics included the fractions ofurban land categories in a given spatial zone (Table2). Connectivity is broadly defined as “how spatiallyor functionally continuous a patch, corridor, networkor matrix of concern is” (Zipperer et al., 2000, page687). The connectivity metrics (road density, medianflow path length, and upstream distance to road)specifically addressed the hydraulic connectivity ofurban land to the channel network within a particu-lar spatial zone, as listed and described in Table 2.


MCBRIDE AND BOOTH

TA

BL

E 1

. Wat

ersh

ed S

izes

an

d L

and

Cov

er D

istr

ibu

tion

s.

Lan

d C

over

Dis

trib

uti

onS

urf

ace

Geo

logy

Wat

ersh

edIn

ten

seG

rass

yF

ores

ted

Gra

ss/

Bar

eG

laci

alG

laci

alS

ize

Urb

anU

rban

Urb

anS

hru

bS

oil

For

este

dW

etla

nd

Til

lO

utw

ash

All

uvi

um

Oth

erS

trea

m(k

m2 )

(per

cen

t)(p

erce

nt)

(per

cen

t)(p

erce

nt)

(per

cen

t)(p

erce

nt)

(per

cen

t)(p

erce

nt)

(per

cen

t)(p

erce

nt)

(per

cen

t)

Juan

ita

17.4

932

396

013

245

460

9

Sw

amp

58.8

1127

288

121

479

164

1

Lit

tle

Bea

r40

.35

1532

71

372

6829

31

Th

orn

dyke

31.0

207

*11

*5

072

377

202

0

*Rec

entl

y lo

gged

are

as in

Th

orn

dyke

Cre

ek a

ppea

r as

gra

ssy

urb

an a

nd

fore

sted

urb

an.

The road density metric (RDD) represents the overallconnectedness of the landscape regardless of the typeof land cover. Roads are typically conduits forstormwater either via pipes or roadside ditches. The

median flow path length metric (MFPL) is a measureof the proximity between urban areas and the streamchannel network, regardless of the road network. Theupstream distance to road metric (UPRD) represents



Figure 2. Conceptual Illustration of the Types of Spatial Scales Used in This Study.

TABLE 2. A List of Landscape Metrics, Their Units, and Detailed Descriptions.

Name Unit Description

Magnitude Metrics

Intense Urban Land (IU) Percent Proportion of intense urban land

Intense and Grassy Urban Land (IGU) Percent Proportion of intense urban land and grassy urban land

Total Urban Land (TU) Percent Proportion of all three urban land categories (intense, grassy, forested)

Connectivity Metrics

Road Density (RDD) km/km2 Total road length within a zone divided by the area of the zone

Median Flow Path Length (MFPL) m Median value of all flow path distances from each pixel of urban landto the closest stream channel

Upstream Distance to Road (UPRD) m Distance between a site and the closest upstream road crossing

how connected a particular stream site is to the near-est significant road crossing, which usually coincideswith a point source of storm water runoff from theroad or an adjacent urban area.

The intactness of the riparian buffer between con-secutive sites was described via spatial analysis toevaluate longitudinal downstream trends. The intact-ness of the riparian buffer was defined by two mea-sures: (1) the proportion of forest and wetland areasremaining in the 100 m buffer between any two sites,and (2) the number of road crossings between any twoconsecutive sites. The number of road crossings wasnormalized by the distance between the consecutivesites.

Analytical Methods

The physical conditions of the study streams wereexplored and compared using descriptive statistics,parametric tests, and nonparametric tests. Descrip-tive statistics such as means and proportions wereused to analyze gradient, morphology, planform, barfeatures, and pool abundance. For ordinal variablesthe median was used to measure the center of the distribution instead of the mean (Afifi et al., 2004).

Analysis of variance (ANOVA) was used to test for dif-ferences in LWD abundance among the four studystreams (Zar, 1984). The Kruskal-Wallis test, a non-parametric ANOVA, was used to test for differences inthe ordinal data, including bank stability, structuralcomplexity, embeddedness, and cementation (Zar,1984). Dunn’s nonparametric multiple comparisontest was used following the Kruskal-Wallis test toinvestigate pairwise differences between the streams(Zar, 1984).

A multimetric index was created to compile themeasurements of the physical attributes into a single,lumped score of physical stream condition. Sixattributes were chosen to be components of the physi-cal stream conditions index (PSCI). Table 3 lists theattributes, their descriptions, and their scoring crite-ria. These attributes were selected because they arewidely observed to vary systematically through a gra-dient of human influence and because they includemany of the responses to urbanization commonlyreported in the literature. Channel size and LWDabundance, the two metrics collected as continuousdata, were ranked to match the ordinal metrics infour categories. Channel size enlargement values fol-lowed a normal distribution, and therefore the rankswere chosen using the mean and standard deviation


MCBRIDE AND BOOTH

TABLE 3. A List of Metrics of the Physical Stream Conditions Index (PSCI) and Their Scoring Criteria.

CorrelationScoring With

Parameter Description 1 2 3 4 PSCI1

Channel Size Rank based on enlargement above an > 90 percent 50 to 90 percent 15 to 50 percent 15 percent 0.26expected channel size given the larger larger larger largerwatershed size2

LWD Abundance Rank based on quantity of LWD pieces < 5 5 to 9 10 to 14 > 14 0.73in the 100 m reach3

Bank Stability Qualitative rank of bank conditions in Unstable Moderately Slightly Stable 0.70the 100 m reach4 Unstable Unstable

Structural Qualitative rank of stream's structural Poor Fair Good Excellent 0.80Complexity complexity5

Embeddedness Qualitative rank of percentage of 75 to 100 50 to 75 25 to 50 < 25 0.59embedded substrate6 percent percent percent percent

Cementation Qualitative rank of compactness of Poor Fair Good Excellent 0.68riffle substrate7

1Spearman's correlation coefficient.2Expected channel sizes calculated using regional regression of nonurban streams (Figure 3; Booth, 1990).3May et al., 1997.4Henshaw and Booth, 2000.5Barbour et al., 1999.6Scholz and Booth, 2001.7McBride, 2001.

values. LWD abundance data did not follow a recog-nizable distribution. Large woody debris abundancewas ranked using equal intervals with the highestrank (> 14) based on the average LWD count for thereference stream, Thorndyke Creek. Lacking any con-ceptual basis to favor one attribute over another, allattributes were ranked with equal weighting, using anumerical scale of 1 to 4, and their individual scorestotaled for the index score. Higher scores indicate bet-ter physical quality of the stream.

The PSCI was analyzed via simple and multipleregressions with landscape metrics using an accept-able error rate of 5 percent. The PSCI and all predic-tor variables were checked for normality via theinspection of normal probability plots. No transforma-tions of the PSCI data or the predictor variables wereneeded. Correlations between the PSCI and its met-rics were identified using Spearman’s correlation coef-ficients, and correlations between the landscapemetrics were identified using Pearson’s correlationcoefficients (Zar, 1984). Longitudinal trends in thePSCI were also explored, particularly in comparisonto the intactness of the riparian buffer between twoadjacent sites. The change in the PSCI score (∆ PSCI)was calculated as the difference in PSCI scorebetween consecutive sites along the stream longitude.Positive values of ∆ PSCI indicate downstreamimprovement, and negative values of ∆ PSCI indicatedownstream decline. Changes in PSCI score betweenconsecutive sites can be used to test for local effectsbecause the watershed characteristics are virtuallyidentical for consecutive sites. All statistical tests andanalyses were performed using SPSS software forWindows (SPSS Inc., 1999).

RESULTS

Physical Stream Conditions

Geomorphic characteristics at all sites were similarin many respects, including gradient, morphologicclassification, planform, bar features, pool abundance,and substrate size (Table 4). Channel gradientsranged from 0.3 percent to 2.5 percent. All sites hadpool-riffle or plane bed morphology. Channel planformwas either meandering or straight; none of the sam-pled sites were braided. Most reaches had storage fea-tures in the form of point or alternate bars. Most ofthe reaches had an average of four pools per 100 m.Substrate size distributions were very similar amongreaches, and the median grain size (d50) ranged from16 to 45 mm.

Other conditions varied substantially, includingbankfull channel dimensions, LWD abundance, bankstability, structural complexity, embeddedness, andcementation (Table 5). Channel dimensions reflecteda characteristic relationship with watershed size – aswatershed size increased, the channel’s cross-section-al area at bankfull increased. The cross-sectionalareas of the sampled sites were plotted against water-shed area (Figure 3). Thorndyke Creek’s channel sizeswere larger than expected given the regional regres-sion of non-urban streams (Booth, 1990), which maybe a result of current or former logging activity. Largewoody debris abundance was significantly differentamong the study streams (p = 0.003, ANOVA). Ranksof bank stability were significantly different amongthe study streams (p < 0.0005, Kruskal-Wallis), butpairwise comparisons showed that several streamshad similar rankings (e.g., Juanita and SwampCreeks; Table 5). Ranks of structural complexity were



TABLE 4. Geomorphic Characteristics of the Four Study Watersheds From Surveys of Multiple Sites.

Proportionof Sites Proportion

Mean Range of With of Sites Mean RangeChannel Channel Pool-Riffle With Bar Pool of Pool SubstrateGradient Gradients Morphology Features Count Counts d50

Stream n (percent) (percent) (percent) (percent) (No./100 m) (No./100 m) (mm)*

Juanita 07 1.1 0.8 to 2.0 100 078 4 3 to 6 22.6

Swamp 28 1.1 0.3 to 2.0 075 084 3 1 to 8 45

Little Bear 22 1.2 0.5 to 2.5 073 058 4 1 to 10 32

Thorndyke 13 1.3 1.0 to 2.0 070 100 4 2 to 6 16

*Substrate size at farthest downstream site.

significantly different among the study streams (p <0.0005, Kruskal-Wallis), but the three urban streams(Juanita, Swamp, and Little Bear Creeks) were indis-tinguishable from each other in pairwise comparisons.Ranks of embeddedness were significantly differentamong the study streams (p < 0.0005, Kruskal-Wallis), but some stream pairs had similar rankings(e.g., Juanita and Little Bear Creeks). Ranks ofcementation were significantly different among thestudy streams (p < 0.0005, Kruskal-Wallis), but twostream pairs (Juanita and Swamp Creeks, Juanitaand Little Bear Creeks) were indistinguishable fromeach other.

Correlations Among Landscape Metrics at DifferentScales

The quantity of urban land cover in the subwater-shed showed very different relationships with thequantity of the urban land in the buffer and localzones. Even though the 100 m buffer zone occupiesonly 16 percent of the subwatershed zone on average,its land cover was nearly indistinguishable from thatof the subwatershed. This strong correlation wasdemonstrated by a correlation of total urban land inthe 100 m buffer zone to total urban land in the sub-watershed (r = 0.99, p < 0.0005, Table 6). Because thequantity of urban land in the 100 m and 200 m bufferzones was so closely correlated with that in the sub-watershed zone, the buffer zone metrics were aban-doned in the subsequent analysis. In contrast, thepercentage of urban land was often considerably dif-ferent between the local zones and the subwatershedzones. Correlations between the subwatershed zonesand the local zones were not significant for theintense urban land metric and the intense and grassyurban land metric, but the total urban land in thesubwatershed and local zones were correlated (r =0.69 for 500 m local zone, r = 0.73 for 1,000 m localzone).

Connectivity metrics were highly correlated withmany of the magnitude metrics (Table 6). Generally,watersheds had no “disconnected” urban land, at leastin the way connectivity was quantified in this study;none of the study sites had high quantities of urbanland and low measures of connectivity. Road densitywas strongly correlated with the amount of totalurban land in the subwatershed by regression analy-sis (r = 0.97, p < 0.0005), and the differences in medi-an flow path lengths between the urban streams wasslight, ranging from approximately 300 m to 400 m.In contrast, the third connectivity metric (UPRD) var-ied considerably, ranging from about 100 m to 1,800m. The UPRD metric was not significantly correlatedwith any other landscape metrics (Table 6).


MCBRIDE AND BOOTH

TA

BL

E 5

. Var

iabl

e P

hys

ical

Ch

arac

teri

stic

s of

Stu

dy S

trea

ms.

Exp

ecte

dC

ross

-C

ross

-M

ean

Ran

geM

ean

Sec

tion

alS

ecti

onal

LW

Dof

LW

DM

edia

n R

ank

s3P

SC

IR

ange

Max

imu

mA

rea

Are

aC

oun

tC

oun

tsB

ank

Str

uct

ura

lS

core

sof

PS

CI

∆PS

CI

Str

eam

n(m

2 )1

(m2 )

2(N

o./1

00 m

)(N

o./1

00 m

)S

tab

ilit

yC

omp

lexi

tyE

mb

edd

edn

ess

Cem

enta

tion

(std

.)4

Sco

res

Sco

re5

Juan

ita

73.

61.

84

1 to

13

2A2D

3EF

2HJ

12.3

(2.

5)

9 to

15.

53.

5

Sw

amp

286.

13.

86

0 to

25

2.5A

B2D

3E2H

14.4

(2.

4)

10.5

to

19.5

7.5

Lit

tle

Bea

r22

5.6

3.0

91

to 2

43B

C3D

4FG

3J16

.7 (

3.5)

12

to

22.5

6.5

Th

orn

dyke

134.

62.

514

0 to

26

3C3.

54G

319

.4 (

1.3)

18 t

o 22

.52.

5

1 Cro

ss-s

ecti

onal

are

a of

far

thes

t do

wn

stre

am s

ite.

2 Exp

ecte

d cr

oss-

sect

ion

al a

reas

cal

cula

ted

usi

ng

regi

onal

reg

ress

ion

of

non

urb

an s

trea

ms

(Fig

ure

3; B

ooth

, 199

0).

3 Cap

ital

lett

ers

(e.g

., A

) de

not

e w

hic

h m

edia

n r

anks

are

not

sig

nif

ican

tly

diff

eren

t by

Du

nn

's n

onpa

ram

etri

c m

ult

iple

com

pari

son

test

(α

= 0.

05; Z

ar, 1

984)

.4 S

tan

dard

dev

iati

on v

alu

es in

par

enth

eses

.5 ∆

PS

CI

is c

alcu

late

d as

th

e P

SC

I sc

ore

of o

ne

site

min

us

the

PS

CI

scor

e of

its

ups

trea

m n

eigh

bor.

PSCI and Urbanization

The mean PSCI scores responded predictably todifferences in urbanization. Measured PSCI valuesranged from 9 to 22.5 out of a total possible range of 6to 24. Correlations between PSCI and its metrics indi-cate that all metrics contributed almost equally toPSCI scores, except for channel size which had alower correlation coefficient (r = 0.26, Table 3). In gen-eral, PSCI scores were greater for watersheds withless urbanization (Table 5). The PSCI showed a signif-icant decline with increasing percent total urban landin the subwatershed zone, though the regression rela-tionship is not compelling (Figure 4a; R2 = 0.42, p <0.0005). When PSCI was regressed with the totalurban land within the local zones, the resulting rela-tionships provide some explanation of the variability(Figure 4b and 4c).

Better relationships between the PSCI and thelandscape metrics were found using multiple regres-sion techniques instead of single regression models. Abetter explanation of the variability in the PSCIscores is given by a multiple regression of percentintense and grassy urban land in the subwatershedzone (IGUSUB) and in the 500 m local zone (IGUL1; R2

= 0.52, p < 0.0005). Other pairings of urban land mag-nitude metrics in the subwatershed and local zonesprovide comparable, statistically significant models.

In an attempt to further explain the PSCI, a connec-tivity metric was added to the regression model. Of allconnectivity metrics, only one, upstream distance to aroad crossing (UPRD), produced a significant regres-sion model (R2 = 0.41, p < 0.0005):

PSCI = 20.1 - 11.8 IGUSUB - 9.4 IGUL1 + 1.7 UPRD

where IGUSUB and IGUL1 are in percent and UPRDis in meters.

The sites from Thorndyke Creek were excludedfrom this regression model because the connectivitymetrics (as defined) were not valid in a watershedlacking true urban land cover. For the three urbanstreams, the regression model in Equation (1) outper-forms the regression model with the two magnitudemetrics (R2 = 0.38, p < 0.0005).

Longitudinal Trends

The PSCI scores were analyzed for longitudinaltrends in the three urban watersheds. The variabilityin PSCI scores among sites in the same urban water-shed was high, as compared to the variability in thereference watershed (see measures of standard devia-tion in Table 5). Swamp and Little Bear Creeks had



Figure 3. Plot of Channel Cross-Sectional Area Versus Watershed Area for the StudySites With a Regional Regression Line for Nonurban Streams (Booth, 1990).

(1)


MCBRIDE AND BOOTH

TA

BL

E 6

. Pea

rson

's C

orre

lati

on C

oeff

icie

nts

Bet

wee

n L

ands

cape

Met

rics

(-

not

sig

nif

ican

t, p

> 0

.05)

.

Lan

dsc

ape

Ab

bre

-M

etri

cvi

atio

nIU

SU

BIU

B1

IUB

2IU

L1

IUL

2IG

US

UB

IGU

B1

IGU

B2

IGU

L1

IGU

L2

TU

SU

BT

UB

1T

UB

2T

UL

1T

UL

2R

DD

MF

PL

UP

RD

Inte

nse

Urb

an L

and

(p

erce

nt)

Su

bwat

ersh

edIU

SU

B1

100

m b

uff

erIU

B1

0.92

1

200

m b

uff

erIU

B2

0.95

0.99

1

500

m lo

cal z

one

IUL

1-

--

1

1 km

loca

l zon

eIU

L2

--

-0.

871 In

ten

se a

nd

Gra

ssy

Urb

an L

and

(p

erce

nt)

Su

bwat

ersh

edIG

US

UB

0.89

0.87

0.87

--

1

100

m b

uff

erIG

UB

10.

830.

870.

86-

-0.

971

200

m b

uff

erIG

UB

20.

840.

870.

86-

-0.

981

1

500

m lo

cal z

one

IGU

L1

--

-0.

820.

73-

--

1

1 km

loca

l zon

eIG

UL

2-

--

0.7

0.8

--

-0.

841

Tot

al U

rban

Lan

d (

per

cen

t)

Su

bwat

ersh

edT

US

UB

0.62

0.65

0.64

--

0.87

0.89

0.91

0.37

0.38

1

100

m b

uff

erT

UB

10.

640.

660.

64-

-0.

880.

910.

910.

360.

370.

991

200

m b

uff

erT

UB

20.

60.

640.

63-

-0.

850.

890.

90.

390.

391

0.99

1

500

m lo

cal z

one

TU

L1

0.21

0.26

0.25

0.47

0.44

0.43

0.44

0.47

0.79

0.68

0.69

0.67

0.7

1

1 km

loca

l zon

eT

UL

20.

230.

290.

270.

410.

480.

460.

490.

510.

710.

790.

730.

720.

740.

911

Con

nec

tivi

ty M

etri

cs*

Roa

d de

nsi

tyR

DD

0.4

0.37

0.36

-0.2

6-0

.32

0.72

0.73

0.76

--

0.97

0.93

0.95

0.44

0.54

1

Med

ian

flo

w p

ath

MF

PL

-0.2

9-0

.26

-0.2

80.

23-

--

-0.

390.

47-

--

0.34

0.49

0.36

1le

ngt

h

Ups

trea

m d

ista

nce

UP

RD

--

--

--

--

--

--

--

--

-1

to r

oad

*Cor

rela

tion

s of

con

nec

tivi

ty m

etri

cs in

clu

ded

data

fro

m t

he

thre

e u

rban

wat

ersh

eds

only

.

the greatest overall range in PSCI score (Table 5).Plots of PSCI scores as a function of channel distancedemonstrate that conditions changed rapidly betweenconsecutive sites and that continuous downstreamtrends were not apparent (Figure 5). The change inPSCI scores of consecutive sites were found to rangefrom no change (∆ PSCI = 0) to substantial change (∆PSCI = 7.5). The change in PSCI scores between con-secutive sites was occasionally as great as the totalrange in PSCI scores within an entire watershed.

The intactness of the riparian buffer explainedsome of the longitudinal changes in the PSCI score.The PSCI scores were found to significantly improvein the downstream direction (∆ PSCI > 0) when the100 m buffer between sites was at least 35 percent

forested (p = 0.05; two-sample t-test with unequalvariance) (Zar, 1984). The differences in ∆ PSCI werehighly significant when sites were grouped using themedian value of forest buffer (50 percent), which facil-itated a two-sample t-test (unequal variance) withequal sample sizes (n = 27, p = 0.002, Figure 6a). Thepresence of road crossings between consecutive siteslikely promoted downstream decline in PSCI scores (∆PSCI < 0; Figure 6b). For consecutive sites with manyroad crossings between them, the ∆ PSCI was oftennegative. These two riparian factors were not signifi-cantly correlated (r = -0.19, p = 0.16) but appeared to act in concert (Figure 6c). When the buffer betweenconsecutive sites was either not fragmented by a road crossing or fragmented by less than three road



Figure 4. Plots of the Relationships Between PSCI Scores and Urbanization in the (A) Subwatershed,(B) 500 m Local Zone, and (C) 1,000 m Local Zone With Linear Regression Lines.

crossings per km, the downstream change in PSCIwas significantly higher for sites with at least a 50percent forested buffer (p = 0.08 and p = 0.03, two-sample t-test with unequal variance). When morethan three road crossings per km were present, aforested buffer was apparently less effective, resultingin a smaller and less significant relative improvementin PSCI scores (p = 0.10, two-sample t-test withunequal variance).

DISCUSSION

Heterogeneity in Physical Stream Conditions

Local instream physical conditions are heteroge-neous and are a function of the geomorphic context,

the urbanization of the watershed, and the landscapeconditions at the local scale. The range of physicalstream conditions was greatest for Little Bear Creekand Swamp Creek (Figure 4a, Figure 5, Table 5), sug-gesting that moderately urbanized watersheds maybe more heterogeneous than highly urbanized water-sheds or forested, nonurban watersheds. The hetero-geneity of moderately urbanized streams is partiallyexplained by the amount of urbanization in the localzone and the intactness of the local riparian buffer.The effects of local urbanization, road crossings, anddeforested riparian buffers may be more pronouncedin stream systems that have not been overwhelmedby the effects of extensive watershed scale urbaniza-tion. Watershed scale urbanization likely sets a maxi-mum attainable best condition, while local andriparian urbanization can further degrade physicalconditions. Road crossings appear to be a key point ofdisruption in urban streams, interrupting the ripari-an zone and providing a point source for storm waterdischarges. Other studies have pinpointed roads askey stressors in urban landscapes (May et al., 1997;Marina Alberti, University of Washington, personalcommunication, December 2003).

Physical Stream Conditions Index

The PSCI effectively integrates a variety of qualita-tive attributes that are strongly influenced by urban-ization into a meaningful, quantitative score. ThePSCI functions well as a general measure of the phys-ical integrity in streams, responding in an intuitivelyreasonable and statistically significant manner togradients of urbanization. The PSCI correlates wellwith the proportion of urban land in the subwater-shed and local zones (Figure 4). To further evaluatethe utility and robustness of the PSCI, however, itneeds further validation with other sampling efforts.

The applicability of the PSCI may be limited by thesampling and geographic scope of this study. Thisindex could be used in other Puget Sound Lowlandsmall order (first-order to third-order) streams with-out hesitation. Applying the PSCI beyond this regionor in larger-order streams, however, would not be rec-ommended without first testing its applicability. Withthat said, most of the PSCI’s individual metrics aremeasures of common symptoms of urban streams inother parts of this country and the world, such asbank instability (Neller, 1988; Trimble, 1997),increased channel size (Pizzuto et al., 2000; Hessionet al., 2003), and the loss of LWD (Booth et al., 1997;May et al., 1997).


MCBRIDE AND BOOTH

Figure 5. Longitudinal Profiles of PSCIScores for the Study Streams.

Measuring Urbanization

The quantity, location, and distribution of urban-ization can be successfully quantified with relativelysimple, GIS based landscape metrics. In someinstances, the variety of landscape metrics explored inthis study provided a more robust characterization ofthe urbanized landscape than more commonly used

lumped measures of urbanization, such as percenttotal impervious area. The urbanization of the localzone and the proximity to road crossings provided fur-ther explanation of the physical stream conditions ofeach site; however, some landscape metrics are soclosely related that they cannot help decipher streamconditions (i.e., urban land in the buffer zone withurban land in the subwatershed and road densitywith urban land). Other studies have uncovered simi-lar correlations between land cover in buffers andwatersheds (Fitzpatrick et al., 2001; Morley and Karr,2002; Wang and Kanehl, 2003). Although not usefulfor better understanding stream conditions, theserelationships between landscape metrics provideinsight to the nature of the urban landscape.

The pattern of urbanization in the Puget Soundlowlands appears to be fairly homogeneous, as isdemonstrated by two key results. First, urban land isevenly distributed throughout the study watershedsin relation to the stream network. The nearly equiva-lent median flow path lengths found in this studyindicate that urban land is not clustered near or farfrom any particular stream channel, which is consis-tent with the finding that the urbanization of riparianbuffers mirrors the urbanization of the entire water-shed. Second, urban areas appear to be equivalentlyconnected to the stream network, as measured by theconnectivity metrics in this study. An increase inurban land leads to an increase in the number ofroads connecting urban areas to stream channels.The minimal variation in median flow path lengthsamong the urban streams also demonstrates thaturban areas have uniform connectivity.

In contrast, other studies have found variations inconnectivity to be an important and influential factor(Bledsoe and Watson, 2000; Wang et al., 2001; Walsh,2004). Bledsoe and Watson (2000) have studied thechange in stream power associated with increasedimpervious areas and have found it to be sensitive tothe connectedness of those impervious areas. A studyof Wisconsin urban streams found that the amount ofconnected impervious area was the best measure ofurban impact to several biotic and physical indicators(Wang et al., 2001). A recent study in the Puget Soundregion has determined that the number of road cross-ings per stream kilometer best predicts biologicalintegrity in streams of 42 drainage basins (MarinaAlberti, University of Washington, personal communi-cation, December 2003). The importance of connectivi-ty in predicting urban stream conditions may be afunction of how connectivity is measured. Road densi-ty, as a metric of connectivity and as a coarse estimateof the extent of hydraulic connections to the channelnetwork, did not provide any additional explanatorypower in this study. If connectivity can be measuredusing more detailed information on storm water



Figure 6. Boxplots of Relationships Between the Change in PSCIScores (∆ PSCI) and (A) the Amount of Forested or Wetland

Buffer Between Consecutive Sites, (B) the Number ofRoad Crossings Between Consecutive Sites, and (C) the

Combined Effect of Buffer Conditions and Road Crossings.

drainage, as in Wang et al. (2001), it may be an impor-tant predictor of stream health.

Important Zones of Influence

Results suggest that physical stream conditions areimpacted by urbanization in the subwatershed andlocal zones to nearly equivalent degrees. The regres-sion of PSCI against subwatershed and local zoneintense and grassy urban land revealed that theselandscape metrics were equally important predictorvariables. The combination of intense and grassyurban land (IGU) had a better regression relationshipwith PSCI than the total urban land (TU) thatincludes forested urban land. Forested urban landswith low total impervious areas likely do not impacturban streams as severely as the other urban landareas. These results mirror those of other studies.Although watershed conditions are undeniably influ-ential, many studies have identified a disproportion-ate influence of the local or riparian zone (Steedman,1988; Wang et al., 2001; Morley and Karr, 2002; Wangand Kanehl, 2003). A similar study of several PugetSound streams found that biological integrity wasequally well predicted by urbanization in the water-shed and by urbanization in the local area (Morleyand Karr, 2002). Wang et al. (2001) found that con-nected impervious area immediately adjacent to astream, within either a local zone or a buffer zone,had the strongest influence on an index of bioticintegrity and base flow.

In this study, the R2 values of the various regres-sion models tested suggest that approximately halfthe variability in physical stream conditions, as mea-sured by PSCI, can be explained by various landscapemetrics. Therefore, landscape metrics should not beexpected to adequately predict stream conditions, andthey cannot be used as a surrogate to instreamassessments. Both GIS based analysis and instreamassessments of physical or biological conditions arerequired to evaluate any particular stream system.

Downstream Recovery

Longitudinal trends in the PSCI scores show thatpartial recovery of physical conditions is possiblewhere a degraded stream flows through an intactforested 100 m riparian buffer. Stream segments withroad crossings and without substantial forested riparian buffers tended to have PSCI scores thatdeclined in the downstream direction. The resultsshowed improved physical conditions where the 100m riparian buffer was at least 35 percent forested.

The greatest downstream improvements in physicalstream conditions were realized in areas that had fewroad crossings and substantial forest or wetlandriparian buffers within a 100 m corridor of the streamchannel.

There are several possible processes acting along astream channel that could improve physical condi-tions. Undeveloped riparian zones in the Puget SoundLowlands typically have active floodplains and ripari-an wetlands. The roughness of a forested riparianzone and wetland areas can attenuate peak stormflows and reduce specific stream power (Bledsoe andWatson, 2000). If the erosive force of peak flows canbe diminished, stream reaches will likely experienceless disturbance in their channels, resulting in morestable streambeds and banks. If forested riparianzones and wetlands can significantly slow peak flowsand temporarily store storm water, fine sediment sus-pended or carried in the water column has the poten-tial to filter out and remain deposited in wetlands oron floodplains or within the channel in bars. An intactforested riparian zone also allows the recruitment ofLWD and, by definition, precludes many directanthropogenic impacts, such as channel straighteningor streambank armoring.

Management Implications

The results of this study have specific managementimplications. The amount of development in a water-shed is extremely influential on the physical and bio-logical conditions in streams, which necessitateswatershed wide land use planning for successful pro-tection of streams. Watershed land use is not the soledeterminant of stream conditions, however, and astrategy that imposes only a watershed wide limit ondevelopment will be inadequate. Local land cover isextremely important to physical stream conditions,and therefore this zone of the watershed should havehigh priority in planning and regulations. If urbandevelopment can proceed while maintaining intact,undeveloped riparian buffers, the impact of urbaniza-tion should be less than from traditional developmentpatterns (Wang et al., 2001). The results also suggestrestoration potential for degraded urban streams. Ifriparian buffers can be reforested and road crossingseliminated or avoided in certain reaches of streams inwatersheds with moderate urbanization, partialrecovery of a stream’s physical integrity is possible.


MCBRIDE AND BOOTH

CONCLUSIONS

This study demonstrates that the effects of urban-ization on physical stream conditions are influencedby spatial scale and landscape patterns. Urbanizationof both the entire contributing watershed and the partof the watershed closest to the stream appear to haveapproximately equal weight in influencing a stream’sphysical conditions, analogous to a prior study of bio-logical conditions in the same region (Morley andKarr, 2002). Urbanization in the watershed is highlyinfluential to streams and likely sets a maximumattainable best condition, yet conditions are stronglymodified by the local landscape conditions. Physicalconditions can improve downstream from degradedstream reaches if the riparian zone is substantiallyforested and devoid of road crossings.

Results also highlight the utility of several method-ologies used in this study. The PSCI effectively inte-grates a set of physical attributes, responding in anintuitively reasonable and statistically significantmanner to gradients of urbanization in the PugetSound lowlands. The GIS based analysis generatedseveral landscape metrics that described the quantity,location, and distribution of urban land in the studywatersheds and explained much of the variability inphysical stream conditions. This study integratedthese methodologies to interpret the effects of spatialscale, connectivity, and longitudinal trends in urbanstreams.

In sum, with better information on the interactionof urbanization and stream ecosystems, policies andmanagement strategies for protecting stream integri-ty in developing areas can be improved. With morerobust knowledge the landscapes can be modified topreserve those streams or stream segments that stillfunction while targeting rehabilitation efforts to thosedegraded portions of streams that have realisticchances for improvement.

ACKNOWLEDGMENTS

Research was supported in part by the National Science Foun-dation Graduate Fellowship Program, the University of WashingtonValle Scholarship, and the Center for Water and Watershed Studies(formerly the Center for Urban Water Resources Management) atthe University of Washington. The authors thank Marina Alberti,Kristina Hill, Sarah Morley, and Daniele Spirandelli for sharingideas on urban patterns. The authors also thank one anonymousreviewer and Cully Hession and his graduate students for theirhelpful reviews.

LITERATURE CITED

Afifi, A., V.A. Clark, and S. May, 2004. Computer-Aided Multivari-ate Analysis (Fourth Edition). Chapman and Hall/CRC, BocaRaton, Florida.

Allan, J.D., D.L. Erickson, and J. Fay, 1997. The Influence of Catch-ment Land Use on Stream Integrity Across Multiple SpatialScales. Freshwater Biology 37(1):149-161.

Barbour, M.T., J. Gerritsen, B.D. Snyder, and J.B. Stribling, 1999.Rapid Bioassessment Protocols for Use in Streams and Wade-able Rivers: Periphyton, Benthic Macroinvertebrates and Fish(Second Edition). EPA 841-B-99-002, U.S. EnvironmentalProtection Agency, Office of Water, Washington, D.C.

Bledsoe, B.P. and C.C. Watson, 2000. Observed Thresholds ofStream Ecosystem Degradation in Urbanizing Areas: A Process-Based Geomorphic View. In: Watershed Management: Scienceand Engineering Technology for the New Millennium, MarshallFlug and Donald K. Frevert (Editors). American Society of CivilEngineers, Reston, Virginia, pp. 1-10.

Booth, D.B., 1990. Stream-Channel Incision Following Drainage-Basin Urbanization. Water Resources Bulletin 26(3):407-417.

Booth, D.B., 1991. Urbanization and the Natural Drainage System.Northwest Environmental Journal 7(1):93-118.

Booth, D.B. and P.C. Henshaw, 2001. Rates of Channel Erosion inSmall Urban Streams. In: Land Use and Watersheds: HumanInfluence on Hydrology and Geomorphology in Urban and For-est Areas, M. Wigmosta and S. Burges (Editors). AGU Mono-graph Series, Water Science and Application Volume 2, pp.17-38. Washington, D.C.

Booth, D.B. and C.R. Jackson, 1997. Urbanization of Aquatic Sys-tems: Degradation Thresholds, Stormwater Detection, and theLimits of Mitigation. Journal of the American Water ResourcesAssociation (JAWRA) 33(5):1077-1090.

Booth, D.B., D.R. Montgomery, and J. Bethel, 1997. Large WoodyDebris in Urban Streams of the Pacific Northwest. In: Effects ofWatershed Development and Management on Aquatic Ecosys-tems, L.A. Roesner, E. Shaver, and R.R. Horner (Editors). Engi-neering Foundation Conference Proceedings, American Societyof Civil Engineers, New York, New York, pp. 178-197.

Caraco, D.S., 2000. The Dynamics of Urban Stream ChannelEnlargement. Watershed Protection Techniques 3(3):729-734.

Center for Water and Watershed Studies, 1998. Landsat Land-Cover Classification. Center for Water and WatershedStudies, University of Washington. Available at http://depts.washington.edu/cwws/Research/Projects/landsat.html. Accessedin March 2005.

Dunne, T. and L.B. Leopold, 1978. Water in Environmental Plan-ning. W.H. Freeman and Co, New York, New York.

Finkenbine, J.K., J.W. Atwater, and D.S. Mavinic, 2000. StreamHealth After Urbanization. Journal of the American WaterResources Association (JAWRA) 36(5):1149-1160.

Fitzpatrick, F.A., B.C. Scudder, B.N. Lenz, and D.J. Sullivan, 2001.Effects of Multiscale Environmental Characteristics on Agricul-tural Stream Biota in Eastern Wisconsin. Journal of the Ameri-can Water Resources Association (JAWRA) 37(6):1489-1507.

Hammer, T., 1972. Stream Channel Enlargement Due to Urbaniza-tion. Water Resources Research 8(6):1530-1540.

Henshaw, P.C. and D.B. Booth, 2000. Natural Restabilization ofStream Channels in Urban Watersheds. Journal of the Ameri-can Water Resources Association (JAWRA) 36(6):1219-1236.

Hession, W.C., J.E. Pizzuto, T.E. Johnson, and R.J. Horowitz, 2003.Influence of Bank Vegetation on Channel Morphology in Ruraland Urban Watersheds. Geology 31(2):147-150.



Hill, K., E. Botsford, and D.B. Booth, 2003. A Rapid Land CoverClassification Method for Use in Urban Watershed Analysis.Water Resources Series, Technical Report No. 173, Departmentof Civil and Environmental Engineering, University of Washing-ton, Seattle, Washington.

Karr, J.R. and E.W. Chu, 1999. Restoring Life in Running Waters:Better Biological Monitoring. Island Press, Washington, D.C.

Knighton, D., 1998. Fluvial Forms and Processes: A New Perspec-tive. Arnold, London, United Kingdom.

Leopold, L.B. and M.G. Wolman, 1957. River Channel Patterns:Braided, Meandering and Straight. U.S. Geological Survey Pro-fessional Paper 282-B. U.S. Government Printing Office, Wash-ington, D.C.

May, C.W., R.R. Horner, J.R. Karr, B.W. Mar, and E.B. Welch, 1997.Effects of Urbanization on Small Streams in the Puget SoundLowland Ecoregion. Watershed Protection Techniques 2(4):483-494.

McBride, M., 2001. Spatial Effects of Urbanization on PhysicalConditions in Puget Sound Lowland Streams. Water ResourcesSeries, Technical Report No. 177, Department of Civil and Envi-ronmental Engineering, University of Washington, Seattle,Washington.

Montgomery, D.R. and J.M. Buffington, 1998. Channel Processes,Classification, and Response. In: River Ecology and Manage-ment, Robert J. Naiman and Robert E. Bilby (Editors). Springer-Verlag, New York, New York, pp. 13-42.

Montgomery, D.R., J.M. Buffington, R.D. Smith, K.S. Schmidt, andG. Pess, 1995. Pool Spacing in Forest Channels. Water ResourcesResearch 13(4):1097-1105.

Morley, S.A. and J.R. Karr, 2002. Assessing and Restoring theHealth of Urban Streams in the Puget Sound Basin. Conserva-tion Biology 16(6):1498-1509.

Neller, R.J. 1988. A Comparison of Channel Erosion in Small Urbanand Rural Catchments, Armidale, New South Wales. Earth Sur-face Processes and Landforms 13(1):1-7.

Pizzuto, J.E., W.C. Hession, and M. McBride, 2000. ComparingGravel-Bed Rivers in Paired Urban and Rural Catchments ofSoutheastern Pennsylvania. Geology 28(1):79-82.

Plafkin, J.L., M.T. Barbour, K.D. Porter, S.K. Gross, and R.M. Hugh-es, 1989. Rapid Bioassessment Protocols for Use in Streams andRivers: Benthic Macroinvertebrates and Fish. EPA 440-4-89-001, U.S. Environmental Protection Agency, Office of Water Reg-ulations and Standards, Washington, D.C.

Rankin, E.T., 1995. Habitat Indices in Water Resource QualityAssessments. In: Biological Assessment and Criteria: Tools forWater Resource Planning and Decision Making, W.S. Davis andT.P. Simon (Editors). Lewis Publishers, Boca Raton, Florida, pp.181-208.

Raven, P.J., N.T.H. Holmes, F.H. Dawson, P.J.A. Fox, M. Everard,I.R. Fozzard, and K.J. Rowen, 1998. River Habitat Quality: ThePhysical Character of Rivers and Streams in the UK and Isle ofMan. Environment Agency, ISBN1 873760 42 9, Bristol, Eng-land.

Roesner, L.A. and B.P. Bledsoe, 2003. Physical Effects of WetWeather Flows on Aquatic Habitats: Present Knowledge andResearch Needs. IWA Pub., London, United Kingdom.

Roth, N.E., J.D. Allan, and D.E. Erickson, 1996. Landscape Influ-ences on Stream Biotic Integrity Assessed at Multiple SpatialScales. Landscape Ecology 11(3):141-156.

Scholz, J.G. and D.B. Booth, 2001. Monitoring Urban Streams:Strategies and Protocols for Humid-Region Lowland Systems.Environmental Monitoring and Assessment 71(2):143-164.

Schuft, M.J., T.J. Moser, P.J. Wigington, D.L. Stevens, L.S. McAllis-ter, S.S. Chapman, and T.L. Ernst, 1999. Development of Land-scape Metrics for Characterizing Riparian-Stream Networks.Programmetric Engineering and Remote Sensing 65(10):1157-1167.

SPSS, Inc., 1999. SPSS Base 10.0 User’s Guide. SPSS, Inc., Chica-go, Illinois, 537 pp.

Steedman, R.J., 1988. Modification and Assessment of an Index ofBiotic Integrity to Quantify Stream Quality in SouthernOntario. Canadian Journal of Fisheries and Aquatic Science45(3):492-501.

Trimble, S.W., 1997. Contribution of Stream Channel Erosion toSediment Yield From an Urbanizing Watershed. Science278(5342):1442-1444.

Turner, M.G. and R.H. Gardner, 1991. Quantitative Methods inLandscape Ecology. Springer-Verlag, New York, New York.

University Libraries, 1999. Washington State Geospatial DataArchive: 10 Meter Digital Elevation Models. UniversityLibraries, University of Washington. Available at http://wagda.lib.washington.edu/data/washdata.html. Accessed in March2005.

U.S. Fish and Wildlife Service, 1987-1989. National WetlandsInventory: Wetlands Digital Data. Available at http://wetlands-fws.er.usgs.gov/download.html. Accessed in March 2005.

Walsh, C.J., 2004. Protection of In-Stream Biota From UrbanImpacts: Minimise Catchment Imperviousness or ImproveDrainage Design? Marine Freshwater Research 55:317-326.

Wang, L., J. Lyons, P. Kanehl, and R. Bannerman, 2001. Impacts ofUrbanization on Stream Habitat and Fish Across Multiple Spa-tial Scales. Environmental Management 28(2):255-266.

Wang, L. and P. Kanehl, 2003. Influences of Watershed Urbaniza-tion and Instream Habitat on Macroinvertebrates in Cold WaterStreams. Journal of the American Water Resources Association(JAWRA) 39(5):1181-1196.

Wear, D.N., M.G. Turner, and R.J. Naiman, 1998. Land Cover Alongan Urban-Rural Gradient: Implications for Water Quality. Eco-logical Applications 8(3):619-630.

Wolman, M.G., 1954. A Method of Sampling Coarse River-BedMaterial. Transactions of the Geophysical Union 35:951-956.

Zar, J.H., 1984. Biostatistical Analysis. Prentice-Hall, EnglewoodCliffs, New Jersey.

Zipperer, W.C., J. Wu, R.V. Pouyat, and S.T.A. Pickett, 2000. TheApplication of Ecological Principles to Urban and UrbanizingLandscapes. Ecological Applications 10(3):685-688.


MCBRIDE AND BOOTH

urban impacts on physical stream condition: effects...

Documents