Spatial Consistency of Chinook Salmon Redd Distribution ...

This article was downloaded by: [University of Washington Libraries]On: 29 April 2013, At: 09:10Publisher: Taylor & FrancisInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,37-41 Mortimer Street, London W1T 3JH, UK

North American Journal of Fisheries ManagementPublication details, including instructions for authors and subscription information:http://www.tandfonline.com/loi/ujfm20

Spatial Consistency of Chinook Salmon ReddDistribution within and among Years in the CowlitzRiver, WashingtonKatherine J. C. Klett a b , Christian E. Torgersen b , Julie A. Henning c & Christopher J.Murray da School of Environmental and Forest Sciences , University of Washington , Box 352100,Seattle , Washington , 98195 , USAb U.S. Geological Survey, Forest and Rangeland Ecosystem Science Center, Cascadia FieldStation, University of Washington , School of Environmental and Forest Sciences , Box352100, Seattle , Washington , 98195 , USAc Washington Department of Fish and Wildlife , 600 Capitol Way North , Olympia ,Washington , 98501 , USAd Pacific Northwest National Laboratory , Post Office Box 999 , Richland , Washington ,99352 , USAPublished online: 28 Apr 2013.

To cite this article: Katherine J. C. Klett , Christian E. Torgersen , Julie A. Henning & Christopher J. Murray (2013): SpatialConsistency of Chinook Salmon Redd Distribution within and among Years in the Cowlitz River, Washington, North AmericanJournal of Fisheries Management, 33:3, 508-518

To link to this article: http://dx.doi.org/10.1080/02755947.2013.778924

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions

This article may be used for research, teaching, and private study purposes. Any substantial or systematicreproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form toanyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation that the contentswill be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses shouldbe independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims,proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly inconnection with or arising out of the use of this material.

North American Journal of Fisheries Management 33:508–518, 2013C© American Fisheries Society 2013ISSN: 0275-5947 print / 1548-8675 onlineDOI: 10.1080/02755947.2013.778924

ARTICLE

Spatial Consistency of Chinook Salmon Redd Distributionwithin and among Years in the Cowlitz River, Washington

Katherine J. C. Klett*School of Environmental and Forest Sciences, University of Washington, Box 352100, Seattle,Washington 98195, USA; and U.S. Geological Survey, Forest and Rangeland Ecosystem Science Center,Cascadia Field Station, University of Washington, School of Environmental and Forest Sciences,Box 352100, Seattle, Washington 98195, USA

Christian E. TorgersenU.S. Geological Survey, Forest and Rangeland Ecosystem Science Center, Cascadia Field Station,University of Washington, School of Environmental and Forest Sciences, Box 352100, Seattle,Washington 98195, USA

Julie A. HenningWashington Department of Fish and Wildlife, 600 Capitol Way North, Olympia, Washington 98501, USA

Christopher J. MurrayPacific Northwest National Laboratory, Post Office Box 999, Richland, Washington 99352, USA

AbstractWe investigated the spawning patterns of Chinook Salmon Oncorhynchus tshawytscha on the lower Cowlitz River,

Washington, using a unique set of fine- and coarse-scale temporal and spatial data collected during biweekly aerialsurveys conducted in 1991–2009 (500 m to 28 km resolution) and 2008–2009 (100–500 m resolution). Redd locationswere mapped from a helicopter during 2008 and 2009 with a hand-held GPS synchronized with in-flight audiorecordings. We examined spatial patterns of Chinook Salmon redd reoccupation among and within years in relationto segment-scale geomorphic features. Chinook Salmon spawned in the same sections each year with little variationamong years. On a coarse scale, 5 years (1993, 1998, 2000, 2002, and 2009) were compared for reoccupation. Reddlocations were highly correlated among years. Comparisons on a fine scale (500 m) between 2008 and 2009 alsorevealed a high degree of consistency among redd locations. On a finer temporal scale, we observed that ChinookSalmon spawned in the same sections during the first and last week. Redds were clustered in both 2008 and 2009.Regression analysis with a generalized linear model at the 500-m scale indicated that river kilometer and channelbifurcation were positively associated with redd density, whereas sinuosity was negatively associated with redd density.Collecting data on specific redd locations with a GPS during aerial surveys was logistically feasible and cost effectiveand greatly enhanced the spatial precision of Chinook Salmon spawning surveys.

*Corresponding author: [email protected] August 10, 2012; accepted February 14, 2013

508

Dow

nloa

ded

by [

Uni

vers

ity o

f W

ashi

ngto

n L

ibra

ries

] at

09:

10 2

9 A

pril

2013

SPATIAL CONSISTENCY OF CHINOOK SALMON REDD DISTRIBUTION 509

Chinook Salmon Oncorhynchus tshawytscha are known tospawn consistently in the same areas year after year, and yetthere have been few published papers that have statisticallyevaluated spatiotemporal consistency in spawning patterns ofsalmon (Oncorhynchus spp.) over decades. Furthermore, previ-ous research on spawning distributions has been conducted atthe reach scale (∼101 m; Geist et al. 2000) and at the basin scale(∼103 m or larger; Isaak and Thurow 2006; Isaak et al. 2007[see a comprehensive review of previous research on salmonidspawning distribution at various spatial scales by Beechie et al.2008]). To fully understand spawning patterns, it is importantto examine patterns at multiple scales, both spatially and tem-porally (Geist and Dauble 1998; Fausch et al. 2002). On a reachscale, Geist et al. (2000) investigated the spatial and tempo-ral patterns of fall Chinook Salmon spawning sites in a spatialgrid with cells that were 20 × 20 m wide in 1994 and 1995 inthe Hanford Reach, Columbia River (Washington). Isaak andThurow (2006) examined basin-scale patterns over an entirewatershed and among tributaries in central Idaho, using 28-km sections for analysis. In contrast to these previous stud-ies, we evaluated spawning patterns of Chinook Salmon in thelower 80 km of the Cowlitz River at the segment scale (∼102

m), which is intermediate to the reach (∼101 m) and basin(>103 m) scales (Frissell et al. 1986).

Previous studies of redd distribution at a reach scale have em-phasized patterns of sediment and water quality parameters inand adjacent to individual redds (Bjornn and Reiser 1991; Geistet al. 2000; Geist et al. 2002; Malcolm et al. 2003; Moir et al.2004). To complement previous work at finer spatial scales,we investigated the effects of geomorphic features at a seg-ment scale. Channel bifurcation (Dauble et al. 2003), sinuosity(Dauble and Geist 2000), tributaries (Martin et al. 2004; Riceet al. 2008), depth discontinuities (Brunke and Gonser 1997),and channel gradient (Dauble and Geist 2000) have been iden-tified as potential segment-scale controls on redd distribution.

Fall and spring Chinook Salmon in the Cowlitz River arepart of the Lower Columbia River Evolutionarily SignificantUnit and are listed as threatened under the Endangered SpeciesAct (Good et al. 2005). The number of wild adult fall ChinookSalmon spawning in the Cowlitz River was estimated at 100,000adults historically but has dropped to less than 2,000 individualsrecently (LCFRB 2004). There were 1,620 redds counted duringaerial surveys in 2007, including both wild and hatchery springand fall Chinook Salmon (Henning 2008). This decline in Chi-nook Salmon has been attributed to numerous causes but is mostlikely due to overfishing and habitat degradation. In the lowerCowlitz basin, there have been many human-caused changessuch as dams, dredging, diking, and straightening of the mainchannel, which have had a negative effect on salmon habitat(LCFRB 2004). This habitat is crucial for juvenile salmon, butit also has a large impact on spawning adults and on the survivalof their eggs (Sear and DeVries 2008).

Determining the patterns of Chinook Salmon spawningwithin the Cowlitz River is necessary for conservation and

restoration. Additional information about where salmon arespawning and why they choose certain areas allows for moreefficient management of the fishery. Fisheries managers can usethis information to identify the locations and geomorphic char-acteristics of reaches that are occupied repeatedly by salmonand to more effectively protect these areas. The objectives ofthis study were to (1) examine temporal consistency (i.e., reoc-cupation) of spawning locations within a single year and amongyears, (2) determine if Chinook Salmon redds were distributedin distinct aggregations throughout the Cowlitz River, and(3) investigate the segment-scale habitat features that may beaffecting redd distribution.

METHODS

Study AreaThe Cowlitz River is located in southwestern Washington

(Figure 1), and the lower Cowlitz basin encompasses approx-imately 1,140 km2. The 82-km study section is located on the

FIGURE 1. Study area on the lower Cowlitz River in southwestern Wash-ington. Study sections 1–8 are demarcated by black bar symbols labeled withthe section number in a circle. River kilometer (rkm) markers at 5-km intervalsindicate the location upstream from the mouth of the Cowlitz River.

Dow

nloa

ded

by [

Uni

vers

ity o

f W

ashi

ngto

n L

ibra

ries

] at

09:

10 2

9 A

pril

2013

510 KLETT ET AL.

main-stem Cowlitz River between Kelso and the Barrier Dam,which is a diversion dam that allows fish to be captured andreleased in the upper river. Above the Barrier Dam, the CowlitzRiver has three hydroelectric dams: Mayfield Dam, MossyrockDam, and Cowlitz Falls Dam. The Cowlitz River Salmon Hatch-ery is located at the Barrier Dam and has been operated by theWashington Department of Fish and Wildlife (WDFW) withsupport from Tacoma Power since 1968. Spring and fall ChinookSalmon and Coho Salmon O. kisutch are produced in the hatch-ery but also spawn naturally in the Cowlitz River and upstreamtributaries (Tipping and Busack 2004). There is no volitionalfish passage through the Barrier Dam, and all fish that spawnin historic spawning areas above the Barrier Dam have beentransported (Fulton 1968). The geologic setting of the lowerCowlitz valley consists of Eocene basalt flows and flow brec-cias and has a typical maritime climate with warm (16–24◦C),dry (17–89 mm precipitation) summers and cool (8–17◦C), wet(100–200 mm precipitation) winters. The majority of precipi-tation occurs as rain between October and March, resulting inpeak flows occurring during these months; however, peak flowsoccasionally occur in the spring due to snowmelt. The flows be-low the Barrier Dam are regulated by hydroelectric dams, andany extremes in the hydrograph are moderated by flow regula-tion. Below Mayfield Dam, 80% of the land use within the basinis commercial timber harvest. Human population increases inCowlitz and Lewis counties have also resulted in increased resi-dential, industrial, and agricultural development along the river.Forest climax species are western hemlock Tsuga heterophylla,Douglas fir Psuedotsuga menziesii, and western red cedar Thujaplicata, while red alder Alnus rubra, black cottonwood Popu-lus trichocarpa, bigleaf maple Acer macrophyllum, and willowSalix spp. dominate the riparian areas (LCFRB 2004).

Data CollectionField surveys.—Aerial surveys of redds on the lower Cowlitz

River have been conducted by WDFW for Tacoma Power sincethe 1940s (Mark LaRiviere, Tacoma Power, personal communi-cation) to evaluate population trends. We used data from 1991to 2009 in this study because they were easily accessible andthe methods from these years were most consistent. Four to sixhelicopter flights a year were conducted on a biweekly basis, de-pending on weather and river conditions, from mid-Septemberthrough December. The accuracy and availability of redd datawere dependent on weather and river conditions, which couldcause very poor visibility due to high runoff during heavy rain-fall. Chinook Salmon redds were easily observed from the airdue to their large size and high visibility (Isaak and Thurow2006). The river was divided into eight sections that ranged inlength from 0.5 to 28.0 km and were demarcated by landmarkseasily identified from the air, such as boat ramps and bridges(Figure 1). For each survey, Chinook Salmon redds were countedand recorded in each section. Each individual redd was countedwhenever possible, but when densities were too high to countaccurately, the number of redds was estimated.

Spring and fall Chinook Salmon both spawn in the main stemof the Cowlitz River, and we did not differentiate between thetwo runs when redds are counted during aerial surveys. Thereis evidence from carcass surveys that spring Chinook Salmontypically spawn in late September, during the time of the firstaerial flight. For fish management purposes, WDFW uses a runtiming date of September 30 to differentiate spring and fall adultsalmon to the hatcheries. Carcass surveys show that the numberof fall Chinook Salmon is consistently much higher than thenumber of spring Chinook Salmon, even during the time pe-riod when the two runs overlap. For example, in 2008 and 2009the population estimates for spring Chinook Salmon were 425and 763, respectively. Estimates for fall Chinook Salmon duringthese two years were 2,100 and 2,800, respectively (Q. Daugh-erty, Washington Department of Fish and Wildlife, unpublisheddata). Because the relative abundance of spring Chinook Salmonwas so low in comparison to fall Chinook Salmon, we refer toChinook Salmon redds in general for the purposes of this study.We do, however, refer to the two different runs in our analysisof the spatial distribution of the first and last redds.

In 2008 and 2009, a digital audio recorder (Olympus Digi-tal Voice Recorder WS-2105) and a global positioning system(GPS; Garmin GPSmap 60CS) with a track log of geographicpositions recorded at 1-s intervals were used to map the loca-tions of each redd or cluster of redds. The recorded accuracyspecified by the GPS at the time of measurement was approxi-mately 15 m. All observations were made while the helicopterwas in flight, but we were not able to hover over each individualredd. The entire flight was recorded in the GPS track log, andthe times when redds were observed were noted on the audiorecorder. When there were too many redds to identify individ-ually, we attempted to count all of the redds in that area andthen associated one GPS point with the aggregation of redds.This was often done in the section between Mill Creek BoatLaunch and the Barrier Dam (Section 8) because there werehigh densities of redds and it was difficult to map each reddaccurately with single GPS points. Visser et al. (2002) docu-mented that redd counts are often underestimated due to highdensities; therefore, these estimates may be slightly lower thanthe actual number. The length of Section 8 was approximately500 m, and this distance was used as the minimum bin length(i.e., spatial resolution) for spatial analysis. Redd locations weregeoreferenced in the laboratory by synchronizing the GPS tracklog with the digital audio recordings from the flight. This in-formation was then transferred into a geographic informationsystem (GIS) database (ArcMap 9.3; Environmental SystemsResearch Institute, Redlands, California).

To characterize the depth patterns throughout the river, aSolinst LTC Levelogger Junior pressure transducer (verticalaccuracy = 0.01 m; Model 3001; Solinst, Georgetown, On-tario) was towed behind a drift boat, near the river bottom, torecord barometric pressure measurements every 2 s (Vaccaroand Maloy 2006). The drift boat floated with the current orwas rowed at approximately 4 km/h in slow-moving sections

Dow

nloa

ded

by [

Uni

vers

ity o

f W

ashi

ngto

n L

ibra

ries

] at

09:

10 2

9 A

pril

2013

SPATIAL CONSISTENCY OF CHINOOK SALMON REDD DISTRIBUTION 511

during the 3-d float of the river. Field measurements of riverdepth were conducted in August 2009. A GPS track log with1-s time stamps was used to georeference the barometric pres-sure measurements. The GPS and the Solinst Levelogger weresynchronized to an atomic clock prior to deployment.

Spatial analysis and GIS.—Geomorphic variables at the 500-m segment scale that were evaluated in relation to redd distri-bution included distance upstream from the river mouth (riverkilometer), channel bifurcation, tributary junctions, sinuosity,channel gradient, and depth discontinuities. To quantify thesegeomorphic variables, we used 1:24,000 U.S. Geological Sur-vey (USGS) topographical maps, 2009 National AgricultureImagery Program aerial photos (1 m resolution), and 10-m digi-tal elevation models in ArcMap. A stream line was created usingthe National Agriculture Imagery Program photos. Linear ref-erencing was used to assign route measures to the entire river,redds, and geomorphic features.

Segment-scale depth discontinuities are similar to transitionareas between pools and riffles but on a larger scale (e.g., 500 m).The average relative depth over each segment was calculatedfrom the barometric pressure measurements. We identified rela-tive differences in the depth profile, where the river transitionedfrom a decreasing depth to an increasing depth. The absoluteamount of change in depth was not calculated. Segments withdepth discontinuities may have increased hyporheic–surface wa-ter interactions, which are known to affect redd site selection byspawning salmon (Brunke and Gonser 1997).

River characteristics that were evaluated included river kilo-meter, channel bifurcation, tributary junctions, sinuosity, andchannel gradient. To examine channel morphology, we mea-sured the length of multiple channels associated with islands ineach section (Figure 2); the total length of channels by sectionprovided a relative measure of the degree of channel bifurcation.Channel bifurcation was defined as the splitting of a channelinto two or more active channels. To identify tributary junc-tions, we used a USGS topographical map and a 10-m digitalelevation model to calculate flow accumulation (ArcMap Spa-tial Analyst toolbox). We included tributaries that appeared inboth the flow accumulation analysis and were indicated on theUSGS topographical maps. To calculate sinuosity, we dividedthe main-stem river route into 500-m segments and then usedHawth’s Tools (Beyer 2004) to measure sinuosity (Rayburn and

FIGURE 2. Schematic diagram of a bifurcated river channel with an islandand two river channels. The black line indicates the length of side channel usedin calculations of channel characteristics per 500-m bin.

Schulte 2009; Yan et al. 2010) for each segment. To calculatechannel gradient, we used a method similar to Dauble and Geist(2000) based on USGS 1:24,000 topographical maps.

Statistical AnalysisFive different statistical analyses were conducted to examine

the reoccupation, randomness, and habitat associations of theredds (Table 1). Fine-scale (500-m) spatial data were only avail-able in 2008 and 2009. The coarse-scale spatial data were usedonly to investigate reoccupation historically. Reoccupation wasexamined at coarse and fine spatial and temporal scales. Thefine-scale spatial data from 2008 and 2009 were used to de-termine whether redds were distributed randomly or in distinctaggregations. Regression analysis was performed to assess as-sociations between habitat features and the number of redds;this analysis incorporated the fine-scale spatial data from 2009.

Reoccupation and spatial aggregation.—Reoccupationbetween the locations of historic redds (starting in 1991)and current redds in 2009 was examined by comparing reddcounts in each section of the river over time (Figure 1). Due toavailability of redd data, statistical analysis was performed forfive nonconsecutive years: 1993, 1998, 2000, 2002, and 2009(Table 1). We chose survey years that had at least five flightsover the entire length of the river. There were five flights in 2004,but because spawning distributions in 2002 and 2004 were verysimilar, data from 2002 was used as that year had more redds. We

TABLE 1. Spatial and temporal scales of data and survey years used in statistical analyses.

Statistical analysis Spatial scale Years Weeks

Reoccupation Coarse-scale redd data 1993, 1998, 2000,2002, and 2009

All

Reoccupation Fine-scale redd data 2008, 2009 AllReoccupation Fine-scale redd data 2008, 2009 First and lastSpatial aggregation Fine-scale redd data 2008, 2009 AllGeneralized linear model (GLM) Fine-scale redd data and

segment-scale habitat data2009 All

Dow

nloa

ded

by [

Uni

vers

ity o

f W

ashi

ngto

n L

ibra

ries

] at

09:

10 2

9 A

pril

2013

512 KLETT ET AL.

did not use data from 2008 because those data were very similarto data from 2009 (as shown by the fine-scale comparisonbetween 2008 and 2009). These 5 years represented a large rangein the annual number of redds (714–5,643 between 2000 and2002). Some years had fewer flights due to weather conditionsor water visibility (i.e., high flows). To statistically analyze re-occupation, we ranked the sections by redd density (number/m2)in each year. We used R (R Development Core Team 2009) tocalculate a correlation matrix with associated P-values and ad-justed P-values, using Holm’s method, to account for multiplecorrelations (Wright 1992). To quantify the linear trend that wasobserved, we also calculated a correlation coefficient betweenthe section number and the rank for each of the 5 years analyzed.

To evaluate reoccupation on a finer spatial scale between2008 and 2009 (Table 1), we divided the river into 500-m sec-tions and determined whether each section was occupied duringany week in 2008 or 2009. We used R to create 2 × 2 con-tingency tables and calculate the G-statistic, which is an analogfor the χ2 test of independence (Zar 1999). Contingency tableswere calculated based on whether a section was (1) occupiedin both 2008 and 2009, (2) never occupied, or (3) occupied inone year or the other. The G-statistic test compared the observeddistribution to a completely random distribution and determinedwhether there was more reoccupation than would have been ex-pected, given a random distribution of spawning locations inthe 2 years. We also examined reoccupation of redd locationswithin years for survey data in 2008 and 2009 (Table 1) with con-tingency tables and the G-statistic (Zar 1999). Because springChinook Salmon typically spawn during the first week and fallChinook Salmon spawn during the later weeks of our survey pe-riod, we examined whether the potential combination of springand fall runs would result in different distributions in the firstversus the last weeks of the survey period.

We compared the spatial pattern of redd distribution in 2008and 2009 to a random distribution using standard and medianruns analysis (α = 0.05; Turechek and Madden 1999; Borradaile2003; Gent et al. 2006). For the standard runs analysis, a sectionwas coded with either “0” if there were no redds, or “1” if therewere redds present. The median runs analysis calculated themedian number of redds per year and then coded a section “0”if n ≤ median or “1” if n > median. To ensure that the samplingdistribution of the total number of runs followed a normal dis-tribution, we counted the zeroes and ones and verified that therewere at least nine of each (Borradaile 2003). Redds were de-termined to be clustered if the number of runs was significantlylower than what would be expected given a random distribution.

Segment-scale habitat associations.—We used multiple re-gression with a generalized linear model (GLM) to assess therelationships between redd locations from 2009 and the six ge-omorphic variables (river kilometer, channel bifurcation, sinu-osity, tributary junction presence, depth discontinuity presence,and channel gradient; Kutner et al. 2004). The analysis was com-pleted at the 500-m scale for redd counts in 2009, which hadhigher redd counts than in 2008. We used the reoccupation re-sults to determine whether it would be sufficient to evaluate only

1 year. Multicollinearity was addressed using variance inflationfactor analysis and correlation matrices (Mansfield and Helms1982). To investigate autocorrelation among residuals, we gen-erated an autocorrelation function plot for the final model resid-uals. Variables were removed using a backwards-eliminationstepwise regression with an elimination value of P = 0.05 (Kut-ner et al. 2004). Distributions that were examined as possibilitiesfor the GLM included the Poisson distribution and the nega-tive binomial distribution or quasi-Poisson distribution, whichare both statistically appropriate when data are overdispersed(Lawless 1987; Ver Hoef and Boveng 2007). Akaike informa-tion criterion (AIC) values and weights were used to validatestepwise variable selection (Buckland et al. 1997; Burnham andAnderson 2002; Torgersen and Close 2004). Alignment betweenpeaks from observed data and fitted model values was visuallyevaluated as a metric of model fit.

RESULTS

Reoccupation and Spatial AggregationChinook Salmon consistently spawned in the same propor-

tions in the same sections of the Cowlitz River during the 5 yearsthat were examined (Figure 3). The two sections closest to theBarrier Dam consistently had the highest density of redds. Therewere two reaches in which sections alternated in rank (i.e., sec-tions 2 and 3 and sections 5 and 6). The lowest correlation coeffi-cient between sections was 0.90, and the adjusted P-values wereall less than 0.002. This indicated that (1) the ranks of sectionsby redd density for all of the years were highly correlated and (2)there was little variation in the ranks of the sections among years.The longitudinal trend in redd density with increasing densityupstream was generally consistent among sections (Figure 3).The lowest correlation coefficient between section number andrank was 0.93, thus supporting the observation of the linear trendin Figure 3.

For the fine-scale reoccupation analysis of redd distributionsin 2008 and 2009, the G-statistic was 17.14 (P < 0.001), indi-cating that redds occurred in the same locations in these 2 years(Figure 4). Spatial patterns of redd density were very similarin 2008 and 2009, even though the number of redds was 120%higher in 2009 (n = 2,728) than it was in 2008 (n = 1,247).Thus, we used data from 2009 for the habitat association study.The redds from the first and last weeks of the survey periodoccurred in the same sections in 2008 (P < 0.02) and 2009(P < 0.001). Redds were clustered in 2008 and 2009 basedon both the standard and median runs analyses (P < 0.001;Figure 4).

Segment-Scale Habitat AssociationsMulticollinearity among explanatory variables in the 500-m

model was not detected, and variance inflation factor values wereslightly above 1 and none were larger than 2.0. Autocorrelationvalues were not larger than the 95% confidence interval lines,nor were there significant longitudinal trends; therefore, we as-sumed that there was no significant autocorrelation (Neumann

Dow

nloa

ded

by [

Uni

vers

ity o

f W

ashi

ngto

n L

ibra

ries

] at

09:

10 2

9 A

pril

2013

SPATIAL CONSISTENCY OF CHINOOK SALMON REDD DISTRIBUTION 513

FIGURE 3. Ranks of redds/m2 by study section for 1993, 1998, 2000, 2002, and 2009. A rank of 8 indicates the greatest number of redds/m2. Section locationsare depicted in Figure 1.

et al. 2003). When using a Poisson distribution for the GLM,the data were overdispersed (Ver Hoef and Boveng 2007). Thequasi-Poisson and negative binomial regressions both resultedin the same significant variables. The results from the negativebinomial regression were used because (1) they included AICvalues, which we used for model comparison purposes, and(2) quasi-Poisson regressions cannot calculate AIC values (VerHoef and Boveng 2007).

Stepwise model selection and AIC values and weights in-dicated that the best model for predicting redd distribution in2009 included river kilometer, channel bifurcation, and sinuosity(Table 2; Figure 5). Redd density was positively associated withriver kilometer and channel bifurcation and negatively associ-ated with sinuosity (Table 3; Figure 5). River kilometer was themost significant variable but not the only important variable, asindicated by changes in AIC values. The 500-m model correctly

Dow

nloa

ded

by [

Uni

vers

ity o

f W

ashi

ngto

n L

ibra

ries

] at

09:

10 2

9 A

pril

2013

514 KLETT ET AL.

FIGURE 4. Spatial distribution of Chinook Salmon redds in 2008 and 2009 in the Cowlitz River downstream of the Barrier Dam. Pointer arrows indicate sectionbreaks and the numbers indicate the section label.

predicted the locations of 74% of the peaks in the observed data,but the model often underpredicted the total number of reddsexpected in those peaks (Figure 6).

DISCUSSIONChinook Salmon in the Cowlitz River spawned in clusters

and reoccupied the same areas at different spatial and temporalscales. To assess redd distribution at these different scales, weused data and analytical methods at scales that were different

from previous studies. For example, our analyses were at the seg-ment scale (80 km) and incorporated relatively high-resolutionspatial data (500 m). Isaak and Thurow (2006) used cumulativecurves and the Shannon–Wiener diversity index to demonstratethat redds were distributed nonrandomly in the Salmon Riverwatershed in central Idaho, but these analyses were conductedat a basin scale. Neville et al. (2006) also examined random-ness in the Salmon River redd data at a number of spatial scales(1, 2, 5, 10, and 20 km) and found that redds were distributed inclusters; however, they used autocorrelation function plots for

Dow

nloa

ded

by [

Uni

vers

ity o

f W

ashi

ngto

n L

ibra

ries

] at

09:

10 2

9 A

pril

2013

SPATIAL CONSISTENCY OF CHINOOK SALMON REDD DISTRIBUTION 515

TABLE 2. Candidate GLMs and corresponding AIC values and weights. Thetop model was selected using backwards-elimination stepwise regression andAIC values and weights. The model with the lowest AIC was identified asthe best model. Greater differences between respective models based on AICweights indicate a better fit. Abbreviations are as follows: RKM = river kilo-meter, CB = channel bifurcation, SIN = sinuosity, TJP = tributary junctionpresence, DDP = depth discontinuity presence, and CG = channel gradient.

AIC AICModel value weight

RKM + CB + SIN 795.43 0.52RKM + CB + SIN + TJP 796.45 0.31RKM + CB + SIN + TJP + CG 798.26 0.13RKM + CB + SIN + TJP + DDP + CG 800.24 0.05

their analysis. Geist et al. (2000) investigated Chinook Salmonspawning at a reach scale using a nearest neighbor analysis andcontingency tables and found that spawning occurred in clus-tered patterns in the same areas for two different years. It is

FIGURE 5. Longitudinal profiles of bifurcated channel length, sinuosity, chan-nel gradient (%), and redds (number/500 m) from 2009 in the lower CowlitzRiver. Gray dashed vertical lines demarcate peaks in the 2009 Chinook Salmonredd distribution relative to spatial patterns of bifurcated channel length, sinu-osity, and channel gradient.

TABLE 3. Estimated coefficients from a GLM of segment-scale stream habi-tat variables explaining the distribution of Chinook Salmon redds in the CowlitzRiver, Washington. Abbreviations are as follows: RKM = river kilometer,CB = channel bifurcation, and SIN = sinuosity.

ParameterVariable estimate SE Z-value P-value

Intercept 4.00 2.01RKM 7.06 × 10−5 8.25 × 10−6 8.56 <0.01CB 1.86 × 10−3 8.29 × 10−4 2.24 0.03SIN −5.16 1.88 −2.75 0.01

generally understood that salmon spawn in clusters in the sameareas each year; however, it is important to statistically exam-ine these assumptions at different spatial and temporal scales toconfirm that they are valid.

By comparing redd distribution between the first versus thelast weeks of the spawning period, we were able to deter-mine whether spring and fall Chinook Salmon were consistentlyspawning in the same areas in the Cowlitz River. The first reddsof the year are built by spring Chinook Salmon, whereas thelater redds are fall Chinook Salmon redds (Henning 2008). Theredds from the first and last weeks of the survey period occurredin the same areas in both 2008 and 2009 (Figure 4). This isunexpected because typically in the Columbia River basin thereis spatial separation between spring and fall Chinook Salmonspawning grounds, with spring Chinook Salmon spawning insmaller tributaries and upper reaches of principal tributaries andfall Chinook Salmon spawning in the lower river tributaries andmain stem (Fulton 1968). For example, fall Chinook Salmon onthe Yakima River only spawn in the lower river, typically belowGranger, Washington, while spring Chinook Salmon spawn atleast 100 km farther upstream, between the Easton Dam and El-lensburg, Washington, and in connecting tributaries (Major andMighell 1969). On the Cowlitz River, spring Chinook Salmonspawn in the same areas as fall Chinook Salmon. Spring Chi-nook Salmon cannot migrate past the Barrier Dam and haveto be transported in order to spawn in historic spawning areas(Fulton 1968). Therefore, some spring Chinook Salmon spawnin the lower portions of the Cowlitz River, which are preferredby fall Chinook Salmon. Our analysis used 500-m sections;therefore, we cannot conclude that redd superimposition is oc-curring in these highly used areas. However, the inability ofspring Chinook Salmon to fully access their historic spawningareas suggests that migration barriers could cause competitionbetween fall and spring Chinook Salmon for quality spawninghabitat.

River kilometer was the most significant predictor of thespawning locations for the 500-m model predicting redd dis-tribution in the Cowlitz River. The linear pattern of increasingspawning density up to the Barrier Dam is partially explained bythe location of the Cowlitz River Salmon Hatchery adjacent tothe Barrier Dam. This pattern also could be caused by the barrier

Dow

nloa

ded

by [

Uni

vers

ity o

f W

ashi

ngto

n L

ibra

ries

] at

09:

10 2

9 A

pril

2013

516 KLETT ET AL.

FIGURE 6. Longitudinal spatial patterns of observed redds (number/500 m)from 2009 and GLM fitted values. Black circles indicate locations where peakswere observed in both the observed redds and those predicted by the model.

to the migration of salmon to their historic spawning grounds.Historically, all spring Chinook Salmon spawned above the Bar-rier Dam. There were also high-quality spawning grounds forfall Chinook Salmon directly above the Barrier Dam and fartherupstream in a 13-km section that was inundated by MayfieldReservoir (Fulton 1968). On the Yakima River, there is a similarlinear trend in redd distribution up to the Easton Dam, whichcuts off access to historic spring Chinook Salmon spawning ar-eas (Dittman et al. 2010). There is fish passage at Easton Dam,but it is only operational during certain flows and most fish nowdo not pass the dam. Martin et al. (2004) also observed largenumbers of Chinook Salmon returning to spawn directly belowa fish barrier on the Green River that prevented migration tohistorical spawning areas.

Another factor that may contribute to the strong associationof redd counts and river kilometer is that habitat quality in-creases on the Cowlitz River from rkm 30–82. In the first 30 kmof the Cowlitz River, no redds were observed in 2008 and 2009;this area did not appear to have high-quality spawning areas,presumably due to high concentrations of silt and residential de-velopment in the lower Cowlitz River. Upstream of these highlydeveloped residential areas, there is substantially less channelmodification (e.g., levees and bank armoring), and the overallquality of spawning habitat is greater. Furthermore, sedimentinputs in downstream reaches of the Cowlitz River have beenhigh historically. For example, the Cowlitz River received largeamounts of fine sediment from the Toutle River, which enters theCowlitz River at rkm 25, due to the eruption of Mt. St. Helens in1980. The volcanic eruption moved large amounts of sedimentinto both rivers, and recovery to preeruption sediment yields inthe Cowlitz basin is not predicted for another 10 years (Majoret al. 2000).

Channel bifurcation was positively associated with the oc-currence of spawning in reaches in the Cowlitz River. Mul-tiple channels are associated with intragravel flow on a largescale that is critical for incubating salmon (Brunke and Gonser1997; Geist 2000). Increased intragravel flow often occurs atthe upstream and downstream ends of channel bars and is-lands, where the river is slower and shallower (Brunke andGonser 1997; Dauble and Geist 2000). Multiple channels alsocan be associated with an increased area of the riverbed avail-

able for spawning (Dauble and Geist 2000). High densities ofsalmon redds in the Cowlitz River were more likely related toislands and associated increased hyporheic flow as opposed tothe increased area of the riverbed available for spawning. Forexample, a single split channel on the Cowlitz River was desig-nated by fisheries managers in WDFW as an entire section foraerial surveys because of the high density of redds in the area.Dauble and Geist (2000) found that Chinook Salmon spawn-ing in the Hanford Reach, Washington, were concentrated inbraided river sections and areas that had complex channel for-mations. Coulombe-Pontbriand and Lapointe (2004) also foundthat large numbers of redds of Atlantic Salmon Salmo salar oc-curred near the upstream margins of channel islands in the Pe-tite Cascapedia and Bonaventure rivers on the Gaspe Peninsula,Quebec.

Sinuosity was negatively associated with spawning locationsin the Cowlitz River. The entire Cowlitz River had a very lowsinuosity (1.6) and was often confined by levees and develop-ment, particularly on the lower river. Previous studies on theColumbia and Snake rivers have shown that areas with highersinuosity have higher Chinook Salmon redd density at a 1.6-kmscale (Dauble and Geist 2000). Additionally, Fukushima (2001)found that Sakhalin Taimen Hucho perryi preferred to spawn insites located below highly sinuous reaches. However, sinuositywas only important as a predictor variable at a small scale(50 m). In higher-sinuosity streams that are not bounded bychannel constraints, sinuosity has been associated with channelmorphology and pool–riffle complexes that are associatedwith higher redd densities (McKean et al. 2008). However, onthe Cowlitz River, sinuosity was not correlated with channelbifurcation or depth discontinuities from pool–riffle complexes.This is a potential explanation for the negative associationbetween sinuosity and redd density in our study.

Tributary junctions, depth discontinuities, and channel gradi-ent were not significant predictors of redd density on the CowlitzRiver. These variables were positively associated with redd loca-tions in previous studies at a reach and larger spatial scales. Trib-utaries can affect sediment inputs and create optimal locationsfor spawning (Rice et al. 2008). On the North Fork StillaguamishRiver, Rice et al. (2008) found that patterns of Chinook Salmonredd distribution were not associated with tributaries at 1.1-km scale. However, tributaries were associated with spawningpatterns at smaller spatial scales. Discontinuities in depth canalso indicate areas of increased hyporheic–surface water ex-change. Previous studies found redds in tailouts of pools andat the boundaries between pools and riffles (Bjornn and Reiser1991), where a change in depth was associated with increasedhyporheic–surface water exchange. These patterns have beendescribed at multiple spatial scales (Baxter and Hauer 2000).We investigated these associations at a 500-m scale and foundthat depth discontinuities were not associated with redd loca-tions in the Cowlitz River. Although we observed no associationwith channel gradient, lower-gradient areas typically have well-developed floodplains and gravel bars, and these areas have been

Dow

nloa

ded

by [

Uni

vers

ity o

f W

ashi

ngto

n L

ibra

ries

] at

09:

10 2

9 A

pril

2013

SPATIAL CONSISTENCY OF CHINOOK SALMON REDD DISTRIBUTION 517

shown to have high densities of Chinook Salmon redds in theColumbia River basin (Fulton 1968; Dauble and Geist 2000;Dauble et al. 2003).

The GLM regression approach that we used effectively pre-dicted areas of peak redd density (74% accuracy; Figure 6),but it underpredicted the number of redds at these peaks. Themodel also predicted that redds would be present in some lo-cations where no redds were observed. The underprediction bythe model of the number of redds at the peaks may be attributedto other factors and the scale(s) at which these factors weremeasured (Torgersen et al. 2012). For example, at the basinscale, the number of returning fish depends on ocean condi-tions, which were not incorporated into the model. Conversely,at small spatial scales, information such as sediment grain size,water chemistry, and velocity measurements, may be neededto obtain more accurate predictions. We found that segment-scale features accurately predicted where redds occurred in theCowlitz River; however, to fully understand the observed pat-terns in redd distribution, a more detailed assessment at reachand basin scales is needed (Lapointe 2012). Further work onthe Cowlitz River could involve spatially continuous surveysof sediment size and other reach-scale variables that could beincluded in predictive models (Brenkman et al. 2012).

Management ImplicationsCollecting location data with a GPS and digital audio

recorder during aerial surveys of redd distribution is cost-effective and allows much greater flexibility for analysis of redddistribution across spatial scales. Many aerial redd surveys stillrely on paper maps and require redd counts to be tallied inflight between landmarks over long sections (0.5–28.0 km) ofriver. With such low-precision techniques, important informa-tion about redd distribution patterns is lost, and analyses canonly be conducted at coarse spatial scales. We found that rel-atively inexpensive GPS and digital audio technology can beused effectively in an aircraft and requires minimal postpro-cessing to analyze in a GIS. Furthermore, this approach can beused by fisheries managers who may have minimal training orexperience with GPS or GIS.

Maps of salmon redd distribution and aquatic habitat are im-portant tools in riverine fisheries management because salmonmay spawn in the same sections of the river year after year,and this information can be used to prioritize habitat conser-vation and restoration efforts. In many rivers with anadromoussalmonids in North America, river reaches with multiple chan-nels provide more suitable areas for conservation and restora-tion. Knowledge of where salmon redds occur in relation tochannel features that can be mapped using the publically avail-able digital maps and data sources described in this study canhelp river managers plan reach- and segment-scale restorationefforts with greater confidence and potentially use engineeredlogjams to increase channel complexity (Roni et al. 2002).Segment-scale information of the type described by this studycan be used in fisheries management to promote long-term pop-

ulation viability of Chinook Salmon in the Cowlitz River and inother similar rivers.

ACKNOWLEDGMENTSWe thank Ryan Klett, Matt Groce, and Jeremy Cram for

assistance with field work, Ethan Welty for GIS expertise andprogramming in R statistical software (linbin package), andMark LaRiviere, from Tacoma Power, for providing logisticalsupport and background information on the Cowlitz River. Wewould also like to thank Gino Lucchetti, Susan Bolton, andtwo anonymous reviewers for their constructive comments andrecommendations. The School of Environmental and Forest Sci-ences at the University of Washington and the U.S. GeologicalSurvey, Forest and Rangeland Ecosystem Science Center pro-vided partial funding for this work. Any use of trade, product, orfirm names is for descriptive purposes only and does not implyendorsement by the U.S. Government.

REFERENCESBaxter, C. V., and F. R. Hauer. 2000. Geomorphology, hyporheic exchange,

and selection of spawning habitat by Bull Trout (Salvelinus confluentus).Canadian Journal of Fisheries and Aquatic Sciences 57:1470–1481.

Beechie, T., H. Moir, and G. Pess. 2008. Hierarchical physical controls onsalmonid spawning location and timing. Pages 83–101 in D. A. Sear andP. DeVries, editors. Salmonid spawning habitat in rivers: physical controls,biological responses, and approaches to remediation. American FisheriesSociety, Symposium 65, Bethesda, Maryland.

Beyer, H. L. 2004. Hawth’s analysis tools for ArcGIS. Spatial Ecology, Toronto.Available: www.spatialecology.com. (April 2010).

Bjornn, T. C., and D. W. Reiser. 1991. Habitat requirements of salmonids instreams. Pages 83–138 in W. R. Meehan, editor. Influences of forest andrangeland management on salmonid fishes and their habitats. American Fish-eries Society, Special Publication 19, Bethesda, Maryland.

Borradaile, G. 2003. Statistics of earth science data: their distribution in time,space and orientation. Springer-Verlag, Berlin.

Brenkman, S. J., J. J. Duda, C. E. Torgersen, E. Welty, G. R. Press, R. Peters,and M. L. McHenry. 2012. A riverscape perspective of Pacific salmonidsand aquatic habitats prior to large-scale dam removal in the Elwha River,Washington, USA. Fisheries Management and Ecology 19:36–53.

Brunke, M., and T. Gonser. 1997. The ecological significance of exchangeprocesses between rivers and groundwater. Freshwater Biology 37:1–33.

Buckland, S. T., K. P. Burnham, and N. H. Augustin. 1997. Model selection: anintegral part of inference. Biometrics 53:603–618.

Burnham, K. P., and D. R. Anderson. 2002. Model selection and multimodelinference: a practical information-theoretic approach, 2nd edition. Springer-Verlag, New York.

Coulombe-Pontbriand, M., and M. Lapointe. 2004. Geomorphic controls, rifflesubstrate quality, and spawning site selection in two semi-alluvial salmonrivers in the Gaspe Peninsula, Canada. River Research and Applications20:577–590.

Dauble, D. D., and D. R. Geist. 2000. Comparison of mainstem spawninghabitats for two populations of fall Chinook Salmon in the Columbia Riverbasin. Regulated Rivers: Research and Management 16:345–361.

Dauble, D. D., T. P. Hanrahan, D. R. Geist, and M. J. Parsley. 2003. Impacts of theColumbia River hydroelectric system on main-stem habitats of fall ChinookSalmon. North American Journal of Fisheries Management 23:641–659.

Dittman, A. H., D. May, D. A. Larsen, M. L. Moser, M. Johnston, and D.Fast. 2010. Homing and spawning site selection by supplemented hatchery-and natural-origin Yakima River spring Chinook Salmon. Transactions of theAmerican Fisheries Society 139:1014–1028.

Dow

nloa

ded

by [

Uni

vers

ity o

f W

ashi

ngto

n L

ibra

ries

] at

09:

10 2

9 A

pril

2013

518 KLETT ET AL.

Fausch, K. D., C. E. Torgersen, C. V. Baxter, and H. W. Li. 2002. Landscapesto riverscapes: bridging the gap between research and conservation of streamfishes. BioScience 52:483–498.

Frissell, C. A., W. J. Liss, C. E. Warren, and M. D. Hurley. 1986. A hierarchicalframework for stream habitat classification: viewing streams in a watershedcontext. Environmental Management 10:199–214.

Fukushima, M. 2001. Salmonid habitat–geomorphology relationships in low-gradient streams. Ecology 82:1238–1246.

Fulton, L. A. 1968. Spawning areas and abundance of Chinook Salmon (On-corhynchus tshawytscha) in the Columbia River basin: past and present. U.S.Fish and Wildlife Service Special Scientific Report Fisheries 571.

Geist, D. R., and D. D. Dauble. 1998. Redd site selection and spawning habitatuse by fall Chinook Salmon: the importance of geomorphic features in largerivers. Environmental Management 22:655–669.

Geist, D. R., T. P. Hanrahan, E. V. Arntzen, G. A. McMichael, C. J. Murray,and Y. J. Chien. 2002. Physicochemical characteristics of the hyporheic zoneaffect redd site selection by Chum Salmon and fall Chinook Salmon in theColumbia River. North American Journal of Fisheries Management 22:1077–1085.

Geist, D. R., J. Jones, C. J. Murray, and D. D. Dauble. 2000. Suitability crite-ria analyzed at the spatial scale of redd clusters improved estimates of fallChinook Salmon (Oncorhynchus tshawytscha) spawning habitat use in theHanford Reach, Columbia River. Canadian Journal of Fisheries and AquaticSciences 57:1636–1646.

Gent, D. H., W. F. Mahaffee, and W. W. Turechek. 2006. Spatial heterogene-ity of the incidence of powdery mildew on hop cones. Plant Disease 90:1433–1440.

Good, T. P., R. S. Waples, and P. Adams, editors. 2005. Updated status offederally listed ESUs of West Coast salmon and steelhead. NOAA TechnicalMemorandum NMFS-NWFSC-66.

Henning, J. A. 2008. Cowlitz River evaluation program annual report for 2007.Washington Department of Fish and Wildlife, Olympia.

Isaak, D. J., and R. F. Thurow. 2006. Network-scale spatial and temporal vari-ation in Chinook Salmon (Oncorhynchus tshawytscha) redd distributions:patterns inferred from spatially continuous replicate surveys. Canadian Jour-nal of Fisheries and Aquatic Sciences 63:285–296.

Isaak, D. J., R. F. Thurow, B. E. Rieman, and J. B. Dunham. 2007. ChinookSalmon use of spawning patches: relative roles of habitat quality, size, andconnectivity. Ecological Applications 17:352–364.

Kutner, M. H., C. J. Nachtsheim, and J. Neter. 2004. Applied linear regressionmodels, 4th edition. McGraw-Hill, New York.

Lapointe, M. 2012. River geomorphology and salmonid habitat: some examplesillustrating their complex association, from redd to riverscape scales. Pages191–215 in M. Church, P. M. Biron, and A. G. Roy, editors. Gravel-bed rivers:processes, tools, and environments. Wiley, Chichester, UK.

Lawless, J. F. 1987. Negative binomial and mixed Poisson regression. CanadianJournal of Statistics 15:209–225.

LCFRB (Lower Columbia Fish Recovery Board). 2004. Lower Columbiasalmon recovery and fish and wildlife subbasin plan, volume IIE. CowlitzRiver subbasin plan. LCFRB, Longview, Washington.

Major, J. J., T. C. Pierson, R. L. Dinehart, and J. E. Costa. 2000. Sediment yieldfollowing severe volcanic disturbance: a two-decade perspective from MountSt. Helens. Geology 28:819–822.

Major, R. L., and J. L. Mighell. 1969. Egg-to-migrant survival of spring ChinookSalmon (Oncorhynchus tshawytscha) in the Yakima River, Washington. U.S.National Marine Fisheries Service Fishery Bulletin 67:347–359.

Malcolm, I. A., A. F. Youngson, and C. Soulsby. 2003. Survival of salmonideggs in a degraded gravel-bed stream: effects of groundwater–surface waterinteractions. River Research and Applications 19:303–316.

Mansfield, E. R., and B. P. Helms. 1982. Detecting multicollinearity. AmericanStatistician 36:158–160.

Martin, D., L. Benda, and D. Shreffler. 2004. Core areas: a framework foridentifying critical habitat for salmon. Report to King County Depart-

ment of Natural Resources and Parks, Water and Land Resources Division,Seattle.

McKean, J. A., D. J. Isaak, and C. W. Wright. 2008. Geomorphic controls onsalmon nesting patterns described by a new, narrow-beam terrestrial–aquaticlidar. Frontiers in Ecology and the Environment 6:125–130.

Moir, H. J., C. N. Gibbins, C. Soulsby, and J. Webb. 2004. Linking chan-nel geomorphic characteristics to spatial patterns of spawning activity anddischarge use by Atlantic Salmon (Salmo salar L.). Geomorphology 60:21–35.

Neumann, D. W., B. Rajagopalan, and E. A. Zagona. 2003. Regression model fordaily maximum stream temperature. Journal of Environmental Engineering129:667–674.

Neville, H. M., D. J. Isaak, J. B. Dunham, R. F. Thurow, and B. E. Rieman.2006. Fine-scale natal homing and localized movement as shaped by sex andspawning habitat in Chinook Salmon: insights from spatial autocorrelationanalysis of individual genotypes. Molecular Ecology 15:4589–4602.

R Development Core Team. 2009. R: a language and environment for statisti-cal computing. R Foundation for Statistical Computing, Vienna. Available:www.R-project.org. (April 2010).

Rayburn, A. P., and L. A. Schulte. 2009. Landscape change in an agriculturalwatershed in the U.S. Midwest. Landscape and Urban Planning 93:132–141.

Rice, S. P., P. Kiffney, C. Greene, and G. R. Pess. 2008. The ecological im-portance of tributaries and confluences. Pages 209–242 in S. Rice, A. Roy,and B. Rhoads, editors. River confluences, tributaries and the fluvial network.Wiley, Chichester, UK.

Roni, P., T. J. Beechie, R. E. Bilby, F. E. Leonetti, M. M. Pollock, and G. R. Pess.2002. A review of stream restoration techniques and a hierarchical strategyfor prioritizing restoration in Pacific Northwest watersheds. North AmericanJournal of Fisheries Management 22:1–20.

Sear, D. A., and P. DeVries, editors. 2008. Salmonid spawning habitat in rivers:physical controls, biological responses, and approaches to remediation. Amer-ican Fisheries Society, Symposium 65, Bethesda, Maryland.

Tipping, J. M., and C. A. Busack. 2004. The effect of hatchery spawning proto-cols on Coho Salmon return timing in the Cowlitz River, Washington. NorthAmerican Journal of Aquaculture 66:293–298.

Torgersen, C. E., C. V. Baxter, J. L. Ebersole, and R. E. Gresswell. 2012.Incorporating spatial context into the analysis of salmonid–habitat relations.Pages 216–224 in M. Church, P. M. Biron, and A. G. Roy, editors. Gravel-bedrivers: processes, tools, environments. Wiley, Chichester, UK.

Torgersen, C. E., and D. A. Close. 2004. Influence of habitat heterogeneity onthe distribution of larval Pacific Lamprey (Lampetra tridentata) at two spatialscales. Freshwater Biology 49:614–630.

Turechek, W. W., and L. V. Madden. 1999. Spatial pattern analysis of straw-berry leaf blight in perennial production systems. Phytopathology 89:421–433.

Vaccaro, J. J., and K. J. Maloy. 2006. A thermal profile method to identifypotential ground-water discharge areas and preferred salmonid habitats forlong river reaches. U.S. Geological Survey, Scientific Investigations Report2006-5136, Reston, Virginia.

Ver Hoef, J. M., and P. L. Boveng. 2007. Quasi-Poisson vs. negative bino-mial regression: how should we model overdispersed count data? Ecology88:2766–2772.

Visser, R., D. D. Dauble, and D. R. Geist. 2002. Use of aerial pho-tography to monitor fall Chinook Salmon spawning in the ColumbiaRiver. Transactions of the American Fisheries Society 131:1173–1179.

Wright, S. P. 1992. Adjusted P-values for simultaneous inference. Biometrics48:1005–1013.

Yan, B., M. D. Tomer, and D. E. James. 2010. Historical channel movement andsediment accretion along the South Fork of the Iowa River. Journal of Soiland Water Conservation 65:1–8.

Zar, J. H. 1999. Biostatistical analysis, 4th edition. Prentice-Hall, Upper SaddleRiver, New Jersey.

Dow

nloa

ded

by [

Uni

vers

ity o

f W

ashi

ngto

n L

ibra

ries

] at

09:

10 2

9 A

pril

2013