Comparing Riparian and Catchment Influences on Stream Habitat in a Forested, Montane Landscape

Embed Size (px)

Citation preview

  • 8/3/2019 Comparing Riparian and Catchment Influences on Stream Habitat in a Forested, Montane Landscape

    1/18

    Ar ne ricd F isheries SocietySymposium48:175-197,2006Q 2006 by the American Fisheries Society

    Comparing Riparian and Catchment Influences onStream Habitat in a Forested, Montane Landscape

    Kelly M. Burnett*, Gordon H. ReevesUSDA Forest Service, Pacific Northwest Research Station, 3200 SW Jefferson Way,

    Corvallis, Oregon 9733 , USASharon E. Clarke

    Roaring Fork Conservancy, Post Office Box 3349, Basalt, Colorado 8 162 , USAKelly R. Christiansen

    USDA Forest Service, Pacific Northwest Research Station, 3200 SW Jefferson Way,Corvollis, Oregon 9733 1, USA

    Abstract.-Multiscale analysis of relationships with landscape characteristicscan help iden-tify areas and physical processes that affect stream habitats, and thus suggest where and howland management is likely to influence these habitats. Such analysis is rare for mountainousareas where forestry is the primary land use. Consequently, we examined relationships in aforested, montane basin between stream habitat features and landscape characteristics hatwere summarized at five spatial scales (three riparian and two catchment scales). Spatialscales varied in the area encompassed upstream and upslope of surveyed stream segmentsand, presumably, in physical processes. For many landscape characteristics, riparian spatialscales, approximated by fuced-width buffers, could be differentiated from catchment spatialscales using forest cover from 30-m satellite imagery and 30-m digital elevation data. Inregression with landscape characteristics, more variation in the mean maximum depth andvolume of pools was explained by catchment area than by any other landscape characteristicsummarized at any spatial scale. In contrast, at each spatial scale except the catchment,varia-tion in the mean density of large wood in pools was positively related to percent area inolder forests and negatively related to percent area in sedimentary rock types. The regres-sion model containing these two variables had the greatest explanatory power at an inter-mediate spatial scale. Finer spatial scales may have omitted important source areas andprocesses for wood delivery, but coarser spatial scales likely incorporated source areas andprocesses less tightly coupled to large wood dynamics in surveyed stream segments. Ourfindings indicate that multiscale assessments can identify areas and suggest processes mostclosely linked to stream habitat and, thus, can aid in designing land management to protectand restore stream ecosystems in forested landscapes.

    INTRODUCTION al. 1986;Naiman e t al. 2000).A catchment con-tains a mosaic of patches an d interconnected

    The condition of a stream ecosystem is largely networks (Pickett and White 1985;Swanson eta function of landscape characteristics in the al. 1997;Jones et al. 2000) that control the rout-surrou nding catchment (Hynes 1975;Frissell et ing of energy and materials to str eams and that

    ultimately control stream ecosystems (Swanson*Corresponding author: [email protected] et al. 1998; Jones et al. 2000;Puth and Wilson

  • 8/3/2019 Comparing Riparian and Catchment Influences on Stream Habitat in a Forested, Montane Landscape

    2/18

    t et al .

    2001). These patches and networks have char-acteristics such as size, shape, type (e.g., forestor paved roads) and location (e.g., ridge top orriparian). Direct effects on streams of landscapecharacteristics in the local riparian area are wellestablished (Osborne and Koviac 1993; Naimanet al. 2000; National Research Council 2002).However, relationships between streams andlandscape characteristics are less well under-stood and agreed upon when landscape char-acteristics are considered upstream along ariparian network (Weller et al. 1998; Jones etal. 1999) or upslope throughout a catchment(Jones and Grant 1996, 2001; Thomas andMegahan 1998; Gergel2005).

    Influences of riparian and catchment char-acteristics on stream ecosystems have been ex-amined predominantly in agricultural andurbanized areas. For example, the abundanceof adult coho salmon Oncorhynchus kisutch inthe Snohomish River, Washington was signifi-cantly related to land cover (expressed as per-cent urban, agriculture, or forest) summarizedfor the loca1,riparian area and for the entirecatchment (Pess et al. 2002). Riparian andcatchment land cover may explain approxi-mately equal proportions of physical (Richardset al. 1996) and biological (Van Sickle et al.2004) variation in agricultural or urbanizedstream systems.

    Conclusions often differ, however, regardingthe relative influence of riparian and catchmentland cover on streams in agricultural and urbanenvironments. Certain in-channel responseswere best explained by land-cover characteris-tics summarized for the local riparian area (e.g.,catch per 100 m of cool- and coldwater fish[Wang et al. 2003al). Others were best explainedby land-cover characteristics summarized forthe entire catchment (e-g., total fish and macro-invertebrate species richness [Harding et al.19981). For water quality parameters, land-cover characteristics explained more variationwhen summarized for the riparian network insome studies (Osborne and Wiley 1988) but for

    the entire catchment in others (Omernik et al.1981), or explained a variable degree of varia-tion depending on data resolution, season orlocation of sampling, and modeling approach(Hunsaker and Levine 1985; Johnson et al.1997). Even when the same response variable(index of biological integrity) was examined inthe same river basin but at different spatial ex-tents, judgments differed about the influencesof riparian and catchment land cover (Roth etal. 1996; Lammert and Allan 1999). Given suchvariability, extrapolating understanding frommultiscale studies in more developed land-scapes to stream systems in forested landscapesmay be ill advised.

    Riparian and catchment land cover have sel-dom been compared for relationships o streamsin mountainous areas where forest uses domi-nate. We are aware of few studies examining ri-parian and catchment influences on streams thatdrain forested regions or areas with minimalhuman development (Hawkins et al. 2000; Wanget al. 2003b; Weigel et al. 2003; Sandin andJohnson 2004). Understanding arising from suchstudies may contribute to conservation of Pacificsalmon and trout, which are widely distributedin North America. Abundances of these fish andconditions of their freshwater habitat have beenrelated to land-cover characteristics at differentspatial scales, including the local riparian area(Bilby and Ward 1991), the riparian network(Botkin et al. 1995), and the catchment (e.g.,Reeves et al. 1993; Dose and Roper 1994; Dun-ham and Rieman 1999; Thompson and Lee2002). Although such studies offered valuableinsights, none directly examined relationshipsbetween salmon, or their habitats, and land-covercharacteristics summarized at more than onespatial scale.Multiscale assessments may identify riparianand upslope areas that help create and maintainsalmon habitats in forested, montane landscapes.Pools and large wood are essential componentsof salmon habitat in such landscapes, providingliving space and cover from predators (Bilby and

  • 8/3/2019 Comparing Riparian and Catchment Influences on Stream Habitat in a Forested, Montane Landscape

    3/18

    Comparing Riparian and Catchment Influences on Stream Habitat 177

    Bisson 1998; McIntosh et al. 2000). Pools are ar-eas of local scour caused by fluvial entrainmentand transport of bed substrates that persist untilsediment inputs to, and outputs from, a poolequilibrate. The creation and morphology(depth, volume, and surface area) of pools aredriven by sediment supply, hydraulic discharge,and presence of flow obstructions (e.g., woodand boulders) (Buffington et al. 2002).AU threefactors are affected by channel-adjacent and hill-slope processes. For example, the amount of sedi-ment and wood supplied to pools can increasewith increases in the frequency of channel-adja-cent processes, such as bank erosion, or of hill-slope processes, such as landsliding. The relativeimportance of channel-adjacent and hill-slopeprocesses can vary with channel type (Mont-gomery and Buffington 1998; Buffington et al.2002) and land cover (e.g., Bilby and Bisson 1998;Ziemer and Lisle 1998; Montgomery et al. 2000),and thus, the potential for land management toimpact pools and large wood varies across thelandscape. Consequently, studying relationshipsat multiple spatial scales can help identify whichprocesses are, and where land management is,likely to alter salmon habitat.

    Our goal was to understand relationships be-tween salmon habitat and landscape character-istics, summarized at multiple spatial scales, in amontane basin where forestry is the dominantland use. Targeted habitat features were the meanmaximum depth of pools, mean volume ofpools, and mean density of large wood in pools.Three riparian scales (segment, subnetwork, andnetwork) and two catchment scales (subcatch-ment and catchment) were considered for eachstream segment where targeted habitat featureswere evaluated (Figure 1).Spatial scales differedin the area included upslope and upstream ofsurveyed stream segments, and presumably invegetative, geomorphic, and flwial processes thatmay affect targeted habitat features. Channel-adjacent processes (e.g., tree mortality in ripar-ian stands and streamside landsliding) andin-channel process (e.g., debris flows and fluvial

    transport) were assumed to dominate at the ri-parian scales. Potential for nonchannelized hillslope processes (e.g., surface erosion andlandsliding) were added at the two catchmentscales. Specific study objectives were to (1) ex-amine differences among spatial scales for land-scape characteristics described with relativelycoarse-resolution data, and (2) compare the pro-portion of variation in stream habitat featuresexplained by landscape characteristics summa-rized within and among different spatial scales.

    STUDY AREAThe study was conducted in tributaries of theupper Elk River, located in southwestern Oregon,USA (Figure 2). The main stem of the Elk Riverflows primarily east to west, entering the PacificOcean just south of Cape Blanco (42"S'N lati-tude and 124O3'W longitude). The Elk River ba-sin (236 km2) is in the Klamath Mountainsphysiographic province (Franklin and Dyrness1988)and is similar to other Klamath Mountaincoastal basins in climate, landform, vegetation,land use, and salmonid assemblage.

    The climate is temperate maritime with re-stricted diurnal and seasonal temperature fluc-tuations (USFS 1998). Ninety percent of theannual precipitation occurs between Septemberand May, principally as rainfall. Peak streamflows are flashy following 3-5-d winter rain-storms, and base flows occur between July andOctober. Elevation ranges from sea level to ap-proximately 1,200 m at the easternmost drain-age divide. Recent tectonic uplift produced ahighly dissected terrain that is underlain by thecomplex geologic formations of the KlamathMountains. Stream densities in these rock typesrange from 3 to 6 kmlkm2(FEMAT 1993).

    Much of the study area is in mixed coniferand broadleaf forests that include tree species ofDouglas fir Pseudotsuga menziesii western hem-lock Tsuga heterophylla, Port Orford cedarChamaecyparis lawsoniana, tanoak Lithocarpus

  • 8/3/2019 Comparing Riparian and Catchment Influences on Stream Habitat in a Forested, Montane Landscape

    4/18

    184

    RockTypes

    Burnett et al.

    Slope Class

    S SN SC N C S SN SC N CSpatial Scale Spatial ScaleFigure 3. Distribution of landscape characteristics among analytical units at each of the five spatial scales intributaries of the Elk River, Oregon . Spatial scales were the segment (S), subnetwork (SN), subcatchment (SC),network (N), and catchment (C). Boxes designate the 25 th and 75th percentiles, the sol id line indicates themedian and the dotted line the mean, whiskers denote the nearest data point within 1.5 times the interquortilerange, and 5th and 95 th percentiles are shown by disconnected points. For a given landscape characteristic,two scales with the same letter label above their box plots have a significant pair-wise difference betweenmedians when the overall type I error rate is controlled at ar = 0.05.

  • 8/3/2019 Comparing Riparian and Catchment Influences on Stream Habitat in a Forested, Montane Landscape

    5/18

    Comparing Riparian and Catchment Influences on Stream Habitat 185ForestCover Forest C w e r

    100 - (&open a d emi-d'osed - 100 - (k)L . i ammr80 - 80

    8 60 - ab cd ef ace bdf4$40 - $"8 0$ $Q&:$ 6+ - 0 -0 - , 0 -

    100 (h) Broadleaf ' 100 - (1) Very large diameter80 - 80 -

    a 6 0 - a4 ab bc c - f 0 - a bc abd de ceg 4 0 - - $40-20 - 2 0 - Q & Q G Q :

    4% -- 0 -100- ' (i) Small diameter- 100 -80 - ~ O - $ Q $ @ Q ~

    4- 60-

    14S 4 0 $ 4 0-20 - abc def ad be cf 20 - (in)Medium to very&-&A0 -- - large diameter0 -100 - (j) Medium diameter - (n)Road Density5 -80 - 3 8 4 - a b c abc -g 60 - ab cd ace ef bdf -2 "2 3 -

    $40 - - E "20 -

    0 -0 -S SN SC N C S SN SC N CSpatial Scale Spatial Scale

    Figure 3. continued

  • 8/3/2019 Comparing Riparian and Catchment Influences on Stream Habitat in a Forested, Montane Landscape

    6/18

    186 Burneti

    Because stream segments were not selectedwith a probability sampling design, we assessedregression residuals from each best among-scalemodel for nonrandom errors that might reflectspatial autocorrelation. For all possible pairs ofstream segments, stream distance and the abso-lute difference between regression residuals werecalculated. These two sets of values were re-gressed to determine the proportion of the varia-tion in the absolute difference between regressionresiduals explained by the stream distance be-tween stream segments.

    RESULTSLandscape Characterization

    Variance across stream segments differed signifi-cantly (df= 4,70; P50 .0 5) among spatial scalesfor all but four landscape characteristics, he per-cent area in (1) igneous intrusive rock types, (2)slopes 5 30%, (3) slopes > 60%, and (4) openand semiclosed canopy forest. The smallest vari-ance was observed at either the network or catch-ment scale for all landscape characteristics exceptthe percent area in forests of small diameter trees.. In one-way ANOVA, the blocking factor,stream segment, was significant (F(,,,,,; P r0.0001 for all landscape characteristics, andmedians differed significantly (F(,,,,;P 5 0.03)among spatial scales for 10 of 14 landscape char-acteristics (Figure 3). Pair-wise differences inmedians were not significantly (P> 0.05) differ-ent between the segment and subnetwork scalesfor any landscape characteristic. For most land-scape characteristics, pair-wise differences be-tween medians were significant ( P r 0.05)between a catchment scale (subcatchment orcatchment) and one or more of the riparian scales(segment, subnetwork, or network) (Figure 3). Toillustrate, for the percent area in slopes 5 30%(Figure 3D), the medians of the subcatchment(12.2%) and the catchment (11.9%) scales, al-though not significantly different from eachother, were significantly different from those ofthe segment (26.2%), subnetwork (2 1.3%), and

    network (23.1%) scales. Pair-wise differencesbetween the riparian scales were not significantfor this landscape characteristic.

    Regression of Stream Ha bitat Featureswith Landscape Characteristics

    Mean maximum depth and mean volume ofpools.-Both of these stream habitat featureswere positively related to catchment area (Table3). In one or more of the within-scale regres-sions, landscape characteristics explained a sig-nificant proportion of the variation in the meanmaximum depth of pools (RZ5 .29;df= 14;Pr 0.04; AAICr 7.3) and in the mean volume ofpools (R2I .48; 14< df 4 3 ; P1 .008; AAICr 20.2). However, no within-scale model metthe reporting criterion of AAIC 5 5 and eachexplained about half or less of the variation ex-plained by catchment area alone. Therefore, abest within-scale regression model was not iden-tified for either the mean maximum depth orvolume of pools.

    The best among-scale regression model forthe mean maximum depth of pools containedonly catchment area (Table3 ) .This was the onlyone of seven models for the mean maximumdepth of pools, which included or were withintwo AIC units of the smallest AIC value, to meetthe reporting criteria. In among-scale regressionfor the mean volume of pools, only one modelmet the reporting criteria (Table 3). However,the F-value of this model was substantially owerthan that of the model containing catchmentarea alone, which was therefore considered thebest among-scale regression model for the meanvolume of pools (Table 3). Stream distance be-tween each pair of stream segments explainedonly a small proportion of the variation in theabsolute differences between residuals from thebest among-scale regression model for the meanmaximum depth of pools (R2= 0.04; df = 104;P= 0.06) or for the mean volume of pools (R2=0.01; df = 104; P = 0.36).

    Mean density of large wood i n pools.-Al-though the mean density of large wood in pools

  • 8/3/2019 Comparing Riparian and Catchment Influences on Stream Habitat in a Forested, Montane Landscape

    7/18

    Comparing Riparian and Catchment Influences on Stream Habitat 187Table 3. Results from among-scale linear regression to explain variation in stream habitat features among15stream segments fortributaries of the Elk River, Oregon. Explanatory variables were catchment area alone andcatchment area plus landscape characteristics summarized at the segment(S), subnetwork (SN), subcatchment(SC), network (N), and catchment (C) scales. For among-scale regressions, the number of models that in-cluded, or were within two AIC units of the smallest AIC value, is given after the stream habitat feature.Reported models hod explanatory variables with signif icant slope estimates(a= 0.05) and little multicollineariiy(VIF < 4). Methods ore fully described in the text for identifying the set of reported models and best among-scale models indicated by '. Direction of relationships with explanatory variables is indicated by+/-. The AAICis relative to the model with catchment area alone for that stream habitat feature.Stream habitat featureExplanatory variable in mode l +/-PzItl VIF Model F P > F RZ (R . M I CMe an maximum depth of po obCatchment area 17.1 +0.001* 0.57Me an volume of poolsCatchment areaMe an density of large wood in poolsCatchment areaM ea n volume of pools (4)Catchment area +

  • 8/3/2019 Comparing Riparian and Catchment Influences on Stream Habitat in a Forested, Montane Landscape

    8/18

    188 . Burnett et al.Table 4 . Results from within-scale linear regression to explain variation in the mean density of large wood inpools among 15 stream segments in tributaries of the Elk River, Oregon. Explanatory variables are landscapecharacteristics summarized at five spatial scales. The number of models that included, or were within two AICunits of, the smallest AIC value is listed after the spatial scale. Reported models had explanatory variables withsignificant slope estimates (a = 0.05) and little multicollinearity (VIF < 4). Methods are fully described in thetext for identifying the set of reported models and the best model for each spatial scale, indicated by '. Directionof relationships with explanatory variables is indicated by +/-. The AAIC is relative to the model with cotch-ment area alone for the stream habitat feature.Spatial scaleExplanatory variable in model + / - P > I t l VIF Model F P > F R2(R2& ) AAICSegment (7)% sedimentary rock types -0.04 1 .OO 4.55 0.03' 0.34 0.0% medium-very large trees +0.05Subnehvork(2)% sedimentary rock types -0.01 1.03 7.47 0.008' 0.48 -3.7% medium-very large trees +0.01Subcatchment (1)% sedimentary rock types -0.004 1.07 10.48 0.002' 0.58 -6.8% medium-very large trees +0.003Network (4)% sedimentary rock types -0.04 1.09 5.94 0.02* 0.41 -1.9% medium-very large trees +0.01% sedimentary rock types -0.04 1.08 5.92 0.02 0.41 -1.9% open and semi-closed -0.01% sedimentary rock types -0.02 1.29 5.63 0.02 0.40 -1.5% road density (km/km2) -0.01Catchment(10)% open and semi-closed 7.31 -0.02* 0.36 -0.2

    I . , . . . . . . . . . . . . . . . . . . . . . . . .0.0 0.2 0.4 0.6 0.8 1.0Road Density (krn/km2)

    Figure 4. Results of linear regression behveen the per-cent area in forests of medium to very large diametertrees and road density at the network scale to explainvariation among stream segments for tributaries ofthe Elk River, Oregon. The linear regression line and95%mean confidence curves are shown (y = 85.7-16.7~; 2 = 0.69; P = 0.0001).

    in forests of medium to very large diameter trees(Table 3) . Stream distance between each pair ofstream segments explained little of the variationin the absolute difference between residuals fromthis among-scale regression (RZ= 0.01; df = 104;P= 0.26).

    DISCUSSION

    This study illustrated the value of multiscaleanalysis in relating stream habitat to riparian andcatchment characteristics in a landscape domi-nated by forest uses. Although ecologists ac-knowledge the importance of matching the scaleof inquiry to the questions posed (Wiens1989,2002), often the "right scale" is not known at theoutset of an investigation. Analysis at multiplescales may be necessary to elucidate linkagesamong stream organisms, heir habitats, and the

  • 8/3/2019 Comparing Riparian and Catchment Influences on Stream Habitat in a Forested, Montane Landscape

    9/18

    Comparing Riparian and Catchment Influences on Stream Habitat 189

    surrounding landscape. Indeed, we found thatrelationships between stream-habitat featuresand specific landscape characteristics differeddepending on spatial scale, enabling us to sug-gest processes responsible for observed variation.Fausch et al. (2002) emphasized that informa-tion most germane to land management deci-sions will likely stem from research in streamecology at intermediate temporal and spatialscales. Our finding that the mean density of largewood in pools of mid-order channels was bestexplainedwith landscape characteristics summa-rized at an intermediate spatial scale seems tobolster their case. We recognize that the scale atwhich stream habitat and landscape character-istics are most tightly coupled is undoubtedlyinfluenced by where examination is focused inthe drainage network. Had we targeted low-order, headwater channels instead of mid-orderchannels, stream habitat features may have beenmore directly affected by landscape conditionsthroughout these smaller catchments, increas-ing the likelihood of more variation being ex-plained at the catchment scale.

    ~if ferences mong Spatial Scalesin Landscape Characteristics

    The smallest variance among analytical units forlandscape characteristicswas generally observedat one of the coarser spatial scales (network orcatchment scale). Because the spatial resolutionof landscape coverages was typically finer thanthe area of analytical units, variance declined asthe area of analytical units increased. Our resultsagree with predictions from landscape ecologythat variability in landscape characteristics de-creases as grain or patch size increases (Formanand Godron 1986; Syrns and Jones 1999).Given that significant pair-wise differences inmedians for landscape characteristics were gen-erally between catchment and riparian scales,riparian areas were distinguished when delin-eated with a fixed width buffer and described by30-m digital elevation data and 30-m LandsatThematic Mapper Satellite imagery. This method

    detected expected geomorphic and ecologicaldifferences between riparian and upslope areasand so appears to be useful for characterizingriparian areas over broad spatial extents in for-ested systems. For example, our buffer charac-terization distinguished low-gradient valleybottoms in that segment, subnetwork, and net-work scales contained greater percentages of thelowest slope class than either of the catchmentscales. Futhermore, among-scale differences inpercentage area of broadleaf forest apparentlyreflect the greater likelihood of red alder occur-rence in the wetter and more frequently dis-turbed areas near streams (Pabst and Spies 1999).

    Previous studies characterizingriparian areasover a broad region generally used a fixed-widthbuffer rather than attempting to delineate theactual riparian area. Some of these studies foundsimilarities between riparian and upslope areasin landscape characteristics (e.g., Richards andHost 1994;Wangetal. 1997;Van Sickleet al.2004),but others did not (e.g.,Lammert and& 1999).Alternative, and potentially more accurate, meth-ods for delineating and characterizing riparianareas include mapping valley bottoms from finer-resolution digital topographic data (e.g.,Hemstrom et al. 2002), classifying digital imag-ery of higher spectral or spatial resolution, inter-preting standard aerial photography, and fieldmapping. The latter two methods are time andlabor intensive, however, and thus may limit thespatial extent reasonably addressed.

    Spatial Autocorrelation in Regressionof Stream Habitat Features

    with Landscape CharacteristicsResiduals from among-scale regression of thethree stream habitat features (mean maximumdepth of pools, mean volume of pools, and meandensity of large wood in pools) suggested littleevidence of spatial autocorrelation,and so we didnot attempt to remove or account for it in regres-sion models (Cliff and Ord 1973; Legendre 1993).However, relatively small sample size may havelimited ourabilityto detect spatial autocorrelation.

  • 8/3/2019 Comparing Riparian and Catchment Influences on Stream Habitat in a Forested, Montane Landscape

    10/18

    190 Burnett

    We are aware of no ideal technique to assess spa-tial dependence for stream networks when usingrelatively coarse-grained analytical units that dif-fer in size and spacing. Consequently,we adaptedan approach that assesses the degree of relation-ship for geographic distances between allpairs oflocations and corresponding differences betweenvalues of variables at those locations (Legendreand Fortin 1989). Geographic distances are usu-ally calculated with x-y coordinates (e.g., Hinchet al. 1994), but we chose stream distance to bet-ter reflect potential connectivity between streamsegments.

    Stream Habitat Featuresand Catchment Area

    Catchment area explained more among-streamsegment variation in the mean maximum depthof pools and the mean volume of pools thanother landscape characteristics at any of the fivespatial scales we examined. Land-cover variablesalso had less explanatory power for channelmorphology than catchment area in agriculturalsystems (Richards et al. 1996) and in a relativelyundegraded forest ecoregion (Wang et al. 2003bj.Catchment area is related to stream powerthrough its direct influence on stream discharge.Streams with higher discharge generally havegreater stream power, an index of the ability totransport materials, and tend to be deeper andwider than those with lower discharge (Gordonet al. 1992). Accordingly, the mean maximumdepth and volume of pools in Elk River tribu-taries increased as catchment area increased,paralleling results of Buffington et al. (2002).

    Although we determined that land cover ex-plained little of the variation in maximum depthor volume of pools, previous studies have dem-onstrated relationships between channel mor-phology and land uselcover. Based on correlativestudies, stream morphology is thought to be af-fected by land uses (Roth et al. 1996; Snyder etal. 2003; Wang et al. 2003a, 2003b), includingtimber harvest (Bilby and Ward 1991; Reeves etal. 1993; Dose and Roper 1994; Wood-Smith and

    Buffington 1996).Ourability to discern relation-ships between land cover and the mean maxi-mum depth of pools may have been hamperedbecause the maximum depths of the deepestpools were estimated and not measured. Giventhe apparent influence of catchment area, asample size larger than ours may be necessary toaccount for catchment area and thus to distin-guish relationships between timber harvest andpool morphology. Scaling by catchment area didimprove the ability to detect anthropogenic ef-fects on IBI metrics in Pacific Northwest coastalstreams (Hughes et al . 2004; Kaufmann andHughes 2006, this volume).

    The mean density of large wood in pools wasalso related to catchment area. The inverse rela-tionship between these two variables likely arisesfrom an increased ability of larger streams totransport wood. An inverse relationship wasfound with stream size in other forestry-domi-nated systems of the Pacific Northwestern UnitedStates (Bilby and Ward 1991; Montgomery et al.1995; Wing and Skaugset 2002) but not in Mid-western agricultural systems (Richards et al.1996; Johnson et al. 2006, this volume) or whendata from mixed-use and silvicultural systemswere combined (Wing and Skaugset 2002). Adirect relationship was found in midwestern ag-ricultural systems (Richards et al. 1996; Johnsonet al. 2006) and in mixed-use silvicultural sys-tems (Wing and Skaugset 2002). As the inten-sity and duration of human-caused disturbanceincreases, he presence of large wood in a streammay be determined more by sources of new re-cruitment than by transport capacity of thestream.

    Wood density and an indicator of stream dis-charge, bank-full stream width, were related inold-growth forests with few human impacts(Bilby and Ward 1989). Bilby and Ward (1989)noted the value of this relationship for determin--ing if wood density at another site was similar tothat expected for a "natural" stream of the samesize. Regression parameters or proportion ofvariation explained by such a relationship maybe useful benchmarks for assessing whether

  • 8/3/2019 Comparing Riparian and Catchment Influences on Stream Habitat in a Forested, Montane Landscape

    11/18

    Comparing Riparian and Catch1nent Influences on Stream Habitat 191

    wood dynamics at broader spatial scales are op-erating naturally (within the range of naturalvariability [Landreset al. 19991). Deviations fromsuch benchmarks may indicate that anthropo-genic disturbanceshave disrupted wood dynam-ics and constrained variability of in-channelwood across a landscape.

    Density of Large Wood in Poolsand Landscape Characteristics

    We found that landscape characteristics at eachspatial scale generally explainedasmuch or moreof the variation in the mean density of largewood in pools as catchment area. The mean den-sity of large wood in pools was negatively relatedto the percent area of sedimentary rock typessummarized at one or more spatial scales whenconsidered in combination with land cover. Theimportance of mass-wasting processes, such asdebris flows, to large wood delivery has been es-tablished in the Oregon Coast Range (Reeves etal. 2003) and the Olympic Peninsula, Washing-ton (Benda et al. 2003).Although possibly moreprevalent in other systems, debris flows occur inthe Elk River basin on all lithologies and deliverto higher order channels (Ryan and Grant 1991).However, less mass-wasting debris reachesstreams of the Elk River basin in sedimentaryrock types than in other rock types (McHugh1986), which is consistent with interpretationsof results from elsewhere in western Oregon(Scott 2002; Kaufmann and Hughes 2006), andmay help explain the negative relationship wefound between sedimentary rock types and themean density of large wood in pools.

    The mean density of large wood in pools waspositively related to stand age. Age or stem diam-eter of forest cover reflects time since timber har-vest in areas such as the Elk River basin, wherelogging dominates the disturbance regime. Thus,the positive associations we found between largewood and the percent area in forests of mediumto very large diameter trees, for example, corrobo-rate negative associations with percent area loggedor harvest intensity in other forested systems

    (Bilby and Ward 1991; Reeves et al. 1993; Mont-gomery et al. 1995; Wood-Smith and Bdhgton1996; Lee et al. 1997). Large wood was also posi-tively related to the amount of forested land insystemswith more agriculturaland urbanizedarea(Richards et al. 1996; Wang et al. 1997; Snyder etal. 2003). The large wood in the stream and indi-cators of timber harvest may not always be related(Lisle 1986; Frissell1992; Ralph et al. 1994),par-ticularly considering ime lags in tree mortality asforests age, decay of in-channel wood from theprevious stand,andwood delivery following epi-sodic disturbances (fires, storms). Because landcover variables had more explanatory power forthe mean density of large wood in pools than forpool morphology, large wood metric5 may be themore sensitive indicators of land managementeffects, especially where logging has been moder-ate as in the Elk River basin.Importance of Spatial Scale in Understanding

    Variation in Large Wood DensityOur use of multiscale analysis suggests areas andprocesses that are most closely linked to largewood in pools. The relatively low proportion ofvariation explained with lithology and forestcover summarized at the segment scale impliesthat wood is delivered from sources in additionto those immediately adjacent to surveyed streamsegments. Explanatory power was greater at thesubnetwork than at the segment scale, possiblybecause the subnetwork scale included many ofthe lower-order tributaries capable of deliveringlarge wood via debris flows to surveyed streamsegments. The most variation was explained atthe subcatchment scale. This scale incorporatesunmapped lower-order tributaries and upslopeareas capable of delivering wood from unchan-nelized hill slope processes. The proportion ofvariation explained by landscape characteristicsdecreased at spatial scales beyond the subcatch-ment, indicating that regression relationshipsmay be less reflective of processes and sourceareas influencing wood dynamics in surveyedstream segments.

  • 8/3/2019 Comparing Riparian and Catchment Influences on Stream Habitat in a Forested, Montane Landscape

    12/18

    192 Burnett et al.

    We did not determine the distance upstreamfrom surveyed segments that explanatory powerbegan to decline. Identification of any such up-stream threshold may help in comparing theimportance of fluvial transport and other wood'delivery processes in these higher-order channelsand, therefore, in designing riparian protectionand timber harvest. To more thoroughly miti-gate negative of effects of logging on wood instreams, our findings indicate that it may be nec-essary to modify management practices alonglow-order tributaries and on hill slopes suscep-tible to mass wasting, as well as along fish-bear-ing channels. This is consistent with theconclusion drawn from other multiscale studiesthat riparian buffers alone may not fully protectstreams from land use impacts (Roth et al. 1996;Wang et al. 1997; Snyder et al. 2003).

    With landscape characteristics summarizedat the network scale, an approximately equalproportion of variation in the mean density oflarge wood in pools was explained by substitut-ing road density (km/km2) or forest cover in re-gression with percent area of sedimentary rocktypes. Dose and Roper (1994) found similar re-sults in the South Umpqua River basin of Oregonwhere the percent area harvested and road den-sity were highly correlated with each other andwere almost equally correlated with change instream width. Road density and forest cover vari-ables (the percent area in forests of medium tovery large diameter trees, the percent area in openand semiclosed canopy forests, and the percentarea in large diameter forests) were correlated atall five spatial scales. The degree of correlation,however, generally increasedwith increasing spa-tial scale, suggesting that roads and forest distur-bances were not always sited together.

    Although road density and forest cover canbe highly correlated, one variable or the othermay have more explanatory power for a particu-lar response (Bradford and Irvine 2000) or at aparticular spatial scale, as we found. Roads andtimber removal share effects on some processesthat shape stream ecosystems (e.g., increasinglandsliding and surface runoff rates) but not all

    (e.g., increasing direct insolation to streams)(Hicks et al. 1991) and may differ in the quality,timing, or magnitude of those effects shared (e.g.,Jones and Grant 1996; Jones 2000). Roads canintercept debris flows that would have otherwisedelivered wood to streams (Jones et al. 2000).However, the amount of wood available for de-livery in our study was probably influenced moreby timber harvest. Two findings suggest this: ( I )more variation in large wood density was ex-plained by a model containing forest cover ateach scale than by the model containing roaddensity; and (2) the only significant relationshipto road densitywas at the network scale, one ofthe two spatial scales that road density and for-est cover were most strongly related. Before oneconcludes that conditions of aquatic habitat orbiota are unrelated to silvicultural activities, itmay be prudent to examine relationships withboth forest cover and road density, particularlywhen these are summarized at finer spatial scales.Additionally, primary influences may be indi-cated by determining if a response variable isrelated to road density or forest cover or bothand at what scales.In conclusion, the spatial scales explored caninfluence interpretations about the importanceof particular landscape characteristics, physicalprocesses, or terrestrial areas to stream ecosys-tems. For example, our finding that variation inthe mean density of large wood in pools was bestexplainedwith landscape characteristics summa-rized at an intermediate spatial scale suggestedthat source areas for important processes wereprobably not fully encompassed at finer scales,but at coarser scales, source areas were includedthat were less connected to large wood dynam-ics in surveyed stream segments. Additionally,had only the catchment scale been examined,wemight have incorrectly concluded that theamount of large wood in pools is unrelated tolithology and forest cover. Although multiscaleanalysis has contributed to exploring land-useeffects on stream ecosystems in urbanized andagricultural settings, this study demonstrated itsbenefits for understanding relationships between

  • 8/3/2019 Comparing Riparian and Catchment Influences on Stream Habitat in a Forested, Montane Landscape

    13/18

    Comparing Riparian and Catchment Influences on Stream Habitat 193

    landscape characteristics and stream habita t ina mountainous area where forestry is the primaryland use. Among-scale similarities and differ-ences in relationships suggested key processesresponsible for those relationships. Conse-quently, analysis at multiple scales may providecritical knowledge about system function andinform land management decisions to betterprotect and restore stream ecosystems.

    ACKNOWLEDGMENTSWe thank Kathryn Ronnenberg and Tami Lowryfor copyediting.Kathryn Ronnenberg isalso grate-fullyacknowledged for her graphics expertise. Weappreciate statistical consulting from GeorgeWeaver and Lisa Ganio. Earlier versions of themanuscript benefited greatly from insightful anddetailed comments by Jason Dun ham, RobertGresswell, Bob Hughes, K.N. Johnson, FredSwanson, John Van Sickle, and three anonymousreviewers. The research is part of the Coastal Land-scape AnalysisandModeling Study (CLAMS)an dwas funded by the U.S. EPA National Health andEnvironmental Effects Research Laboratory inCorvallis, Oregon and the U.S.D.A. Forest ServicePacific Northwest Research Station.

    REFERENCESAkaike,H. 1969. Fitting autoregressive models for pre-

    diction. Annals of the Institute of Statistical Math-ematics 2 1:243-247.

    Akaike, H. 1974. A new look at the statistical modelidentification. IEEE Transactions on AutomaticControl 19:716-723.

    Benda, L., C.Veldhuisen, and J. Black. 2003. Debrisflows as agents of morphological heterogeneity atlow-order confluences,Olympic Mountains, Wash-ington. Geological Society of America Bulletin115:lllO-1121.

    Bilby,R. E., and P. A. Bisson. 1998. Function and dis-tribution of large woody debris. Pages 324-346 inR. J. Naiman and R. E. Bilby, editors. River ecologyand management: lessons from the Pacific coastalecoregion. Springer-Verlag, New York.

    Bilby,R. E., and J. W. Ward. 1989. Changes in charac-teristics and function of woody debris with increas-ing size of streams in western Washington.Transactions of the American Fisheries Society118:368-378.

    Bilby, R. E., and J. W.Ward. 1991. Characteristics andfunction of large woody debris in streams drain-ing old-growth, dear-cut, and second-growth for-ests in southwestern Washington. Canadian Journalof Fisheries and Aquatic Sciences 48:2499-2508.

    Botkin, D., K. Cummins, T. Dunne, H. Regier, M.Sobel,L. Talbot, and L. Simpson. 1995. Status andfuture of salmon of western Oregon and northernCalifornia: findings and options, Appendices, Re-port #8. The Center for the Study of the Environ-ment, Santa Barbara, California.

    Bradford,M. J., and J. R. Irvine. 2000. Land use, fishing,climate change, and the decline of Thompson River,British Columbia, coho salmon. Canadian Journalof Fisheries and Aquatic Sciences57: 13-16.

    Buffington, J.M., T. E. Lisle, R. D. Woodsmith, and S.Hilton. 2002. Controls on the size and occurrenceof pools in coarse-grained forest rivers. River Re-search and Applications 18:507-53 1.

    Burnham,K. P., and D. R. Anderson. 1998. Model se-lection and inference: a practical information-theo-retic approach. Springer-Verlag,New York.Burnett, K. M. 2001. Relationships among juvenileanadromous salmonids, their freshwater habitat,and landscape characteristics over multiple yearsand spatial scales in the Elk River, Oregon. Doc-toral dissertation. Oregon State University,Corvallis.

    Cliff,A. D., and J.K. Ord. 1973. Spatial autocorrelation.Pion Limited, London.

    Cohen, W. B., T.K. Maiersperger,T.A. Spies, and D. R.Oetter. 2001. Modeling forest cover attributes ascontinuous variables in a regional context withthematic mapper data. International Journal ofRemote Sensing 22:2279-23 10.

    Dose, J.I., and B. B. Roper. 1994. Long-term changesin low-flow channel widths within the SouthUmpqua watershed, Oregon. Water Resources Bul-letin 30:993-1000.

    Dunham, J.B., and B. E. Rieman. 1999. Meta-population structure of bull trout: influences of

  • 8/3/2019 Comparing Riparian and Catchment Influences on Stream Habitat in a Forested, Montane Landscape

    14/18

    194 Bumett et al.

    physical, biotic, and geometrical landscape char-acteristics. EcologicalApplications 9:642455.

    Fausch, K. D., C. E. Torgersen, C.V. Baxter, and H. W.Li. 2002. Landscapes to riverscapes: bridging thegap between research and conservation of streamfishes. BioScience 52:483-498.

    FEMAT (Forest Ecosystem Management AssessmentTeam). 1993. Forest ecosystem management: anecological, economic, and social assessment. U.S.Government Printing Office,Report of the ForestEcosystem Management Assessment Team, 1993-793-071, Washington,D.C.

    Flack, V P., and P. Chang. 1987. Frequency of select-ing noise variablesin subset regression analysis: asimulation study. American Statistician 41:84-86.

    Forman, R. T. T., and M. Godron. 1986. Landscapeecology. John Wiley and Sons, New York.

    Fox, J. 1991. Regression diagnostics. Sage Publications,Quantitative Applications in the Social SciencesSeries No. 79, Thousand Oaks, California.

    Franklin, J. F,, and C. T. Dyrness. 1988. Natural veg-etation of Oregon and Washington. Oregon StateUniversity Press, Cowallis.

    Frissell, C. A. 1992. Cumulative effects of landuse onsalmon habitat in southwest Oregon coastalstreams. Doctoral dissertation. Oregon State Uni-versity, Corvallis.Frissell,C.A.,W. J. Liss, C.E. Warren, and M. D. Hurley.1986.A hierarchical framework for stream habitatclassification: viewing streams in a watershed con-text. Environmental Management 10: 199-2 14.

    Gergel, S. 2005. Landscape indicators and watershedloading models: can simple spatial theory guidetheir refinement?Landscape Ecology 20:177-189.

    Gordon, N. D., T. A. McMahon, and B. L. Finlayson.1992. Stream hydrology: an introduction for ecolo-gists. John Wdey and Sons, New York.

    Hankin,D.G., and G. H. Reeves. 1988. Estimating o-tal fish abundance and total habitat area in smallstreams based on visual estimation methods. Ca-nadian Journal of Fisheries and Aquatic Sciences452334444.

    Hardiig, J. S., E. F Benfield, P.V. Bolstad, G. S. Helfman,and E. B. D. Jones, 111. 1998. Stream biodiversity:the ghost of landuse past. Proceedings of the Na-tional Academy of Science 95:14843-14847.

    Hawkins,C .F J. L. Kershner,P A. Bisson,M.D.Bryant,L. M. Decker, S.V.Gregory, D. k McCullough, C.K. Overton, G. H. Reeves, R. J. Steedman, and M.K. Young. 1993. A hierarchal approach to dassify-ing stream habitat features. Fisheries 18(6):3-12.

    Hawkins, C. P., R. H. Norris, J. N. Hogue, and J. W.Feminella. 2000. Development and evaluation ofpredictive models for measuring the biologicalintegrity of streams. Ecological Applications 10:1456-1477.

    Hemstrom, M. A., T. Smith, D. Evans, C. Clifton, E.Crowe, and M. Aitken. 2002. Midscale analysis ofstreamside characteristics in the upper GrandeRonde subbasin, northeastern Oregon. U.S. ForestService, Research Note PNW-RN-534, Portland,Oregon.

    Hicks, B. J., J. D. Hall, P. A. Bisson, and J. R. Sedell.1991. Responses of salmonids to habitat changes.Pages 48 1-5 18 in W. R. Meehan, editor. Influencesof forest and rangeland management on salmonidfishes and their habitat. American Fisheries Sod-ety, Special Publication 19, Bethesda, Maryland.

    Hinch, S. G., K. M. Somers, and N. C. Collins. 1994.Spatial autocorrelation and assessment of habitat-abundance relationships in littoral zone fish. Ca-nadian Journal of Fisheries and Aquatic Sciences51:701-712.Hughes, R. M., S. Howlin, and P. R. Kaufmann. 2004.A biointegrity index (IBI) for coldwater streamsof western Oregon and Washington. Transactionsof the American Fisheries Society 133:1497-1515.

    Hunsaker, C. T., and D. A. Levine. 1995. Hierarchicalapproaches to the study of water quality in rivers:spatial scale and terrestrial processes are importantin developing models to translate research resultsto management practices. BioScience45:193-203.

    Hurlbert, S. H. 1984. Pseudoreplication of the designof ecological field experiments. Ecological Mono-graphs 54: 187-2 11.

    Hynes, H. B. N. 1975. The stream and its valley.Verhandlungen der Internationalen Vereinigungftir Theoretische und Angewandte Limnologie19:l-15.

    Johnson,L. B., G. E. Host, J. H. Olker, and C. Richards.2006. Landscape- and reach-scale predictors oflarge wood abundance in low-gradient streams.

  • 8/3/2019 Comparing Riparian and Catchment Influences on Stream Habitat in a Forested, Montane Landscape

    15/18

  • 8/3/2019 Comparing Riparian and Catchment Influences on Stream Habitat in a Forested, Montane Landscape

    16/18

    196 Burnett et al.

    National Research Council. 2002. Riparian areas: func-tions and strategies for management. NationalAcademy Press, Washington, D.C.

    Omernik, J.M., A.R Abernathy, and L. M. Male. 1981.Stream nutrient levels and proximity of agricul-tural and forest lands to streams: some relation-ships. Journal of Soil and Water Conservation 36:227-231.

    Osborne, L. L., and M. J. Wiley. 1988. Empirical rela-tionships between land uselcover patterns andstream water quality in an agricultural watershed.Journal of Environmental Management 269-27.

    Osborne, L. L., and D. A. Koviac. 1993. Riparian veg-etated buffer strips in water quality restoration andstream management. Freshwater Biology 29:243-258.

    Pabst, R. J., and T. A. Spies. 1999. Structure and com-position of unmanaged riparian forests in thecoastal mountains of Oregon, U.S.A. CanadianJournal of Forest Resources 29: 1557-1 573.

    Pess, G. R., D. R. Montgomery, E. A. Steel, R. E. Bilby,B. E. Feist, and H. M. Greenberg. 2002. Landscapecharacteristics, land use, and coho salmon(Oncorhynchus kisutch) abundance, SnohomishRiver, Wash., U.S.A. Canadian Journal of Fisheriesand Aquatic Sciences 59613-623.

    Pickett, S. T. A., and P. S. White. 1985. The ecology ofnatural disturbance and patch dynamics.AcademicPress, Orlando, Florida.

    Puth, L. M., and K. A. Wilson. 2001. Boundaries andcorridors as a continuum of ecological flow con-trol: lessons from rivers and streams. ConservationBiology 15:2 1-30.

    Ralph, S. C., G. C. Poole, L. L. Conquest, and R. J.Nairnan. 1994. Stream channel morphology andwoody debris in logged and unlogged basins ofwestern Washington. Canadian Journal of Fisher-ies and Aquatic Sciences 51:37-5 1.

    Reeves, G. H., F H. Everest, and J. R. Sedell. 1993. Di-versity of juvenile anadromous salmonid assem-blages in coastal Oregon basins with different levelsof timber harvest. Transactions of the AmericanFisheries Society 122:309-3 17.

    Reeves, G. H., K. M. Burnett, and E.V. McGarry. 2003.Sources of large wood in the main stem of a fourth-

    order watershed in coastal Oregon. Canadian Jour-nal of Forest Resources 33: 1363-1370.

    Richards, C., and G. Host. 1994. Examining land useinfluences on stream habitats and macroinverte-brates: a GIs approach. Water Resources Bulletin30:729-738.

    Richards, C., L. B. Johnson,and G. E. Host. 1996.Land-scape-scale nfluences on stream habitats and biota.Canadian Journalof Fisheries and Aquatic Sciences53(Supplement 1):295-3 11.

    Roth, N. E., J. D.Allan, and D. L. Erickson. 1996. Land-scape influences on stream biotic integrityassessedat multiple spatial scales. Landscape Ecology11:141-156.

    Ryan, S. E., and G. E. Grant. 1991. Downstream ef-fects of timber harvesting on channel morphologyin ElkRiver basin, Oregon. Journalof Environmen-tal Quality 20:60-72.

    Scott, L. 2002. Relationships among in-stream physi-cal habitat, land use, and geology in small coastalstreams of northern Oregon. Master's thesis. Or-egon State University, Corvallis.

    Sandin, L., and R. K. Johnson. 2004. Local, landscape,and regional factors structuring benthicmacroinvertebrateassemblages n Swedish streams.Landscape Ecology 19:501-5 14.

    Snedewr, G. W., and W.G .Cochran. 1980. Statisticalmethods. Iowa State University Press, Ames.

    Snyder, C. D., J.A.Young, R.Villella, and D. P. Lemarie.2003. Influences of upland and riparian land usepatterns on stream biotic integrity. Landscape Ecol-ogy 18:647-664.

    State of Oregon. 1997. Coastal salmon restoration ini-tiative: the Oregon plan for salmon and watersheds.Available: http:/lwww.oregon-plan.org/ (August2004).

    Strahler, A. N. 1957. Quantitative analysis of water-shed geomorphology. Transactions of the Ameri-can Geophysical Union 38:9 13-920.Swanson, F J., J. A. Jones,and G .E. Grant. 1997. Thephysical environment as a basis for managing eco-systems. Pages 229-238 in K. A. Kohm and J. FFranklin, editors. Creating a forestry for the 2 1stcentury: the science of ecosystem management.Island Press, Washington, D.C.

  • 8/3/2019 Comparing Riparian and Catchment Influences on Stream Habitat in a Forested, Montane Landscape

    17/18

    Comparing Riparian and Catchm ent Influences on Stream Habitat 197Swanson, F J., S. L. Johnson, S. V.Gregory, and S.A.

    Acker. 1998. Flood disturbance in a forested moun-tain landscape. BioScience 48:68 1-689.

    Syms, C., and G. P. Jones. 1999. Scale of disturbanceand the structure of a temperate fish guild. Ecol-ogy 80:921-940.

    Thomas, R. B., and W. F Megahan. 1998. Peak flowresponses to clear-cutting and roads in small andlarge basins, western Cascades, Oregon: a secondopinion. Water Resources Research 34:339>3403.

    Thompson, W. L., and D. C. Lee. 2002. A two-stageinformation-theoretic approach t o modeling land-scape-level attributes and maximum recruitmentof chinook salmon in the Columbia River basin.Natural Resource Modeling 15:227-257.

    USFS (U.S. Forest Service). 1998. Elk River watershedanalysis, Iteration 2. Powers, Oregon.

    USFS and USBLM (U.S. Forest Service and U.S. Bu-reau of Land Management). 1994. Record of deci-sion for amendments to Forest Service and Bureauof Land Management planning documents withinthe range of the northern spotted owl. Portland,Oregon.

    Van Sickle, J., J. Baker, A. Herlihy, P. Bayley, S. Gre-gory, P Haggerty, L. Ashkenas, and J. Li. 2004. Pro-jecting the biological conditions of streams underalternative scenarios of human land use. Ecologi-cal Applications 14:368-380.

    Walker,G.W., andN. S. MacLeod. 1991. Geologic mapof Oregon. 1:500,000; Lambert conformal conicprojection; colored; 2 sheets.U.S. Geological Sur-vey, Reston, Virginia.

    Wang, L., J. Lyons, P. Kanehl, and R. Gatti. 1997. In-fluences of watershed land use on habitat qualityand biotic integrity in Wisconsin streams. Fisher-ies 22(6):6-12.

    Wang, L., J. Lyons, and P. Kanehl. 2003a. Impacts ofurban land cover on trout streams in Wisconsin

    and Minnesota. Transactions of the American Fish-eries Society 132:825-839.

    Wang, L., J. Lyons, P. Rasmussen, P. Seelbach,T. Simon,M.Wiley, P Kanehl,E. Baker,S.Niemela, and P M.Stewart. 2003b. Watershed, reach, and riparian in-fluences on stream fish assemblages in the North-ern Lakes and Forest ecoregion, U.S.A. CanadianJournal of Fisheries and Aquatic Sciences 60:491-505.

    Weigel,B.M., L.Wang, F W.Rasmussen, J. T.Butcher,P. M. Stewart,T. P Simon, and M. J. Wiley. 2003.Relative influence of variables at multiple spatialscales on stream macroinvertebrates in the North-ern Lakes and Forest ecoregion, U.S.A. FreshwaterBiology 48:144C-146 1.

    Weller, D. E., T. E. Jordan, and D. L. Correll. 1998.Heuristicmodels for material dischargefromland-scapes with riparian buffers. Ecological Applica-tions 8:1156-1169.

    Wiens, J.A. 1989. Spatial scaling n ecology. FunctionalEcology 3:385-397.

    Wiens, J. A. 2002. Riverine landscapes: taking land-scape ecology into the water. Freshwater Biology47:501-515.

    Wing, M. G., and A. Skaugset. 2002. Relationships ofchannel characteristics, land ownership, and landuse patterns to large woody debris in western Or-egon streams. Canadian Journal of Fisheries andAquatic Sciences 59:79&807.

    Wood-Smith, R. D., and J. M.Buffington. 1996.Mul-tivariate geomorphic analysis of forest streams:implications for assessment of land use impacts onchannel condition. Earth Surface Processes andLandforms 2 1:377-393.

    Ziemer, R. R., and T. E. Lisle. 1998. Hydrology. Pages43-68 in R. J. Nairnan and R.E. Bilby, editors. Riverecology and management: lessons from the Pacificcoastal ecoregion. Springer-Verlag,New York.

  • 8/3/2019 Comparing Riparian and Catchment Influences on Stream Habitat in a Forested, Montane Landscape

    18/18