Response of vegetation, shade and stream temperature to debris torrents in two western Oregon...

This article appeared in a journal published by Elsevier. The attachedcopy is furnished to the author for internal non-commercial researchand education use, including for instruction at the authors institution

and sharing with colleagues.

Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third party

websites are prohibited.

In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further information

regarding Elsevier’s archiving and manuscript policies areencouraged to visit:

http://www.elsevier.com/copyright

Author's personal copy

Forest Ecology and Management 261 (2011) 2157–2167

Contents lists available at ScienceDirect

Forest Ecology and Management

journa l homepage: www.e lsev ier .com/ locate / foreco

Response of vegetation, shade and stream temperature to debris torrents in twowestern Oregon watersheds

Lana E. D’Souzaa,∗, Maryanne Reiterb, Laura J. Sixa, Robert E. Bilbya

a Weyerhaeuser Global Timberlands Technology, Federal Way, Washington 98063, USAb Weyerhaeuser Global Timberlands Technology, Springfield, OR 97478, USA

a r t i c l e i n f o

Article history:Received 19 November 2010Received in revised form 8 March 2011Accepted 9 March 2011

Keywords:Debris torrentRiparian vegetationStream temperatureRed alderDisturbance

a b s t r a c t

This study examines watershed patterns of riparian vegetation, shade, and stream temperature eightyears after extreme storm events triggered numerous debris torrents throughout the Pacific Northwest.We examined twelve impacted streams in two western Oregon watersheds: the Calapooia River in thewestern Cascades and the Williams River in the Coast Range. Red alder (Alnus rubra) and willow (Salixspp.) were the dominant species on debris torrented areas in both watersheds. Post-disturbance vegeta-tion recovery was significant in both watersheds, impacting shade and stream temperatures. However,red alder density, basal area, and height were significantly greater along streams in the Williams Riverwatershed than along streams in the Calapooia River watershed. Willow density, basal area and heightwere similar between the watersheds. Stream shading levels mirrored red alder growth, with greateraverage shading in the Williams River watershed. The greater shade translated into lower summer max-imum stream temperatures and maximum diurnal stream temperature fluctuations in the WilliamsRiver as compared to the Calapooia River watershed. Minimum stream temperatures were not differ-ent between the two watersheds. The rapid re-growth of red alder along the Williams River watershedultimately lead to a rapid decline in maximum summer stream temperatures for that watershed com-pared to the Calapooia River watershed. The location where the disturbance occurred had an importantrole in determining the rate and pathway of stream recovery.

© 2011 Elsevier B.V. All rights reserved.

1. Introduction

In steep forested terrain of the Pacific Northwest, debris tor-rents are a common disturbance impacting headwater streams andriparian vegetation (e.g., Hassan et al., 2005; Swanson et al., 1998;May, 2007; Benda et al., 2005). Debris torrents are typically initi-ated during large storm events when a failure of a stream-adjacenthillside enters a streams channel and creates a fast moving mixtureof water, soil and debris. Debris torrents frequently remove vege-tation and upper soil layers along steep channels, fundamentallyaltering the structure and function of streams and riparian envi-ronments (Pabst and Spies, 2001; Lamberti et al., 1991). Impactscan include changes in nutrient cycling, thermal characteristics,detention storage of water, sediment, and organic matter andrecruitment of coarse woody debris (Lamberti et al., 1991; Swansonand Franklin, 1992; Cover et al., 2010).

∗ Corresponding author. Tel.: +1 253 924 6891; fax: +1 253 924 6736.E-mail addresses: lana.dsouza@weyerhaeuer.com (L.E. D’Souza),

maryanne.reiter@weyerhaeuser.com (M. Reiter), laura.six@weyerhaeuser.com(L.J. Six), bob.bilby@weyerhaeuser.com (R.E. Bilby).

Stream temperature is an important aquatic ecosystem param-eter because of its influence on ecological functions such as fishphysiology, interspecies competition, pathogens, chemical pro-cesses, and gas solubilities (Beschta et al., 2006; Coutant, 1999).Riparian vegetation along small streams strongly influences heatenergy transfer processes (e.g., solar radiation flux, wind speed,air temperature) at the water surface (Beschta et al., 1987; Caissie,2006). Of these heat energy transfer processes, solar radiation fluxexerts the greatest influence on maximum stream temperaturesand is predominantly controlled by the amount of shade providedby streamside vegetation in forested environments (Johnson andJones, 2000; Danehy et al., 2005). Recognition of the role of shadein controlling stream temperature has led to the adoption of reg-ulations over the last several decades requiring the retention ofvegetation along streams in commercial forest, agricultural anddeveloped areas in the Pacific Northwest (Young, 2000).

When vegetation is removed or diminished by disturbancessuch as debris torrents, recovery of shade and stream temperatureis dependent upon the re-establishment and growth of overstoryvegetation. Patterns and rate of re-vegetation are key factors inthe stream temperature recovery and restoration of temperature-mediated functions of aquatic ecosystems such as fish physiology(Beschta et al., 2006). Riparian vegetation is influenced by local and

0378-1127/$ – see front matter © 2011 Elsevier B.V. All rights reserved.doi:10.1016/j.foreco.2011.03.015

Author's personal copy

2158 L.E. D’Souza et al. / Forest Ecology and Management 261 (2011) 2157–2167

broad-scale factors, but few studies have examined both. Climateand geology influence riparian vegetation patterns at relativelylarge spatial scales (Sarr and Hibbs, 2007; Ohmann and Spies, 1998).At the local scale, debris torrents cause scour and deposition (Benda,1990; Costa, 1984), creating fine-scale variation in substrate condi-tions and local landform distribution, which, in turn, influences thecomposition of plant communities establishing on these disturbedsites (Pabst and Spies, 2001; Gecy and Wilson, 1990), successionalprocesses and the rate of recovery of aquatic systems (Miles andSwanson, 1986; Pabst and Spies, 2001).

Local scale factors affect the influence of broader scale factors(Denslow, 1980; Pabst and Spies, 1998; Rot et al., 2000a; Canhamet al., 1994; Hupp and Osterkamp, 1996; Bendix and Hupp, 2000).Many studies have examined recovery of vegetation and streamtemperature following timber harvest and found that recovery cantake 5–15 years; variation in recovery rates include a variety offactors such as location, site conditions, and stream size (Mooreet al., 2005a). For example, Summers (1982) found riparian canopycover following timber harvest and burning recovered faster alongstreams in regions at lower elevation and with wetter climatesas warmer temperatures and higher moisture availability accel-erated vegetation growth. Although many studies have examinedrecovery following major disturbance in the Pacific Northwest, fewhave quantified broad-scale recovery, and meta-analysis is difficultbecause of incommensurable metrics and high degree of variabilityin individual studies (Moore et al., 2005a). Despite the understand-ing of multi-scale controls on riparian vegetation patterns (Ohmannand Spies, 1998; Sarr and Hibbs, 2007), there is little researchexamining local and broad-scale factors following disturbance.This study examines riparian and stream temperature recoveryfollowing debris torrents in headwaters of two watersheds in west-ern Oregon, the Calapooia River in the Western Cascades and theWilliams River in the Coast Range.

In 1996, high magnitude storm events triggered numerousdebris torrents in small streams in western Oregon. To assess long-term water temperature recovery, we established five monitoringstations in untorrented and torrented streams in the CalapooiaRiver watershed, in the Oregon Cascade Mountains, and theWilliams River watershed in the Coast Range Mountains. Whileconducting this monitoring, we observed that vegetation appearedto be recovering at different rates in the two watersheds, suggest-ing that rate of stream temperature recovery also would differ. Thisobservation prompted the establishment of a study to examine veg-etation, stream shade and stream temperature approximately eightyears following debris torrents in multiple small streams in bothwatersheds, including the monitored sites. This study addressedseveral questions regarding the patterns and rate of vegetation,shade and water temperature change post-disturbance: (1) Whatis the relationship between vegetation and local landform andsubstrate types along the study streams? (2) Does vegetation com-position and structure, stream shade and water temperature indebris torrented streams differ between the two watersheds? and(3) How does recovery of stream temperature relate to vegeta-tion and shade recovery and does this differ through time betweenwatersheds?

2. Materials and methods

2.1. Study area and data collection

This study examined multiple headwater streams impacted bydebris torrents in 1996 in managed forests in two watersheds:the Williams River, a tributary to the Coos River in the southernOregon Coast Range (latitude 43.22◦N, longitude 123.70◦W), andthe Calapooia River in the central western Oregon Cascades (lati-

tude 44.30◦N, longitude 122.63◦W). Both study watersheds have aMediterranean climate regime with mild, wet winters and warm,dry summers (Franklin and Dyrness, 1973). Soils in both water-sheds are moderately deep, well drained, loams (Langridge, 1987;Haggen, 1989). Topography is steep with rugged terrain. The studyareas are within the western hemlock (Tsuga heterophylla) forestzone (Franklin and Dyrness, 1988); red alder (Alnus rubra) is a com-mon tree species in riparian areas, while Douglas-fir (Pseudotsugamenziesii) is the dominant tree species in the managed stands ups-lope from the riparian areas.

In 1996, the two study watersheds were impacted by twohigh magnitude storm events: extreme precipitation in the southCoast Range in November and rapid snowmelt in conjunction withheavy rainfall in the western Cascades in February. Both of thesestorms initiated numerous debris torrents. The torrents removedsoil and vegetation from riparian areas along impacted streamsand deposited gravel and cobble in the former riparian areas. Thestreambeds and banks were also completely restructured.

Immediately following the storm events, five streams wereselected to monitor long-term stream temperature recovery. In theCalapooia River watershed, two disturbed streams and one control(undisturbed) stream were monitored; in the Williams River water-shed, one disturbed and one control stream were monitored. Watertemperature data were collected at the five long-term monitor-ing sites from the time of disturbance through summer 2004 usingcontinuously recording thermistors (Onset Computer Corporation,Pocasset, Mass.). Thermistors recorded hourly water temperatureand were calibrated according to protocols established in the WaterQuality Monitoring Technical Guide Book (OWEB, 1999).

Eight years post-disturbance, we selected 12 disturbed studystreams (n = 6 per watershed) to examine relationships betweenriparian vegetation, shade and stream temperature. These studystreams included the disturbed streams on which temperature datawere being collected. Additional study streams were selected fromthe area near the monitored debris torrented study streams: fouradditional streams were selected in the Calapooia River watershedand five in the Williams River watershed. These nine streams wererandomly selected from debris -torrent impacted streams identi-fied in aerial photographs or from landslide inventories.

Study stream reaches in both watersheds were characterized bynarrow widths, averaging 3.5 m wide, with small drainage areas,averaging 4.0 km2 (400 ha). Elevation of the study streams aver-aged 350 m, with a range of 200–500 m. Data from remote accessweather stations (Western Regional Climate Center, 2008, Fig. 1)near the study locations indicated that the Williams River water-shed was characterized by greater annual and growing seasonprecipitation (2036 and 462 mm, respectively) than the CalapooiaRiver watershed (1300 and 312 mm, respectively) over the periodfrom 1996 through 2004. Both watersheds have similar annual solarradiation indices (117.9 and 114.9 kW/m2, respectively).

Data were collected from a 300 m stream reach, randomly estab-lished in depositional areas, at each of the 12 study streams. Fourparallel transects were established within each 300 m area, runningperpendicular to the stream channel and 100 m apart. Transectsoriginated at the stream center and extended 10 m from the chan-nel edge into the adjacent riparian area in both directions. Twocircular plots (2 m radius, 12.6 m2) were randomly located alongeach transect 2–10 m from the stream edge, with one plot on eachside of the channel (n = 8 plots/study stream).

Landform, substrate, and vegetation data were measured in eachcircular plot. Landform type was identified for each plot usingthe riparian landform classification scheme of Rot et al. (2000b):floodplain (<1 m above the channel), low terrace (1–3 m above thechannel), high terrace (>3 m above the channel) and slope (>20%slope). Substrate cover was estimated to the nearest percent infour 1 m2 subplots placed around the circular plot center. Mean

Author's personal copy

L.E. D’Souza et al. / Forest Ecology and Management 261 (2011) 2157–2167 2159

Fig. 1. Location of study watersheds in western Oregon.

cover was estimated by averaging the four plot values for each sub-strate category: soil, rock (sand, gravel, cobble and boulder), wood,organic matter, herbaceous vegetation and moss cover. Total sub-strate cover for a plot could exceed 100% due to overlap of substrateclasses (e.g., herbaceous vegetation growing above a soil substrate).

Vegetation data were collected in each 2 m, circular riparianplot (n = 8 per stream). The density (as individual stems or discreteclumps if multi-stemmed) of trees (≥1.37 m tall) and seedlings(<1.37 m tall) were tallied for each species. Tree and sapling datawere combined by species for total stem count. Diameter of all treeswas measured at breast height (DBH; 1.37 m) for single-stemmedspecies and at ground level for multi-stemmed species. Heightwas estimated using a telescopic measuring pole or with a rangefinder (Laser Technology Impulse 200LR) for each individual tree.To make the calculation of basal area values comparable betweensingle-stem and multi-stem species, diameters for single-stemmedspecies (primarily red alder) were converted to basal diameters

using a regression developed from data collected on a subset ofred alder at each site. Both DBH and basal diameters were mea-sured on these trees and regression equations were developed foreach site (mean R2 = 0.97, n = 20).

All woody plants were identified to species with the exceptionof the genus Salix (willow species). One Salix tree specimen andmost Salix seedlings could only be identified to genus because theylacked the necessary reproductive parts or were too immature todistinguish from congeners. At least two Salix species were presentin the study plots: of those identified to species, S. sitchensis domi-nated (Sitka willow, 94.3%), with scattered S. lutea (yellow willow,5.3%). All Salix species were combined for analyses. Nomenclaturefollows Hitchcock and Cronquist (1973).

Stream shade was estimated using hemispherical photographstaken 1 m above the stream center for each transect during summerand winter months (n = 4 per study stream). Images were capturedusing a Nikon COOLPIX 995 digital camera with an FC-E8 fisheye

Author's personal copy

2160 L.E. D’Souza et al. / Forest Ecology and Management 261 (2011) 2157–2167

converter lens (Nikon Corporation, Tokyo, Japan) fitted in a Self Lev-eling Mount, type SLM2 (Delta-T Devices Ltd, Cambridge, UK) andoriented to magnetic north. The camera was set on a wide zoomwith the focus set to infinity; images were taken at dawn, dusk orduring overcast days to minimize glare. Summer and winter imageswere taken to enable an estimate of shade provided by vegetationestablished on areas impacted by the debris torrent and shade fromvegetation or topographical features outside of the disturbed area.Distinguishing between these shade sources was accomplished byediting winter hemispherical images using Adobe® Photoshop®

Elements v. 5.1 (Adobe Systems, Inc., San Jose, CA, USA) to digitallyremove deciduous branches over the stream channel. Shade fromvegetation growing along the debris torrent track (almost entirelyfrom deciduous trees) was calculated as the difference betweenthe total shade measured in the summer photographs (all vegeta-tion and topographic shade) and shade values edited in the winterphotographs (provided by topography and conifers beyond the tor-rent margins). Solar radiation indices based on latitude, longitude,and elevation were derived using HemiView® software version 2.1(Delta-T Devices, Ltd., Cambridge, UK). Global site factor (GSF) wasour chosen shade metric to describe shade: 1-GSF is the estimatedpotential solar energy obscured or reflected by vegetation or topog-raphy (e.g., Davies-Colley and Payne, 1998; Kelley and Krueger,2005, Delta-T Devices, Ltd., 1999).

Stream temperature data were collected at all study streamsduring July and August of 2004 using continuously recordingthermistors (Onset Computer Corporation, Pocasset, Mass.). Ther-mistors were installed at the lower end of each study stream.Temperature readings were recorded hourly from July 1 to August31, 2004. Daily temperature metrics were compiled from the hourlydata and included maximum, minimum, diurnal fluctuation (dailymaximum–daily minimum) and the mean 7-day maximum (thehighest 7-day moving average of the daily maximum tempera-tures). Absolute maximum and minimum values for the two-monthmeasurement period were determined from the daily values foreach site. Thermistors were calibrated according to protocol estab-lished in the Water Quality Monitoring Technical Guide Book(OWEB, 1999). Ultimately, complete stream temperature recordswere obtained from only 10 of the 12 sites because one tempera-ture monitor in the Williams River watershed became exposed toair and one in the Calapooia River watershed malfunctioned. Anal-yses of temperature data on the study streams were limited to the10 sites with complete records.

2.2. Data analysis

Response variables were compared between the Williams Riverand Calapooia River watersheds using the twelve study streams(n = 6 per watershed) as analysis units. Data for all plots (sub-samples) were averaged to generate a value for each of the studystreams. SAS software was used for all analyses (SAS version 9.2,SAS Institute Inc., 2008), which included: t-tests (PROC TTEST), Chi-square tests (PROC FREQ), simple linear regression (PROC REG),Pearson’s correlation coefficient (PROC CORR), and analysis ofcovariance (PROC GLM).

Watershed differences in riparian vegetation variables (seedlingdensity, tree density, stem density, basal area, and height) weretested using a t-test. Red alder and willow were the only two taxaanalyzed as they constituted the vast majority of the woody plantsencountered in the study plots. Vine maple (Acer circinatum) andcoyotebrush (Baccharis pilularis) were present but were rare (den-sity < 1%).

Distribution of landforms in the two watersheds was testedusing a Chi-square test. Because half of the cell counts were small,we used the exact version of the Chi-square test (SAS InstituteInc., 1999). Watershed differences in cover for each substrate cat-

egory were compared using t-tests. Data were proportional andwere transformed with an arcsine-square-root transformation tostabilize the variance and improve normality (Zar, 1999). Pear-son’s correlation coefficient was used to examine the relationshipbetween mean plot values for substrate category and vegetationvariables (densities and basal areas of alder and willow species) ateach study stream.

Total and vegetation-related shade and several stream temper-ature variables were compared between watersheds using a t-test.Stream temperature metrics included: maximum daily stream tem-perature, minimum daily stream temperature, maximum diurnalstream temperature fluctuation and the mean 7-day maximumstream temperature.

Linear regression was used to evaluate the relationship betweenvegetation, shade and water temperature using the average shadeand vegetation values for each study stream. We examined boththe relationship between vegetation metrics (tree density, basalarea, and height) and vegetation-related shade, and the relation-ships between vegetation-related shade and stream temperaturemetrics (daily maximum, 7-day maximum, and maximum dailyfluctuation). Shade values used in the regression analyses werelimited to those provided by the vegetation established in the areaimpacted by the debris torrent, rather than total shade since our pri-mary interest was post-disturbance recovery of vegetation, shadeand temperature.

At the streams where long-term temperature monitoring wasconducted, differences between disturbed and control streams inseveral stream temperature metrics (i.e., stream temperature ofdisturbed streams minus stream temperature of control streams)were examined for temporal trends and compared betweenwatersheds. Linear regression models were used to describe therelationship between temperature differences and time for eachwatershed. If the watershed slopes were not equal to zero, weused analysis of covariance to test for equal slopes between water-sheds. The non-linear relationships were fit using Lowess curves(Venables and Ripley, 2002) to suggest the general shape of therelationship.

3. Results

3.1. Site factors

Landform distributions were similar between watersheds(p = 0.1729); floodplains (50.7%) and low terraces (39.1%) com-prised the majority landforms at plots. Distributions by substratetype were similar between watersheds, with the exception ofrock and moss covers. Rock cover was significantly higher at theCalapooia River sites, while moss cover was higher at the WilliamsRiver sites (Table 1). A significant but weak correlation was foundbetween substrate and tree basal area and density. Correlationsranged from −0.28082 to 0.38580; the strongest negative corre-lation was between alder density and rock (p = 0.0204) and thegreatest positive correlation between alder basal area and moss(p = 0.0012).

Table 1Mean cover for each substrate types in Calapooia River and Williams River sites.

Substrate type Cover (%)

Calapooia Williams p-Value

Soil 9.4 25.1 0.2461Rock 52.4 33.4 0.0335Wood 8.2 10.1 0.6703Organic matter 30.5 28.3 0.7642Herbs and shrubs 29.1 36.7 0.4732Moss 4.9 12.9 0.0008

Author's personal copy

L.E. D’Souza et al. / Forest Ecology and Management 261 (2011) 2157–2167 2161

Table 2Relative density (%) of trees and seedlings in the Calapooia River (CR) and WilliamsRiver (WR) sites. Species are separated by type: conifer, hardwood, or shrub.

Species Relative density (%)

Trees Seedlings

CR WR CR WR

Conifer treesDouglas-fir (Pseudotsuga menziesii) 0.0 0.0 1.6 2.6Western redcedar (Thuja plicata) 0.0 0.0 0.4 37.1Western hemlock (Tsuga heterophylla) 0.0 0.0 16.4 14.3

Hardwood treesBigleaf maple (Acer macrophyllum) 0.0 0.0 5.0 1.4Red alder (Alnus rubra) 60.5 94.2 48.5 30.9California laurel (Umbellularia californica) 0.0 0.0 0.0 1.4

ShrubsVine maple (Acer circinatum) 0.4 0.3 2.2 0.0Coyotebrush (Baccharis pilularis) 0.0 0.3 0.0 1.2Cascara (Frangula purshiana) 0.0 0.0 8.5 0.0Beaked hazelnut (Corylus cornuta) 0.0 0.0 0.0 0.0Willowa (Salix species) 39.1 5.3 17.5 11.0

Total conifer 0.0 0.0 18.4 54.1Total hardwood 60.5 94.2 53.5 33.7Total understory 39.5 5.9 28.1 12.2Total all taxa 100 100 100 100

a Sitka and yellow willow (Salix sitchensis and S. lutea).

3.2. Vegetation patterns

We found twelve tree and shrub species at our study streams;species varied in abundance by watershed and life stage (Table 2).In both watersheds, red alder and willow were the dominanttrees in the overstory, but the seedlings comprised a variety ofspecies, distributed among conifer, hardwood, and shrub species.The majority of seedlings in the Williams River sites were shade tol-erant conifers (Thuja plicata and T. heterophylla), while CalapooiaRiver sites were dominated by hardwood and shrub species(Table 2).

Tree density and basal area differed between the watersheds.Red alder had a significantly greater density and basal area inthe Williams River sites, whereas willow density was greater(marginally significant) in the Calapooia River sites (Fig. 2a and d).Red alder and willow tree height and frequency were similar in bothwatersheds (Fig. 2e and f). Willow seedling density was greater inthe Calapooia River sites while red alder seedling density was simi-lar (Fig. 2b). However, when tree and seedling data were combinedfor each species, total stem densities were similar between water-sheds for red alder, but willow species were still more abundant inthe Calapooia River sites (Fig. 2c).

3.3. Shade and stream temperature

Total shade was greater for the Williams River sites comparedto the Calapooia River sites (Fig. 3a). Vegetation-related shade wasgreater at Williams River sites than Calapooia River sites, but resid-ual shade was similar between watersheds (Fig. 3b and c). All redalder variables (density, basal area, and height) showed a significantrelationship with vegetation-related shade (Fig. 4a–c), while noneof the willow species variables showed a significant relationshipwith vegetation-related shade.

Maximum daily summer stream temperature, mean 7-day max-imum stream temperature and maximum diurnal fluctuations instream temperature in the summer of 2004 were significantlylower for the Williams River streams than the Calapooia Riverstreams (Fig. 3d–f), but minimum stream temperatures were simi-lar (averaged 11.5 ◦C, p = 0.7222). Vegetation-related shade showeda significant negative relationship with maximum daily streamtemperature, mean 7-day maximum stream temperature, and

maximum diurnal fluctuations in stream temperature at bothwatersheds (Fig. 4d–f).

Data from our long-term sites indicated a significant declinein differences between disturbed and control streams in sum-mer maximum stream temperatures and maximum diurnal streamtemperature fluctuations from soon after the debris torrents toeight years following disturbance, but minimum stream temper-ature patterns were not linear (Fig. 5a–c). Rate of decline in streamtemperature differences between torrented and control streamswas more rapid in the Williams River streams than Calapooia Riverstreams for both maximum daily stream temperature and maxi-mum diurnal fluctuations in stream temperature.

4. Discussion

4.1. Vegetation recovery patterns

We found red alder and willow species dominated areasimpacted by debris torrents in both study watersheds eight yearspost-disturbance, but red alder distribution was more extensivethan willow. Both species exhibit early successional life historytraits (e.g., rapid growth rate and prolific reproduction with small,wind dispersed seeds) that make them ideal species for coloniz-ing and dominating disturbed sites. Willow distribution is oftenrestricted to streamside areas due to high moisture requirementsfor establishment and growth (McBride and Strahan, 1984) and itsneed for frequent disturbance to aid in regeneration and compe-tition reduction (Antos and Zobel, 1984; Bliss and Cantlon, 1957;Zasada et al., 2003).

Although red alder stem density (trees and seedlings) was simi-lar between watersheds, the size of the trees differed. Study streamsin the Williams River watershed supported larger alder trees thanthose at the Calapooia River watershed. In addition, the WilliamsRiver study streams had fewer willow trees and seedlings whilethe Calapooia River study streams had more willow trees and moreseedlings of both willow and alder. This pattern suggests that oncered alder dominance is established, suppression of the understorytrees and seedlings occurs. As red alder grows, it reduces light tolevels too low to support willow (Hawk and Zobel, 1974; Walkeret al., 1986). Red alder was generating significant shade at our studysites eight years after disturbance; it seems likely that this processwas already well underway but further advanced at the WilliamsRiver than the Calapooia River.

The larger size of the red alder at the Williams River studystreams suggests that colonization and/or growth was more rapidat this watershed than the Calapooia River following debris tor-rents. The higher density of shade-tolerant conifer seedlings at theWilliams River watershed also suggests that succession was moreadvanced at these study streams than at the Calapooia River sitesas these species commonly establish after canopy closure occurs(Franklin et al., 2002). The future condition of the riparian commu-nities at our study streams is not entirely clear. The high diversityof seedlings below the alder canopy at most study streams, how-ever, suggests that the plant community will diversify over time.However, diversification of the overstory community at these studystreams will depend upon the frequency with which these streamsare impacted by future disturbances. Red alder could persist asthe dominant species at sites frequently impacted by severe dis-turbance; under such conditions this species can dominate a siteindefinitely (Nierenberg and Hibbs, 2000; Barker et al., 2002). With-out disturbance, colonizing species such as red alder and willoware succeeded by an overstory of shade-tolerant conifers or shrubs(Hawk and Zobel, 1974; Beach and Halpern, 2001b; Hibbs andGiordano, 1996; Nierenberg and Hibbs, 2000), if seed sources arepresent (Beach and Halpern, 2001a).

Author's personal copy

2162 L.E. D’Souza et al. / Forest Ecology and Management 261 (2011) 2157–2167

Fig. 2. Mean red alder and willow (a) tree density, (b) seedling density, (c) stem density (trees + seedlings), (d) tree basal area, (e) tree height, and (f) tree plot frequency.Error bars represent standard errors. Red alder (RA) and willow species (W) p-values are listed.

4.2. Site factors and vegetation

We examined landform and substrate to evaluate the effectof these site factors on vegetation recovery. The dominance offloodplain and low terrace landforms in both watersheds did notallow for a thorough assessment of the influence of local land-forms on patterns of post-disturbance vegetation recovery. The

lack of landform diversity could have been related to the selec-tion of our study streams. Our study streams were placed in thearea near the terminus of the debris torrent, where material trans-ported by the torrent was deposited. Highly constrained channels,bordered by high terrace and slope landforms, are less likely loca-tions for a debris torrent to terminate than are less constrainedstream reaches (Benda, 1990). Therefore, it is likely that the pre-

Author's personal copy

L.E. D’Souza et al. / Forest Ecology and Management 261 (2011) 2157–2167 2163

Fig. 3. Mean (a) total shade, (b) vegetation-related, (c) residual shade, (d) maximum daily stream temperature, (e) mean 7-day maximum stream temperature, and (f)maximum diurnal stream fluctuation (DF) for the Calapooia (CR) and Williams River (WR) study streams eight years after disturbance. Error bars represent standard errors.

Author's personal copy

2164 L.E. D’Souza et al. / Forest Ecology and Management 261 (2011) 2157–2167

Fig. 4. Linear regression models of mean red alder (a) tree density, (b) basal area, and (c) height and vegetation-related shade; and (d) maximum daily stream temperature,(e) mean 7-day maximum stream temperature, and (f) maximum diurnal stream fluctuation (DF). Symbols represent Calapooia River (shaded circles) and Williams River(hollow circles) study streams.

dominance of floodplain and low terrace landforms along our studystream reaches was due, in part, to establishing the study sitesat the torrent terminus. Our findings agree with current litera-ture that deciduous trees with high moisture requirements andtolerance to disturbance, like alder and willow, typically dom-inate on floodplain landforms (Rot et al., 2000b; Mollot et al.,2008).

Substrate influences a variety of factors related to vegeta-tion colonization and growth. These include seedling germinationand establishment (McBride and Strahan, 1984; Gecy and Wilson,1990), potential rooting area (Adams and Sidle, 1987), and depo-sition of alluvial material (Hawk and Zobel, 1974; Fonda, 1974).We found a negative correlation between red alder density androck and a positive correlation between alder basal area and moss.

Author's personal copy

L.E. D’Souza et al. / Forest Ecology and Management 261 (2011) 2157–2167 2165

These relationships suggest that denser overstory vegetation cre-ated more suitable understory conditions for mesic, nonvascularplant species. Moss substrate was most common at the WilliamsRiver sites, which also supported high red alder density. At theCalapooia River study sites, the smaller size and lower density of redalder created more xeric conditions, and bare rock was more abun-dant. These were the only substrate relationships we observed.However, substrate quality may have affected vegetation more dra-matically immediately following the debris torrents. We do notknow initial substrate conditions or early post-disturbance vege-tation characteristics at our study streams, but it is possible thatsubstrate was more varied soon after disturbance than we observedeight years later. It is likely that deposition of material over theinitial debris torrent surface by winter floods and organic matterproduced by colonizing vegetation homogenized surficial substratecharacteristics over time (Hawk and Zobel, 1974; Pabst and Spies,2001).

4.3. Vegetation, shade and temperature recovery

While the amount of light beneath a canopy depends on severalfactors, including the sun’s position in the sky, the atmosphere, sur-rounding topography and the characteristics of the canopy (Weltyet al., 2002; Comeau, 2000), in single species stands the amountof light reaching the forest floor can generally be related to thebasal area of the stand (Comeau, 2000). While not the only speciespresent, red alder was the only species that significantly con-tributed to channel shade. As a result, we found that red alderbasal area was the vegetation metric that most strongly predictedtotal channel shade. Thus, the sites that had greater red alder basalarea had greater shade and lower maximum stream temperatures.Minimum stream temperatures were not related to either channelshade or total shade. This is similar to the findings of Johnson (2004)who found that maximum stream temperatures were affected byshading, but mean and minimum stream temperatures were not.The lack of relationship between minimum temperature and shadeindicates that factors other than canopy cover are more significantin determining this metric.

Our long-term temperature monitoring sites indicated thatduring the first summer after the debris torrents, the summer max-imum stream temperature increased 7 ◦C in the Williams Riverwatershed and increased 5 ◦C in the Calapooia River watershed.Increases of this magnitude are comparable to other ranges foundafter total canopy removal due to forest harvesting prior to streambuffer requirements (2.5–11.6 ◦C) (Gomi et al., 2006; Moore et al.,2005a,b). Faster rates of stream temperature recovery occurred inthe Williams River watershed despite the initially greater tem-perature increase. Six years post-disturbance, summer maximumstream temperatures at the Williams River monitoring site hadreturned to pre-disturbance levels. However, the Calapooia Riversite exhibited maximum summer stream temperatures about 2 ◦Chigher than the reference site in this watershed eight years afterdisturbance, indicating that re-establishment of pre-disturbancethermal regimes had not yet occurred. However, rate of tempera-ture recovery at both watersheds was comparable to that reportedfor recovery following riparian timber harvest, which ranges 5–15years (Moore et al., 2005a,b).

Differences found in shade and stream temperature between thestudy watersheds were caused by greater red alder density and sizeat the Williams River study streams. This difference suggests thatconditions at the Williams River were more conducive to red alderestablishment and growth than those at the Calapooia River sites.As noted earlier, the Williams River sites received greater annualprecipitation and growing season precipitation than the CalapooiaRiver over the eight years following the disturbance. More rapidshade recovery in locations with higher moisture availability also

Fig. 5. Difference in stream temperature metrics for (a) maximum stream tem-perature, (b) minimum stream temperature, and (c) maximum diurnal streamfluctuation of torrented and control (untorrented) streams within the Calapooia andWilliams River sites. R2 and p values are results from the analysis of covariance testfor equal slopes between watersheds, nonlinear data were fit using Lowess curves.

was reported by Summers (1982) for a series of sites where ripar-ian canopy was removed by timber harvest and slash burning. Hefound red alder growth rate higher in the Coast Range than thewestern Cascades following disturbance and proposed that thesedifferences were due to differences in climate. Likewise, Haeussleret al. (1995) examined a climatic moisture gradient within the CoastRange and found that at sites with high moisture, red alder estab-lished over a broader range of seedbed conditions and that mild,wet coastal sites were more conducive for red alder establishment,growth, and survival compared to interior sites. Since we exam-ined only one watershed in the Cascade Mountains and one in theCoast Range, we have a limited scope of inference and cannot con-

Author's personal copy

2166 L.E. D’Souza et al. / Forest Ecology and Management 261 (2011) 2157–2167

clude that the relative rate of vegetation and shade recovery is morerapid in the Coast Range. Nonetheless, our results are consistentwith those obtained by other studies of shade recovery in thesetwo geographies, suggesting a spatially consistent pattern in rateof riparian vegetation recovery following disturbance.

Debris torrents tend to be more severe than many other typesof disturbances that impact streams in that these events frequentlyremove not only vegetation but also much of the soil from riparianareas. The lack of organic soil will limit the plant species capa-ble of utilizing the site and might be expected to limit vegetationgrowth, which would imply that the recovery of shade and watertemperature would be impeded. Surprisingly, we found that therates of shade and temperature recovery we observed for smallstreams subjected to debris torrents were similar to those reportedfor timber harvest (Summers, 1982; Moore et al., 2005a,b), an activ-ity that does not remove organic soil. For example, Brown andKrygier (1970) reported thermal recovery following timber harvestoccurred after six years for a stream in the Oregon Coast Range andJohnson and Jones (2000) found recovery required 15 years for astream in the Oregon Cascade Mountains. The similarity in rate ofrecovery is primarily due to the fact that red alder is capable ofcolonizing and growing rapidly on the mineral substrate createdby debris torrents. As a result, even with total soil removal at ourstudy streams, we found that shade and temperature recovery hadadvanced considerably over the study period at both watersheds.Temperatures in impacted streams in the Williams River watershedmatched those seen in a stream that had not experienced a debristorrent after six years. Therefore, even with the disadvantage ofpoor initial soil conditions, riparian areas impacted by debris tor-rents in western Oregon can achieve rapid recovery of shade andthermal control due to the capacity of red alder to thrive at theselocations.

5. Conclusion

Debris torrents can dramatically alter stream ecosystems byremoving riparian vegetation and soil and re-structuring streamchannels. The recovery of stream ecosystem function followingthese disturbances is governed largely by the recovery rate ofriparian vegetation. In much of the Pacific Northwest red aldercan play a significant role in initiating the recovery of riparianvegetation and restoring stream ecosystem components such asshade and litter input. Therefore, factors that influence red aldergrowth rate also affect the rate of recovery of riparian and aquaticsystem processes. Although we found that the composition of ripar-ian vegetation eight years after debris torrents in the Williamsand Calapooia River watersheds was similar, with red alder andwillow the dominant species, the density and basal area of thespecies differed between watersheds. The more rapid growth ofred alder in the Williams River watershed led to greater streamshade levels and subsequent reduction in various water temper-ature metrics to values observed in reference streams in aboutsix years. At the Calapooia River watershed, temperatures in dis-turbed streams were still greater than an undisturbed referencesite eight years after disturbance, although substantial reductionsin temperature were noted. The Williams River watershed sitesexperienced higher total and growing season precipitation rates,which suggest that differences between the watersheds in recov-ery rate may be due to greater moisture availability at the WilliamsRiver sites.

Thus, the geography of a disturbance may strongly influencethe rate of recovery of key ecological stream functions such asshade, stream temperature, litter inputs and wood delivery. Thisobservation suggests that impacts of disturbance events on streamecosystem processes in small streams in regions where climatic

conditions restrict vegetation growth would be expected to persistconsiderably longer than those in more moist environments. Thisfact suggests that retention of shade along headwater streams inwatersheds where the protection of temperature-sensitive aquaticfauna is a management objective might be of greater significance inregions with slower shade recovery rates than in those areas whererecovery occurs rapidly.

Understanding the factors that determine riparian recoveryalong small streams can facilitate the formulation of water-shed restoration and protection strategies, since small headwaterstreams represent 60–80% of the channel length in many water-sheds (Benda et al., 2005). As a result, a majority of the riparianhabitat in a watershed may be associated with these small streamchannels. These systems also are major contributors of water andmaterials to downstream reaches (Gomi et al., 2002), which oftensupport the primary aquatic resources of public concern. Variousdisturbance types can impact small streams, so the expectation thatmanagement mechanisms can be devised that eliminate severe dis-turbance of these channels is unrealistic. Therefore, ensuring thelong-term integrity of these systems will require maintenance ofthe mechanisms that contribute to post-disturbance recovery. Animportant determinant of the resiliency of small streams is theavailability of a nearby source of seeds from those plant speciesthat initiate the recovery process.

Acknowledgments

We thank Steve Duke for his statistical assistance, and JohnHeffner, Storm Beech, Russell Langshaw, Mark Heffner, and ChrisDuke for their excellent field support. We also thank an anonymousreviewer for useful comments on this manuscript.

References

Adams, P.W., Sidle, R.C., 1987. Soil conditions in three recent landslides in southeastAlaska. Forest Ecology and Management 18, 93–102.

Antos, J.A., Zobel, D.B., 1984. Ecological implications of belowground morphology ofnine coniferous forest herbs. Botanical Gazette 145, 508–517.

Barker, J.R., Ringold, P.L., Bollman, M., 2002. Patterns of tree dominance in coniferousriparian forests. Forest Ecology and Management 166, 331–329.

Beach, E.W., Halpern, C.B., 2001a. Controls on conifer regeneration in managed ripar-ian forests: effects of seed source, substrate, and vegetation. Canadian Journalof Forestry Research 31, 471–482.

Beach, E.W., Halpern, C.B., 2001b. Controls on conifer regeneration in managed ripar-ian forests: effects of seed source, substrate, and vegetation. Canadian Journalof Forest Research 31, 471–482.

Benda, L., 1990. The influence of debris flows on channels and valley floors in theOregon Coast Range, USA. Earth Surface Processes & Landforms 15, 457–466.

Benda, L., Hassan, M.A., Church, M., May, C.L., 2005. Geomorphology of steeplandheadwaters: the transition from hillslopes to channels. Journal of the AmericanWater Resources Association 41, 835–851.

Bendix, J., Hupp, C.R., 2000. Hydrological and geomorphological impacts on riparianplant communities. Hydrological Processes 14, 2977–2990.

Beschta, R.L., Bilby, R.E., Brown, G.W., Holtby, L.B., Hofstra, T.D., 1987. Streamtemperature and aquatic habitat. In: Salo, E.O., Cundy, T.W. (Eds.), Stream-side Management: Forestry and Fishery Interactions. University of Washington,Institute of Forest Resources, pp. 191–232.

Beschta, R.L., Bilby, R.E., Brown, G.W., Holtby, L.B., Hofstra, T.D., 2006. Stream tem-perature and aquatic habitat: fisheries and forestry interactions, 191–232.

Bliss, L.C., Cantlon, J.E., 1957. Succession on river alluvium in northern Alaska. Amer-ican Midland Naturalist 58, 452–469.

Brown, G.W., Krygier, J.T., 1970. Effects of clear-cutting on stream temperature.Water Resources Research 6, 1133–1139.

Caissie, D., 2006. The thermal regime of rivers: a review. Freshwater Biology 51,1389–1406.

Canham, C.D., Finzi, A.C., Pacala, S.W., Burbank, D.H., 1994. Causes and consequencesof resource heterogeneity in forests: interspecific variation in light transmissionby canopy trees. Canadian Journal of Forest Research 24, 337–349.

Comeau, P.G., 2000. Measuring Light in the Forest. Extension Note 42. MC Ministryof Forests Research Program, Victoria, BC, pp. 1–7.

Costa, J.E., 1984. Physical geomorphology of debris flows. Developments and Appli-cations of Geomorphology, 268–317.

Coutant, C.C., 1999. Perspectives on Temperature in the Pacific Northwest’s FreshWaters. Oak Ridge National Laboratory, Oak Ridge, Tennessee, p. 44.

Author's personal copy

L.E. D’Souza et al. / Forest Ecology and Management 261 (2011) 2157–2167 2167

Cover, M.R., de la Fuente, J.A., Resh, V.H., 2010. Catastrophic disturbances in headwa-ter streams: the long-term ecological effects of debris flows and debris floods inthe Klamath Mountains, northern California. Canadian Journal of Fisheries andAquatic Sciences 67, 1596–1610.

Danehy, R.J., Colson, C.G., Parrett, K.B., Duke, S.D., 2005. Patterns and sources of ther-mal heterogeneity in small mountain streams within a forested setting. ForestEcology and Management 208, 287–302.

Davies-Colley, R.J., Payne, G.W., 1998. Measuring stream shade. Journal of the NorthAmerican Benthological Society 17, 250–260.

Denslow, J.S., 1980. Patterns of plant species diversity during succession under dif-ferent disturbance regimes. Oecologia 46, 18–21.

Fonda, R.W., 1974. Forest succession in relation to river terrace development inOlympic National Park, Washington. Ecology 55, 927–942.

Franklin, J.F., Dyrness, C.T., 1973. Natural Vegetation of Oregon and Washington. U.S. Forest Service GTR-PNW-8, Washington, DC.

Franklin, J.F., Dyrness, C.T., 1988. Natural Vegetation of Oregon and Washington.Oregon State University Press, Corvallis, OR.

Franklin, J.F., Spies, T.A., Van Pelt, R., Carey, A.B., Thornburgh Dale, A., Berg, D.R.,Lindenmayer, D.B., Harmon, M.E., Keeton, W.S., Shaw, D.C., Bible, K., Chen, J.,2002. Disturbances and structural development of natural forest ecosystemswith silvicultural implications, using Douglas-fir forests as an example. ForestEcology and Management 155, 399–423.

Gecy, J.L., Wilson, M.V., 1990. Initial establishment of riparian vegetation after dis-turbance by debris flows in Oregon. American Midland Naturalist 123, 282.

Gomi, T., Moore, R.D., Dhakal, A.S., 2006. Headwater stream temperature response toclear-cut harvesting with different riparian treatments, coastal British Columbia,Canada. Water Resources Research, 42.

Gomi, T., Sidle, R.C., Richardson, J.S., 2002. Understanding processes and downstreamlinkages of headwater systems. BioScience 52, 905–916.

Haeussler, S., Tappeiner II, J.C., Greber, B.J., 1995. Germination, survivial, and earlygrowth or red alder seedlings in the central Coast Range of Oregon. CanadianJournal of Forestry Research 25, 1639–1651.

Haggen, J.T., 1989. Soil survey of Coos County, Oregon. USDA Soil ConservationService.

Hassan, M.A., Hogan, D.L., Bird, S.A., May, C.L., Gomi, T., Campbell, D., 2005. Spatialand temporal dynamics of wood in headwater streams of the Pacific Northwest.Journal of the American Water Resources Association 41, 899–919.

Hawk, G.M., Zobel, D.B., 1974. Forest succession on alluvial landforms of the McKen-zie River Valley, Oregon. Northwest Science 48, 245–265.

Hibbs, D.E., Giordano, P.A., 1996. Vegetation characteristics of alder-dominatedriparian buffer strips in the Oregon Coast Range. Northwest Science 70, 213–222.

Hitchcock, C. Leo, Cronquist, Arthur, 1973. Flora of the Pacific Northwest. Universityof Washington Press, Seattle and London.

Hupp, C.R., Osterkamp, W.R., 1996. Riparian vegetation and fluvial geomorphic pro-cesses. Geomorphology 14, 277–295.

Johnson, S.L., 2004. Factors influencing stream temperatures in small streams: sub-strate effects and a shading experiment. Canadian Journal of Fisheries andAquatic Science 61, 913–923.

Johnson, S.L., Jones, J.A., 2000. Stream temperature responses to forest harvest anddebris flows in western Cascades, Oregon. Canadian Journal of Fisheries andAquatic Sciences 57, 30–39.

Kelley, C.E., Krueger, W.C., 2005. Canopy cover and shade determinations in riparianzones. Journal of the American Water Resources Association 41, 37–46.

Lamberti, G.A., Gregory, S.V., Ashkenas, L.R., Wildman, R.C., Moore, K.M.S., 1991.Stream ecosystem recovery following a catastrophic debris flow. Canadian Jour-nal of Fisheries and Aquatic Sciences 48, 196–208.

Langridge, R.W., 1987. Soil survey of Linn County. Oregon.May, C.L., 2007. Sediment and wood routing in steep headwater streams: an

overview of geomorphic processes and their topographic signatures. Forest Sci-ence 53, 119–131.

McBride, J.R., Strahan, J., 1984. Establishment and survival of woody riparian specieson gravel bars of an intermittent stream. The American Midland Naturalist 112,235–245.

Miles, D.W.R., Swanson, F.J., 1986. Vegetational composition on recent land slidesin the Cascade Mountains of western Oregon. Canadian Journal of ForestryResearch 16, 739–744.

Mollot, L.A., Bilby, R.E., Chapin, D.M., 2008. A multivariate analysis examining theeffect of landform on the distribution of riparian plant communities of Wash-ington, USA. Community Ecology 9, 59–72.

Moore, R.D., Spittlehouse, D.L., Story, A., 2005a. Riparian microclimate and streamtemperature response to forest harvesting: a review. Journal of the AmericanWater Resources Association 41, 813–834.

Moore, R.D., Sutherland, P., Gomi, T., Dhakal, A., 2005b. Thermal regime of a head-water stream within a clear-cut, coastal British Columbia, Canada. HydrologicalProcesses 19, 2591–2608.

Nierenberg, T.R., Hibbs, D.E., 2000. A characterization of unmanaged riparian areasin the central Coast Range of western Oregon. Forest Ecology and Management129, 195–206.

Ohmann, J.L., Spies, T.A., 1998. Regional gradient analysis and spatial patternof woody plant communities of Oregon forests. Ecological Monographs 68,151–182.

OWEB. Water Quality Monitoring Technical Guide Book, 1999. Oregon Plan forSalmon and Watersheds, 5-6-2008.

Pabst, R.J., Spies, T.A., 1998. Distribution of herbs and shrubs in relation to land-form and canopy cover in riparian forests of coastal Oregon. Canadian Journalof Botany 76, 298–315.

Pabst, R.J., Spies, T.A., 2001. Ten years of vegetation succession on a debris-flowdeposit in Oregon. Journal of the American Water Resources Association 37,1693–1708.

Rot, B.W., Naiman, R.J., Bilby, R.E., 2000a. Stream channel configuratioin, landform,and riparian forest structure in the Cascade Mountains, Washington. CanadianJournal of Fisheries and Aquatic Science 57, 699–707.

Rot, B.W., Naiman, R.J., Bilby, R.E., 2000b. Stream channel configuration, landform,and riparian forest structure in the Cascade Mountains, Washington. CanadianJournal of Fisheries and Aquatic Science 57, 699–707.

Sarr, D.A., Hibbs, D.E., 2007. Woody riparian plant distributions in western Ore-gon, USA: comparing landscape and local scale factors. Plant Ecology 190,291–311.

SAS Institute Inc., 1999. Procedural Guide, Version 8. SAS Institute Inc., Cary,NC.

Summers, R.P., 1982. Trends in Riparian Vegetation Regrowth Following TimberHarvesting in Western Oregon Watersheds. OSU.

Swanson, F.J., Franklin, J.F., 1992. New forestry principles from ecosystem analysisof Pacific Northwest forests. Ecological Applications 2, 262–274.

Swanson, F.J., Johnson, S.L., Gregory, S.V., Acker, S.A., 1998. Food disturbance in aforested mountain landscape. BioScience 48, 681–689.

Venables, W.N., Ripley, B.D., 2002. Modern Applied Statistics with S. Springer, Veglag,New York.

Walker, L.R., Zasada, J.C., Chapin III, F.S., 1986. The role of life historyprocesses in primary succession on an Alaskan floodplain. Ecology 67,1243–1253.

Welty, J.J., Beechie, T.J., Sullivan, K., Hyink, D.M., Bilby, R.E., Andrus, C., Pess, G.,2002. Riparian aquatic interaction simulator (RAIS): a model of riparian forestdynamics for the generation of large woody debris and shade. Forest Ecologyand Management 162, 299–318.

Western Regional Climate Center, 2008. Remote Automated Weather Station USAClimate Archive, 6-1-2008.

Young, K.A., 2000. Riparian zone management in the Pacific Northwest: who’s cut-ting what? Environmental Management 26, 131–144.

Zar, J.H., 1999. Biostatistical Analysis. Prentice Hall, Upper Saddle River,NJ.

Zasada, J.C., Douglas, D.A., Buechler, W., 2003. In: willow, Salix L. (Ed.), Woody PlantSeed Manual. , third revision. USFS Natural Tree Seed Laboratory.