Flood Disturbance in a Forested Mountain Landscape

September 1998 681

Flood Disturbance in a ForestedMountain Landscape

Interactions of land use and floods

Frederick J. Swanson, Sherri L. Johnson, Stanley V. Gregory, Steven A. Acker

Floods trigger cascades

of physical processesthat alter streams

and riparian zones ofmountain landscapes,

yet affected speciesare resilient

Recent flooding in the Pacific North-west vividly illustrates the complex-ity of watershed and ecosystem re-sponses to floods, especially in steepforest landscapes. Flooding involvesa sequence of interactions that beginswith climatic drivers. These drivers,generally rain and snowmelt, inter-act with landscape conditions, suchas vegetation pattern and topography,to determine the capability of a wa-tershed to deliver water, sediment,and organic material to downstreamareas (Figure 1). Land-use practicescan affect watershed responses toflooding through the influences ofmanaged vegetation patterns androads on delivery of water, sediment,and wood to streams. Watershed re-sponses to floods include geophysi-cal processes, such as landslides andchannel erosion, and related distur-bances of aquatic and riparian or-ganisms and their habitats. We ex-plore these geophysical-ecologicalinteractions using a recent flood in

Frederick J. Swanson (e-mail: [email protected]) is a research geologist at the USDAForest Service, Pacific Northwest ResearchStation, Corvallis, OR 97331. Sherri L.Johnson (e-mail: johnsons@fsl. orst.edu) is apostdoctoral fellow in the Department of Geo-sciences, Oregon State University, Corvallis,OR 97331. Stanley V. Gregory (e-mail:gregorst@ccmail. orst. edu) is a professor inthe Department of Fisheries and Wildlife, Or-egon State University, Corvallis, OR 97331.Steven A. Acker (e-mail: [email protected])is an assistant professor in the Department ofForest Science, Oregon State University,Corvallis, OR 97331.

the Pacific Northwest as an exampleof flood effects in a managed moun-tainous landscape.

Floods in forested mountain land-scapes are distinctly different fromlowland floods. Flood peaks in moun-tain streams are brief (hours to daysin duration) and high, reflecting rapidmovement of water down steephillslopes and channels. Steep to-pography facilitates triggering of de-bris slides down hillslopes, and steepchannel gradients permit rapid move-ment of coarse sediment and woodydebris through stream networks. Con-sequently, flood disturbances in moun-tain landscapes are dominated by me-chanical damage to stream and ripar-ian habitats. Despite the continuouspassage of a flood peak throughstream networks, disturbance patternsare very patchy. Where soil, boulders,trees, and large woody debris do notmove during floods, they provide ref-uge for diverse aquatic and riparianspecies.

In large, lowland rivers, on theother hand, floods are commonly

more tranquil, seasonally predictable,and of much longer duration (weeksto months). These features have per-mitted aquatic and riparian species toundergo evolutionary adaptations toflooding to such an extent that the ab-sence of flooding becomes a distur-bance (Bayley 1995).

The biota of mountain and lowlandstream and riparian systems respondto flooding as a disturbance processin many ways. Physical processes oferosion and deposition during a floodcreate disturbance patches and refugesin which aquatic and riparian organ-isms either recolonize or survive(Townsend 1989). Biotic responsesare characterized by both resistance tochange during the event and resilience(recovery) after the event (Sousa1984, Pickett and White 1985, Reice1994). Postdisturbance biological re-sponses are determined by the distri-bution of disturbance patches and ref-uges across the landscape; byspecies-habitat relations; by dispersalamong patches in the river network;by reproductive strategies; by bioticinteractions, such as competition; andby the availability of food resources.

A major flood in February 1996 inthe Pacific Northwest affected areasof long-term ecological and geo-physical research, providing a his-torical context for interpreting floodeffects in mountain landscapes. Ourdetailed observations at the H. J.Andrews Experimental Forest, a Na-tional Science Foundation-sponsoredLong-Term Ecological Research site,and in other parts of the upperMcKenzie River basin in Oregon (Fig-

682 BioScience Vol. 48 No. 9

ure 2 ) are representa-tive of flood effectsobserved elsewhere inthe Pacific Nortwestand in other mountainous regions, suchas the central Appalachians (Hack andGoodlett 1960).

In this article, we examine selectedeffects of the February 1996 flood ona forested mountain landscape in west-ern Oregon. We describe geophysical

Figure 2. The Willamette River watershed inOregon. (a) Map of Oregon showing theWillamette River. (b) Area of enlargementshowing the west-flowing McKenzie River, atributary to the Willamette River, and theAndrews Experimental Forest (shaded), whichdrains into the Blue River.

processes in stream and riparian net-works.and flood effects on taxa thatrepresent a variety of types and ratesof response. Finally, we consider theinteractions of land use and floods.These observations derive from alarger set of studies of flood effectson ecosystems and interactions offloods with forest land-use practices.

The setting and the flood

The upper McKenzie River basin isrepresentative of much of the CascadeRange of Oregon and Washington, bothin general terms and with respect to re-sponses to the February 1996 flood.This study area is characterized by el-evations from 300 to over 1600 m; bytall, native conifer forests ranging inage from 80 to over 500 years sincethe last major wildfire; and by the de-velopment of scattered conifer planta-tions after clear-cut logging of the areaduring the past 50 years. The steephillslopes are underlain by soils derivedfrom volcanic bedrock, which in someareas are subject to small-scale soilmovement by debris slides. On aver-age, the area receives more than 2500

mm of precipitation annually, 80% ofwhich falls in winter, mainly in theform of snow at elevations above 1000m and as rain at elevations below 400m. The transitional transient-snow zonecan experience high levels of waterdelivery to soil and streams as snow-melt augments rainfall runoff (Harr1981). Runoff flows rapidly throughthe steep stream network.

The 6-8 February 1996 flood in thePacific Northwest involved a sequenceof events typical of major floods in theregion (Harr 1981). Following an earlywinter period of little snow accumula-tion, prodigious snowfall in late Janu-ary brought the snowpack for theWillamette River basin to 112 % of thelong-term average. On February 6, astrong jet stream delivered subtropicalmoisture to the northwest, resulting infour days of intensive rainfall (290mm) while the air temperature was wellabove freezing. This one-two punch ofrain and snowmelt triggered floodflows with return periods of 30 to morethan 100 years in many river systemsin Oregon, Washington, and Idaho. Ona more local scale, flood magnitudesvaried with precipitation patterns, wa-tershed structure, and snow hydrology.For example, cold, dry snow in the up-per elevations stored rainfall duringearly periods of the storm, while warm,rain-saturated snow at lower elevationsmelted rapidly.

These large quantities of water mov-ing through this steep, forested land-scape set off movement of soil, sedi-ment, and large pieces of wood. Theinitiation of this movement, the trans-port of these materials, and their ulti-mate deposition commonly involvedinteractions between the mobile mate-rial and standing forest vegetation. Theresulting complex patterns of flood dis-turbance, interspersed with refuge sitesexperiencing minor flood effects, weresubstantially influenced by vegetationconditions in watersheds at the time ofthe flood.

Geophysical disturbanceprocesses and patterns

Geophysical processes that alterstreams and riparian zones duringfloods operate with highly heteroge-neous patterns of disturbance sever-

September 1998 683

ity. These patterns can be interpretedin terms of downstream variation inphysical processes and geographicvariation in landscape susceptibilityto key processes. Flood waters flowprogressively through the streamnetwork, yet physical disturbanceprocesses, a hallmark of flooding inmountain environments, vary in theirproperties and effects along the gra-dient from hillslopes, through smallstreams, to large channels. Whenconsidering either physical pro-cesses or the biology of mountainstream systems, therefore, it is use-ful to distinguish steep headwaterstreams draining 1-100 ha fromlarger, lower-gradient s t reams(drainage areas of 1-1000 or morekm2) because some key processes(e.g., debris flows or movement offloating logs) and biota are restrictedlargely to one or the other.

Water, soil, sediment, and woodydebris move down hillslopes andstream channels by a cascade of pro-cesses, following the gravitationalflow path. Moving solid materialmay have variable disturbance ef-fects. Sediment and wood, for ex-ample, may move as individual par-

ticles with little ecological impact oras large mass movements with theforce to substantially disturb ecosys-tems. Soil mass movement by debrisslides originating on hillslopes mayenter steep headwater channels andchange into debris flows. Varying insize from hundreds to thousands ofcubic meters, debris flows are water-charged masses of sediment and or-ganic matter that move down head-water channels at velocities of up to10 m/s or more (Hack and Goodlett1960, Sidle et al. 1985), scouringchannel sediment and riparian veg-

etation (Figures 3a and 3b). Debrisflows may enter larger, lower-gradi-ent channels that carry sufficient wa-ter to float large logs on the water sur-face, while gravel and boulders rollaudibly along the streambed.Generally, in wet landscapes, an in-creasing amount of water is availablealong this flow path to dilute the sedi-ment in transit and to float l argerpieces of woody debris. Movingwoody debris can become a signifi-cant agent of disturbance in largerchannels as floating logs ram into orlodge against standing trees, acting

684 BioScience Vol. 48 No. 9

as levers to increase the water’s forceuntil trees topple (Figure 3c).

Phenomena operating at severalgeographic scales create the complexpatterns of disturbance severity ob-served among steep headwaterstreams. Some streams may experi-ence high flows but escape major dis-turbance entirely, whereas otherstreams and riparian zones are se-verely scoured by debris flows (Fig-ures 3a and 3b). The disturbance cas-cade may be interrupted if the debrisslides or flows pile up on roads or onthe edges of the floodplains, for ex-ample, obstructing further flow. In theBlue River watershed, areas ofslide-prone soils or high rates of wa-ter delivery to soils in thetransient-snow zone create predictablegeographic zones of high slide anddebris flow frequency (Figure 4a;Hack and Goodlett 1960, Swansonand Dyrness 1975). Geographic pat-terns of debris flows in the 1996 floodwere nearly identical to patterns trig-gered by floods in the 1950-1995 pe-riod, indicating that the geography ofcontrols on debris-flow occurrencecauses some but not all headwaterstreams to experience repeated, severedisturbance.

On a finer scale, debris flows com-monly affect only parts of the streamnetworks of small watersheds, evenwithin a landscape with a high inci-

dence of debris flows (Figure 4b). Thesmall tributaries that do not experi-ence debris flows may serve as ref-uges for organisms that can contributeto the recolonization of channels thatwere severely affected by debrisflows.

Much of the heterogeneity of dis-turbance severity in larger channelsoccurs along lateral gradients from thechannel axis to the floodplain andfrom reach to reach along the mainchannel. Both lateral and along-streamvariation in flood disturbance areregulated in part by the width of thevalley floor: Narrow valley floors con-fine flood waters, limiting lateralchannel migration and extent of dis-turbance, whereas wide valley floorshave room for both the zone of severedisturbance and areas of more tranquilflow. In wide valley floor areas (un-constrained reaches; sensu Swansonand Sparks 1990, Gregory et al. 1991,Grant and Swanson 1995), sections ofthe main channel experience severedisturbance by complete reworking ofthe streambed and removal or topplingof riparian vegetation, commonly redalder (Alnus rubra), that had estab-lished after previous major floods.Flood waters may also inundate areasof riparian forest in which water veloc-ity and the momentum of entrainedwood and coarse sediment are not

sufficient to damage standing veg-etation. Narrow valley floor areaswith steep, rocky stream banks mayrecord fewer effects of flooding sim-ply because they have less floodplainand riparian vegetation, althoughphysical disturbance can be intense.

Biotic response to flooding

Landforms and geophysical pro-cesses establish the physical templatewithin which aquatic and riparianecosystems operate (Gregory et al.1991). The disturbance history of thelandscape strongly influences pat-terns of upland, aquatic, and riparianbiota in Pacific Northwest landscapes(Schoonmaker and McKee 1988,Morrison and Swanson 1990, Lam-berti et al. 1991, Swanson et al.1992). Floods are the most frequentand intense natural physical distur-bances that alter communities ofaquatic organisms.

A fundamental ecological questionrelated to flood disturbance is: Howdo spatial patterns of flood distur-bance and refuges in a river networkaffect the survival and recovery ofaquatic and riparian organisms? Spe-cies differ greatly in their responsesto floods, depending on the type ofrefuge available to them, their dis-persal capabilities, their mode of re-production, and other life-historytraits that affect persistence throughfloods and subsequent recovery(Table 1). We address the variabilityof biotic response to floods by exam-ining several types of taxa that rep-resent a range of interactions withfloods-riparian vegetation and sev-eral groups of aquatic vertebrates.

Riparian vegetation. Natural ripar-ian forests in many Cascade Moun-tain landscapes are commonly com-posed of narrow bands of red alderthat established after flooding in pre-vious decades. Adjacent to theseflood-reset alder stands are taller,older conifer forests that typicallyestablished after wildfire (Swansonet al. 1992). Thus, spatial patterns ofspecies and age classes of treesstrongly reflect past flooding andother disturbances.

Surveys in major tributaries of theMcKenzie River after the 1996 floodrevealed heterogeneous patterns ofdisturbance to riparian forests. Nu-

September 1998 685

Table 1. Some hypothetical species response mechanisms and timing of responses to flood disturbances as constrained by refuges, dispersal,and reproduction.

Taxon Refuges Dispersal mechanism Reproductive strategy Recovery time

Conifers Upland; undisturbed Fall seed dispersal Seeds More than 30 yearsriparian patches

Giant salamanders Secondary channels; stream- Limited crawling; Nest building, guarding More than 5 years(Dicamptodon tennebrosus) bed interstices terrestrial phase (slow egg develop-

ment)

Sculpins (Cottus spp.) Streambed interstices Swimming (weak) Spawning-broadcast More than 5 years(low fecundity)

Red alder (Alnus rubra) Upland; low-disturbance Fall seed dispersal Seeds Less than 5 yearsriparian areas

Cutthroat trout Secondary channels; Swimming (strong) Spawning-nest building 1-3 years(Oncorhynchus clarki) channel margins (high fecundity)

Caddisflies (Trichoptera) Shallow stream margins; Behavioral drift; Terrestrial mating (26-52- 1-3 yearsfloodplains catastrophic drift week generation time)

Midges (Diptera) Crevices in rocks; shallow Behavioral drift; catastroph- Terrestrial mating (4-12- 3-6 monthsstream margins ic drift; aerial dispersal week generation time)

Aquatic algae Crevices in rocks Sloughed cells Vegetative reproduction; Less than 3 monthssexual reproduction

merous steep headwater streams ex-perienced channel-scouring debrisflows that damaged riparian forests,but other channels did not (Figures2, 3a, 3b, and 4a). The larger tribu-taries also exhibited a heterogeneouspattern of riparian disturbance se-verity. For example, zones of com-plete toppling of riparian alder treeswere interspersed with areas of stand-ing alder or mixed standing andtoppled alder (Figure 5). The mostcomplete disturbance of riparian treescommonly occurred in sites wherethe channel is confined within a nar-row valley floor area containing ri-parian stands dominated by young(less than 30 year) alder (Figure 5,zone b). Standing alder remained inwide valley floor areas, where sec-ondary channels and extensive flood-plain area could accommodate floodwaters outside the zone of severe dis-turbance (Figure 5, zones a and d).In the 1996 flood, these wide valleyfloor areas experienced disturbanceto riparian vegetation by lateral chan-nel changes; the common occurrenceof large, floated logs on toppled al-der trees (Figure 3c) implicatesfloated wood as another disturbancemechanism. Few alder trees occur inconstrained, bedrock-defendedreaches because of limited rootingmedium and frequent scouring (Fig-ure 5, zone c).

Long-term riparian vegetationplots reveal fine-scale details of treeresponse to the 1996 and earlier

floods. A 2.4 ha vegetation plot (Ref-erence Stand 38) in the AndrewsForest, for example, contains dis-tinct zones of differing forest com-position and disturbance severity.Areas of conifer-dominated old-growth forest on upper terrace andfloodplain areas (collectively termed“upland” in Figure 6) are above thedamaging flood waters. On the ac-tive valley floor areas, young alder

stands dominate the forest estab-lished after floods in 1964, 1972,and other years (Figure 6) becausethey have the potential to establishprofusely on the fresh gravel sub-strates left by receding flood waters.Alder trees in these near-channel ri-parian areas experienced approxi-mately 20% mortality in the 1996flood as a result of channel erosionand toppling by floating woody de-

686 BioScience Vol. 48 No. 9

bris. The quiet valley floorarea was inundated by floodwaters, but flow was sotranquil that mortality waslimited to several large co-nifers that fell after theirroot systems were undercutby bank erosion. This pat-tern of alder patches domi-nating in areas of recentflood disturbance and coni-fers dominating other valleyfloor forests is typical of ri-parian zones across the re-gion.

The varied interactions ofstanding trees and downedwoody debris with geo-physical features and pro-cesses produced complexpatterns of disturbance andrefuge. Although floatedwood pieces caused majordisturbance in some stands, toppledand standing vegetation combs floatedwood from flood waters in otherplaces, creating zones of low distur-bance in their lee despite inundationby flood waters. These interactionsproduced a heterogeneous patchworkof zones of high and low severity ofdisturbance, providing both refuge forsurviving species and new sites for es-tablishment.

Aquatic vertebrates. The impact ofthe flood in tributaries of the McKenzieRiver varied by species for the mostabundant vertebrates: cutthroat trout(Oncorhynchus clarki), Pacific giantsalamanders (Dicamptodontennebrosus), and sculpins (Cottusbeldingi and C. bairdi). The highlymobile cutthroat trout are most abun-dant in pool habitats and are found inthe water column, although during win-ter high flows they move into relativelyquiet, shallow water along the streammargin or adjacent to wood and boul-der accumulations. Trout are strongswimmers that are capable of movingquickly in the stream, even in areas ofhigh water velocities. Sculpins, whichare benthic dwellers, are most abundantin the spaces among boulders on thebottoms of pools and riffles. These fishmove down into the streambed duringhigh winter flows. In addition, they areless mobile and swim more slowly

than trout and are therefore less ableto move to side-channel refuges dur-ing floods. Pacific giant salamanders,like sculpins, are bottom dwellers,disperse primarily by crawling, andare not strong swimmers. Duringwinter high flows, giant salamanderlarvae seek cover under boulders andgravel, while the adults seek cover inthe litter and soil of adjacent forests.

Long-term population studies ofaquatic vertebrates in and near theAndrews Forest provided a contextfor measuring the responses of theseaquatic vertebrates to the flood of1996. Densities of adult cutthroattrout following the flood were withinthe range of year-to-year variabilityobserved before 1996 (n = 7-24sample years) in a sample of ninestream reaches in which flood con-dit ions ranged from only highstreamflow and gravel movement tosevere disturbance by debris flowdeposition. Although the averagedensities in the summer of 1996were slightly less than in the prefloodyear (approximately 8% lower afterthe f lood), the average lengthsand weights of adult cutthroat wereslightly greater. The percentageof marked trout recaptured inan old-growth and harvested reachof Mack Creek in 1996 (30%)was similar to year-to-year survivalover winters without major floods(23%). The Mack Creek site experi-enced high flows and move-

ment of sediment, but mostmajor habitat features, suchas large logs and boulders,remained in place. Overall,these observations suggestthat cutthroat trout exhibitedhigh resistance to flood dis-turbance.

Sculpin densities in themain stream of the AndrewsForest were more adverselyaffected by the flood, declin-ing by an average of 70%compared to the previousyear (for reference, the aver-age range of variability be-tween years is 30% [n = 7years]). Although it is notpossible to study the behav-ior of these vertebrates dur-ing a flood, it is known thattheir preferred habitat duringhigh winter flows is in

spaces among cobbles and boulders.When flood discharge is great enoughto cause movement of cobbles and boul-ders along the streambed, organismslimited to that habitat maybe killed bymoving particles.

Densities of giant salamander aremuch more variable among years (av-erage range of variability between yearsis 70% [n - 7 years]), so it is difficult toassess the impacts of the flood on thatspecies. The adults are terrestrial andhave the potential to disperse to othersites, possibly leading to highyear-to-year variability in densities oflarval forms within a stream. In somesmall, steep streams, where the onlyvertebrates are salamanders, debrisflows severely damaged salamanderhabitats and densities were very low. Buteven in these sites, some salamanderswere observed. Perhaps these were newcolonists or survivors from riparian orhillslope soil refuges.

Flood responses can differ among ageclasses of a species. Other studies havefound that young of the year are mostvulnerable to flood flows (Harvey1987). However, the 1996 flood in thePacific Northwest occurred prior to thecutthroat spawning season; therefore,young-of-the-year fry observed in thesummer of 1996 were not exposed tothe high flows in February and were theprogeny of adult survivors. Cutthroat frybiomass after the flood was, on aver-

September 1998 687

age, 43 % greater than biomass be-fore the flood. Fry populations afterthe 1996 flood were the highest ob-served in the 24-year period ofinvestigation at Mack Creek. Sev-eral mechanisms may account forthese increases in fry population,including reduced competition fromspecies that were more negativelyaffected by the flood and enhancedreproductive success as a result ofthe flood flushing fine sedimentfrom spawning gravels.

Ecosystem interactions. The com-plexity of responses of individualtaxa to floods is mirrored by thecomplexity of the sets of ecosystemand community interactions that oc-cur over subsequent years and de-cades. Interactions develop amongspecies with slow responses and spe-cies with more rapid responses tofloods (Table 1). Aquatic communi-ties often recover to predisturbancedensities more quickly than stream-side riparian vegetation, which mayhave provided a major food resourcebefore the flood. Most in-streamfood resources are sensitive to flood-ing (Fisher et al. 1982, Power andStewart 1986) and to channel com-plexity (Bilby and Likens 1979,Golladay et al. 1987). The relativelyslow recovery of woody riparianvegetation and the shade it casts overstreams favor aquatic species able totake advantage of altered food re-sources, such as increased aquaticprimary productivity and reducedterrestrial leaf litter, that occur af-ter removal of the riparian canopy.Community composition may shiftdue to differential responses of spe-cies to floods. Biotic interactions ofcompetition and predation might bereduced initially by flood distur-bance, which might counteract thechanges in aquatic species compo-sition due to changing food re-sources.

Long-term studies of severe de-bris flow impacts in 1986 on a tribu-tary of the McKenzie River providean example of longer-term ecologi-cal responses and interactions (Lam-berti et al. 1991; Stanley V. Gregory,unpublished data.) Debris-flow dis-turbance of r iparian vegetat ionopened up the canopy, resulting inincreased light levels in the stream(Lamberti et al. 1991), which led inturn to several years of increased

primary productivity by aquaticplants and increased secondary pro-ductivity in communities of inverte-brates that graze on aquatic plants.Trout populations were initiallygreatly reduced in the most severelydisturbed reaches, but they recoveredto higher than preflood abundancewithin three to five years. Enhancedfood resources and possibly in-creased foraging efficiency underhigh light levels (Wilzbach et al.1986) may have contributed to thetemporary increase in trout popula-tion and biomass, apparently offset-ting the possible detrimental effectsof elevated temperature and of theaccumulation of fine sediment deliv-ered from upstream areas for a fewyears after the 1986 debris flow. Ul-timately, we expect that riparian veg-etation cover will increasingly shadethe stream and return stream produc-tivity to preflood levels.

Such patterns of ecosystem re-sponse will play out over the next fewdecades at the sites disturbed by the1996 flood, which are scattered overthe flood-affected region. These im-mediate responses, longer-term re-covery processes, and the patchinessof disturbance create a complex mo-saic of habitats and biotic diversityalong stream and riparian networks.

Land-use effects, assessment,and mitigation

Floods can be ranked along a con-tinuum from largely natural, as on theTanana River in Alaska (Marie et al.1998), to managed, such as the ex-perimental flood on the ColoradoRiver initiated by water release fromthe Glen Canyon Dam (Schmidt et al.1998 ). In the February 1996 flood inthe Pacific Northwest, many forestedmountain headwater basins experi-enced a hybrid event-wild (unregu-lated) flooding in a managed land-scape. The watersheds we have stud-ied have experienced road con-struction, and 20-30% of the naturalforest cover has been converted toplantations after clear-cutting. Forestmanagement affects many of the pro-cesses involved in flooding, re-sponses to a flood, and effects of sub-sequent floods on ecosystems andhuman systems (Figure 1). Forestcutting and road development can

increase the delivery of water to soiland streams, increasing streamflows(Jones and Grant 1996, Wemple et al.1996), the initiation of debris slidesand debris f lows (Swanson andDyrness 1975, Sidle et al. 1985), andthe availability of sediment (Grantand Wolff 1991) and coarse, woodydebris in streams.

Despite forest land use in manytributaries of the upper McKenzieRiver basin, the aquatic and ripariantaxa considered in this article perse-vered through the 1996 flood at awide range of sampled sites. We be-lieve that several factors mitigateland-use effects on biota during ma-jor floods in this landscape:

• The heterogeneity of habitats in thismountain landscape provides numer-ous diverse, widely distributed ref-uges. Where boulders, logs, and soilmove during floods, disturbance canbe severe; however, many areas of theflooded landscape were not severelydisturbed.• Forests of the McKenzie River ba-sin and other Cascadian landscapeshave experienced extensive past dis-turbances, such as large-scale wild-fire (Agee 1993). Clearly, these wa-tersheds have repeatedly experiencedthe interactions of floods and otherforest disturbance processes. Manynative species are well adapted toflooding, and some species may ben-efit from flooding at specific pointsin their life histories because it mayhelp them to reproduce and recruitsuccessfully.• Management activities in PacificNorthwest forest lands have influ-enced primarily the frequency andmagnitude, rather than the types, ofprocesses and materials that affectstream and riparian ecosystems. Pro-cesses and materials completely ex-otic to the ecosystem, such as exoticchemicals or species, could havemore detrimental effects on nativebiota.• Some management effects may in-crease or decrease impacts of somedisturbance processes. For example,we have observed that roads on up-per and middle hillslope areas maybe sources of debris flows, but roadson valley floors may trap debris flowsbefore they encounter larger chan-nels.

How will future floods interactwith land use to determine overall

688 BioScience Vol. 48 No. 9

watershed responses to flooding?Some legacies of forest roads, forestcutting, and other land-use activitiesof past decades appear to diminishwith time for many important pro-cesses, such as frequency of debrisslides and debris flows (Swanson andDyrness 1975). However, other man-agement effects may last a long time,such as hydrologic effects of roadsand reductions in large, woody debrisin streams, an important componentof forested stream ecosystems (Gre-gory et al. 1991). New federal policyfor watershed management directssubstantial reductions in the rate offorest cutting and major efforts in wa-tershed restoration, with objectivesthat include mitigation of road effectsand enhanced quality of stream habi-tats (FS and BLM 1994). Successfulimplementation of this policy, whichis contingent on many factors, wouldprobably reduce but not eliminateland-use effects on future floods.

Recognition of the ecological im-portance of the natural dynamics ofecosystems in response to such pro-cesses as streamflow and fire isleading to new approaches to as-sessing and mitigating land-use ef-fects (Poff et al. 1997). Land use isnow often evaluated in terms ofhow it has altered natural, histori-cal disturbance regimes. In the caseof flooding, for example, damsmost commonly reduce peak flows,but extensive road developmentmay increase them. River restora-tion projects have increasingly in-cluded some degree of return to amore natural flow regime (Poff etal.1997). This change represents animportant about-face in approachesto managing dis turbance-proneecosystems from suppressing dis-turbances to accepting them as in-tegral to ecosystem integrity.

Implications for research andresource management

Floods, like other large-scale distur-bance processes (Turner et al. 1997),are highly complex in their physicalaspects and ecological effects, re-flecting the overlapping factors ofpredisturbance ecosystems, the dis-turbance processes themselves, andpatterns of recolonization and geo-morphic adjustment. Floods in moun-

tain watersheds have particularly het-erogeneous spatial patterns of trans-port of soil, sediment, and large logsdown steep hillslopes and throughstream channels. Biotic responses toflood-created mosaics of disturbedsites and refuges are species specific,depending on life-history attributes,mobility, availability and accessibil-ity of refuges, and other factors. Thecomplex mix of positive and negativeflood effects and their interactionswith land use points to the need forlong-term, interdisciplinary studies tounderstand ecological and geophysi-cal roles of floods in natural and man-aged landscapes.

River and riparian ecology andmanagement will benefit by consid-ering several points that overlap bothmountain and lowland environments.It is important to understand the func-tion of habitat complexity across thefull range of riverine environments.Information is also needed about eco-logical and other functions of natu-ral and controlled-flow regimes. Inboth of these cases-habitat and flowregime-reference to natural, historicvariability will provide useful infor-mation for watershed management.The importance of natural habitatcomplexity in the response of biotato flooding implies that managersshould seek to maintain natural typesand levels of habitat complexity sothat flooding can provide its ecologi-cal benefits (Bayley 1995, Poff et al.1997).

Acknowledgments

This work has been supported in partby National Science Foundationfunding of the Andrews ForestLong-Term Ecological Research pro-gram, the USDA Forest Service, thePacific Northwest Research Station,and Oregon State University. Wethank Linda Ashkenas, Ted Dyrness,Matt Hunter, Julia Jones, Judy Li,George Lienkaemper, Art McKee,Kai Snyder (for Figure 4), BeverleyWemple, and Steve Wondzell forsharing their insights. We especiallythank Gordon Grant, who has beeninvolved in this flood study, forstimulating discussion and valuableinsights. We appreciate reviews byRebecca Chasan, Penny Firth, SethReice, and an anonymous reviewer.

References citedAgee JK. 1993. Fire ecology of Pacific North-

west Forests. Washington (DC): Island Press.Bayley PB. 1995. Understanding large river-

floodplain ecosystems. BioScience 45:153-158.

Bilby RE, Likens GE. 1979. Effect of hydro-logic fluctuations on the transport of fine par-ticulate organic carbon in a small stream.Limnology and Oceanography 24: 69-75.

Fisher SG, Gray LJ, Grimm NB, Busch DE.1982. Temporal succession in a desert streamecosystem following flash flooding. Ecologi-cal Monographs 52: 93-110.

[FS and BLM] US Forest Service and US Bu-reau of Land Management. 1994. Record ofdecision for amendments for Forest Serviceand Bureau of Land Management planningdocuments within the range of the northernspotted owl. Washington (DC): US Govern-ment Printing Office.

Golladay SW, Webster JR, Benfield EF. 1987.Changes in stream morphology and stormtransport of seston following watershed dis-turbance. Journal of the North AmericanBenthological Society 6: 1-11.

Grant GE, Swanson FJ. 1995. Morphology andprocesses of valley floors in mountainstreams, western Cascades, Oregon. Pages83-101 in Costa JE, Miller AJ, Potter KW,Wilcock P, eds. Natural and AnthropogenicInfluences in Fluvial Geomorphology: TheWolman Volume. Washington (DC): Ameri-can Geophysical Union. Geophysical Mono-graph no. 89.

Grant GE, Wolff AL. 1991. Long-term patternsof sediment transport after timber harvest,western Cascade Mountains, Oregon, USA.Pages 31-40 in Peters NE, Walling DE, eds.Sediment and Stream Water Quality in aChanging Environment: Trends and Expla-nation. Oxfordshire (UK): International As-sociation of Hydrological Sciences. Publi-cation no. 203.

Gregory SV, Swanson FJ, McKee WA,Cummins KW. 1991. An ecosystem per-spective of riparian zones. BioScience 41:540-551.

Hack JT, Goodlett JC. 1960. Geomorphologyand forest ecology of a mountain region inthe central Appalachians. Washington (DC):US Geological Survey. Professional Paperno. 347.

Harr RD. 1981. Some characteristics and con-sequences of snowmelt during rainfall inwestern Oregon. Journal of Hydrology53:277-304.

Harvey, BC 1987. Susceptibility of young-of-the-year fishes to downstream displacementby flooding. Transactions of the AmericanFisheries Society 116: 851-855.

Jones JA, Grant GE. 1996. Peak flow responsesto clear-cutting and roads in small and largebasins, western Cascades, Oregon. Water Re-sources Research 32: 959974.

Lamberti GA, Gregory SV Ashkenas LR,Wildman RC, Moore KMS. 1991. Streamecosystem recovery following a catastrophicdebris flow. Canadian Journal of Fisheriesand Aquatic Sciences 48: 196208.

September 1998 689

Morrison PH, Swanson FJ. 1990. Fire historyand pattern in a Cascade Range landscape.Portland (OR): US Department of Agricul-ture Forest Service, Pacific Northwest Re-search Station. General Technical Reportno. PNW-GTR-254.

Pickett STA, White PS, eds. 1985. The Ecol-ogy of Natural Disturbance and Patch Dy-namics. Orlando (FL): Academic Press.

Poff NL, Allan JD, Bain MB, Karr JR,Prestegaard KL, Richter BD, Sparks RE,Stromberg JC. 1997. The natural flow re-gime. BioScience 47: 769-784.

Power ME, Stewart AJ. 1986. Disturbance andrecovery of an algal assemblage followingflooding in an Oklahoma stream. AmericanMidland Naturalist 117: 333-345.

Reice SR. 1994. Nonequilibrium determinantsof biological community structure. Ameri-can Scientist 82: 424-435.

Schmidt JC, Webb RH, Valdez RA, MarzolfGR, Stevens LE. 1998. Science and valuesin river restoration in the Grand Canyon.BioScience 48: 735-747.

Schoonmaker P, McKee A. 1988. Species com-position and diversity during secondary suc-cession of coniferous forests in the westernCascade Mountains of Oregon. Forest Sci-ence 34: 960-979.

Sidle RC, Pearce AJ, O Loughlin CL. 1985.Hillslope stability and land use. Washing-ton (DC): American Geophysical Union.Water Resources Monograph no. 11.

Sousa WP. 1984. The role of disturbance innatural communities. Annual Review of

Ecology and Systematics 15: 353-391.Swanson FJ, Dyrness CT. 1975. Impact of

clear-cutting and road construction on soilerosion by landslides in the western Cas-cade Range, Oregon. Geology 3: 393396.

Swanson FJ, Sparks RE. 1990. Long-term eco-logical research and the invisible place.BioScience 40: 502-508.

Swanson FJ, Wondzell SM, Grant GE. 1992.Landforms, disturbance, and ecotones.Pages 304-323 in Hansen AJ, di Castri F,eds. Landscape Boundaries: Consequencesfor Biotic Diversity and Ecological Flows.New York: Springer-Verlag.

Townsend CR. 1989. The patch dynamics con-cept of stream community ecology. Journalof the North American Benthological Soci-ety 8: 36-50.

Turner MG, Dale VH, Everham EH III. 1997.Fires, hurricanes, and volcanoes: Com-paring large disturbances. BioScience 47:758-768.

Wemple BC, Jones JA, Grant GE. 1996. Chan-nel network extension by logging roads intwo basins, western Cascades, Oregon.Water Resources Bulletin 32: 1195-1207.

Wilzbach MA, Cummins KW, Hall JD. 1986.Influence of habitat manipulations on in-teractions between cutthroat trout and in-vertebrate drift. Ecology 67: 898-911.

Yarie J, Viereck LA, Van Cleve K, Adams PC.1998. Flooding and ecosystem dynamicsalong the Tanana River. BioScience 48:690-695.

690 BioScience Vol. 48 No. 9