landscape patterns and legacies resulting from large ... · 1980 eruption of mount st. helens, and...

Landscape Patterns and LegaciesResulting from Large, Infrequent

Forest Disturbances

David R. Foster,1* Dennis H. Knight,2 and Jerry F. Franklin3

1Harvard Forest, Harvard University, Petersham, Massachusetts 01366; 2Department of Botany, University of Wyoming, Laramie,Wyoming 82071-3165; and 3College of Forest Resources, University of Washington, Box 352100, Seattle, Washington 98195, USA

ABSTRACTWe review and compare well-studied examples offive large, infrequent disturbances (LIDs)—fire, hur-ricanes, tornadoes, volcanic eruptions, andfloods—in terms of the physical processes involved,the damage patterns they create in forested land-scapes, and the potential impacts of those patternson subsequent forest development. Our examplesinclude the 1988 Yellowstone fires, the 1938 NewEngland hurricane, the 1985 Tionesta tornado, the1980 eruption of Mount St. Helens, and the 1993Mississippi floods. The resulting landscape patternsare strongly controlled by interactions between thespecific disturbance, the abiotic environment (espe-cially topography), and the composition and struc-ture of the vegetation at the time of the disturbance.The very different natures of these interactions yielddistinctive temporal and spatial patterns and de-

mand that ecologists increase their knowledge ofthe physical characteristics of disturbance processes.Floods and fires can occur over a long period,whereas volcanic eruptions and wind-driven eventsoften last for no more than a few hours or days.Tornadoes and floods produce linear patterns withsharp edges, but fires, volcanic eruptions, and hurri-canes can affect broader areas, often with gradualtransitions of disturbance intensity. In all cases, theevidence suggests that LIDs produce enduring lega-cies of physical and biological structure that influ-ence ecosystem processes for decades or centuries.

Key words: disturbance; forest ecosystem; land-scape; legacy; fire; hurricane; tornado; volcano;flood; Mount St. Helens; Yellowstone fires; Missis-sippi River.

INTRODUCTION

Large infrequent disturbances (LIDs) play an impor-tant role in many forested landscapes, in part be-cause the resulting landscape pattern influences therate and pattern of energy flow, nutrient cycling,wildlife and human responses, and susceptibility tosubsequent disturbance (Turner and Dale 1998).Recognition of the control of landscape pattern andprocess by natural disturbance regimes has funda-mentally altered the interpretation and manage-ment of forest ecosystems (Peterken 1997). Nonethe-less, despite extensive research on disturbances,

there have been few attempts to compare thelandscape characteristics of large disturbances or theimplications of regional differences in disturbanceregimes for understanding forest dynamics andcharacteristics [cf. Rogers (1996)]. In this article, weexamine five LIDs that influence forest ecosystems—hurricanes, tornadoes, floods, volcanoes, and fire—and we evaluate the complex landscape patternsthat result from the interaction of these physicalprocesses with the abiotic and biotic environment(Swanson and others 1990; Foster and Boose 1992).To highlight the distinctive characteristics and dam-age patterns of each LID, we present examples ofwell-studied historical events, contrasting the con-trols over their distribution and effect and evaluat-ing the biological legacies that influence subsequent

Received 14 July 1998; accepted 6 October 1998.*Corresponding author; e-mail: [email protected]

Ecosystems (1998) 1: 497–510 ECOSYSTEMSr 1998 Springer-Verlag

497

ecosystem process. By legacies, we mean physicalstructures (for example, sediment layers, windthrowmounds, and lava flows) and biotic remnants (forexample, seeds, rhizomes, and coarse woody debris)that persist after disturbance. We argue that thespecific landscape patterns and legacies of distur-bance are critical to understanding the contrastingcharacteristics of ecosystem recovery from differenttypes of disturbance. Furthermore, a long-termpattern of LID events comprises characteristic distur-bance regimes that exert an enduring influence onthe landscape-level arrangement of vegetation andecosystem process [cf. Franklin and Halpern (1989)].

HURRICANES

Hurricanes (known as typhoons in the westernPacific region) develop in tropical oceanic regionswhere warm surface water temperatures (26°C orwarmer) and an atmosphere conducive to the devel-opment of cloud towers coincide with an externalstarting mechanism (Foster and Boose 1994). In thenorthern hemisphere, hurricanes often follow aparabolic path, drifting westward and polewardwith the tropical trade winds and then recurvingeastward with the prevailing westerlies at higherlatitudes (Dunn and Miller 1964; Simpson and Riehl1981). Hurricane season extends from June toNovember, with storms most common in Septem-ber, when ocean temperatures are warmest. InEastern North America, hurricane frequency andimportance decrease with latitude and from coastalareas inland.

A mature hurricane is a revolving vortex (counter-clockwise in the Northern Hemisphere) that ex-tends upward 7–8 km and to a radius of 1000 km.The central eye is characterized by low pressure andlow wind speeds, and is surrounded at a radius ofapproximately 30 km by the eyewall, an area ofintense convection where the highest wind speedsand precipitation occur (Foster and Boose 1994).Beyond the eyewall wind speeds and precipitationnormally decline exponentially, such that severewind damage seldom extends beyond 100 km (Frank1977; Anthes 1982). Hurricanes weaken upon land-fall due to increased surface friction and removal ofthe energy source of warm, moist sea air. However,tropical storms may regain their intensity if theymove back out to sea, though colder oceans dimin-ish their power.

The patterns of damage caused by hurricanewinds on forested landscapes are complex and mayappear from field observations to be random andindecipherable (Webb 1958). However, studies ofthe interaction of meteorological, physiographic,

and biotic factors provide insights on the importantcontrolling processes (Smith 1946; Foster and Boose1992; Boose and others 1994). Regional scale (about100–500 km) patterns of forest damage are con-trolled by (a) wind gradients resulting from hurri-cane size, intensity, and storm track; (b) largetopographic features such as coastlines and moun-tain ranges that weaken storms; and (c) regionalvariation in vegetation. Landscape-scale (about 10km) patterns of forest damage are controlled byinteractions among (a) gradients of wind velocityand direction and finer-scale meteorological phe-nomena such as gusts, downbursts, and tornadoes;(b) topographic exposure; and (c) differential standsusceptibility to wind as determined by composi-tion, structure, site conditions, and stand history. Asa consequence of such interactions, the potentialrange in patterns of damage and resulting landscapestructure is enormous. The impact is also highlydependent on the history of natural disturbance andhuman activity (Foster and others 1999).

Hurricane impact across a gradient of spatialscales is illustrated by the 1938 hurricane, one of themost catastrophic windstorms in US history. At aregional scale, the storm created a pattern of vari-able damage across a swath 100 km wide and nearly300 km long that extended from Long Island north-ward into northern Vermont (Figure 1) (Smith1946; Foster 1988a). Damage was concentrated tothe right side of the storm track, where the forwardmovement of the storm (greater than 50 kph)coincided with the rotational velocity, and damagewas influenced by regional variation in forest typeand extent (Foster 1988b). At a landscape scale,vegetation height and composition, controlled inlarge part by prior land use, interacted with windexposure determined by slope, aspect, and land-scape position to produce a patchy damage patterncharacterized by a highly irregular border (Fosterand Boose 1992). At the stand scale, uprooting wasa predominant mode of windthrow, with the inten-sity of damage strongly correlated with tree heightand species (Foster 1988a). Conifers were muchmore susceptible than hardwoods. The resultingpattern was highly variable at all scales, with smallindividual and multiple tree gaps and branch break-age occurring along with extensive windthrow overlarge areas [cf. Naka (1982)].

Hurricanes produce many important and long-lasting legacies, including a patchwork of forest ageand height structure, uproot mounds and downedboles, standing broken snags, and leaning and dam-aged trees (Foster 1988a). The survival, releafing,and sprouting of windthrown and damaged treesmay provide important biotic control on subsequent

498 D. R. Foster and others

ecosystem development (Foster and others 1997;Scatena and others 1997; Zimmerman and others1997). Also, increased loadings of fine woody debrisand leaves may increase the probability of fire andmultiple LIDs in the same area (Patterson and Foster1990; Lynch 1991; Paine and others 1998). Floodsand landslides resulting from the intense and ex-tended precipitation that may accompany hurri-canes can have additional profound effects on eco-system processes and trophic structure (Scatena andLarson 1991; Covich and others 1997). The soillegacies created through tree uprooting and land-slides triggered by windthrow and intense precipita-tion are an inherent part of soil dynamics andlandscape pattern in many landscapes (Stephens1956), and exert a strong, though fine-scaled, con-trol over patterns of regeneration and variation inbiogeochemical processes (Bowden and others 1993;Carlton and Bazzaz 1998).

The meteorological and oceanic controls overtropical storm frequency and intensity interact withcoastline geography to produce strong variation inhurricane disturbance regimes along coastal regionsof the Atlantic and Pacific, including striking differ-ences among relatively adjacent locations. In theCaribbean, some locations experience intense andslow-moving storms with such frequency that hurri-canes may be considered essentially a chronic distur-bance and the vegetation is continually recoveringfrom windthrow (Lugo and Scatena 1989), whereasin northern latitudes powerful storms may occur atcentury or longer intervals. In many hurricane-prone areas, the long-term consistency in stormmovement and the direction of intense winds mayinteract with topography to produce persistent land-

scape-level patterns of vegetation (structure andcomposition) and legacies of hurricane disturbance(Foster and Boose 1994).

TORNADOES

Tornadoes are among the most violent and unpre-dictable of meteorological phenomena and, as short-lived, relatively small, and complex storms, they aredifficult to study directly, impact limited areas, andhave a very long return interval (Fujita 1971;Peterson and Pickett 1991). A tornado is a relativelyorderly, though violently rotating, column of air inwhich the azimuthal velocity drops off rapidly withdistance from the center of the vortex. At groundlevel, the air converges rapidly into the funnel,often creating a convergent pattern of damage andgiving the wind tremendous lifting and movingcapability. Tornado winds greatly surpass hurricanewinds in intensity, reaching an estimated maximumexceeding 400 kph [F5 on the Fujita scale (Fujita1987)].

Tornadoes develop under three meteorologicalconditions that control their geographical, seasonal,and diurnal distribution. The largest and most dam-aging tornadoes, and the most extensive clusters oftornadoes, are spawned in long-lived supercell thun-derstorms. Isolated tornadoes may also develop inordinary thunderstorms, and small clusters of torna-does are frequently produced in the right-frontquadrant of hurricanes following landfall (Bluestein1996). Supercell thunderstorms are highly orga-nized meteorological systems with an intense, rotat-ing updraft that can produce a series of mesocy-clones, each of which is capable of generating a

Figure 1. Landscape andregional scale patterns offorest disturbance resultingfrom five contrasting largeinfrequent disturbances: the1938 Hurricane in New En-gland, the Yellowstone firesof 1988, the eruption ofMount St. Helens in 1980;the tornado at the TionestaScenic Area in Pennsylva-nia, and floods in the Missis-sippi River in 1993. The ar-eas of greatest disturbanceare shown in black. Lesserdisturbance is shown ingrey.

Landscape Patterns of Disturbance 499

tornado that may develop through a five-phasecycle: (a) dust-whirl stage, when dust movementindicates cyclonic circulation from the cloud to thesurface; (b) organizing stage, as condensation fromlowered air pressure fills the expanding funnel; (c)mature stage, when the funnel reaches peak size andintensity and a near vertical orientation; (d) shrink-ing phase, when the funnel decreases in diameterand begins to tilt; and (e) decaying stage, when thevortex shrinks and becomes ropelike. Due to theirassociation with supercell thunderstorms and forma-tion at the boundary between relatively moist mari-time and dry warm continental air masses, torna-does are most common in spring in late afternoonand are concentrated in interior continental re-gions, decreasing toward the coasts (Fujita 1987).

Tornado funnels move with the associated weathersystem, which in North America is predominantlyto the northeast (59%) and east (19%). Violentstorms (F5) have an even greater tendency (72%)to follow a northeasterly direction (Fujita 1987).Tornadoes are extremely variable in intensity, pathlength, width, and continuity. Fujita (1987) hasdocumented a very strong positive relationshipbetween the path length or area impacted and windspeed, with F0 (65–115 kph) storms having a meanpath length of 1.9 km compared with 60 km for F5storms (350–510 kph). Funnel movement and stormcharacteristics such as intensity and width dependon atmospheric and meteorological conditions. Thus,in contrast to hurricanes, tornado impact is notstrongly related to physiography or vegetation (Gra-zulis 1993). Due to this extreme variability inindividual storms, the potential for multiple touch-downs, and the high frequency of multiple storms inclusters, tornadoes can create complex damage pat-terns. The great range in storm intensity leads bothtornadoes and hurricanes to create gradients ofdamage ranging from branch break and single treegaps to extensive areas of complete blowdown.Unlike hurricanes, tornadoes are not generally ac-companied by extensive rainfall or flooding and,due to the extremely long return interval to anyparticular location, the concept of disturbance re-gime is much less useful.

Tornadoes are often associated with, and con-fused with, downbursts and microbursts, which cangenerate very similar patterns of landscape damage(Fujita 1971; Grazulis 1993). Downbursts are strongdowndrafts of air from a thunderstorm or largecumulus cloud that generate an outflow of windthat travels in straight or curved lines outward fromthe point of ground initiation, in contrast to theoften convergent pattern of tornado winds (Grazulis1993). Downbursts exhibit an extreme range in size

from less than 50 m to greater than 10 km in width(microbursts, less than 4 km in width; macrobursts,more than 4 km in width). Downbursts causedextensive damage to old-growth forests along Flam-beau River in Wisconsin (Dunn and others 1983)and in the Five Ponds Wilderness Area of theAdirondacks in New York. At Five Ponds, the dam-age extended across an area exceeding 20,000 haand included widespread blowdown in the typicalstraight-line downburst pattern [cf. Grazulis (1993)].

The tornado ‘‘outbreak’’ of May 31, 1985, thatimpacted Pennsylvania, Ohio, New York, and On-tario illustrates the landscape pattern and ecologicalimpact of severe tornadoes (Peterson and Pickett1995). The outbreak was the worst in the history ofthe region and was comprised of 41 separate torna-does, including eight of category F4 and F5 thatkilled 42 people and injured 1047 (Grazulis 1993).Impact was concentrated between Lake Ontario andLake Huron, with tracks oriented eastward andnortheastward (Figure 1). More than 800 km oftornado damage was mapped, with an average pathwidth of 500 m and range from less than 200 m tomore than 2750 m. The F4 tornado at the TionestaScenic Area had a path length of more than 19 kmand width exceeding 1 km, and it blew downsubstantial areas of second-growth and old-growthforest. The damage pattern within the old-growtharea was remarkable for the sharpness of the edgebetween intact forest and completely windthrownareas (D. Foster unpublished data; Peterson andPickett 1995). Uprooted and broken trees werearranged in both convergent and divergent patternswith wide variation in orientation. The absence ofphysiographic or biotic control was indicated bydamage across all slopes and aspects. Thus, withregard to the factors controlling landscape pattern,the sharpness and extent of the damage pattern, andthe spatial orientation of the damage, these torna-does were strikingly different from the 1938 hurri-cane. Notably, at a stand level the biological andphysical legacies in intensively damaged areas werequite similar for both storms (Foster 1988a; Peter-son and Pickett 1995).

Hurricanes, tornadoes, and downbursts representextremes in the gradient of size and severity of winddamage. In many regions, however, other meteoro-logical phenomena, including frontal storms, canapproach such magnitudes (Swanson and others1988). In addition, all storm types have a gradient ofintensities and many individual storms vary greatlyduring the course of their evolution in terms ofseverity and spatial extent of damage. Most ecologi-cal studies emphasize broad-scale and catastrophiceffects (Yih and others 1991). However, the major


role of tropical storms and tornadoes as low-intensity (for example, gap-forming) events is impor-tant, though often overlooked (Bormann and Lik-ens 1979; Naka 1982; Shaw 1983).

FIRES

Fires typically are caused by lightning and humans,and they generally burn less than a few hectaresbefore they are extinguished because of rain, a fuelcomplex that is not sufficiently flammable, or sup-pression by human activities. However, during longperiods of drought and strong winds, fires can burnout of control, sometimes for months and overmillions of hectares. Therefore, like tornadoes, hur-ricanes, and floods, they are a climatically drivendisturbance. In northern and montane conifer for-ests, large, infrequent fires account for less than 3%of all fires, but more than 95% of the land areaburned (Agee 1993). Such fires are important (orwere important before suppression) in the conifer-ous and mixed conifer–hardwood forests of north-ern Eurasia and northern and western areas ofNorth America, leaving an irregular, long-lastingmosaic of burned and unburned vegetation (Knight1987; Heinselman 1996). The return interval forlarge, infrequent forest fires varies from about 70years to over 300 years (Hemstrom and Franklin1982; Foster 1983; Whelan 1995). Forest structuremay be abruptly changed by intense canopy fires,which burn leaves and small branches and whichare accompanied by surface fires that consumeforest floor and understory vegetation. Notably,most of the larger tree boles are not burned, even ifkilled, and they often remain standing for 10–50years after a fire. Such snags and residual living anddead organic matter are enduring legacies of firethat are important for wildlife habitat, nutrientdynamics, ecosystem function, and forest recovery(Harmon and others 1986).

The spread of fire across a landscape involves thedrying of fuels and the production of gases (pyroly-sis) that are ignited and burned in visible flames, aprocess known as flaming combustion (Johnson 1992;Agee 1993; Whelan 1995). The flaming front movesmost rapidly (up to 10 kph) with strong windsbecause the flames are pushed closer to the un-burned fuels ahead. After flaming combustion hasignited and burned most of the volatiles in the fuelbase, the remaining carbon is subjected to glowingcombustion, a process that is much slower, is noteasily extinguished, and enables the fire to continueburning for several days or weeks, even in dampweather. As the hot air rises, cooler air is drawn intothe base of the flames, providing oxygen. This

upward air movement, combined with the addi-tional turbulence created by winds, often carriesglowing embers to new points of ignition, a processknown as spotting. Thus, although a fire starts at oneplace, it spreads by flaming combustion, glowingcombustion, and spotting. Moreover, multiple igni-tions are likely to occur when climatic conditionsare favorable. If not extinguished, the spread of asingle fire occurs at variable rates, often shiftingdirections as the prevailing winds change (Foster1983). The pattern of a burned area will be morelinear if there have been only a few ignition pointsand strong directional winds.

Because of the upward movement of hot air,flames burn uphill more readily than down. There-fore, leeward slopes often burn with less intensity, ifthey burn at all, illustrating one way that topogra-phy influences fire and the postfire landscape pat-tern that develops. Ridges can function as fire-breaks, along with rivers, lakes, and areas of less fuelor less flammable fuels [for example, see Melansonand Butler (1984), Foster and King (1986), andGrimm (1984)]. The abundance of fuel may affectfire spread when conditions contributing to flamma-bility are not extreme, but high-intensity firestorms(crown fires, or wildfires) will typically burn a largerportion of the landscape more evenly. For example,young as well as old forests burned during the 1988‘‘firestorms’’ in the Greater Yellowstone area, whichoccurred primarily during the drier, windier periodof late summer. Even then, however, some tracts offorest were not burned, creating an irregular land-scape mosaic (Figure 1) (Christensen and others1989). As the intensity of a fire lessens during theevening when wind is low and humidity is high, onleeward slopes, or at the edge of a burn, theproportion of trees and other plants that are killed isreduced. Thus, the perimeter of fires typically has adiffuse border, where some of the trees are killedand some survive. Trees that survive but are dam-aged or scarred may be more susceptible to insectsand fungi (Gara and others 1984), illustrating howfires can lead to secondary effects on vegetationdynamics. The complex of dead trees, organic mat-ter, and surviving organisms, and the pattern ofvariable-sized patches of unburned vegetation andareas of different burn intensity, comprise the lega-cies of fire that are so important in diversifyingforest landscapes (Heinselman 1973; Swanson 1981).Variation in fire size, intensity, type, and returnfrequencies as controlled by weather, vegetation,and the distribution of firebreaks produces character-istic fire regimes that vary on a broad scale. In turn,these contrasting regimes produce distinctive long-term patterns in forest age and height structure,


composition and the distribution, and abundance offire-generated legacies such as standing dead trees,coarse downed debris, and charcoal.

FLOODS

Floods occur when the rate of water supply exceedsthe capacity of stream-channel drainage. Such con-ditions develop during periods of heavy rains andrapid melting of snow and ice, making most floods aclimatically driven phenomenon. Other floods oc-cur when dams break, whether the dams are con-structed by humans or were caused by ice jams,landslides, or beavers, or by tectonic and othergeological processes, including volcanoes. In nar-row, upper-watershed streams, floods can exert apronounced geomorphic impact that may shapelocal landscapes, vegetation patterns, and ecosystemfunction for decades (Baker and others 1988). How-ever, in large, lower-watershed rivers, the longhistory of regular flooding is indicated by the pres-ence of geomorphically and vegetationally distinctfloodplains, and floods play a fundamental role inthe maintenance and function of the ‘‘river–floodplain’’ ecosystem (Bayley 1995; Sparks 1996;Bornette and Amoros 1997). Notably, aside fromsevere economic damage to buildings, roads, bridges,and cropland that were developed with inadequateregard to the inevitability of floods, the ecologicaldamage of regular flooding may be minor. Flood-plain plants and animals are well adapted to periodicinundation, and consequently riverine ecologistsagree that most ‘‘flood pulses’’ should not be viewedas a disturbance (Bayley 1995). Indeed, many organ-isms in the river and on the floodplain depend onfloods (for example, many species of fish for spawn-ing and plants for continued establishment on depos-its of nutrient-rich sediment). Typically, the mechani-cal force of floodwater is not adequate to increaseplant mortality rates, especially when flooding oc-curs in spring before leafout of deciduous trees.Geomorphic processes such as oxbow formation, icescouring, and bank erosion may cause the death ofsome trees and changes in the landscape mosaic, butvegetation structure and resource distribution aregenerally affected only over limited areas. Consider-ing the range of benefits that floods can bring toriver–floodplain ecosystems, the major disturbanceto many river ecosystems is flood suppression (Wil-liams and Wolman 1984; Bayley 1995).

Exceptional events such as the great floods of1993 in the Mississippi River drainage system (pri-marily the lower Mississippi, Illinois, and MissouriRivers) may create extensive, though highly vari-able, damage patterns (Figure 1). Tree mortality

following the 1993 event was widespread, reaching50%–90% in some areas. The primary cause ofmortality was excessively long periods of inunda-tion during the middle of the growing season thatkilled the roots of many species due to the develop-ment of anaerobic conditions. As noted by Sparks(1996), some mortality may result, paradoxically,from drought stress due to the inability of thedepleted or damaged roots to replace the transpiredwater. Thus, a flood-caused disturbance can behighly selective, with severity related to flood dura-tion, geomorphology, species composition, and spe-cies physiology (Sparks 1996). Slightly higher groundon the floodplain, such as that of older terraces,would have shorter periods of inundation, and thusit is not surprising that topography is often corre-lated with the mosaic of stands with high and lowplant mortality (Sparks 1996). The constrained na-ture of river–floodplain ecosystems leads to smallerand more linear disturbed areas than following largefires and hurricanes. However, the kinds of floodsthat cause disturbances typically last for months,like large fires, rather than minutes or hours (as isthe case with volcanic eruptions, tornadoes, andhurricanes).

VOLCANIC ERUPTIONS

Volcanic eruptions are usually complex events thatspan a wide gradient of intensity, size, and durationand involve the interactions of multiple distur-bances, including explosive blasts, thermal and toxicchemical waves, landslides, glowing avalanche de-posits, debris flows (lahars), lava flows, and airfallsof volcanic ejecta (tephra). Each disturbance typehas unique attributes with regard to impact, spatialcharacteristics, interactions with existing landforms,and resultant landscape-level patterns. Interactionsbetween these disturbance types are important indetermining the nature of the postevent landscape.Geographically, volcanic activity is strongly con-trolled by tectonic processes such as occur along thePacific Rim of western North America. Earthquakes,another consequence of the tetonic activity charac-teristic of this region, also produce widespreadgeomorphic changes that are critical to the under-standing of forest growth and patterns (Veblen andothers 1992).

Eruptive-related disturbances can be very large(for example, 1,000,000 km2 or more), particularlyin the case of airfall depositions, although a scale of10–1000 km2 is more common. Patterns are typi-cally either isodiametric, where there is an epicenterfor the disturbance, such as with an explosive blast,or linear to dendritic where a flow is involved, such


as with a landslide or lava flow. Edge definitionvaries widely but is generally most abrupt with themost intense disturbances, such as glowing ava-lanches. Volcanic disturbances tend to be short termand high intensity, although debris flow and airfalldeposition vary substantially in intensity and mostexhibit severity gradients related primarily to dis-tance from source. Volcanic disturbances interactwith and are influenced by physiography and exist-ing ecosystem conditions to varying degrees, andthus prediction of landscape patterns is moderate tohigh for volcanic disturbances (given a known pointof origin and magnitude of the event). Flow-typedisturbances are most predictable because of topo-graphic controls on their direction and boundaries.Airfall depositions are least predictable because oftheir dependence upon wind speed and direction inthe upper and lower atmosphere. A unique andimportant characteristic of volcanic eruptions istheir lack of seasonality and predictability of returninterval. Furthermore, seasonal timing of the erup-tive event has a very profound effect on its impactsand the resulting postdisturbance landscape pat-tern.

The eruption cycle of Mount St. Helens thatbegan on 18 May 1980 provides a well-studiedexample of at least seven volcanic disturbance types(Figure 1). Initiated with a series of small explosionsand minor airfall deposition during early 1980, alandslide on 18 May unroofed the core of thevolcano and led to a blast and thermal wave thatcreated a roughly semicircular ‘‘blast’’ zone. Glow-ing avalanches subsequently occurred on the moun-tain itself, and debris flows occupied major drain-ages. Tephra fell on hundreds of thousands of squarekilometers, but magma was extruded only withinthe exposed mountain crater. Thus, unlike thefamiliar tectonic processes in Hawaii, lava flow wasinsignificant. The blast zone was subjected to at leastthree different disturbances types—explosive blast,thermal wave, and tephra deposition—and someareas, such as drainages, were subjected to up tothree other disturbance types—landslide, debris flow,and glowing avalanche deposit. Topography was animportant factor influencing the distribution andintensity of gravity-based events, such as the land-slide, mudflows, and glowing avalanches. Topogra-phy had less effect on the explosive blast, althoughmountaintops and ridges provided protection forsome locations and influenced the patterns of treeblowdown.

Biological legacies were present throughout mostof the blast zone despite the intensity and multiplic-ity of disturbance events (Franklin and others 1985).Surviving animals, perennating plant parts, and

other belowground organisms varied widely in abun-dance after the eruption. None were associated withglowing avalanche deposits; few on the landslidedeposits along the margins of the flow; moderateabundance associated with debris flows; and greatabundance on the blast zone. Most surface waterswithin the blast zone also retained many fish,amphibians, and plankton. Death of organisms inthe areas subjected only to tephra deposition wasgenerally selective and left the structure and func-tion of the affected ecosystems relatively intact.

The combined effects of predisturbance landscapepatterns, multiple disturbances, physiography, andvarying levels of biological legacies resulted in verycomplex spatial patterns within the blast zone. Forexample, rapid recovery of vegetation on recentlyclear-cut areas contrasts with slow recovery of areasoccupied by virgin forests (Franklin and others1985; Franklin and Halpern 1989). Similarly, organ-isms within persistent late-spring snowpacks wereprotected from the blast and development of a hardtephra crust. Overall, the presence of residual organ-isms ensured that recovery within the large blastzone developed from many well-distributed focirather than through recolonization from the mar-gin. The importance of the seasonal timing of theMount St. Helens event is clear. A comparableeruption during midwinter with a deeper, morewidespread snowpack would have greatly extendedmudflows and would have provided a protectivecover for more of the blast zone, whereas a similarevent during the late summer would have hadreduced mudflows and snowbank refuges. Signifi-cantly different posteruptive landscapes would re-sult, depending on the season of eruption.

COMPARISON OF LIDS

IN FOREST LANDSCAPES

The complexity of disturbance patterns resultingfrom LIDs has led to their characterization as ran-dom, chaotic, and uninterpretable (Shaw 1983).However, an understanding of their interactionswith the physical structure and biological character-istics of the landscape greatly enhances our ability tointerpret and characterize the patterns of distur-bance, to reconstruct their occurrence, and to pre-dict their distribution in time and space (Fujita1987; Arno and others 1993; Boose and others1994). Notably, this approach has led to the recogni-tion that variation in the landscape-level distribu-tion of disturbance type and intensity helps toexplain spatial patterns of ecosystem structure andfunction (Foster 1983; Grimm 1984; Lugo andScatena 1989; Motzkin and others 1996). Over long


periods, the established disturbance regime of LIDsin a region may be instrumental in determining thelandscape patterns of forest stands and processesthat characterize a regime. Thus, an understandingof specific disturbance events and the legacies thatthey create, coupled with information on the char-acteristic distribution of these events in time andspace, is critical in the interpretation of forest land-scape patterns (Table 1).

The physical characteristics of a disturbance typeand the nature of an individual event clearly estab-lish the initial limits set on the spatial extent,pattern, and temporal distribution of damage. Impor-tant attributes include the climatic and meteorologi-cal constraints on the process, the extent and man-ner in which damage is controlled by predisturbancevegetation and physiography, and variation in theduration, intensity, and types of disturbance effects.The geographical distribution of the LIDs that wehave discussed is largely determined by their control-ling meteorological drivers, for example, supercellthunderstorms for tornadoes, warm oceans andcloud towers for hurricanes, drought and lightning

for natural fires, and extreme precipitation andrunoff for floods. In turn, these connections be-tween many LIDs and broad-scale weather patternsmay impose constraints on subsequent landscapepatterns. Thus, tornadoes move with thunderstormsin prevailing frontal systems and tend to formwest-to-east damage tracks in North America (Fu-jita 1987). Atlantic hurricanes have a northwesterlymovement that combines with their dependence onwarm surface water and counterclockwise rotationto produce a decreasing gradient of damage inlandfrom the eastern US coast and generate strongestwinds from the right-hand (easterly) side of thestorm (Foster and Boose 1992). Even on relativelysmall islands such as Puerto Rico, this northwesterlytracking and the decline in wind intensity over landhas led to historical gradients of hurricane impactsthat control canopy structure and forest composi-tion (Scatena and Larsen 1991; Boose and others1994). Of the LIDs reviewed, only volcanoes areindependent of climate, although their patterns ofimpact and vegetation recovery may be quite sensi-tive to seasonal timing.

Table 1. General Characteristics of the Large Infrequent Disturbances (LIDs) Discussed in This Paper

LID CharacteristicVolcanicEruptions Tornadoes

ForestFires Hurricanes

RivervineFloods

Duration of event Hours Minutes Weeks/months Hours Weeks/monthsReturn interval (years) 100–1000 100–300 75–500 60–200 50–100Size of event (km2) 5–100 5–100 50–20,000 50–100,000 50–50,000Geographic location Volcanic mts Inland Inland Warm coasts RiparianSpatial predictability High Low Moderate Moderate HighTemporal predictability Moderate Low Moderate Moderate ModerateEnergy point source or flow Variable Flow Flow Flow FlowMultiple initiations No Yes Yes No Yes

Cause of disturbanceBurning Localized No Extensive No NoIntense heat Extensive No Extensive No NoInundation (flooding) Localized no No Localized ExtensiveStrong wind No Yes Periodic Yes NoLandslides Localized No Localized Localized Bank erosionExplosive blast Extensive No No No NoGlowing avalanche Localized No No No NoDebris flow Localized No No No PeriodicLava flow Extensive No No No NoToxic chemicals Extensive No No No NoAirfall deposition Variable No No No No

Variables affecting severityDistance from energy source Yes Yes Yes Yes YesClimatic factors No Yes Yes Yes YesSeason Yes Yes Yes Yes YesTopographic factors Yes No Yes Yes YesVegetation structure Yes No Yes Yes Yes


Climatic control also constrains the seasonal occur-rences of many disturbances, for example, hurri-canes to August to November, tornadoes to July andAugust, floods to the season of precipitation orsnowmelt, and fires to periods of low fuel moisture.Seasonal timing, in turn, controls the potentialrange of impact. For example, hurricanes and torna-does generally occur during the growing seasonwhen the presence of leaves makes trees moresusceptible to blowdown. However, the occasionallate-season storms that occur after leaf fall in temper-ate areas exert quite distinctive effects, includingdifferential damage to evergreens and a lack ofmajor organic and nutrient inputs from massive leaffall. Other interesting interactions occur betweenthe seasonal timing of an LID and the physiologicalstate of the vegetation or seasonal nature of thelandscape. Volcanic impact may be strongly influ-enced by the presence and distribution of a snow-pack, flooding by the physiological condition of theinundated trees, and fire by the microclimate andfuel-moisture conditions controlling seasonal leaf-out. Such seasonal effects can result in contrastingimpacts and patterns of ecosystem dynamics follow-ing similar disturbances. Climatic control over distur-bance regimes and subsequent ecosystem responseyields strong variability in disturbance frequencyand long-term patterns of ecosystem processesthrough cyclical and longer-term meteorologicalchanges. Thus, Atlantic hurricane frequency hasundergone multidecadal cycles associated withchanges in weather on and adjacent to Africa, and arecent increase in Pacific tropical storms has beenlinked to ENSO activity and warming of near coastalwaters. Over longer periods, natural or human-generated changes in climate will certainly drivemajor changes in fire, hurricanes, and tornadoes.Due to the longevity of trees and the persistentlegacies of disturbance in a landscape relative to therate of climate change, it is likely that many land-scape patterns of forest conditions and structures ina landscape have arisen as a consequence of adisturbance regime that differs from that prevailingtoday.

LIDs vary considerably in the nature and extent ofinteraction with physiography and vegetation. Inturn, the degree of interaction exerts a stronginfluence over the predictability of the resultingdamage patterns and the extent to which long-termpatterns in landscape-level vegetation emerge as aconsequence of LIDs. Tornadoes are unique in thenear absence of topographic control, as they mayimpact all slopes and aspects in their erratic path. Incontrast, the broad landscape-level patterns of hur-ricane impact appear to be explained by relatively

simple exposure relationships between slope, as-pect, and wind direction (Foster and Boose 1992;Boose and others 1994), and flood duration isdirectly related to elevation and drainage position(Sparks 1996). Topography and landform (for ex-ample, lake and wetland) distribution control themovement, intensity, and landscape pattern of fire(Heinselman 1996). Complexity emerges, however,in the case of multiple disturbance types associatedwith an LID, for example, gusts, tornadoes, ordownbursts that may accompany a hurricane, ormajor shifts that occur in wind direction during afire. Similarly, volcanic activity produces explosiveblasts, thermal waves, and landslides that may bevery sensitive to site exposure, whereas tephradeposition is independent of physiography.

Predisturbance vegetation may influence the be-havior of a disturbance, such as fire, and with littleexception influences the severity and type of dam-age. In particular, because of the ‘‘top heavy’’ natureof trees, and the lateral loading pressure generatedby many LIDs, there is a tendency for damage toincrease with tree age and height (Foster 1988b).Fire probability often increases with stand age dueto the general increase in fuel (Clark 1989; Heinsel-man 1973), but individual tree suceptibility todamage or mortality from fire often declines withtree size due to increasing bark thickness and aseparation of foliage from the ground, which re-duces the likelihood of crown fires. Specific damagepatterns following fire are determined by structuralcharacteristics of the trees and vegetation and mayvary widely within and between regions and foresttypes (Heinselman 1996).

LIDs differ in duration, ranging from minutes tohours for tornadoes, hurricanes, and (some) volca-nic eruptions, to months or even longer for somefloods, fires, and volcanic lava flows (Figure 2). Thelong duration of some fires enables fire changes inintensity and direction. Wind shifts may causecomplex fire patterns that do not follow topography,as a leeward slope on one day becomes a windwardslope on the next. In contrast, brief fires driven byone wind direction may generate narrow, linearpatterns. Flood duration often is correlated withseverity, as the primary cause of mortality is theextended lack of oxygen caused by long-lastingfloodwaters. Tornadoes, fires, and floods commonlyhave multiple initiations with numerous eventsoccurring simultaneously in relatively small areasunder favorable climatic conditions. Moreover, thelandscape patchiness that develops among torna-does and fires is due, in part, to their tendency tojump from place to place, by funnel skipping andspotting, respectively.


Considerable variation exists in the area andpattern of damage resulting from different LIDs(Figure 1). In general, however, large disturbancesdo not create large homogeneous patches of dam-age. Rather, due to the aforementioned interplay ofbiological and physiographic factors and distur-bance processes, the resulting damage tends to beheterogeneous in terms of internal patterns of mor-tality and landscape patterns of damage patches(Turner and Dale 1998). With the exception oftornadoes, undamaged patches are typically locatedwithin the perimeter of the disturbance, and there isa tendency for the overall pattern to be elongated inthe direction of damage flow.

The number, size, and distribution of unaffectedpatches within a severe disturbance are of interestbecause of their potential impact on a variety ofecosystem characteristics. Intense tornadoes typi-cally leave only severely impacted patches. Volca-noes, such as Mt. St. Helens, often leave few intactpatches, mostly on the perimeter of the blast zone.Fires and hurricanes often leave large intact areas,as controlled by firebreaks and windbreaks andvegetation structure. In general, although hurri-canes have the potential for impacting extremelylarge areas with severe damage, patches of interme-diate damage and broad areas with scattered canopygaps usually predominate within the damage area.The greatest number and extent of minimally dis-turbed patches revealed in our review was createdby the 1993 floods on the Mississippi River in which

broad forest areas were only thinned by extendedinundation.

BIOLOGICAL LEGACIES OF LARGE

INFREQUENT DISTURBANCES

Remnants of the previous forest ecosystem, that is,biological legacies, persist following a disturbanceregardless of whether the event is large or small,frequent or infrequent (Franklin and Halpern 1989).For example, following fires, a large amount ofunburned material remains, including seeds, largewood in the form of snags and coarse debris, andunderground plant parts that often replace theaboveground biomass that was burned. Similarly,following hurricanes, tornadoes, and floods, most ofthe biomass remains in redistributed accumulationsand is essential to the process of ecosystem recovery(Peterson and Pickett 1991; Cooper-Ellis and othersforthcoming). Of the various disturbances, volcaniceruptions have the potential of leaving the leastamount of residual material from the previousecosystem. The intense force of the blast, and thelarge amount of earth that is either moved orcovered with various kinds of ejecta, lead to adisturbance that is more severe than the hottest fireor the most intense windstorm. Of course, theeffects diminish as distance from the volcano in-creases. Notably, whereas fires, windstorms, andfloods move across the landscape according to thespread of heat, wind, and water, the force of volcaniceruptions is a combination of an intense pressure wavecombined with scorching and the subsequent flowingof various earthen materials (including water if alarge snowpack is present at the time).

The question remains whether or not LIDs lead todifferent legacies than smaller, frequent distur-bances. Several factors must be considered. First, ifthe disturbance is frequent enough, then organismscomprising the ecosystem will be adapted to survivethe disturbance and little change may occur. Ex-amples would include fires that occurred at 1- to5-year intervals in the native grasslands of NorthAmerica and floods of decadal intervals. In bothcases, changes in ecosystem structure and functionare so minor that a disturbance is hardly indicated.Of the LIDs examined for this article, the 1993floods on the Mississippi River appear to have hadthe least amount of damage and may be inappropri-ately included in the LID category, despite theirsevere effects on human structures. In contrast, thereturn interval of large fires, such as the 1988Yellowstone fires, or major windstorms and volcaniceruptions, is sufficiently long and the impact is sogreat that they almost invariably lead to clearchanges in ecosystem structure and function.

Figure 2. Comparison of the extent and duration of largeinfrequent disturbance by fire, flood, hurricane, tornado,and volcano.


We have not identified the size that is required fora disturbance to be classified as an LID, but unlikethe 1938 hurricane, the Yellowstone fires, the Ti-onesta tornado, or Mount St. Helens, some infre-quent disturbances affect just a few hectares. Thelandscape pattern of legacies from small distur-bances probably is of much less significance becauseno portion of the disturbed area is far removed fromthe adjacent, undisturbed portion of the ecosystem.The proximity of the undisturbed land leads to acomparatively rapid recovery. The shape of a distur-bance is another landscape variable that seemsespecially important in determining the role ofbiological legacies and patterns of ecosystem recov-ery. If a large disturbance can be characterized aslong, narrow, linear, or dendritic, as usually in thecase with windstorms and floods (including volcanicdebris flows some distance away from the blastzone), then ecosystem recovery will be compar-atively rapid because, as with small disturbances, allportions of the disturbed area will be in closeproximity to undisturbed remnants at the edges ofthe disturbance. Moreover, the amount of biologicalmaterial remaining after windstorms and floods islarger than after fires and some volcanic eruptions,where combustion occurs.

Slower rates of recovery can be expected if theshape of a disturbance is both large and circular, asoccurs following volcanic eruptions and some fires.Circular or oval disturbances seem to result fromeither an extraordinarily intense force, such as anisodiametric volcanic blast, or a disturbance of longduration, such as fires in western coniferous forestswhere the flames often burn first in one directionand then another as the wind shifts. Two otherfactors that surely contribute to oval-shaped burnsare extreme fuel flammability and the multiple

ignitions that occur during unusually dry years,such as in 1988 in the Yellowstone areas.

In a large disturbed area, the degree of patchinessis a factor influencing ecosystem recovery. Patchi-ness within the disturbed area probably increases asthe severity of the disturbance declines or as topo-graphic diversity increases. Long-lasting fires willburn some days with great intensity, leaving fewpatches as a legacy, but will burn more slowly onother days or during the night (because of coolertemperatures or less intense wind). The result is thateven fires that are perceived as highly dangerousand severe will leave unburned patches. Such patchi-ness is less likely to develop during a volcaniceruption, as the heat and intensity of the blastdestroys much of the ecosystem regardless of topo-graphic position, except on the fringes of the blastzone. For both fires and volcanic eruptions thatcreate broad disturbance patterns, patches that re-main can be considered as part of the biologicallegacy that contributes to the rate of ecosystemrecovery. The critical variable may not be the size ofa disturbance, but rather the mean distance of aseries of random points within a disturbed area to anundisturbed patch. Such patches can be expected toamplify greatly the other forms of biological legacythat inevitably remain (Table 2).

ENDURING IMPRINTS OF LIDS

ON FOREST ECOSYSTEMS

Over extended periods encompassing many epi-sodes of disturbance, landscape-level patterns ofdisturbance susceptibility, controlled by the interac-tion of the disturbance, physiography, and biota,may lead to long-lasting patterns of species andecosystem distribution that represent enduring im-

Table 2. Landscape Pattern Characteristics Resulting from Five Specific Large, InfrequentDisturbances (LIDs)

LIDCharacteristic

Mount St. HelensVolcanic Eruption

TionestaTornado

YellowstoneFires

Hurricane1938

MississippiRiver Floods

Year of occurrence 1980 1985 1988 1938 1993Duration of event 12 h 1 h 90 days 8 h ,90 daysTime of year 18 May 31 May May–October September June–AugustOverall shapea Isodiametric, broad Linear, narrow Broad Linear, broad Linear, narrowEdge transition Diffuse Abrupt Diffuse Diffuse DiffuseMinimally affected patches

Distribution Distant from crater None Throughout Ridges; near edges ThroughoutSeverely affected area

Proportion of trees killed 100% 50% 85%–100% 50% 65%–50%

aShape varies with the different causes, for example, the explosive blast at Mount St. Helens led to a broad, isodiametric shape, while glowing avalanches and debris flows led to amore liner shape (see Figure 1).


prints of disturbance (Whitmore 1974; Foster andothers 1999). This long-term landscape conse-quence of disturbance regimes has not been investi-gated extensively but may be a common effect ofsome LIDs due to their size and predictable distribu-tions at a landscape scale (Smith 1946; Swanson andothers 1988; Foster and Boose 1994). For fire,directionality of prevailing winds during thedroughty fire season and the distribution of physi-ographic firebreaks may lead to a well-developedlandscape pattern of frequency and intensity. Overrepeated episodes, the landscape variation in distur-bance regime may control forest distribution (Hein-selman 1973), organic matter accumulation andwetland development (Foster 1983), and soil andecosystem processes (Goodlett 1956; Viereck andothers 1983) that in turn may have importantfeedbacks to the disturbance process (Grimm 1984;Foster and King 1986). In regions prone to cata-strophic wind events, it has been suggested thatpersistent landscape-scale variation in site suscepti-bility may lead to important spatial variation in thefrequency and intensity of damaging winds thatover long periods may control canopy structure(Webb 1958; Lugo 1983), the distribution of under-story plants (A. Sabato, personal communication),the spatial pattern and characteristics of succes-sional and old-growth forests (Boose and others1994), and primary production, hydrology, andnutrient retention (Swanson and others 1988; Scat-ena and others 1997). In all cases, these effects arenot the consequence of individual events or suddenchanges but the legacies of long-term predictabilityin the landscape distribution of a LID that has led tomultiple outcomes and repeated patterns on similarsites.

ACKNOWLEDGMENTS

The development of the ideas in this report ben-efited greatly from the input of all members of theLIDs workshop held at the National Center ofEcological Analysis and Synthesis and supported bythe National Science Foundation. Helpful com-ments were provided by M. Turner, V. Dale, M.Tegner, W. Baker, P. Harcombe, T. Veblen, C. Can-han, and members of the Harvard Forest ecologygroup. Additional support was provided by the A.W. Mellon Foundation and Harvard Forest LongTerm Ecological Research program (NSF-BSR-9411764).

REFERENCES

Agee JK. 1993. Fire ecology of Pacific Northwest forests. Washing-ton (DC): Island.

Anthes RA. 1982. Tropical cyclones: their evolution, structureand effects. American Meteorological Society, Boston: Meteo-rological Monographs 19.

Arno SF, Reinhardt ED, Scott JH. 1993. Forest structure andlandscape patterns in the subalpine lodgepole pine type: aprocedure for quantifying past and present conditions. Wash-ington (DC): USDA Forest Service Intermountain ResearchStation; General Technical Report INT-924.

Baker VR, Kochel RC, Patton PC, editors. 1988. Flood geomor-phology. New York: Wiley.

Bluestein HB. 1996. Tornadoes. In: Schneider SH, editor. Encyclo-pedia of climate and weather. Oxford: Oxford University Press,p 764–8.

Boose ER, Foster DR, Fluet M. 1994. Hurricane impacts totropical and temperate forest landscapes. Ecol Monogr 64:369–400.

Bormann H, Likens G. 1979. Pattern and process of a forestecosystem. New York: Springer-Verlag.

Bornette G, Amoros C. 1997. Disturbance regimes and vegetationdynamics: role of floods in riverine wetlands. J Vegetation Sci7:615–22.

Bowden R, Castro MC, Melillo JM, Steudler PA, Aber JD. 1993.Fluxes of greenhouse gases between soils and the atmospherein a temperate forest following a simulated hurricane blow-down. Biogeochemistry 21:61–71.

Carlton GC, Bazzaz F. 1998. Resource congruence and forestregeneration following an experimental hurricane blowdown.Ecology 79:1305–19.

Christensen NL, Agee JK, Brussard PF, Hughes J, Knight DH,Minshall GW, Peek JM, Swanson FJ, Thomas JW, Wells S,Williams S, Wright HA. 1989. Interpreting the Yellowstonefires of 1988. BioScience 39:678–86.

Clark JS. 1989. Ecological disturbance as a renewal process:theory and application to fire history. Oikos 56:17–30.

Cooper-Ellis S, Foster D, Carlton G, Lezberg A. Forest responseand reorganization following experimental hurricane. Ecology.Forthcoming.

Covich A, Crowl TA, Johnson SL, Pyron M. 1997. Distributionand abundance of tropical freshwater shrimp along a streamcorridor: response to disturbance. Biotropica 28:484–92.

Dunn CP, Guntenspargen GR, Droney JR. 1983. Catastropic winddisturbance in an old-growth hemlock–hardwood forest, Wis-consin. Can J Bot 61:211–7.

Dunn GE, Miller BI. 1964. Atlantic hurricanes. Rev. ed. BatonRouge: Louisiana State University Press.

Foster DR. 1983. The history and pattern of fire in the borealforest of southeastern Labrador. Can J Bot 61:2459–71.

Foster DR. 1988a. Disturbance history, community organizationand vegetation dynamics of the old-growth Pisgah Forest,southwestern New Hampshire, USA. J Ecol 76:105–34.

Foster DR. 1988b. Species and stand response to catastrophicwind in central New England, USA. J Ecol 76:135–51.

Foster DR, Aber J, Bowden R, Melillo J, Bazzaz F. 1997.Temperate forest response to catastrophic disturbance andchronic stress. BioScience 47:437–45.

Foster DR, Boose ER. 1992. Patterns of forest damage resultingfrom catastrophic wind in central New England, USA. J Ecol80:79–98.

Foster DR, Boose ER. 1994. Hurricane disturbance regimes intemperate and tropical forest ecosystems. In: Coutts M, editor.Wind effects on trees, forests and landscapes. New York:Cambridge University Press. p 305–39.


Foster DR, Fluet M, Boose ER. 1999. Natural or human distur-bance: landscape dynamics of the tropical forests of PuertoRico. Ecol Appl. Forthcoming.

Foster DR, King GA. 1986. Vegetation pattern and diversity inS.E. Labrador, Canada: Betula papyrifera (birch) forest develop-ment in relation to fire history and physiography. J Ecol74:465–83.

Frank WM. 1977. The structure and energetics of the tropicalcyclone. I. Storm structure. Mon Weather Rev 105:119–35.

Franklin JF, Halpern CB. 1989. Influence of biological legacies onsuccession. In: Ferguson DE, Morgan P, Johnson FD, editors.Proceedings: land classifications based on vegetation—applica-tions for resource management. Washington (DC): USDAForest Service, Intermountain Research Station. p 54–5.

Franklin JF, MacMahon JA, Swanson FJ, Sedell JR. 1985.Ecosystem responses to the eruption of Mount St. Helens. NatGeogr Res 1:198–216.

Fujita TT. 1971. Proposed characterization of tornadoes andhurricanes by area and intensity. Chicago: Department ofGeophysical Sciences, University of Chicago; Satellite andMesometeorology Research Project Paper 91.

Fujita TT. 1987. U. S. Tornadoes: part one—70-year statistics.Chicago: University of Chicago Satellite and MesometeorologyResearch Project.

Gara RT, Little WR, Agee JK, Geiszler DR, Stuart JD, Driver CH.1984. Influence of fire, fungi and mountain pine beetles ondevelopment of a lodgepole pine forest in south-central Or-egon. In: Lodgepole pine: the species and management. p35–7.

Goodlett JC. 1956. Vegetation and surficial geology. In: Surficialgeology and geomorphology of Pottern County, Pennsylvania.

Gray WM. 1990. Strong association between West Africanrainfall and US landfall of intense hurricanes. Science 249:1251–6.

Grazulis TP. 1993. Significant tornadoes, 1680–1991. St. Johns-bury (VT): Environmental Films.

Grimm EC. 1984. Fire and other factors controlling the BigWoods vegetation of Minnesota in the mid-nineteenth cen-tury. Ecol Monogr 54:291–311.

Harmon ME, Franklin JF, Swanson FJ, Sollins P, Gregory SV,Lattin JD, Anderson NH, Cline SP, Aumen NG, Sedell JR,Linekaemper GW, Cromack K Jr, Cummins KW. 1986. Ecologyof coarse woody debris in temperate ecosystems. In: Mac-Fadyen A, Ford ED, editors. Advances in ecological research.Orlando (FL): Academic. p 133–302.

Heinselman ML. 1973. Fire in the virgin forests of the BoundaryWaters Canoe Area, Minnesota. Quat Res (NY) 3:329–82.

Heinselman ML. 1996. The boundary waters wilderness ecosys-tem. Minneapolis: University of Minnesota Press.

Hemstrom MA, Franklin JF. 1982. Fire and other disturbances ofthe forests in Mount Rainier National Park. Quat Res (NY)18:32–51.

Knight D. 1987. Parasites, lightning, and the vegetation mosaic inwilderness landscapes. In: Turner MG, editor. Landscape hetero-geneity and disturbance. New York: Springer-Verlag. p 59–83.

Lugo A. 1983. Ecological aspects of catastrophes in CaribbeanIslands. Acta Cient 2:24–31.

Lugo AE, Scatena FN. 1989. Ecosystem properties of the LuquilloForest of Puerto Rico. In: Proceedings of the ITF 50th anniver-sary symposium.

Lynch JF. 1991. Effects of Hurricane Gilbert on birds in a drytropical forest in the Yucatan Peninsula. Biotropica 23:488–96.

Melanson GP, Butler DR. 1984. Avalanche paths as fuel breaks:implications for fire management. J Environ Manage 19:229–38.

Motzkin G, Foster D, Allen A, Harrod J, Boone R. 1996.Controlling site to evaluate history: vegetation patterns of aNew England sand plain. Ecol Monogr 66:345–65.

Naka K. 1982. Community dynamics of evergreen broadleafforests in southwestern Japan. 1. Wind damaged trees andcanopy gaps in an evergreen oak forest. Bot Mag Tokyo95:385–99.

Paine RT, Tegner MJ, Johnson EA. 1998. Compounded perturba-tions yield ecological surprises. Ecosystems 1:535–545.

Patterson WA, Foster DR. 1990. ‘‘Tabernacle Pines’’: the rest ofthe story. J For 89:23–5.

Peterken GF. 1997. Natural woodland: ecology and conservationin northern temperate regions. New York: Cambridge Univer-sity Press.

Peterson CJ, Pickett STA. 1991. Treefall and resprouting follow-ing catastrophic windthrow in an old-growth, hemlock–hardwood forest. For Ecol Manage 42:205–17.

Peterson CJ, Pickett STA. 1995. Forest reorganization: a casestudy in an old-growth forest catastrophic blowdown. Ecology76:763–74.

Rogers P. 1996. Disturbance ecology and forest management: areview of the literature. Washington (DC): USDA ForestService General Technical Report INT-GTR-336.

Scatena F, Larsen M. 1991. Physical aspects of Hurricane Hugo inPuerto Rico. Biotropica 23:317–23.

Scatena FN, Maya S, Estrada C, Chinea J. 1997. The first fiveyears in the reorganization of aboveground biomass andnutrient use following Hurricane Hugo in the Bisley Experimen-tal Watersheds, Luquillo Experimental Forest, Puerto Rico.Biotropica 28:424–41.

Shaw WB. 1983. Tropical cyclones: determinants of pattern andstructure in New Zealand’s indigenous forests. Pac Sci 37:405–14.

Simpson RH, Riehl H. 1981. The hurricane and its impact. BatonRouge: Louisiana State University Press.

Smith DM. 1946. Storm damage in New England forests [MSthesis]. New Haven: Yale University. 173 pp.

Sparks RE. 1996. Ecosystem effects: positive and negative out-comes. In: Changnon SA, editor. The great flood of 1993. (CO):Westview. p 132–62.

Stephens EP. 1956. The uprooting of trees: a forest process. SoilScience Society proceedings. p 113–6.

Swanson FJ. 1981. Fire and geomorphic processes. In: MooneyHA, editor. Proceedings of the conference on fire regimes andecosystems. Washington (DC): USDA General Technical Re-port.

Swanson FJ, Franklin JF, Sedell JR. 1990. Landscape patterns,disturbance, and management in the Pacific Northwest, USA.In: Zonneveld IS, Forman RTT, editors. Trends in landscapeecology. New York: Springer-Verlag. p 191–213.

Swanson FJ, Kratz TK, Caine N, Woodmansee RG. 1988. Land-form effects on ecosystem patterns and processes. BioScience38:92–8.

Turner MG, Dale VH. 1998. Comparing large, infrequent distur-bances: what have we learned? Ecosystems 1:493–496.

Veblen TT, Kitzberger T, Lava A. 1992. Disturbance and forestdynamics along a transect from Andean rain forest to Patago-nian shrubland. J Veg Sci 8:507–20.


Viereck LA, Dyrness CT, Van Cleve K, Foote MJ. 1983. Vegetationsoils and forest productivity in selected forest types in interiorAlaska. Can J For Res 13:703–20.

Webb LJ. 1958. Cyclones as an ecological factor in tropicallowland forest, North Queensland. Aust J Bot 6:220–8.

Whelan RJ. 1995. The ecology of fire. Cambridge: CambridgeUniversity Press.

Whitmore TC. 1974. Change with time and the role of cyclonesin tropical rain forest on Kolombangara, Solomon Islands.Oxford: Holywell; Commonwealth Forestry Institute Paper 46.

Williams GP, Wolman MG. 1984. Downstream effects of dams onalluvial rivers. Washington (DC): United States GeologicalSurvey; Professional Paper 1286.

Wood DM, Morris WF. 1990. Ecological constraints to seedlingestablishment on the Pumice Plains, Mount St. Helens, Wash-ington. Am J Bot :1411–8.

Yih K, Boucher D, Vandermeer J, Zamora N. 1991. Recovery ofthe rain forest of southeastern Nicaragua after destruction byHurricane Joan. Biotropica 23:106–13.

Zimmermann J, Willig MR, Walker LR, Silver WL. 1997. Distur-bance and Caribbean ecosystems. Biotropica 28:414–23.


landscape patterns and legacies resulting from large ... · 1980 eruption of mount st. helens, and...

Documents