11111111 Emulating Natural

Forest LandscapeDisturbances

Concepts and Applications

Edited by

AJITH H. PERERA

LISA J. BUSE , AND

MICHAEL G. WEBER

0- COLUMBIA UNIVERSITY PRESSNEW YORK

DISCLAI M ER

The opinions expressed in this text are those of the individualauthors and do not necessarily reflect offcial views orpolicies of the funding agencies or the institutions withwhich the authors are affliated.

Columbia University Press

Publishers since 1893

New York Chichester, West Sussex

If 2004 Columbia University PressAll rights reserved.

LIBRARY OF CONGRESS CATALOGING-IN-PUBLICATION DATA

Emulating natural forest landscape disturbances: conceptsand applications / edited by Ajith H. Perera , LisaJ. Buse,and Michael G. Weber.

p.

cm.Includes bibliographical references (p. ).ISBN 0-231- 12916-5 (cloth: alk. paper)1. Forest management-United States. 2. Forest

ecology-United States. 3. Ecological disturbances-United States. 4. Forest management-Canada. 5. Forestecology-Canada. 6. Ecological disturbances-Canada.1. Perera, Ajith H. II. Buse, LisaJ. III. Weber, Michael G.

SD143.E52004634. dc22 2003068813

Columbia University Press books are printed on durable andacid-free paper.

Printed in the United States of America

10 9 8 7 6 5 4 3 2

Ii'

Using Criteria Based on the Natural Fire Regime to

Evaluate ForestManagement in the Oregon Coast Range

of the United States

CHAPTER 12

Dii1111

fir

th,Jig

na-vol

MICHAEL C. WIMBERLY, THOMAS A. SPIES , and ETSUKO NONAKA

Forest landscapes in the Oregon Coast Range havechanged considerably since settlers first arrivedin the mid-nineteenth century. Clearing of

forestsfor agriculture and development, ignition of ex-tensive forest fires by settlers and loggers

, andconversion of natural forests to managed planta-tions have all contributed to the fragmentationof late-successional forests (Ripple et al. 2000;Wimberly et al. 2000). These rapid and wide-spread changes have led to concern for popula-tions of native species, particularly those associ-ated with old-growth forests and aquatic habitats.Some threatened or endangered species such asthe northern spotted owl (Strix occidentalis cau-rina (Merriam)) have been the subject of inten-sive research, resulting in the development ofdetailed conservation plans (Thomas et al.

1990).There are many other species, however, forwhich comprehensive infolmation on habitat

requirements and demography is not available.Because of limited knowledge and resourcesdeveloping and implementing individual man-agement plans for every species is not feasible.Instead , the coarse filter approach has been pro-posed as an alternative for conserving native bio-diversity at the landscape scale (Noss

1987). Thismethod entails preselving ecological diversity atthe community level based on the assumptionthat a representative array of habitats wil besufficient to meet the needs of most species (seeThompson and Harestad, this volume , chapter 3).Studies of landscape dynamics under naturaldisturbance regimes can playa critical role inthe development of coarse-

fiter conservationstrategies (Landres et al. 1999; Swetnam et al.1999). Disturbance-based ecological,assessmentsare grounded on the assumption that historical

146

CUI

scacorfirethewit

landscapes were subjected to both natural and

human disturbances for milennia prior to the ar-rival of Europeans and other nonindigenous set-tlers. Despite these perturbations

, pre-Europeansettlement landscapes sustained the native speciesthat we currently wish to preserve. Comparisonsbetween present-day and pre-European sette-ment disturbance regimes can therefore serve asindicators of the potential for conserving speciesand sustaining ecosystem processes in the mod-ern landscape. Analyses of natural disturbance

regimes and the forest patterns they generate canalso provide targets for future landscape restora-tion efforts and can suggest approaches to main-taining habitat diversity in dynamic landscapes.Because landscapes influenced by natural distur-bance regimes seldom approach a steady stateequilbrium, historical patterns must be charac-terized as a distribution of possible system statesrather than as an arbitrary "snapshot" taken at asingle point in time (Sprugel 1991).

Fire was a ubiquitous landscape-scale distur-bance in the Pacific Northwest for thousands ofyears prior to the arrival of settlers

(Agee 1993;see also Hessburg et al. , this volume , chapter 13).Because fire has been almost entirely Suppressedin the Oregon Coast Range for the past 50 yr

, ourability to observe and study such fundamentalprocesses as fire ignition, fire spread , and fuelconsumption is limited. However

, paleoecolog-ical, dendroecological, and historical researchhas yielded considerable information about thefrequencies, sizes , and severities of past fires(Teensma et al. 1991; Impara 1997; Long et al.1998; Weisberg and Swanson 2003). Thereforewe have adopted a primarily historical and em-pirical approach to characterizing and modeling

(2)patJtlen(3) c

bastin tworecOI

Stud

The23,Pacijtheand1EleviIOGC

is chsteep

Forest

both natural andlOia prior to the ar-

onindigenous set-)ns pre-Europeanj the native speciesrve. ComparisonsEuropean settle-therefore serve as

onserving species:esses in the mod-tural disturbancethey generate canandscape restora-)roaches to main-lamic landscapes.by natural distur-ch a steady state; must be charac-ible system statespshot" taken at a91).:ape-scale distur-

for thousands oftlers (Agee 1993;lme , chapter 13).tirely suppressedIe past 50 yr, ourch fundamental;pread, and fuel

, paleoecolog-torical researchation about thees of past fires)97; Long et al.)03). Thereforetorical and em-g and modeling

fire regimes in this chapter. For the purposes ofour study, we assumed that the "natural" fireregime is equivalent to the historical fire regimethat existed prior to settlement. Fires initiated byJightning, as well as those ignited by NativeAmericans, are all considered to be part of thisnatural fire regime (Suming and Perera, thisvolume, chapter 4).

The goal of this research was to reexaminecurrent forest management practices and land-scape patterns in the Oregon Coast Range byconsidering them in the context of the naturalfire regime. Specific objectives were to (I) contrastthe frequencies , sizes , and effects of historical fireswith those of forest management disturbances;(2) examine differences in the abundance andpattern of seral stages between pre-European set-tlement and present-day forest landscapes; and(3) assess the potential for applying managementbased on pre-European settlement fire regimesin the Oregon Coast Range within the frame-work of present-day forest conditions and socio-economic constraints.

predominantly well-drained Andisols and Incep-tisols (based on the Soil Science Society of Amer-ica Classification System). Parent materials aremostly marine sandstones and shales, along withsome basaltic volcanics and related intrusives.The climate is generally wet and mild , with mostprecipitation fallng between October and March.Precipitation is highest and summer tempera-tures are lowest near the coast, resulting in lowmoisture stress during the growing season andhigh forest productivity. Decreasing precipita-tion and increasing temperature with increasingdistance from the coast create a predominantlywest -to-east gradient of increasing moisture stress(color plate 16).

Major coniferous species include Douglas-fir(Pseudotsuga menziesii (Mirb.) Franco) and west-ern hemlock (Tsuga heterophylla (Raf.) Sarg.with Sitka spruce (Picea sitchensis (Bong.) Carr.prevalent along a narrow coastal strip. Hard-woods, including red alder (Alnus rubra Bong.and bigleafmaple (Acermacrophyllum Pursh), areoften found in mixed stands with young conifersand dominate many riparian areas. Long-livedconifers and a favorable climate combine to pro-duce some of the largest accumulations of live-tree biomass in the world (Waring and Franklin1979). The presence of large live and dead trees(often roo cm in diameter at breast height(dbh)), along with a diverse multilayered canopyand spatial heterogeneity created by canopygaps, are characteristics of Pacific Northwest old-growth forests (Franklin et al. 1981). These fea-tures tyically require 200 yr or more to develop

FIGURE 12. 1. Maps of the OregonCoast Range study area. (a) Bound-ary of the study area , with majorhistorical fires of the past 200 yr.(b) Location of the study areawithin the Pacific Northwestregion. (c) Location ofthe studyarea within North America.

NATURAL FIRE REGIME AND EVALUATING FOREST MANAGEMENT 147

Study Area

The Oregon Coast Range encompasses more than23,000 km in western Oregon, bounded by thePacific Ocean to the west, the Coquile River tothe south, the Wilamette Valley to the eastand the Columbia River to the north

(figue 12.1).Elevations range from sea level to more thanrooo m at the highest peaks. The physiographyis characterized by highly dissected terrain withsteep slopes and high stream densities. Soils are

following a stand-replacing disturbance, al-though old-growth structure may develop morequickly following partial disturbances that leavea cohort of live remnant trees.

Major groups of forest landowners in the Ore-gon Coast Range include private industry, pri-vate nonindustrial owners, the state of Oregonand the federal government (the U.S. Depart-ment of Agriculture Forest Service and the Bu-reau of Land Management). Landowner goalsforest management strategies, and regulatoryconstraints vary considerably among theseownership classes. Private industrial ownershipscomprise 38% of the study area, concentrated inblocks in the northern, central, and southernportions of the Coast Range (color plate 16b).Private industrial lands are managed primarilyfor wood production within the regulatory con-straints imposed by the Oregon State Forest Prac-tices Act (Oregon Department of Forestry 2002).The predominant management practice on pri-vate industrial land is clearcutting, with rotationsof 30 to 50 yr. Private nonindustrial ownershipscover 24% of the study area, primarily along theWilamette Valley margin and in the large rivervalleys. Private nonindustrial owners operatewithin the same regulatory framework as privateindustry, but have a broader range of goals anduse a wider variety of management practices.

Federal lands managed by the U.S. Departmentof Agriculture Forest Service and the Bureau ofLand Management also occupy substantial areasof the Coast Range (rr and 14% of the study arearespectively). A significant portion of the landcontrolled by the Bureau of Land Managementis interspersed with private industrial land ina checkerboard pattern (color plate 16b). Thefederal lands are presently managed under theNorthwest Forest Plan for primarily ecologicalgoals related to the preservation and restorationof old-growth forests , conservation of threat-ened and endangered species , and protectionof aquatic ecosystems (Forest Ecosystem Man-agement and Assessment Team 1993). Timberharvests are restricted in a large network of late-successional and riparian reserves, and harvestsin the remaining matrix lands are required toleave residual live trees , snags, and downed trees.State forest lands cover 12% of the study area.Current plans for the state forests propose activemanagement for timber production, protectionof the forest's health, and species conservationby combining retention of large trees , snagslogs , and other habitat elements following tim-ber harvests with long rotations to create a range

148 MICHAEL C. WIMBERLY etai.

of forest structures across the landscape (OregonDepartment of Forestry 2001). fir

Methods

Comparison of Disturbance Regimes

Historical fire regimes in the Coast Range werecharacterized through a literature review andan analysis of published data. Paleoecological re-search based on high-resolution charcoal analy-ses provided long-term fire history informationfor the watersheds surrounding Little Lake (Longet al. 1998) in the central Coast Range and TaylorLake (Long and Whitlock 2002) in the north-western Coast Range. Fire frequency data fromthese studies were summarized for different timeperiods as mean fire-return intervals , defined asthe mean number of years between fires at a par-ticular location (Agee 1993). Dendroecologicalresearch based on tree-ring and fire-scar datafrom a I375- study area in the central CoastRange provided a more detailed picture of ratesand patterns of burning over the past 500 yr (lm-para 1997). Using published data from Imparastudy, we computed the natural fire rotation toserve as a metric of fire frequency for the coastal-interior and Wilamette Valley margin portionsof the study area. The natural fire rotation repre-sents the total number of years required to burna particular landscape (Heinselman 1973), andis computed as:

si,ti,

Ie;

YEARS(12.NFR=

NFIRESFSIZE TAREA

;=1

where NFR is the natural fire rotation YEARS

the length of the fire history record NFIRESis the total number of fires in the record FSIZEthe size of the ith fire in the record , and TAREA

is the total area of the landscape.We chose this metric to measure fire frequency

because it does not depend on a particular statis-tical model of fire frequency and therefore doesnot assume that the fire regime remains constantover time and space (Fall 1998). However, theestimated natural fire rotation wil be biased up-ward if evidence of the oldest fire events is over-written by more recent fires. We therefore usedan alternative method, adapted from Morrisonand Swanson (1990), to correct for our reducedabilty to sample the oldest fires:

ti,

NFR=YEARS

(12.NFIRES

(FSIZE RAREA

iscape (Oregon where RAREA is the area available to record eachfire, which is computed as the total area that didnot experience a stand-replacing fire more re-cently than the occurrence of the recorded fire.

Fire-size distributions were computed usingestimates of fire size from historical maps offorest vegetation in 1850 , 1890 , 1920 , and 1940

(Teensma et a1. 1991). Most of the fires repre-sented in these maps occurred after settlementof the study area had begun, but before the es-tablishment of an effective fire suppression pro-gram in the latter half of the twentieth century.Human influences and a warm climate bothcontributed to widespread fires between 1850

and 1940 (Weisberg and Swanson 2003), and weassumed that the frequency distribution of firesizes from this period was generally representa-tive of the pre-European settlement fire regime.These historical fire regimes were contrastedwith modern disturbance regimes measured in aremote-sensing study of landscape change from197 to 1995 (Cohen et al. 2002). A map of thelocations and sizes of clearcuts that occurredduring this time was used to determine the sizedistribution of the disturbances caused by tim-ber harvesting and to calculate a harvest rotationusing the same approach as used in equation 12.Because the time period of the study of land-

scape change (23 yr) was less than the harvestrotations commonly used in the Coast Range

(~3 yr), we assumed that no clearcuts had beenoverwritten by subsequent disturbances.

.ast Range wereure review andeoecological re-charcoal analy-

)ry informationjttle Lake (Longange and Taylor

:) in the north-Lency data from)r different timervals , defined asen fires at a par-endroecologicalld fire-scar datahe central CoastI picture of rates~past500yr (Im-ta from ImparaI fire rotation toy for the coastal-margin portionsre rotation repre-required to burnlman 1973), and

(12. Simulation of Pre- EuropeanSettlement Landscape Dynamics

Pre-European settlement landscape dynamicswere simulated using the Landscape Age-Class

Dynamics Simulator (LADS), a spatial simula-tion model that predicts the initiation spreadand effects of fires (Wimberly 2002). Multiplespatial data layers delineated simulation areaboundaries , topography, and major climaticzones as raster data layers with a 9-ha resolution

x 300-m2 cells). The model used a statistical

approach to simulate fire regimes, and modeledthe frequencies , sizes, and severities of individ-ual fires as probabilty distributions derived fromempirical data (Gardner et al. 1999). Differentprobabilty distributions were derived for thecoastal-interior and Wilamette Valley margin cli-matic zones to reflect broad-scale spatial trendsin the study area s fire history (see figure 12.Ia).Fire ignition and spread were modeled using astochastic cellular-automata algorithm that al-lowed fire to occur more frequently on more

)tation YEARS

record, NFIRES

le record, FSIZE

cord, and TAREA

)e.

ure fire frequencya particular statis-nd therefore doesremains constant

18). However, thewil be biased up-

fire events is over-Ne therefore used2d from Morrisonct for our reduced2S:

(12.

susceptible landforms and in stands with highlevels of fuel accumulation. Stand dynamics weresimulated using an age-driven model of forestcohorts, which assumed that structural diversityrecovered more rapidly after moderate-severityfires than after stand-replacing fires. Forest vege-

tation was mapped using five structure classes

(table 12.1): early successional , young, matureold growth (early transition), and old growth(late transition-shiftng mosaic). For the pur-poses of our analysis and to permit comparisonwith the present-day landscape , the two old-growth classes were combined into a single struc-ture class. Additional details on the develop-ment, parameterization, and sensitivity analysisof the LADS model are provided by Wimberly(2002).

Previous research has demonstrated that thedisturbance regimes and landscape dynamicsof the Coast Range are scale dependent. Pre-European settlement landscape variabilty mustbe studied over relatively large areas (~IOO,OOO

ha) and long time intervals (~500 yr) because theoccurrence of occasional large fires makes land-scape dynamics unpredictable at finer scales(Wimberly et al. 2000). Our simulations of land-scape dynamics therefore encompassed thespatial extent ofthe Coast Range (approximately23,000 km ) and were carried out based on thefire regime that characterized the IOOO-yr periodprior to settlement. This period was deemed mostrelevant to present-day management questionsand was found in a previous study to representan era in which climate , fire regimes , and species

composition were generally similar to those ofthe present day (Wimberly 2002). The model wasrun for a total of 50 000 simulation years follow-ing a rooo-yr initialization period. Landscapepatterns were output as digital maps at 200-intervals, generating 250 sample landscapes.Running the model for 50 000 simulation yearsdoes not imply that the simulated fire regimewas representative of the actual 50 000 yr priorto settlement. Instead, the long simulation

period was needed to generate a large numberof independent landscapes that represented therange of variabilty that could have occurredover the rooo-yr time frame of the model.

The spatial pattern of each sample landscapewas summarized by computing the total areaand largest patch size for each structure class.In addition, we computed a spatial index thatmeasured the relative isolation of each structureclass from the other classes. In equation 12.3, thisisolation index was calculated for a given

NATURAL FIRE REGIME, AND EVALUATING FOREST ' MANAGEMENT 149

TABLE 12. 1. Classification of Forest Patches into Structure Classes based on AGE (Time Since StandInitiation) and TFIRE (Time Since Last Fire)

AGE (yr) TFIRE (yr) Structure Class Description

Any 0(30 Early successional Open canopy to dense , closed canopy with high tree densities.May have residual live trees , snags , and downed trees from theprefire stand.

;,30 30- Young Closed canopy with low levels of understOIY tree regenerationshrubs , and herbs. May have residual live trees , snags , anddowned trees from the prefire stand.

80-200 80-200 MatuIe Canopy gaps allow light to reach the forest floor, faciltatingreestablishment of the understory layer. Typically has lowvolumes of dead wood compared with older and younger ageclasses.

;,200 80-500 Old growth (early Characterized by large living trees (;,100 cm dbh);transition) accumulations of large snags and downed trees; a diverse

multilayered canopy dominated by shade-tolerant trees in themid-canopy layers; and a heterogeneous spatial pattern ofcanopy gaps and forest patches in various age classes.

;,500 ;,80 Old growth (late Similar to the old growth (early transition) phase, but withtransition- deCIeasing numbers of living remnant trees from theshifting mosaic) establishment phase and increasing dominance of canopy-

gap dynamics.

Notes: AGE = TFIRE for even-aged stands, but AGE;, TFIRE for stands that have had one or more fires of moderate sever-ity. See Wimberly et al. (2000) for a detailed explanation of this classification scheme. Descriptions of stand structureclasses are from Spies and Franklin (r991) and Spies (1997).

structure class (s) by first computing the distancesin km from every cell (i) in the raster map

to its nearest neighbor

(j)

belonging to class s.The isolation index for structure class s was thencomputed as:

Lcr. d..i=l I

cr,i=l l

(12.

where N is the total number of cells in the sim-ulated landscape , cri = 0 for cells belonging tostructure class s, and = I for cells that do notbelong to structure class s. This index was similarto the GISfrag index proposed by Ripple et al.(1991), except that distance values of 0 were notincluded in our computation. The isolation in-dex computed for a particular structure class re-flected the mean distance from each cell that wasnot a member of that structure class to its nearestneighboring cell that was a member. For exam-ple, computing a low isolation index for the old-growth class would indicate that forests in otherclasses tended to be located in close proximity toold-growth patches , suggesting a high potentialfor dispersal of organisms from old-growth refu-

150 MICHAEL C. WIMBERLY etai.

gia into the younger forests. In contrast, a highisolation index computed for old-growth forestwould suggest a Jimited potential for organismsto disperse from old-growth refugia into the SUI-

rounding landscape.Simulations of the historical landscape were

contrasted with a 1996 map of forest vegetationpatterns derived from forest inventory plotsLandsat TM imagery, and GIS layers describingownership, topography, and climate (Ohmannand Gregory 2002). This map provided a set ofcompositional and structural measurements foreach forested patch , including an estimate ofstand age. Patches in the 1996 map were reclas-sified into the same age-based structure classesused in the simulation model. The 1996 vegeta-

tion map initially had a 25-m spatial resolution,and was rescaled to a 300-m resolution to matchthe resolution of the simulation model's out-put. The total area, largest patch size , and isola-tion index were computed for each of the fourstructure classes based on the 1996 map. Thesevalues were compared with the probabiltydistributions of the same indices derived fromsimulation of the pre-European settlement fireregimes.

Result

Compo

The p

have Ileast 1

ever, Iably a

000climaclimaltIe LaLake)coole)tween2750 JValue:Lake

betwe

(LongUsi)

a natlCoasttlemeldata 1

of apfporticWilaIarea. -were sand

ducedsubsecsimul,causefor thclose 1

pOItecLong,tionsshortea consforests

Nat1the sel

ablybwith iJ1997).flecteddurinson 2C

creaselfire suing 33

Results

Comparison of Disturbance Regimes

densities.ees from the

The paleoecological records showed that fireshave burned in the central Coast Range for atleast the past 9000 yr (Long et al. 1998). How-ever, mean fire-return intervals varied consider-ably as climate changed over this epoch. From000 to 6850 yr before the present (YBP), the

climate was warmer and drier than the presentclimate and the mean fire-return interval at Lit-tle Lake averaged no yr. The interval at LittleLake lengthened over time as climate becamecooler and more humid, averaging 160 yr be-tween 6850 and 2750 YBP and 230 yr between275 YBP and the present day (Long et al. 1998).Values of the mean fire-return interval at TaylorLake showed similar trends, averaging 140 yrbetween 4600 and 2700 YBP and increasing to

yr between 2700 YBP and the present day(Long and Whitlock 2002).

Using equation 12. , Impara (1997) reporteda natural fire rotation of 452 yr for the centralCoast Range over the 367-yr period prior to set-tlement in 1845. Wimberly (2002) used the samedata to compute natural fire rotation valuesof approximately 200 yr for the coastal-interiorportion of the study area and roo yr for theWilamette Valley margin portion of the studyarea. This calculation included only fires thatwere severe enough to initiate tree regenerationand used equation 12.2 to correct for bias intro-duced by the erasure of evidence of older fires bysubsequent burns. We used these values in oursimulations of historical landscape dynamics be-cause the 200-yr natural fire rotation computedfor the coastal-interior region was reasonablyclose to the mean fire-return interval values re-ported for the past rooo yr (Long et al. 1998Long and Whitlock 2002). Running the simula-tions with natural fire rotation values at theshorter end of the range of estimates provideda conservative estimate of the amount of olderforests in the pre-European settlement landscape.

Natural fire rotation decreased to 78 yr duringthe settlement period from 1845 to 19IO , prob-ably because of the warming climate combinedwith increased fire ignitions by settlers (Impara1997). This increase in the rate of burning re-flected a regional trend of increased fire frequencyduring the settlement era (Weisberg and Swan-son 2003). In contrast, natural fire rotation in-creased considerably in the twentieth century as

fire suppression became more effective, averag-ing 335 yr between 19IO and 1994 (Impara 1997).

enerationags , and

:il ta ting'has lowTOunger age

diverseIt trees in theattern of;ses.

utwiththeIf canopy-

oderate sever-

tand structure

trast, a highrowth forestIr organismsinto the sur-

dscape were;t vegetationntory plotsS describing:e (Ohmannided a set oflIements forestimate ofwere reclas-cture classes

1996 vegeta-II resolution,on to matchnodel' s out-

, and isola-1 of the four, map. Theseprobabilty

lerived fromttlement fire

Stand-replacing fires burned only 0.06% of theCoast Range between 1972 and 1995 (Cohen et al.2002). If this rate of burning is extrapolated overlong time periods, it is equivalent to a natural firerotation of greater than 1600 yr.

Most of the fires reconstructed by Impara(I997)conformed to one of two distinctive patterns.Widespread fires were large burns that encom-passed both the coastal-interior and WilametteValley margin climate zones and were predom-inantly stand-replacing disturbances. Histori-cally documented examples of such events (Loyet al. 1976) include the Nestucca fire of 1848(120 000 ha); the Siletz fire of 1849 (325,000 ha);the Yaquina fire of 1853 (195,000 ha); the CoosBay fire of 1868 (n8 000 ha); and the Tilamookfires of 1933 (96,000 ha), 1939 (76 000 ha), and1945 (72 000 ha). ln contrast, WilametteValleymargin fires were smaller, predominantly mod-erate- and low-severity burns that left significantcohorts.of remnant live trees (Impara 1997). Fire-size data from reconstructed forest age mapsfor the Oregon Coast Range in 1850 , 189and 1940 (Teensma et al. 1991) also suggest a pat-tern of a few large, widespread fires combinedwith smaller, more numerous fires along theWilamette Valley margin. Mean fire size in thecoastal-interior climate zone was 12 309 hawhereas the mean fire size in the Wilamette Val-ley margin climate zone was only 2576 ha (Wim-berly 2002). These fire-size distributions wereheavily skewed, with fires smaller than IO OOO haaccounting for 89% of the total number of fires.However, fires larger than 10 000 ha accountedfor 84% ofthe total area burned (figure 12.2). Thefive largest fires, all larger than 50 000 ha, ac-counted for 71% of the total burned area.Between 1972 and 1995, 27% of the Coast

Range was clearcut, equivalent to a harvest rota-tion of 85 yr (Cohen et al. 2002). In contrast withfires, most clearcuts retain few live trees , snagsor large downed trees within their boundaries.Densities of large trees and snags (~50 cm dbh)can be three to five times higher in postfire standsthan in logged stands of similar age (Hansenet al. 1991). The mean size of clearcuts was lessthan 200 ha, with no units larger than 5000 ha(Cohen et al. 2002). Spatial variabilty in harvestrotations and the sizes of timber harvests ispresently controlled by the management regimesused in different classes of forest ownership (fig-ure 12.3). Between 1972 and 1995, the harvest ro-tation on private industrial lands was 51 yr, com-pared with roo yr on private nonindustrial lands189 yr on state-owned lands , and an average of

NATURAL FIRE REGIME AND EVALUATING FOREST MANAGEMENT '5'

0 HistoricalFIG U R E 12. 2. Size-class distribution

FIGUREII Modern for historical wildfires (1850- 1940) and(pre-Eunmodern wildfire and timber harvestof varia!:disturbances (1972- 1995), expressed asthe land.percentages of the total area disturbedsuccessicHistorical fire data are from Teensmaand (d) cet al. (1991), and timber harvest dataOregon (are from Cohen et al. (2002).represenlands cap

the presE10-100 "100

200

150

100

PNI State BLM USFS PNI

Ownership

Disturbance size (1000 ha)

160 yr on federal lands managed by the Bureauof Land Management and the U.S. Departmentof Agriculture Forest Service. Similarly, the meansize of clearcuts was more than twce as large onprivate industrial lands as in any other owner-ship class.

The harvest rotations presented in figure 12.must be interpreted with the caveat that they arebased on only 23 yr of data. On state lands, litteharvesting occurred from 1972 to 1995, because

much of the forest was below harvestable age.The calculated harvest rotation on state landswil decrease in the future, as stands reach mer-chantable age and new management plans areimplemented. On federal lands

, the reportedharvest rotations mostly reflect the period priorto the implementation of the Northwest ForestPlan. Under current policies, future rates of clear-cutting on federal lands wil be lower than thosereported here.

/""

Simulation of Pre- EuropeanSettlement Landscape Dynamics

Old growth was the most common structureclass in the simulated pre-European settement

landscape (color plate 17); in our simulations(figure 12-4), it occupied a median of 42% of theCoast Range , with values ranging from 29% (fifthpercentile) to 52% (ninety-fifth percentile). Incontrast, median abundances were 21% for theyoung class, 17% for the early-successional classand 16% for the mature class. In the present-

dayCoast Range, the percentages of mature and old-growth forests are smaller than would be expectedunder the pre-European settlement fire regimeand the percentages of early-successional andyoung forests are greater than expected. In thesimulated pre-European settlement landscapesthere was at least one large (~IOo ooo-ha) patchof old-growth forest present at all times , but thelargest old-growth patch in the present-day land-scape occupies only 650 ha (figure 12.

5). In con-trast, a mosaic of early-successional and youngforests dominates the present-day landscape, withthe largest patch of early-successional forest largerthan 500 000 ha and the largest patch of youngforest larger than 300 000 ha. In the simulatedpre-European settlement forests , most early-successional , young, and mature forests werewithin I km of an old-growth forest, as shown by

PI State BLM USFS

FIGURE 12. 3. Distribution of (a) timber harvest rotations and (b) clearcut sizes in the OregonCoast Range from 1972 to 1995. PNI = private nonindustrial

, PI = private industIial , State =state of Oregon, BLM = Bureau of Land Management , USFS = U.S. DEpartment of AgricultureForest Service. Based on data from Cohen et al. (2002).

MICHAEL C. WIMBERLY eta!.

the isolcwhich spa tial i

creasedscape , wand you

Discussic

"'-

Changesand Their

One hunuse in tha wholedynamicthe CoasdominatSmaller fwidely d

3 '

0, 4 i3 I2 I1 I

class distribution FIGURE 12. 4. Simulated historical fires (1850-1940) and (pre-European settlement) rangesnd timber harvest of variabilty for the percentages of

1995), expressed as the landscape covered by (a) early-, total area disturbed. successional , (b) young, (c) mature1 are from Teensma and (d) old-growth forests in theimber harvest data Oregon Coast Range. Dashed linestal. (2002). represent the percentage of the

landscape covered by each class inthe present-day landscape.

in our simulations

edian of 42% of theing from 29% (fifth'ih percentile). Ins were 21% for the

'-successional class. In the present-day; of mature and old-rl would be expectedlement fire regimey-successional and10 expected. In the

lement landscapes

'IOo ooo-ha) patchat all times, but theIe present-dayland-figure 12.5). In con-ssional and youngdaylandscape , withessional forest largergest patch of younga. In the simulated)rests, most early-Iature forests vvere

l forest , as shown by

the isolation index values for old-growth forestwhich were mostly less than I (figure 12.6). Thespatial isolation of old-growth patches has in-creased considerably in the present-day land-scape , whereas the isolation of early-successionaland young patches has decreased.

Discussion

Changes in the Disturbance Regimeand Their Ecological Implications

One hundred and fifty yr of postsettement landuse in the Oregon Coast Range has brought abouta wholesale change in the spatial pattern anddynamics of the forest landscape. Historically,the Coast Range was a shifting landscape mosaicdominated by large patches of old-growth forest.Smaller fragments of old-growth forest were alsowidely distributed in blocks of younger forest.

10 20 30 50

10 20 30 40 50 Percentage of landscape

In contrast, early-successional and young forestsdominate the current landscape, with old-growthforest present only as small , fragmented patches.These changes have occurred because disturbance

rates over the past ISO yr have been consistentlyhigher than under the pre-European settlementdisturbance regime. Increased burning in the

late nineteenth and early twentieth centurieswas at least partially the result of an increase inignitions by settlers (Weisberg and Swanson2003). Although fire suppression has greatly re-duced the incidence of forest fires in the latterhalf of the twentieth century, disturbance fromtimber harvests in the modern landscape has oc-curred at a more rapid rate than did disturbanceby fire in the pre-European settlement landscape.Clearcutting has tyically removed the majorityof trees from a site , in contrast to the variablenumbers of remnant live trees left by the range

,,':: -,

4000 8000 12000

4 ,3 '

1 ;,

0 '4000 8000 12000

5 I 4 ;:3 "

2 :'1 !

0 '

" '""

4000 8000 12000

20 '15

' '

10 ,

- '

05 ,

" , ; ' .""

00 :' ,,

: ' . , '"' . , "","'

4000 8000 12000

the Oregon, State =

19ricultureLargest patch (km

FIGURE 12. 5. Simulated historical(pre-European settlement) rangesof variabilty for the largest patch of(a) early-successional, (b) young,(c) mature , and (d) old-growthforests in the Oregon Coast Range.Dashed lines represent the largestpatch in the present-day landscape.

NATURAL FIRE REGIME AND EVALUATING FOREST MANAGEMENT 153

30

FIGURE 12. 6. Simulated historical (pre-European settlement) ranges of

(f;variability for the isolation indexof (a) early-successional , (b)young,(c) mature , and (d) old-growthforests in the Oregon Coast Range.Dashed lines represent the value ofthe isolation index in the present-day landscape.

Isolation index (km)

of fire severities that characterized historical fireregimes. The small sizes and spatial dispersion ofmost harvested areas compared with the histor-ical pattern of large fires have also altered thespatial pattern of the forest landscape.

These changes in landscape patterns haveseveral ecological implications. Approximatelyone-third of all vertebrates and one-half of allamphibians inhabiting forests in western Ore-gon have been characterized as "closely associ-ated" with old-growth habitats (Olson et al. 2001).Although many of these species are also found inyounger forests , it is reasonable to hypothesizethat the significant decline in their preferred

habitat has affected their populations. Species

that require large concentrations of mature orold-growth habitat, such as the northern spot-ted owl, may be particularly sensitive to the de-creased size and dispersed pattern of patches ofmature and old-growth forest (McComb et al.2002). Although small patches historically rep-resented only a minor portion of the total area ofold-growth forest, they may stil have faciltatedthe persistence of disturbance-sensitive, dispersal-limited species such as the lichen Lobaria oregana(Tuck.) Mull. Arg. This species can survive and re-

produce in young forests, but is tyically elimi-nated by stand-replacing disturbance and mustrecolonize regenerating stands through dispersalfrom undisturbed forest (Silett et al. 2000). Lossof the numerous, scattered old-growth patchesthat were historically left by fires may greatlyreduce the number of sources of propagules forreestablishing disturbance-sensitive , dispersal-limited species in young forests (Wimberly andSpies 2002).

154 MICHAEL C. WIMBERLY etai.

from anmore , thlby this dilower voand old-Europemmental c

driven bsolely at Isustain tltat (Reev

Changes in the temporal and spatial patternsof stand-replacing disturbance may also affectbroad-scale ecosystem processes , particularly thewatershed-level dynamics of water, wood, andsediment. The large fires that occurred under thepre-European settlement disturbance regime of-ten burned substantial areas within a watershedreducing slope stabilty and initiating landslidesand debris flows that delivered large pulses ofwood and sediment to streams (Benda et al. 1998).In the short term , these disturbances Jikely de-stroyed habitat for Pacific salmon (Oncorhynchusspp.) and other native fish species. Over timehowever, the wood and sediment inputs that oc-curred after fires provided raw materials for thedevelopment of diverse aquatic habitats. Longfire-return intervals provided time for fluvialprocesses to generate complex habitats that in-cluded pools , large pieces of in-stream wood , anda variety of different substrate tyes (Reeves et al.1995). These fire-free periods also allowed forestson the surrounding slopes to reestablish them-selves and develop into late-successional standswith large living and dead trees.

In contrast, the present-daydisturbance regimeis dominated by timber harvesting, which pro-duces disturbances that are more frequent

smaller, and more widely dispersed than histori-cal fires. Although individual timber harvestsonly affect small portions of a particular water-shed, a larger number of disturbances occursthan takes place under the natural fire regime,and these disturbances are distributed moreevenly in time and space than were historicalfires. Consequently, wood and sediment inputsresulting from these disturbances have shifted

ProspectsBased on

CurrentCoast Ra

the pre-

Timber 1

generatecreated bof uncu1

I999). Htern of h

bution ojand late-deferralsfurther c

In mana:could in(Curtis I'acteristictened thls il vicul 11

Tappeimsome elestands bon the siold-growof model

DespitfOIest m:and sodplementiturbanceecologiC(nipulatein the neof disturalready tfOIests , a

pattern Ihasten tlet al. 19'

fOIests c

forests th

ulated historicallement) ranges ofsolation indexional, (b) young,, old-growthon Coast Range.

sent the value ofc in the present-

from an episodic to a chronic process. Further-more , the predominantly young forests producedby this disturbance regime have smaller trees andlower volumes of dead wood than the matureand old-growth forests that dominated the pre-European settlement landscape. These funda-mental changes suggest that disturbance regimesdriven by forest management activities aimedsolely at maximizing timber production may notsustain the production of high-quality fish habi-tat (Reeves et al. 1995).

spatial patternsmay also affectparticularly the

ater, wood, and:urred under the)ance regime of-hin a watershedlating landslidesl large pulses ofend a et al. 1998).)ances likely de-

(Oncorhynchuscies. Over time,It inputs that oc-naterials for the= habitats. Longtime for fluvialhabitats that in-

tream wood , andpes (Reeves et al.o allowed forestsestablish them-

:cessional stands

Prospects for Forest ManagementBased on Natural Disturbance Regimes

Current forest management practices in theCoast Range could be altered to better emulatethe pre-European settlement disturbance regime.Timber harvests could be spatially clustered togenerate large disturbed patches similar to thosecreated by fire, while retaining contiguous blocksof uncut forest (Gustafson 1996; Cissel et al.1999). However, changing only the spatial pat-tern of harvesting wil not influence the distri-bution of forest age classes. Protecting old-growthand late-successional forests through long-termdeferrals , reserves , and wilderness areas can limitfurther declines in the amount of older forests.In managed landscapes, longer rotation lengthscould increase the proportion of older forests(Curtis 1997). Development of old-growth char-acteristics in young managed stands could be has-tened through thinning, underplanting, or othersilvicultural treatments (McComb et al. 1993;Tappeiner et al. 1997). It is also possible to retainsome elements of old-growth structure in youngstands by leaving residual trees, logs , and snagson the site after timber harvests in mature andold-growth stands, thereby emulating the effectsof moderate-severity fires (Franklin et al. 1997).

Despite the many opportunities for modifyingforest management practices, major ecologicaland social obstacles limit the potential for im-plementing management based on natural dis-turbance regimes in the Coast Range. The majorecological impediment is that our abilty to ma-nipulate landscape structures and compositionin the near term is constrained by the past ISO yrof disturbance history. Fires and logging havealready eliminated the majority of old-growthforests , and simply shifting the frequency andpattern of timber harvesting may do little tohasten the recovery of this forest type (Wallnet al. 1994; Baker 1995). Whereas old-growthforests can be rapidly converted into youngforests through disturbance, the development of

,turbance regimeting, which pro-more frequentsed than histori-timber harvests

particular water-urbances occurs

ural fire regimeistributed more1 were historicalsediment inputsces have shifted

old-growth forests occurs over hundreds of yearsthrough the relatively slow processes of forestsuccession and stand dynamics. Thus, any at-tempt at landscape management based on natu-ral disturbance regimes and historical landscapepatterns must also consider the protection ofexisting old growth, as well as of mature foreststhat wil develop into old growth over the nextcentury.

The complex pattern of forest ownership isperhaps an even greater obstacle to implement-ing management based on natural disturbanceregimes in the Coast Range. Forest managementwithin each ownership type is influenced by aunique set of management goals and regulatoryconstraints, limiting the range of conditions thatcan be produced within a particular ownershipclass. For example, management on most privatelands is driven by predominantly economic con-siderations, which prescribe clearcutting at fairlyshort rotations (40-60 yr) as the prevalent man-agement practice. If present forest managementregimes are extrapolated 100 yr into the futureprivate lands wil continue to consist primarilyof early-successional forest and young, closed-canopy forest (Spies et al. 2002).

In contrast, the U.S. Department of Agricul-ture Forest Service and the Bureau of Land Man-agement operate under regulations imposed bythe Northwest Forest Plan, which in the CoastRange allows only minimal timber harvestsoutside of late-successional and riparian reserves(Forest Ecosystem Management and AssessmentTeam 1993). If fire suppression policies continueto be effective , most federal lands wil succeedto old-growth conditions. The Oregon StateDepartment of Forestry intends to manage stateforest lands in the northern Coast Range using astructure-based management approach, in whichstands are harvested on 80- to 130-yr rotationswith active management to provide for largetrees , snags , downed trees , and other old-growthstructural components. Current plans call foreventually maintaining 20-30% of the land basein an older forest structure, which is intended toemulate the structural features and functionalcharacteristics of old-growth forests (OregonDepartment of Forestry 20or).

If Coast Range forests continue to be managedunder existing policies, increasing the amountsof old-growth forest on public lands will moveoverall landscape patterns closer to the range ofhistorical variabilty. However, the disparate dis-turbance regimes on federal and private landsmay eventually leave few forests in the mature

NATURAL FIRE REGIME AND EVALUATING FOREST MANAGEMENT 155

structure class. Harvest rotations on most privatelands are too short to allow forests to reach themature stage, and the continued exclusion of fireand timber harvesting on federal lands wil allowmost forests to pass through the mature stageand develop into old growth. Under current for-est management policies, this divergence in theage structure of forests under public and privateownership is predicted to occur gradually overthe next roo yr (Spies et al. 2002). Some newmature forests wil develop on state and privatenonindustrial lands managed under long rota-tions, but the total area of mature forests willikely account for only a small portion of theCoast Range. If old growth is lost in the futureas a result of fire, wind, or other natural distur-bances , this gap in the forest age structure couldlimit the potential for rapid development of newold-growth habitats.

The spatial pattern of ownership also limitsfuture possibilties for replicating large, histori-cal fire events through timber harvesting. Forexample , the boundaries of the largest blocksof public land in the Coast Range correspond tothose of large historical fires (see figure 12.color plate 16b). The southern block of the Sius-law National Forest includes a large area burnedby the Siletz fire , and the northern block en-compasses a significant portion of the Nestuccafire. The Tilamook State Forest includes themajority of land burned by the Tilamook firesand the Ellot State Forest covers most of the areaof the Coos Bay fire. Arbitrary human-imposedboundaries dominate in other areas, such as thecheckerboard pattern of ownership in the south-eastern portion of the Coast Range, where Bureauof Land Management lands and private lands areinterspersed in 2. blocks. Also, many largeblocks of private land actually have multiplelandowners, each acting largely independentlyto manage their land.

Our present social, political, and regulatorysystems do not provide a framework for manage-ment coordinated among multiple private land-owners or between private and public ownerships.Therefore, the size of a particular ownership setsan upper boundary on the spatial scale of theforest management regime that can be appliedtherein. Because the largest ownership blocks inthe Coast Range are artifacts of past fires, man-agement strategies aimed at maintaining a rangeof forest structures in a single ownership blocknecessitate timber harvests that are considerablysmaller than the largest historical burns. Thesemanagement strategies may eventually produce

156 MICHAEL C. WIMBERLY eta!.

novel landscape patterns that contain a habitatmosaic with finer resolution than was presentunder the pre-European settlement fire regime.The relative importance of habitat amount ver-sus habitat pattern for species conservation ispresently a subject of active scientific debate(Schmiegelow and Monkkonen 2002), and theimplications of altering the spatial scale of theforest habitat mosaic are unclear.

Conclusions

The shift from a disturbance regime dominatedby fire to one dominated by timber harvestinghas had significant impacts on the pattern anddynamics of forest habitats in the Oregon CoastRange. Clearcuts are smaller and occur at shorterrotations than did historical fires. In additionclearcutting removes a larger proportion of bothlive and dead trees from the site than do manywildfires. Whereas pre-European settlement fireregimes varied primarily along a climatic gradi-ent, forest management policies vary amonglandowners. In turn, the pattern of land owner-ship reflects a legacy of past wildfires as well as

the social , economic , and poliical history ofthe region. The present-day forest landscape hassmaller areas of mature and old-growth foreststhan would be expected under the pre-Europeansettlement disturbance regime. Whereas large(~IOo ooo-ha) patches of old-growth forest werecommon in pre-European settlement landscapesold growth is currently distributed in small(-:650-ha) fragments that are isolated from otherhabitat types.

Any future forest management policy basedon emulating natural disturbance in the CoastRange wil necessarily represent a compromisebetween our scientific understanding of dis-turbance and its ecological effects, and the eco-nomic, social, and political reaJities of the re-gion. Given the history of human land use andthe constraints imposed by current ownershippatterns , attempting to closely replicate the tem-poral and spatial patterns of natural disturbancemay be an unrealistic goal, and could actuallymove landscape patterns further from the rangeof historical variabilty if the remaining old-growth forests are disturbed. Instead , a moretenable short-term goal might be to maintainkey elements of historical landscape processesand structures at multiple scales. Examples ofthis approach include maintaining a few largepatches of old-growth forest at the regional scale,numerous small patches of old growth at a land-scape scale , and individual old-growth struc-

tures (e. , laat a stand scowner goalsbroad-scale ithrough a rcombinationbased approa

Acknowledgm

We thank Jarfeedback on

)ntain a habitat

Lan was presentlent fire regime.tat amount ver-conservation iscientific debate2002), and thetial scale of the

ime dominatedIlber harvesting:he pattern ande Oregon Coastoccur at shorteres. In additionportion of boththan do many

, settlement fireclimatic gradi-

s vary amongof land owner-

jfires as well astical history oft landscape hasgrowth forests

e pre-EuropeanWhereas largewth forest wereent landscapesmted in small

lted from other

It policy based:e in the Coasta compromisemding of dis-

and the eco-

ities of the re-n land use andent ownership)licate the tem-ral disturbancecould actuallyfrom the rangeemaining old-stead , a moreIe to maintaincape processes

;. Examples ofng a few largeregional scale

)wth at a land-growth struc-

tures (e. , large trees, snags, and downed trees)at a stand scale. Given the wide range of land-owner goals and regulatory constraints, thesebroad-scale goals would have to be achievedthrough a mixture of strategies, including acombination of active management and reserve-based approaches.

Two anonymous reviewers also provided usefulcomments. David Boughton, Norm JohnsonStephen Lancaster, and other researchers work-ing with us on the Coastal Landscape Modelingand Assessment Study (CLAMS), provided valu-

able critiques throughout all stages of this proj-ect. Support for this work was provided by theNorthwest Forest Plan Program of the U.S. De-partment of Agriculture Forest Service PacificNorthwest Research Station and by the Univer-sity of Georgia.

NATURAL FIRE REGIME AND EVALUATING FOREST MANAGEMENT 157

Acknowledgments

We thank Janet Ohmann for providing data andfeedback on earlier versions of the manuscript.