Reprinted from BioScience

Tree Death as anEcological Process

The causes, consequences, and variability of tree mortality

Jerry F. Franklin, H. H. Shugart, and Mark E. Harmon

T

ree death is so commonplacethat the casual observer mightlogically assume it to be well

understood by biologists. Somecauses of tree mortality are obviousand even spectacular events, such aswildfires and hurricanes. But overallthe patterns and causes of tree deathtypically are complex, and we areonly beginning to appreciate thecomplexities.

Understanding and predicting treemortality is critical in both appliedand basic ecology. Practically speak-ing, information on mortality is es-sential in calculating forest standyields and allocating efforts in tendingand protecting forests. A thoroughknowledge of tree death is also neces-sary to interpret correctly the dyinghack of forests. Yet, despite its longhistory, forest husbandry lacks acomprehensive understanding of treemortality.

In basic ecology, tree death is rele-vant to a broad array of topics. Ecolo-

Jerry F. Franklin is the chief plant ecolo-gist, USDA Forest Service, Pacific North-west Research Station, Forestry SciencesLaboratory, Corvallis, OR 97331, andBloedel Professor of Ecosystem Analysis,College of Forest Resources, University ofWashington, Seattle, WA 98195. H. H.Shugart is Corcoran Professor of Environ-mental Sciences, University of Virginia,Charlottesville, VA 22903. Mark E. Har-mon is a research associate, Forestry Sci-ences Laboratory, Oregon State Universi-ty, Corvallis, OR 97331. © 1987American Institute of Biological Sciences.All rights reserved; the US governmentretains a nonexclusive royalty-free licenseto publish or reproduce this article.

We are only beginning toappreciate the complexityof patterns of tree death

gists focusing on ecosystems, commu-nities, populations, physiology, andevolution all find tree death signifi-cant to their perspectives. Tree mor-tality exemplifies several importantprinciples of ecological complexity.

Tree death can be used to illustratethe variability of an ecological proc-ess in terms of rates, as well as causalfactors or mechanisms; the necessityfor defining the spatial and temporalscales of interest; and the importanceof the natural history of species andecosystems in understanding ecologi-cal processes. Studies of tree deathcan also illustrate the relevance andvalidity of differing viewpoints—those of different disciplines orscales—on the same process.

An emphasis on these general fea-tures of ecological processes and sys-tems is especially appropriate in viewof the all-too-human tendency ofecologists to seize upon one view-point to the exclusion of all others.The problem is compounded by at-tempts to define many ecologicalproblems rigidly in terms of either/orhypotheses. Many , Of the ecologicalproceSses and systems are not suffi-ciently well understood, or are toocomplex, to be described in such lim-ited terms.

It is tempting to use simple systemsand models to circumvent the com-

plexity introduced by the varied natu-ral histories of species and naturalecosystems. Unfortunately, such sim-plifications can also mislead the un-wary about important ecologicalprocesses.

In this article, intended to provide acontext for the other articles in thisissue of BioScience, we provide anoverview of tree death as a rich eco-logical process. We include its conse-quences and causes, its variability,and the importance of species' naturalhistories. We also use tree death toillustrate some general aspects of eco-logical processes.

Consequences of tree deathTree death's importance in ecologyreflects the multiple roles that a treeplays. It is a primary producer, astorage compartment, and a supportstructure. Tree death removes a ge-netically distinct individual from thestand, but it also provides additionalresources to the ecosystem. In thisway, the death process itself doesimportant work (Table 1).

The function of dead trees in theecosystem has rarely received the con-sideration that it deserves. At the timea tree dies, it has only partially ful-filled its potential ecological function.In its dead form, a tree continues toplay numerous roles as it influencessurrounding organisms. Of course,the impact of the individual tree grad-ually fades as it is decomposed and itsresources dispersed, but the woodystructure may remain for centuriesand influence habitat conditions formillenia.

550 BioScience Vol. .37 No. 8

Figure I. Logs provide habitat for tree seedlings and other higher plants. In ecosystemssuch as the alluvial Sitka spruce-western hemlock forests of the Olympic Peninsula,Washington, essentially all tree reproduction is confined to rotten wood seedbeds,primarily nurse logs.

•.4

Figure 2. Even death can contribute important ecological work. Uprooting of trees incoastal Alaskan Sitka spruce-western hemlock forests is an important factor retardingdevelopment of iron pans (impervious layers of mineral deposits) and consequentpaludification of forest sites.

Table 1. Some ecological changesbrought about by the death of a tree.

Altered tree population structure

Altered community structure

Shift from biomass to necromass

Resources released (light, nutrients, moisture)

Resources stored by decomposers

New resources createdStructures (snags or logs) for wildlifeHabitat for decomposer organismsComplex organic compounds

Work carried outKills other trees or organisms by crushingMixes soil (in case of uprooting)

While many organisms display acontinuum of ecological roles be-tween the living and dead forms, witha gradual fading of influence afterdeath, in trees this continuum is moreapparent because of their size, dura-bility, and multiplicity of roles in theecosystem.

Although from an ecosystem per-spective the tree is shifted from thecategory of living to dead matter,physiologically, some (even most) of atree (e.g., the heartwood) could beconsidered dead much earlier and sig-nificant portions of a live tree mayalready have been decomposed. In alive conifer, only about ten percent ofthe cells are actually alive: the leaves(three percent), inner bark (phloemand cambium, five percent), and raycells in sapwood (two percent). Someprocesses associated with dead treesbegin while the tree is still alive. Forexample, fungi are already at workrotting the woody material, and ani-mals excavate the dead parts of livingtrees. In contrast, a dead tree or log inan advanced state of decay may in-clude a considerable number of livingcells, as much as 35% of the biomassmay be live fungal cells alone (Swift1973).

Tree death substantially increasesthe resources (e.g., light, nutrients,water, and energy) available to otherorganisms in the ecosystem. Theamount of resources made availabledepends on the number and size oftrees that die. The resources may bemade available instantaneously (e.g.,light) or very slowly (e.g., nutrientsand energy contained within theboles). The dead tree may also func-tion as a sink where nutrient re-sources brought in by the decomposer

organisms are immobilized for a peri-od of time.

The dead tree is itself a major newresource for the ecosystem, whetheras a snag (standing dead tree) or as adowned log. The importance of deadwood structures to the geological andecological functions of forest andstream ecosystems has been thor-oughly reviewed (Harmon et al.1986, Maser and Trappe 1984).

With the large array of organismspresent in the decaying log, it may bemore "alive" than a living bole. Inaddition to being the habitat of de-composer organisms, dead trees pro-vide critical habitat for sheltering andfeeding a variety of animal species

streams, they contribute to develop-ment of stepped stream profiles, re-duce channel erosion, and createmore retentive stream reaches (Har-mon et al. 1986).

Tree death may itself do importantmechanical work. Falling trees orsnags often kill other trees or otherorganisms. More than 15% of themortality in mature and old-growthDouglas-fir stands in the PacificNorthwest consists of trees knockedover, broken, or crushed by fallingtrees. The uprooting of trees lifts andmixes forest soil, an important eco-logical process (Figure 2). For exam-ple, in the Sitka spruce-western hem-lock (Picea sitchensis-Tsuga hetero-phylla) forests of southeasternAlaska, soil churning by the roots ofwindthrown trees retards develop-ment in the soil of impervious layersof mineral deposits, known as iron pan.Without this process, standing pools ofwater would eventually produce swampyforest sites (Ugolini et al. 1987).

Causes of tree death

Although tree death is sometimesabrupt, it is more frequently a com-plex and gradual process with multi-ple contributors (Waring, p. 569, thisissue). For example, the proximate

causes of death (e.g., an insect ordisease) may be simply the coup degrace, whereas the primary factors(e.g., starvation) may not be obvious.Tree death often represents an arbi-trary point on a continuum.

Causes of tree death can be catego-rized in a variety of ways, includingsuch dichotomies as abiotic and biotic(Table 2), allogenic and autogenic,and extrinsic and intrinsic. All theseclassifications fail to portray the com-plicated interactions among trees,their environment, and various agentsof mortality. In part, the interactive,sequential nature of tree mortalitylimits the value of these dichotomies.For example, the phenomenon calledsuppression, the limiting of one tree'sgrowth by the presence of another,usually larger, tree may reduce thesuppressed tree's rooting strength andthus increase its susceptibility towind. Suppression also may reducetree size and bark thickness, therebyincreasing vulnerability to surfacefires.

Abiotic causes of tree death arealso, in large measure, allogenic andextrinsic in nature. Environmentalstresses, such as flooding, drought,heat, low temperatures, ice storms,and excess sunlight, tend to be partic-ularly important in the death of tree

seedlings. We place most pollutantstresses (e.g., acid precipitation,ozone, and acid-forming oxides ofnitrogen and sulfur) into the abioticcategory, although the proximatecause of death may be biotic agents orphysiological failure. Most of the abi-otic agents could be strong selectiveforces in evolution, but at least one,volcanic eruptions, may be too ran-dom in its timing and impact to havean evolutionary effect. Research atMount St. Helens, for example, hasshown dramatic differences, depend-ing on the season of eruption, in thesurvival of organisms and the rateand composition of posteruptive for-est recovery (Franklin et al. 1985).

Biotic factors are highly variableand difficult to define as being eitherextrinsic or intrinsic. Most effects ofcompetition fall in the category ofstarvation, where light, nutrients, orwater limit photosynthesis. The mostdrastic effects of herbivory occurwhen insects, ungulates, or humanseat tissues essential for growth (e.g.,cambium). Herbivory of roots maykill trees or predispose them to me-chanical failure. But trees can toleratesignificant herbivory of photosynthet-ic tissue and sap.

Diseases may also kill trees or maypredispose them to mechanical fail-ure. For example, a large percentageof windthrown old-growth Douglasfirs (Pseudotsuga menziesii) containsignificant butt rot (Polyporusschweinitzii). Both insects and diseasemay be the proximate agent of deathin trees already weakened by otherfactors; as such, they often areblamed for deaths more properly as-

Table 2. Some common contributors toor causes of tree death.

AbioticFireLightningChemical pollution (e.g., ozone)Environmental stress (e.g., flooding,

drought, heat)WindVolcanic eruptionClimatic change

BioticOld age, senescenceMechanical imbalance (e.g., top heavy)Starvation (inadequate photosynthesis)Consumption (includes insects, ungulates,

and humans)Disease

(Brown 1985, Thomas 1979). Snagsand logs also provide habitat forplants of higher orders. Indeed, theseedbeds provided by "nurse logs"may be the primary sites for treereproduction in some ecosystems(Figure 1) (Harmon et al. 1986).Along with the nutrients and energyreleased by the decomposition proc-ess, there is also significant nitrogenfixation by organisms living within(in terrestrial habitats) and on (instream habitats) the wood itself (Har-mon et al. 1986).

Woody structures also influencegeomorphic processes. For example,they serve as erosion barriers on for-est slopes and, in smaller forest

September 1987

551552

BioScience Vol. 37 No. 8

HEALTHY TREE

BLUE-STAIN FUNGUS

DEFOLIATIONSUPPRESSION

pitchdefense •

DEATH }

BARK BEETLES

dominance recovery

Table 3. Changes in causes and rates of tree mortality during forest successional stagesin the Douglas-fir region of the Pacific Northwest.

Stage

Prevegetative Full vegetative Closed tree Mature Oldclosure cover canopy forest forest

Approximateperiod(years)

0-5 5-20 20-100 100-200 >200

Mortality Very high High High to Medium to Medium torate medium low low

Typicalmortalitycauses

Environmentalstress,

predation,pathogens

Interspecificcompetition,

environmentalstress,

pathogens,predation

Intraspecificcompetition,pathogens,

wind

Pathogenswind,

competition

Wind,pathogens

physiologicaldisorders

COMPETITION

Figure 3. Mortality spiral for a Douglas-fir tree illustrates the series of events leading toits death. The spiral is based upon the decline disease spiral of Manion (1981). In thisexample, a healthy tree is suppressed by larger trees. If not released from competition,the tree is predisposed to attack by defoliators. Once partially defoliated, the weakenedtree is attractive to bark beetles (Wickman 1978) and is unable to resist the beetles,which carry blue-stain fungus (Berryman 1982). The fungus blocks the transpirationstream in the tree and causes dessication of the leaves. As the tree progresses along thisspiral, its opportunities to escape death become more limited.

A

signed elsewhere. Humans are, ofcourse, a major biotic cause of treedeath, acting both directly (tree cut-ting) and indirectly in influencing al-most all other agents.

The issue of tree senescence and itscontribution to mortality is a veryinteresting one. Does old age—an in-trinsic physiological alteration in thetree—contribute to tree death? Somebiologists have proposed that treeshave the potential to be immortal,i.e., they must be killed by some ex-ternal agent (Molisch 1938). But, in-herent declines in vigor and in growthdo detract from a tree's ability toresist a variety of damaging agents.Families and genera differ markedlyin the timing of these declines; forexample, trees in the families Cupres-saceae and Taxodiaceae typically canlive for millenia. There is also anenvironmental component of tree lon-gevity. Some trees live longer wheresite conditions restrict growth rates.Bristlecone pine, Pinus aristata, grow-ing under stressful conditions onmountains, can live for more than3000 years.

Tree death thus is generally theresult of complex interactions amongmultiple factors. As Shigo (1985)notes, most "disease-causing agentsinjure organisms that have been pre-disposed to diseases." He emphasizesthe importance of energy reserves indisease resistance of trees and thecumulative effect of many seeminglyunimportant injuries. Manion's(1981) "decline disease spiral" can begeneralized to a mortality spiral re-flecting the cumulative, sequentialcontributions of various events andfactors (Figure 3).

A probability transition matrixmay provide a useful mathematicalconstruct of the mortality spiral.While at first glance this approachwould seem to allow an infinite num-ber of causal combinations, it is likelythat there is a limited number ofmortality spirals (or transitions in thematrix), each with strongly linkedfactors. We have already described anumber of well-known spirals, suchas butt rot and wind. Other spiralsmay be less strongly linked and lesspredictable. This approach also al-lows biologists to consider the mecha-nisms allowing trees to escape variousspirals. In addition, the optimal pointto apply management actions could

be identified. This perspective shouldreceive more attention in future mor-tality studies.

The way in which a tree diesstrongly influences its subsequent ef-fect on the ecosystem. For example,instantaneous death may result inrapid understory response, whereas aslow decline would allow gradual ad-justments to freed resources. Treesthat die standing upright provide dif-ferent animal habitats than downedlogs and also differ in rates and domi-nant processes of decomposition. Thedifferences in decay rates have impor-tant ecosystematic implications. Forexample, Douglas-fir snags and theirdebris in the Pacific Northwest maydisappear at two to three times therate of comparable windthrown treesdue to high rates of fragmentation(Graham 1982).

Temporal variability

The timing of tree death, like manyother ecological processes, is highlyvariable and unpredictable. The tem-poral variation is influenced by physi-ology, such as the age of the individ-ual or of entire cohorts (Mueller-Dombois, p. 575, this issue); succes-sion, that is, the progression of thecommunity through time; andchance.

The rates and causes of tree mortal-ity show dramatic variation withsuccession (Harcombe p. 557, andPeet and Christensen, p. 586, thisissue). This variation is well illustrat-ed by a successional stage after wild-fire or clearcutting in the Douglas-firregion of the Pacific Northwest(Table 3).

Mortality rates, primarily due to

heat and drought, are high amongcolonizing tree seedlings. Herbivoryand pathogens may also be consid-ered important factors in tree mortal-ity at this stage. As herb and shrubcoverage increases, competition be-tween trees and other plants becomesimportant; this process may dominatefor 10-20 years.

Competition among trees becomesimportant with development of aclosed forest canopy, and so-calledthinning mortality begins. This com-petition dominates for a long periodin Douglas fir; for example, stands100-150 years old still exhibit thispattern of mortality,' although theeffects of pathogens may also be sig-nificant. The period of competitivemortality may occur earlier and bebriefer for tree species that maturemore rapidly, such as loblolly pine(Pinus taeda) (Peet and Christensen,p. 586, this issue). Finally, wind andpathogens tend to become the proxi-mate causes of death, to which senes-cence, competition, and environmen-tal stresses may contribute.

Some important generalities can beassociated with tree death along thissuccessional gradient (Table 3). First,mortality rates generally decline, andcauses of death appear to becomemore complex throughout the succes-sional stage. Second, the forest is atan approximate equilibrium for longperiods of time, that is, mortalityrates are essentially constant andchange in forest structure and compo-

sition is slow.'` An organism or anecologist observing such a forest overa 50-year period would appropriatelyview it as in equilibrium; one with a500-year perspective would not.Third, probability functions based onsuch factors as species and relativesize are required for predicting indi-vidual and stand mortality.

Much tree death is episodic andirregular. Catastrophic destruction ofstands represents the extreme form ofa mortality episode, in which rates ofmortality rise above the backgroundlevels for the stand or cohort of trees.But mortality episodes also occurwithin stands remaining essentiallyintact. Such episodes may involve in-dividual trees scattered through thestand or together in small groups.Criteria defining such episodes maybe either quantitative (e.g., mortal-ity beyond two standard deviationsfrom the long-term mean) or moresubjective. The episodic mortalityrate varies with tree species, foresttype, and successional stage. Someepisodes of mortality are predict-able, based for example, on develop-ing stand structure, whereas otherepisodes, such as those associatedwith unusual climatic events, are de-termined by chance.

Episodes of tree death are often dueto bark beetle epidemics, wind, cli-matic stresses (e.g., drought), or an-thropogenic influences (e.g., pollut-ants). At least some of these episodes,such as bark beetle outbreaks, appear

to be related to successional stage andother predisposing factors. Standsmay develop beyond the long-termcarrying capacity of the site duringfavorable periods, undergo stress dur-ing an unfavorable climatic period,and then be subject to a major out-break of insects. In the Douglas-firregion an extensive windthrow creat-ed conditions for epidemic outbreaksof the Douglas-fir bark beetle (Den-droctonus pseudotsugae) and subse-quent elevated tree mortality (Wrightand Lauterbach 1958).

Spatial variationMortality is not evenly distributed inspace. The rates and mechanisms oftree death differ dramatically over alandscape. Systematic differences mayexist on different landforms or onadjacent sites of different productivitylevels. In the Pacific Northwest, forexample, mortality rates in matureand old-growth stands are generallyhigher on habitats of high productivi-ty than on those of low productivity.'A regional gradient is also apparent(Table 4), reflecting, at least in part,differences in productivity. Mortalityis higher for the coastal Sitka spruce-western hemlock than for the Cas-cade Range Douglas fir, which in turnhas higher productivity than the inte-rior ponderosa pine (Pinusponderosa).

Many agents of tree mortality havedistinct spatial patterns at the land-scape level. Windthrow is most im-portant on wet soils where rootingzones are restricted and in particulartopographic positions (Gratkowski1956). Regional gradients in intensityof wind damage also occur, as in thePacific Northwest where wind-relatedmortality drops from about 80% incoastal spruce-hemlock stands to lessthan 15% in interior ponderosa pinestands (Table 4).

Wildfires are known to occur withdifferent frequencies in different partsof a landscape or region (Hemstrom1982, Kessell 1979). Mortality fromatmospheric pollutants is often mostsevere in ridgetop, cloud, or fog for-ests (Johnson and Siccama 1983, Lo-vett et al. 1982, Manion 1981). Mor-tality caused by fluvial processes,such as bank cutting or flooding, also

'See footnote 1.'J. E Franklin, M. Klopsch, K. J. Luchessa, and J. E Franklin and D. S. DeBell, 1987. Maim-M. E. Harmon, 1987. Manuscript submitted. script submitted.

September 1987

553

554

BioScience Vol. 37 No. 8

Table 4. Differences in rates of mortality and percentage of tree death associated withwind in mature and old-growth conifer forests along a geographic gradient from coastalSitka spruce-western hemlock to interior ponderosa pine forests.

Annualmortalit■.,

rate

Wind-related

mortalityForest (% of cohort) (% of total)

Sitka spruce- Cascade Head, OR. 2.95 83western hemlock Olympic Peninsula, WA* 0.81

Douglas fir-western hemlock

H. J. Andrews ExperimentalForest, OR'

0.70 33

Mount Rainier, WA' 0.52 41Wind River Experimental 0.75 46

Forest, WA*

Ponderosa pine Metolius Research Natural 0.31 10Area, OR.

Pringle Falls Research 0.52 18Natural Area, OR.

Data courtesy of S. Greene, Forestry Sciences Laboratory, Corvallis, OR 97331.*Data on file at the Forestry Sciences Laboratory, Corvallis, OR 97331.'Franklin et al. submitted manuscript.*Franklin and DeBell 1987.

has a strong spatial pattern (Swansonand Lienkaemper 1982). Becausesuch landscape-level patterns in ratesand causes of mortality are so com-mon, positions of study sites need tobe recognized and defined in researchon tree death.

The spatial patterning in tree mor-tality is not completely accounted forby the landscape perspective. Know-ing the degree to which tree death isdispersed or aggregated within thestudy area is often important. If in astand, for example, dying trees scat-tered rather than clustered, the conse-quences of the tree death may be quitedifferent. This patterning will deter-mine, for instance, the size of gapsthat are created.

Similar information is needed at thelandscape level; is tree death dis-persed throughout or is it occurring inpatches or even as complete stands?The number of trees that die may beidentical, but the ecological implica-tions very different.

Importance of natural historyKnowledge of the ecology and naturalhistory of individual species and eco-systems is essential for understandingand predicting what will occur inforests. The only way to contemplatethe near-infinite array of possibilitiesof species and systems is through sys-tematic observations under naturalconditions. The study of ecosystemdynamics requires such natural his-tory research.

The variety in patterns of deathamong tree species reflects such fac-tors as differences in life-spans, vul-nerability to various agents, and dis-tribution in the landscape. Thisvariety is apparent in regional pat-terns of tree mortality (Buchman et al.1983) as well as among species in thesame stand.

Tree species can be grouped intotypes with similar patterns of death(Franklin and Hemstrom 1981, Shu-gart 1984); such functional groupsmay be particularly useful in model-ing exercises where detailed species-specific information is not available.One paradigm often used in model-ing-in which an individual's proba-bility of death increases as it ap-proaches the maximal size or age forthe species-should probably beavoided whenever possible. There is

evidence that rates of mortality differ(Hibbs 1979) and may actually de-cline with size and age in at least somespecies (Harcombe, p. 557, thisissue).

Near-simultaneous death of entiretree cohorts, as reported by Mueller-Dombois (p. 575, this issue), is anextreme case exemplifying the impor-tance of considering individual spe-cies and their ecologies. Synchrony insenescence of even-aged tree cohortsis probably widespread but rarely tothe degree seen in the Metrosiderosspp. of Hawaii and New Zealand(Allen and Rose 1983, Mueller-Dom-bois 1983, Stewart and Veblen 1983),the Nothofagus forests of New Zea-land (Jane and Green 1983, Wardleand Allen 1983), and the Abies spp.of Japan and New England (Iwakiand Totsuka 1959, Sprugel 1975).

Tree mortality has important impli-cations for succession because the in-dividuals it removes may not be re-placed. In an attempt to add scientificstructure and mathematical rigor tothe understanding of ecological suc-cession, Tansley (1929) proposedclassifying ecological successions ac-cording to apparent causes of thechange in the vegetation pattern.Thus he identified allogenic succes-sions, in which change was drivenmostly by external features, and con-trasted these with autogenic succes-sions in which the change was driven

from within. Unfortunately, the logicof this dichotomy did not alter thefact that, in reality, both autogenicand allogenic causes seem to fre-quently share (or alternate) the con-trol of the successional direction ofany given ecological system. Tansleycame to recognize this and retractedthe idea of the dichotomy in the clas-sic paper (Tansley 1935) in which hedefined the ecosystem concept.

The Gordian knot of intertwinedcausality that made it difficult forTansley to design a clean, useful di-chotomy in the successional dynamicsof ecosystems in the 1920s and 1930sremains tightly tied today. Perhapsrecognizing that such dynamics arethe consequence of multiple contrib-uting factors is more useful thanclinging to an artificial autogenic-al-logenic dichotomy.

ConclusionsThe consequences of tree death, interms of effects on other ecosystemcomponents and processes, dependon many variables including the spe-cies, mortality agent, position, spatialpattern (dispersed or aggregated), andnumbers that have died. Tree death isan important indicator of ecosystemhealth and can assist recognition ofstresses caused by pollutants, such asacid rain and ozone. However, thevalue of tree death as an indicator of

anthropogenic disturbance dependson a thorough understanding of pat-terns of tree death under natural con-ditions. At the present time, adequateunderstanding of this is woefullylacking.

Tree death also demonstrates someprinciples of ecological processes: theimportance of defining the spatial andtemporal context of a study, the im-portance of stochastic processes, thefact that most ecological processes aredriven by multiple mechanisms andthat the relative importance of thesemechanisms changes in time andspace, and the importance of species'and ecosystems' natural histories.Tree death illustrates that many validand useful perspectives on a single,presumably simple process exist. Fur-ther, it makes clear that we need togive more consideration to the biolo-gy of organisms and ecosystems indeveloping, evaluating, and applyingtheoretical constructs.

References citedAllen, R. B., and A. B. Rose. 1983. Regenera-

tion of southern rata (Metrosideros umbel-lata) and kamahi (Weinmannia racemosa) inareas of dieback. Pac. Sci. 37: 433-442.

Berryman, A. A. 1982. Mountain pine beetleoutbreaks in Rocky Mountain lodgepolepine forests. J. For. 80: 410-413, 419.

Brown, E. R., ed. 1985. Management of wild-life and fish habitats in forests of westernOregon and Washington. Part 1-Chapternarratives. R6-F&WL-192-1985. USDAForest Service, Pacific Northwest Region,Portland, OR.

Buchman, R. G., S. P. Pederson, and N. R. Wal-ters. 1983. A tree survival model with appli-cation to species of the Great Lakes region.Can. J. For. Res. 13: 601-608.

Franklin, J. F., and M. A. Hemstrom. 1981.Aspects of succession in the coniferous for-ests of the Pacific Northwest. Pages 212-229in D. C. West, H. H. Shugart, and D. B.Botkin, eds. Forest Succession: Concepts andApplication. Springer-Verlag, New York.

Franklin, J. F., J. A. MacMahon, E J. Swanson,and J. R. Sedell. 1985. Ecosystem responses

to the eruption of Mount St. Helens. Natl. Peet, R. K., and N. L. Christensen. 1987. Corn-Geogr. Res. 1: 198-216. petition and tree death. BioScience 37: 586-

Graham, R. L. 1982. Biomass dynamics of 595.dead Douglas-fir and western hemlock boles Shigo, A. L. 1985. Wounded forests, starvingin mid-elevation forests of the Cascade trees. J. For. 83: 668-673.Range. Ph.D. dissertation. Oregon State Uni- Shugart, H. H. 1984. A Theory of Forestversity, Corvallis. Dynamics. Springer-Verlag, New York.

Gratkowski, H. J. 1956. Windthrow around Sprugel, D. G. 1975. Dynamic structure ofstaggered settings in old-growth Douglas-fir. wave-regenerated Abies balsamea forests inFor. Sci. 2: 60-74. the northeastern United States. J. Ecol. 64:

Harcombe, P. A. 1987. Tree life tables. BioSci- 889-911.ence 37: 557-568. Stewart, G. H., and T. T. Veblen. 1983. Forest

Harmon, M. E., et al. 1986. Ecology of coarse instability and canopy tree mortality in West-woody debris in temperate ecosystems. Pages land, New Zealand. Pac. Sci. 37: 427-432.133-302 in A. MacFadyen and E. D. Ford, Swanson, E J., and G. W. Lienkaemper. 1982.eds. Advances in Ecological Research, vol. Interactions among fluvial processes, forest15. Academic Press, Orlando, FL. vegetation, and aquatic ecosystems, South

Hemstrom, M. A. 1982. Fire in the forests of Fork Hoh River, Olympic National Park.Mount Rainier National Park. Pages 121- Pages 30-34 in E. E. Starkey, J. E Franklin,126 in E. E. Starkey, J. E Franklin, and J. W. and J. W. Matthews, eds. Ecological Re-Matthews, eds. Ecological Research in the search in National Parks of the PacificPacific Northwest. Oregon State University, Northwest. Oregon State University, ForestForest Research Laboratory, Corvallis. Research Laboratory, Corvallis.

Hibbs, D. E. 1979. The age structure of a Swift, M. J. 1973. The estimation of mycelialstriped maple population. Can. J. For. Res. biomass by determination of the hexoamine9: 504-508. content of wood tissue decayed by fungi. Soil

Iwaki, H., and T. Totsuka. 1959. Ecological Biol. Biochem. 5: 321-332.and physiological studies on the vegetation Tansley, A. G. 1929. Succession: the conceptof Mt. Shimagare 11. On the crescent-shaped and its values. Proc. Int. Congress Plant Sci.,"dead trees strips" in the Yatsugatake and 1926: 667-686.the Chichibu Mountains. Bot. Mag. Tokyo 1935. The use and abuse of vegetation-72: 255-260. al concepts and terms. Ecology 16: 284-

Jane, G. T., and T. G. A. Green. 1983. Vegeta- 307.tion mortality in the Kaimai Ranges, North Thomas, J. W., ed. 1979. Wildlife habitats inIsland, New Zealand. Pac. Sci. 37: 385-390. managed forests in the Blue Mountains of

Johnson, A., and T. Siccama. 1983. Acid depo- Oregon and Washington. Agric. Handb. 553.sition and forest decline. Environ. Sci. Tech- USDA Forest Service, Washington, DC.nol. 17: 294-306. Ugolini, E C., B. Bormann, and F. H. Bowers.

Kessell, S. R. 1979. Gradient Modeling Re- 1987. The role of tree windthrow on forestsource and Fire Management. Springer-Ver- soil development in southeast Alaska. Can. J.lag, New York. For. Res., in press.

Lovett, G., W. Reiners, and R. Olson. 1982. Wardle, J. A., and R. B. Allen. 1983. DiebackCloud droplet deposition in subalpine bal- in New Zealand Nothofagus forests. Pac.sam fir forests: hydrological and chemical Sci. 37: 397-404.inputs. Science 218: 1303-1304. Waring, R. H. 1987. Characteristics of trees

Manion, P. D. 1981. Tree Disease Concepts. predisposed to die. BioScience 37: 561-Prentice-Hall, Englewood Cliffs, NJ. 583.

Maser, C., and J. M. Trappe, eds. 1984. The Wickman, B. E. 1978. Tree mortality and top-seen and unseen world of the fallen tree. kill related to defoliation by the Douglas-firGen. Tech. Rep. PNW-164. USDA Forest tussock moth in the Blue Mountains out-Service, Pacific Northwest Forest and Range break. Res. Pap. PNW-233. USDA ForestExperiment Station, Portland, OR. Service, Pacific Northwest Forest and Range

Molisch, H. 1938. The Longevity of Plants. Experiment Station, Portland, OR.E. H. Fulling, New York. Wright, K. H., and P. G. Lauterbach. 1958. A

Mueller-Dombois, D. 1983. Canopy dieback 10-year study of mortality in a Douglas-firand successional processes in Pacific forests. sawtimher stand in Coos and Douglas Coun-Pac. Sci. 37: 317-326. ties, Oregon. Res. Pap. PNW-27. USDA For- 1987. Natural dieback in forests. Bio- est Service, Pacific Northwest Forest and

Science 37: 575-583. Range Experiment Station, Portland, OR.

556 BioScience Vol. 37 No. 8September 1987 555