CONSEQUENCESOF TREE MORTAliTYTO THE GLOBALCARBONCYCLE

Mark E. Harmon, Sandra Brown,and Stith T. Gower

ABSTRACT

Tree mortality and accumulations of woody detritus are important, but unstudied facets of the global carbon (C) cycle.Ecological studies from undisturbed temperate and tropical forests indicate woody litter inputs associated with tree mortalityrange between 0.16-5.0 Mg Cjha/yr. Mortality rates in forests appear to be positively comlated with ecosystem productivityand time since disturbance. United States Forest Service Continuous Forest Inventory data indicate that tree mortality in

the United States increased over the pastfive decades from 52 to 65 Tg C/yr. However, mortality is increasing in some regions(eastem United States). while decreasing in others (Pacific Northwest) because of changes in forest-age structure. Globally,there are few measurements of tree mortality. Stand and Continuous Forest Inventory data were used to estimate tumoverrates of live woody biomass for forests. Applying these ratios to cumnt live biomass estimates indicates that 1.2-9.1 Pg C/yr is added globally to detrital pools by dying trees within intact forests. An additional 0.9-1.8 Pg of woody C may be addedto the detrital cycle by catastrophic disturbances. The combined dead tree input is 7-39% of the 28 Pg C/yr added by finelitterfall. At these rates of input, coarse woody debris would reach a steady-state mass of 60-232 Pg C. However, becausepast land conversions reduced mortality and woody detritus, considerable carbon could be sequestered in forests recoveringfrom clearing.

INTRODUCTION

Examination of any current global carbon bud-get (Post et al., 1990;Houghton and Woodwell,1989) reveals a very startling point: dead treesdo not exist! The tacit assumption inmost, ifnotall,global carbon budgets isthat although> 80%of the globe's livingbiomass iswoody(Woodwellet al., 1978),very little of this material becomeslitter. We will demonstrate that this assump-tion is unfounded and has led to a major under-estimate of global detrital stores.

Excludingdead trees has severalprofoundim-pacts upon understanding global carbon dy-namics. First, the sizeofthe terrestrial detritalpool has been underestimated (Harmon andChen Hua, 1991). ITwoody detritus was in

equilibrium, this might not be a major concern;however, land use changes have placed woodydetritus in disequilibrium (Schiffman and John-son, 1989). For example, loggingwithin PacificNorthwest forests has halved carbon storage inthis region (Harmon et al., 1990). Fully one-third of the difference between carbon stores in

a plantation forest and an old-growth forest isdue to the reduction of woody detritus (deadtree) carbon (Harmon et al., 1990). This reduc-tion is not unique to this region, but probablyapplies to most boreal, temperate and tropicalforests. IT this is true, past calculations ofcarbon flux from forest clearing are underesti-mates. Conversely, this would imply that alarge, unaccounted for carbon sink could beoccurring in forests recovering from past har-

167

CONSEQUENCES OF TREE MORTAUI'Y TO mE GLOBAL CARBON CYCLE

vest (Harmon and Chen Hua, 1991;Lugo andBrown,in press).

Another problem with excluding dead treesfrom global carbon budgets involves the futurepattern of detrital stores. Most assessments ofclimate change upon detrital stores are com-parisons of steady-state solutions under currentand projected future climates. These studiesgenerally indicate a greater storage of carbonunder a warmer climate (Chapin and Matthews,1992),primarily as a result ofhigher productivity.In contrast, analysis of transient responses toclimate change indicates a large pulse of carbonmay be injected into the atmosphere during thetransition between these two steady-states(Neilson, personal communication). To a largedegree, the temporal dynamics of this transientpulse willbe controlled by the decomposition ofwoodyplantskilledbycatastrophicdisturbances(Le., fire), or increased stress.

In the followingpaper, we will examine the roledead trees play in the current global carboncycle. Coarse and fine woody detritus areformed from the death of twigs,branches, tootsand boles of trees and other woody plants. Thediameter used to separate coarse- and fine-fraction woody detritus is 10 em. For thepurposes of this review, we will exclude fineroot turnover from our analysis, although thefine roots of many trees and shrubs are in factwoody and can exceed lead litterfall in someforest ecosystems (Vogt et al., 1986). We firstreviewthe characteristicsofwoodydetritus thatmake its behavior different from other forms of

litter. We then examine factors controlling treemortality at the stand and regional levels. Fi-nally, we estimate the input of carbon via dyingtrees to the global detrital system and examineits implications for the overall global carboncycle.

CHARACfERISTICS OF WOODYDETRITUS

The decay and input dynamics of woody detri-tus are quite distinct from those of leafy detri-tus. Decay rates of woody detritus can be morethan an order of magnitude less than leafy litter(Harmon et al., 1986). This has several impli-cations for successional dynamics and C stores.First, fine litter and woody litter will store equalamounts of carbon at steady-state if woodyinputs are an order of magnitude lower thanfine litter inputs. These differences in decayrate also implywoody litter will take more thanten times as long to reach steady-state onceinput rates have stabilized.

In addition to differences in decay rates, theperiod required to reach maximum input ratesfollowingdisturbance is considerably longerforwoody than leafy litter. Input rates for leafylitter usually peak at the time of crown closure,which is as little as five years in tropical anddeciduous forests (Bormann and Likens, 1979;Brown and Lugo, 1990) and up to 50 years inconiferous forests (Gessel and Turner, 1976).Incontrast,woodylitterinputs donotpeakuntilthe old-growth stage of succession, and this maytake > 100years even in tropical forests (Brownand Lugo, unpublished). This implies thatstores of woody detritus will not reach steady-state levels for at least a century followingdisturbance.

Despite the fact that disturbances, includinglogging, create a large quantity of woody de-tritus, it is not unusual for this component to beoverlooked in successional or ecosystem pro-cess models or carbon-sequestration assess-ments of climate change. We suggest that theconservation of mass applies even to dead trees.In the case of natural catastrophic disturbances(Le., wind and fire), woody detritus can be thelargest single carbon pool immediately follow-ing disturbance. In a Douglas-fir forest, forexample,woodydetritus comprised25%ofthetotal ecosystem carbon before a crown fire and75% afterward (Agee and Huff, 1987). Formost disturbances, this large increase is obvi-

168

Ponderosa pine

Lodgepole pine

Korean pine

Red spruce-Fir

Western White pine

Virginia i>ine

Silver-fir

Douglas-fir

Spruce Hemlock

Whi te fir

M.B. HARMON, S. BROWN, AND S.T. GOWER

o 2

I

. i

4 6 8 10

_ MEAN INPUT (Mg/ha/y)studies with> 5 ha-years of observation

Figure 1. Woody detritus input associated with tree death in conifer forests. Only mature to old-growth forest withat least five hectare-years of observation are displayed.

ously a result of the large mass in trees beingsuddenlytransferred to thewoodydetritalpool.However, in the case of fires, woody detritus isless apt to burn than leafy litter. This is es-pecially true for the coarse- fraction material,which even under the most severe fire condi-

tions is rarely burned completely (Sandbergand Ottmar, 1983).

STANDLEVEL STUDIES

Ecologistsand foresters have been measuringforestgrowthfor centuries;however,there arevery few published mortality measurementsbased on direct observation (Franklin et al.,1987). The most reliable data are from largepermanent plots (> 1ha area) withindividuallymarked trees observed for at least five years.Although numerous data sets meeting thesecriteria exist, many of them remain unpub-lished or analyzed solely from a populationperspective. It is hoped that this trend willnotcontinue.

Although it is unlikely stand level studies willever be numerous enough to estimate globalmortality rates directly, they are quite useful in

understanding the causes of mortality and howstand structure, age, and productivity affectrates. Perhaps the best set of stand level plotsavailable at this time is from the Pacific North-

west, a region where large-scale (>0.5 ha)permanent plots have been observed for morethan 20 years (Franklin et al., 1987). Woodydetritus inputs (including all tree parts exceptleaves) associated with tree death range anorder of magnitude from <0.16 Mg C ha/yr inponderosa pine to >4 Mg C ha/yr in white firforests of California (Figure 1). In contrast, finelitterfall in these stands only differs by a factorof two from 1-2Mg C/ha/yr(Vogt et aI., 1986).Both fine litterfall and tree mortality in matureto old-growth coniferforests appears positivelycorrelated with site quality. Plotting total above-ground NPP (net primary production, equal tothe sum of biomass increment, mortality andfine litterfall) against tree mortality for matureto old-growth forests indicates that there is ageneral correspondence (Figure 2). For a givenamount of above ground NPP, however, coniferforests produce far more woody detritus thandeciduous or tropical forests. In particular,Pacific Northwest conifer forests produce al-most three times the "expected" amount of tree

JII

..

! IIi

I

Ji

I,i

t!

169

CONSEQUENCES OFTREB MORTAUIY TO 'mE GLOBAL CARBON CYCLE

mortality litter. This difference may be attrib-uted to, in part, from the greater longevity, andthus biomass, of these forests; a greater amountof biomass may therefore offset lower NPP.

The lower mortality rate in less productiveforests does not necessarily mean that deadwoody detrital stores will be lower in theseforests. Dead wood stores in tropical forestsmight be quite s~ar to those in boreal forestsbecause productivity and decomposition ratesare positively correlated on a global scale-bothincreasing toward the tropics. There are excep-tions to this pattern as some temperate regions(Le., the Pacific Northwest) can also have highproductivity. Woody decomposition rates arestrongly influenced by fungi-toxic compoundsin heartwood and excess moisture, factors thatmay not be correlated to latitude. Few borealgenera contain fungi-toxiccompounds, whereasthey are extremely common in tropical hard-woods. Moreover, excess moisture may limitwood decay in some tropical forests, but it isfrequent in cool humid regions such as thePacific Northwest.

,-

Another factor that profoundly influences theamount of tree mortality is stand age (Figure 3).Although the number of stems dying is oftenhighest during the middle stages of succession(Knox et al., 1989), it would appear that woodylitter input increases with forest age (Harcombeet al., 1990). This discrepancy may result fromdifferences of population and detrital inputdynamics. The large number of stems dyingduring the middle stages of succession are smallindividuals with little mass. In contrast, fewertrees die in older forests, but they are muchlarger and not biased toward the smallest indi-viduals of the population (Knox et al., 1989).

The non-linear patterns of tree mortality inputsduring succession, coupled with slow decompo-sition, mean that woody detrital stores rarelyreach a steady-state. Disturbances add a largeamount of woody detritus, yet the followingyoung forests do not produce much woodydetritus. This leads to a large decrease in woodydetrital stores early in succession that mayequal or exceed the carbon sink created byregrowing trees (Harmon et al., 1990). If mor-tality rates remain low during the middle stages

170

TREE MORTALITY (Mg/ha/year)10

8 -------- 00

6 ______m __.____ ----- u0

0

4 I- ._-- ---- --. 0

0

2 I- --- --8-;- 0DoO

Z X0 I ,

00 5 10 15 20 25 30

ABOVE-GROUND NPP (Mg/ha/year)

0 CONIFER Do MIXED x DECIDUOUS o TROPICAL

F1gUre 2. Woody detritus input associated with tree death versus above-ground Net Primary Production (NPP).The NPP was calculated as the sum of biomass increment, littedall and tree mortality.

..

o , ;..;. j;.;um u u ;--......o

Mortality (Mg/ha/year)14

12

10

8

6

4

M.B. HARMON. S. BROWN. AND S.T. GOWER

2

100 120 14020 40 60 80

Time (years)after Harcombeet al. (1990)

. mean - - 1 std + 1 std

F'IgUI'e3. Changes in tree mortality as a function of stand age for a western hemlock/ sitka spruce ecosystem. Two series

of permanent plots were used.

of succession, woody detrital stores may de-crease below old-growth values (Spies et al.,1988). Eventually, the mortality rates return toold-growth amounts, leading to an additionalsink of carbon in the ecosystem. For example,in the sitka spruce/western hemlock forest,another 100 to 150 Mg C/ha may be stored indead wood in the next 200 to 300 years. Thisamounts to an additional sink of 0.5 Mg C/ha/yr in a forest that has reached a "steady-stateUliving biomass (Harcombe et al., 1990).

REGIONAL STUDIES

Stand level ecological plots are not numerousor representative enough to be unbiased esti-mates of mortality rates on a regional or conti-nental scale. Timber inventories generally fitthese criteria; however, few countries appar-ently include non-catastrophic mortality. Thisprobably results from a methodologicallimita-tion because permanent plots are needed toestimate mortality, and most timber invento-ries are one-time surveys. An exception is theU.S. Forest Service Continuous Forest Inven-

Table 1. Expansion factors used to convert the volume ofgrowing-stockmortality from Waddell et ale(1987)to totalwoody mortality.

'Expansion factor Conifer Deciduous

Growing stock to 1.053 1.667

Total volume (a)

Wood density (Mg/m3) 0.45 0.55

Wood to Total mass (b) 1.62 1.62(a) Baocd upoD miDI or lOcal 10 puwiD& IIOCIr.""Iume (BecII1old, 1984).

(b) JDdudel but, bl'8DCbel, 1'00II.

tory (CFI), which has recorded mortality inpermanentlymarkedplotssincethe 1950's.Wehave converted these data (Waddell et aI.,1987) from loss in cubic volume of growingstockto carbonusingvariouscorrectionfactorsfor non-merchantable trees, wooddensityandnon-merchantable parts (Table 1). We havealsousedthe fme-litterinputrates estimatedbyMeentemeyer et al. (1982) and the area inforest land (Waddell et al., 1987)to estimatefine-litter inputs for the same area examinedfor tree mortality.

171

9

1:"I ,.,;'...

.,

,./.~,

I

I

IJ!;

CONSEQUENCES OF TREE MORTALITY TO TIIE GLOBAL CARBON CYCLE

Total Input USA (Tg carbon/year)70

60

50

40

30

20

10

o1952 1962 1970 1976

Inventory Year

1986

_ conifers _ hardwoods

Figure4. Woody detritus input associated with tree death in U.S. forests in the past 50 years.

Continuous Forest Inventory data indicate thatthe rate of tree mortality in the United Stateshas increased in the last five decades from 52 to

65 Tg C/year (Figure 4). This is < 10% of thetotal above-ground inputs, which at first glancewould support the notion that tree mortalityinput could be overlooked with little conse-quence. This perspective is very misleading,however, because woody litter has a very slowdecay rate and consequently can form a largeamount of detrital stores. For example, ap-plyingthe known differences in leafyand woodydecay rates for two typical forests (i.e., an east-ern deciduous and a cold coniferous forest) wefind that woody detritus comprises 40 to 50% ofthe above-ground detrital stores in a steady-state condition.

The cause of the slight increase in tree mortality. in the United States is difficult to unravel giventhe large-scale changes in pollution stress andage-structure that have occurred. Examinationof regions indicates that the overall increase isdriven by increases in the eastern United Statesthat offset large decreases in the Pacific North-west. We feel this pattern parallels shifts inforest age structure; regions with decreasingmortality rates are being converted to younger

forests, whereas regions with increasing mor-tality rates are generally increasing in age.

Regardless of the current causes of increasingmortality, it is important to note that currenttree mortality rates are far below pre-Europeanlevels. The effect of forest harvest on mortalityrates is most clearly seen in Oregon and Wash-ington, a region where the area extent of old-growth forests has decreased in the last 50years. The CFI data show mortality has de-clined two-fold during this period (Figure 5).Extrapolating back to the turn of the century,when these forests were first beginning to bedeveloped, indicates there has been a three- tofive-fold decrease in the mortality rate.

GLOBALESTIMATES

Ideally, one would base global es~imates of treemortality upon forest inventories; however, un-til more government agencies collect or publishthese types of data, direct estimates will not bepossible. We have therefore used an alterna-tive approach to estimate mortality in intactstands based upon the ratio of mortality andgrowing stock volume (Figure 6). This ratioamounts to a turnover rate for the living canopy

172

M.E. HARMON, S. BR.OWN, AND S.T. GOWER.

Mortality ( Tg Organic MaUer)60 --

50 -----

- --.40

30

20 '- =~_:h ::-~-=::h"_h_=:::~

10 ~ ~-- --- - - -- -- - -~-~ -.---

oo 40 50 60

Year70 80 90 10010 20 30

- CFI data ~ mU(blum ,.etmate minimum estimate

based upon Waddell et a1.(i§S7) and Hiitnwft et fit. (1986)

Figure 5. Estimated historical changes in tree mortality during the last 00yeus in the Douglas-fir region of Oregon andWashington.

that can be multiplied bythe livingbiomass datato crudely estimate the mass of moruility. Notethat in our current estimates, forest age strue..ture is not used; more accurate estimates couldbe made if age-class specific turnover rates

wereusedinconjunctionwithage-classspecificbiomassdata. For the temperate and borealbiomes, we used the U.S. CFI data, and for1rumfdtropical biomes,weused data from Brown

_ Mortality (million cubic feet/year)350 - -

300~ n- ..----

250 -- ----

100

- __0--

200 --- - - .-- - -------

150 ~- -----

-

o ~ = =. c o50

oo 5 10 15 20 25 30

Growing Stock Volume (billion cubic ft)35

o CFI DATA cold conifer - warm coniferbased on CFI data (Waddell et aI., 1987)

FJgU1'e6. Cubic mortality of growing stadt versus growiDg stadt for conifers in the United States. Each point represents

173

"

J,I

CONSEQUENCES OF TREE MORTAUTY TO mE GLOBAL CARBON CYCLE

tions: eastern pine-type species with 0.95%jyrand cold coniferous forests with 0.43% jyr. Thisis consistent with the stand level results thatindicate higher mortality with increasingpro-ductivity. The opposite appears true for de-ciduousforests in that cooler, less productiveforests have a higher turnover rate (0.87jyr)than warmerdeciduousforests (0.64jyr). Thispattern mayreflect the overall advantage thatconifershaveincoolerclimates. Alternatively,this pattern may reflect the switch in succes-sional status of conifers and deciduous treeswith latitude. That is, conifers are generallyshort-livedearlyseral species in lowlatitudes,whereasdeciduoustrees playthis role inhigherlatitudes.The tropicalVenezueladata indicatethat both dry and moist tropical forests haveperhapstwicethe rate ofturnoveroftemperateforests, suggestinga positive correlation be-tween turnover rate and productivity.

(a) Cakulated as the moo between the DIISS avdumc cillD11ality aI¥I the

grovqsta:k DIISSavdumc.

(b) 8Iscd upca U; Rxcst Service Q1 data fairxlividua1 states; r.mge is the

bvcstal¥lhighcstvaluc fastates in given f<nst type.

(c) 8Iscd upcastaOO data franlbvnal¥l Lugo(l1~

and Lugo (unpublished) to estimate these turn-over rates. The living biomass data were takenfrom Woodwell et al. (1978) and Brown et al.(1989).

Examiningmortality turnover rates from fivedifferent regions indicates considerablevaria-tion within and between regions (Table 2).Coniferous forests have two distinct popula-

Table 3. Global detrital input associated with tree mortality.

."-- .- -- -----

Applying these ratios to the biomass data indi-cates that globally 1.2-9.1 Pg Cjyr is added todetritus via tree mortality within intact stands(Table 3). Byfar the largest amount appears to

(a) Based upon WoodweU et aI., 1978; Brown et aI., 1989.

(b) Mean (minimum-maximum), calculated using the mean, minimum and maximum

mortality turnover rates Table 2.

(c) Calculated by dividing the range of return intervals in years into the live biomass.

(d) Based on Sanford et aI., 1985.

(e) Based upon Kilgore (1978) and Christensen (1978).

(f) Based upon Seischab and Orwig, 1991.

(g) Based upon Heinselman, 1978.

(h) Based upon Lanly, 1982.

174

Table 2. Mortality turnover rates of major forest regionsof the world.R:giJt Mx1aity tuux:wenate(%/yr) (a)

mean mininun 11DXimum

. ICokJcalifer (b)

0.43 0.17 1.00

CokJdccidlxu (b) 0.87 0.52 1.48

Wanncalifer(b) 0.95 0.52 1.28

Wanndccidlxu (b) 0.64 0.42 0.93

ITropcal ck:sc (c)

1.58 0.13 4.93

ITrq:ical qJCII(c) 0.22 0.113 0.35

Catastrophic Mortality Input

Ecosystem Area (a) Live carbon (a) return interval Normal (b) Catastrophic (c)

(106 km2) (Pg) (centuries)

ropical closed 12.0 (h) 122 7-15 (d) 1.91 (0.16-6.04) 0.08-0.18

ropical open 7.3 (h) 26 7-15 (d) 0.06 (0.02-0.1) 0.02-0.04

emperate evergreen 5.0 79 2-5 (e) 0.75 (0.41-1.01) 0.16-0.39

emperate deciduous 7.0 95 10-15 (f) 0.61 (0.40-0.88) 0.06-0.09

Boreal 12.0 108 1-2 (g) 0.46 (0.18-1.08) 0.54-1.08

otal 48.5 743 3-7 3.79 (1.17-9.11) 0.86-1. 78

be dyingin closed tropical forests that have bothhigh turnover rates and high biomass stores(due to extensive areal extent). Boreal forestsare less important globally, but still may beadding as much as 1Pg C/yr from intact stands.

These estimates are low, of course, as they donot include contributions from catastrophicdisturbances. Unfortunately, there are fewestimates of natural rates of catastrophic dis-turbance on a global scale upon which to baseour calculations. We approach this problem intwo ways. First, one might consider the returninterval required to have catastrophic mortalityequal to within stand mortality. This wouldindicate whether setting non-catastrophic andcatastrophic mortality rates equal is a reason-able assumption. The return interval would bethe reciprocal for the estimated turnover ratesof intact stands. This indicates a catastrophic-disturbance return interval as short as 43-200

years in tropical forests and as long as 200-600years in boreal forests. The return interval forboreal forests seems reasonable in that it im-

plies 2 to 6 x lQ6ha/yr are disturbed, less thanthe 8 x lQ6ha/yr estimated for a decade withhigh fire occurrence (Stocks, 1990). In contrast,the return intervals for temperate and tropicalforests is unrealistically short. A more reason-able estimate from these regions can be madeby using published return intervals (Table 3).This indicates catastrophic disturbances maybe adding another 0.85 to 1.78 Pg/yr globally.Thus total tree mortality input for the globewould appear to be in the range of2.03-1O.89PgC/yr.

Global estimates of fine litterfall (Meentemeyeret al., 1982) are about 28 Pg/yr, thus woodydetrital inputs are relatively small comparedwiththatwell-knowninput.Nonetheless,woodydetrital flowsare importantin tworegards. The

- -- - - --

M.E. HARMON, s. BROWN, AND S.T. GOWER

first is that this flow is about equal to fossil fuelburning. The idea that flowsthis large are beingignored, as ecologists "balance" the global car-bon budget (Post et aI., 1990), is mind-boggling.The second point is that while the input ratemay be small relative to leafy litter, the storagein woody detritus may far exceed that of leafylitter because of the much lower decompositionrate of woodymaterial. We have calculated thepotential steady-state carbon store in woodydetritus using the range of input rates (Table 3)and estimated decay rates (Le.,tropical = 10%/yr, temperate.evergreen = 3%/yr, temperatedeciduous = 7.5%/yr, andboreaI2%/yr). Thiscalculation indicates that given an input of 2 to11Pg C/yr, a steady-state store of 64 to 232 Pgof woody detritus carbon results. Ecologistshave thus been ignoring a detrital pool that atleast equals and most likely far exceed the 70Pgof fine litter carbon (Post et aI., 1990).

CONCLUSIONS

Our estimates of the detrital input associatedwith tree mortality are preliminary, and to anorder of magnitude only. They do indicate,however, that a major flow in the global carboncycle has been ignored. We question whetherthe global carbon budget can be balanced whenflows as large 2-10 Pg C/yr are ignored. Ouranalysis indicates global detritus may be under-estimated by an amount roughly equal to thecarbon stores in peats. Given that past forestharvest and clearing have greatly reduced woodydetrital stores globally, this pool may have beena significant source of carbon in the past and apotentially significant future carbon sink in thepresent. However, until we have better esti-mates of tree mortality and woody detritaldecay rates, the exact contribution from woodydetritas to the global carbon cycle will remain amystery.

175

......

REFERENCES

CONSEQUENCES OF TREE MORTALITY TO 1HE GLOBAL CARBON CYCLE

Agee, J.K and M.H. Huff. 1987. Fuel succession in awestern hemlockjDouglas-fIr forest. Canadian Journal ofForest Research 17: 697-704.

Bechtold,WA. 1984.Forest statistics for North Carolina.Resource BulletinSE-RB-78, US. Forest Service,Ashville,NC, p. 60.

Bormann,F. H. and G.E. Likens. 1979. Pattern andProcessin a Forested Ecosystem.Springer-Verlag,Inc.,NewYork.

Brown, S. and A.E. Lugo. 1991. Production of coarsewoody debris and its role in organic matter budgets intropiqi!.forests of Venezuela. Journal ofTropical Ecology(submitted)

Brown,S. and AE. Lugo. 1990. Tropical secondaryforests. Journal of TropicalEcology6:1-32.

Brown, S. and AJ.R. Gillespie, and AE. Lugo. 1989.Biomass estimation methods for trpoical forests wilappli-cation to forest inventory data. Forest Science35:881-902.

Chapin, F.S.and E. Matthews. 1992.Boreal carbon pools:approaches and constraints in global extrapolations. (thisvolume).

Christensen, N.L. 1978. FIre regimes in southeasternecosystems. In: FIre regimes and ecosystem properties.General Technical Report WO-26, US Forest Service, pp.112-136.

Franklin, J.F., H.H. Shugart, and M.E. Harmon. 1987.Tree death as an ecological process. Bioscience 37:550-556.

Gessell, S.P. and J. Turner. 1976. Litter production inwestern Washington Douglas fIr stands. Forestry 49:63-72.

Harcombe, PA., M.E. Harmon, and S.E. Greene. 1990.Changes in biomass and production over 53 years in a

. coastalP;ceas;tchens;s- Tsuga heterophyl/a forest approach-ing maturity. Canadian Journal of Forest Research20:1602-1610.

Harmon, M.E., J .F.Franklin, FJ. Swanson,P. Sollins,S.V.Gregory, J.D. Lattin, N.H. Anderson, S.P. Cline, N.G.Aumen, J.R. Sedell, G.W. Lienkaemper, K Cromack Jr.,and K W. Cummins. 1986. Ecology of coarse woodydebris in temperate ecosystems. Advances in EcologicalResearch 15:133-302.

Harmon, ME., W.K Ferrell, and J.F. Franklin. 1990Effects of carbon storage of conversion of old-grov.othforests to young forests. Science 247:699-702.

Harmon M.E. and C. Hua. 1991. Coarse woody debrisdyn~ics in twoold-growthecosystems. Bioscience41:6Q4..610.

Heinselman, M.L. 1978. rIre intensity and frequency asfactors in the distribution and structure of northern ec0-systems. In: rue regimes and ecosystem properties.General Technical Report WO-26, US Forest Servite, pp.7-57.

Houghton, RA. and G.M. Woodwell. 1989. Globalclimate change. Scientific American 260:36-44.

Kilgore, B.M. 1978. FIre in ecosystem distribution andstructure: western forests and scrublands. In: FIre re-gimes and ecosystem properties. General Technical Re-port WO-26, US Forest Service, pp. 58-69.

Knox, R.G., R.K Peet, and N. Christensen. 1989. Popu-lation dynamics in loblolly pine stands: changes in skew-ness and size inequality. Ecology 70:1153-1166.

Lanely, J.P. 1982. Tropical forest resources. FAQForestryPaper 30. FAO, Rome. 106pp.

Lugo, AE. and S. Brown Tropical forests as sinks ofatmospheric carbon. in press. Forest Ecology and Man-agement.

Meentemeyer, V., E.O. Box, and R. Thompson. 1982.World patterns and amounts of terrestrial plant litterproduction. Bioscience 32:125-128.

Neilson, R. 1992. Personal communication.

Post, W.M., T.H. Peng, W.R. Emanue~ AW. King, V.H.Dale, and D.L. DeAngelis. 1990. The global carbon cycle.American Scientist, 78:310-326.

Sandberg,D.V.andR.D.Ottmar. 1983.Slash burning andfuel consumption in the Douglas-fIr subregion. SeventhConference on rIre and Forest Meterology, Fort Collins,Co. pp. 90-93.

Sandford, R.L., jr., J. Saldarriaga, KE., C. U~ and R.Herrera. 1985.Amazon rain-forest fIres. Science 227:53-55.

Schiffman, P.M. and W.C.Johnson. 1989. Phytomass and

detrital carbon storage during forest regrowth in the

176

southeastern United States. Piemont. Canadian Journalof Forest Research 19:69-78.

Seischab,F.K. and D. Orwig. 1991. Catastrophic distur-bancesin the presettlement forests of western New York.Torrey Botanical Club 118:117-122.

Spies, T A., J .F. Franklin, and T.R Thomas. 1988. Coarsewoody debris in Douglas-fIr forests of western Oregon andWashington. Ecology 69:1689-1702.

Stocks,BJ. 1990. The extent and impact of forest fIresin northern circumpolar countries. In: Atmospheric,climatic and biospheric implications, Proceedings of theChapman Confernece on global biomass burning.

_ --- ----

M.E. HARMON, S. BROWN, AND S.T. GOWER-Vogt, K.A., C.C. Grier, and DJ. Vogt. 1986. Production,turnover, and nutrient dynamics of above- and below-ground detritus of world forests. Advances in EcologicalResearch 15:302-377.

Waddell, K.L., D.D. Oswald, and D.S. Powell. 1987.Forest statistics of the United States. Resource BulletinPNW-RB-I68, U.S. Forest Service, Portland, OR, 1089.

Woodwell, G.M., R.H. Whittaker, WA. Reiners, G.E.Likens, C.C. Delwiche, and D.B. Botkin. 1978. The biotaand the world carbon budget. Science 199:141-146.

177

,I

.I,~

J

.1

I

jJ

I