Successional diversity and forest ecosystem function

INTRODUCTION

Over the past decade numerous experiments havebeen conducted to examine the relationshipbetween species diversity and ecosystem function.In most ecosystems, the relationship betweendiversity and productivity has been shown to bepositive (Tilman et al. 1997). In a few ecosystems,however, no significant relationship has beenobserved, while in others increasing species diver-sity has been shown to decrease productivity(Hooper & Vitousek 1997). These divergentresults have stimulated a vociferous debate thatrevolves around a single question: what factorsdetermine whether the relationship between diver-sity and productivity is positive?

Recent syntheses of the experimental resultssuggest that a positive relationship between diver-sity and productivity depends on a positive rela-tionship between productivity and competitiveability (Kinzig et al. In press). If a species’ abilityto maximize carbon gain in a particular environ-

ment also confers a competitive advantage in thatenvironment, competition will maximize produc-tivity by favoring the most productive species.Thus, a polyculture will be more productive thana monoculture because the most productive specieswill prevail in each environment within the ecosys-tem. However, there may be exceptions in whicha species’ ability to maximize carbon gain does notconfer competitive dominance.

In forests, for example, fast-growing early suc-cessional species are replaced by slower growinglate successional species (Horn 1974). Thus, pro-ductivity is not maximized by competition becausestands dominated by slow-growing late succes-sional species fix less carbon than forests domi-nated by fast-growing early successional species.Nor is productivity necessarily enhanced by suc-cessional diversity; stands containing a mix of earlyand late successional species may fix less carbonthan stands dominated by a single fast-growingearly successional species.

Forests have scarcely been mentioned in the biodiversity debate due to the difficulty of con-ducting experiments in ecosystems with slowdynamics. Yet, there is a vast store of observationaldata on forests that can be used to examine whetherforest ecosystems differ in some fundamental way from other kinds of ecosystems. While

Ecological Research (2001) 16, 895–903

Successional diversity and forest ecosystem functionJohn P. Caspersen* and Stephen W. Pacala

Department of Ecology and Evolutionary Biology, Princeton University, Princeton, New Jersey,United States of America

Forest inventory data was used to examine the relationship between successional diversity and forestecosytem function. The inventory data show that stands composed of early successional species aremore productive than stands composed of late successional species, whereas stands composed of latesuccesssional species have lower turnover than stands composed of early successional species. Takenalone, these results would suggest that forests should be managed in a way that favors the most pro-ductive early successional species or longest-lived late successional species, depending on whetherthe goal is to maximize productivity or maximize carbon storage. However, the inventory data alsoshow that stands with low successional diversity fix and store less carbon than stands with high suc-cessional diversity. This result suggests that forests should be managed in such a way as to retainspecies diversity while also favoring species that maximize the ecosystem function of interest.

Key words: carbon storage; diversity; forest; productivity; succession.

*Author to whom correspondence should beaddressed. Email: [email protected]

Accepted 15 October 2001.

observational data cannot be used to establishcausality, they can be used to document whetherthe relationship between diversity and productiv-ity is positive, negative, or neutral.

In this paper, we use inventory data to examinerelationships between successional diversity, pro-ductivity, and carbon storage in forests. To moti-vate the analysis, we first present a simple nullmodel that is similar to previous Markov modelsof forest succession (Horn 1975; Van Hulst 1979;Binkley 1980). The null model predicts that pro-ductivity is highest in low diversity stands of earlysuccessional species, and that carbon storage ishighest in low diversity stands of late successionalspecies. We then analyze the forest inventory datato evaluate the predictions of the null model. Con-trary to the predictions of the null model, analy-ses showed that stands with low successionaldiversity fix and store less carbon than stands withhigh successional diversity.

NULL MODEL

Consider a forest that consists of a mosaic ofpatches at different stages of succession, each patchbeing dominated by a single tree belonging to oneof n species. To model the successional dynamicsof this mosaic, we assume that there is a simplecompetitive hierarchy among the n species. Upondeath, an individual of species 1 is replaced by anindividual of species 2, which is replaced by species3, and so on up to species n, which is replaced byspecies 1. Thus, if X1 is the fraction of patchesoccupied by species 1, the rate of change for species1 is:

(1)

where M1 is the mortality rate for species 1 and Mn

is the mortality rate for species n, the last speciesin the successional sequence. Similarly, the rate ofchange for any of the remaining species is:

(2)

where Xi is the fraction of patches occupied byspecies i, X(i – 1) is the fraction of patches occupiedby the preceding species in the successionalsequence, and M(i – 1) is the mortality rate for thepreceding species in the successional sequence.

dX dt M X M Xi i i i i= - + -( ) -( )1 1

dX dt M X M Xn n1 1 1= - + ( )

896 J. P. Caspersen and S. W. Pacala

Equations 1 and 2 can be expressed in matrixform as follows:

(3)

This system of equations can then be solved toobtain the equilibrium abundance of each species,expressed here in terms of species longevity. First,we express the average lifespan of each species asthe reciprocal of its mortality rate: Li = 1/Mi. Then,replacing the left-hand side of equation 3 withzero, it is easy to show that the area occupied byspecies i is proportional to its longevity:

(4)

where SL is the total length of the successionalsequence.

We can also solve for the equilibrium produc-tivity, mortality and storage, provided we maketwo simplifying assumptions about the age-dependence of productivity and mortality. First, ifwe assume that production is constant with respectto tree age, then the productivity of the succes-sional mosaic is calculated as a simple area-weighted average:

(5)

where Pi is the productivity of species i, and the productivity of species i is weighted by thefraction of area occupied by species i; that is, Li/SL.

PL

Pi

L

ii

n

∫=

Â S1

X Li i L* = S

dX dt

dX dt

dX dt

dX dt

M M

M M

M M

M

X

X

X

n

n

n

1

2

3

1

1 2

2 3

1

2

3

0 0

0 0

0 0

0 0 0

∑∑∑

È

Î

ÍÍÍÍÍÍÍÍ

˘

˚

˙˙˙˙˙˙˙˙

=

- ∑ ∑ ∑- ∑ ∑ ∑

- ∑ ∑ ∑∑ ∑ ∑ ∑ ∑∑ ∑ ∑ ∑ ∑∑ ∑ ∑ ∑ ∑

∑ ∑ ∑ -

È

Î

ÍÍÍÍÍÍÍÍ

˘

˚

˙˙˙˙˙˙˙˙

∑∑∑∑

È

Î

ÍÍÍÍÍÍÍÍ

˘

˚

˙˙˙˙˙˙˙˙

Xx

Second, if we assume that mortality is randomwith respect to age, the age distribution of treespecies i is:

(6)

the total biomass stored in trees of species i is:

(7)

the total biomass stored in the mosaic is:

(8)

and the fraction of biomass lost to mortality is:

(9)

because the mortality rate is equal to productivityP at equilibrium.

To illustrate the predictions of this model, weuse a five-species system and assume an inverserelationship between longevity and productivity.From species 1–5, longevity increases from 20 to100 years and productivity decreases from 3 to 2 tons ha–1 year–1. Later we show that theseassumptions are qualitatively consistent with theinventory data. Here we simply note that byassuming longevity increases from the bottom to

MP

PL∫

PLL

P Li

Li

n

i i∫=

Â S1

q a P a da P Li i i i( )( ) =

•

Ú0

q aL

ei

i

a

Li( ) =-1

the top of the competitive hierarchy, we are alsoassuming that competitive ability is correlatedwith longevity.

In a five-species system, there are 31 possiblecombinations of species, including five differentmonocultures, one five-species polyculture, fivefour-species polycultures, 10 three-species poly-cultures, and 10 two-species polycultures. Figure 1shows the productivity and total biomass for each of the 31 possible mosaics. The commonpattern in both of these graphs is that the 31mosaics form a triangle of points in which theupper boundary decreases with increasing diver-sity. The reason is that and are averaged community values calculated by weighting thecontribution of each species by its equilibriumabundance. Thus, the average community valuesfor polycultures are necessarily intermediatebetween the extreme values for monocultures. Forexample, the species 1 monoculture is more pro-ductive than any of the polycultures simplybecause species 1 is the most productive species.Conversely, the species 5 monoculture is less pro-ductive than any of the polycultures becausespecies 5 is the least productive species. The tri-angular pattern observed in Fig. 1 holds true nomatter how the P and L values vary among thespecies.

To facilitate comparison with the inventorydata, we now divide the 31 mosaics into early andlate successional categories based on the averagelifespan L of the species in each mosaic:

PLP

PLP

Successional diversity and forest ecosystem function 897

Fig. 1. (a) Equilibrium productivity P–

for each of the 31 successional mosaics. The highest value in the upper left-hand corner represents the species 1 monoculture. The lowest value in the lower left-hand corner represents thespecies 5 monoculture. (b) Equilibrium biomass PL

—for each of the 31 successional mosaics. The highest value in the

upper left-hand corner represents the species 5 monoculture. The lowest value in the lower left-hand corner repre-sents the species 1 monoculture.

(10)

where SLi is the total timespan of the successionalsequence and N is the number of species in the suc-cessional sequence. The average lifespan is longestin mosaics composed of long-lived late successionalspecies. Thus, the late successional categoryincludes mosaics in which the average lifespan is> 60, and the early successional category includesmosaics in which the average lifespan is £ 60.

To illustrate how productivity and mortalityvary with successional composition, we calculatethe average and for both of these categories.Early successional mosaics are more productivethan late successional mosaics (Fig. 2a); however,early successional mosaics also have a higherturnover than late successional mosaics (Fig. 2b).Later we show that these patterns are qualitativelyconsistent with the inventory data.

To illustrate how productivity and mortality varywith successional diversity, we further divide thesetwo categories into four categories: (i) early succes-sional with low diversity; (ii) early successional with

MP

LL

N

ii

N

= =Â

1


high diversity; (iii) late successional with low diver-sity; and (iv) late successional with high diversity.The low diversity category includes mosaics withone or two species and the high diversity categoryincludes mosaics with more than two species.

Figure 3a,b shows the average productivity and mortality for each of these four categories.The low-diversity early successional mosaics aremore productive than either of the high diversitymosaics (Fig. 3a). Similarly, the low-diversity latesuccessional mosaics have a lower turnover thaneither of the high diversity mosaics (Fig. 3b). Thenet result of these differences in productivity andmortality is that the low-diversity late successionalmosaics store more carbon than either of the highdiversity mosaics (Fig. 3c).

If correct, the null model would have importantimplications for the biodiversity debate and themanagement of forests as carbon sinks. First, thenull model suggests that forest productivity wouldbe maximized by planting monocultures of themost productive early successional species. Second,the null model suggests that carbon storage wouldbe maximized by planting monocultures of thelongest-lived late successional species.

MP

Fig. 2. (a) Average produc-tivity of early and late succes-sional mosaics. (b) Averagemortality of early and late suc-cessional mosaics.

Fig. 3. (a) Average productivity, (b) mortality, and (c) biomass of successional mosaics in each of four categories:(i) early successional with low diversity; (ii) early successional with high diversity; (iii) late successional with lowdiversity; and (iv) late successional with high diversity.

METHODS

To test the predictions of the null model, we ana-lyzed inventory data obtained from the ForestInventory and Analysis (FIA) database. The FIAdata are collected by the United States ForestService using a standardized methodologydescribed by Hansen et al. (1992) and Birdsey andSchreuder (1992). We will now give a briefdescription of the FIA sampling methods and amore detailed description of the methods we usedto analyze the data.

Sampling methods

Inventories are performed separately for each stateand follow a two-phase sampling procedure knownas double sampling for stratification. In the firstphase, a random sample of points is located onaerial photographs and classified by land cover andforest type. In the second phase, a subsample isselected from each of the classes or strata in thefirst-phase sample. This stratified subsample servesas the second-phase sample of plots that are visitedin the field.

For each plot, the classification is verified in thefield and a number of additional plot classificationvariables are recorded if the plot is forested, includ-ing stand age and stand origin (plantation vsnatural). Trees are sampled at a cluster of pointscovering an area of approximately 1 acre (2.47 ha).Trees between 2.5 cm and 12.7 cm in diameter atbreast height (d.b.h.) are sampled in a circularsubplot centered on each of the points. Trees wider than 12.7 cm d.b.h. are sampled at eachpoint using horizontal point sampling. For eachtree sampled, a number of observations arerecorded, including species, status (live, cut, ordead from natural causes), current d.b.h., and pre-vious d.b.h. (if the tree was measured in the pre-vious inventory). The data can be downloadeddirectly from FIA database(http://www.srsfia.usfs.msstate.edu).

Indices

We analyzed data from 24 670 natural forest plotsin 11 states, including Indiana, Michigan, Illinois,Wisconsin, Minnesota, Iowa, Missouri, Kansas,Nebraska, South Dakota, and North Dakota. For

each of the species occurring in this sample, we cal-culated a successional niche index:

(11)

where NIi is the niche index for species i, J is thenumber of plots in the sample, Kji is the numberof trees of species i with a crown in the canopy ofplot j, Ni is the total number of trees of species iin the sample, and Agej is the age of the stand inplot j. The niche index is a measure of a species’position in the competitive hierarchy of species.

For each plot in this sample, we then calculatedthe mean niche index of the trees occurring in thatplot:

(12)

where is the mean niche index for plot j and Sis the total number of species, NIj is a measure ofthe successional composition of a plot and is analo-gous to the average lifespan L calculated above.

Finally, for each plot we calculated the standarddeviation from the mean niche index:

(13)

where SNI is a measure of the successional diversityof a plot; that is, the farther the species in a plotdeviates from the mean niche index, the greater thesuccessional diversity of the plot. In other words,the SNI of a plot that contains a mix of early andlate successional species will be greater than the SNI

of a plot that is dominated by early successionalspecies or a plot that is dominated by late succes-sional species.

Growth, mortality and biomass

From the 24 670 plots analyzed earlier, we selectedstands > 30 years in age that had no evidence ofharvesting in the recent past. We excluded stands< 30 years old because they do not have closed

SNI NI K

KNI

i j jii

S

jii

S=-( )

-=

=

Â

Â

2

1

1

1

NI j

NINI K

Kj

i jii

S

jii

S= =

=

Â

Â1

1

NIK Age

Ni

ji ij

J

i

= =Â

1


canopies and would therefore introduce unduevariation in growth unrelated to the variables ofinterest. Stands with evidence of harvesting wereexcluded for the same reason.

For each tree in each of these plots, we calcu-lated aboveground dry biomass. The abovegroundbiomass was calculated using allometric equationsrelating tree mass to d.b.h. (Schroeder et al. 1997).The aboveground biomass of each living tree wasthen summed to obtain the aboveground biomassof each plot, following the methods of Caspersenet al. (2000).

We also used the methods of Caspersen et al.(2000) to calculate growth and mortality for eachof the remeasured plots. Growth is the change inbiomass density of living trees measured in the


first inventory that survived to the second inven-tory. Mortality is the biomass density of livingtrees measured in the first inventory that died bythe second inventory. Mortality is expressed on apercentage basis by dividing by the biomass of theplot at the time of the first inventory.

Data–model comparison

To facilitate comparison between the model pre-dictions, we divided the plots into the same cate-gories used earlier. First, we divided the plots intotwo successional categories: (i) plots composed ofearly successional species; and (ii) plots composedof late successional species. The late successionalcategory included plots with a mean niche indexNIj > 60, and the early successional categoryincluded plots with a mean niche index NIj £ 60.These categories are comparable to the categoriesused in Fig. 3a,b. We then further divided thesetwo categories into four categories: (i) early suc-cessional with low diversity; (ii) early successionalwith high diversity; (iii) late successional with lowdiversity; and (iv) late successional with highdiversity. The high diversity categories includedplots in which the standard deviation SNI was> 7.0, and the low diversity categories includedplots in which the standard deviation SNI was£ 7.0. For each of these four categories, we calcu-lated the average biomass, the average growth, andthe average mortality. These averages can be com-pared to the averages presented in Fig. 4.

RESULTS

Indices

The species included exhibit an approximatelynormal distribution along the successional nicheaxis (Fig. 4). For most species, the niche index NIi

falls between 30 and 90, indicating that thesespecies are most commonly found in standsbetween the age of 30 and 90 years. The plotsincluded show an approximately normal distribu-tion along the successional niche axis (Fig. 5). Formost plots, the average niche index NIj fallsbetween 40 and 80, indicating that the speciesfound in the plot are most commonly found instands between the age of 40 and 80 years.

Fig. 4. Niche index of each of the species in thesample, which contains 11 states.

Fig. 5. Average niche index of each of the stands inthe sample containing 11 states.

Successional composition

Figure 6 shows the average growth and mortalityof early and late-successional stands. Early succes-sional stands are more productive than late suc-cessional stands (Fig. 6a), whereas late-successionalstands have lower turnover than early successionalstands (Fig. 6b). These results are qualitativelyconsistent with the null model (Fig. 2).

Successional diversity

Figure 7 shows the average growth, mortality, andbiomass for each of the four categories used inFig. 3. These results contradict the null model in two ways. First, growth is highest in high-diversity early successional stands (Fig. 7a), not inlow-diversity early successional stands (Fig. 3a).Second, biomass is highest in high-diversity latesuccessional stands (Fig. 7c), not in low-diversitylate successional stands (Fig. 3c). Otherwise, Figs 3and 7 exhibit the same qualitative patterns.

Considered together, the growth and biomassdata suggest that species diversity enhances pro-

ductivity and carbon storage. Yet, there may bealternative reasons that inventory data are incon-sistent with the predictions of the null model. Forexample, low diversity plots may have experiencedmore selective harvesting prior to the first censusinterval. However, it is unlikely that correlationsbetween stand age and successional compositionare responsible for the discrepancies betweenFigs 3 and 7. Transient solutions of the null modelgive the same qualitative patterns seen in Fig. 7.Furthermore, restricting the data analysis toinclude only plots within 10 year age classes givesthe same qualitative patterns as those seen inFig. 7.

DISCUSSION

One of the dominant paradigms in forest ecologyis that the principal axis of functional variationamong tree species is a successional axis. Thus, itis commonly assumed that functional diversity issynonymous with successional diversity. Indeed,this assumption underlies many gap models as well


Fig. 6. (a) Average growth(± 2 SE) of early and late suc-cessional stands. (b) Averagemortality (± 2 SE) of early andlate successional stands.

Fig. 7. (a) Average growth, (b) mortality, and (c) biomass of forest stands in each of four categories: (i) early suc-cessional with low diversity; (ii) early successional with high diversity; (iii) late successional with low diversity; and(iv) late successional with high diversity.

as the null model presented in this paper. If thisassumption were true, then productivity would behighest in monocultures of the most productiveearly successional species, not in stands with highsuccessional diversity. However, the inventory datashow that stands with high successional diversityfix and store more carbon than stands with low successional diversity, regardless of successionalcomposition.

Why, then, are productivity and storage posi-tively correlated with successional diversity? Themost likely explanation is that there is another axisof functional variation among tree species that iscorrelated with the successional axis. Indeed, theindex of successional diversity SNI presented in thispaper is positively correlated with other indicesthat measure species diversity per se. For example,SNI is positively correlated with the number of treespecies Q with at least one tree in the canopy ofthe plot (r = 0.23, P < 0.05). Thus, it is possiblethat stands with high successional diversity aremore productive because there is additional func-tional variation among species that enhances theproductivity of high diversity stands.

To examine this possibility, we assessed whetherproductivity is positively correlated with diversityper se. Figure 8 shows the relationship betweengrowth and Q, a simple index of species diversity.Growth increases almost twofold from monocul-tures to the highest diversity polycultures. Thesedata suggest that there is in fact additional func-tional variation among species that enhances theproductivity of high diversity stands. Of course, it


is also possible that causality runs in the oppositedirection; that is, more productive stands maysimply permit the coexistence of more species.Unfortunately, cause and effect cannot be disen-tangled from observational data. Nonetheless, thedata do demonstrate a strong positive relationshipbetween productivity and diversity per se.

There are two primary conclusions that can bedrawn from the data analysis. First, successionalcomposition has a significant effect on forestecosystem function; that is, early successionalstands are more productive than late successionalstands, whereas late successional stands have alower turnover than early successional stands. Thisresult has clear implications for the managementof forests as carbon sinks. If the goal is to sequestercarbon in forest products, then forests should bemanaged in such a way as to favor fast-growingearly successional species over slow-growing latesuccessional species. On the other hand, if the goalis to sequester carbon in living trees, then forestsshould be managed to favor long-lived late suc-cessional species over short-lived early successionalspecies.

Second, successional diversity is positively cor-related with productivity, as is species diversity perse. If increased diversity does in fact enhance productivity, then this result also has importantimplications for the management of forests ascarbon sinks. In particular, this result wouldsuggest that forests should be managed in such away as to maintain species diversity, while alsofavoring species that maximize the function ofinterest. Conversely, if the correlation reflects thefact that more productive stands permit the coex-istence of more species, then this result challengesour understanding of the mechanisms that main-tain species diversity. Clearly, disentangling causeand effect presents a considerable challenge toforest ecologists interested in the biodiversitydebate.

ACKNOWLEDGEMENTS

This research was conducted under the auspices ofthe Carbon Modeling Consortium (CMC), whichis supported by the Office of Global Programs ofthe National Atmospheric and Oceanic Adminis-tration. Support was also provided by the Andrew

Fig. 8. Relationship between stand growth andspecies diversity per se.

W. Mellon Foundation. We thank Dr P. Moorcroftfor his help in processing the FIA data.

REFERENCES

B inkley C. S. (1980) Is succession in hardwoodforests a stationary Markov process? Forest Science26: 566–570.

B irdsey R. & Schreuder H. (1992) Anoverview of forest inventory and analysis estima-tion procedures in the eastern United States – withemphasis on the components of change. GeneralTechnical Report RM-214. USDA Forest Service,Rocky Mountain Forest and Range ExperimentStation, Fort Collins, CO.

Caspersen J., Pacala S., Jenkins J., HurttG., Moorcroft P. & B irdsey R. (2000) Con-tributions of land-use history to carbon accumula-tion in U.S. forests. Science 292: 2316–2320.

Hansen M., Frieswyk T., Glover J. & KellyJ. (1992) The eastwide forest inventory database:User’s manual. General Technical Report NC-151.USDA Forest Service, North Central ExperimentStation, St Paul, MN.

Hooper D. & V itousek P. (1997) The effects of plant composition and diversity on nutrientcycling. Ecological Monographs 68: 121–149.

Horn H. S. (1974) The ecology of secondary suc-cession. Annual Review of Ecology and Systematics 5:25–37.

Horn H. S. (1975) Markovian properties of forestsuccession. In: Ecology and Evolution of Communities.(eds M. Cody & J. Diamond) pp. 196–211.Belknap, Cambridge, MA.

K inzig A., T ilman D. & Pacala S. (In press)Biodiversity and Ecosystem Function. Princeton Uni-versity Press, Princeton, NJ.

Schroeder P., Brown S., Mo J., B irdseyR. & C ieszewski C. (1997) Biomass estimationfor temperate broadleaf forests of the United States using inventory data. Forest Science 43: 424–434.

T ilman D., Knops J., Wedin D., Reich P.,R itchie M. & S ieman E. (1997) The influenceof functional diversity and composition on ecosys-tem processes. Science 227: 1300–1302.

Van Hulst R. (1979) On the dynamics of vegeta-tion: Markov chains as models of succession. Veg-etatio 40: 3–14.


Documents

Successional diversity and forest ecosystem function