Group Title: Productivity and herbivory in high and low diversity tropical successional ecosystems in Costa Rica /
Title: Productivity and herbivory in high and low diversity tropical successional ecosystems in Costa Rica
Full Citation
Permanent Link:
 Material Information
Title: Productivity and herbivory in high and low diversity tropical successional ecosystems in Costa Rica
Physical Description: viii, 292 leaves : ill. ; 28 cm.
Language: English
Creator: Brown, Becky Jean, 1948-
Publication Date: 1982
Copyright Date: 1982
Subject: Herbivora -- Ecology -- Costa Rica   ( lcsh )
Plant succession -- Costa Rica   ( lcsh )
Ecological succession   ( lcsh )
Botany thesis Ph. D
Dissertations, Academic -- Botany -- UF
Genre: bibliography   ( marcgt )
non-fiction   ( marcgt )
Thesis: Thesis (Ph. D.)--University of Florida, 1982.
Bibliography: Bibliography: leaves 249-265.
General Note: Typescript.
General Note: Vita.
Statement of Responsibility: by Becky Jean Brown .
 Record Information
Bibliographic ID: UF00099501
Volume ID: VID00001
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: alephbibnum - 000334715
oclc - 09491282
notis - ABW4358


This item has the following downloads:

PDF ( 12 MBs ) ( PDF )

Full Text








This research was part of the University of Florida/CATIE

cooperative study, Natural Succession as a llodel for the

Design of New Tropical Agroecosystems. The research was

supported by NSF grants DEB 78-10721 and DEB 80-11136, Dr.

John J. Evel, Principal Investigator. A pilot study to

investigate herbivory measurement techniques was funded by a

Research Initiation and support (RIAS) grant from the

National Science Foundation, awarded by the Organization for

Tropical Studies. Data were analyzed using the facilities

of the Northieast Regyional Data Center, University of


I am grateful to Jack Ewel for invaluable guidance and

support throughout the project. Ariel lugo, Howard T. Odum,

Edward Deevey, and Dana Griftin provided encouragement and

many useful suggestions. I thank Hon Harrell for

collaboration in developing herbivory rate equations; Martin

Artavia L., Cory Berish, Chantal Blanton, Don Antonio Coto

M., Luis Coto n., Richard Hawkins, and Norm Price for

generous assistance in the field and laboratory; Grace

Russell fcr logistical support; Dawn Green, Laura Jimenez,

Chris McVoy, and Doris Pandolph for assistance in data

processing; and George Fuller for illustrations.


ACKNOWLEDGMENTS ... ... . ... .. .. . 1

ABSTRACT .. .. ... ... . ... .. .. .. vi


I. INTRODUCTION .. ... .. ... .. .. .. . 1

Belated Research .. ... .. .. ... 1
The Diversity-Stability Issue .. ... 1
Impacts of Herbivory ... .. .. .. 6
Direct impacts on net primary productivity 7
Impacts on species composition and
diversity .. .. .... .. . 8
Diversity Effects on Herbivory .......11
Research Questions .. . .. .. .. ... 12

II. METHODS .. . .. .... .. .. .. . . 13

The Study Site .. .. .. .. .. .. .. 13
Site Preparation .. .... . ... .. 15
Elain Treatments ... .. .. .. .. .. 16
Natural Succession .. .. .. .. ... 16
Mimic of Succession ... .. .. . ... 16
Enriched Succession ... .. .. .. .. 18
Successional Monoculture .. . .. ... 18
Plot Layout and Variables Measured ... .. 20
Measurements of Vegetation Structure .. .. 23
Leaf Area Index .. ... .. ... .. 23
Species Composition .. .. ... .. .. 24
Vegetation Height . ... .. .. . 25
Productivity Measur~emets .. .. ... .. 25
Above-Ground Biomnass . .. .. .. ... 28
Litterfall ... .. .. .. .. .. .. 30
Herbivory Rates .. . ... .. .. .. 31
Estimation of Hole expansion .. . ... 38
Subtreatments .. ... ... .. .. .. 45
Background Herbivory . .. ... .. .. 45
Decreased Herbivory .. .. ... .. .. 45
Increased Hertivory ... .. .. .. .. 47

III. RESULTS ... . .. .. .. .. .. .. .. .. 50

Vegetation Structure . ... .. .. ... 50
Species Composition ... . ... ... 50
Loaf Area Index .. ... .. .. .. .. 65
Herbivory Rates . .. ... .. ... .. 74
Above-Ground Biomass ... .. ... .. 111
Litter .. .. .. .. .. . .. .. ... 122
Above-Ground Productivity .. .. . ... 126
Effects of Decreased Herbivory . ... .. 133
Rates of Herbivory in Tasecticide Plots . 133
Species Composition .. ... .. .. 145
Leaf Area Index .. .. .... .. 150
Above-Ground Biomass ... . .. .. 156
Litterfall . . . . . . . . 162
Above-Ground Productivity 166
Responses to Artificial Defoliation .. .. 172
Results of Preliminary Study .. .. .. 172
Responses to Repeated Defoliation .. .. 174
Changes inl leaf productiviy .. .. 170
Changes in vegetation structure . .. 182
Changes in species composition . ... 188
Cassava biomass .. .. . ... .. 200

IV. DISCUSSION .. .. .. .. . .. . 202

Net Primary Productivity . ... .. 202
Relationship Between Net Primary Productivity
and Diversity .. .. . .. 202
Continuous Biomass Accumulation in Diverse
Systems .. .. .. .. . .. 205
Continuous Biomass Turnover in Diverse
Systems .. ... .. .. .. .. 210
Importance of Standing Dead Biomass . .. 210
Herbivory . .. .... .. .. .. .. 214
Low Herbivory Bates .. ... .. .. 214
Absolute Losses and Diversity Not Correlated 222
Percent Losses Correlated witu LAI . .. 224
Effects of Plant Species Composition .. 228
Plant Herbivore Defenses .. .. .. .. 232
Structural Complexity ... .. . 233
Herbivory, Diversity and Energy Flow .. .. 234
Energy Flow Elodel ....... ..235
Resilience of High and Low Diversity'
Ecosystems ... .. ... .. 241

LITERATURE CITED .. .. .... .. .. . ... 249



BIOGRAPHICAL SKETCH . ... .. .. .. .. .. .. 292

Abstract of Dissertation Presented to the Graduate Council
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy



Becky Jean Brown

December 1982

Chairman: Dr. Johni J. Ewel
Major Department: Department of Botany

Above-ground net primary productivity (NPP), herbivory

and vegetation structural characteristics were measured in

high and low diversity successional and agricultural

ecosystems at a vet tropical site near Tucrialta, Costa

Rica. Insecticide and defoliation experiments were

performed to evaluate the etfects of herbivory on NPP in

high and low diversity ecosystems.

The four expecimental ecosystems were enriched succession

(natural regeneration augmented by propaigule additions),

natural succession (control), successional minic (an

ecosystem with investigator-contro~lld species composition

designed to imitate natural succession), and successional

monoculture (two maize crops followed by cassava). Plant

species richness and leaf area index (LAI) were highest in

the enriched, high in the natural succession, intermediate

in the mimic, and low in the monoculture at 1.5 yr.

Net primary productivity, estimated from biomass

increments adjusted for turnover, was not related to

ecosystem complexity. The NPP was highest in the most

diverse (enriched) anid least diverse (monoculture) systems.

More than 82% of the above-ground production was lost

annually through litterfall, plant mortality and herbivory.

Standing dead biomass that did not fall into litter traps

accounted for a significant fraction of total turnover in

all ecosystems.

Herbivores consumed approximately the same amount of leaf

tissue per mZ of ecosystem in each of the three diverse

systems (54-61 cmr mR-2 ground day-i). Consumption expressed

as a percent of total leaf area was higher in the ecosystem

with lower LAI (the mimic). Absolute and percent losses

were lower in the monoculture than in the other ecosystems.

In the less diverse systems containing cultivars, herbivory

had high temporal variability. Species' herbivory rates

ranged from <1 to 131 cmr m-2 leaf day-i and appeared to be

related to palatability, ecosystem LAI and species


Herbivory stimulated NPP over a wide range of herbivory

levels in both the diverse system and the monoculture. The

stimulatory effect was greater, and maximum stimulation

occurred at a higher herbivory level, in the diverse system.

The resilience of the diverse system, due to compensatory

fluctuations in dominance of co-occurring species, has

important implicatious for agroecosystem design.



Complex traditional agroecosystems in the hiumid tropics

have persisted for many years without the use of pesticides,

while introduced monocultures have often been plagued by

pest attacks that lead to decreased crop productivity. The

magnitude of pest problems in an agrcecosystem may be

related to the degree of similarity between the

agroecosystem and the natural system it replaces. The

hypothesis is that the natural ecosystem possesses

structural and functional characteristics that allow it to

survive in its environment, and the more similar the

agroecosystem is to the natural system, the greater is its

chance for success. The objective of this study was to

investigate herbivory and primary productivity in ecosystems

structurally similar and dissimilar to a diverse tropical

successional system.

Related Research

The D~riversit-tabilijty Issu~e

In addition to the goal of maximizing production per unit

of energy input, tropical agriculturists are interested in

two other properties of agroscosystems: stability and

sustainability. A stable agroecosystem lacks fluctuations

in productivity (or variability in yield) over time, and a

sustainable agroecosystem has the ability to persist in the

face of perturbations (Conway 1982). Many complex

traditional agroecosystems have high sustainability and high

stability, and it has been suggested that these

characteristics are a function of their diversity

(Soemarwoto and Soemarwoto 1979, Gliessman et al. 1981).

Interest in the stabilizing effect of diversity in

agroecosystems is reflected in the expressed need for

development of complex agricultural systems for the humid

tropics (Holdridge 1959, Dickinson 1972, Trenbath 1975, Hart

1980), and in the current agronomic emphasis on polyculture

cropping systems research (Dalrymple 1971, Kass 1978).

A large body of literature on the theory of

diversity-stability relationships in ecological systems

bears directly on the question of agricultural

diversification as a means of reducing pest problems. The

traditional belief for many years among ecologists was that

diverse systems were more stable than simple ones. Strong

support of this view was expressed by most contributors to a

symposium volume on the topic (Woodwell and Smith 1969).

Subsequent work, including empirical studies and development

of mathematical models (see work cited by Goodman 1975), did

not support the original hypothesis. Goodman (1975)

reviewed the development of the diversity-stability theory

in detail and concluded that there is no clear relationship

between ecosystem diversity and stability. Empirical

studies have yielded inconsistent and contradictory results,

partly due to disagreement among ecologists both on the

definition of the term "stability"' and on appropriate

criteria for measuring it.

Many empirical studies to test the relationship between

diversity and stability have considered fluctuations in

numbers of individuals within a single population or trophic

level; fewer studies have considered the effects of

diversity on ecosystem properties such as energy flow and

nutrient cycling. Holling (1973) distinguished between

stability (small fluctuations around an equilitrium point)

and resilience (ability of a system to persist by moving

between multiple equilibria). Using these definitions the

spruce-fir forest of eastern Canada is an unstable system

that fluctuates widely in plant and animal species

composition, However, because of the instability of

populations and the resulting ettects on competition,

regeneration and forest growth rates, this system has very

high resilience (i.e., it persists).

In YcNaughton's (1977) restatement of the

diversity-stability hypothesis, the emphasis was on

stability of ecosystem processes rather than stability of

population numbers. Process stability and population

stability are not necessarily related. As Margalef (1975,

page 160) stated, "A system which is highly unstable in

species composition may be stable with relation to the

energy flowing through it." In general, a system will tend

toward the configuration of species that best processes the

available energy, thus maximizing energy flow (0dua and

Pinkerton 1955).

Odum (1975) proposed that the optimal diversity of a

system is a function of the sources and quantities of

available ene rgies. Hre calculated diversity indices from

empirical data on plant and animal species abundances in a

variety of ecosystems. The frequency distribution of the

diversity indices was bimodal. Stressed, selectively

managed and subsidized ecosystems had low diversity indices;

natural ecosystems where solar radiation was the primary

energy source had high diversity indices.

Lugo (1978) emphasized the importance of energy drains,

as well as energy sources, in determining system complexity.

It is generally accepted that ecosystem complexity and

efficiency of energy use are positively correlated (see

Margalef 1968), and it has been hypothesized that plant

diversity is positively associated with primary productivity

(Connell and Orias 1964, Margalef 1968, H. T.Odum 1971).

However, the development and maintenance of diversity

requires energy expenditures and the complexity of an

ecosystem is determined by the balance between energy inputs

and energy drains (a. T. Odus 1971, Lugo 1978). For

example, very productive systems with low energy drains have

high diversity (e.g., a coral reef), while very productive

systems with high energy drains have low diversity (e.g., an

estuary with tidal exports of organic matter).

In a natural ecosystem, high diversity of components

provides many possible pathways for the flow of energy.

When a high diversity system is stressed, either by a

fluctuation in the energy inputs to the system or by an

increase in energy drains from the system, the dominant

energy pathways change, but the system may still be able to

process the available energy. High diversity results in

more alternative equilibrium states of the system (Holling

1973), which provide more options for maximizing energy flow

under fluctuating conditions. Diversity, then, is a

homeostatic mechanism operating at the ecosystem level that

insures continuous energy flow through the system (Reichle

et al. 1975). Species abundances change when a perturbation

occurs, the decreases in some species are compensated for by

increases in other species, and by this mechanism ecosystem

functional properties are stabilized (McNaughton 1977).

Lugo (1978) proposed that the ability of a system to respond

to a perturbation depends on the dynamics of the system's

enecqy pathways, the type and intensity of the perturbation,

and the kinds and numbers of pathways altered.

Jaats of Herbigg

Herbivory stresses the ecosystem by draining energy from

plant biomass. In natural ecosystems, herbivory is a normal

or background stress to which the system is usually well

adapted (Lugo 1978). In ecosystems that are not well

adapted to herbivore stress (e.g., many agricultural systems

and natural systems with introduced pests), herbivory may

ultimately effect the ability of the system to persist

through its impact on energy flow.

Herbivory may alter energy flow through the primary

producers in two ways: (1) directly, by reducing the amount

of photosynthetic tissue and by stimulating compensatory

growth in remaining tissue, and (2) indirectly, by affecting

structural and functional characteristics of the system,

which in turn alter the primary productivity rate.

Although insects generally consume only a small fraction

of the leaf tissue in a terrestrial ecosystem, the effects

of herbivores dre greater than simply loss of leaf area

(Harper 1977, Whittaker 1979, Lubchenco and Gaines 1981).

Herbivory influences ecosystem structure and function by

increasing light penetration and reducing competition for

nutrients, water, and light. Herbivory may accelerate

nutrient cycling through increased nutrient leaching from

damaged foliage and increased decomposition rates (Mattson

and Addy 1975, Golley 1977, Bormann and Likens 1979, Barbour

et al. 1980). Herbivores act as ecosystem regulators

through direct and indirect feedback loops to the autotrophs

(Odum and Ruiz-Beyes 1970, Chew 1974, Mattson and Addy 1975,

Lee and Inman 1975). The effects of herbivores on system

processes ady be positive or negative, depending on the

characteristics and state of the system (Lugo 1978).

Direct impactoqnts on setprimary ro d~uctivity. Moderate

amounts of herbivory may stimulate plant productivity under

certain conditions (McNaughton 1979a), and compensatory

growth following defoliation has been well documented

(Alcock 1962, Pearson 1965, Hodqkinson ft al. 1972, Gifford

and Marshal 1973, ncNaughton 1976, Det~ling et al. 1979,

Painter and Detling 1981). Many plants normally

photosynthesize at less than their maximum rates. It has

been suggested that the relationship between herbivory and

net primacy productivity (NJPP) is nonmonotonic, and there is

an optimum grazing level at which NEP is maximized

(McNaughten 1979a). Although herbivory is usually

considered a stress to the plant community, stress may

accelerate processes and in some caSEs benefit the system

(Lugo 1978). Stimulation of plant productivity by grazing

is an example of a positive feedback loop within the system

that amplifies energy flow (Odum 1977). Feedtdck may be

negative rather than positive at high herbivory levels, and

there is a threshold herbivory level above which plant

productivity decreases (Vickery 1972, Dyee 1975, Noy-neir

1975, Caughley 1976).

IEpacts on species c omposition and diversity.Idvda

plant responses to herbivory may be positive or negative,

depending on plant genetics, intensity and frequency of

defoliation, the tissues affected, plant developmental stage

at the time of attack, and environmental factors (McNaughton

1979a) .

Herbivory may lead to a variety of physiological

responses in the individual plant. These include (1) plant

mortality and reduced growth (Kulman 1971); (2) alteration

of plant resource partitioning (Gifford and Marshal 1973,

Detling et al. 1979); (3) stimulation of compensatory growth

in residual tissue (Pearson 1965, Hodgkinson et al. 1972,

Dyer 1975, McNaughton 1976, 1979a, Detlingq et al. 1979,

Painter and Detling 1981); (4) increases or decreases in

plant reproductive output (Jameson 1963, Cavers 1973,

Rockwood 1973, Harris 1974, Owen and Eiegert 1976, Boscher

1979, Pinter and Kalman 1979, Benticy et al. 1980s

Stephenson 1981); (5) changes in plant growth patterns, such

as increased branching or tillering (Oppenheimer and Lang

1969, Youngner 1972, Saunders 1978, Simberloff et al. 1978,

Owen 1980); (6) increased or decreased root growth

(Troug~hton 1960, Alcock 1962, Jameson 1963, Taylor and

Bardner 1968, Dunn and Engel 1971, Whittaker 1979); (7)

delay of plant senescence (Chew 1974, McNaughton 1976); (8)

increased water use efficiency, due to reduced transpiration

area (Daubeamire and Colwell 1942, Baker and Hunt 1961) ; and

(9) reduced nutritive? quality of remaining leaf tissue

(bchu3tz and Baldwin 1982).

Plant response; to herbivory reflect a complex

interaction of factors. The net result of herbivory at the

community level is a change in competitive advantage among

species. As Whittaker (1979) pointedl out, the competitive

balance among species is altered by h~erbivory regardless of

whether an individual plant is damaged or benefited.

Results of numerous studies. (e.g., Malone 1969, Rafes 1970,

Harris 1973, McNaughton 1979b, Linhart and Whelan 1980)

support the generalization that herbivory shapes the plant

species composition of an ecosystem by altering the

competitive balance among species. Instances of successful

biological control of plant pests by introduced insects are

examples of the impact that herbivory can have on plant

species composition (see De~ach 1974).

By affecting competition, herbivory may regulate plant

diversity in an ecosystem. It has been suggested that

herbivory may maintain local species diversity by keeping

plant populations at low densities dnd by increasing nicho

differentiation (Whittaker 1965, Connell 1971, Huffaker

1971, Harris 1973). Grime (1973) predicted that

herbivore-susceptible species would be outcompeted at high

grazing rates, herbivoro-resistant species would be

outcompeted at Low grazing rates, and therefore highest

species diversity would occur at intermediate grazing


intensities. Lubchenco and Gaines (1981) hypothesized that

diversity would be a maximum at low or intermediate

herbivore levels, depending on the nature of the competitive

interactions between plants. Harper (1969) and Caughley and

Lawton (1981) suggested that the! effects of predation were

determined by herbivore abundance and feeding

characteristics and that herbivore activity might increase

or decrease plant diversity.

Regardless of the direction of the change, the effects of

herbivory-induced shifts in diversity on ecosystema processes

may be important determinants of ecosystem stability.

McNaughton (1977, page 516) reiterated the idea developed

within the framework of diversity-stability theory that

"compensatory fluctuations in the abundances of co-occurring

system elements (species populations) in a variable

environment can stabilize aggregate system properties." He

presentJ; empirical data from a grazing experiment in high

and low diversity ecosystems that supported this idea. In

the highr diversity system, grazing resulted in a change in

plant species diversity, but had little effect on the total

plant biomass. In the low diversity system, an equal amount

of grazing did not affect species diversity, but

significantly reduced plant biomass. Thus high diversity

provided a homeostatic mechanism that allowed functional

stability maintenancee of plant biomass) in the face of a

perturbation (grazing).

Dversit Effect on He vrbigy

The relationship between herbivory and plant diversity is

a two-way interaction. In addition to the effects of

herbivory on ecosystem processes, the structural

characteristics of the system also influence herbivory


It has been suggested that increased plant diversity

results in decreased herbivory, and many investigators have

reported fewer herbivores and/or less herbivore consumption

in floristically diverse than in floristically simple

systems (Burleigh et al. 1973, Root 1973, Dempster and

Coaker 197((, Smith 1976, Altieri et al. 1977s Altieri et al.

1978, Bach 1980, Risch 1981). Hlerbivory reduction in

diverse systems has been attributed to the presence of

alternative hosts that divert plant pests, greater abundance

and diversity of insect predators, and/or structural

complexity that interferes with insect movements and makes

host plants harder to find (Root 1973, Atsatt and O'Dowd

1976, Pimentel 1977).

These studies may lead to the conclusion that by

increasing plant species diversity, one increases the

resistance of an ecosystem to herbircre attack. However,

attempts to relate ecosystem diversity to herbivory patterns

have not always yielded consistent results. There is

evidence that the buffered environment of a complex

ecosystee may support certain pests not able to survive in a

more open monoc ul ture, and that some pest problems may

increase with ecosystem complexity (Hart 1974, van Euden

1977, Way 1977). For example, some investigators have

reported fewer predaceous insects (Pimental 1961b, Pollard

1971), lower insect predator efficiency (Price et al. 1980),

and greater abundances of some herbivores (Cromartie 1975,

Thompson and Price 1977) in diverse systems.


The primary objective of this study was to investigate

net primary productivity and hecbivory in hign diversity and

low diversity tropical successional ecosystems. The work

was done as a part of a larger study designed to test the

feasibility of using natural succession as a model for the

development of new tropical agroecosystems. Experimental

successional ecosystems that lacked, imitated, and exceeded

the floristic complexity of the natural successional system

provided the framework for investigating four questions:

(1) Does net primary productivity differ in high and low

diversity systems? (2) Do herbivore consumption rates

differ in high and low diversity systems? (3) How does

herbivory affect net primacy productivity in high and low

diversity systems? () Are high diversity systems more

homeostatic than low diversity systems when partially



The Study Site

The research was carried out in the Florencia Norte

Forest of the Centro Agronomico Tropical de Investigacion y

Ensenanza (CATIE), at Turcialba, Costa Rica. The site,

located dt go 53' N, 83o 40' W, lies at the eastern edge of

the central plateau of Costa Rica at an elevation of 650 m.

The topography is gently undulating, and the vegetation of

the area falls into the tropical premontane wet forest life

zonel (sensu Holdridge 1967, Tosi 1969).

Long term mean annual rainfall foc the area is

approximately 2700 mm, with a pronounced dry season from

January through larch. Mean annual rainfall for 1979-1980

(2169 am) was somewhat lower` than the long term average.

Monthly rainfall amounts ranged from 14 mm in March 1980 to

4160 mam in December 1980 (Fig. 1). Temperatures ranged from

an average maximum of 28.40 C to an average minimum of 17.10

C, with a median temperature of 22.70 C.

The 2.11 ha study site is typical of large areas in the

mid-elevation warm humid tropics that have been deforested

for agricultural use. At the start of the study, the

vegetation on the site consisted of 8-9 ye old second growth


1 10


r -


.O or



interplanted with timber trees, and remnants of a 56-60 yr

old secondary forest dominated by Goethalsia maiantha. The

immediate study area was surrounded by diverse second

growth, pasture, and experimental forestry plantings, and

overlapped with some of the land where Harcombe (1977a,

1977b) did earlier studies on tropical succession.

The soil at the study site, classified as a Typic

Dystrandept (Soil Conservation Service 1975), was an upland

soil overlying upper Miocene or lower Pliocene rock

(Harcombe 1973). This deep, freely drained soil is

characterized by low bulk density, <50% base saturation, and

a moderate to high cation exchange capacity.

Site Preparation

During the fist week of January 1979, the vegetation was

felled on six 33 x 33 m plots and several smaller plots,

using machetes and a chain saw. Border strips of Living

vegetation at least 5 m wide were left between plots.

Firewood was removed from the site, and the remaining

vegetation was left on the ground through the dry season.

On 22 March 1979, the plots were burned. The turn was

intense and complete, and lcft the site with a Iiniform cover

of white ash. The impacts of the slash and burn process on

nutrient budgets, soil carbon dioxide evolution, soil seed

storage, and plant growth vere studied and are reported

elsewhere (Evel et al. 1981). Immediately after the burn,

the four experimental manipulations were initiated.

Main Treatments

Three experimental successional ecosystems, plus a

natural successional system, were studied. The experimental

systems were designed to represent two types of

floristically diverse successional ecosystems and one

floristically simple system. Natural succession provided

the baseline witn which the other systems were compared.

The four main treatments are described below.

Natural succession

In this system natural regeneration began after the burn,

and secondary succession was allowed to proceed with no

experimental manipulations. The natural succession provided

an estimate of what nature does during early tropical

succession. This treatment was used as a control for

comparison of structural and functional characteristics of

the other three main treatments.

Himic of succession

In this treatment a diverse successional system vas

experimentally constructed and maintained. The idea was to

try to imitate the structure and function of the natural

successional system by substituting species mnorphologJically

similar to those found in the natural succession. The

species co mposit ion of the minic was completely

investigator-controlled. Both careful observation of the

natural succession plots and prior knowledge of tropical

successional trends provided guidelines for selection of

species to be included in the mimic. For e xa mple,

herbaceous vines (e.g., Vlqgn uniculata, several varieties

of Eggaselu _v ygfaris, Cucurbita eepo and Sechium edule)

imitated early successional vines in the Cucurbitaceae

(e.g., Frant~zia ettieri, Momordica char~anta) and

Leg~uminosae (e.g., ghynch~osi a graidati~s, Vi~n~a vexill;ata).

Castor bean (Ricillus comm~unis) and papaya (Caygiga paagya)

were substituted for fast-growing pioneer tree species

(Cecropia spp. and Bocconia frutescens). Large monocots

such as plantains (Musa paradisiac) were imitations of

common early succession aonocots (e.g., Calathea insignis,

Helic~ona latlispatha,. and Ischooshoon Eittieri)

Cultivated herbs (e.g., Cagsicum sp.) replaced

mor phologically similar native herbs (e. g., Solanum


Both cultivars and non-cultivated species that were not

present in the area were included in the mimic. Continuous

evaluation of the mimic and regular additions of new species

occurred during the 1.5 yr study period. The plots were

periodically weeded to remove natural colonizers.

The mimic was a key ecosystem for testing whether it was

possible to imitate succession in such a way that the

productivity and hlomeostasis of the natural system was


Enriched Succession

The enriched succession was a system in which the natural

regeneration was supplemented by continuous inputs of

propaquies of many species not present in the vicinity of

the study site. This was a self-design treatment in which

nature controlled the selection process in an ecosystem in

which the limitations of seed accessibility had been

reduced. This system was used to determine whether or not

the removal of some biogeographical constraints would result

in an ecosystem iore diverse than the natural succession,

and whether the resulting ecosystem would differ

structurally or functionally from the natural successional


Propagules of both cultivars and non-cultivars were added

to the auriched succession plots at approximately bi-weekly

intervals. Seeds were scattered on the ground, and stem

cuttings and seedlings were planted at randomly located

points within the plots. During most months, a minimum of

10,000 propagules of at least 30 species were added to each


Succession~~al Moonoctultu

A single species system was included in the study for

comparison with the high diversity systems. A series of

three monocultures was planted, with the species chosen (1)

to resemble the life forms of dominant successional species


at that stage in succession, and (2) to represent important

cropping systems in the area.

Maize (gea _mays var. Turpeno), an herbaceous monocot

similar tc some early successional qcasses, was planted

immediately after the burn (late March 1979). The first

maize crop was harvested in mid-July 1979 and was followed

by a second maize planting. After the second maize harvest

(November 1979), cassava (Manihot esculenta var. Japonesa)

was planted. Cassava is a tuber crop important throughout

the tropics. Cassava was chosen for the monoculture because

its uoody ycowth form was similar to the growth form of the

shrubs that were rapidly becoming dominant in the 7 mo old

natural succession. The cassava was harvested in

mid-September 1980 and was followed by a planting of Cordia

alliedfora, an important timber species. Data on the Cordia

monoculture are not included in th~is study.

The planting procedures and management of the monoculture

plots followed as closely as possible the methods used by

local farmers. Maize was planted at 1.0 x 0.5 a spacing,

two seeds per hole. The cassava was grown from stem

cuttings planted at 1 x 1 m spacing. At plant maturity, the

harvestable crop (ears or tubers) was removed from the

plots, and the remaining plant material was left on the

ground. All monoculture plots were periodically weeded.

Plot Layot nd V ariale Mesued

The treatment plots were arranged in a randomized

complete block design, with six replications of each of the

four main treatments (Fig. 2). Each study plot measured 14

x 14 m (196 mr) within permanent metal markers. An

additional border strip approximately 1 m wide was left

around each plot, making the actual plot size 16 x 16 m (256

m2). The study plots within each replication were separated

by I m wide access trails. Buffer strips at least 5 a wide

of original, uncut vegetation were left between replications

to serve as a source of seeds for the experimental plots.

Specific areas within each study plot were designated for

particular types of investigations, including the work

reported here and the work of other researchers (Fig. 3).

Variables monitored during the 1.5 ye study period in the

C main treatments fall into two categories: (1) vegetation

structural characteristics, such as leat area index, species

composition, and vegetation height, and (2) productivity

measurements. The methods employed for each type of

measurement are described in detail belcw.

-'-:--- -- r.-~
.--:-.-- --.-r
-.=-.-IIT';-~-t.~ -.~-- li- I;
_r- Y. _.- -- ---_-~
_i__ii """~ ;=' ~--.--
roor rarlL sr~ru~~ =;1..;...
"""' ~~~~ ~-~-- --paruaE
I .-I
-;-;;~o.. r ~~LTV~ ~onm i~i-i-
;- ~~ ---
"""""""" ;- lulmrloa ,urroc~N Gn tI_-;
j CI C
..... ~-. 9
""'- -.-r.::_. -r
GYEU* :::::r IDIL slT CLICI~I (li Ti
................ UETED~OLmlCIL I
"""""' I~STCUYENTar ?rr
""""""'~'~iiii~~~~.-:~~~:i ..I.YYL*I. ICI~O~~I~~.~C.I
:::::::MUNI::::r-~~ --r3~r--~ ~~ _~;__~_ _~~~~ F 17 r rji Ci C_'~
- n
n3~. ~.~;
....~- r -I~: r:'~ ':'~r '- i- m
::::::::::::IP~ ~ ;-C~-rcZ-,
"""""''*' '"' It;c~-~~
--,, ~I 'r
- ~n
r~- .--,
.r X. r -IL iT-, r-:---r
':::' .r r-n
rr ~C
r..i. -;a =II~ r ry-r
. ~- ?r II ,,,
r ''
_~ji~J i..ri~;_~
''~'' ., L ~T Y~L ~5~~
~- 'c'3 .
-. `' '=1. ~ j~f~~; :...... -~"~
~''~ .?I u, C-.

"'~ C~7: jr:
~LL~;r~C ii
"~ ~LL~, %~-T.
~ II
c*- I--cr I i 7t'l
---~ c. i
.J'~iiL .:--
--L -e
' ~

s to to ~o rs~u~ s~rju~uor~o ~j;~i3L~t~qcr~ ~J;
"""' j3~~RnY~~:~ rClj3iJry Ji-
C~ rg~~,(J.i iV OP^?."~cJ i

Figure 2. Map of the study site.

1 I



16 m








figure 3. Diagram of study plot.


I. .o .
cooo SOIL PITco


Measurements of eetatio Structr

Leaf Area Index

seaf area index (LAI) is defined as lear area per unit

ground area. Values are usually reported as m2 of leaf

tissue (one side of leaf) per mt of ground. In this study

LAI was measured using a plumt-bob method similar to the

method used by Benedict (1976). A th:in line is lowered

vertically from thle top of the vegetation canopy to the

ground and the number of leaves touching the line is

counted. This method reduces the sampling area to a single

point, and the number of leaves above a point (i.e., the

number of intersec tions of line and leaf) is a direct

measure of LAI. The intersections were recorded by species

and height above the ground.

The instrument used to measure LAI was constructed from a

rigid extendable metal rod. A fishing reel was connected at

its base and a pulley at the tip. A thin nylon twine

attached to the rod with a small weight at its end could

then be easily lowered vertically through the vegetation.

The twine was knotted at 25 cm Intervals, and alternate

intervals were painted for easy reading in the field. This

instrument could be used in vegetation up to 8 m in height.

In taller vegetation, it was necessary in a few cases to

estimate the numoer of leaves above the rod.

Leaf area index measurements were made in all main plots

during May 1979, July 1979, Novembec 1979, April 1980, and


October 1980. In May 1979, 20 LAI measurements were made in

each study plot of each replication. Five locations were

chosen randomly in each plot, and four LAI readings were

taken at each location by dropping the line vertically

through the vegetation four times. For all other sample

dates, 30 LAI measurements were made in each plot. Ten 1 mt

quadrats were systematically located in each plot and

permanently marked. Three LAI measurements were made in

each of these quadrats on each date.

The uniform spacing of crop plants in rows created

special problems in use of the plumb-bob method to measure

LAI, especially in systems with very low LAI. For this.

reason, LAI of the maize monoculture in November 1979 was

calculated using leaf biomass/leaf area regressions rather

than by using the plumb-bob method.

Species data from the leaf area measurements were used to

calculate LAI for individual species, and percent of total

LAI was used as an estimate of relative species dominance.

Species Compositiog

Species inventories were done in the natural succession,

enriched succession, and simic plots during July 1979,

November 1979, April 1980 and October 1980. For each plot a

list was made of all flowering plants and ferns encountered

in each of the telr 1 mZ quadrats described above. From

these data, diversity indices were calculated. In addition,

a complete species inventory was made in each 16 x 16 mr

plot in Octobne 19)80. Plant specimens were identified at

the National Museum of costa Rica.

yggetation Height

At the same time that the species composition and LAI

measurements were made, the height of the tallest plant in

each of the ten 1 mt quadrats in each plot was measured.

Average canopy height for each plot was then calculated.

Also, the species and height of the tallest plant in the

entire 10 x 16 m2 plot was recorded.

Productivity Measurem~ents

Net primary productivity is one of the principal response

variables that was used to compare the four experimental

ecosystems. A common method for estimating net primary

productivity is by using periodic biomass measurements to

calculate changes in standing crop over time. However, in

fast-growing tropical successional vegetation, the

measurement of changes in living biomass underestimates

actual net primary production because of rapid turnover of

plant parts and losses to herbivores during the time

intervals between harvests. Litterfall and insect

consumption are two losses of net productivity that cannot

be measured by biomass harvests. In this study,

measurements were made of plant mortality, rates of


litterfdll, and cates of harbivory, in addition to periodic

measurements of above-ground living biomass. The values

obtained were used to estimate above-ground net primary


Mean rates of biomnass inicrement (g a-' day-i) were

estimated for intervals between biomass harvests as

B = ----------- Eq. 1

where 8(i) = above-ground living biomass at harvest(i) in

g/m2, E(i-1) = above-ground living biomass at harvest(i-1)

in g/mz, and t(i)-t(i-1) = number of days between biomass

harvests. These cates were plotted at the mid-points of the

intervals between harvests, and the points were connected by

straight lines. Linear regressions were then used to

estimate daily biomass increments.

Increments of standing dead biomass (g m-r day-1) were

estimated as

D = ----------- Eq. 2

where D(i) = standing dead biomaass at harvest(i) in q/mz,

D(i-1) = standing dead biomass at harvest(i-1) in g/mz, and

t~i-t~-1)= number of days between harvests. As above,

the rates were plotted at the mid-points of the intervals

between harvests, the points were conneccted by straight

iines, and linear regressions were used to estimate daily

increments in standing dlead biomass. The turnover rate of

standing dead biomass was not known. The conservative

assumption was made that turnover was neqLigible. Positive

daily increments in the standing dead biomass category were

used as tsti~tute of daily production of standing dead

biomass. If the turnover cate was high, production of

standing dead and net primary productivity would both be

underestimated by these methods.

Litterfall rates (g m-' ddy-') were estimated for each

ecosystem as

L = ----------- Eq. 3

where L(i) = amount of litter collected during a 4) wk

interval (q/mn), and titi1)= number of days in

interval. These rates were plotted at the mid-points of the

intervals, the points were connected by straight lines, and

linear regressions were used to estimate daily litterfall


Daily herbivory rates for each ecosystem were estimated

from three 1 mo sampling periods. Linear regressions were

used to estimate daily herbivory rates.

Daily net primary productivity rates were calculated as

NPP(i) = b(i) + 1(i) + h(i) + d(i)

Eq. r(

where NPP(i) = niet above-ground productivity on day(i) in

1 mu-r day-i, b(i) = biomass increment on day(i) in g as-p

day-i, 1(i) = litterta-ll on day~i) in q m--2 day-i, h(i)=

herbivory rate on day(i) in g m-2 day-i, and d(i)

production of standing dead biomass on day(i) in g a-r


Above-Ground Biomass

Immediately after the burn, randomly located subplots

were marked with string and metal stakes in the area of each

study plot designated for biomass harvests. Fourteen

biomass harvests were made during the 1.5 yr study period.

Early harvests in the natural succession, enriched

succession, and mimic of succession were done at frequent

intervals (approximately bi-weekly) on, small (0.24r mr)

plots, and later harvests were at less frequent intervals on

larger plots. Dates and plot sizes for each of the harvests

were 14 May 1979, 31 May-5 June 1979, 20 June 1979, 9-10

July 1979 (0.20 az); 1-2 August 1979, 10-12 September 1979,

8-10 October 1979, 19-21 November 1979, 17-19 December 1979,

21-23 January 1980 (1.60 mp); 17-19 Carch 1980, 19-21 M~ay

1980, 8-11 July 1980, 28-31 October 1980 (4.00 mZ). At the

time of each harvest, one randomly selected subplot was

harvested inl each study plot (total number of subplots

harvested per treatment = 6).


It was decided that the harvest of individual plants and

plant density data, rather than the harvest of vegetation in

random subplots, would yield better estimates of biomass in

the monoculture treatment where plants were uniformly

spaced. Therefore, from one to four randomly chosen plants

of the moncculture species were harvested per plot at each

sampling date. Harvests of the monoculture were made at

each date listed above. Additional harvests were made at

crop maturity (29 October 1979 and 10-12 September 1980) and

during the early growth stage of the second maize

monoculture (16 August 1979). At maturity of each

monoculture, samples of the harvestable crop vere used to

estimate economic yield.

Above-ground biomass was harvested by clipping all

vegetation within subplot boundaries at ground level. All

plants rooted inside the plot were included, even if parts

of the plant extended outside the sample area. Likewise,

all plants rooted outside the plot were excluded. Vines

were clipped at the plot boundary. The vegetation from each

plot was separated into four classes: leaves, stems,

reproductive parts, and standing dead. Vegetation samples

were weighed in the field. Subsamples of each vegetation

class were taken to the laboratory, weighed to the nearest

0.1 q, dried to a constant weight at 700 C, and reweighed to

obtain fresh to dry weight conversions.

Data for each vegetation component (leaves, stems,

reproductive parts and standing dead) and total above-ground

biomass were analyzed using a randomized complete block,

fixed effects statistical model with four treatments and six

blocks replicationss). The biomass data did not meet the

homogeneity of variance assumption of analysis of variance.

Means and variances were not independent; in most cases,

variance was proportional to the square of the mean. The

biomass data were transformed using the following log

transformation: y=1n(x+1). All analyses of variance and

Duncan's multiple range tests were done on the transformed

data, using the General Linear Models (GLM) program of the

Statistical Analysis System (SAS). Reported means and

standard deviations are of original untransformed data.


Three 0.25 mt litter collectors were located near the

soil surface in each replicate of each treatment. Each

collector was 1.00 x 0.25 x 0.15 m (length x width x height)

and was supported approximately 2 cm above the soil surface

by metal brackets. The collectors had wooden sides and

fine-mesh screen bottoms for drainage. The shape and small

size of the collectors allowed the successional vegetation

to grow up and over the collectors rapidly.

The collectors were positioned 1 m from the access trail

in the portion of each plot designated for litterfall


studies (see Plot Map, Fig. 3). Litter was collected from

the baskets at 2 wk intervals throughout the 1.5 yr study

period. The litter from the three collectors in each plot

was combined into one composite sample, oven dried at 700 C

to a constant weight, and weighed to the nearest 0.1 g.

The baskets collected both autochthonous and allocthonous

litter inputs to the plots. To calculate net primary

productivity of the vegetation in the plots, a measure of

autocht honous litter production was needed. Alloct honor us

inputs were estimated from a single collector (0.25 mt)

placed near the other three collectors in each monoculture

plot. For eaca of these 'control' baskets, leaves of the

monoculture species in the basket at each sampling date were

discarded. All other material in the basket was collected,

dried and weighed.


Losses of plant tissue due to herbivory were estimated by

monitoring amounts of damage incurred on taqqed leaves of

dominant species in each treatment. It was not possible to

separate losses due to plant diseases (fungal, viral,

bacterial) from losses to herbivorous insects, so loss

estimates include damage due to plant diseases as well as

losses to herbivores.

At each of three sampling periods (October 1979, February

1980, and June 1980) the most recent LAI data were used to

select the species to be tagged. The species of each

treatment were ranked from highest to lowest LAI, and those

more comacn species that jointly accounted for at least 80

percent of the total LAI of that treatment were selected for

herbivory measurements.

In the portion of the study plots designated for

non-destructive sampling, five plants of each species (three

in insecticide plots) were arbitrarily chosen for tagging.

Usually no more than one individual of each species was

tagged per replication. In a few cases, patchy distribution

of a species made it necessary to taq more than one

individual of that species within a single replication.

A plant stem was considered eligible for tagging if it

was unbroken, unbranched, and bore at least four leaves.

One eligible stem was chosen on each plant. From four to

eight consecutive leaves were selected along the stem, and

these individual leaves were numbered from youngest to

oldest. Small plastic bands marked with yellow tape were

looped around the stem at two places. Positions of leaves

relative to these bands were used to identify individual

leaves at the time of harvest. When the leaves were tagged,

the holes present in each leaf were measured by placing a

sheet of ma-ruled graph paper under the leaf and counting

the uncovered squaces. Brown spots on each leaf were

estimated visually, and total damage (holes + brown spots)

was recorded for each leaf.


The length of each leaf was measured to the nearest am at

the time of tagging. Leaf length/leaf area regressions for

each species (developed from a sample of at least 50 leaves

per species) were used to estimate the initial leaf area of

each leaf (Table 1). For each species, the best curve fit

was obtained by using a quadratic equation for all but very

small leaves, and a linear equation through the origin for

very small leaves. These initial leaf area estimates,

together with direct measurements of leaf area at the time

of harvest, were used to estimate leaf expansion during the

interval. In grasses and some herbaceous species with small

leaves (mature leaves <40 cm in length), leaf lengths were

not measured, and leaf expansion was niot estimated.

After 3 to 7 wk, the tagged leaves and all new leaves

produced on the marked stems during the interval were

harvested. Mortality of tagged leaves and number of new

leaves were recorded for each plant. In the laboratory, the

area of damage on each leaf was trdCEd on a sheet of clear

plastic and filled in using a permanent black marking pen.

Two categories of damage, holes (H) anid brown spots (B),

were drawn separately. All missing tissue, plus damage that

left only a transparent layer of leaf tissue, was recorded

as holes. All other damage, including damage by leaf-mining

insects, damage by gasping insects, fungal and viral damage,

plus the ne-rotic tissue around holes, was recorded as brown


Regression Equations

x>125: y=0.00203x2 + 0.303x 47.779
x 125: y=0.174x

x>24.: y=0.000856x2 + 0.0667x 0.927
x 24: y=0.0469x

x>23: y=0.00431x2 + 0.0475x 1.909
x 23: y=0.0624x

x>81: y=0.0172x2 1.837x + 56.231
x 81: y=0.249x

x>58: y=0.00440x2 0.228x + 6.995
x 58: y=0.147x

x>34: y=0.00318x2 0.101x + 1.895
x 34: y=0.0624x

x>22: y=0.00637x2 + 0.0803x 3.050
x<22: y=0.0772x

x>32: y=0.0124x2 0.0163x 6.809
x 32: y=0.149x

x>53: y=0.00473x2 + 0.289x 20.419
x 53: y=0.148x

x>171: y=0.00913x2 1.061x + 82.992
x 171: y=0.980x

x>77: y=0.0115x2 0.573x + 27.339
x 77: y=0.661x

x>19: y=0.00361x2 + 0.0633x 1.042
x 19: y=0.0731x

x>10: y=0.00402x2 0.0148x + 0.339
x 10: y=0.0591x

x>29: y=0.00655x2 + 0.201x 7.644
x 29: y=0.125x


Table 1. Leaf length:1eaf area regression equations for common
species. In the equations, x = leaf length in mm and
y = leaf area in cm2

















Borreria laevis

Cajanus cajan

Carica Eapaya

Clibadium aff.

Cordia inermis


Cucurbita pepo

Canavalia sp.




Hyptis vilis

Ipomoea batata

Table 1--continued.














Ipomoea sp.

Iresine diffusa


tube rosa




Solanum torvum


Vernonia patens

Vigna sp.

Regression Equations

x>28: y=0.0117x2 0.341x + 4.392
x_28: y=0.142x

x>30: y=0.00445x2 0.110x + 2.373






x 50:

x 26:

x 50:




y=0.0117x2 0.784x +

y=0.00733x2 0.0228x

y=0.0135x2 0.960x +

y=0.00267x2 + 0.0271x

y=0.00748x2 0.227x

y=0.00352x2 0.00506x

y=0.00125x2 + 0.117x

y=0.00154x2 + 0.221x

y=0.00568x2 0.0873


- 10.987




S- 0.522






The leaf remnants and plastic sheets were run through a

Lambda Instruments LL-COB (LL-3000) area moter, which

measures the surface area of opaque surfaces to the nearest

0.01 cm= with an accuracy of + 1%. In a few cases, leaves

from a plant were processed as a group rather than


For each leaf (or group of Leaves), total damage present,

D(t(f)), and gross leaf area, G(t(f)), at the time of

harvest were calculated as

D(t(f)) = H + B Eq. 5


G(t(f)) = H + H Eq. 6

where t(f! = time of leaf harvest, H = holes present at

t(f), 8 = brown spots present at t(f)s and R = residual leaf

area at t(f).

Herbivory ra tes (i.e., Loss of leaf tissue per unit area

of leaf per unit time) were calculated for each leaf of each

species. Two factors contribute to the total loss due to

herbivory: (1) actual consumption by herbivores and (2)

loss of potential photosynthetic leaf area due to expansion

of damaged areas after consumption has occurred. Since the

rate of expansion of holes in a leaf is equal to the rate of

expansion of the Leaf (Reichle et al. 1973, Coley 1980),

estimates of percent consumption are not affected by leaf


expansion during the sampling interval. Percent consumption

(LOSS) was estimated for individual leaves by the following


D(t(f)) D(t(0))100
LOSS = -------- --- ---- X --- --- -- Eq. 7
G~~t~~f)) Ge () (f) -t(0)

where D (tli)) = damage present at t (i), G(t(i)) = gcoss leaf

area at t (i), t (0) = time of leaf tagging, and t(f) = time

of leaf harvest. An absolute consumption rate was then

calculated for each species by multiplying mean percent

consumption of the species by LAI of the species.

The area of 50 leaves of each species was measured using

the LI-COR (LI-3000) area meter. The leaves of each species

were pooled, oven dried to constant weight at 700 C, and

weighed. Leaf specific mass (mass per unit area of leaf)

was then calculated so that herbivory rates could be

expressed oL a mass basis as well as on an area basis.

Three non-parametric statistics (Wilcoxon 2-sample rank

sums test, Kruskal-Wallis test, and median test) were used

to test for differences in herbivery rates between

ecosystems for: several plant species. These statistical

procedures make no assumptions about the distribution of the

data, but do require homogeneity of variance. The level of

significance of ordinary 2-sa~nple procedures is not

preserved if the variances of the tuc populations differ

(Pratt 1964(). The robustness of the tests under departure


from the assumption varies with test used, sample size of

the populations, and magnitude of departure from the

assumptions. The homogeneity of variance assumption was not

met by the herbivory data. In general, means and variances

were proportional; large variances were associated with

large means, and small variances with small means.

Therefore the levels of significance associated with test

results ace not exact.

Estimation of Hp~k l~glole Exp nso

For those species in which initial leaf area was

estimated (using regression equations), it was possible to

estimate the loss of potential photosynthetic leaf area due

to expansion of the holes inr leaves. The mathematical

equation derived to estimate consumption and expansion is

based on three assumptions: (1) the damage expansion rate

equalled the leaf expansion rate; (2) the consumption rate

was constant during the time interval in which herbivory was

monitored; and (3) for a group of leaves on a single stem,

leaf growth rate was a constant function. The validity of

each of these assumptions is discussed below.

The first assumption (that hole expansion rate = leaf

expansion rate) is generally assumed to be valid and has

been verified experimentally by Reickle et al. (1973) for a

temperate deciduous forest species (Liriodendro~n _tuligifera)

and by Coley (1980) for several tropical forest species. In

an unpublisned study of a common successional species

(Conostegia eittjesi) in a tropical premontane uet forest at

Elonte Verde, Costa Rica, I found that hole expansion rate

and leaf expansion rate did not differ significantly (n = 70


although herbivory on individual leaves does not occur at

a constant rate, the cate of damage accumulation may be

assumed to be constant for a population of leaves

(assumption 2). Likewise, altuouga tae growth curve of an

individual least is probably signoidal rather than Linear,

the average leaf growth rate or a population of leaves of

varying ages may remain constant over time (assumption 3).

Although these assumptions seem intuitively reasonable, they

have not been verified experimentally.

If the assumptions are not met, bias is introduced into

the estimation of the relative proportion of the total

herbivory loss attributable to consumption and expansion.

The results of several types of possible deviations from

assumptions 2 and 3 are presented in Table 2. If

consumption rate (c) and le~af growth rate (G;') are both

constant, then assumptions 2 and 3 are met, and the method

used in this study accurately estimates percent of total

damage due to consumption and expansion. If c and/or G' are

increasing or doctoasing functions, losses due to expansion

(e) may be overestimated or underestimated by the methods

used in this study.

Table 2. Comparison of estimated (e*) and actual (e)
to expansion, for several consumption rate
growth rate (G') functions; t = time.

losses due
(c) and leaf

Case 1
(G' constant)

Case 2
(G' decreasing)

Case 3
(G' increasing)

' /

e*=e e*>e




Case 1

Case 2
(C Decreasing)

Case 3
(c increasing)



Using the assumptions listed above, percent consumption

rate (c) and percent expansion rate (e), bJoth in smz a-2

day-i, were estimated for each plant by the following


D(t(f)) (t(0)) -------j
GIt (0) ) 10000
C = ----------------------------- X ------- Eq. 8
n-1 G(t(f))
m 1

n n-i(1-r)

D(t(f)) D(t (0)) (cXm) 10000
e = -------------------------- X ------- Eq. 9
m G It (f))

where t (0) = time of leaf tagging, tlf) = time of leaf

har ves t, a = t If) t(0) = number of days leaves were

tagged, D(t(0)) = damaged area ot t(0) in cmz, D(t~f))

damaged area at C(f) in cmz, G(t(0)) = gross leaf area at

t(0) in cm2z, G(t(f)) = gross loaf area at t(f) in cmr, r

G It (0)/G(t(f)), and ni = the number of sub-intervals

(t(11),tj))into which the time interval (t(0),t(f)) is

divided. TIhe derivation of Equation 8 is given in Appendix


In the equation above, D(t(0)), D(t(f)), G(t(0)), and

G(t~f)) are totals of all tagged leaves on a given plant,

excluding taqqed leaves that died during the interval and


new leaves produced during the interval. Calculations of

losses due to hole expansion were made using plant totals

rather than individual leaf data for two reasons. (1) The

precision of the regression estimates of initial leaf areas

was not high enough to allow individual leaf expansion to be

estimated. Although the least lengthyleaf area regressions

for most species were quite good (9r > 0.94 for 19 of 25

species, Table 1), in some cases overestimates of initial

leaf area led to negative leaf growth rates for individual

leaves during the interval. (2) The assumption that leaf

growth was a constant function is better fit by qcoups of

leaves of vacying ages than for individual leaves.

The herbivory rate calculated using plant totals is

mathematically equivalent to the mean of the hectivory rates

calculated for individual leaves if all of the leaves are

equal in size; if damage areia:leaf area is ccustant for all

leaves (i.e., herbivory is evenly distributed among leaves);

if tne sums of damage area:Leaf area ace the same foc groups

of equal-sized leaves; or if total leaf areas are thle same

in groups of leaves with equal percent damage. None of the

sufficient conditions listed for equality of the 2 methods

are necessarily set by the data. Thus pocling individual

leaf data for analysis mcly introduce a source of erroc. To

evaluate the magnitude of the error, herbivory rates

calculated from individual leaf data and from plant totals

were compared for six species (Table 3). Although herbivory

Table 3. Comparison of mean consumption rates calculated from individual leaf data and
from pooled leaf data for selected species.

Rate Based on
No. of Rate Based on Individual
Plants Plant Totals Leaf Rates Difference Value
Species (n) (cm2/m2 leaf/day) (cm2/m2 leaf/day) D (sD) of t

Bocconia 12 11.05 9.21 1.84 (5.76) 1.10

Cajanus cajan 8 32.15 13.33 0.82 (3.91) 0.59

Carica papaya 3 2.36 2.46 -0.10 (0.29) -0.61

Manihot 36 8.90 11.07 -2.17 (5.12) -2.54*

Phytolacca 24 11.62 11.39 0.23 (2.45) 0.45

Vernonia patens 13 27.33 25.61 1.72 (3.48) 1.78


rates calculated by the two methods differed considerably

for some plants, the two methods yielded significantly

different mean species herbivocy rates for only one species

Mniho~t _esculents).

Consumption rates were estimated by an iterative process

in which the time interval (t(0),t (f)) was divided into a

smaller sub-intervals (t(j-1),t(1)), and consumption and

expansion were calculated for each of these sub-intervals.

In this method, both the expansion of damage present on the

leaves at t(0) and the expansion of damage that occurred

during the interval (t(0),t(t)) were excluded from the

estimate of consumption. As the number of iterations (n)

was increased, the precision of the estimate of c also

increased. To select an appropriate value of n, consumption

rates were estimated using various n values for nine plants.

For each of the plants, an n value of 55 was sufficiently

large to insure that the consumption rates (cm2 plant-'

day-1) were accurate to the nearest 0.01 cmZ. For most of

the sample plants, the required n value for this level of

accuracy was much less than 55. On the basis of these

prelimina ry tests, calculations of damage expansion were

done with n = 55. Computer programs to calculate damage

expansion were developed using the Statistical Analysis

System (SAS). One program wac developed for use with

alternate-leaved species. A modified version of this

program was used for opposite-leaved species, in which data

were pooled for opposite leaf pairs.


In addition to main treatment comparisons, a major

objective of the study was to evaluate the effects of

herbivory on net primary productivity, vegetation structure,

and species composition in high and 10w diversity tropical

successional ecosystems. To do this, comparisons were made

between high diversity systems (natural succession and

enriched succession) and low diversity systems (maize and

cassava monocultures) at three levels of herbivory: (1)

background or naturally occurring level, (2) decreased level

of herbivory, and (3) increased level of herbivory.

Background HerbivorY

Rates of herbivory naturally occurring in the enriched

succession, the natural succession, and the monocultures

were measured using the methods described earlier (Chrapter

II, 'Herbivory ratess'. Net primary productivity and

vegetation structure measurements in these treatments

provided baseline data for comparison with plots

experiencing artificially induced high and low levels of


Dereasged erbivory

To compare high and low diversity systems experiencing

low herbivore pressure, three auxiliary plots of the

enriched succession and the monoculture were maintained at

lower than normal levels of herbivory by use of


Each insecticide study plot was 11.5 x 14 m, with a border

strip approximately 0.5 m wide around each plot. The two

plots in each replication were separated by a 1 m wide

access trail. The insecticide plots were separated from the

main plots by strips of uncut vegetation at least 5 m wide,

and were located such that other study plots would not be

contaminated withi insecticide residues through runoff and/or

drainage. Within each plot, specific areas were designated

for biomass harvests and for non-destructive sampling such

as litter collection and herbivory measurements.

In all insecticide plots, above-ground plant parts were

sprayed with Diazi non, a broad spectrum i nsec tic ide.

Diazi non is a short-li ved org ano phos phate with few

phytotoxic effects that is effective against most sucking

and chewing insects. The plots were sprayed weekly during

the dry season and twice-weekly during the rainy season,

using a backpack sprayer. Diazinon powder (25X active

ingredient) and Pegafix (a wetting agent that increases

adhesion of the insecticide to least surfaces) were mixed

with water (1 ml1 Diazinon and 1.5 ml Pegafix per liter of

water), and plants were sprayed until thoroughly wetted.

Aldrin, a persistent chlocinated hydrocarbon effective

against root-feeding insects, was applied to the soil in the

insecticide plots twice yearly at the rate of 10 kg active

ingredient per ha. Dates of Aldrin application were 31

March 1979, 1 November 1979, and 26 May 1980.

Small ditches (2b cm wide and 10 cm deep) were dug around

the insecticide plots and sprinkled with 25% ALdrin powder

approximately every 2 me to prevent leaf-cutter ants (Atta

cephalotes) from entering the plots. These channels were

kept clear of fallen leaves and twigs that might act as

passageways for ants. No leaf-cutter activity was observed

in the insecticide plots.

All vegetation structure and productivity measurements

made in the main treatment plots were also made in the

insecticide plots. Species present in four systematically

located, permanently marked I mz quadrats per plot were

recorded at four sampling dates during the study period.

Three LAI measurements were made in each quadrat (total

number of LAI measurements per plot = 12) at each sampling

date, and vegetation height was measured in each of the four

quadrats at each date. Three litter collectors were placed

in each plot. Litter collections, biomass harvests, and

herbivory measurements were made at the same frequency and

using the same methods as in the main treatments.

Increase edHerbiv~ory

To study the relative abilities of simple and complex

systems to respond to high levels of insect attack,

artificial defoliation experiments were performed in the

natural succession, enriched succession, and monoculture


A preliminary series of defoliations was performed in

October 1979. Defoliations were done in designated 4.5 x 14

m subplots in replications 2, 5, and 6 of the enriched

succession and the maize monoculture. Approximately 50% of

the total leaf area on each plot was removed, by clipping

(at the petiole) alternate leaves along each stem. Leaf

tissue removed was weighed in the field, subsampled, and

returned to the plots. Three least subsamples (approximately

0.5 kg each) from each plot were taken to the laboratory,

weighed to the nearest 0.1 g, dried to constant weight at

70o C, and cewe~ighed to determine fresh to dry weight

conversions. Biomass harvests were made before the

defoliation (May-September 1979), for 8 mo after defoliation

in the enriched succession (October 1979-May 1980), and

until the maize harvest (November 1979) in the monoculture.

A second defoliation study was carried out during

April-June 1980 in replicatious 1, 2, and 3 of the natural

succession and the cassava monoculture. Defoliation plots

were 4.5 x 9.5 m, and defoliation techniques were the same

as those used in the pilot study. In this study, a series

of three defoliations was pernormed at n wk intervals. At

eaca defoliation, approximately 50% of the total leaf area

of each plot was removed.


Rate of recovery of leaf area, as measured by changes in

LAI after defoliation, was the response variable used to

compare the high and low diversity systems in the second

defoliation study. The LAI measurements were made in each

of thle def olia tion plo ts at the following times: (1)

impme('iately before each of thie three defoliations, (2)

immediately after each of the three detoliations, and (3)

after 2 wk of regrowth following each defoliation. The LAI

measurements were made from 15 equally-spaced locations

along the perimeter of each plot, five measurements per

location (total per plot = 75). The LAI measurements were

recorded by species and height above the ground. The

non-destructive sampling areas (see diagram of study plot,

Fig. 3) in replications 1, 2, and 3 of the natural

succession and the cassava monoculture were used as control

plots for the second defoliation experiment, and LAI was

measured in the control plots on the same dates that the

defoliated plots were measured (15 sampling locations x 5

measurements per location = 75 LAI measurements per control



Vegetation Structure

Seven factors related to vegetation structure and species

composition were estimated in each of the four experimental

ecosystems: species richness, species evenness, overall

species diversity, relative species abundance, species

changes through time, leaf area index, and vertical leaf

distribution. Based on these measurements, the natural

succession and enriched succession were structurally very

similar; the mimic, although similar in many ways to the

natural succession, had several important structural

dif ferences; and the monoculture was completely dissimilar

to the other ecosystems.

Segecies Compg~sjtlo9

Species data irom the LAI measurements were used to

calculate species diversity, evenness, and rate of species

turnover in the experimental ecosystems (Table 4). The

number of species intersected by 180 LAI measurements was

approximately equal in the natural and enriched succession

at each~ date; fewer species were intersected in the mimic.

Species richness increased during the study period in all

Table 4. Changes in number of species, diversity, and evenness in four ecosystems.

Vegetation Natural Enriched Mimic of
Characteristic Age (mo) Succession Succession Succession Monoculture

Number of leaves 3 734 788 321 153
intersected by 7 654 671 317 90a
180 LAI measurements 12 415 466 193 520a
18 782 905 545 524b

Number of species 3 37 35 10 1
intersected by 7 39 40 17 1
180 LAI measurements 12 36 39 15 1
18 53 63 32 1

Number of species 26 21 6 0
intersected both at
3 mo and 18 mo

Number of species 27 42 26 1
gained from 3 mo
to 18 mo

Number of species 11 14 4 .1
lost from 3 mo
to 18 mo

Species diversity 3 1.02 1.04 0.88 0.00
(R')c 7 1.17 1.09 0.90 0.00
12 1.15 1.04 0.58 0.00
18 1.26 1.24 0.92 0.00

Evennessd 3 0.65 0.67 0.88 0.00
7 0.73 0.68 0.73 0.00
12 0.73 0.65 0.49 0.00
18 0.73 0.69 0.61 0.00

Vegetation Natural Enriched Mimic of
Characteristic Age (mo) Succession Succession Succession Monoculture

Community similarity 0.41 0.60 0.15 0.00
(C) between age 3 mo
and age 18 moe

aNot measured directly. Value estimated from leaf biomass data and leaf weight/1eaf area

bSeptember 1980 measurement (mature cassava).

CH' = -C(n ,AT)log(ni/N), where ni is the number of leaf intersections for species i, and
N is the total number of leaves intersected (Shannon index).

dEvenness = H'/10g S, where H' is Shannon diversity index and S is number of species.

eC = a(1) + a(2) +...+ a(i) +...+ a~n), where i is a species present at 3 mo and/or 18 mo,
a(i) is the lesser percent LAI value for species i from the two dates, and n is the total
number of species.

Table 4--continued.


ecosystems except the monoculture. Species richness at 18

me, (based on a total inventacy of all plots) was highest in

the enriched succession (159 plant species present on 1536

m2), followed by the natural succession (121 species), mimic

of succession (82 species), and monoculture (1 species).

The Shannon diversity index (H') was calculated as a

simple measure to compare overall diversity (richness and

evenness) of the experimental ecosystems. An evenness index

based on the Shannon index (evenness = H'/log S, where S is

the number of species) was also calculated. The diversity

index increased over time in the natural succession and

enriched succession, but not in the minic (Table 4).

Diversity at 18 mo was higher in the natural succession and

enriched succession (1.24 and 1.26 respectively) than in the

mimic (0.92). Of the possible range of evenness values from

0 to 1, the values in the natural succession and enriched

succession were approximately equal (from 0.65 to 0.73),

with little change over time. Evenness values in the mimic

were more variable (from 0.49 to 0.88).

The species composition of thle natural succession and

enriched succession was very similar early in succession (at

3 mo), Lut less similar at 18 mo. The natural succession

and enrichea succession had 86 species in common at 18 mo.

Thirty-five at the species present in the natural succession

at 18 mo were not present in the enriched succession.

Seventy-three species were present in the enriched


succession but not In the natural succession, and of these

at least 24 were invsetigator-introduced.

Some of the species differences between the natural anrd

enriched succession may be due to random differences in seed

availability of native species and to random

micro-environmental differences among plots. However, at

least 9% of the 264 species introduced into the enriched

succession had become successfully established by the end of

the study period. It was possible to increase species

richness by propagule additions, and these data suggest that

species richness was limited by propagule accessibility

during the earliest stage of succession. This result may be

a temporary phenomenon due to the stochastic nature of early

succession (Uebb et al. 1972,. Horn 1974) and to the

continuous capid changes in vertical and horizontal plant

distribution that allowed colonization by new species.

Longer-term results of the study will verify whether or not

the higher species richness of the enriched succession can

be maintained.

To compare the degree of similarity in species

composition between ecosystems, a community similarity index

was calculated for each pair of ecosystems at four dates

(Table 5). The index (Gleason 1920) was C = a(1) + a(2) +

... a~) +...+ a(n), where i is a species present in at

least one of the two ecosystems being compared, a(i) is the

lesser percent LAI value from the two ecosystems for species

Enriched Mimic of
Date Succession Succession Monoculture

July 1979 Natural succession 0.66 0.00 0.00
Enriched succession 0.01 0.00
Mimic of succession 0.14

November 1979 Natural succession 0.68 0.00 0.00
Enriched succession 0.06 0.00
Mimic of succession <0.01

April 1980 Natural succession 0.63 0.00 0.00
Enriched succession 0.05 <0.01
Mimic of succession 0.03

October 1980a Natural succession 0.69 0.00 0.00
Enriched succession 0.06 <0.01
Mimic of succession 0.14

Table 5. Community similarity indices (C).
per ecosystem on each date.

Values are based on 180 LAI measurements

aSeptember 1980 for monoculture.


i, and n is the total number of species present in the two

ecosystems. C may range in value from 0 to 1. Community

similarity was high between the natural succession and

enriched succession. The values ranged from 0.66 to 0.69,

with no significant change during the 18 mo period.

Community similarity values for other pairs of ecosystems

were 0 or very low, indicating little or no species overlap.

The natural succession and enriched succession were

comprised of a few abundant species and many race species

Figs. 4 and 5). Most of the abundant species in the natural

succession at 3 mo (July 1979) were also abundant in the

enriched succession. Of the species individually accounting

for L22 of total LAI in the natural succession (number of

species = 9) Ind in the enriched succession (number of

species = 9) at 3 mo, seven were common to the ecosystems

(Table 6). By 18 mo (October 1980) the similarity in

dominant species between the enriched succession and the

natural succession had decreased. Of the 12 abundant

species: (those comprising L2% of total LAI) in the natural

succession, only five were also abundant in the enriched

succession. One of the abundant species in the enricned

succession at 18 mo was an introduced species, plantain

(MUSI earadi~siaca)

The species composition of eacl of the ecosystems changed

during the 1.5 yr study period. The turnover of abundant

species from 3 to 18 mo differed in the enriched succession

~UUu~ ii _U I










.2 .4 .6

1.0 1.2

Tiqure d. Number of soecies in the natural succession by
LIT class. Values are based on 180 THI
neasurements on each date.




1 0l ~ l l

O .2 .4 .6 .8 1.0 1.2 1.4


Figure 5. Number of species in the enriched succession
by LLrI class. Values are based on 180 LAI
measurements on each date.


Table 6. Species accounting for 12% of LAI in four ecosystems.
A dash (-) indicates that a species comprised <2% of
ecosystem LAI.



Mimic of

Phytolacca rivinoides
Momordica charantia
Solanum nigrescens
Borreria laevis
Bocconia frutescens
Clibadium aff. surinamense
Panicum maximum
Hymenachne amplexicaulis
Trema micrantha
Frantzia pittieri
Acalypha macrostachya
Panicum trichoides
Vernonia patens
Mikania sp.

Phytolacca rivinoides
Momordica charantia
Solanum nigrescens
Borreria latifolia
Bocconia frutescens
Clibadium aff. surinamense
Panicum maximum
Vernonia patens
Ipomoea neei
Musa paradisiaca
Ipomoea sp.

Vigna sinensis
Cucurbita pepo
Phaseolus vulgaris
Ipomoea batata
Oryza sativ
Cajanus cajan
Zea mays
Cymbopogon citratus
Manihot esculenta
Crotalaria micans
Musa paradisiaca
Hyptis suaveolens


- .









Table 6--continued.

% of LAI
3 mo 18 mo





Zea may
Manihot esculenta

alncludes at least six species of grasses that were indistinguish-
able by vegetative parts.

blncludes at least four species of sedges that were indistinguish-
able by vegetative parts.

and the natural succession. Two woody species (Bocconia

frutescens and Clibadium aff. suriunamense) and two grass

groups (Pa~nicum maiximum and a group of 10 grass species)

were abundant in both ecosystems at 3 mo and 18 mo (Table

6). However, the enriched succession gained fewer new

dominant species (Table 6), but more species overall

(including all species encountered in the LAI measurements,

Table 4) than did the natural succession from 3 to 18 mo.

The community similarity index between the 3 mo old

vegetation and 18 mo old vegetation was higher in the

enriched succession (C = 0.60) than in the natural

succession (C ; 0.41). This is due both to the addition of

fewer new dominant species and to smaller relative changes

in species abundance over time in the enriched succession.

The 82 species present in the mimic plots at the time of

the October 1980 species inventory represent 46% of the 178

species introduced into the mimic plots from March 1979 to

October 1980. During the first 3 mo of succession, plant

growth and structural development in the mimiC of SUCCeSSion

equaled or exceeded that Lf the natural succession. This

was due primarily to the early and rapid development of

herbaceous species (mainly cultivars) in the mimic. In

subsequent months, development of the mimic was slower. At

18 mo, species richness and plant diversity were lower in

the mimic than in the natural succession. In general, the

mimic was much more similar structurally to the natural

succession than to the monoculture. The structural

differences between the mimic and the natural succession

indicate that (1) there was a time lag between the

devclopment oi the natural succession and the development of

the mimic, and/or (2) some of the species introduced into

the mimic treatment, although morphologically similar to the

native successional species, were not good functional mimics

of the native species.

Large numbers of relatively uncommon species were present

in the natural succession, but not in the aimic, at 3 mo

(Figs. 4 and 6). This probably reflects the initial pattern

of species introductions in the mimic by the investigators.

This difference between the mimnic and the natural succession

elucidates an important characteristic of the natural

succession that was difficult to imitate. The many care

species in the natural succession formed a pool of

potentially important ecosystem components that could

increase in dominance as microenviroornental factors and the

competitive balance of the system changed. In managing the

mimic ecosystem, anticipation of the types of species needed

and introduction of such species at appropriate times to

insure establishment and to maintain a pool of rare species

was difficult.

Several structural characteristics of the mimic at 18 mo,

including species abundance, were similar to characteristics

of the natural succession and enriched succession at a much

') '

O .2 .4 .6 .8 1.0 1.2 1.4


I l

( 1 l1 1l i/





Figure 6. Number of species in the mimic of succession
by LAI class. Values are based on 180 LAI
measurements on each date.



earlier age (3 mo). The number of species intersected by

LAI measurements in the mimic at 18 20 (32 species) is

similar tc the numbers intersected in the natural succession

and enriched succession at 3 mo (37 and 35 species

respectively). This indicates slower development of the

'investigator-controlled' treatment (the mimic) than of the

'nature-controlled' treatments (natural succession and

enriched succession). For example, there was a time lag

between the appearance of woody species in the natural

succession and the selection and introduction of similar

woody species in the mimic. It is expected that longer-term

results will show convergence of structural characteristics

of the minic and natural succession.

The mimic of succession had higher turnover of species

than the enriched or natural succession (Taoles q and 6).

The species composition of the 18 mo old minic was very

dissimilar to that of the 3 ma old mimic (C = 0.15). The

Jull 1979 monoculture and the October 1980 monoculture had

no species in common (C = 0.00). Changes in species

composition in the monoculture were not gradual as in the

other ecosystems; instead, composition changed completely as

one monoculture species replaced another. If community

similarity (C) is used as a measure of rate of species

turnover in each ecosystem, with lower C values indicating

greater changes in species composition during the first 18

mo of succession, then the systems may be ranked by

magnitude of change as follows: monoculture > mimic >

natural succession > enriched succession.

Leaf Area Index

Leaf area index developed rapidly in both the natural

succession and the enriched succession (Fig. 7). The LAI

increased rapidly in all ecosystems during the first 2 mo,

but thereafter was lower in the mimic than in the natural

succession and enriched succession. Seasonal LAI

fluctuations were similar in the natural succession,

enriched succession, and mimic, with maximum values during

the rainy season and minimum values during the dry season.

Increase in LAI was rapid during the growth of the first

maize moncculture (ILA = 1.22 at 2 mo), but leaf area

development of the second maize crop was poor (maximum LAI

0.5). Cassava LAI after 9 mo of growth (mean + 1 s.d. = 2.9

+ 2.0) was not significantly different from LAI in the 7 mo

old natural succession (3.7 + 2.0). A decrease in LAI

occurred during the dry season in the natural succession,

enriched succession, and mimic. At 18 mo, mean LAI (+ 1

s.d.) was n.l( + 2.8 in the natural succession, 5.0 + 3.4 in

the enriched succession, and 3.6 + 3.0 in the aimic.

Vertical distribution of leaf area was similar in the

natural succession and enriched succession (Pigs. 8 and 9),

except in tie lowest (0-25 cm) stratum. In this stratus

near the ground LAI was consistently higher in the enriched




figure 7. LAI in natural succession, enriched succession, mimic of succession,
and monoculture. Values are x + 1 s.e.

( 3.6 )

(2.3 )


O .2 .4 .6 O .2 .4 .6 O .2

LAl (m /m2 ground )

Figure 8. Vertical distribution of leaf area in the
LAI at each age is in parentheses.

natural succession.


(3.7 )


( 5.0 )

O .2

O .2 .4 .6

LAl (m /m2 ground )

.2 .4 .6

Figure 9. Vertical distribution of leaf area in the enriched
LAI at each age is in parentheses.

succession. Total


succession than in the natural succession. This may be due

to the abundance of introduced propagules in the enriched

succession, leading to increased numbers of seedlings. In

the 0-25 cm stratum, 0%, 3.9%, and 6.5% of the LAI was

comprised of introduced species at 8 mo, 13 mo, and 18 mo,

respectively. The LAI in the mimic was concentrated
from the soil surface at 8 mo and 13 mo, and leaf

development higher in the canopy was patchy. By 18 mo the

height of the canopy had increased in the mimic, although

more than half the leaf area was still concentrated <1 m

from the ground (Fig. 10). vertical distribution of leaf

area in the monoculture reflected the growth form of a

single species rather than the interactions among a large

array of species. In the mature cassava monoculture, leaf

tissue was concentrated at 1-3 m above the ground (Fig. 11).

All eccsystems were characterized by rapid growth to an

average canopy height of 3-4 m at 18 mo (Fig. 12). The

natural succession and enriched succession contained some

emergent plants with heights of up to 10.8 m at 18 mo (Table


(2. 1) -1 (I.3) -t (3.6)

O .2 .4 .6 .8 O .2 .4 .6 O .2 .4 .6 .8

LAl (m2/m2 ground)

Figure 10. F'ertical distribution of leaf area in the mimic of succession. Total
LAI at each age is in parentheses.

O .2 .4
LAl (m2/m2 ground )

Figure 11. Vertical distribution of leaf area in the cassava monoculture. Total
LAI is in parentheses.


1979 1980

Figure 12. Vegetation height in natural succession, enriched succession, mimic
of succession, and monoculture. Values are x + 1 s.d.

Table 7. Tallest plants in natural succession, enriched suc-
cession, and mimic at 18 mo, and in cassava monocul-
ture at 10 mo.

Height of tallest
individual (m)











Natural succession

Enriched succession

Mimic of succession


Ochroma pyramidale

Vernonia patens

Bocconia frutescens

Trema micrantha

Vernonia patens

Musa paradisiaca

Manihot esculenta

Ricinus communis

Manihot esculenta

Herbivory Bates

nean herbivory rates varied widely among species, and

among sampling dates for some species (Table 8). For most

species, herbivory rates were not normally distributed. The

Kolomogorov-Smirnov statistic to test the nu~ll hypothesis

that the data were a random sample from a normal

distribution was significant in 50 of 59 tests. Sample

distributions were skewed to the right in most species

studied (Fig. 13). Median losses were lower than mean

Icsses for all species (Table 9). In three species (Panicua

trichoides, Erlthrina costaricensis, and Tanihot esculenta),

damage distribution was dependent on the type of ecosystem

in which the species was found (Fig. 14 Fig. 16).

Of the eight species monitored in both the natural

succession and the enriched succession, one species (Panicum

trichoides) had different herbivory rates in the two

ecosystems. This species had a lower rate in the enriched

succession than in the natural succession (Table 9). For

the two species monitored in the enriched succession and in

the mimic of suc-cession (Eryhrina co~staricensis) and

MIan ihot esuet) both had lower herbivory rates in the

enriched succession. Manihot also had lower rates in the

monoculture than in the mimic.

Some ecosystem characteristics that may affect the

herbivory rate on an individual species are species

diversity, LAI, and species composition. In addition, the

Table 8. Mean herbivory losses by species and ecosystem. Losses are x (s.d.), in
m2/m2 leaf/day; n is number of leaves (alternate-leaved species), or number
of leaf pairs (opposite-leaved species).

Natural Enriched Mimic of
Succession Succession Succession Mlonoculture
Species Date n loss n loss n loss n loss



Clibadium aff.

Oct. 79

Feb. 80

June 80

Oct. 79

Feb. 80

June 80

Oct. 79

Feb. 80

June 80

37 16.5

33 12.7

26 9.8

13 30.5

25 11.9

14 9.2

6 13.7

17 16.0

9 13.5

21 14.4

34 27.1

27 3.5

10 25.6

24 15.0

13 51.0

13 17.9

14 9.2

16 16.1

Table 8--continued.

n loss

n loss

Mimic of
n loss

n loss



Panicum maximum


Cordia inermis



4 16.3

5 6.8

9 13.0

9 14.7

8 12.8

9 15.5

3 8.4

Oct. 79

Feb. 80

June 80

Oct. 79

Feb. 80

June 80

June 80

Oct. 79

Feb. 80

June 80

32 6.3

30 12.5

19 21.4

16 3.5

3 0.6

6 7.6

9 13.5

9 29.7

11 36,4

Table 8--continued.

n loss

n loss

11 46.3

29 34.1

20 77.9

Mimic of
n loss

n loss

Vernonia patens







Feb. I

June 1

Feb. I

June 1

June I

June I


Feb. I


34 24.2

6 131.4

15 15.6

9 21.5

13 7.8

21 2.0

5 2.3

13 11.5

7 10.6

Table 8--continued.

n loss

n loss

12 1.4

3 51.6

Mimic of
n loss

n loss








Zea mays













10 4.8

18 13.2

24 14.8

22 57.5

16 26.2

10 5.6

25 44.9

14 6.2

26 7.3



















n loss


Table 8--continued.

n loss

17 4.0

3 0.8

Mimic of
n loss

241 35.7

1 1.7

4' 27.7

27 12.5

13 62.9

5 0.5

19 1.4

15 0.9

3 1.1

3 0.6

n loss

68 11.5

4 9.1

Natural Enriched Mimic of
Succession Succession succession Mlonoculture
Species Date n loss n loss n loss n loss

Musa June 80 4 3.8
paradisiaca (2.9)

Ipomoea Oct. 79 19 28.4
batata (26.3)

Feb. 80 3 103.7

Carica papaya June 80 11 3.2

Crotalaria June 80 6 1.9
micans (1.9)

alncludes at least six species of grasses that were indistinguishable by vegetative

blncludes at least four species of sedges that were indistinguishable by vegetative

Table 8--continued.

~_3 _L,


40- CyperoCeae



LO55 (cm /m /doy)


40. POnicum maximum

3 0 -



O a l I
0 50 100
LOSS Icm2/m /doy)

Cirbadlum off. surinomense

10 t

0 01010

LOSS (cm2/m2/day)

150 200


20 Baccoml



0 50

SS5 (cm2/m2/day

50 100 150 200
LOSS (cmll2/m day)

Figure 13. Distribution of loss to herbivores among
leaves in six common species.


Vernonla potens

O 50 100 150
LOSS (cm2 Im2 /day )


Ph~ooC rivmoaldes

Table 9. Mean and median herbivory rates on selected species in different ecosystems.
Number of leaves (n) includes samples from all dates for which treatment
comparisons could be made.

Species Ecosystem n x (s.d.) Median p Valuea


Clibadium aff.




Natural succession
Enriched succession

Natural succession
Enriched succession

Natural succession
Enriched succession

Enriched succession
Mimic of succession

Natural succession
Enriched succession

Enriched succession
Mimic of succession

Natural succession
Enriched succession

Natural succession
Enriched succession

Natural succession
Enriched succession

52 15.8 (17.2)
47 27.2 (40.8)

32 14.9 (25.5)
43 14.4 (13.0)

9 21.5 (39.1)
7 10.6 (12.7)

24 14.9 (20.2)
22 57.5 (40.8)

20 33.4 (63.8)
18 9.4 (17.0)

.70, .69, .77

.32, .31,.72

.83, .79, .63

<.01, <.01, <.01

.19, .18, .05

--,<.01, <.01

.18, .17, .54

<.01, <.01, .01

.82, .82, .76












18 12.0 (18.6)
26 14.4 (12.4)

19 21.4 (26.1)
16 3.5 (5.0)

96 13.4 (26.2)
82 16.1 (33.0)

Species Ecosystem n : sd. edian p Valuea

Vernonia paes Natural succession 34 24.2 (29.4) 11.2 .06, .06, .06
Enriched succession 29 34.1 (34.9) 21.8

aWilcoxon 2-sample rank sums test, Kruskal-Wallis test, median test.

bpor this species, n is number of opposite leaf pairs.

cIncludes at least four species of sedges that were indistinguishable by vegetative

dIncludes at least six species of grasses that were indistinguishable by vegetative

Table 9--continued.

1 I

1 rmnrl nn






> 20-

Panicum tricholdes


1 i 1 lI i l


l i I
0 50 100

LOSS ( cm2 /m2 /doy )

Figure 14. Loss distribution among leaves of Panicum
trichoides .

z 4
a 4

30 -



Erythrina cosarcensis



O 50 100 150
LOSS (cm2/m2/day)
Figure 15. Loss distribution among leaves of Erythrina costaricensis.

1 I ,nmn

Monihot esculento



Z 20- 60-

I '10 50-

*io0 100 150

~50 30-

40- 20-

301 MIMIC 10-

o 50 100 150 300

LOSS (cm /rn /day)

Figure 16. Loss distribution among leaves of Manihot esculenta.


abundance and spatial distribution of a particular species

within the system may affect its herbivory rate. Few

differences in herbivory rates between natural and enriched

succession were expected, because these systems were very

similar; in species diversity, LAI, and species composition.

Panicum tricholdes had relatively low LAI in both systems

(0.04 in natural succession, 0.07 in enric-hed succession).

Thus differential plant abundance was probably not an

important factor affecting herbivory rate for this species.

Plant spatial distribution and/or small sample size may

explain the observed difference.

Several factors may contribute to the higher herbivory

rates on Erythrina in the mimic than in the enriched

success ion. Abundance of Egythrina was similar in the two

systems. Although both systems had relatively high species

diversity, the species similarity between the systems was

low. In addition, the LAI of the simic was lover than the

LAI of the enriched succession. This suggests that the

kinds of species that succound a given plant, as well as

their abundance, may affect the herbivory rate on that

plant. Manihot and Eryth~rina (both cultivars) had lower

apparency and greater protection trom herbivores when

surrounded by native successional species in the enriched

succession plots, than when planted in plots containing a

different array of species including many cultivars.

Manihot, a relatively unpalatable; species, had its

highest herbivory rate in the ecosystem with intermediate

species diversity and LAl (the mimic). The herbivory rate

on this species was not linearly related to species

diversity. This result suggests that species composition,

rather than diversity p~er _se, was an important factor

influencing herbivory on Manihot.

There was no simple relationship between LAI of a species

and that species' herbivory rate. However, the data

indicate that in the natural succession, enriched succession

and mimic, the very high rates of herbivory occurred on the

less common species, and a11 Of the Very COmmOR species (LAZ

S0.5) had relatively low herbivory rates (Fig. 17). The

loss rate for each species (cmz m-2 leaf day-i) was

multiplied by the LAI of the species to obtain the species'

loss rate in cmz ra-2 ground day-i. Some relatively uncommon

species contributed significantly to the total ecosystem

loss to herbivores (Fig. 18).

The coefficient of variation (CV = s.d./mean) of

heroivory rates was used to identify trends in the spatial

distribution of damage among leaves and plants of several

species. A large coefficient of variation (i.e., s.d. >

mean) indicates high variability in herbivory rate among

leaves or plants, and implies aggregation of damage, with

some leaves or plants receiving very high levels of damage

and others receiving very low levels. A low CV value (i.e.,

.4 .6 .e 1.0 1.2





O .2 .4 .6 .8

0 .2 .4

.9 1.0 1.2


I .


1.4 1.6

3 0


I I I l l I I i a 1s
.2 .4 .6 .8 1.0 l.2

1.4 1.6 !.8 2.0 2.2 2.4

ores by LAI. Each point

Figure 17.

Losses to herbive

represents one sE







5-0 o


8 IO 12 14 16 18 20 22 24


Figure 18. Herbivory rates per unit ground area by LAI. Each point represents
one species.


s.d. < mean) indicates that spatial variability of damage is

low and implies that damage tends to be evenly distributed

among leaves or plants. The CV calculated using mean leaf

herbivory rates reflects the damage distribution among

leaves of a given species; the CV calculated using mean

plant herbivory rates reflects the damage distribution among

plants. The CV values calculated from leaf herbivory rates

were higher, on the average, than the values calculated from

plant herbivory rates (Table 10). This implies that

leaf-to-leat damage variability was higher than

plant-to-plant variabilit". In other words, most damage

from herbivores tended to be aggregated on a subset of the

leaves of a species, but all plants of the species in the

same ecosystemn were equally likely to have some leaves

heavily damaged by herbivores.

Both leaf-to-leaf and plant-to-plant variability were

high in cassava. This result reflects the foraging pattern

of one of cassava's major herbivores, the leaf-cutter ants

(Atta c~eehalotes). These ants selected a few plants of

cassava for consumption (leaving many other individuals

untouched), and removed some (but not all) leaves of each

selected plant almost entirely, leaving only the mid-ribs.

Young leaves and old leaves of most species were consumed

at equal rates. Percent leaf expansion during the

monitoring period was used as an inaicator of leaf age (high

percent expansion = young leaf; low percent expansion = old

Table 10. Coefficients of variation (CV) of herbivory rates
by species. Coefficient of variation is calculated:
(1) based on individual leaf data and (2) based on
plant data.

Based on Leaf Data Based on Plant Data
Number Number
of of
Species Leaves CV Plants CV

Cordia inermis 62 4.89 10 2.07

Manihot esculenta 140 2.57 33 1.95

Gramineaea 38 2.21 9 0.81

Phytolacca 178 2.02 27 1.53

Cajanus cajan 40 1.94 7 1.74

Momordica charantia 21 1.87 5 1.70

Hymenachne 46 1.71 11 1.21

Panicum trichoides 35 1.61 10 1.27

Bocconia frutescens 99 1.46 25 1.41

Frantzia pittieri 28 1.39 6 0.68

Phaseolus vulgaris 25 1.26 5 0.81

Panicum maximum 44 1.12 13 0.76

Cymbopogon citratus 39 1.10 11 0.71

Erythrina 46 1.08 10 0.91

Vernonia patens 94 1.01 16 0.70

Ipomoea batata 22 0.98 7 0.79

alncludes at least six species of
guishable by vegetative parts.

grasses that were not distin-

University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - - mvs