Community disassembly, biodiversity loss, and the erosion of an ecosystem service

Abstract

Distinguishing the mechanisms responsible for the relationship between bio-diversity and ecosystem services requires knowledge of (1) the functional roles played by individual species and (2) the sequence with which species are added to or lost from communities in nature (i.e., ''community assembly'' and ''community disassembly,'' re-spectively). Rarely, if ever, are both these issues understood with certainty in any given ecosystem. We used an empirically based simulation model to assess the degree to which the sequence of species loss from vertebrate communities influences risk of human exposure to Lyme disease, as measured by the proportion of ticks infected with the etiological agent. Dramatic differences in the shapes of the curves relating vertebrate biodiversity to disease risk (which we consider an ecosystem service) were observed. Randomized sequences of species loss resulted in a decrease in disease risk with reduced biodiversity, a result that is contradicted by both empirical observations and model results from nonrandomized sequences of species loss (i.e., specific ''disassembly rules''). All potentially realistic dis-assembly rules resulted in increases in disease risk with decreasing biodiversity, although shapes of the curves varied considerably. Our results highlight the importance of both species identity and the order by which species are lost, in understanding the mechanisms by which biodiversity affects ecosystem functioning.

Full text

1421
Reports
Ecology,
84(6), 2003, pp. 1421–1427
q
2003 by the Ecological Society of America
COMMUNITY DISASSEMBLY, BIODIVERSITY LOSS, AND THE EROSION
OF AN ECOSYSTEM SERVICE
R
ICHARD
S. O
STFELD
1
AND
K
ATHLEEN
L
O
G
IUDICE
2
Institute of Ecosystem Studies, Box AB, 65 Sharon Turnpike, Millbrook, New York 12545 USA
Abstract.
Distinguishing the mechanisms responsible for the relationship between bio-
diversity and ecosystem services requires knowledge of (1) the functional roles played by
individual species and (2) the sequence with which species are added to or lost from
communities in nature (i.e., ‘‘community assembly’’ and ‘‘community disassembly,’’ re-
spectively). Rarely, if ever, are both these issues understood with certainty in any given
ecosystem. We used an empirically based simulation model to assess the degree to which
the sequence of species loss from vertebrate communities influences risk ofhuman exposure
to Lyme disease, as measured by the proportion of ticks infected with the etiological agent.
Dramatic differences in the shapes of the curves relating vertebrate biodiversity to disease
risk (which we consider an ecosystem service) were observed. Randomized sequences of
species loss resulted in a decrease in disease risk with reduced biodiversity, a result that
is contradicted by both empirical observations and model results from nonrandomized
sequences of species loss (i.e., specific ‘‘disassembly rules’’). All potentially realistic dis-
assembly rules resulted in increases in disease risk with decreasing biodiversity, although
shapes of the curves varied considerably. Our results highlight the importance of both
species identity and the order by which species are lost, in understanding the mechanisms
by which biodiversity affects ecosystem functioning.
Key words: biodiversity; blacklegged tick; community assembly; community disassembly;disease
ecology; ecosystem function and ecosystem services; functional redundancy; habitat fragmentation;
Lyme-disease risk; vertebrate ecology; white-footed mouse.
I
NTRODUCTION
The predominant approach to assessing the ecolog-
ical consequences of biodiversity loss is to assemble
experimental communities by drawing species random-
ly from a species pool and determine how community
properties or ecosystem functioning respond to varia-
tion in species richness. Such an approach, although
rigorous experimentally, leaves open the question of
whether enhanced ecosystem functioning in more di-
verse systems is caused by biodiversity per se or by
the chance inclusion of particular species with domi-
nant roles (Loreau et al. 2001). In addition, such studies
cannot provide insight into the importance of the se-
quence of species addition (community assembly) or
loss (community disassembly).
Distinguishing the mechanisms by which biodiver-
sity influences a specific ecosystem property or service
in natural systems requires knowledge of (1) the func-
tional roles played by individual species, and (2) the
sequence with which species are added to or lost from
communities in nature. Rarely, if ever, are both these
issues understood with certainty in any given ecosys-
tem (Tilman et al. 1997, Knops et al. 1999, Chapin et
Manuscript received 30 September 2002; revised 5 February
2003; accepted 6 February 2003. Corresponding Editor: O. V.
Schmitz.
1
E-mail: Rostfeld@ecostudies.org
2
Present address: Department of Biology, Union College,
Schenectady, New York 12308 USA.
al. 2000, Cottingham et al. 2001, Loreau et al. 2001,
Schwartz et al. 2000).
Recent empirical and modeling studies have revealed
a novel ecosystem property associated with biodiver-
sity—the reduction in transmission of vector-borne dis-
ease (Ostfeld and Keesing 2000
a
,
b
, Gilbert et al. 2001,
Schmidt and Ostfeld 2001). According to this model,
termed the ‘‘dilution effect,’’ high species diversity in
the community of vertebrate hosts for tick vectors re-
duces risk of human exposure to tick-borne infections
by diluting the impact of highly competent disease res-
ervoirs. The model was developed for the Lyme disease
system in eastern North America, in which the tick
vector (
Ixodes scapularis
) feeds from a diverse assem-
blage of mammals, birds, and reptiles, but acquires the
etiological agent (
Borrelia burgdorferi
) most efficient-
ly from the white-footed mouse (
Peromyscus leuco-
pus
); however, the dilution effect may also operate in
many other vector-borne disease systems (Ostfeld and
Keesing 2000
b
). The model predicts that the presence
of alternative (non-mouse) hosts in communities with
high species diversity results in fewer tick meals being
taken from mice, a lower infection prevalence in the
tick population, and perhaps a smaller tick population
due to poorer feeding success on non-mouse hosts (Ost-
feld et al. 2002). The dilution effect is enhanced when
the species added to increasingly diverse communities
directly or indirectly reduce population density of
white-footed mice, or reduce tick burdens on mice via
competition for parasites (Schmidt and Ostfeld 2001).

Reports
1422
RICHARD S. OSTFELD AND KATHLEEN L
O
GIUDICE
Ecology, Vol. 84, No. 6
We (LoGiudice et al. 2003) recently parameterized
a model that accounts for the role of each potentially
important vertebrate species in determining the nymph-
al tick infection prevalence (or NIP), which is an epi-
demiologically relevant risk factor for Lyme disease
(Ostfeld et al. 2002). Lyme disease is transmitted pri-
marily by the nymphal stage of the tick vector (Barbour
and Fish 1993), and risk of human exposure is reflected
by the proportion of host-seeking nymphs that harbors
a
B. burgdorferi
infection. Parameterization of the
model required that field estimates be made for each
species of vertebrate host regarding: (1) larval tick bur-
den (average number of larvae per host individual); (2)
host population density; and (3) reservoir competence
(the proportion of larvae that acquire an infection from
that species of host). After exhaustively trapping and
netting the potentially important mammalian and avian
hosts at a local field site in southeastern New York
State (USA), we generated an estimate of NIP that was
matched closely by empirical data from naturally oc-
curring populations of host-seeking nymphs, indicating
that the model and parameter values were representa-
tive and accurate (LoGiudice et al. 2003). Our data,
therefore, provide reliable estimates for the functional
role that each species of host plays in influencing NIP,
and they pave the way for understanding mechanisti-
cally the role of biodiversity in influencing disease risk.
Little is known about the sequence with which ver-
tebrate species are added to or lost from native eco-
logical communities that vary in diversity. The liter-
ature on species assembly rules (e.g., Fox and Brown
1993, Stone et al. 1996) suggests that the order in which
species are added to increasingly diverse communities
over evolutionary time scales is nonrandom, but this
literature is of limited relevance to communities influ-
enced by highly accelerated, anthropogenic alterations
of species diversity. Although forest fragmentation is
known to reduce species richness of forest vertebrates
(Laurance et al. 2000, Crooks 2002, Donovan and
Flather 2002), little is known about either the specific
sequence of loss, or about the species-specific traits
that might influence their likelihood of being lost from
a disturbed or fragmented ecosystem. We term the pos-
sible relationship between species-specific life-history
traits and the probability of local extinction from com-
munities experiencing disturbance or fragmentation,
‘‘community disassembly rules.’’ In this paper we use
our empirically parameterized model of the individu-
alistic roles of vertebrate hosts in determining NIP to
assess, using computer simulations, the impact of a set
of plausible community disassembly rules on Lyme
disease risk. The primary objective is to evaluate
whether different sequences of species loss cause dif-
ferences in the shape of the relationship between bio-
diversity and this particular ecosystem function. If so,
the importance of the disassembly rules that govern
patterns of species loss will be underscored as relevant
to the biodiversity–ecosystem function debate, and the
relative importance of species composition vs. species
number (e.g., Wardle et al. 1999) highlighted.
T
HE
M
ODEL
The model is an extension of Giardina et al. (2000),
which calculates the predicted nymphal tick infection
prevalence (NIP) as the sum across all species of the
nymphs infected by each species divided by the total
number of nymphs fed by all species. Parameters of
the model include: density of host species,
N
i
; species-
specific body burdens,
B
i
; and species-specific reservoir
competence,
C
i
. Therefore,
m
i
5
N
i
B
i
, where
m
i
is the
number of larval meals taken from species
i
;
I
i
5
m
i
C
i
,
where
I
i
is the number of nymphs infected from their
larval meal on species
i
; and the total number of
nymphs infected from their larval meal (
I
T
)is
I
T
5S
m
i
C
i
. The number of nymphs not infected in their larval
meal,
U
i
5
m
i
(1
2
C
i
) and the total number of nymphs
uninfected is
U
T
5S
m
i
(1
2
C
i
). Thus, the total nymph-
al infection prevalence (NIP
T
) is NIP
T
5
I
T
/(
I
T
1
U
T
).
This allows us to describe the contribution of each
species to NIP. Assumptions of this model, and methods
for determining parameter values empirically for each
species of host, are described in LoGiudice et al.
(2003). The model was validated by a close fit between
the NIP predicted by the set of species-specific param-
eter values as determined in a local, relatively intact
community consisting of 13 species of mammalian and
avian hosts and the NIP determined empirically by as-
saying field-collected, host-seeking nymphal ticks
(LoGiudice et al. 2003). Sensitivity analyses showed
that the close fit between predicted and observed NIP
was robust (
6
1–3% change) to 20% variation in host
population density estimates—the parameter with the
largest potential error (LoGiudice et al. 2003).
S
IMULATIONS OF
C
OMMUNITY
D
ISASSEMBLY
We used the model to explore the outcomes (values
of NIP [nymphal tick infection prevalence]) predicted
by disassembling host communities according to spe-
cific rules. Initially, we evaluated the hypothesis that
NIP increases with decreasing vertebrate diversity
when species are removed randomly to deconstruct an
initially intact community. Beginning with an intact
community of 13 species of mammals and birds known
to comprise a nearly complete set of tick hosts at our
field site in Dutchess County, New York, USA (Ap-
pendix), a specified number of species was drawn in
random order for removal to create communities with
species richness values of 12, 11, 10 ...1(100 iter-
ations for each community size). We compared this
completely random disassembly procedure to model
runs in which we assumed that the white-footed mouse
is present in all communities, and a third set of runs
assuming that mice and white-tailed deer are always
present. These latter simulations were intended to more
closely reflect the real world. Strong evidence supports
the assertion that white-footed mice are ubiquitous

June 2003 1423
COMMUNITY DISASSEMBLY, ECOSYSTEM SERVICE
Reports
members of vertebrate communities, occurring in both
highly fragmented landscapes and pristine, intact com-
munities (Nupp and Swihart 1996, 2000, Krohne and
Hoch 1999, Rosenblatt et al. 1999). In addition, be-
cause of the importance of white-tailed deer as hosts
for adult
Ixodes scapularis
(Barbour and Fish 1993),
endemic Lyme disease may require the presence of this
host. Thus, the comparison between a completely ran-
dom removal of species on the one hand, and the as-
sumption that either mice only, or mice plus deer, are
always present, allows us to evaluate the consequences
of a simple, empirically based rule. In this case, white-
footed mice are known to be a high-impact species for
Lyme-disease risk, and we can ask how a null model
(complete randomness) compares with somewhat more
realistic models of community disassembly.
In both cases, we set the population density (
N
i
)of
non-mouse members of the community at either em-
pirically determined values for our site (chipmunks,
deer, birds) or a value determined from the literature
to represent an average for oak/mixed-hardwood for-
ests of the northeastern United States (see LoGiudice
et al. [2003] for details). However, to reflect the dra-
matic population fluctuations that typify white-footed
mice (e.g., Ostfeld et al. 2001) we ran separate simu-
lations at levels of mouse density between 25 and 100
individuals/ha. From the 100 simulations at each com-
bination of species richness and mouse abundance we
calculated the mean (
6
1
SE
) nymphal infection prev-
alence expected from that level of host community di-
versity.
We followed simulations of randomized community
disassembly with simulations employing strict disas-
sembly rules. We removed species in the orders indi-
cated below until only white-footed mice (set at 25
individuals/ha to reflect long-term averages at local
sites; Ostfeld et al. 2001) and white-tailed deer re-
mained. To simulate plausible sequences of species loss
from landscapes subjected to forest destruction and
fragmentation, we removed species according to the
following rules: species are lost (1) in decreasing order
of body mass; (2) in decreasing order of home-range
size; (3) from highest to lowest trophic level; and (4)
in the approximate order described for Midwestern U.S.
mammals in forest patches in an agricultural matrix
(Rosenblatt et al. 1999, Nupp and Swihart 2000; see
Appendix 1 for rankings). Our intention was to also
simulate loss in decreasing order of ecological spe-
cialization, but data were insufficient for reliable rank-
ing.
We also wished to evaluate the degree to which in-
teractions among the various species in our virtual com-
munities might influence the shape of the relationship
between diversity and ecosystem functioning. How-
ever, assessing interactions among vertebrate species
that might affect their role in feeding and infecting tick
populations is not straightforward. As a first attempt
at this assessment, we allowed species added to virtual
communities to interact with white-footed mice either
by reducing their numbers or by reducing their tick
burdens (e.g., by competition for parasites). Data do
not exist that would allow us to estimate empirically
the net effects of each species added on either popu-
lation size or tick burdens on white-footed mice.
(Moreover, in real communities, the species added
could potentially affect both population sizes and tick
burdens on all other species present, not only mice.)
Given these severe data limitations, we chose to assign
species to categories according to the expected inten-
sity of their reduction in mouse numbers and tick bur-
dens on mice. Thus, strong competitors such as sciurid
rodents and
Blarina
shrews were each expected to re-
duce mouse abundance by 1%, generalist predators
such as skunks and raccoons by 0.5%, and weak com-
petitors such as birds,
Sorex
shrews, and opossums by
0.1%. Similarly, based on empirical demonstration of
strongly reduced average tick burdens on mice with
increasing population density of chipmunks (Schmidt
et al. 1999), we expected small mammals (shrews,
sciurid rodents), which overlap strongly with mice in
microhabitat use, to exert the strongest reduction in
tick burdens on mice (16%), whereas mesomammals
(skunks, raccoons, opossums) would reduce ticks on
mice by 5%, and birds by 1%. We emphasize that the
assignment of these values represents a crude first at-
tempt to assess the potential for interactions among
species, defined relative to a specified ecosystem func-
tion, to influence model outcome.
R
ESULTS
Randomized community disassembly
The relationship between vertebrate diversity and
Lyme-disease risk (represented by bacterial infection
prevalence in nymphal ticks) differed dramatically be-
tween the simulated communities that were disassem-
bled randomly vs. those in which white-footed mice,
or mice plus deer, were always present. When all host
species, including mice and deer, were removedin ran-
dom sequence, species richness of hosts had a curvi-
linear, positive effect on nymphal infection prevalence
(NIP), such that species-poor communities had the low-
est Lyme-disease risk (Fig. 1). In contrast, when mice
were present in all communities, and the removal se-
quence of non-mouse hosts was randomized, a strong
negative relationship was observed between host spe-
cies richness and NIP (Fig. 1, mouse density set at 25
individuals/ha). Simulations in which mice plus deer
were always present were similar to those in which
mice were always present; in those two scenarios, pop-
ulation density of mice (25, 50, or 100 individuals/ha)
was positively correlated with NIP, as has been ob-
served empirically (Ostfeld et al. 2001). In simulations
with completely random removal sequences, the high-
est NIP typically occurred at intermediate mouse den-
sities, although mouse density had only a modest effect

Reports
1424
RICHARD S. OSTFELD AND KATHLEEN L
O
GIUDICE
Ecology, Vol. 84, No. 6
F
IG
. 1. Results of simulations assessing the effects of reduced species richness on Lyme disease risk, as measured by
nymphal tick infection prevalence (NIP). Data are means and 1
SE
of 100 replicates Three types of simulations were run
with communities that began as intact assemblages of 13 host species for which we (LoGiudice et al. 2003) empirically
determined contributions to NIP: (1) species were removed one by one by a random-selection criterion; (2) white-footed
mice were present in all communities, but otherwise removal was in random order; and (3) mice and white-tailed deer were
present in all communities, but otherwise removal was in random order.
(data not shown). For all simulations, little change in
NIP occurred as species richness declined from 13 to
8 species, after which NIP changed dramatically (Fig.
1).
Community disassembly with rules
The application of different rules by which com-
munities disassemble under habitat destruction or frag-
mentation caused dramatic variation in the relationship
between host species richness and Lyme-disease risk
(Fig. 2A). When species were lost in order of largest
to smallest home-range size or body mass, the pattern
of change in NIP was characterized by a gradual in-
crease as species richness declined from 13 to 4 species,
followed by a rapid increase as richness declined from
4 to the final 2 species (mice plus deer). Because home-
range size scales strongly to body size, differences be-
tween these two models were subtle. In stark contrast,
when species were lost in order of highest to lowest
trophic level, a
decrease
in NIP was observed in com-
munities as richness declined from 13 to 4 species,
followed by a major irruption in NIP as the community
lost 2 of its last 4 species. The most gradual pattern of
changing NIP with community disassembly was seen
for communities losing species in approximately the
order observed in forest patches within agricultural ma-
trices observed in Indiana and Illinois (Nupp and Swi-
hart 1996, 2000, Rosenblatt et al. 1999) (Fig. 2A).
Qualitatively similar patterns were observed in mod-
el runs in which non-mouse species were assigned in-
teraction coefficients that affected either population
density of mice (via predation or competition) or tick
burdens on mice (via competition for generalist para-
sites) (Fig. 2B). As would be expected from the inclu-
sion of interaction coefficients, NIP was always lower
for interaction models than for corresponding runs of
non-interaction models. Inclusion of interaction coef-
ficients did not change the shapes of the biodiversity/
NIP curves under any of the disassembly rules sce-
narios.
D
ISCUSSION
As habitat destruction and fragmentation reduce spe-
cies diversity in human-impacted communities, eco-
system properties and the services derived from them
might be altered. However, change in ecosystem prop-
erties might be modest or nonexistent in the face of
biodiversity loss if the remaining species compensate
for the lost contributions of missing species, or if the
species likely to be lost are also likely to have little
effect on the ecosystem property of interest. For any
given ecosystem property, determining the effects of
reduced species diversity requires knowledge of (1) the
functional roles played by individual species in gov-
erning the property of interest; (2) the degree to which
species interactions affect functional roles; and (3) the
likelihood of local extinction of individual species.
For the Lyme disease system, we have determined
the roles that individual host species play in influencing
nymphal infection prevalence (NIP) (LoGiudice et al.
2003), which, due to its association with human disease
risk, we consider an important ecosystem function. We

June 2003 1425
COMMUNITY DISASSEMBLY, ECOSYSTEM SERVICE
Reports
F
IG
. 2. Results of simulations assessing the effects of reduced species richness on Lyme disease risk, as measured by
nymphal tick infection prevalence (NIP). Specific community disassembly rules were applied to reduce richness from com-
munities that began as intact assemblages of 13 host species for which we (LoGiudice et al. 2003) empirically determined
contributions to NIP. Species were removed in decreasing order of: (1) body size; (2) home-range size; and (3)trophic level;
or (4) according to observations made in fragmented landscapes in the Midwestern United States. In (A) non-mouse species
did not affect population density of mice or tick burdens on mice; in (B) interaction coefficients were incorporated (see
Results: Community disassembly with rules
).
do not know, but can estimate, the degree to which host
species interact to affect NIP. Our main goal was to
ask whether different rules governing the sequence by
which species are lost from human-impacted commu-
nities (community disassembly rules) result in different
shapes of the biodiversity–NIP relationship. In our ini-
tial simulations of vertebrate host communities ranging
in species richness from 13 to 1 species, we found that
removing species randomly from an initially intact
community resulted in no relationship between species
richness and NIP until only
;
4 species remained, fol-
lowed by a strong decrease in NIP as richness declined
from 4 to 1 species. Under the more realistic scenario
of white-footed mice being ubiquitous members of all
communities, followed by random draws for species
removal, we found a strong, saturating decline in NIP
with increasing species richness (cf. Schwartz et al.
2000). Assuming that both mice and deer are always

Reports
1426
RICHARD S. OSTFELD AND KATHLEEN L
O
GIUDICE
Ecology, Vol. 84, No. 6
present gave similar results. This exercise illustrates
that the incorporation of even rudimentary knowledge
about the sequence by which natural communities are
assembled or disassembled can result in profound dif-
ferences in the postulated relationship between biodi-
versity and an ecosystem service, such as risk of ex-
posure to an infectious disease.
Our simulation of community disassembly under dif-
ferent plausible sets of rules governing the sequence
of species loss demonstrated that these rules might in-
deed cause dramatic changes in the way NIP declines
with the loss of species. When species were removed
according to patterns observed in fragmented agricul-
tural landscapes of the Midwestern United States, the
resulting increase in NIP was gradual. When species
were lost in decreasing order of body size or home-
range size, increasing NIP underwent abrupt transitions
in the decline from 10 to 7 species and again in the
decline from 4 to 2 species. When species were re-
moved in order of highest to lowest trophic level, a
counterintuitive decrease in NIP occurred as richness
declined from 13 to 4 species, followed by a more than
doubling of NIP as the final two species were lost.
Differences in the shape of the biodiversity–NIP
curves with different removal sequences resulted from
two processes. First, the vertebrate host species dif-
fered strongly in their contributions to total NIP. Host
species can be arrayed along two axes that determine
their impact on NIP. The first axis is the number of
ticks fed by the population of any given host species
(which is the product of the host population density
and average tick burden per individual). This value
determines the magnitude of the effect of the loss (or
inclusion) of that species. The second axis is the in-
fectivity (or reservoir competence) of that host species
for feeding ticks, a value that determines the direction
of the effect (higher or lower NIP) when that species
is lost or added. Some species feed and infect many
ticks (e.g., white-footed mice, eastern chipmunks, and
shrews), others feed many ticks but tend not to infect
them (‘‘dilution hosts’’—e.g., the tree squirrels), and
others feed and infect few ticks (e.g., carnivores, birds).
Variation in model outcome arising from different
disassembly rules results from changes in the timing
of loss of species of high impact, either positive or
negative. For example, for the body-size and home-
range-size rules, squirrels are lost in the second ex-
tinction wave (going from 10 to 7 species) and both
shrews were lost in the fourth cut (from 4 to 2 species).
These were the transitions associated with the greatest
change in NIP. In contrast, for the trophic-level rule,
the loss of all shrews in the first cut caused a decrease
in NIP, and the persistence of squirrels until the last
cut prevented NIP from increasing until the final 2
species were lost. We suggest that such idiosyncratic
effects of particular extinction sequences will typify
communities in which species differ in the direction
and magnitude of effects on ecosystem function.
An additional important mechanism behind the di-
verse responses to disassembly rules is the contingent
nature of species impacts on NIP. The degree to which
the loss of any particular species in our system in-
creased or decreased NIP depended on the identities of
the other species present. For example, if a moderately
reservoir-competent host, such as the chipmunk or
Blarina
shrew, is lost from a low-diversity community
with high NIP, the result will be an increase in NIP.
But if these same species are lost from a high-diversity
community with low NIP, the result will be a strong
decrease in NIP.
Our analyses have some important limitations. First,
although we found that including interaction terms be-
tween white-footed mice and other members of the
community had only a minor effect on NIP, we did not
assess the impact of allowing all species to interact
with one another. Real ecological communities incor-
porate networks of interactions such that the inclusion
or exclusion of one species can, via indirect pathways,
cause unexpected changes in abundance of others
(Lawton 2000, Pimm 1993). Although a food-web ap-
proach to assessing net effects of species loss on abun-
dance and tick burdens of each host species would add
realism, the data that would allow parameterization of
such a model are lacking. Second, our set of plausible
disassembly rules is incomplete, owing to poor infor-
mation on what factors influence species vulnerability
to local extinction. Degree of habitat or trophic spe-
cialization, initial population density, sensitivity to hu-
man artifacts, or other features could potentially influ-
ence the sequence of loss.
Our study has potentially important implications for
recent studies that have examined the relationship be-
tween biodiversity and other ecosystem properties such
as primary production, resistance to invasion, and al-
bedo (Knops et al. 1999, Cottingham et al. 2001, Lo-
reau et al. 2001). The relationships between biodiver-
sity and both NIP and the more traditional ecosystem
functions are dependent on the same factors—the in-
teractions among species and their role in governing
the functions under investigation. Such relationships
are likely to be equally sensitive to the pattern of dis-
assembly or assembly, but studies to date have not
examined this possibility.
In conclusion, results of our simulations suggest that,
because individual species have effects that are both
idiosyncratic and strongly contingent on the identities
of the other members of the community, the sequence
of species loss (or addition) is crucial to the relationship
between biodiversity and ecosystem processes, repre-
sented by Lyme-disease risk. Determining the rules that
govern the process of species disassembly and assess-
ing the net effects of species losses on the abundance
of other species should be included in any deliberations
over the role of biodiversity in governing ecosystem
properties and functioning and the services we derive
from them.

June 2003 1427
COMMUNITY DISASSEMBLY, ECOSYSTEM SERVICE
Reports
A
CKNOWLEDGMENTS
We are grateful to C. Canham, V. Eviner, P. Hudson, F.
Keesing, K. Schmidt, O. Schmitz, and an anonymous re-
viewer for advice and comments on a draft, and to F. Keesing
for stimulating discussions and insights. Financial support
was provided by NIH (R01 AI40076), and NSF (DEB
9807115 and 0075277). This is a contribution to the program
of the Institute of Ecosystem Studies.
L
ITERATURE
C
ITED
Barbour, A. G., and D. Fish. 1993. The biological and social
phenomenon of Lyme disease. Science 260:1610–1616.
Chapin, F. S., III, E. S. Zavaleta, V. T. Eviner, R. L. Naylor,
P. M. Vitousek, H. L. Reynolds, D. U. Hooper, S. Lavorel,
O. E. Sala, S. E. Hobbie, M. C. Mack, and S. Diaz. 2000.
Consequences of changing biodiversity. Nature 405:234–
242.
Cottingham, K. L., B. L. Brown, and J. T. Lennon. 2001.
Biodiversity may regulate the temporal variability of eco-
logical systems. Ecology Letters 4:72–85.
Crooks, K. R. 2002. Relative sensitivities of mammalian car-
nivores to habitat fragmentation. Conser vation Biology 16:
488–502.
Donovan, T. M., and C. H. Flather. 2002. Relationships
among North American songbird trends, habitat fragmen-
tation, and landscape occupancy. Ecological Applications
12:364–374.
Fox, B. J., and J. H. Brown. 1993. Assembly rules for func-
tional groups in North American desert rodent communi-
ties. Oikos 67:358–370.
Giardina, A. R., K. A. Schmidt, E. M. Schauber, and R. S.
Ostfeld. 2000. Modeling the role of songbirds and rodents
in the ecology of Lyme disease. Canadian Journal of Zo-
ology 78:2184–2197.
Gilbert, L., R. Norman, K. Laurenson, H. W. Reid, and P. J.
Hudson. 2001. Disease persistence and apparent compe-
tition in a three-host community: an empirical and analyt-
ical study of large-scale, wild populations. Journal of An-
imal Ecology 70:1053–1061.
Knops, J., D. Tilman, N. M. Haddad, S. Naeem, C. E. Mitch-
ell, J. Haarstad, M. E. Ritchie, K. M. Howe, P. B. Reich,
E. Siemann, and J. Groth. 1999. Effects of plant species
richness on invasion dynamics, disease outbreaks, insect
abundances and diversity. Ecology Letters 2:286–293.
Krohne, D. T., and G. A. Hoch. 1999. Demography of
Per-
omyscus leucopus
populations on habitat patches: the role
of dispersal. Canadian Journal of Zoology 77:1247–1253.
Laurance, W. F., H. L. Vasconcelos, andT. E. Lovejoy. 2000.
Forest loss and fragmentation in the Amazon: implications
for wildlife conservation. Oryx 34:39–45.
Lawton, J. H. 2000. Community ecology in a changing world.
Ecology Institute, Oldendorf/Luhe, Germany.
LoGiudice, K., R. S. Ostfeld, K. A. Schmidt, and F. Keesing.
2003. The ecology of infectious disease: effects of host
diversity and community composition on Lyme disease
risk. Proceedings of the National Academy of Sciences
(USA) 100:567–571.
Loreau, M., S. Naeem, P. Inchausti, J. Bengtsson, J.P. Grime,
A. Hector, D. U. Hooper, M. A. Huston, D. Raffaelli, B.
Schmid, D. Tilman, and D. A. Wardle. 2001. Biodiversity
and ecosystem functioning: current knowledge and future
challenges. Science 294:804–808.
Nupp, T. E., and R. K. Swihart. 1996. Effect of forest patch
area on population attributes of white-footed mice (
Pero-
myscus leucopus
) in fragmented landscapes. Canadian Jour-
nal of Zoology 74:467–472.
Nupp, T. E., and R. K. Swihart. 2000. Landscape-level cor-
relates of small-mammal assemblages in forest fragments
of farmland. Journal of Mammalogy 81:512–526.
Ostfeld, R. S., and F. Keesing. 2000
a.
Biodiversity and dis-
ease risk: the case of Lyme disease. Conservation Biology
14:722–728.
Ostfeld, R. S., and F. Keesing. 2000
b.
The function of bio-
diversity in the ecology of vector-borne zoonotic diseases.
Canadian Journal of Zoology 78:2061–2078.
Ostfeld, R. S., F. Keesing, E. M. Schauber, and K. A. Schmidt.
2002. The ecological context of infectious disease: diver-
sity, habitat fragmentation, and Lyme disease risk inNorth
America. Pages 207–219
in
: A. Aguirre, R. S. Ostfeld, C.
A. House, G. Tabor, and M. Pearl, editors. Conservation
medicine: ecological health in practice. Oxford University
Press, New York, New York, USA.
Ostfeld, R. S., E. M. Schauber, C. D. Canham, F. Keesing,
C. G. Jones, and J. O. Wolff. 2001. Effects of acorn pro-
duction and mouse abundance on abundance and
Borrelia
burgdorferi
infection prevalence of nymphal
Ixodes sca-
pularis
ticks. Vector Borne and Zoonotic Diseases 1:55–
64.
Pimm, S. L. 1993. The balance of nature? University ofChi-
cago Press, Chicago, Illinois, USA.
Rosenblatt, D. L., E. J. Heske, S. L. Nelson, D. M. Barber,
M. A. Miller, and B. MacAllister. 1999. Forest fragments
in east-central Illinois: islands or habitat patches for mam-
mals? American Midland Naturalist 141:115–123.
Schmidt, K. A., and R. S. Ostfeld. 2001. Biodiversity and
the dilution effect in disease ecology. Ecology 82:609–619.
Schmidt, K. A., R. S. Ostfeld, and E. M. Schauber. 1999.
Infestation of
Peromyscus leucopus
and
Tamias striatus
by
Ixodes scapularis
(Acari: Ixodidae) in relation to the abun-
dance of hosts and parasites. Journal of Medical Ento-
mology 36:749–757.
Schwartz, M. W., C. A. Brigham, J. D. Hoeksema, K. G.
Lyons, M. H. Mills, and P. J. van Mantgem. 2000. Linking
biodiversity to ecosystem function: implications for con-
servation biology. Oecologia 122:297–305.
Stone, L., R. Dayan, and D. Simberloff. 1996. Community-
wide assembly patterns unmasked: the importance of spe-
cies’ differing geographical ranges. American Naturalist
148:997–1015.
Tilman, D., J. Knops, D. Wedin, P. B. Reich, M. E. Ritchie,
and E. Siemann. 1997. The influence of functional diversity
and composition on ecosystem processes. Science 277:
1300–1302.
Wardle, D. A., K. J. Bonner, G. M. Barker, G. W. Yeates, K.
S. Nicholson, R. D. Bardgett, R. N. Watson, andA. Ghani.
1999. Plant removals in perennial grassland: vegetation
dynamics, decomposers, soil biodiversity, and ecosystem
properties. Ecological Monographs 69:535–568.
APPENDIX
A community-disassembly table for the 13 species of mammals and birds at our field site in Dutchess County, New York,
USA, giving parameters used in models relating vertebrate community composition to nymphal tick infection prevalence
(NIP), is available in ESA’s, Electronic Data Archive,
Ecological Archives
E084-035-A1.