Altered resource availability and the population dynamics of tree species in Amazonian secondary ...

12 Pages·2010·0.34 MB·English

Oecologia(2010)162:923–934
DOI10.1007/s00442-009-1524-5
POPULATION ECOLOGY - ORIGINAL PAPER
Altered resource availability and the population dynamics
of tree species in Amazonian secondary forests
Lucas Berio Fortini • Emilio M. Bruna •
Daniel J. Zarin • Steel S. Vasconcelos •
Izildinha S. Miranda
Received:19March2009/Accepted:18November2009/Publishedonline:9December2009
(cid:1)Springer-Verlag2009
Abstract Despite research demonstrating that water and across years. Furthermore, most of these transient treat-
nutrient availability exert strong effects on multiple eco- ment effects had little effect on population growth rates.
system processes in tropical forests, little is known about Our results suggest that despite major experimental
the effect of these factors on the demography and popu- manipulations of water and nutrient availability—factors
lation dynamics of tropical trees. Over the course of considered critical to the ecology of tropical pioneer tree
5 years,wemonitoredtwocommonAmazoniansecondary species—autogenic light limitation appears to be the pri-
forest species—Lacistema pubescens and Myrcia sylvati- mary regulator of tree demography at early/mid succes-
ca—in dry-season irrigation, litter-removal and control sional stages. Indeed, the effects of light availability may
plots. We then evaluated the effects of altered water and completely override those of other factors thought to
nutrient availability on population demography and influence the successional development of Amazonian
dynamics using matrix models and life table response secondary forests.
experiments. Our results show that despite prolonged
experimental manipulation of water and nutrient avail- Keywords Amazonia (cid:1) Regrowth forests (cid:1) Succession (cid:1)
ability, there were nearly no consistent and unidirectional Water stress (cid:1) Nutrient limitation
treatmenteffectsonthedemographyofeitherspecies.The
patterns and significance of observed treatment effects
werelargelydependentoncross-yearvariabilitynotrelated Introduction
to rainfall patterns, and disappeared once we pooled data
Water and nutrient availability influence multiple ecosys-
tem processes in tropical forests. For instance, limited
CommunicatedbyPeterClarke.
water availability can increase rates of tree mortality
L.B.Fortini(&)(cid:1)D.J.Zarin (Condit 1998; Chazdon et al. 2005; Nepstad et al. 2007),
SchoolofForestResourcesandConservation,University reduce photosynthesis (Nepstad et al. 2002; Fortini et al.
ofFlorida,P.O.Box110760,Gainesville,FL32611-0760,USA 2003; Araga˜o et al. 2005), change leaf and reproductive
e-mail:lfortini@ufl.edu
phenology (Kitajima et al. 1997; Malhi et al. 1998), and
increase the susceptibility of forests to fire (Nepstad et al.
E.M.Bruna
DepartmentofWildlifeEcologyandConservationandCenter 2004). Similarly, observational and manipulative studies
forLatinAmericanStudies,UniversityofFlorida,POBox have shown that nutrient availability alters patterns of tree
110430,Gainesville,FL32611-0430,USA
growth (Davidson et al. 2004), fine litterfall production
(Mirmanto et al. 1999), above-ground biomass (Laurance
S.S.Vasconcelos
Laborato´riodeEcofisiologiaePropagac¸a˜odePlantas,Embrapa et al. 1999), and resource allocation to below-ground
AmazoˆniaOriental,Bele´m,PA66095-100,Brazil structures (Gower 1987; Giardina et al. 2004). Despite
considerable work investigating the individual- or ecosys-
I.S.Miranda
tem-level consequences of nutrient and water availability
UniversidadeFederalRuraldaAmazoˆnia,CP917,
Bele´m,PA66077-530,Brazil fortropicalforests,however,littlehasbeendonetoexplore
123
924 Oecologia(2010)162:923–934
how these factors influence the population dynamics of populations to environmental variability in ways that sim-
tropical trees. Elucidating these relationships is critical, ply comparing the sizes, structures, or growth rates of
since population dynamics link individual responses to populations cannot (e.g., Bruna and Oli 2005).
abioticconstraintswithpatternsofcommunitycomposition In this study we used LTRE to investigate how a long-
and ecosystem processes. term, stand-level experimental manipulation of nutrient
‘‘Secondary’’or‘‘regenerating’’forestsaccountformore availability (via litter removal) and water availability (via
than 40% of tropical forest worldwide (Brown and Lugo irrigation) conducted in an Amazonian secondary forest
1990) and play a critical role in regional C dynamics, the influenced the demography and population growth of two
maintenance of biodiversity, and income generation common tree species. These experimental manipulations
(Chazdon and Coe 1999; Hughes et al. 1999; Aide et al. have been shown at our site and elsewhere to significantly
2000; de Jong et al. 2001; Smith et al. 2003; Gavin 2004; alterecosystem-level(Vasconcelosetal.2004,2007,2008;
Olschewski and Benitez 2005). There are many models of Vasconcelos 2006; Veluci-Marlow 2007) and individual-
succession that attempt to explain how secondary forest levelprocesses(Fortinietal.2003;Araga˜oetal.2005).We
composition and structure change over time (Clements monitoredthedemographyoftwoearlyshort-livedpioneer
1916; Yarranton and Morrison 1974; Pickett 1976; Noble species past their peak abundance and under decline,
and Slatyer 1980). A commonality of these models is that Lacistema pubescens (Lacistemataceae) and Myrcia sylv-
populations of species reach peak abundances at different atica (Myrtaceae), in dry-season irrigation, litter-removal
stages of succession, resulting in easily recognizable spe- and control plots for a 5-year period. We then used these
cies turnover. While light availability is a key driver of demographic data to estimate k using matrix models and
thesesuccessionaldynamics(HustonandSmith1987),tree applied two-way LTREs to evaluate whether changes in
populationsmayalsobeparticularlysusceptibletochanges nutrient and water availability alter the rate of population
in water and nutrient availability. For instance, some sec- decline.
ondary forest species may be very susceptible to drought
because they allocate more fine roots to shallower soil,
while others may face nutrient limitation due to depletion Materials and methods
from past anthropogenic activities (Juo and Manu 1996;
Hopkins et al. 1996; Ayuba et al. 2000; Pavlis and Jenik Site description
2000;McGrathetal.2001).However,itisunclearhowthe
rate and magnitude of changes in abundance during suc- The experimental site is located at the research station of
cession are altered by the availability of water and the Universidade Federal Rural da Amazoˆnia (UFRA),
nutrients. located near the city of Castanhal, Para´, Brazil. Annual
Unfortunately, the high cost and logistical issues asso- precipitationintheregionis2,000–2,500 mm,witharainy
ciated with establishing experiments at suitable spatial season that extends from December to May. Mean daily
scales, along with the need to monitor permanent forest temperatures fluctuate between 24 and 27(cid:2)C. Soils in the
plots for multiple years, have resulted in few studies that siteareclassifiedasdystrophicyellowlatosols,stonyphase
rigorouslytesttheeffectsofwaterandnutrientavailability I (concretionary, lateritic) in the Brazilian soil classifica-
on the population dynamics of trees in tropical secondary tion system, which corresponds to Sombriustox in US soil
forests. Descriptions and comparisons of stands, chrono- taxonomy (Teno´rio et al. 1999). At a 0–10 cm depth, soil
sequences, and observational longitudinal studies provide pHis5.0,totalNis0.15%,andMehlich-1extractablePis
limited inference regarding mechanisms underlying com- 1.58 mg kg-1 (Rangel-Vasconcelos et al. 2005). The
plexforestdynamicsandtheirpotentialabioticconstraints. landscape surrounding the field station is characterized by
However, when long-term data are available, life table secondary forests, annual crops, and active and degraded
response experiments (LTREs) provide a powerful pastures.
approachtodisentanglehowexperimentalfactorsinfluence Our experiment wasinitiatedin 2001in a2-ha patchof
the growth and demography of populations (Burns 2008). secondary forest that had been fallow for 14 years. Prior
These analyses allow for the decomposition of each to being abandoned, the site had undergone multiple
experimental treatment’s effects on the growth rates of cycles of shifting cultivation that began about 60 years
populations (i.e., k); specifically, they measure the contri- ago when the old-growth forest was first cleared. Each
butions to the differences in k between treatments arising cycle included cultivation of corn, manioc, and beans for
from changes in size-class specific vital rates (Caswell 1–2 years, followed by a fallow stage. Interviews with
2001). Because even similar population growth rates can local residents and field station personnel suggest typical
result from very different demographic mechanisms (e.g., shifting cultivation cycles lasted 7–10 years (G. Silva
BrunaandOli2005),LTREscanelucidatetheresponsesof e Souza and O. L. Oliveira, personal communication).
123
Oecologia(2010)162:923–934 925
In 2001 stand basal area was 16.7 m2 ha-1, average stand 3,568 ± 136 g m-2, leading to changes in soil nutrient
density was 208 (±33 SE) individuals[1 cm diameter at statusandreductionsinlitterfallNcontent(Veluci-Marlow
breast height (DBH), and average species richness was 24 2007; Vasconcelos et al. 2008).
(±1 SE) species 100 m-2 (Arau´jo et al. 2005). Nested in the center of each 20 9 20-m plot were
10 9 10-mpermanentinventoryplots(hereafterreferredto
Study species as ‘‘overstory plots’’). Within these plots all tree individ-
uals of the two study species with DBH[1 cm had their
LacistemapubescensMart.andMyrciasylvatica(G.Mey.) height and DBH measured annually between July and
DC. are early short-lived pioneer species prevalent in August. We randomly located four 1 9 1-m quadrats in
secondary forest stands across the eastern Amazon. each plot to monitor individuals C10 cm high and\1 cm
L. pubescens is one of the earliest tree species to establish DBH (hereafter referred to as ‘‘understory plots’’).
in fallows; it typically has a single upright stem and is Although we monitored these quadrats every 2 months
shade intolerant. M. sylvatica generally appears in fallows from2001to2006,forouranalysesweuseddatafromthe
after L. pubescens and other early pioneer species have surveys conducted in July–August of each year to best
become established (Coelho et al. 2004). It is noticeably match inventory dates for the individuals [1 cm DBH.
moreshadetolerantthanitspredecessorsandtypicallyhas Here we report on results from five measurement intervals
a highly branched architecture. L. pubescens and M. sylv- included in the 2001–2006 yearly inventories.
atica are the most abundant overstory species in our
experimental plots accounting for 36 and 22% of all stems Matrix model construction
[1 cm DBH, respectively (Arau´jo et al. 2005).
We observed a strong relationship between height and
Experimental design DBHforbothspecies(r2 = 0.71and0.73forL.pubescens
and M. sylvatica, respectively). However, because of the
The study was conducted in twelve 20 9 20-m plots sep- difficulty in obtaining accurate height measurements for
arated by 10-m buffer strips. We randomly selected four some individuals and the die-off of some shoots, we used
plotsfordry-seasonirrigation,litterremoval,anduntreated DBHmeasurementsasthestatevariableforallindividuals
controls (n = 12 plots total). The irrigation treatment with DBH C1 cm. Height was the only choice for state
consisted of providing the equivalent of 5 mm daily pre- variable of all individuals with DBH\1.0 and height
cipitationduringrainlessdry-seasondayswithanirrigation C10 cm. Following state variable definition, we created a
tape system, corresponding to regional estimates of daily four lifecyclemodel for L.pubescens andafivelife cycle
evapotranspiration (Shuttleworth et al. 1984; Lean et al. model for M. sylvatica. We defined the number of stages
1996; Jipp et al. 1998). Irrigation was initiated in July andtheirsizecut-offsbasedonpatternsofmortalityacross
2001, which corresponded to the start of the dry season. size,limitedreproductivephenologydata,heighttocanopy
Measurements of gravimetric soil moisture content indi- calculations, field observations, and available number of
cated that during the dry season the irrigated plots had individuals.
slightlymorethandoublethemoistureofcontrolplots(22
vs. 10%). Wet-season gravimetric soil moisture content Calculating vital rates
was27%forbothtreatments(Vasconcelosetal.2002).We
established four litter-removal plots in an attempt to break Prior to parameterizing our matrices, we calculated sep-
the tight nutrient cycling of our forest and thus decrease arate estimates of survival, growth, regression and fertility
nutrient availability experienced by plants within these rates for each size class in each species 9 treat-
plots.Whileweacknowledgethatlitter-removaltreatments ment 9 year combination. Since understory individuals
may result in several secondary effects (e.g., leaching, were often difficult to locate in the field, data for a small
changes in soil microbial biomass, soil water content, or number of them were missing in some years. Since the
soil organic matter), we focused our analyses on the pri- exclusion of these individuals from the calculation of
maryeffect ofreducingnutrientavailabilityasmostofthe mortality rates for a given transition year would artifi-
secondaryeffectsalsoexacerbatenutrientlimitation(Sayer cially depress survival, we replaced all single-year gaps in
2006). In August 2001, all leaf and branch litter was data with the average size from adjacent years. However,
removed from the litter-removal plots with plastic rakes; we did exclude the very small number of individuals with
this process was subsequently repeated every 2 weeks. more than 1 year of missing data (n\10) from the cal-
Standinglittercropwaslow,butnotentirelyabsent,inthe culations of vital rates because of the potentially greater
litter-removal plots. Total new non-woody litterfall errors that could result from interpolating data for longer
removed from August 2001 to December 2005 was time periods.
123
926 Oecologia(2010)162:923–934
We calculated survival rates for each matrix using Demographic analyses
logistic models parameterized with all overstory and
understorydata.Thisapproachavoidedthe‘‘over-parsing’’ Using our 30 matrices, we calculated k for each spe-
of ourdata (sensuMorris and Doak 2002) andperfect sur- cies 9 year 9 treatment combination, as well as the sen-
vivalinsizeclasseswithlowsamplesizesandhighsurvival sitivities of population growth toall underlying vital rates.
rates.Aseparatelogisticfitwasusedfortheunderstoryand We also bootstrapped our data sets using the reducibility
overstory data, with resulting size class survival values fix described above to estimate bias-corrected 95% confi-
basedonthelogisticfitevaluatedatmedianindividualsize dence intervals for ks per species,year,and treatment.We
observed within each size class. To avoid poor fits to the used the 30 matrices to perform standard two-way full
data,weusedthestandardcount-basedmethodsofsurvival factorial LTRE analyses to determine the effects of the
calculations whenn\10 for either data set. experimental treatments and study year on each species
We calculated growth and regression transition proba- (Caswell 2001, p. 263). Because our experimental design
bilities following standard count-based methods where the does not consider the interaction between irrigation and
probabilityofgrowth(regression)ofanygivensizeclassis litter removal, irrigation and litter-removal effects were
the proportion of individuals alive at time t that grew evaluated independently through separate LTRE analyses
(regressed) to another size class at time t ? 1. Growth inthefollowingcomparisons:controlversusirrigationand
transition probabilities between understory and overstory controlversuslitterremoval.Foralltests,wecomputedthe
size classes were based on actual observed recruitment to ‘‘reference matrices’’ (sensu Caswell 2001) and related
theoverstorysizeclassesattimet ? 1dividedbythearea- vital rates as the average from control plot data for all
scaled number of understory size class individuals present measurement intervals. For each factor level for each
at time t. Seedling recruitment rates were calculated based LTRE, the midpoint matrix used to evaluate the contribu-
on the observed number of new individuals from one tionofobservedvitalratetokwascomputedastheaverage
measurementtothenext.Lastly,weusedCaswell’s(2001) matrix between the reference matrix and related factor
approach to calculate anonymous fertility rates with equal matrix. Lastly, we examined the significance of treatment
reproductive weights among overstory size classes. effects statistically by bootstrap-derived confidence inter-
valsofthedifferencebetweentreatmentandreferencevital
Analyzing differences among vital rates rate estimates.
To evaluate variability in estimates of vital rate, we per- Precipitation analysis
formed a bootstrapping procedure on the overstory and
understory data sets (1,000 runs). We used the resulting To determine whether precipitation patterns influenced
bootstrapped estimates to create 95% confidence intervals species demography or treatment effects, we performed a
foreachvitalrateforeachspecies,sizeclass,treatmentand multiple correlation analysis between precipitation and
year. demographic analysis results. Precipitation variables
includeddry-seasonrainfallamount,maximumandmedian
Creating population matrices rainless interval durations from current and previous year
to account for potential time-lagged responses (Table 1;
We created a total of 30 matrix models (2 species 9 3 Vasconcelos 2006). Our demographic variables included
treatments 9 5 time intervals) based on lower level vital demographic rates by species and size class from control
rates (Morris and Doak 2002). To avoid problems associ- plants to evaluate precipitation effect on species demog-
ated with reducible matrices (Caswell 2001), we substi- raphy and the difference between treatment and control
tuted the rare zero values for growth and survival demographic rates to evaluate potential links between
probabilities with the lowest bootstrapped non-zero value yearly rainfall patterns and treatment response. Results
of the vital rate for the respective treatment among all from the multiple comparisons where evaluated using
measurement years. Fecundity values of zero also yield standard and Bonferroni-corrected a-values.
reducible matrices. Therefore, if fecundity for any matrix
was zero we only substituted fecundity from the largest
size class with the lowest bootstrapped non-zero fecundity Results
value for the respective treatment among all measurement
years. Preliminary comparisons showed that while this ad Species demographic trends
hoc reducibility fix improved the results of subsequent
analysis, it had only minor effects on estimates of popu- Lacistema pubescens andMyrcia sylvatica had contrasting
lation vital rates and population growth. patterns of survivorship for the smallest size classes, but
123
Oecologia(2010)162:923–934 927
Table1 Rainfallpatternsfor
Year Season Total Medianrainless Maximumrainless
UniversidadeFederalRuralda
rainfall(mm) periodlength(days) periodlength(days)
Amazoˆniaresearchstation,
Castanhal,Para´,Brazil
1999 Dry 352 3 10
2000 Wet 1,922.9 1 9
2000 Dry 223.5 4.5 11
2001 Wet 3211 2 7
2001 Dry 438.9 5 32
2002 Wet 1,813.2 2 7
2002 Dry 679.4 5 16
2003 Wet 1,718.6 1 11
2003 Dry 696.8 5 21
2004 Wet 2,881.5 2 11
2004 Dry 444.9 4 25
2005 Wet 2,177.2 2 11
2005 Dry 225.2 9 19
2006 Wet 2,234.9 2 4
similarly high rates of survivorship for overstory size Bothspeciesshowedcontrastingpatternsofrecruitment
classes. L. pubescens survival in the smallest two size acrossyears.WhileL.pubescensshowedconstantandvery
classes was low for the entire study period, typically rang- low yearly recruitment below 20 individuals 100 m-2,
ingfrom 0.4to0.8 (Fig. 1a). Survival inthese sizeclasses M. sylvatica presented recruitment above 400 individuals
only peakedabove theselow values duringthe 2002–2003 100 m-2 in the first time interval (2001–2002). However,
time interval when both dry seasons showed total rainfall M. sylvatica recruitment dropped across all years to 25
amountsconsiderablygreaterthanusual.Ontheotherhand, individuals 100 m-2 by the 2005–2006 time interval
nearlyallofthesurvivalratesobservedforM.sylvaticaare (Table 2).
above 0.8 and present no large differences among years or
size classes (Fig. 1b). The only minor temporal trend Population growth patterns across time and treatments
observed for M. sylvatica suggests that the survival of the
largest size class is the moststable across allyears. We found that with one exception—the irrigated L. pu-
For both species, the probability of growth from bescenspopulationin2003–2004—therewasnosignificant
understorytooverstorysizeclasseswasextremelylowand differenceinkamongpopulationsofthetwostudyspecies
there were no clear differences in growth transition prob- indifferenttreatments(Fig. 2).Furthermore,wefoundthat
abilities among overstory size classes (Fig. 1c, d). Allo- only the control M. sylvatica population in the 2001–2002
metric height/DBH equations reveal that neither species transition year included positive, albeit non-significant,
had understory individuals with heights close to the populationgrowth(i.e.,k[1).Thepopulationgrowthrate
equivalent to the minimum overstory size cut off (1 cm for L. pubescens ranged from 0.8909 (population in the
DBH), lending further evidence to a clear overstory irrigation treatment in the 2005–2006 period) to 0.9913
recruitment limitation for both species. For M. sylvatica (population in the control treatment in the 2003–2004
individuals in the understory, large growth transition period). Results for M. sylvatica were similar, with k
probabilities from size class 1 and regression probabilities ranging from 0.8906 to 1.006 and little difference among
for size class 2 reflect the large variability of measured treatments and years.
heights that results from the difficulties in evaluating
growth based on height (e.g., shoot die-backs, new leaf Two-way LTRE results
flushes, etc.; Fig. 1d, f). However, these rates are excep-
tionally high for the 2002–2004 period, potentially related For both species, treatment differences in vital rates were
tothehigh dryseasonrainfallyearsof2002and2003.All small and mostly fell within 0.05 of control values. Fur-
other regression probabilities for both species do not show thermore, LTRE analysis showed that because of the dif-
any clear temporal or size-related patterns, although ferential impacts of vital rates on population growth,
regression probabilities for overstory size classes of treatment effects on growth and mortality often did not
M.sylvaticawerenoticeablyhigherandmorevariablethan resultincommensuratepopulationgrowtheffectsforeither
that of L. pubescens (Fig. 1e, f). of the two studied species (Fig. 3). For both species, the
123
928 Oecologia(2010)162:923–934
Fig.1 Sizeclass(Sz)annual
ratesofsurvival,growthand
regressionwith0.975and0.025
quantilesofLacistema
pubescens(a,c,e)andMyrcia
sylvatica(b,d,f).HgtHeight,
DBHdiameteratbreastheight
Table2 Offspringper100m2forLacistemapubescensandMyrcia effects, only the irrigation growth of M. sylvatica’s seed-
sylvaticaacross timeintervals.Givennotreatmentdifferences, only lingsizeclass 1showedaconstantunidirectional response
controldataareshown over the course of the experiment.
Yearinterval L.pubescens M.sylvatica While vital rates varied considerably from one time
intervaltothe next, wefound that—forboth species—pat-
2001–2002 6.25 406.25
ternsof‘‘time’’effectsonvitalrateswereidiosyncratic.The
2002–2003 0 181.25
onlyconsistenttemporalpatternobservedwasadecreasein
2003–2004 12.5 143.75
the survival of all size classes of M. sylvatica during the
2004–2005 18.75 100
2003–2004measurementinterval.Surprisingly,inter-annual
2005–2006 6.25 25
rainfall patterns were not significantly correlated to size-
classspecificspeciesandtreatmentvariabilityofvitalrates.
survival and regression probabilities make the largest
contributions to Dk across time and treatments, with tran-
sition probabilities describing growth having a relatively Discussion
smallerimpactonpopulationgrowth.Inaddition,posthoc
bootstrapping of differences in vital rates between the To our knowledge, our study is the first large-scale and
experimental and control treatments indicate that no long-termexperimentalinvestigationofhowirrigationand
treatment differences in vital rates were significantly dif- litterremovalinfluencethepopulationdynamicsoftropical
ferent when averaging data across years. However, the trees. We found that although previous work has docu-
evaluation of the interaction effect between treatment and mented large ecosystem-level (Vasconcelos et al. 2004,
time suggests significant treatment impacts for specific 2007, 2008; Vasconcelos 2006; Veluci-Marlow 2007) and
time intervals (Fig. 4). Of these time-dependent treatment individual-level effects (Fortini et al. 2003; Araga˜o et al.
123
Oecologia(2010)162:923–934 929
Our results suggest that the ability to intuit the conse-
quences of environmental change for population dynamics
on the basis of short-term experiments is limited. For
instance, the high unexplained temporal variability in
demographic rates resulted in time-dependent treatment
differences;wereitnotforourlong-termanddemographic
approach we could have easily misinterpreted these dif-
ferences as persistent and biologically meaningful. Fur-
thermore, the potential disconnect between changes in
individual vital rates and actual changes in population
dynamics is a cautionary tale for experimental studies that
either utilize only a subset of life stages or analyze
demographic rate responses separately (see also Halpern
andUnderwood2006).Ourresultsalsohighlighttheutility
of matrix models, whose integrative nature reveals long-
term patterns that would not have been detected by com-
parison of changes in population size and the distribution
or a limited subset of life history stages (Caswell 2007).
These models allow us to understand the relative impor-
tance to population dynamics of different vital rates; in
doingsotheylinkobservedshiftsinpopulationdistribution
and abundance to underlying demographic mechanisms.
Furthermore,estimatesofkcanbeinterpretedasameasure
of relative fitness (McGraw and Caswell 1996), allowing
onetobetterunderstandsuccessionaltrendsresultingfrom
altered competitive balance between species.
Fig.2 Populationgrowthrates(k)ofaL.pubescensandbM.sylv- Previousresearchintropicalforestshasshownthatthere
atica in stands with water addition and litter removal. Values are can be major physiological responses to water availability
mean±95%confidenceinterval (Engelbrecht and Kursar 2003; Bunker and Carson 2005;
Nepstadetal.2007;Tanner andBarberis 2007;Yavittand
2005) of altering dry season water availability and litter Wright2008).However,ourunderstandingofhowdrought
abundance in our experimental plots, the patterns and impacts tropical tree populations is rudimentary. Nepstad
significance of these manipulations on demography over et al. (2007) argue that the effects could be large, in part
our 5-year experiment were generally negligible. Further- because large trees are at greater risk due to greater
more, the transient effects of treatment on demography evaporative demand of canopy exposure. This idea is
were neither related to inter-annual rainfall patterns nor corroborated by their dry season rainfall-exclusion studies
drove significant changes in population growth. conducted in Amazonia, which resulted in substantial
Fig.3 Impactsofirrigationand
litter-removaltreatmentson
survival,growthandregression
probabilitiesonpopulation
growthofaL.pubescensand
bM.sylvatica
123
930 Oecologia(2010)162:923–934
direct sunlight, resulting in increased evaporative demand
under drought conditions. They also grew in shallow lat-
eritic soil formations that appear to limit the development
ofdeeprootsystems.Nevertheless,theseconditionsdidnot
translate into reduced growth and survivorship. On the
other hand, our study shows only limited evidence sup-
porting the notion that seedlings are highly drought sensi-
tive.Only the growth ofthe smallest M.sylvaticaseedling
size class showed a constant increase in response to irri-
gation, which probably resulted from large changes in C
assimilation(Fortinietal.2003).However,thiseffectonly
lastedfor3ofthe5 yearsofmonitoringandwasshownto
beinconsequentialintermsofcontributionstoM.sylvatica
population growth.
Large-scale, long-term natural litter-removal experi-
ments designed to explore the role of nutrient constraints
are rare in tropical ecosystems (Sayer 2006). Other
research at our study site has found relatively rapid
responses of nutrient availability, possibly reflecting the
quickandtightcyclingofnutrientscommoninthetropical
forests (Vitousek and Sanford 1986; Veluci-Marlow 2007;
Vasconcelos et al. 2008). Our analyses suggest secondary
forest tree populations are very tolerant to these changes.
Although this tolerance is supported by fertilization
experiments demonstrating limited effects of reduced
nutrient availability on tropical forest vegetation (Mayor
and Roda 1994; Pearson and Vitousek 2001), our results
areneverthelessatoddswithseveralotherstudiesthathave
shown clear increases in growth and survival following
nutrient addition (Gehring et al. 1999; Lawrence 2001;
Ceccon et al. 2003; Davidson et al. 2004; Yavitt and
Wright2008).Indeed,ourresultsareparticularlysurprising
Fig.4 Statisticallysignificanttime9treatmenteffectsonvitalrates giventherelativelylowvaluesofsoilextractablePpresent
of a L. pubescens, b M. sylvatica. Up arrow Increase in vital rate, at our site (Rangel-Vasconcelos et al. 2005).
downarrowdecreaseinvitalrate.sSurvival;ggrowth;rregression
(vital rate numbers indicate size class, e.g., g3 growth rate of size
class 3); I significant irrigation effect; R significant litter-removal Exploringthelimitedeffectsofresourcelimitationonk
effect
Ourresultsbegetthequestionofwhytherewerenostrong
mortality of large trees and canopy-bound lianas (Nepstad andunidirectionaleffectsondemographyandkinourtwo
et al. 2007). On the other hand, irrigation experiments in focal species. First, it is possible that despite our large
Panama and elsewhere have suggested it is actually the experimental manipulations of water and nutrients, the
seedlingstagesoftropicaltreesthataremostvulnerableto treatmentsdidnotresultinsignificantchangesinwaterand
droughtdue totheir limited root systems (Engelbrecht and nutrient availability. However, recent studies strongly
Kursar2003;YavittandWright2008).Ifso,thelong-term suggest that this is not the case. Past studies from our site
effectsofoccasionaldroughtsonpopulationdynamicsmay have linked changes in water availability to several eco-
be limited, owing to the generally low elasticity values of system and individual-level responses, including increased
seedling-related vital rates (e.g., Bruna 2003). above-ground net primary productivity (ANPP; Fortini
Ourexperimentalanddemographicdesignallowedusto et al. 2003; Vasconcelos et al. 2004, 2007; Vasconcelos
explore the responses to drought of stages spanning the 2006; Veluci-Marlow 2007). Veluci-Marlow (2007) has
seedling-adult continuum. Our results suggest the impacts alsodemonstratedthatourlitter-removaltreatmentresulted
of drought on tree demography are far more complex than in decreased N mineralization, phosphatase activity, and
simply influencing primarily seedlings or adults. On the NH ? availability, as well as increased availability of
4
one hand, many of our larger adult trees were exposed to NO -. Furthermore, Vasconcelos et al. (2008) found a
3
123
Oecologia(2010)162:923–934 931
significant decrease in litterfall N content, indicative of
responses to nutrient limitation by individual trees at our
litter-removalplots.Althoughtherearenoresultsfromthis
system quantifying the increased magnitude of nutrient
limitation resulting from litter removal, the fact that the
removal of 4–5 times the initial P stocks in above-ground
finelitterovera5-yearperioddidnotappreciablyinfluence
the demography of our study species at such a nutrient-
poor site is surprising and merits further investigation.
Second, it is important to consider the possibility that
species with different life history strategies could be more
susceptible to changes in water and nutrient availability
than our focal taxa (McCook 1994). Past research con-
ducted at our study site suggest that other species may be
more responsive to changes in water availability than our
twofocalspecies.Astudyofchangesinwhole-standANPP
in response to water availability found that whole-stand
diameter increment was linked to changes in water avail-
ability resulting from irrigation and previous dry sea- Fig.5 ChangesinthediameterdistributionofaL.pubescens,bM.
son rainfall (Vasconcelos 2006). Although the growth of sylvaticaandcallotherspeciesfromcontrolplotsatthe14-year-old
L.pubescens individualsinsizeclass 3tracked thechange secondary forest in Apeu´, Para´, Brazil. Irrigation and litter-removal
plotsexhibitedsimilardistributionalshifts
in ANPP for the whole stand, no other vital did. Further-
more,shiftsinthesizedistributionofthetwofocalspecies
resulted in a drastic decrease in their population numbers, personal observation), seedling recruitment by the end of
and suggest the dynamics of these two species may be the study period was appreciably lower; this suggests that
differentfromothersinthestand(Fig. 5).Nevertheless,itis the possibility of a delayed recruitment response to either
important to recognize that L. pubescens and M. sylvatica treatmentisunlikely.Thisdynamism,andthemyriadother
represent very distinct species in the community of suc- factorsthatalsoinfluencedemography,workinacomplex
cessionaltreesfoundinourstudysite,andarethedominant fashion to influence a species’ interactions with competi-
successional species in the region’s secondary forests tors; this competitive balance in turn alters population and
(L.Fortini,unpublisheddata).Hence,theirresponsetoour stand structure, leading to feedback changes in light
experimental manipulations—or lack thereof—will have availability and other abiotic constraints.
important implications for the trajectory ofsuccession. While understory light availability did not vary with
Third,itisimportanttorememberthatthedemographic respect to experimental treatment (L. Fortini, unpublished
responsesofaspeciestochangesinabioticconstraintswill data),thereisevidencesuggestinglightplaysanimportant
be influenced by how other species—which can be either role in the demography of our focal species as seen else-
competitors or facilitators—respond to the same changes where (Kobe et al. 1995; Graham et al. 2003). Although
(Goldberg 1996; Choler et al. 2001; Liancourt et al. 2005; neither study species was suffering from high overstory
Suter et al. 2007). In secondary forests this is further mortality or low reproductive output (L. Fortini, unpub-
complicated by changes in the strength and direction of lisheddata),bothspeciesshowedclearbottlenecksintheir
competitive interactions over the course of a stand’s suc- understorysizeclasses.AlmostnoL.pubescensindividuals
cessional trajectory (McCook 1994; Kobe et al. 1995; Mal recruited in spite of ample fruiting, including in litter-
et al. 1997). The relative importance of different limiting removal plots where the elimination of the litter layer
factors can also change as succession proceeds, further should have resulted in increased understory recruitment
shifting thecompetitive balance between species.Assuch, (Facelli and Pickett 1991). Furthermore, there was also an
theresponseoffocalspeciestothesamechangesinabiotic understory bottleneck in M. sylvatica—despite ample
constraints may have differed if we had conducted these understory recruitment, nearly no individuals survived to
experiments at earlier or later successional stages. At our reach the smaller reproductive size classes. Given the
study stand, the shifts in diameter distribution for all spe- previously documented impacts of water and nutrient
ciespresentinthestandduringthestudyperiodindicatean availability on stand- and individual-level processes at our
ongoing dynamic development of the forest (Arau´jo et al. study site and elsewhere, we expected controls over forest
2005). Further support for this conclusion is that despite succession were potentially more complex than overriding
continued fruiting by the two study species (L. Fortini, autogenic control through light availability (Huston and
123
932 Oecologia(2010)162:923–934
Smith1987).Ourresultssuggestthat,forsomespeciesand forest regrowth is flourishing (Vester and Cleef 1998;
at certain successional stages, light availability may still Coomesetal.2000;Johnsonetal.2000;Moranetal.2000;
overshadow prolonged alterations of other ecosystem deJonget al.2001;GuariguataandOstertag2001;Capers
properties posited critical for forest function. et al. 2005; Vasconcelos 2006). While our results suggest
Our results suggest a set of research directions and the population growth rates of secondary forest tree
related experiments that merit further exploration. First, it species are resilient to altered precipitation regimes or
is necessary to explore the role of autogenic successional land-use-induced nutrient limitation, the complex mecha-
factors (e.g., light availability, root competition) in medi- nisms underlying this response clearly indicate a need for
atingpopulationresponsetoexogenousabioticconstraints. workinothersystemsandwithabroaddiversityofspecies.
This could be done by monitoring of seedlings to adults If climate change results in increases in precipitation for
fromanumberofspeciesofcontrastinglifehistorytraitsin the region (Marengo 2004), our results suggest that the
similar manipulative experiments, therefore ensuring a demographyandpopulationdynamicsofsomespecieswill
better representation of species under varied strengths of beminimallyaffected byincreasesinthe frequencyofdry
autogenic control from stand structure. Alternatively, one seasonrainfall.However,ifclimatechangeresultsinmore
could explore the effects of experimental manipulations of intense dry seasons (Harris et al. 2008), rainfall-exclusion
abiotic factors along a chronosequence representing dif- experimentsaremorelikelytotakeplantsintowaterstress
ferent stages of succession. Second, applied treatments levels beyond those commonly experienced in typical dry
shouldideallyincludemultidirectionalmanipulations(e.g., seasons and thus may be valuable complements to our
irrigationandexclusion)toproperlyevaluatelimitationsto present study (Nepstad et al. 2007). In effect, while
typical and atypical stress conditions. Finally, such exper- addressing the consequences of climate change on
iments should monitor how treatments impact competitive tree populations in the Amazon, both irrigation and rain-
ability/relative fitness of individuals over time to begin fall-exclusion experiments only offer partial answers.
understanding the potentially complex response of popu- Similarly, nutrient limitation comparable to that in our
lations in a diverse community setting. Such experimental experimental manipulations would also be predicted to
approaches, coupled with recent theoretical advances have limited effects on demography.
(Caswell 2007), could help investigate the potentially
important role of transient population dynamics. Acknowledgments WethankOsorioOliveira,GlebsonSousa,and
Evandro da Silva for their assistance in the field, and Raimundo
Lastly, it is also worth noting how the contrasting
Nonato da Silva (UFRA) and De´bora Araga˜o for logistical support.
demographic patterns for the two focal species clearly
Thisresearchwasconductedundercooperative agreementsbetween
reflect their distinct successional roles. L. pubescens the University of Florida, Universidade Federal Rural da Amazonia
establishes stand dominance early on, but soon declines in andEmbrapa Amazoˆnia Orientaland supportedby agrant from the
Andrew Mellon Foundation to D.J.Z., a grant from Conselho
abundance. We found no consistent temporal patterns in
Nacional de Desenvolvimento Cient´ıfico e Tecnolo´gico—CNPQ to
L.pubescensdemographicrates,suggestingthepatternsof
I.S.M. and an EPA STAR fellowship to L.B.F. All conducted
demography we observed reflect ecological influences experimentscomplywithcurrentBrazilianlaws.
that preceded our monitoring of the stand. In contrast,
M. sylvatica’sinverse-J population structure, high survival
References
of smaller size classes, and initially high understory
recruitment are consistent with the observation that its
AideTM,ZimmermanJK,PascarellaJB,RiveraL,Marcano-VegaH
dominance in the successional community lags behind
(2000) Forest regeneration in a chronosequence of tropical
that of L. pubescens. Nevertheless, the near absence of
abandonedpastures:implicationsforrestorationecology.Restor
recruitment into M. sylvatica’s understory stages and the Ecol8:328–338
smallest overstory size class implies the long-term persis- Araga˜oDV,FortiniLB,MulkeyS,ZarinDJ,AraujoMM,Carvalho
CJR (2005) Correlation but no causation between leaf nitrogen
tence of M. sylvatica is unlikely. If so, this gives rise to
andmaximumassimilation:theroleofdroughtandreproduction
increased dominance of other species in the successional
in gas exchange in anunderstory tropical plantMiconia ciliata
progression of these stands, although it does not exclude (Melastomataceae).AmJBot92:456–461
the possibility of M. sylvatica’s return from the seed bank Arau´jo MM, Tucker JM, Vasconcelos SS, Zarin DJ, Oliveira W,
Sampaio PD, Rangel-Vasconcelos LGT, Oliveira FDA, Coelho
orfromyoungerneighboringstandsfollowingdisturbance.
RFR, Araga˘o DV, Miranda I (2005) Padra˘o e processo suces-
sionaisemflorestassecunda´riasdediferentesidadesnaAmazoˆ-
Conclusion niaoriental.CieˆnciaFlorestal15:343–357
AyubaHK,AwetoAO,AbubakarSM(2000)Soilnutrientdynamics
under small-holder agricultural practices in Konduga, north-
With the growing interest in the mitigation of anthropo-
easternNigeria.TropAgric77:116–118
genic disturbance and changes in global C dynamics,
Brown S, Lugo AE (1990) Tropical secondary forests. J Trop Ecol
research on processes and factors constraining tropical 6:1–32
123