Survival, Growth, and Ecosystem Dynamics of Displaced ... · viduals growing on palms and within 2 m of the ground were sampled. Transplanted bromeliads were carefully removed from

211

BIOTROPICA 34(2): 211–224 2002

Survival, Growth, and Ecosystem Dynamics of DisplacedBromeliads in a Montane Tropical Forest1

Jennifer Pett-Ridge and Whendee L. SilverDepartment of Environmental Science, Policy, and Management, 151 Hilgard Hall Room 3110, University ofCalifornia, Berkeley, California 94720, U.S.A.

ABSTRACTEpiphytes generally occupy arboreal perches, which are inherently unstable environments due to periodic windstorms,branch falls, and treefalls. During high wind events, arboreal bromeliads are often knocked from the canopy anddeposited on the forest floor. In this study, we used a common epiphytic tank bromeliad, Guzmania berteroniana (R.& S.) Mez, to determine if fallen bromeliads can survive, grow, and reproduce on the forest floor and evaluate thepotential impact of adult dispersal on plant and soil nutrient pools. Bromeliads were transplanted to and from treestems and the forest floor and monitored intensively for six months; survival, growth, and impacts on ecosystemnutrient pools were followed on a subset of plants for 16 months. Six months after transplanting, bromeliad mortalitywas low (3%), and 19 percent of study individuals had flowered and produced new juvenile shoots. Mortality on thesubset of plants followed for 16 months was 14–30 percent. Although survival rates were relatively high in all habitats,bromeliads transplanted to trees grew significantly more root length (x̄ 6 SE: 189 6 43 cm) than those moved tothe forest floor (53 6 15 cm) and experienced lower rates of leaf area loss. All transplanted bromeliads rapidly alteredthe substrate they occupied. Individuals transplanted to and among trees rapidly decreased base cation concentrationsbut significantly increased P concentrations of their underlying substrate. On the ground, bromeliads increased C, N,and P concentrations within nine months of placement. Our results suggest that in this montane tropical forest,bromeliads respond rapidly to displacement, locally modify their substrates, and can access the resources needed forsurvival regardless of habitat.

Key words: epiphytes; Guzmania berteroniana; hurricanes; nutrient cycling; Puerto Rico; tropical montane forest.

EPIPHYTES ARE A CONSPICUOUS COMPONENT of manytropical forests and have been shown to play animportant role in ecosystem-level processes (Matel-son et al. 1993). Most epiphytes occupy arborealperches where they are subjected to differentedaphic and microclimatic conditions than terres-trially rooted plants. The arboreal habitat isthought to supply epiphytes with higher inputs ofatmospheric resources and light than the soil sur-face, but it also has lower storage capacity for waterand nutrients, higher winds, and less stability. Epi-phytes are particularly abundant in montane trop-ical forests, which can have high atmospheric in-puts from rain and clouds but are also often high-wind environments (Asbury & McDowell 1994).In forests exposed to frequent high winds and hur-ricanes, epiphytes can fall from the canopy to theforest floor, either individually or with branch- andtreefalls (Migenis & Ackerman 1993). These fallenepiphytes may not die immediately. Matelson et al.(1993) reported high survival rates for fallen epi-phytic mats associated with branch falls in a cloudforest of Costa Rica. The consequences of displace-ment from the canopy for growth, reproduction,

1 Received 30 August 2000; revision accepted 6 Novem-ber 2001.

and ecosystem nutrient pools have not been de-scribed.

Epiphytes contribute to ecosystem-level nutri-ent cycling by capturing atmospherically derivednutrients that constitute a net gain for terrestrialnutrient pools (Benzing 1990, Matelson et al.1993). This is particularly true for bromeliads,which concentrate airborne nutrients via uptakethrough specialized trichrome cells located in leafsubaxils. Tank bromeliads can also access nutrientsthrough a variety of symbiotic relationships withother organisms, some of which have been shownto fix atmospheric N (Bermudes & Benzing 1991).Soils that develop immediately beneath epiphytesmay have high nutrient concentrations due to thecombined nutrient inputs from decomposing epi-phytic litter, impounded organic matter, and leak-age of nutrient-rich waters (Paoletti et al. 1991).Most researchers agree, however, that bromeliadsrarely access these soil resources via root uptake andinstead have adapted to use leaves as their solemeans of carbon, moisture, and ion gain (Benzing1980).

The epiphytic community of the upper mon-tane forests of Puerto Rico has developed under theinfluence of frequent high winds (Weaver 1994)and exposure to periodic hurricanes (Walker et al.

212 Pett-Ridge and Silver

1991, 1996, 1999). These wind events often resultin the displacement of arboreal bromeliads fromcanopy perches to the forest floor. In theory, dis-placed epiphytes should have the ability to survive,grow, and reproduce in new habitats as long as leaftissue, the primary tissue for resource acquisition,remains undamaged and sufficient light resourcesremain available. We tested this theory by deter-mining the effects of relocation on a common bro-meliad Guzmania berteroniana (R. & S.) Mez. Wealso hypothesized that bromeliad growth on theforest floor would have a differential effect on planttissue chemistry (reflecting a change in resourceavailability or uptake strategy) and/or substrate nu-trients (through leaf senescence and leaching). Totest these hypotheses, we used a reciprocal trans-plant experiment to address the following ques-tions: (1) What are the effects of physical relocationon bromeliad growth and survival? (2) How doesa change in substrate/habitat affect bromeliad tissuechemistry? and (3) How do bromeliads affect nu-trient cycling in the substrate they occupy?

MATERIALS AND METHODS

The study was conducted in the palm forest-typeof the Luquillo Experimental Forest (LEF), PuertoRico (188109N, 658509W), as part of the NSFsponsored Long Term Ecological Research Pro-gram. This forest type is dominated by a singlespecies of palm, Prestoea montana (R. Grah) Nich-ols, which grows to an average canopy height of 15m (Weaver 1994) and occurs between 700 and 900m elevation (Beard 1949, Brown et al. 1983). Theaverage annual rainfall is 4450 mm, mean annualtemperature is 18.78C, and average relative humid-ity is 98 percent, with little or no seasonality(Weaver 1994). Tropical storms occur in this re-gional annually, and the hurricane return intervalis approximately every 60 years.

We stratified our sampling on leeward andwindward slopes where the tank bromeliad G. ber-teroniana is the dominant bromeliad species andabundant both above ground and at ground level.On palm stems, bromeliad density ranges from1600 individuals/ha in the canopy, 3700 individ-uals/ha at the stem midpoint, to 800 individuals/ha at the stem base (Silver et al., pers. obs.). In themontane forests of the LEF, ground-dwelling G.berteroniana occur in densities of up to 700 adultindividuals/ha and are typically root-anchored inorganic and/or mineral soil horizons. We refer tothese ground-dwelling bromeliads as ‘‘terrestrial’’throughout this paper in reference to the physical

habitat these plants occupy, and the term does notimply a change in life history strategy. Adult repro-ductive G. berteroniana range in size from 40 to120 cm across the longest leaf axis. Like most bro-meliads, this species utilizes a dual-reproductionstrategy: vegetative axillary offshoots are producedas well as brightly colored inflorescences (Fig. 1),which bear seeds that appear to be wind-dispersed.

RECIPROCAL TRANSPLANT EXPERIMENT. Forty arbo-real and 60 fallen individuals were randomly se-lected within two 450 m2 plots, one leeward andone windward facing. Effort was taken to establishthese plots under a continuous canopy of P. mon-tana, without large gaps. Since no differences werefound by aspect in any of the demographic orchemical variables we measured, data from leewardand windward sites were pooled by treatment. Eachbromeliad was assigned to one of nine treatments,including six control categories (Table 1). Weavoided using very small epiphytes (,20 cm diam.)in an attempt to eliminate immature plants fromthe data set. For aboveground samples, only indi-viduals growing on palms and within 2 m of theground were sampled. Transplanted bromeliadswere carefully removed from tree stems or forestfloor with tweezers, a flathead screwdriver, and/ora trowel to minimize damage to roots and the re-moval site. Transplanted bromeliads were relocatedwithin plots using a stratified random design. Tocontrol for the effect of severing connections of thebromeliad to its substrate, 20 plants (10 on treesand 10 on the forest floor) were replaced in theiroriginal position after connections were severed. Anadditional 20 control plants (10 on trees and 10on the forest floor) were not manipulated. Wefound no differences between the two types of con-trols, and so they were pooled for analyses by hab-itat location. All bromeliads transplanted to theforest floor were supported by small PVC stakesand a single nylon thread to keep them from tip-ping (Fig. 1). On trees, bromeliads were supportedby a 1 cm wide, 10 cm diameter PVC ring thatsupported the base of the plant and was attachedto the tree with a bungee cord (Fig. 1). As plantswere reattached, a square piece of thin plastic meshwas placed between the substrate and the trans-planted bromeliad base to facilitate measurementsof root growth. We also located plants along road-sides, where bromeliads are rare or absent, in aneffort to determine the effects of canopy removal.These roadsides were covered with grasses and lowsecond-growth vegetation and characterized ashigh-light environments. Photosynthetically active

Substrate and Habitat Effects on Bromeliads 213

FIGURE 1. Guzmania berteroniana specimen experimentally transplanted from the forest floor to a tree stem in anupper montane rain forest, Luquillo Experimental Forest, Puerto Rico. The PVC ring that supports this bromeliad isattached to the stem by an elastic cord. Between the bromeliad and the tree stem, blue plastic mesh has been insertedso that new root growth (extending through the mesh) can be measured.

TABLE 1. Experimental and control treatments employed with Guzmania berteroniana specimens in the palm forest-typeof an upper montane rain forest in the Luquillo Experimental Forest, Puerto Rico.

Category Label N Description

FF controlFF (in place)FF → FFFF → RoadFF → TreeTree controlTree (in place)Tree → TreeTree → FF

FCFFFFRFTTCTTTTF

101010201010101010

forest floor controluprooted from forest floor and replaced in same location (control for uprooting)transplanted from one forest floor location to anothertransplanted from the forest floor to open roadsidetransplanted from the forest floor to tree stemtree stem controlremoved from tree stem and replaced in same stem location (control for removal)transplanted from one tree stem location to anothertransplanted from tree stem to forest floor

radiation (PAR) (mml/m2/h) in nearby forest androadside plots was measured continuously for oneweek during our study period at 0.05 and 1 mheights. PAR measurements within the forestranged from 1.3 to 4.7 mml/m2/h at 0.05 m offthe ground to between 2.9 and 9.3 mml/m2/h at1 m off the ground, while roadside plots averaged34.8 and 33.0 mml/m2/h for the two heights, re-spectively (Olander et al. 1998).

Treatment and control bromeliads were sur-

veyed every 2 weeks for 6 months to measure in-dices of health, growth, and mortality. Measure-ments included horizontal plant length and width(longest x- and y-axis); number of live leaves; num-ber of attached dead leaves; leaf damage by eitherherbivory or sun scorch; reproduction (floweringand number of juvenile shoots); vine, litter, or liveleaf contact; and mortality. Under the canopy, her-bivory caused the majority of leaf tissue damage,whereas in open roadsides we observed leaves that


had dried and shriveled due to higher light inten-sity and temperature. We continued less intensivemonitoring on a subset of plants for flowering andreproduction until month 9 (N 5 73) but followedmortality and growth for the entire 16-month pe-riod (N 5 46).

At 6, 9, and 16 months from the start of theexperiment, we randomly selected three individualsper treatment to be harvested. All plants were re-measured just prior to harvesting. The harvestedplants were separated into root and shoot tissues,washed individually to remove adhering soil andorganic litter, dried at 658C, weighed for biomass,and ground for nutrient analyses. For transplantedindividuals, new root length was measured for rootsthat had grown through the mesh squares. For con-trols, total root length was measured. We used thefollowing equation to estimate relative foliargrowth rates assuming a constant leaf width: Rel-ative foliar growth 5 [LL * Lnl * (1 2 Ld)] Tf /[LL * Lnl * (1 2 Ld)] To, where LL is the averageleaf length calculated as half the average plant di-ameter (measured from leaf tip to leaf tip on bothx- and y-axes), Lnl is the number of live leaves andLd is the percent of damage to leaves as estimatedby visual inspection. Plant growth or dieback wascalculated as the relative change between the initialmeasurements (To) and those at the end of themonitoring period (Tf).

The soil, forest floor, or epiphytic organic mat-ter (Ingram & Nadkarni 1993) beneath bromeliadsselected for transplanting was sampled at the startof the experiment. Soils were sampled directly un-der bromeliads from 0 to 10 cm depth using a 2.5cm wide soil corer. Soils were air-dried, sievedthrough a 2 mm mesh screen, and ground with aWiley Mill. Forest floor litter was removed from a10 3 10 cm2 area directly beneath the bromeliad,dried at 658C, weighed to determine mass, andground in a Wiley mill. Samples of epiphytic or-ganic matter, hereafter referred to as arboreal sub-strate, included bryophyte mats (which consistedof live and dead bryophytic tissue and arboreal soil)growing on tree stems in a 10 3 15 cm2 area underthe bromeliad base.

NUTRIENT ANALYSES. We determined the ex-changeable soil nutrient concentrations under bro-meliads transplanted to and around the forest floor(including roadsides) and total nutrient content offorest floor and tree substrates under transplantedbromeliads. Both the original substrate (To) andthe final substrate (Tf) were measured for brome-liads harvested within the first 9 months. No sub-

strate samples were taken at 16 months. Exchange-able Ca, Mg, K, Mn, Fe, Zn, and Al concentrationswere determined on a Perkin Elmer ICP at YaleUniversity after extracting 5 g of air-dried andground soil with 50 ml of NH4Cl on a verticalvacuum extractor (Silver et al. 1994). Extractablesoil P was determined after two successive extrac-tions of 2 g air-dried and ground soil in 20 ml ofBray solution (0.003 M NH4F and 0.025 M HCl;Olsen & Sommers 1982). Analyses were conductedon a Technicon Auto Analyzer at Yale University.Total C and N were determined for soil sampleson a LECO CHN analyzer at Yale University. Soilmoisture content was determined gravimetricallyafter drying at 1058C. All data reported here areon an oven-dry soil basis unless noted. Arborealsubstrate samples and plant tissues were analyzedfor total C and N at the University of Californiaat Berkeley on a CE Elantec CN analyzer. For Cand N analyses, three 0.3 g replicates of each sam-ple were analyzed. Organic matter content was de-termined by loss on ignition (LOI) at 5008C. Allanalyses included blanks, reference samples, andsample replicates for quality control.

The total elemental composition of tree sub-strate, forest floor, and bromeliad tissue sampleswas estimated using a modified Kjeldahl procedure(Parkinson & Allen 1975). All samples were driedat 658C and ground. Approximately 0.25 g sampleswere then digested in 8 ml H2SO4 and LiSO4 us-ing a block digestor. Solutions were analyzed on aPerkin Elmer ICP at Yale University and a Spec-traspan V GCP at the International Institute ofTropical Forestry (IITF) for total Ca, Mg, K, P,Mn, Fe, Zn, and Al. A subset of solutions wasanalyzed both at IITF and Yale University and gaveresults within 5 percent of one another. All runsincluded blanks, standard reference material, andsample replicates for quality control.

STATISTICAL ANALYSES. Data were analyzed usingSystat 7.0 (Wilkinson 1991). Likelihood ratio chi-square tests were used to analyze effects of movingbromeliads on life history characteristics such asmortality and reproduction. We focus on the 6-month intensive data set. Analysis of covariance(ANCOVA) was used to compare regressions ofplant growth over time. ANOVAs were used to de-termine (1) effects of transplanting bromeliadswithin and between forest floor, stems, and road-sides on root growth and nutrient content and (2)effects of location (stem vs. forest floor vs. roadside)on initial and final substrate nutrient concentra-tions and tissue mass and nutrient content. The


TABLE 2. Life history characteristics of bromeliads used for transplant experiments in an upper montane rain forest,Luquillo Experimental Forest, Puerto Rico. See Table 1 for abbreviations. Numbers in parentheses indicateplants missing from a treatment and presumed dead at the end of 16 months.

Treatment N Dead/absent specimens Flowering plants Plants w/juveniles

6 Months; N 5 100FF controlFF (in place)FF → FFFF → RoadFF → TreeTree controlTree (in place)Tree → TreeTree → FF

101010201010101010

000200001

320731111

311521111

16 Months; N 5 46FF controlFF (in place)FF → FFFF → RoadFF → TreeTree controlTree (in place)Tree → TreeTree → FF

444

1244443

0 (1)0 (1)0 (2)5 (2)

0000

1 (1)

110410011

111320012

Tukey–Kramer least significant difference (LSD)multiple comparison test was used to determinewhere significant differences occurred, and logtransformations were performed if necessary tomeet the assumptions of ANOVA. We report dataon life history characteristics for the 9- (N 5 73)and 16-month (N 5 46) samples, but we cautionthat sample sizes for these data are reduced. Weused a power analysis to evaluate our ability to de-tect significant effects with a given sample size. Sig-nificance was determined as P , 0.05 unless oth-erwise noted.

RESULTS

HEALTH, GROWTH AND MORTALITY. Total brome-liad mortality for the study was 3 percent after 6months (N 5 100) and 30 percent after 16 monthsof monitoring (N 5 46; Table 2). The total mor-tality at 16 months included 7 individuals, fromvarious treatments, that could not be located afterthe hurricane season of 1996. If these individualsare not included in the mortality calculations, totalmortality was 14 percent for the 16-month period.At 6 months, transplant location had not signifi-cantly affected bromeliad survivorship (Table 3);however, location did have a significant effect by16 months (P , 0.01, power 5 0.84). Mortalityoccurred in individuals that were forest floor con-trols or transplants to the forest floor or roadside.

There was no mortality of individuals on trees, andthere were no significant differences in mortalitybetween forest floor and roadside plants.

During the first 6 months, 19 percent of thestudy bromeliads produced flowers (Table 2). In allcases, flowering was initiated between August andNovember and lasted for several weeks. None ofthe individuals we monitored flowered more thanonce during the experiment. While treatment ap-peared to have a weak effect on flowering duringthe initial 6 months of the study (P , 0.1), thispattern was not apparent for the subset of plantsfollowed for 9 and 16 months (power 5 0.70).Flowering was significantly correlated with produc-tion of juvenile offshoots for all time periods ex-amined (P , 0.01, power 5 1.0); the majority ofjuveniles were fully formed 2–4 months after ces-sation of flowering. Individuals transplanted toroadsides had slightly higher rates of flowering (P, 0.1) than those of other treatments (Table 2).

Herbivore damage was estimated by measuringloss of foliar tissue as a percent of total leaf area.Herbivory losses varied significantly with experi-mental treatment (P , 0.01) but were unaffectedby measurement date for the first 6 months (Table3). In particular, forest floor and tree stem controls,and plants moved from one forest floor site to an-other experienced higher rates of herbivore damagethan other transplanted individuals (Table 4).Plants transplanted to roadsides were relatively un-


TABLE 3. The effects of treatment and time on growth and life history measures of Guzmania berteroniana in an uppermontane rain forest, Luquillo Experimental Forest, Puerto Rico. No significant differences existed betweenuntouched controls and ‘‘in place’’ controls (FC and F; TC and T); so these treatments were pooled by location(T or F) for analyses. (A) chi-square results for bromeliad status at 6-month harvest date. G2 is the likelihoodchi-square statistic. (B) ANOVA and ANCOVA results for herbivory and plant growth after 6 months ofmonitoring.

(A) Source of variation G2 N df P

Mortality 3 TreatmentFlowering 3 TreatmentJuvenile prod. 3 TreatmentJuvenile prod. 3 Flowering

7.4410.63

5.3939.00

100100100100

66

122

0.280.100.940.00

(B) Source of variation F N df P

Herbivory (%)TreatmentDateTreatment 3 Date

9.530.793.94

122012201220

61117

0.000.650.00

Relative plant growthTreatmentDateTreatment 3 Date

5.711.332.82

115611561156

61218

0.000.200.00

TABLE 4. Growth and health characteristics of bromeliads used for transplant experiments in an upper montane rainforest, Luquillo Experimental Forest, Puerto Rico. Standard errors are indicated by parentheses. Treatments withdifferent lowercase letters are significantly different at the 95 percent level using a Tukey-Kramer LSD multiplecomparison test. No significant differences existed between untouched controls and ‘‘in place’’ controls (FC andF; TC and T); so these treatments were pooled by location (T or F) for analyses. Root growth was measuredfor new roots only at 6, 9, and 16 months on harvested specimens. No root growth data (nd) are included forforest floor and tree stem controls because new growth could not be determined on undisturbed individuals. SeeTable 1 for abbreviations.

Treatment N

Herbivory at6 months (%)

x̄ (SE)

D Plant leaf areaat 6 months (%)

x̄ (SE) N

Root growth rate(cm/d)x̄ (SE)

FF controlFF → FFFF → RoadFF → TreeTree controlTree → TreeTree → FF

20102010201010

6.6 (0.3)ab7.1 (0.4)a5.2 (0.4)bc5.9 (0.3)c7.3 (0.3)a4.9 (0.5)c5.4 (0.5)c

218.5 (7.0)a224.2 (9.1)ab222.5 (8.0)b

0.4 (8.3)c20.3 (6.1)c23.7 (7.4)c

213.9 (6.0)a

167

13102010

8

nd0.28 (0.13)b0.56 (0.15)a0.71 (0.18)ab

nd0.62 (0.15)a0.23 (0.05)bc

affected by herbivory over the study period but ini-tially experienced a high degree of leaf damage dueto desiccation from sun exposure. The majority ofthese individuals responded by growing new leavesor juvenile shoots soon after sun damage occurred,and thereafter appeared unaffected by high lightintensity. At the 9- and 16-month harvest dates,leaf area affected by herbivory had increased.

There were no statistically significant differenc-es in the size of bromeliads growing on the groundor in trees at the start of the experiment. After 6months, forest floor bromeliads had lost a signifi-cant amount of leaf area relative to bromeliads

growing on trees (Tables 3, 4). This trend contin-ued to be evident on the subset of individuals mon-itored for 16 months. Plants moved to the roadsidealso lost considerable leaf area (x̄ 6 SE: 222.5 68%), primarily due to sun scorching. In the forest,the majority of leaf area loss was due to physiolog-ical senescence as opposed to herbivore or sundamage. Relative foliar growth of individuals trans-planted to or among forest floor sites was less than1, indicating a loss of leaf area, while individualstransplanted to or among trees had higher relativefoliar growth rates close to or greater than 1, in-dicating net growth (Fig. 2). The size of individuals


FIGURE 2. The relative bromeliad growth rate (RGR)in seven experimental treatments six months after trans-planting in an upper montane rain forest, Luquillo Ex-perimental Forest, Puerto Rico. See Table 1 for abbrevi-ations. No significant differences existed between un-touched controls and ‘‘in-place’’ controls (FC and F; TCand T); so these treatments were pooled by location (Tor F) for analyses. When RGR 5 1, there was no changefrom initial size. Different lowercase letters indicate sig-nificant differences at the 95 percent level.

FIGURE 3. Mean total new root length (cm) for bro-meliads harvested after 6 to 9 months of experimentaltreatments in an upper montane rain forest, Luquillo Ex-perimental Forest, Puerto Rico. See Table 1 for abbrevi-ations. No significant differences existed between un-touched controls and ‘‘in place’’ controls (FC and F; TCand T); so these treatments were pooled by location (Tor F) for analyses. Different lowercase letters indicate sig-nificant differences at the 95 percent level.

moved to tree stems and tree controls stayed rela-tively constant over the study period (Table 4).

For the combined pool of bromeliads harvestedon all three dates (N 5 84), average root growthrates 6SE ranged from 0.23 6 0.05 cm/day to0.71 6 0.18 cm/day (Table 4) and total new rootlength was significantly affected by treatment (P ,0.05, power 5 0.79). Bromeliads moved to theforest floor from trees and those moved to a newlocation on the forest floor showed significantly lessnew root growth than individuals moved to treesor to roadsides, regardless of harvest date (Fig. 3).After 6 months, bromeliads transplanted to the for-est floor had a low amount of total new root length(x̄ 6 SE: 53 6 15 cm), while those moved to treestems grew an average 189 6 43 cm new rootlength.

TISSUE AND SUBSTRATE CHEMISTRY. Overall, ele-ment concentrations were significantly higher in fo-liar tissues than in roots; the exceptions were C andFe, which were lower in leaves than in roots, andAl, which did not differ significantly between thesetwo tissue types (Table 5). Foliar element concen-trations were not significantly influenced by treat-ment in most cases; however, control plants grow-

ing on trees contained significantly higher leaf Cthan plants from all other treatments (P , 0.1). Inaddition, foliar Mg was 29 percent lower in plantsgrowing by roadsides than those growing in theforest. In root tissues, forest floor controls had sig-nificantly lower Mg, K, and P than plants frommost other treatments, but had higher concentra-tions of Al and Fe relative to arboreal controls (Ta-ble 5). Bromeliads transplanted from forest sites toroadsides gained Ca, Mg, K, and P (P , 0.01)relative to forest floor controls. While root C waslow in roadside plants, it was also highly variableand differed significantly only from arboreal con-trols (Table 5). There were no additional patternsin foliar or root element content when plant masswas considered.

Arboreal substrate P concentrations were higher(Fig. 4a) and K was lower in samples collected be-neath bromeliads than in samples collected fromunoccupied substrates sampled at the beginning ofthe study (P , 0.1; Table 6a). We also found thatarboreal substrates under control plants containedsignificantly higher Fe (P , 0.1) and lower Ca andMg than initially unoccupied transplant sites (Ta-ble 6a). Arboreal substrates were composed of de-caying bryophytes, sloughed tree bark, and undis-


TA

BL

E5.

Tota

lel

emen

tal

conc

entr

atio

nsin

folia

ran

dro

otti

ssues

ofG

uzm

ania

bert

eron

iana

tran

spla

nted

amon

gfo

rest

floor

,tr

ee,

and

road

side

loca

tion

sin

anup

per

mon

tane

rain

fore

st,Lu

quill

oE

xper

imen

tal

Fore

st,Pu

erto

Ric

o.Va

lues

are

mea

nsof

6-an

d9-

mon

thha

rves

ts(N

55–

7).

No

signi

fican

tdi

ffere

nces

exist

edbe

twee

nun

touc

hed

cont

rols

and

‘‘in

plac

e’’co

ntro

ls(F

Can

dF;

TC

and

T);

soth

ese

trea

tmen

tsw

ere

pool

edby

loca

tion

(Tor

F)fo

ran

alys

es.

Stan

dard

erro

rsar

ein

pare

nthe

ses.

Diff

eren

tlo

wer

case

lett

ers

wit

hin

anu

trie

ntam

ong

sam

ple

type

sin

dica

tesig

nific

ant

diffe

renc

esat

the

95pe

rcen

tle

vel

usin

ga

Tuke

y–K

ram

erLS

Dte

stan

dar

ein

clud

edw

hen

the

over

all

AN

OVA

mod

elw

assig

nific

ant

atth

e95

perc

ent

leve

l.A

n*

indi

cate

sm

odel

signi

fican

ceat

the

90pe

rcen

tle

vel.

Upp

erca

sele

tter

sde

note

signi

fican

tdi

ffere

nces

betw

een

aver

age

conc

entr

atio

nsof

folia

ran

dro

otsa

mpl

es(t

reat

men

tsco

mbi

ned)

.Se

eTa

ble

1fo

rab

brev

iati

ons.

Sam

ple

type

Tota

lC

(%)

Tota

lN

(%)

Nut

rien

tco

ncen

trat

ion

(mg/

g65

8Cdr

yw

t.)

Ca

Mg

KA

lP

Fe

Folia

geFF

cont

rol

FF→

FFFF

→R

oad

FF→

Tree

Tree

cont

rol

Tree

→Tr

eeTr

ee→

FF

45.6

1(0

.32)

a*45

.74

(0.3

9)a

45.3

9(0

.33)

a43

.62

(1.2

1)a

48.2

8(2

.59)

b45

.55

(0.4

1)a

44.9

7(0

.36)

a

1.07

(0.1

7)1.

09(0

.14)

0.86

(0.1

2)0.

94(0

.13)

0.88

(0.5

4)0.

74(0

.08)

0.87

(0.0

3)

3.22

(0.3

3)3.

16(0

.22)

3.14

(0.3

6)3.

16(0

.13)

3.39

(0.3

7)2.

95(0

.13)

2.98

(0.2

4)

3.42

(0.3

6)ab

3.57

(0.3

1)ab

2.53

(0.0

7)c

3.82

(0.2

5)b

3.92

(0.3

9)ab

3.41

(0.3

2)ab

3.67

(0.1

7)ab

16.0

4(1

.37)

18.0

5(0

.83)

14.0

0(0

.79)

17.3

4(1

.19)

17.8

1(1

.38)

16.3

6(1

.82)

16.7

5(1

.00)

0.12

(0.0

2)0.

20(0

.07)

0.08

(0.0

2)0.

14(0

.01)

0.12

(0.0

3)0.

13(0

.02)

0.15

(0.0

4)

2.27

(0.5

7)2.

83(0

.52)

1.52

(0.1

8)2.

42(0

.41)

2.39

(0.3

2)1.

84(0

.28)

2.22

(0.2

4)

0.09

(0.0

3)0.

19(0

.11)

0.04

(0.0

1)0.

12(0

.01)

0.12

(0.0

3)0.

07(0

.01)

0.09

(0.0

2)x̄

45.5

9(0

.80)

A0.

92(0

.17)

A3.

17(0

.27)

A3.

52(0

.29)

A16

.69

(1.2

3)A

0.13

(0.0

2)A

2.24

(0.3

8)A

0.10

(0.0

3)A

Roo

tsFF

cont

rol

FF→

FFFF

→R

oad

FF→

Tree

Tree

cont

rol

Tree

→Tr

eeTr

ee→

FF

47.4

5(0

.60)

ab*

46.3

2(0

.76)

ab45

.18

(2.6

4)a

46.0

9(2

.52)

ab49

.68

(1.5

2)c

47.9

5(0

.84)

ac48

.83

(0.9

5)bc

0.74

(0.0

4)0.

82(0

.09)

0.83

(0.0

7)0.

74(0

.08)

0.91

(0.1

0)0.

81(0

.08)

0.83

(0.0

8)

1.58

(0.2

3)a

1.96

(0.1

9)ab

2.94

(0.4

9)c

1.75

(0.1

9)ab

1.56

(0.2

7)a

1.89

(0.4

2)ab

2.28

(0.2

1)b

0.32

(0.0

8)a

0.42

(0.1

0)ab

0.58

(0.0

8)bc

0.52

(0.1

5)bc

0.51

(0.0

9)b

0.57

(0.1

8)b

0.72

(0.1

3)c

0.90

(0.1

4)a*

1.48

(0.3

8)b

1.67

(0.3

1)b

1.31

(0.3

1)ab

1.49

(0.3

4)b

1.86

(0.4

1)b

1.35

(0.2

1)b

0.50

(0.2

0)a*

0.58

(0.1

6)a

1.25

(0.8

7)a

0.17

(0.0

8)ab

0.16

(0.0

6)b

0.13

(0.0

8)b

0.32

(0.0

8)ab

0.72

(0.0

5)a

0.93

(0.0

8)b

0.96

(0.1

0)b

0.91

(0.1

1)b

0.94

(0.1

0)b

1.03

(0.1

0)b

0.96

(0.1

1)b

0.36

(0.1

6)a

0.61

(0.1

1)a

1.47

(0.9

6)a

0.23

(0.1

2)ab

0.11

(0.0

4)b

0.11

(0.0

6)b

0.32

(0.1

0)a

x̄47

.36

(1.4

0)B

0.81

(0.0

8)B

1.89

(0.2

8)B

0.50

(0.1

1)B

1.38

(0.2

9)B

0.42

(0.2

0)A

0.90

(0.0

3)B

0.41

(0.1

9)B


FIGURE 4. (a) Total arboreal substrate P (mg/g); (b)total terrestrial substrate C (%); (c) total terrestrial sub-strate N (%); and (d) terrestrial substrate exchangeable P(mg/kg) in samples collected beneath bromeliads trans-

←

planted among forest floor, tree stems, and roadsides inthe Luquillo Experimental Forest, Puerto Rico. ‘‘Control’’substrate was collected beneath non-transplanted plantsat the end of the study. ‘‘Original’’ substrate was collectedprior to transplanting; ‘‘Transplant’’ substrate was sam-pled in the new transplant location; and ‘‘Final Trans-plant’’ substrate was collected at the transplant locationat the end of the study. Different lowercase letters indicatesignificant differences among substrate groups at the 95percent level.

tinguishable organic matter, and had significantlyhigher organic matter concentrations in sites be-neath bromeliads than in tree substrates with nobromeliad colonization. When we corrected forhigh variability in organic matter content of arbo-real substrates (by calculating concentrations on anash weight basis), we still found higher P (P ,0.01) in substrates of control and transplanted bro-meliads relative to substrates collected in unoccu-pied sites. Total K concentrations continued toshow the opposite trend (P , 0.01) and were high-est in unoccupied substrates sampled at the begin-ning of the study.

We compared total C and N and exchangeableelements in terrestrial bromeliad substrates (soilplus litter) with similar substrate materials fromunoccupied sites. Substrates sampled from beneathcontrol plants and plants prior to removal con-tained significantly more C (P , 0.01), N (P ,0.01), and extractable P (P , 0.05) than unoccu-pied sites (Fig. 4b, c, d). Terrestrial substrate sam-ples collected at the end of the study beneath trans-planted bromeliads had gained C (Fig. 4b) and N(Fig. 4c) and were no longer significantly differentthan either control or unoccupied sites. Similarly,substrates under terrestrial bromeliads had slightlyhigher exchangeable P (Fig. 4d) compared to un-occupied sites. Phosphorus concentrations weresubstantially higher than reported literature valuesfor soils in the mountains of the LEF (Frangi &Lugo 1985; Silver et al. 1994, 1999; Table 6b). Ofthe other exchangeable elements we measured, onlyAl varied consistently (P , 0.1), and was higher insubstrates of controls and transplant bromeliadsprior to their removal (Table 6b).

DISCUSSIONThere is relatively little known about the life his-tory characteristics of most bromeliad species, andG. berteroniana, the most common bromeliad inthe upper montane forests of the LEF, is no excep-tion. Most authors describe bromeliads as obligate


TA

BL

E6.

(A)

Tota

lel

emen

tan

dor

gani

cm

atte

rco

ncen

trat

ions

asm

easu

red

bylo

sson

igni

tion

(LO

I)of

arbo

real

subs

trat

esan

d(B

)ex

chan

geab

leel

emen

tsin

terr

estr

ial

subs

trat

es.

‘‘Ori

gina

lsit

eT

o’’su

bstr

ate

sam

ples

colle

cted

atsta

rtof

study

atin

itia

lbr

omel

iad

loca

tion

s.‘‘T

rans

plan

tsit

eT

o’’su

bstr

ate

sam

ples

colle

cted

atsta

rtof

study

atpr

evio

usly

unco

loni

zed

tran

spla

ntlo

cati

ons.

‘‘Con

trol

Tf’’

subs

trat

esa

mpl

esco

llect

edbe

neat

hco

ntro

lsdu

ring

6-an

d9-

mon

thha

rves

ts.‘‘T

rans

plan

tTf’’

subs

trat

esa

mpl

esco

llect

edbe

neat

htr

ansp

lant

edbr

omel

iads

afte

r6

and

9m

onth

sof

grow

thin

new

loca

tion

.Va

lues

are

mea

nsw

ith

stand

ard

erro

rsar

ein

pare

nthe

ses.

Diff

eren

tlo

wer

case

lett

ers

wit

hin

anel

emen

tam

ong

subs

trat

egr

oups

indi

cate

signi

fican

tdi

ffere

nces

atth

e95

perc

ent

leve

lus

ing

aTu

key–

Kra

mer

LSD

mul

tipl

eco

mpa

riso

nste

stan

dar

ein

clud

edw

hen

the

over

all

AN

OVA

mod

elw

assig

nific

ant

atth

e95

perc

ent

leve

l.A

n*

deno

tes

mod

elsig

nific

ance

atth

e90

perc

ent

leve

l.

(A) Su

bstr

ate

grou

pN

Arb

orea

lsu

bstr

ate

tota

lel

emen

tco

ncen

trat

ion

(mg/

g)

Ca

Mg

KA

lFe

Ppe

rcen

tLO

I

Con

trol

Tf

Ori

gina

lsi

teT

oTr

ansp

lant

site

To

Tran

spla

ntsi

teT

f

16 28 22 14

2.98

(0.3

1)a

4.49

(1.2

9)b

3.05

(0.4

3)b

2.85

(0.3

0)a

1.10

(0.1

1)a

2.61

(0.5

8)b

1.55

(0.1

3)c

1.13

(0.1

5)a

3.60

(0.3

2)a*

4.76

(0.8

6)ab

5.33

(0.5

3)b

3.57

(0.4

7)a

0.65

(0.3

2)a*

0.35

(0.0

7)a

0.23

(0.0

5)ab

0.22

(0.0

9)b

0.52

(0.2

9)a*

0.22

(0.0

5)b

0.16

(0.0

5)b

0.18

(0.0

8)b

2.19

(0.1

9)a*

2.13

(0.3

3)a

1.56

(0.1

5)b

1.93

(0.1

2)a

91.5

8(1

.37)

a90

.01

(1.3

0)a

86.2

0(1

.53)

b92

.06

(1.1

8)a

(B) Su

bstr

ate

grou

pN

Terr

estr

ial

subs

trat

eex

chan

geab

leel

emen

ts(c

mol

/kg)

Ca

Mg

KA

lFe

Exc

hang

eabl

eP

(mg/

kg)

Con

trol

Tf

Ori

gina

lsi

teT

oTr

ansp

lant

site

To

Tran

spla

ntsi

teT

f

11 31 40 19

6.87

(2.3

1)7.

47(1

.52)

6.79

(1.0

2)6.

70(1

.43)

3.77

(1.0

2)3.

02(0

.69)

2.15

(0.4

8)2.

43(0

.39)

0.70

(0.1

2)a

1.03

(0.1

9)b

0.61

(0.1

9)a

0.54

(0.1

3)a

4.59

(1.3

2)a*

2.83

(0.2

2)a

2.12

(0.3

8)b

2.02

(0.4

4)b

0.52

(0.1

2)0.

26(0

.05)

0.21

(0.0

4)0.

35(0

.07)

68.8

8(4

3.35

)a64

.06

(19.

82)a

10.8

4(6

.03)

b13

.33

(7.6

2)ab


arboreal epiphytes (Kress 1986, Benzing 1990).While bromeliads have been noted on the forestfloor (Matelson et al. 1993, Richardson et al.2000), it is often assumed that they cannot persist,and thus are functionally equivalent to litterfall.The common assumption that tropical tank bro-meliads are adapted only to an arboreal environ-ment is misleading, however, because their require-ments for growth (atmospheric nutrients, humidity,and low to medium light levels) may be availablein both arboreal and terrestrial habitats. This is par-ticularly true in short-statured montane forests thatare frequently characterized by high humidity andinputs from cloud- and rainwater, but this may alsooccur in other tropical forests.

In the LEF, mature bromeliads commonly oc-cur on the forest floor, achieving densities as highas 700 individuals/ha. The absence of single juve-nile bromeliads on the forest floor, however, indi-cates that G. berteroniana is unable to germinateon the forest floor; that very small plants are notas susceptible to displacement; and/or that seedsand small propagules suffer lethal disturbancecaused by falling litter and woody debris. Adultbromeliads on the forest floor probably establishedthere after displacement from arboreal sites duringsemiannual tropical storms or hurricanes. Theirability to persist long after being knocked from thecanopy may be due to the relatively low canopy(15 m) and steep slopes that allow light to enterfrom the side during part of the day. The large sizeand the ability to expand through vegetative repro-duction contribute to the abundance and high sur-vivorship of adult G. berteroniana. Indeed, the lowmortality rate of transplanted and control brome-liads during our 16-month experiment suggest thatthese epiphytes are not obligately dependent on ar-boreal habitats for survival and growth. Althoughbromeliads on trees had higher survival and rootgrowth rates than forest floor bromeliads, we ob-served many G. berteroniana that maintained sub-stantial leaf area, produced flowers, and reproducedvegetatively (producing juvenile offshoots) after be-ing moved to the ground. It is possible that weoverestimated growth and survival by loosely sup-porting the bromeliads we transplanted to theground. Under natural disturbance conditions,some plants are likely to fall inverted and may suf-fer higher mortality or have to allocate more re-sources to root growth to right themselves. Ourdisturbance control treatments (‘‘FF in place’’ and‘‘Tree in place’’) did not have significantly differentmortality or relative growth from undisturbed con-

trols, suggesting that this species is fairly resilientto mechanical perturbation.

Longevity of arboreal and terrestrial epiphyteshas not been well documented, although Ober-bauer et al. (1996) noted that several Tillandsiaspp. survived at least 10 months following hurri-cane disturbance in southern Florida. Our resultssuggest that bromeliad longevity can exceed 16months following a major disturbance event. In ad-dition, the G. berteroniana that we moved to newhabitats quickly grew new securing roots on thescale of tens of millimeters per day. Plants that weremoved to tree stems, which represents a very un-stable habitat, grew the most new root length. Theability to quickly prioritize C allocation to new tis-sues such as roots has been suggested as a responseto a lack of specific resources (Bloom et al. 1985).In this case, physical stability may be the limitingresource stimulating plant responses. Individualsthat were moved to roadsides, characterized by dri-er and sunnier conditions, responded by allocatingresources to new shoots and roots instead of estab-lished tissues. Roadside plants also produced moreflowers than forest plants, suggesting that in thisspecies, allocation to reproduction is possible understressful conditions. Other bromeliad species alsohave been shown to respond to disturbance withthe production of juvenile shoots (Oberbauer et al.1996). The correlation we found between flower-ing and axillary shoot production supports the as-sumption that individual bromeliad shoots are de-terminate (Richardson et al. 2000). Our resultsshowed that after being moved to the forest floor,G. berteroniana did not exhibit a shade-stress re-sponse by increasing leaf area, but instead lostmany older leaves to senescence and herbivory.This is partly because light levels are not signifi-cantly different in this forest type between the for-est floor and mid–low canopy levels (Olander et al.1998) and partly because the higher degree of dis-turbance on the forest floor may overwhelm theplant’s leaf turnover capacity.

The effects of herbivory on our study brome-liads varied with habitat. In particular, plants grow-ing along roadsides experienced less herbivory thancontrol bromeliads on the forest floor. Bromeliadsthat were moved to or relocated among trees andthose moved from trees to the forest floor had lessherbivore damage than stationary controls. Thismay have been due to a lag time following relo-cation before plants were discovered by herbivores.If this is the case, the ability to successfully relocateor ‘‘be mobile’’ may have an added benefit of re-duced risk of herbivory, at least over the first year;


however, some of the events that lead to bromeliaddisplacement can also result in canopy openings,which increase light and temperature and can leadto leaf desiccation (as evidenced by the high degreeof sun damage to roadside bromeliad).

We hypothesized that habitat (terrestrial or ar-boreal) would have a significant impact on C andnutrient concentrations in bromeliad tissues. Somebromeliads have been shown to switch uptake andallocation strategies as resource availability changes(Nadkarni & Primack 1989). Here, G. berteronianagenerally allocated more nutrients (except Fe) toleaves than to root tissues. Bromeliads did not re-spond to different microenvironments by changingnutrient allocation to leaves. Arboreal bromeliads,however, contained higher levels of Mg, K, and Pin root tissues than the forest floor controls. Thisresource allocation to support structures may berelated to bromeliads’ increased need for stabilitywhen growing suspended against a tree stem. Whilethese patterns of tissue concentrations could becontrolled by differential root uptake in arborealversus terrestrial habitats, to show this conclusively,further study using substrate fertilization and/orisotopic tracers needs to be conducted. Higher Feand Al in root tissues on the forest floor is likelyto result from mineral plaques precipitating on rootsurfaces of newly rooted bromeliads.

In addition to altering plant nutrient content, dis-placed bromeliads also changed the nutrient contentof their substrate. Tank bromeliads, with their abilityto impound water and floral and faunal detritus, areeffective repositories for decomposing organic matter,which often leads to the formation of arboreal soil.We found that arboreal substrate (a combination ofbryophytes and arboreal soil) that developed directlybeneath G. berteroniana had higher organic mattercontent, P, and Fe, and lower K than similar materialfrom sites uncolonized by bromeliads. These patternsresult both from nutrient capture by bromeliads andthe effects of nutrient inputs from bromeliads to theirsubstrate. Bromeliads influence arboreal soil by cap-turing nutrients from atmospheric sources, intercept-ing tree litterfall and promoting decay in tank im-poundments, leaking nutrients from tanks, disruptingstemflow, and producing bromeliad leaf/root litter.Nutrients that are mineralized in arboreal environ-ments may be taken up by microorganisms, otherepiphytes, or leached to other parts of the ecosystem(Nadkarni & Matelson 1991). Epiphyte nutrient cy-cling is difficult to quantify at an ecosystem scale dueto high spatial and temporal variability. For example,in a Costa Rican cloud forest, epiphyte-derived litterdecay and nutrient dynamics were highly sporadic in

space and time but generally exhibited fast decay ratesand slow nutrient turnover times (Nadkarni & Ma-telson 1992).

Fallen epiphytes affected ecosystem nutrient cy-cles differently than litterfall. Beneath terrestrialbromeliads, we measured increased substrate C, N,and P. This was likely a result of nutrient-rich bro-meliad litter inputs (Nadkarni & Matelson 1992)and leakage from the tank. It is not surprising thatfor forest floor-dwelling G. berteroniana, the ma-terial impounded in the plant’s tank, together withits senesced tissues, will eventually contribute to theterrestrial ecosystem nutrient cycle when the plantdies (Veneklaas 1990, Nadkarni & Matelson1992). Our results, however, indicate that underterrestrial bromeliads, increased soil C and nutrientpools may develop while the plants are alive, aug-menting the eventual inputs that will occur withplant death. Thus, fallen bromeliads not only con-tribute appreciably to the forest floor nutrient cap-ital through their eventual death and decay but alsoaffect the timing of nutrient release because of thelag time between falling and plant death.

The high concentrations of P we found in bro-meliad substrates are particularly intriguing. InPuerto Rican floodplain forests, Frangi and Lugo(1985) found that concentrations of P were eighttimes higher in epiphyte-derived arboreal soil thanin terrestrial soil. This may have been a result ofhigh organic matter content, as well as high at-mospheric nutrient inputs to these systems fromtrade winds containing high concentrations of Sa-haran dust (Talbot et al. 1986). It could also havebeen caused by lower Fe in arboreal soil relative toterrestrial soil. Phosphorus can be rapidly occludedin tropical soils by Fe and Al oxides and hydrox-ides, removing it from the soil exchange complex(Sanchez 1976). Phosphorus is commonly a lim-iting element in tropical forests. Phosphorous add-ed through epiphytic soil results in hot spots of Pavailability on the landscape.

Our results showed that bromeliads are capableof surviving drastic changes in their habitat con-ditions caused by displacement and relocation.Most plants are static with regard to locationthroughout their life cycle. In high wind environ-ments, however, epiphytes can move to new loca-tions, effectively dispersing as adults and colonizingnew habitats. In this study, the successful coloni-zation of new habitats was augmented by the abil-ity to allocate resources to new root growth, as wellas the production of new shoots. As with all plants,bromeliads modified their environment; but unlikemost plants, which tend to deplete nutrients in


their rhizosphere, transplanted bromeliads locallyconcentrated C, N, and P in their substrate over arelatively short period of time. These resourceswere probably derived from atmospheric sourcesand can be considered a net input to the ecosystemas opposed to a recycling of resources already pre-sent. In summary, our results showed that brome-liads can successfully disperse as adults surviving toreproduce vegetatively, and that the presence ofbromeliads can locally enrich substrate nutrientcapital. We found that G. berteroniana is adaptedto the dynamic nature of montane tropical forestswith its ability to relocate to new environments andsuccessfully reestablish.

ACKNOWLEDGMENTS

This research was supported by grants from the A. W.Mellon Foundation (to W. Silver) and the Institute forTropical Studies (to J. Pett-Ridge). Additional supportwas provided by NSF grant no. BSR-8811902 as part ofthe Long Term Ecological Research program and theUSDA Forest Service International Institute of TropicalForestry. We would like to thank M. Salgado-Ramirez, L.Kueppers, L. Olander, and C. Garcia for help with field-work. M. J. Sanchez, E. Lopez, and F. Scatena at theInternational Institute for Tropical Forestry, T. Siccama,J. Tilley at Yale University, and M. McGroddy at U.C.Berkeley provided facilities and help during laboratoryanalysis. Thanks also to J. Kirchner and C. D’Antoniofor advice on statistical analyses and the manuscript.

LITERATURE CITED

ASBURY, C. E., W. H. MCDOWELL, R. T. PIZARRO, AND S. BERRIOS. 1994. Solute deposition from cloud water to thecanopy of a Puerto Rican montane forest. Atmos. Environ. 28: 1773–1780.

BEARD, J. S. 1949. Natural vegetation of the Windward and Leeward Islands. Oxf. For. Mem. 21: 1–192.BENZING, D. H. 1980. The biology of the bromeliads. Mad River Press, Eureka, California.———. 1990. Vascular epiphytes: general biology and related biota. Cambridge University Press, Cambridge, England.BERMUDES, D., AND D. H. BENZING. 1991. Nitrogen fixation in association with Ecuadorean bromeliads. J. Ecol. 7:

531–536.BLOOM, A. J., F. S. CHAPIN, AND H. A. MOONEY. 1985. Resource limitation in plants—an economic analogy. Annu.

Rev. Ecol. Syst. 16: 363–392.BROWN, S., A. E. LUGO, S. SILANDER, AND L. LIEGEL. 1983. Research history and opportunities in the Luquillo

Experimental Forest. U. S. For. Serv. Gen. Tech. Rep. SO-44, South. For. Exp. Stn., New Orleans, Louisiana.FRANGI, J. L., AND A. E. LUGO. 1985. Ecosystem dynamics of a subtropical floodplain forest. Ecol. Monogr. 55: 351–

369.INGRAM, S. W., AND N. M. NADKARNI. 1993. Composition and distribution of epiphytic organic matter in a Neotropical

cloud forest, Costa Rica. Biotropica 25: 370–383.KRESS, W. J. 1986. A symposium: the biology of tropical epiphytes. Selbyana 9: 1–22.MATELSON, T. J., N. M. NADKARNI, AND J. T. LONGINO. 1993. Longevity of fallen epiphytes in a Neotropical montane

forest. Ecology 74: 265–269.MIGENIS, L. E., AND J. D. ACKERMAN. 1993. Orchid–phorophyte relationships in a forest watershed in Puerto Rico.

J. Trop. Ecol. 9: 231–240.NADKARNI, N. M., AND T. J. MATELSON. 1991. Fine litter dynamics within the tree canopy of a tropical cloud forest.

Ecology 72: 2071–2082.———, AND ———. 1992. Biomass and nutrient dynamics of epiphytic litterfall in a Neotropical montane forest,

Costa Rica. Biotropica 24: 24–30.———, AND R. B. PRIMACK. 1989. The use of gamma spectrometry to measure within-plant nutrient allocation of a

tank bromeliad, Guzmania lingulata. Selbyana 11: 22–25.OBERBAUER, S. F., K. VON KLEIST III, K. R. T. WHELAN, AND S. KOPTUR. 1996. Effects of Hurricane Andrew on

epiphytic communities within cypress domes of Everglades National Park. Ecology 77: 964–967.OLANDER, L. P., F. N. SCATENA, AND W. L. SILVER. 1998. Impacts of disturbance initiated by road construction in a

subtropical cloud forest in the Luquillo Experimental Forest, Puerto Rico. For. Ecol. Manage. 109: 33–49.OLSEN, S. R., AND L. E. SOMMERS. 1982. Phosphorous. In A. L. Page (Ed.). Methods of soil analysis, part 2: Chemical

and microbiological properties, agronomy no. 9, pp. 403–430. Am. Soc. Ag., Inc. and Soil Sci. Soc. Am.,Inc., Madison, Wisconsin.

PAOLETTI, M. G., R. A. J. TAYLOR, B. R. STINNER, D. H. STINNER, AND D. H. BENZING. 1991. Diversity of soil faunain the canopy and forest floor of a Venezuelan cloud forest. J. Trop. Ecol. 7: 373–383.

PARKINSON, J. A., AND S. E. ALLEN. 1975. A wet oxidation procedure suitable for the determination of nitrogen andmineral nutrients in biological material. Comm. Soil Sci. Plant Anal. 6: 1–11.

RICHARDSON, B. A., M. J. RICHARDSON, F. N. SCATENA, AND W. H. MCDOWELL. 2000. Effects of nutrient availabilityand other elevational changes on bromeliad populations and their invertebrate communities in a humidtropical forest in Puerto Rico. J. Trop. Ecol. 16: 167–188.

SANCHEZ, P. A. 1976. Properties and Management of Soils in the Tropics. J. Wiley Press, New York, New York.SILVER, W. L., A. E. LUGO, AND M. KELLER. 1999. Soil oxygen availability and biogeochemistry along rainfall and

topographic gradients in upland wet tropical forest soils. Biogeochemistry 44: 301–328.


———, F. N. SCATENA, A. H. JOHNSON, T. G. SICCAMA, AND M. J. SANCHEZ. 1994. Nutrient availability in a montanewet tropical forest—Spatial patterns and methodological considerations. Plant Soil 164: 129–145.

TALBOT, R. W., R. C. HARRIS, E. V. BROWELL, G. L. GREGORY, D. I. SABACHER, AND S. M. BECK. 1986. Distributionand geochemistry of aerosols in the tropical North Atlantic troposphere: relationship to Saharan dust. J.Geophys. Res. 91: 5173–5182.

VENEKLAAS, E. J., R. J. ZAGT, A. VAN LEERDAM, R. VAN EK, A. J. BROEKHOVEN, AND M. VAN GENDEREN. 1990.Hydrological properties of the epiphyte mass of a montane tropical rain forest, Colombia. Vegetatio 89: 183–192.

WALKER, L. R., N. V. L. BROKAW, D. J. LODGE, AND R. B. WAIDE. 1991. Ecosystem, plant, and animal responses tohurricanes in the Caribbean. Biotropica 23: 1–521.

———, J. K. ZIMMERMAN, D. J. LODGE, AND S. GUZMAN GRAJALES. 1996. An altitudinal comparison of growth andspecies composition in hurricane-damaged forests in Puerto Rico. J. Ecol. 84: 877–889.

WALKER, L. R., AND M. R. WILLIG. 1999. An introduction to terrestrial disturbances. In L. R. Walker (Ed.). Ecosystemsof the world; ecosystems of disturbed ground, pp. 116. Elsevier Science Inc., New York, New York.

WEAVER, P. L. 1994. Bano de Oro Natural Area, Luquillo Mountains, Puerto Rico. U.S. For. Serv. Gen. Tech. Rep.SO-111. South. For. Exp. Stn., New Orleans, Louisiana.

WILKINSON, L. 1991. Systat. The system for statistics. Systat, Inc., Evanston, Illinois.

Survival, Growth, and Ecosystem Dynamics of Displaced ... · viduals growing on palms and within 2 m of the ground were sampled. Transplanted bromeliads were carefully removed from

Documents