Quantifying the Deciduousness of Tropical Forest Canopies Under Varying Climates Author(S): Richard Condit, Kristina Watts, Stephanie A

International Association for Vegetation Science

Quantifying the Deciduousness of Tropical Forest Canopies under Varying Climates Author(s): Richard Condit, Kristina Watts, Stephanie A. Bohlman, Rolando Pérez, Robin B. Foster and Stephen P. Hubbell Reviewed work(s): Source: Journal of Vegetation Science, Vol. 11, No. 5 (Oct., 2000), pp. 649-658 Published by: Wiley Stable URL: http://www.jstor.org/stable/3236572 . Accessed: 10/12/2012 14:51

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Wiley and International Association for Vegetation Science are collaborating with JSTOR to digitize, preserve and extend access to Journal of Vegetation Science.

http://www.jstor.org

This content downloaded by the authorized user from 192.168.52.67 on Mon, 10 Dec 2012 14:51:47 PM All use subject to JSTOR Terms and Conditions Journal of VegetationScience 11: 649-658, 2000 ? IAVS;Opulus Press Uppsala. Printed in Sweden 649

Quantifying the deciduousness of tropical forest canopies under varying climates

Condit, Richardl*; Watts, Kristina2; Bohlman, Stephanie A.3; Perez, Rolando1; Foster, Robin B.4 & Hubbell, Stephen P.5

1CenterforTropical Forest Science, SmithsonianTropical Research Institute, Unit 0948, APO AA 34002-0948, USA; 2SouthamptonCollege (currentaddress: Schoolfor Public and EnvironmentalAffairs, Indiana University,Bloomington IN 47401, USA); 3College of Forest Resources, Universityof Washington,Seattle WA98192, USA;4Botany Depart- ment,The Field Museum,Roosevelt Rd. at LakeShore Dr., Chicago,IL 60605, USA;5Department of Botany, University of Georgia, Athens, GA 30602, USA; *Correspondingauthor; Fax +5072325978;E-mail [email protected]

Abstract. Deciduousness is an importantfunctional attribute Introduction of tropical trees, reflecting climatic conditions. Precisely and deciduousness in forests quantifying mapping tropical Deciduousness is a key functional trait both for indi- will be for remote necessary calibrating sensing images vidual trees and for forest communities. Partial or full which attempt to assess canopy properties such as carbon leaf loss during part of the year is an eco- cycling, productivity,or chlorophyll content. We thus set out important since it indicates that forest to assess the degree of canopy deciduousness in three moist, system trait, productivity semi-deciduous tropical forests in central Panama.One site is seasonally reduced, and deciduousness may correlate is a 6-ha research plot near the Atlantic coast of Panama, with important ecosystem-level characteristics, such as where rainfall is 2830 mm/yr. The second site is a 50-ha plot tree height, biomass, and diversity. Since deciduousness on Barro Colorado Island, near the center of the isthmus of changes canopy reflectance, it can be detected in satel- where rainfall is 2570 and the final Panama, mm/yr, site is a lite images (Reed et al. 1994; Bohlman et al. 1998), and 4-ha near the Pacific coast of Panama,where rainfall is plot thus could potentially be mapped over wide areas. Re- 2060 mm/yr. At each site, a randomsample of trees from all flecting its importance in tropical ecosystems, deciduous- canopy species (those with individuals > 30 cm DBH) were ness often provides a main axis for vegetation classifica- visited and scored for deciduousness three times during the tion Hall & Swaine Box 1997 dry season. The estimated peak fraction of deciduous (Beard 1944; 1981; 1996): individuals in the canopy at the wetter site was 4.8%, at the tropical dry forests are deciduous, wet forests are ever- intermediate site, 6.3%, and at the drier site, 24.3%. The green, and forests in between are called semi-deciduous. estimated fraction of crown area deciduous peaked at 3.6%, Deciduousness is thus an important and easy-to-meas- 9.7%, and 19.1% at the wet, medium, and dry sites respec- ure indicator, and models of how tropical forests re- The of tively. percentage canopy species that was deciduous spond to future climatic changes (Steffen et al. 1996) - 14%,28%, and 41 %- was much thanthe higher percentage might be calibrated by assessing how well they predict of deciduous because not all individuals of de- individuals, deciduous behavior under different climates. ciduous species were deciduous. During the 1999 dry sea- This requires calibration sites, where deciduousness son, every individual of all the deciduous species was visited and climate are at the two driersites, and the total numberof deciduous trees carefully measured, and despite the observed closely matched the estimated numbers based on importance of this single character in forest classifica- the smaller 1997 samples. tion and ecosystem function, there are very few quantitative estimates of deciduousness in tropical forests. The only indication of the degree of deciduousness typically Keywords: Climate gradient; Deciduousness; Phenology; reported is the fraction of species (e.g. Frankie et al. 1974; Tropicalforest. Reich 1995). Quantitative studies on the fraction of canopy cover lost during seasons of leaf fall are rare. We thus embarked on a straightforward effort to map and Abbreviation: BCI = BarroColorado Island. quantify deciduousness in three forests in central Panama. We sampled every canopy species for deciduous behavior, and then made a complete census of all deciduous species. The three sites, though close together, have different precipitation patterns, so we were able to quantify the relationship between climate and deciduousness.

This content downloaded by the authorized user from 192.168.52.67 on Mon, 10 Dec 2012 14:51:47 PM All use subject to JSTOR Terms and Conditions 650 Condit,R. et al.

Study sites on the Smithsonian Tropical Research Institute's research reserve on Barro Colorado Island was set up in The study was carriedout in lowland forests (< 200 1981 (Hubbell& Foster 1983, 1986, 1992; Conditet al. m elevation) adjacent to the Panama Canal. Annual 1995, 1996a, c). Here we use datafrom the 1995 census rainfall even in the driest sites in central Panama ex- of the plot. A 6-ha plot near the Atlantic coast at Fort ceeds potential evapotranspirationand is ample to sus- Shermanwas establishedin 1996, and a 4-ha plot near tain tall, moist, high-biomassforest. Thereis a long dry the Pacific, near the Cocoli River, was established in season, though,running from mid-Decemberuntil mid- 1994. The latter two sites are on former US military April, during which many trees lose all their leaves installations. (Leigh et al. 1982), and there is considerablevariation across the isthmus in the length of the period when potentialevapotranspiration exceeds precipitation(Fig. Methods 1). At the Pacific side of the Canal, near PanamaCity, total annualrainfall is 2060 mm and the dry season lasts All stems > 1 cm diameter-at-breast-height(DBH) 129 days; at BarroColorado in the centerof the isthmus, were identified to species, measured at breast height, rainfallis 2570 mm and the dry season 118 days; while and mapped,throughout all threeplots. Condit (1998b) nearthe Atlantic,rainfall is 2830 mm andthe dry season describes in detail the methods used. We are primarily is 106 days (Fig. 1). Condit (1998a) reportedslightly interestedin deciduousness of trees forming the upper differentestimates for dry season duration,but based on canopy, since this is where leaf-loss would have the an approximationfor potentialevapotranspiration (100 greatestimpact on ecosystem function, and where it is mm/month);here we use directmeasurements of poten- visible remotely. In addition, understory species are tial evapotranspiration(Fig. 1). All three sites have seldom deciduous(Frankie et al. 1974; Wright 1991): to typical lowland-tropical temperatureregimes, with a our knowledge, just one understorytree at BarroColo- mean close to 27 ?C throughoutthe year. rado - Erythrina costaricana - is (Condit et al. 1996b). We have establishedthree forest census plots across Thus, we only surveyed species which reached a DBH the isthmus,spanning this rainfallgradient. A 50-ha plot of 30 cm; we refer to these as canopy species. At Barro

500 - --PET - Sherman BCI Cocoli 400

300

-C

E 200

100

0 0 5 10 15 20 25 30 35 40 45 50 week of year

Fig. 1. Precipitation and potential evapotranspiration (PET) at the three plot sites. Both were calculated starting with weekly means over 12 - 25 yr, taking a 5-week running mean at each point, then adjusting to monthly by multiplying the weekly mean by 4.35 (the number of weeks in a month). One station is 6 km from the Sherman plot, the second is 1 km from the BCI plot, and the last is 4 km from the Cocoli plot. PET data are from Barro Colorado Island only. Dry season length was estimated as the interval each precipitation curve spent below the PET curve; for example, at Cocoli, precipitation rises above PET at week 15.8 and falls below at week 49.5, which works out to 129 days below PET. The difference between dry seasons is probably even more extreme than this indicates, because there is undoubtedly greater solar radiation and thus higher PET at the drier sites. (BCI data courtesy of the Smithsonian Environmental Sciences Program; Sherman and Cocoli data courtesy of the Panama Canal Commission.)

Colorado, 139 of the 144 canopy species in the 50-ha (nearlyevery individualwas scored as 3 or 4). plot were censused; at Sherman, 64 of the 66 canopy We tallied the fraction of trees in each category, 0 species were censused; and at Cocoli, 34 of 39 species through4, but we simplify most of our presentationby were censused. The species not censused included two reportingall individualswith a score of 0 or 1 as decidu- palms at BCI and Cocoli (known to be non-deciduous), ous (< 35% of leaves remaining),and all others as non- plus a few species having only a single individual in deciduous. Although this reflects an arbitrarydivision each plot. of a continuousscale (leaf loss) into discretecategories, For each species, three individuals> 30 cm DBH at we believe it most closely reflects everyday use of the each plot were selected at random from the computer termdeciduous. We also reportan estimatetrees scoring database,and each selected tree was visited duringthe 2, to indicate the numberof individuals that were par- last two weeks of January,1997. For species which did tially deciduous. not have threeindividuals, all individuals> 30 cm DBH We report several alternateindices of forest deci- were censused. Subsequentlyat BCI, butnot at the other duousness at each of the three plots: the fraction of two plots, additionalindividuals of common deciduous canopy species that were deciduous, the fraction of species were addedto the census - up to seven trees > 30 individualsin the canopy that were deciduous, and the cm DBH, and seven additionaltrees 5-30 cm DBH. All fraction of crown area that was deciduous. From the selected trees were revisited during the last half of 1997 census, we had to estimate these fractions from Februaryand the last half of March, 1997, and then small samples for each species, using the mean fraction againin September,1997. If a tree was founddead or the of deciduous individuals per species in the samples, crown could not be observed due to lianas, it was re- then weighting this by the abundanceof each species in moved from the list and anotherrandomly selected tree the entireplot. Thatis, let ns be the abundanceof species chosen. s, so N= 5sns is the total abundance of all species, and let Duringthe second half of March, 1999, we extended fs be the fractionof deciduous individualstallied in our this approachby visiting every individualof all species samplefor species s. An estimateof the total abundance recordedas deciduousat BCI or Cocoli duringthe 1997 of deciduousindividuals in species s is thusfn,s, and the surveys (but not at Sherman, where few trees were estimated fraction deciduous for the whole canopy is deciduous). Non-deciduous species were not checked, found by summing over all species. We call this the with two exceptions - Jacaranda copaia and Tabebuia index of deciduousness: rosea - which have been seen deciduous(but not during the 1997 census). All Tabebuia were visited, but we EsL,fs (1) checkedonly 22 Jacarandabecause it is extremelyabun- N dant and most had full crowns of leaves in March, 1999 Deciduousness of crown area was calculated as well, (one of 22 was deciduous). One species, Cavanilesia based on allometricequations relating crown radiusto platanifolia, was excepted from the Cocoli census in DBH. Crown area refers to the area of the horizontal 1999, since it is abundantand always deciduous in the dry projection of each individual crown, based on the as- season.We savedtime by omittingit, andassumed in the sumptionthat crowns are circularso theirarea isp times data analysis that all Cavanilesiawere deciduous.Each the radiussquared. We have equationsfor 81 individual tree visited was scoredfor leaf loss on a scale from 0 - 4: species (Bohlman et al. submitted);for species lacking the allometric a based on datafor 4 = > 95% of leaves remaining; data, generalequation 3 = 65-95% of leaves remaining; all 81 species combined was used. We tallied the total 2= 35-65% remaining; crownarea of each species in the canopyby summingall = 5-35% 1 remaining; individualsalive in the main 1995 census,then used Eq. 1 0 = < 5% remaining. andthe 1997 sampleto get totaldeciduous crown area. In This semi-quantitativesystem is a typical approachfor 1999, since we checked every deciduousindividual, we assessing phenology (Borchert 1980; Bullock & Solis- summedthe crown areaover each thatwas deciduous. Magallanes1990). Yellow or brownleaves were counted In 1999, we had a near-completecount of all decidu- as missing. Leaves were considered missing if there ous individualsin the BCI and Cocoli plots at a single were branchesthat had small branchletsbut no leaves. pointin time. But trees thatwere dead,had badly broken Branches that were clearly rotting or had no small crowns, or could not be observed due to liana cover (or branchlets were considered dead and not counted as occasionally stranglers),cause this to be an underesti- missing leaves. There were undoubtedlya few cases mate. At BCI duringMarch, 1999, 720 trees were vis- where recently dead branchesor even trees were scored ited, but 118 could not be scored (43 definitely dead,59 incorrectly as deciduous, but the September census with lianas and some possibly dead, 16 with broken proves that nearly all trees we censused were alive crowns).At Cocoli, 14 of 72 trees visited likewise could

This content downloaded by the authorized user from 192.168.52.67 on Mon, 10 Dec 2012 14:51:47 PM All use subject to JSTOR Terms and Conditions 652 Condit, R. et al. not be scored. We made a simple assumptionthat the ha); and plateau,which is flat, non-flooded terrain.The dead trees had been replaced by an equal number of plateau was subdivided into three sections: a 6.8 ha others growing past 30 cm DBH since 1995, and that block in the easternpart of the plot thatis > 150 m above these and the ones with liana cover had a fraction of sea level, a 2.1 ha section of young forest, and a 24.8 ha deciduoustrees matchingthe fractionin those observed. westernblock that is < 150 m elevation. The remaining For example, at BCI in March 1999, we scored 265 of 1.8 ha of the 50-ha plot consist of quadrats with a 602 trees as deciduous.To accountfor the 118 individu- mixtureof habitats,and these were not used in analyses. als missed, we multiply 265 by 720/602, giving 317 The habitatsdiffer in soil moisturebecause of a perched deciduous trees. We report this as our 1999 BCI esti- water table that meets the surface along the slopes; the mate and use a similarcalculation for the Cocoli plot. high plateau is furthestfrom the water table and thus Errorrates for the percentageof deciduousindividu- driest (Condit 1995, 1996a, c; Harms 1997). als in 1997 were calculatedin a simple and approximate Species names match those given in Condit et al. way. The total number of individuals sampled at one (1996b). Some of these names have now been revised, plot was taken as the sample size, and 95% confidence for example, the generaScheelia andBombacopsis, and limits on the estimatedpercent deciduous (from Eq. 1) eventually we will update Condit et al. (1996b) with a were takenfrom Sokal & Rohlf (1969). For example, at table showing old and new names for all changes. Here BCI duringthe March, 1997 census, 428 individuals> we use names reportedearlier so thatdata from different 30 cm DBH were censused.The estimatedpercentage of reportson the BCI plot are compatible. the canopy that was deciduous was 6.3 %, and 95 % confidence limits on 6 % of 421 - interpolatedfrom Sokal and Rohlf's appendix - are ca. 4 - 9 %. This is a Results roughapproximation of confidencelimits, since it effec- assumes thatthe was uniform tively sample statistically Deciduous species (that is, that all species were identical). We did not calculate confidence limits on the percentageof crown The fraction of canopy species recordedas decidu- area, since there was no obvious way to do so and the ous followed the precipitationgradient, with the wet site errorwould presumablybe close to thatfor individuals. at Fort Sherman having the fewest deciduous species We did not calculateerrors around the percentof species and the drierCocoli site having the most. At BCI, there because we had complete lists of species in the plots - were 37 species scored as deciduous at least once in they were not samples. Neither did we calculate error 1997:20 thathad at least one individualscoring 0 and 17 for the 1999 census, since it was a near complete count more scoring a 1 but not 0. In 1999, two additional of all deciduous individuals. species were recordedas deciduous, bringing the total An index of deciduousnesswas calculatedfor each numberof deciduous canopy species in the BCI plot to of the five census dates, including only those individu- 39, or 28%of all the canopy species in the plot (Table 1). als observeddeciduous at the time of the given census.In At FortSherman, 14.1% of canopy species were decidu- addition,we estimatedthe index by including all indi- ous, and at Cocoli, 41.2% (Table 1). No 1999 census vidualtrees that were observeddeciduous at least once. If was done at FortSherman, and no new deciduousspecies treesare not synchronizedin leaf loss, this numberwould were added in 1999 at Cocoli. Table 2 lists all the be higher than the fractiondeciduous during any single species recordedas deciduous. census. Forexample, if a thirdof the treesof a species are The 12 most common deciduous species at BCI deciduousin January,a differentthird deciduous in Feb- contributedmore than75% of all the deciduoustrees in ruary,and the last thirddeciduous in March,then we say that 100 %of the individualsof thatspecies were deciduous at least once, while only 33% were deciduousat any Table1. Numberof deciduousspecies observed in eachplot. Thetotal > one time. We could not make this calculationin 1999, numberof speciesis basedon allindividuals 30 cm DBHalive the 1994census at Cocoli,1995 at BCI, and since we had only one census thatyear. during 1996at Sherman. The number of deciduousspecies includes all Finally, the degree of deciduousnesswas calculated whichhad at least one deciduousindividual during 1997 or in differenthabitats within the 50-ha plot at BarroColo- 1999. rado, following habitatdefinitions given in Conditet al. (1995, 1996ac). Each 20 m x 20 m quadratwas classi- Sherman BCI Cocoli fied into one of the the a 1.2 following regions: swamp, Total in 66 144 39 ha thatis flooded with most of the species plot region standingwater No. of speciescensused 64 139 34 year; streamsides,regions within 20 m of small streams No. of deciduousspecies 9 39 14 (1.9 ha); slopes, quadratsinclined by 7? or more (11.4 %of speciesdeciduous 14.1 28.1 41.2

Table 2. Fractionof individualsthat was deciduousduring four differentcensuses (January-March,1997, and March, 1999) at three plots. The column 'any 97' gives the fractionof individualsthat was deciduous in at least one of the 1997 censuses. Only species recordedas deciduousat least once, at one site, areincluded; blanks indicate a species was not presentat the site (or had no individuals > 30 cm DBH). If all entries at one plot are 0.00, the species was present(and censused) at the plot but not observedto be deciduous, while it was deciduousat anotherplot (for example,Andira inermis was recordedas deciduousat BCI, but not at Cocoli or Sherman). The numberof individualscensused is given in parenthesesat each site (in 1997, not all individualswere censused in every month, because some trees were added after January).

Cocoli BCI Sherman Species Jan Feb Mar any Mar Jan 97 Feb Mar any Mar Jan Feb Mar any 97 97 97 97 (N) 99 (N) 97 97 97 97 (N) 99 (N) 97 97 97 97 (N)

Andira inermis 0.00 0.00 0.00 0.00 (2) (0) 0.00 0.00 1.00 1.00 (1) 1.00 (1) 0. .00 0.00 0.00 0.00 (2) Astronium graveolens 0.00 0.00 0.33 0.33 (3) 0.00 (7) 0.20 0.20 0.20 0.40 (5) 0.31 (16) Cavanillesia platanifolia 1.00 0.80 1.00 1.00 (5) (0) 1.00 1.00 1.00 1.00 (5) 1.00 (17) Ochroma pyramidale 0.00 0.00 1.00 1.00 (1) 0.00 (1) 0.00 0.00 0.50 0.50 (2) 0.00 (2) Bombacopsis quinata 1.00 1.00 1.00 1.00 (1) 1.00 (1) 0.00 0.00 1.00 1.00 (1) 1.00 (1) Bombacopsis sessilis 1.00 0.00 0.00 1.00 (1) 1.00 (1) 0.67 0.33 0.33 0.67 (3) 0.00 (3) Pseudobombax septenatum 0.33 1.00 1.00 1.00 (5) 1.00 (9) 1.00 0.80 0.80 1.00 (5) 1.00 (6) Spondias mombin 0.00 0.60 0.40 0.60 (5) 0.88 (8) 0.00 0.83 1.00 1.00 (6) 0.77 (13) Spondias radlkoferi 0.00 0.00 0.00 0.00 (1) (0) 0.00 0.67 0.67 0.67 (3) 0.91 (23) .2 0.2 .0 0.40.... Sterculia apetala 0.50 0.00 0.50 0.50 (2) 0.00 (2) 0.80 0.00 0.00 0.80 (5) 0.00 (14) Tabebuia rosea 0.00 0.00 0.00 0.00 (1) (0) 0.00 0.00 0.00 0.00 (3) 0.61 (18) Zuelania guidonia 0.50 1.00 1.00 1.00 (2) 1.00 (2) 0.50 1.00 0.50 1.00 (2) 1.00 (2) Apeiba membranacea -_-_ 0.40 0.33 0.14 0.57 (7) 0.66 (79) 0..60 1.00 1.00 1.00 (5) Aspidosperma cruenta 0.00 0.00 0.00 0.00 (3) (0) 0..00 0.20 0.20 0.20 (5) ___ .00 1.00 1.00 Cassipourea elliptica -_-_ 0.00 0.00 0.00 0.00 (5) (0) 1. 1.00 (1) Jacaranda copaia ___ 0.00 0.00 0.00 0.00 (3) 0.05 (22) 0..20 0.20 0.00 0.40 (5) Lonchocarpus latifolius -_-_ ___ 0.20 0.00 0.00 0.20 (5) 0.04 (24) 0..00 0.00 0.00 0.00 (1) ___ 0.00 0.00 .00 1.00 1.00 1.00 Trichospermum galeottii -_-_ 0.00 0.00 (1) (0) 1. (1) Virola sebifera 0.00 0.00 0.00 0.00 (3) (0) 0..25 0.25 0.00 0.25 (4) Annona 0.00 --- 1.00 1.00 spraguei ___ (1) (0) tibourbou 0.00 0.33 0.67 1.00 Apeiba ___ 0.67 (3) (3) Ceiba pentandra 0.14 0.00 0.14 0.14 (7) 0.18 (28) ...... -- - - Cupania latifolia ___ 0.00 0.00 0.33 0.33 (3) 0.00 (3) ...... -- - - - ___ 0.33 Dipteryx panamensis 0.00 0.67 0.67 (6) 0.14 (28) ...... -- Ficus 0.25 0.00 trigonata 0.00 0.25 (4) 0.00 (5) ...... -- 0.00 0.29 0.00 0.29 Guapira standleyana (7) 0.06 (35) ...... - - - 0.00 0.00 0.67 Guazuma ulmifolia 0.67 (3) 0.00 (18) ...... - - - - - ___ 0.00 0.00 Hampea appendiculata 0.50 0.50 (2) 0.00 (2) ___------Hirtella americana --- 1.00 0.00 1.00 (1) (0) ...... -- - - - Hura 0.43 0.67 0.67 0.89 0.93 crepitans ___ (9) (67) Myrospermumfrutescens 0.50 1.00 1.00 1.00 (4) 1.00 (4) ...... -- - - - Nectandra cissiflora 0.33 0.33 0.33 0.33 (3) 0.00 (2) ...... -- - - - Ormosia macrocalyx ___ 0.50 0.00 0.00 0.50 (2) 0.00 (2) ...... -- - - Platymiscium pinnatum 0.00 0.00 0.67 0.67 (3) 0.44 (25) ...... -- - - - Platypodium elegans 0.00 0.33 1.00 1.00 (3) 0.86 (22) ...... -- Pterocarpus rohrii 0.00 0.33 1.00 1.00 (3) 0.60 (5) ...... -- - - Sapium aucuparium 0.83 1.00 1.00 1.00 (6) 1.00 (8) ...... -- - - Sapium spec. 0.00 0.00 0.50 0.50 (2) 0.00 (2) ...... - - - Tabebuia guayacan 0.00 0.00 0.25 0.25 (4) 0.71 (17) ...... - - - - - Trema micrantha 0.00 0.17 0.33 0.33 ...... - - - -_-_ (6) 0.00 (5) Triplaris cumingiana 0.00 0.67 0.67 0.67 (3) 0.33 (6) ...... -- - - Zanthoxylum belizense 0.25 0.00 0.00 0.25 (4) 0.00 (65) ...... -- - - - Zanthoxylumpanamense 0.00 0.00 0.50 0.33 (3) 0.22 (9) ...... -- - - - Bursera simaruba 0.67 1.00 1.00 1.00 (3) 1.00 (7) ------_ - - __------Cedrela odorata 0.00 1.00 1.00 1.00 (2) 1.00 (3) ...... -- - - - .....------Dalbergia retusa 0.00 0.80 0.80 1.00 (5) 0.33 (6) ...... -- - - - .00 1.00 1.00 1.00--- Jacaranda caucana 0.00 1.00 1.00 1.00 (1) 0.00 (1) ...... ------Trichilia hirta 0.00 0.67 1.00 1.00 (3) 0.29 (7) ...... - - - Jacaratia spinosa ...... ------1..00 1.00 1.00 1.00 (1) Matayba apetala ...... ------0..00 0.20 0.20 0.20 (5) Schizolobium parahybum ...... ------1..00 1.00 0.00 1.00 (1)

the forest (Table 3). At Cocoli, four species accounted terms of crown area(> 30 cm DBH) at Sherman-Brosimum for close to 73% of the deciduous individuals. At utile (Moraceae),Manilkara zapote (Sapotaceae), Vochysia Sherman, two species accounted for nearly all decidu- ferruginea (Vochysiaceae), and Calophyllum longifolium ous trees (Table 3). At all three plots, canopy dominants (Guttiferae) - were not deciduous. At BCI, the four top were mostly non-deciduous. The four top ranking in ranking were Trichilia turberculata (Meliaceae), Quara-

This content downloaded by the authorized user from 192.168.52.67 on Mon, 10 Dec 2012 14:51:47 PM All use subject to JSTOR Terms and Conditions 654 Condit, R. et al. ribea asterolepis (Bombacaceae), Hura crepitans and Cocoli (Table 4) in 1997. Complete censuses in 1999 Tetragastrispanamensis (Burseraceae);only Hura was gave similarfigures: 11.8%at BCI and 18.2%at Cocoli. deciduous.Even at Cocoli, threeof the top four ranksin Estimates of deciduousnessbased on species, indi- crownarea were non-deciduous-Anacardium excelsum viduals, andcrown area,differ for two reasons.First and (Anacardiaceae), Calycophyllum candidissimum foremost,not all individualsof deciduous species were (Rubiaceae),and Scheelia zonensis (Palmae);the highly deciduous (Table 2, Table 3). At BCI, in the 37 species deciduous Cavanilesia platanifolia ranked second in recordedas deciduous in 1997, we censused 142 indi- crown area at Cocoli. viduals. Of these, only 88 were recordedas deciduousat least once; the other 54 individuals in these deciduous Deciduousness of the canopy species were never observeddeciduous. At Sherman,14 of 28 individuals of the deciduous species were not The estimated fraction of the canopy that was de- recordedas deciduous, and at Cocoli, five of 39 indi- ciduous followed the same trend, with Shermanbeing viduals were not. the least andCocoli the most deciduous.But the fraction Second, deciduous species had different average of deciduousindividuals was much lower than the frac- abundanceand crown sizes thannon-deciduous species. tion of deciduous species. The estimated fraction of At BCI, 26% of the species were deciduous (Table 1), canopy individuals which were deciduous reached a but these accountedfor just 21% of the crown areaand peak of 4.8 % at Fort Sherman,in the February,1997, 16%of the individualsin the entireplot (> 30 cm DBH). census (Table 4). At BCI, the peak deciduousness in Three common deciduoustrees in the BCI plot - Hura, 1997 was 6.3 % (in March),and at Cocoli, 24.3 % (also Cavanilesia, and Platypodium - are large canopy in March).In March, 1999, deciduousnesswas slightly emergents, with immense crowns but few small indi- higher at BCI thanin 1997, and slightly lower at Cocoli viduals in the plot (Conditet al. 1998), whereas several (Table 4), but the complete census closely confirmed of the dominant non-deciduous species have narrow the estimates from small samples in 1997. crowns (especially Trichiliatuberculata). At Sherman, The same trends were reflected by estimates of the 14%of the species were deciduous,and these accounted fraction of total crown area that was deciduous. The for 9% of the crown area and 14% of the individuals; estimated peak deciduousness of crown area in 1997 deciduous species at Shermanall have small crowns, was 3.6 % at Sherman, 9.7% at BCI, and 19.1% at but were numericallyquite common. At Cocoli, 41% of

Table 3. Dominantdeciduous species. Numberof deciduoustrees > 30 cm DBH in the threeplots, duringMarch, 1997 and March, 1999. The total numberof individualsis based on the full plot censuses of 1994-1996. The 1997 estimatesfor deciduousindividuals are based on small samples, extrapolatingto the total numberalive in the full census (from Eq. 1). The calculationfor Hura provides an example: the 50-ha plot at BCI had (in the 1995 census) 77 individuals> 30 cm DBH of Hura; that is not an estimate,all 77 of those trees were tagged and mapped. Duringthe Marchcensus in 1997, we visited nine of those trees, and found six deciduous.The estimate 50.8 is 0.66 x 77. The 1999 figures are actual numbersof observedtrees, with no estimateincorporated. species Family Total individuals Individuals deciduous Individuals deciduous > 30 cm DBH (March 97) (March 99) Sherman Apeibamembranacea Tiliaceae 8 8.0 Aspidospermumcruenta Apocynaceae 34 6.8 -- BCI Huracrepitans Euphorbiaceae 77 50.8 62 Platypodiumelegans Leguminosae 25 25.0 19 Platymisciumpinnatum Leguminosae 33 22.1 11 Dipteryxpanamensis Leguminosae 28 18.8 4 Spondiasradlkoferi Anacardiaceae 26 17.2 21 Cavanilesiaplatanifolia Bombacaceae 17 17.0 17 Apeibamembranacea Tiliaceae 106 14.8 52 Spondiasmombin Anacardiaceae 14 14.0 10 Guazumaulmifolia Sterculiaceae 20 13.4 0 Tabebuiaguayacan Bignoniaceae 19 4.8 12 Tabebuiarosea Bignoniaceae 21 0.0 11 Sapiumaucuparium Euphorbiaceae 8 8.0 8 Cocoli Cavanilesiaplatanifolia Bombacaceae 29 29.0 -- Pseudobombaxseptenatum Bombacaceae 10 10.0 9 Trichiliahirta Meliaceae 10 10.0 2 Burserasimarouba Burseraceae 9 9.0 7 Spondiasmombin Anacardiaceae 11 4.4 7

This content downloaded by the authorized user from 192.168.52.67 on Mon, 10 Dec 2012 14:51:47 PM All use subject to JSTOR Terms and Conditions - Quantifyingthe deciduousnessof tropicalforest canopies undervarying climates - 655 the species were deciduous,but these accountedfor just (thoseare not estimates, but trees actually observed). These 22% of the crown areaand 29% of the individuals.Two represent7.7% and 14.0% of all canopy trees in each large-crowned deciduous Bombacaceae are abundant region(based on the 1995 census). there (Cavanilesia and Pseudobombax),but they have many more smaller individualsthan they do at BCI; in Smaller individuals addition, Cocoli has common deciduous species with small crowns (Trichilia hirta, Bursera; Table 2). In Some trees between 5 and 30 cm DBH were decidu- terms of numericalabundance, Cocoli is dominatedby ous at BCI, but a lower fractionthan among trees larger the non-deciduousAnacardium excelsum. than30 cm DBH. In 23 deciduousspecies, we censused some individuals< 30 cm DBH and some > 30 cm DBH. Partial deciduousness In 16 of these, smallertrees had a lower fractiondeciduous thanlarger trees, while in three species, largertrees At all threeforests, many trees scored 2 on the scale had a lower fraction (four species had equal fractions). of deciduousness, with 35 - 65 % of leaves remaining. For all deciduousspecies combined,88 of 142 individu- The peak fractionof individualsin this categoryin 1997 als (61 %) above 30 cm DBH were recordedas decidu- was 6.2 % at Sherman, 10.2 % at BCI, and 20.7 % at ous during at least one census in 1997, whereas in the Cocoli. At BCI and Fort Sherman,these are higherthan smallerDBH class, 32 of 75 (43%) were. the fractionswe definedas deciduous(scoring 0 or 1, with The total fraction of individuals and crown area < 35 % of leaves remaining;Table 4). The fractionof between 10 and 30 cm DBH that was deciduous at BCI canopyindividuals that were fully deciduous- reaching never exceeded 1%. This is partly because deciduous a score of 0 - was 1.5 % at Sherman,2.9 % at BCI and species have a lower percentage deciduousness in the 20.9 % at Cocoli. Most of the deciduousnessat any one smallersize class. But it is mostly because severalof the time, especially at BCI and Sherman,was in trees which prominentdeciduous species at BCI - Hura, Platypo- only droppedsome of their leaves. dium, Dipteryx, and Cavanilesia - have populations consisting of very few small individuals (this is docu- in detail in Condit et al. Habitats in the BCI plot mented 1998).

The differenthabitats of the BCI plot were not mark- Wetseason census edly differentin deciduousness,with the single exception of the smallpatch of young forest.This 2.1 ha sectionhad In September, 1997, 572 individuals that had been a peak in 1997 of 14.8 % of its individualsand 15.2 % of censusedduring the dry season of 1997 (January-March) its crown areadeciduous. The adjacenthigh plateauwas were revisited. Of these, 34 could not be observed (due 6.7 %deciduous in termsof individuals,9.3 %in termsof to lianas, or having died). Only one individual was crown area. The 1999 count of deciduous individuals deciduous (a Cavanilesia platanifolia), and 12 more confirmedthis difference:45 individualswere observed scored 2 on the deciduous scale. The remaining 525 deciduouson the high plateau,and 16 in the young forest trees scored 3 or 4.

Table 4. Fractionof deciduousnessin the canopy at three plots. All entriesrefer only to trees > 30 cm dbh. The 1997 figures are estimatesfrom small samples, as calculatedfrom Eq. 1; confidence limits (see Methods)are given in parentheses. The 1999 figures are based on near complete samples, but were adjustedupward slightly to account for trees not observed due to liana cover or mortality(see Methods). Sherman BCI Cocoli Frequency % Frequency % Frequency % Individualscensused for deciduousnessin 1997 201 421 69 Individualscensused for deciduousnessin 1999 0 720 72 Totalindividuals in plot 516 100.00 4135 100.00 325 100 Deciduousindividuals Jan97 13.7 2.7 (1-6) 158.6 3.8 (2-7) 42.3 13.0 (6-25) Feb 97 24.7 4.8 (2-11) 201.1 4.9 (3-8) 67.2 20.7 (11-33) Mar97 18.8 3.6 (1-10) 261.1 6.3(4-9) 78.8 24.3 (14-37) Mar99 ------317 7.7 71 21.9

Totalcrown area (ha) 2.570 100.00 28.106 100.00 2.33 100 Deciduouscrown area (ha) Jan97 0.056 2.2 1.450 5.2 0.243 10.4 Feb 97 0.093 3.6 1.928 6.9 0.386 16.6 Mar97 0.066 2.6 2.729 9.7 0.444 19.1 Mar99 ------3.323 11.8 0.423 18.2

This content downloaded by the authorized user from 192.168.52.67 on Mon, 10 Dec 2012 14:51:47 PM All use subject to JSTOR Terms and Conditions 656 Condit, R. et al.

45 -+- Australia 1 - - Australia 2 40 - Panama - -0- Jamaica 35

In 30 ._> 25 o

I 20 8? 15

5 s + - O? 0 I,- I, -- 0 500 1000 1500 2000 2500 3000 annual rainfall

Fig. 2. Percentdeciduous individuals recorded in varioustropical forests. Site 1 in Australiais poor in nutrients,while site 2 is rich (both from Webb 1968). The Jamaicansite is on limestone, andtaken from Kelly et al. (1988); thereis a questionmark near the lower point because no exact figure was given for this site, only that one or two species were deciduous.Webb's data from Australiaare also very poorly documented,with no dates, tree sizes, or sample sizes given. The Panamadata are those presentedin this paper.

Discussion species were deciduous. The gradientof deciduousnessin our study matched This is the most complete, quantitativeassessment the gradientof moisture,as predictedby physiological of deciduousness in a tropical forest that we know of. models, and as observed qualitatively,for example, in Otherstudies have tallied only the fractionof deciduous Africanforests (Hall & Swaine 1981; Veenendaalet al. species in a flora, for example, Guanacastein Costa 1996). Kikuzawa (1991, 1995) showed how a simple Rica has 63 - 83 % deciduoustree species, La Selva has model of leaf value predictsincreasing deciduousness in 15 %, and BCI has been reportedto have 10 % (Frankie forests with longer periods unfavorable for growth. et al. 1974; Reich 1995; Wright 1991). Some tropical Deciduousnessis usually taken to be an adaptationto dry forests have essentially 100 % deciduous species reducewater stress during dry periods in the tropics (Borchert (Bullock & Solis-Magallanes 1991), but others have 1980, 1992;Reich 1984;van Schaiket al. 1993). some evergreenspecies (Sobrado 1991). However, we must consider that some of the gradi- From the perspective of ecosystem function, the ent in deciduousnessin our samplesis caused by differ- percentof species thatare deciduousis not as important ences in forest age, not climate. The Cocoli forest is as the percentof the canopythat is deciduous.This is the clearly a younger forest than the other two, as the index that would reflect seasonality in carboncycling, biggestAnacardiumand Cavanilesia there are not nearly primaryproductivity, or leaf areaindex; it is whatwould the size as those on BCI; it has probablyregrown in 100 be detectablein aerialimages of the canopy (Reed et al. yr or less. In young forest near the canopy crane at 1994; Bohlman et al. 1998). We found that the percent PanamaCity (not far from Cocoli), evergreen species of the canopy that was deciduous was much lower than appear to be replacing deciduous ones as succession the percent of species that was deciduous, at all three proceeds (J. Wrightpers. comm.). On BCI, we estimated plots. At the wettest site, the peak deciduousness of the small patch of young forest in the 5-ha plot to be crown area was 3.6 %, while 14 % of the species were about 15% deciduous, while the old part of the 50-ha deciduous. At BCI, crown area deciduousnessreached plot (> 500 yr old) was 6- 8 %deciduous. Thus, second- 9.7 %,while 28% of species were deciduous,The figure ary forest at BCI is still less deciduous than secondary of 28 % for species is much higher than that given by forest at Cocoli. We estimate that if the Cocoli forest Wright (1991) and Croat (1978), probablybecause we remainsundisturbed for three more centuries,its deci- are only considering large trees; understorytrees are duousnesswill declinefrom its currentlevel close of 20- seldom deciduous (Frankieet al. 1974; Wright 1991). 25 % down to 12- 15 %.These changeswould be caused At the driestplot, 19 %of the crown areabut 41 %of the by a decline in the abundanceof deciduous species,

This content downloaded by the authorized user from 192.168.52.67 on Mon, 10 Dec 2012 14:51:47 PM All use subject to JSTOR Terms and Conditions - Quantifyingthe deciduousnessof tropicalforest canopies undervarying climates - 657 especially Cavanilesiaand Pseudobombax,during suc- we may have missed individuals that were deciduous cession. outside our censuses, our estimates of the degree of Geldenhuys(1993) did a preciselydocumented analy- deciduousness at four points in time are unbiased. We sis of deciduousnessat dry, subtropicalsites in South are also certain that our censuses captured the peak Africa, where rainfall was about 1000 mm/yr. He re- deciduousness at each plot, or very nearly the peak, portedthat 13.8 % of the stems, 27.7 % of the basal area, duringFebruary and March, toward the end of the dry and 28 % of the species were either deciduousor semi- season, when soil moistureis most limiting. deciduous.Kelly et al. (1988) reportedthat 24 % of the What we would like to do as a next step is test individuals, 37 % of the basal area, and 27 % of the whether these differences in leaf loss can be detected species were deciduousin a drytropical forest (1200 mm using remote sensing. Seasonal changes in remotely- rainfall)in Jamaica.In both studies, though,the figures senseddata have been tracedto treephenology (Bohlman refer to the total number of individuals of deciduous et al. 1998) but proving this connection will require species, not the numberactually deciduous at one time. quantitativeestimates of deciduousness. Ultimately, it Webb (1968) reportedquantitative figures for the may be possible to assess deciduousness over wide proportionof deciduous individuals in forests all over spatial scales using remote-senseddata. If so, it would eastern Australia.In Fig. 2, we compare his results as be possible to monitor changes through time in the well as those from Jamaicawith ours. The driest site in degree of deciduousness.Models predictinghow plant Panama is much more deciduous than the Australian functional types change as climates change (Steffen et and Jamaican sites, which are slightly drier. Interest- al. 1996) could be tested at large scales by observing ingly, if we are correctthat Cocoli's forest will fall to 12 shifts in the distributionand abundanceof deciduous - 15 % deciduousness as it reaches maturity (and we species, a majorfunctional type of tropicaltrees. made that estimate before seeing Webb's or Kelly's data), it would align quite reasonably with the high- nutrientsite in Australia.BCI also has a rich soil (Leigh Acknowledgements.We wouldlike to thankJ. Wrightfor 1999), suggesting that perhapsBCI and Australianfor- many comments on deciduousness, and S. de Lao for data ests are similar in their deciduousness-climaterelation. management.S. Patonprovided the climaticdata from the Smithsonianand the PanamaCanal Commission. The BCI Data are far too to this now, but Mackey sparse pursue hasbeen theNational Science Foundation, used Webb's data to construct a plot supportedby (1993) quantitative the SmithsonianScholarly Studies Program, the John D. and model predictingthe occurrenceof deciduousand semi- CatherineT. MacArthur Foundation, the World Wildlife Fund, deciduous forest as a function of climate and soil in the EarthwatchCenter for Field Studies,the GeraldineR. Australia.We are in a position to develop the same sort Dodge Foundation,and the Alton Jones Foundation.The of model in both CentralAmerica and in Australia,once Cocoliand Sherman plots have been funded primarily by the we develop common means for describingclimate. Departmentof DefenseLegacy Program. We wouldalso like An intriguingfeature of deciduousnessin the Panama to thankthe SmithsonianTropical Research Institute, espe- I. for the for two forests is its variationand inconsistency. Most species at cially Rubinoff, supporting largeplots nearly the field workerswho andmeas- all three the were decades,and many tagged plots, including canopy dominants, uredover a thirdof a milliontrees. This is a scientificcontribu- Of the of that were not evergreen. minority species tionfrom the Center for Tropical Forest Science, supported by evergreen,only some individualswere deciduous,and it theJohn D. andCatherine T. MacArthurFoundation, and we was typical for individualsof a species to be deciduous thankI. Rubinoffand E. Losos for consistentsupport and at different times and to different degrees. There were leadershipof the Center. also many individuals in both forest that were partly deciduous (35 - 65 % of leaves remaining),more than those we scored as deciduous. This variationhas been References describedcarefully in individualspecies (Borchert1980) and may be typical of tropical trees in wet seasonal Beard,J.S. 1944. Climaxvegetation in tropicalAmerica. environments,where wateravailability in the soil varies Ecology 25: 127-158. M.O. & D.L. from year to year and from site to site. Bohlman, S.A., Adams, J.B., Smith, Peterson, Because of the variationin the of deciduous- 1998. Seasonal foliage changes in the Eastern Amazon timing basin detectedfrom Landsatthematic ness between andbetween it is clear mapperimages. species individuals, 30: 376-391. that we could have missed deciduous individuals and Biotropica Borchert,R. 1980.Phenology and ecophysiology of tropical that leaves before or between our cen- species dropped trees:Erythrinapoeppigiana O.F. Cook.Ecology 61: 1065- suses. Still, we reiteratethe pointthat what really matters 1074. at the ecosystem level is the deciduousnessat any one Borchert,R. 1992.Computer simulation of treegrowth perio- moment, which is exactly what we censused. Although dicity and climatichydroperiodicity in tropicalforests.

This content downloaded by the authorized user from 192.168.52.67 on Mon, 10 Dec 2012 14:51:47 PM All use subject to JSTOR Terms and Conditions 658 Condit, R. et al.

Biotropica 24: 385-395. mattersto tropical conservationand management.Oikos Box, E.O. 1996. Plant functional types and climate at the 63: 48-61. global scale. J. Veg. Sci. 7: 309-320. Kelly, D.L., Tanner, E.V.J., Kapos, V., Dickinson, T.A., Bullock, S.H. & Solis-Magallanes, J.A. 1990. Phenology of Goodfriend, G.A. & Fairbaim, P. 1988. Jamaicanlime- canopy trees of a tropical deciduous forest in Mexico. stone forests:floristics, structureand environment of three Biotropica 22: 22-35. examples along a rainfall gradient.J. Trop.Ecol. 4: 121- Condit,R. 1998a.Ecological implications of changesin drought 156. patterns:shifts in forest compositionin Panama.Climatic Kikuzawa,K. 1991. A cost-benefit analysis of leaf habit and Change 39: 413-427. leaf longevity of trees and their geographicpattern. Am. Condit, R. 1998b. Tropical Forest Census Plots. Springer Nat. 138: 1250-1263. Verlagand R.G. Landes Company. Berlin and Georgetown, Kikuzawa,K. 1995. Leaf phenology as an optimalstrategy for TX. carbongain in plants. Can. J. Bot. 73: 158-163. Condit,R., Hubbell,S.P. & Foster, R.B. 1995. Mortalityrates Leigh, E.G., Jr. 1999. Tropicalforest ecology: A view from of 205 neotropicaltree and shrubspecies andthe impactof Barro Colorado Island. Oxford University Press, New a severe drought.Ecol. Monogr. 65: 419-439. York, NY. Condit,R., Hubbell,S.P. & Foster,R.B. 1996a. Assessing the Leigh, E.G., Jr.,Rand, S.A. & Windsor,D.M. (eds.) 1982. The response of plant functional types in tropical forests to ecology of a tropicalforest: Seasonal rhythmsand long- climatic change. J. Veg. Sci. 7: 405-416. term changes. SmithsonianInstitution Press, Washington, Condit, R., Hubbell, S.P. & Foster, R.B. 1996b. Changesin a DC. tropicalforest with a shifting climate:results from a 50 ha Mackey, B.G. 1993. A spatial analysis of the environmental permanentcensus plot in Panama.J. Trop.Ecol. 12: 231- relations of rainforest structuraltypes. J. Biogeogr. 20: 256. 303-336. Condit, R., Hubbell, S.P., LaFrankie, J.V., Sukumar, R., Reed, B.C., Brown, J.F., VanderZee, D., Loveland, T.R., Manokaran,N., Foster,R.B. & Ashton,P.S. 1996c.Species- Merchant,J.W. & Ohlen, D.O. 1994. Measuringpheno- areaand species-individualrelationships for tropicaltrees: logical variabilityfrom satellite imagery. J. Veg. Sci. 5: a comparisonof three 50 ha plots. J. Ecol. 84: 549-562. 703-714. Croat, T.R. 1978. Flora of Barro Colorado Island. Stanford Reich, P.B. 1995. Phenology of tropical forests: patterns, University Press, Stanford,CA. causes, and consequences. Can. J. Bot. 73: 164-174. Frankie,G.W., Baker,H.G. & Opler,P.A. 1974. Comparative Sobrado,M.A. 1991. Cost-benefitrelationships in deciduous phenological studies of trees in tropical wet and dry for- and evergreenleaves of tropicaldry forest species. Funct. ests in the lowlands of Costa Rica. J. Ecol. 62: 881-915. Ecol. 5: 608-616. Geldenhuys,C.J. 1993. Observationsof the effects of drought Sokal, R.R. & Rohlf, F.J. 1969. Introductionto biostatistics. on evergreen and deciduous species in the eastern Cape W.H. Freemanand Co., San Francisco,CA. forests. S. Afr. J. Bot. 59: 522-534. Steffen, W.L., Cramer,W., Plochl, M. & Bugmann,H. 1996. Hall, J.B. & Swaine, M.D. 1981. Distributionand ecology of Globalvegetation models: incorporating transient changes vascularplants in a tropical rainforest: Forest vegetation to structureand composition.J. Veg. Sci. 7: 321-328. in Ghana. Junk,The Hague. van Schaik, C.P., Terborgh,J.W. & Wright, S.J. 1993. The Harms,K.E. 1997. Themaintenance ofdiversity in a neotropical phenology of tropical forests: adaptive significance and tree communityof Panama. PrincetonUniversity. Prince- consequences for primary consumers.Annu. Rev. Ecol. ton, NJ. Syst. 24: 353-377. Hubbell, S.P. & Foster, R.B. 1983. Diversity of canopy trees Veenendaal,E.M., Swaine, M.D., Blay, D., Yelifari, N.B. & in a neotropicalforest and implications for conservation. Mullins, C.E. 1996. Seasonal and long-term soil water In: Sutton,S.L., Whitmore,T.C. & Chadwick,A.C. (eds.) regime in West Africantropical forest. J. Veg. Sci. 7: 473- Tropicalrainforest: Ecology and management,pp. 25-41. 482. Blackwell Scientific Publications.Oxford. Webb, L.J. 1968. Environmentalrelationships of the struc- Hubbell, S.P. & Foster, R.B. 1986. Canopy gaps and the tural types of Australianrain forest vegetation. Ecology dynamics of a neotropicalforest. In: Crawley, M.J. (ed.) 49: 296-311. Plant ecology, pp. 77-96. Blackwell Scientific Publica- Wright, S.J. 1991. Seasonal drought and the phenology of tions, Oxford. understoryshrubs in a tropical moist forest. Ecology 72: Hubbell, S.P. & Foster, R.B. 1992. Short-termpopulation 1643-1657. dynamicsof a neotropicalforest: why ecological research

Received 27 July 1999; Revision received 22 February2000; Accepted 14 March2000. CoordinatingEditor: Ch. Leuschner.

This content downloaded by the authorized user from 192.168.52.67 on Mon, 10 Dec 2012 14:51:47 PM All use subject to JSTOR Terms and Conditions