You are on page 1of 10

Forest Ecology and Management 260 (2010) 12–21

Contents lists available at ScienceDirect

Forest Ecology and Management
journal homepage: www.elsevier.com/locate/foreco

Local and regional environmental variation influences the growth of tropical
trees in selection trials in the Republic of Panama
Andrew Park a,∗ , Michiel van Breugel b , Mark S. Ashton c , Mark Wishnie b,1 , Emilio Mariscal b ,
José Deago b , Diogenes Ibarra b , Norma Cedeño b , Jefferson S. Hall b
a
Department of Biology and Centre for Forest Interdisciplinary Research (CFIR), University of Winnipeg, 515 Portage Avenue, Winnipeg, Manitoba, Canada R3B 2E9
b
PRORENA, Smithsonian Tropical Research Institute, 401 Ave Roosevelt, Balboa Ancon, Panama
c
Yale School of Forestry & Environmental Studies, Sage Hall, 205 Prospect Street, New Haven, CT 06511, USA

a r t i c l e i n f o a b s t r a c t

Article history: Native Neotropical trees are being increasingly planted for restoration purposes and timber production,
Received 8 January 2010 but we lack species-specific data on growth responses to different regional climates and local environ-
Received in revised form 22 March 2010 mental variation. We used regression trees and variance components to quantify the effects of within-
Accepted 24 March 2010
and among-site environmental variation on the basal area (BA) of 21 Neotropical and two exotic tree
species at three selection trials in the Republic of Panama. Sites represented distinct regional climates in
Keywords:
which annual rainfall varied from 1100 to 2226 mm, with dry seasons of 4.1–6.7 months. Local environ-
Local environmental variation
mental variables included measures of slope steepness and position, soil texture, soil color, and indicators
Regional climate
Neotropical trees
of soil condition, such as subsoil rockiness.
Regression tree analysis Basal area in 17 species responded primarily to regional differences among sites, and explained between
Panama 32% and 72% of species BA. Low BA plots of most species were located in the driest site, while high BA plots
Species selection trials were found in the two wetter sites (mean BA difference = 117 ± 20.6 cm2 ). Local variables also influenced
the growth of 12 species, with percent slope, soil texture at 5–10 cm, subsoil rockiness and Munsell value
from 20 to 50 cm explaining between 0.2% and 24.5% of within-site variance in BA. For these variables, BA
differences across adjacent branches of regression trees ranged from 59 ± 23.4 cm2 (subsoil rockiness) to
176 ± 45.9 cm2 (Munsell value). Our results support the growing evidence that local as well as regional
environmental variation influences tree community composition, growth and survival in mature forests.
Furthermore, the heterogeneity of responses to local variables among environmentally sensitive species
allowed us to make some preliminary site and species-specific silvicultural recommendations. For site
generalists, future research should extend the current trials to multiple sites within each regional climate
to separate climatic influences from those of the local environment.
© 2010 Elsevier B.V. All rights reserved.

1. Introduction Carpenter et al., 2004). Native tree species are promoted to enhance
rural livelihoods (Arnold and Dewees, 1998; Murray and Bannister,
In the humid and seasonal tropics, the ongoing conversion of for- 2004), restore biological diversity (Leopold et al., 2001; Lamb and
est to farmland and pasture has left a legacy of over 3.5 million km2 Gilmour, 2003), sequester carbon (Silver et al., 2000), and com-
of degraded lands on which natural forest regeneration cannot bat soil erosion (Scott et al., 2005). Renewed attention is also being
occur (ITTO, 2002). Deforestation is particularly pervasive in Cen- paid to mixed-species plantations (Kelty, 2006; Nichols et al., 2006),
tral America, where it is estimated that seven countries lost from raising the prospect that desirable species traits might be combined
7% to 37% of their remaining forest cover between 1990 and 2000 in a single forest stand.
alone (FAO, 2001). In response to this challenge, a renewed interest Some old world tropical trees are widely used in restoration
in reforestation and restoration using native tree species is emerg- and reforestation. Both teak (Tectona grandis) (Healey and Gara,
ing (Butterfield, 1995; Ashton et al., 2001; Leopold et al., 2001; 2003) and Acacia mangium (Norisada et al., 2005) are commonly
deployed in afforestation and reforestation projects across Central
America and Latin America. Despite widespread interest in planting
native species in Central and South America, silvicultural recom-
∗ Corresponding author. Tel.: +1 204 786 9407.
mendations for Neotropical rainforest trees are, for the most part,
E-mail address: a.park@uwinnipeg.ca (A. Park).
1
Current address: Equator Environmental, 250 Park Avenue South, New York, NY qualitative in nature (e.g. Quírico Jimenéz et al., 2002). Unexplained
10003, USA. variations in species performance (Butterfield, 1995), inadequate

0378-1127/$ – see front matter © 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.foreco.2010.03.021

be amplified into Seedlings were planted in three completely randomized blocks major size differences. 1999) affect clay.. 2001. Stanley and Montagnini. personal observation). live crown length. 2009) with a 5. Methods spacing. Soils at Soberania and Los site characteristics. 1996. Tree measurements of Forestry and Environmental Studies. ing. Seed trees were located in relative importance of local (within-site) environmental variation mature forest sites across the Isthmus of Panama.87 ± 0. the native species reforestation project of the Smithsonian Tropical Research Institute (STRI) and Yale School 2. Twenty-one Neotropical and two exotic tree species were tions are therefore needed. semi-dry (Azuero) and dry (Rio Hato) variety of tree species? Secondly.177 for the Each site has distinctive topography and edaphic features 22 species. n = 27) (José Deago. 2004).. Detailed biomass measurements were also in which less than 100 mm rain fell in each month.. such as altitude and slope position (Asner et Santos are primarily clays and silty clay loams. and vegetation varied from low thorny scrub to pasture poorly studied. indicated by erosion pavements covered by small ing sites is a major concern in silviculture (Evans. A. Of from 2000–2004) (Griscom et al. A site at Rio taken from 6 to 12 individuals of each species after 2 years of growth Hato in Panama’s “Dry Arc” received 1107 mm of rainfall annually. 20 seedlings per plot established at 3 m × 3 m 2. At the time of plant- detailed provenance trials (Butterfield. Haggar and 2.. Soberania had not been farmed for at least 10 years. Basic pebbles and a lack of organic matter. Montagnini et al. 2009) All sites were cleared of forest before 1960. Hall et al. Isthmus of Panama (Wishnie et al.. or they may. plots were thinned to 50% of their original density (i. 1999) are common. we explore the relative planted in 2003. We used linear regression to test with a 6. Seeds for the differences in tree size that were mediated by local environmen. and had subsequent or environment by genotype interactions have been observed in histories of grazing or small scale agriculture. and..1. seasonally inundated lowlands. Study sites growth. Davidson et al. ative Saccharum spontaneum. in the case of tural data (Guariguata and Pinard. 2007). Los Santos soils had relatively high total nitrogen (N). local soil management practices.g.. 1998. soils and rainfall on the estab.g. Baker et al. All plots were sprayed biweekly to control insects during the first 2 years of 2. at each site. live fencing and fruit production (Wishnie et al. magnesium (Mg). Rio Hato has flat or gen- potentially impede the widespread adoption of native species. 2009). yielding an average r2 of 0. and was graphic effects connected to soil quality and erosion has also been dominated by a near-uniform cover of the invasive sugar cane rel- found in a few restoration trials (Carpenter et al. 1996). Mg. with a 4. . potential (e. who had an intimate knowledge of the local geog- We sought to answer two principal questions.3.e. nia sp. 1997. 2007). trunk volume Azuero Peninsula occupied an intermediate regional climate with modeled as a cylinder. and was therefore used as our response variable in this (Table 1). log10 BA was consistently the best predictor of log10 season (Craven et al. tions of N. one species over another.. the influence of local low intensity cattle grazing of its understory dominated by Helico- and regional variation in topography. and grasses. nursery staff. tly sloping terrain in which substantial patches have suffered from Achieving an appropriate match between species and plant. and of Rio Hato’s mineral soils ranges from loamy sand through to heavy land use history (Kalinganire. seedlings were acclimated to full sunlight over tistically significant environmental effects may not produce enough 3 weeks by being progressively moved into areas of the nursery that variation in growth rates to seriously affect the decision to plant enjoyed increasing light levels (Wishnie et al. 1999). height.. erosion control). whereas the texture al. we asked whether intraspecific regions between December 2002 and June 2003. K.7-month dry season. Rio Hato had been free of livestock for at least lishment and growth of planted trees in restoration trials remains 3 years. Los Santos continued to experience The above-cited studies notwithstanding. Soberania and Los Santos are characterized by rolling study. and phosphorus (P). 2002. what is the raphy of seed trees of each species. Condit with considerably lower cation exchange capacities and concentra- et al. Urich and Reeder. Seeds of native species were collected by the PRORENA Republic of Panama. In this paper. micro-topography. 1998. of apparent environmental influences on growth. 2003) and the influence of trees on soil properties (e. Lamb and Gilmour. 2007).. These sites were selected to be representative of regional climates and soils across the Field technicians measured basal diameter at the root collar. 2007). Prior to outplanting. Provenances collected at different sites were also allo- cated at random among sites and subplots (three replicates per species per block. and a growth index (GI) based on the sum of a 10-year average annual rainfall of 1800 mm (but only 1467 mm basal area and crown volume as potential predictors of biomass. mesic (Soberiana).. of species in mature tropical forests (Clark et al. The third site at Los Santos in the basal area (BA) derived from basal diameter.. Soberania National Park total height.2. over time. First. submitted). lished by PRORENA. Seedlings were cultivated for 2–8 was motivated by the need to investigate the practical importance months (depending on growth rate) in 125 mm root pruning pots. Svenning et al.. sheet erosion. two exotic tree species in three large species selection trials in the 2007). mined site properties on the growth of 21 native Neotropical and use for fodder. Park et al. Climate sium (K) calcium (Ca). and hilly terrain punctuated by moist swales.1-month dry season basis from 2004 to 2006. even though investigations of survival and growth grasses.. and in the wet versus regional climate (among-site variation) to the growth of a (Chagres). Furthermore..g. Leopold et al. timber value. (Bastién Henri et al..2-month dry these metrics. competition from crown closure. and two crown diameters (along (hereafter “Soberania”) received an average of 2226 mm annual the widest axis and perpendicular to it) of all trees on an annual rainfall between 1987 and 2002. For example. exotics (Acacia mangium and Tectona grandis) were obtained from tal variables were any larger than those among tree species in local suppliers and are the standard provenances used in commer- which no local environmental effects were detected? Question 2 cial plantations across Panama.. Studies to relate tree growth to environmental factors in restoration planta. 2003. nitrogen fixation. (e. / Forest Ecology and Management 260 (2010) 12–21 13 knowledge of species-site relationships. significant clone by site (Naik et al.. 1999. transformed biomass. sta. 2004. potas- the productivity of secondary forests and tree plantations. to 10 stems per plot) to avoid intraspecific Our research sites were three species selection trials estab. After 2 years of growth. P than at the other two sites (Craven et al. Species selection and planting Ewel. 2004). Significant topo. 2003) Los Santos. and a lack of basic silvicul. The and edaphic factors also modify the distribution and growth rates soils at Rio Hato were the least fertile soils among the three sites. Species selection criteria included restoration influence of local environmental variables and regionally deter.

Sample points were arranged using Eq. 2003..141) 0. and Di = the depth value to be inversely correlated with soil organic matter.09 (0.9 (25. (1996).58 (0. 5.65 (2. gullied plots.976) 11.68 (0.66) ExpInd 0. Site Rio Hato Los Santos Soberania N (%) 0.360) K (ppm) 47.5 and 10.957) 3.013) 0.013) P (ppm) 1.73 (1. see also Healy et al.48) Plinth 0.0 (81.09 (0.125) Slopos 4.59) 2373.301) 10. Instead. and most variables. fore grows in inverse proportion to the cumulative depth of rocky We therefore used Munsell-value as a surrogate measure of organic impacts and with the number of impacts achieved. sion trees are often superior to linear models for exploring complex .501) Tex50 11.479) 0.14 A. and the included slope position (Slopos). carbon content in our statistical analyses. microsite descriptions were therefore gathered in each monocul. Tex10.68 (1. Val10. 2 = plinthite at 20–40 cm and 3 = plinthite at represented less than 1% cover. 2008) for use in our predominantly yellow-red tropical soils.5YR.24 (0. observations of soil conditions.225) Mottle 0..150) 7.0 (302. Surfst: surface stones.653) 2. 7.056) 11.92 (0.3 (15.468) 3. upper slope and and therefore were not part of a mixed-species ecological commu- ridge crest plots.13 (0.62 (1.40 (0.40) 58. and plots with high values of SurfSt nity whose members shared common values of the environmental and/or RocInd.978) Surfst 1.54) 3.001).3 (0.4.1.38 (0. Growth–environment the variables collected in the sampled plots. together with standard deviations. Soils were sampled for each species to experience somewhat different combina- to a depth of 50 cm in between 6 and 12 randomly selected plots tions of environmental values.10 Slope (%) 9. Chri represents 8 D i=1 i Munsell-chroma.55.050) 11. where C is the number of points in which rocks were hit by the 1992. Val50: Munsell-value at 10 and 50 cm. 2. Sub. and combined into one index (RRav ) plot to a maximum depth of 50 cm.671) RocInd 1. Regression trees stratification. which corresponds to Munsell- RocInd = × C (1) hues 10R. Acronyms for environmental variables—ExpInd: exposure index. technique developed by Thomson et al. The index there. 1984). Basic topographic data and determined texture using the pipette method (Spearman R = 0.91) Ca (ppm) 599.903) 0.669) 1.36 (0. Slopos: slope position. (1): 4 C 50 × C where Hi = 0.g.482) 1. (2): in four pairs (one in each quadrant of the plot) with each pair being 4 about 1 m apart.893) 2.8 (6.5 ha of a subsample of soils (n = 50) were calibrated against laboratory varied topography and soil conditions. Strata contained between 8 and 51 plots.).16 (1. A 1-m long piece of Hue. 2008). / Forest Ecology and Management 260 (2010) 12–21 Table 1 Summary of environmental characteristics of the three major research sites.0.600) Tex10 6.632) 1.957) 3. we expected the set of monoculture plots plots were within 15–25 m of a sampling point. 2007 found Munsell- rebar.1 (12. and 5 stood for 15% or more.193) 0.008) 0.40 (0. and expressed on a 13-point ordinal scale where 1 is pure sand and 13 represents heavy clay.390) Val50 4.6 (0. Regres- lead author at 0–5.02 (1.4 (52. presence of depth to plinthite (indurated sesquioxide and iron rich horizons) gullies (Gully). relationships (question 1) were instead explored separately for Soil texture was assessed manually (Bates et al. unpublished data).5. surface stones was assessed visually on a 5-point scale where 1 1 = plinthite at ≥40 cm..17 (0. value and chroma were incorporated into a redness rating (RR) 1 cm diameter rebar was hammered into the soil at 8 points in each for each soil sampling depth.0. 30 at Rio Hato.546) Val10 3.13 (0.2 4.28 (1. Mottle: mottle code. inundated bottomlands. and the results of preliminary soil sampling to guide 2.43 (3.88 (0. Park et al. respectively.12) 4. Plinth: plinthite code. (2007).7 (0. Soil nutrient and rainfall data were adapted from Craven et al.04) Annual rainfall (mm) 1107 (56) 1946 (65) 2226 (67) Dry season (months) 6. All plots in each stratum were assigned the average values of was therefore inappropriate for our data. Rainfall figures are averages for the years 1987–2002 (ANAM.2) 1246. lower slope. multivariate ordination) site). that was corroborated at our sites (Spearman R = −0. Statistical analyses geneous sampling units using topography.38) 562.06) 5.0. We therefore stratified each site into reasonably homo.24) 6. Soils data could not be gathered from every one of the 648 tree plots. and V is Munsell-value (Munsell soil color Charts.27 (0. Healy surface rockiness was also estimated in each plot by adapting a et al. 7.11 (0.81.49 (0.5.847) 3. Strata were riparian areas/valley bottoms.5.7 (17.76) 28. Manual texture estimates for The planted area at each site covered approximately 4.920) 4.840) Texture range Loam to clay Silty clay loam to clay Sandy clay to heavy clay RI 3.05 (0.470) 0. We modified a soil redness index (Levula et al.74) 346.001). The data gathered Munsell soil color was also assessed at each depth class.135) 0.2 (5.38 (0. and 9 at Los Santos.07 (0.. Environmental measurements etc. The cover of was recorded as a 4-point plinthite code (Plinth) (0 = no plinthite. finding at which rock was encountered at each point i..58 (1.82) Mg (ppm) 99. 5YR.38 (0.46 (0.86 (0. A rockiness index (RocInd) was then calculated i=1 [(10 − Hi ) × Chri ]/Vi RRav = (2) using Eq. P < 0. 50 is the maximum depth that was probed.22 (7. 50: Texture class at 10 and 50 cm. mid-slope.0 (13.5YR and 10YR. RocInd: sub-surface stoniness index.87) pH 5.76 (0.41 (1.340) 0. 5–10. 2. Tex10. RI: redness index averaged across soil depths.7 5.89 (0.136) Soil nutrient data are based on 45 samples at Soberania. seasonally Species were planted in spatially separated monoculture plots.50 (0.89 (1. ture plot on the three sites (n = 648.80 (11. 2. P < 0. and cover of surface stones (Surfst). 2. 20–30 and >45 cm in depth (Tex5. see Table 1). (2007) and Wishnie et al. 1988) by the each species using regression trees (Breiman et al.66 (1. Ensemble analyses of the pooled per stratum using a 5 cm diameter bucket auger (n = 53–56 plots per species-environment relationships (e.79) 143.. percent slope (Slpct).53) 23. Wills et al. <20 cm).19 (0.

01 is often used to specify when to stop growing the tree. With some minor variation in rankings. 2002) in R version 2. C. 2007). (Mortality = −10. but this Basal area of the pooled species varied by more than two orders rule often results in an overfitted tree with some trivial splits. and was dispro- portionately concentrated in plots of C. and 8 plots sion trees. The most common Site-mediated BAmax − BAmin splits placed low BA plots from Rio Hato on one branch and higher where BAr is the transformed value of basal area for a particular BA plots from Los Santos and Soberania on its opposite (12 species). Two-tailed Type I error rates were calculated to tributions can be relaxed. Each random sample was split into paired variables. BAi − BAmin BAr = (3) P ≤ 0. 106 ± 22.4 to 51 ± 30.87 for Luehea seemanii (mean relative vide a common interspecific standard of comparison for the effects R2 = 0.7 cm2 (G. plinthite code sites. and in 10 of these it was the only explanatory variable (t-test.4 cm2 for Slpct. we used the minimum average error from 100 V-fold species had mean BAs between 445 ± 159. Twenty-two out of 23 tree species produced pruned regres- We also used randomization tests to compare environmentally sion trees with between one and four hierarchical split levels mediated BA differences within species to those among species that (Table 2). At each bifurca. amazonia achieved higher BA in Differences in BAr produced by local environmental variables Soberania compared to plots in Rio Hato and Los Santos. of environment on species size: Site was the most important determinant of BA for 17 species. 59 ± 23. / Forest Ecology and Management 260 (2010) 12–21 15 ecological data.6 cm2 for Tex10.7. 2000). percent slope smallest species were also consistent.2. 1). also the primary factor dividing high and low BA groups in 7 of dom samples of tree species that did not respond to variations in the 12 species that also responded to within-site environmental the local environment. mangium.3). The sizes of random subsamples of environmentally unre. the explanatory variable that minimizes the residual error in the response variable is used to define the split. R2 = 0. Twenty of these tested. Basal area differences produced by these local variables on the opposite branch with Tex10 ≥ 8. a critical alpha of recursively into a hierarchy of paired branches using explanatory 0. Appendix S2). sepium had the largest values of BA in every site. Because sample sizes were relatively small. mangium appeared to exhibit less within. 26 splits were statistically significant (t-test. (Plinth) and average Munsell-values for 0–10 and 20–50 cm (Val10 Cumulative mortality by the end of year 3 averaged 16 ± 23. pyramidale permutations to select the size of the final tree (further details in at Soberania) and 55 ± 18.6.001. For example. All regression trees were generated using the mvpart of stems per plot (not including mortality due to second library (De’ath. that appeared in at least two regression trees. alliodora and D. we calculated a t-test (assuming unequal variances among branches) on the mean differences in basal area (BA) between branches. ranked species (BA = 2 ± 1. Because growth at zero splits. species. Acacia Appendix S1). A regression tree is “grown” by splitting the response variable 1998). Regression trees by Site and by environmental variables. BA was considerably greater at Soberania or Los Santos than branches. cia mangium. (128 ± 51.2% and Val50). All calculations were variables to define bivariate splits in the data. I. BAi is the ith value of BA for that species. cross-validation permutation procedure to estimate “honest” pre- Ochroma pyramidale. Tex10 (6 splits). Results parameter (CP) measures the progressive reduction of heterogene- ity within adjacent subgroups of the data as they are split into 3. Erythrina fusca. Park et al. tion. Corresponding paired ran. For almost every regression tree is therefore “pruned” to produce a tree with fewer species. Cordia alliodora. with Albizzia adinocephala. These top-ranked In our case. Species characteristics smaller subsets on subordinate branches of the tree. in that assumptions about data formats and dis. A. can therefore be included in subsamples were as large or larger than mean BAr differences across the same model (Urban.5. Predictor variables were Site. Table 2. A.1 was used to assess statistical significance.and between-site vari- Pruned trees were created for each species using BA as the ation in BA than most of the other large trees. We also calculated the variance proportions explained 3. Cordia alliodora had a minimum cross-validation error displayed no apparent environmental response. Guazuma ulmifolia to 0. Continuous. panamensis 2. (n = 8).001. trees. Mortality was inversely proportional to BA 2007). Dipteryx face rockiness (Surfst). nested within Site..07 for ranging transformation of BA (Legendre and Legendre.4 cm2 dom samples would therefore satisfy Tex10 < 8 (n = 5) and Tex10 ≥ 8 for RocInd. Appendix S3). and therefore no stable tree could be formed for this rates. whereas Aca- variable BAr takes value from zero to one. Site was were then compared to a null model derived from 9999 paired ran. average mineral soil texture from 0 to 10 cm panamensis and Terminalia amazonia being among the six lowest (Tex10). P ≤ 0. and in Cedrela odorata. Significance of regression tree splits and variance explained at all three sites. The of magnitude within and among sites (Fig. and 176 ± 45. The rankings of the response variable. 1998) to pro. such as those in our data.2. De’ath and Fabricius. sepium at Rio Hato). we used a cross-validation error) for the pruned trees varied from 0.6.9 cm2 forVal50) were comparable in mag- .047.1. odorata – achieved higher BAmin are its maximum and minimum basal area.3 Ln(BA) + 62. RocInd (4 splits) the original regression tree. a split produced by Tex10 and Val50 (3 splits) influenced the formation of two or more regres- might have 5 plots on the branch produced by Tex10 < 8. species.8 cm2 (O. (Slopct). P ≤ 0. respectively. sub-surface rockiness (RocInd). 2000). Relative coefficients of determination (R2 = 1 − minimum fered greatly among species (Wishnie et al. performed in R 2. sur- Astronium graveolens. Pachira quinata and Gliricidium diction errors for trees of different size (De’ath and Fabricius. slope position (Slopos). subsamples based on the values of the environmental variables Seven local environmental variables contributed 26 splits to the that produced the observed BA differences in the species being branching hierarchies of 12 pruned trees (Table 2).2 (R Development Core Team. The branches to be pruned are determined using a V-fold in Rio Hato. A complexity 3. and T. and categorical test the null hypothesis that mean BAr differences between random variables. stem and crown morphologies and potential final size dif. aromatica and Inga punc- For each bivariate split produced in the pruned regression tata at Rio Hato and Los Santos. 1999. punctata. Appendix S3). Copaifera aromatica. year thinning). ordinal. branches of the empirical regression tree (Legendre and Legendre. and BAmax and Two species – Albizzia guachapele and C.1.3 cm2 ) in at least two of three average redness rating across all soil depths (RRav ).42 ± 0. average mineral soil texture from 30 to 50 cm (Tex50). A CP of 0. The transformed BA at Los Santos than in Rio Hato and Soberania. sponsive trees were made identical to those from the branches of Four variables—Slpct (10 splits).2.

Species acronyms are the first four letters of the genus followed by the first two letters of the species name (e. C.3) on gentle slopes (Slpct < 13. local environment 3. response to slope conditions at both sites. and reached its maximum BA on steeper slopes with dark soils at both Soberania and Los Santos. These results confirmed that the effect sizes contingent upon the values of variables that determined higher produced by local variables on the BAr of environmentally sensi- level splits. C. even in (BA = 355 ± 17. saman and 24.5). 1. P ≤ 0. achieved lower BA under RocInd ≥ 0. and Tex10 explained 22. we distinguished five groups of species. Low BA plots of A.8–19. BA in Enterolobium cyclocarpum responded to 4. BAr differences 126 ± 31.16 A.1 ± 6. but to Slpct at Los Santos and Soberania. Permutation tests Environmentally mediated differences in BAr from empirical Based on the combined results of exploratory regression trees regression trees were larger than those produced by the equiv. For example. was smaller on gentle 3.1. although it did not contribute to Terminalia amazonia had opposite responses to Slpct at different the regression tree for the latter species. small T. and variance analysis. odorata. alent null models for 22 of 26 environmentally mediated splits characterized by different responses to between. / Forest Ecology and Management 260 (2010) 12–21 Fig. Three out of the four exceptions variables.05). vary significantly from their randomized equivalents from 12 of tertiary and quaternary divisions of a regression tree were often 19 permutation tests. and Rio Hato.14% for textured soils (Tex10 < 7. Relative importance of site vs. the site where it was grown. Tex10. Park et al. regional differences in Site.5% of total variance in E. however. rosea were within-site generalists.2% in L seemanii. However. and RocInd sites. but lower BA also contributed substantial (10.6% in A. and Val50 – were nested within Site in to calculate BA (73 ± 69.98% for Tex10. Val50.and within-site (P ≤ 0.1 on 9999 permutations).9 ± 6.86% for Val50.2. mangium on coarse.25) than on explained variance.1. glan- to this general pattern also failed to produce significant t-tests for dulosa. as were plots of I. 5. having higher BA on gentle slopes at Soberania. and 3. responding to mean BA differences observed in the empirical regression trees. or (Table 3). but was larger on clay loams and silty clay loams portion of variance in BA (mean = 60 ± 10. Acacma = Acacia mangium. 8. Tectona grandis had lower Tex10. and T. 4. Ochroma pyramidale. S. suggesting a common Variables producing two or more regression tree splits – Slpct. fusca. Slpct explained 24.0 cm2 ) on coarse-textured soils (Tex10 < 6. graveolens.3 to For species that responded to Site alone. Finally.1 ± 4.4 cm2 . C.29 with Slpct ≥ 13. nitude to those produced by Site.50% for Slpct. candidissimum. measured across branches of the empirical regression tree did not The responses of species to variables that produced secondary.2%) variance components to on gentle slopes on the tertiary splits from RocInd at Los Santos the total BA variance in several other species (Table 3).5 (Table 2). Albizzia guachapele. those species whose regression trees had no site-mediated splits The direction of a species’ response to a given local variable.g.2 cm2 ) than on heavy clays (BA = 217 ± 36. on the other hand.75) were all located at Rio Hato.6 ± 4.1 cm2 ).85. and species measured under the same conditions. Plots of RocInd. Site consistently explained the largest pro- fine-textured soils. Ochrpy = Ochroma pyramidale). but not to local environmental variation . grandis on coarse-textured soils were also located at Rio 23. Environment nested within Site generally explained a even the identity of the variable responded to. RocInd.1% of BA variance in Hato. Discussion Tex10 within Rio Hato.2%. punctata with low BA and Tex10 < 11. Variance components slopes. could also depend on relatively small percentage of total variance: 7. Variability of basal area among individual plots for 23 tree species.3. which ranged from 89 ± 30. Samanea saman responded positively tive species were significantly greater than those of Site responsive to darker soils (Val50 < 3.

subsoil O. 2000). The second group comprised of Acacia mangium. RocInd: subsoil rockiness index. A. grandis were those that had their smallest BA plots on the the general observation that T. Expressed on a per and which may have limited the growth of many species.4 m2 ha−1 Low BA plots of T. icantly depress growth in these species. cyclocarpum. and S. L. At this site. **≤0.1. Sob: Soberania. Statistical significance of the splits. / Forest Ecology and Management 260 (2010) 12–21 17 (Table 1). Tex10: Mineral soil texture between 5 and 10 cm depth.5 ha site. Park et al. . ***≤0. (E. grandis grows optimally on rela- clay loams and loams that were almost exclusively a localized fea. assuming unequal variances between groups. E.65—pure clay). Park. as determined by t-test. mombin displayed BA differences and surface plinthite. fusca) across a 4. conditions that did of 24–558 cm2 in response to local environmental variables that not exist at Soberania or Los Santos (A. Lower slope positions at A third group.01. cyclocarpum) and 31 m2 ha−1 (E. these BA differences would be between 1. Slpct: Percent slope. seemanii. but grows poorly in soils may have augmented the effects of the dry climate to signif. dense clays and shallow soils (Weaver. respectively. pyramidale. I. saman. punc. hectare basis. and erosion pavements. RRav: average redness index. fusca. tively fine soils that do not suffer restricted drainage or excessive ture at Rio Hato.001. grandis were also found on the finest tex. A. LS: Los Santos. RH: Rio Hato. personal observation). Val10 and Val50: Munsell-value between 5–10 and >45 cm. also explained significant variance proportions. *≤0. excessive drainage of coarse-textured resistance to water uptake during dry periods. Rio Hato were also characterized by high subsoil rockiness. on sandy clay loams and silty clays (Table 2). S. with the best growth being Species from this group were particularly sensitive to variations in Table 2 Pruned regression trees for 20 Neotropical and two exotic tree species. graveolens. These findings reflect tata and T. tured soils (Tex10 ≥ 11. E.

fusca achieved higher BAs on gentler slopes. at Los Santos were located on a seasonally flooded stratum where dark soils (Val50 ≤ 3) penetrated to depths of at least 50 cm (A. but at least one local variable explained significant vari- (Table 2). fine-textured soils at Soberania and Los Santos.2. the location of high BA plots of E. but these variables explained rela- tion on these particular microsites. Comparisons with other studies personal observation). G.18 A.62. rockiness and percent slope. pyramidale for relatively steep PRORENA plantations was consistent with findings from studies microsites with darker soils. P ≤ 0. non-significant variance proportions in these species. Park. D. saman also grew well ance proportions. By contrast. panamensis. saman with high BA tively small. and P. as well as the affinity of O. Between 7% and 25% of species in large forest plots . infiltration rates. may have reflected a preference for relating species abundances to environmental variation in mature better drained microsites. Healy et al. Park et al. fusca on steeper slopes with coarse-textured The proportion of environmentally sensitive species in soils. Finally. and loss of water tropudalph soils were prone to waterlogging. cycloparpum and E. amazonia separated on dark. Astronium graveolens. Regression trees for five other species – A. E. quinata – were defined by also usually lower slopes (Slpct/Slopos correlation = −0. 4. seemanii and S. perhaps into low and high BA subgroups based on different values of subsoil reflecting superior nutrient availability or better soil water reten. which were matica. (2008) tropical forests. to lower slope positions. and we hypothesize found that plantation microsites on steeper slopes were more pro- that their growth may have been affected by the balance between ductive than those on gentle/lower slopes at a site where the aquic water inputs from upslope. Plots of S. C. aro- and E.01) Site alone. Echoing these results. G. adinocephala. sepium. / Forest Ecology and Management 260 (2010) 12–21 Table 2 ( Continued ) percent slope and shallow mineral soil texture. High BA plots of L. ulmifolia and T. cylindrocarpum.

4 0.2 7.2 3. These size differences some- eral nutrients or dynamic movements of soil water.2 1.3* 7. and had the potential to such variables is demonstrated in studies where both descriptive produce commercial or practical consequences for plantation own- and more detailed physical and chemical measurements are taken.1*** 1..2* 15.0 1.5 28.2 14.01.2 17.7 16.0 6.1*** 15.4 Erythrina fusca 49.4 5. ers or small farmers planting individual trees.8 9.7 15.1*** 7.0* 0. 2004).5 23.1** 0.5 Calycophyllum candidissimum 71.5 Terminalia amazonia 62.8 Luehea seemannii 57.0*** 0. weak relationships between species and topography in primary for- est (Valencia et al.6 7.1 Tabebuia rosea 59.8 16..1 3.1*** 0.5 3. 2002).9 6. . ated with variations in soil nutrient concentrations. Species Site Tex10 Slpct Val50 RocInd Residual *** * Acacia mangium 64.4* 0. soil matric potential on Barro Colorado Island corre- responses to local variables implies that species-specific planting lates with slope position (Daws et al.6 1. The heterogeneity of For example.5 15. eroded slopes.1 Samanea saman 48.1 0. Panama (Healy et al.2 Spondias mombin 69..3 19.4 2.3*** 1. Other studies have found plantations.7 4.4 Albizia guachapele 62. and high rainy season prescriptions will be required to locate environmentally sensitive seedling mortality may be linked to wet soils or enhanced pathogen species in microsites where they display superior growth.7 Copaifera aromatica 48.3 1.8 1.4 0. soil texture and rock 4.9 Ochroma pyramidale 62. used carefully stratified experimen- three large (>25 ha) Neotropical forest plots were spatially associ.6* 7.9 12. John et al.4 1.3 0.7 Enterolobium cyclocarpum 69.4*** 24.3*** 2. but improved growth under amendments of NH4 .9*** 5. and PO4 .1 0. Park et al.7 Pachira quinata 65. 2004).4 5. that local variables can have important effects on tree growth in specific combinations of local variables.1 4.6 16.6 10.0 10. The value of times rivaled those produced by Site.3 18. seemanii on dark colored soil Table 3 Variance components for environmental variables (nested within “Site”) that produced more than one split on different regression trees.4 Cedrela odorata 70. shallow mineral soil study are often used as surrogates for complex distributions of min- texture.1 1.5 3. tal blocks to demonstrate decreased growth of T amazonia on steep.. In a site restoration 2001). or with gap size.9 24.3 10. soil moisture and nutrient variables explained on steeper sites with darker soils.2 13. In this study.1 1. at Sardinilla.0* 7.2 4.5* 8. 2008).001.1 12.9 35.2 1.6*** 5.6 2.0*** 7.5 22.4 14. 2005).0** 0. we could recommend prescriptions that place O..3 Gliricidia sepium 76. L.1*** 9..4 2.0* 17.5*** 12. An ensemble ingly.5 2. 2005).1 17. rockiness and Munsell-value. We have demonstrated that substantial size differences can be The descriptive and semi-quantitative variables used in our associated with variations in percent slope.6 2.6*** 22.0*** 10. The last two studies support our general conclusion at each site.5 Astronium graveolens 32.0 1.8*** 3.6*** 9.2 23. A.2*** 7..8 Colubrina glandulosa 67. pyramidale of 18 topographic.1.5 Tectona grandis 60.3.4* 2.0 3. 1999.3 12.5 Albizia adinocephala 48.5 7.7 5. underscoring the potential influence of unique. Harms et al.3* 5.1 11.5 5.6** 5. *** Significance value of the associated F-test is <0.9 2. / Forest Ecology and Management 260 (2010) 12–21 19 Table 2 ( Continued ) in Central America were non-randomly distributed across different about 20% of the productivity variance in a 5-ha biodiversity trial soil types or topographic positions (Clark et al.2 3. nutrients that appeared to determine species niches were different NO3 . Accord- activity in these wet microsites (Daws et al.0 Dipteryx panamensis 52.0* 6.6* 24..0** 1.4 Inga punctata 60.9 5.9 1. (2007) report that 36–51% of all tree species at study (Carpenter et al.6 8.4 7. ** Significance value of the associated F-test is <0. site. Silviculture and recommendations outcrops in logging gaps (Park et al.4*** 2.7 6.0 Guazuma ulmifolia 56.3*** 4.7 12.9*** 7.0 11.2*** 0.8 * Significance value of the associated F-test is <0.

J. Although our results provide initial indications of envi. 21– production or for soil restoration. in planting arrangements and scales of operation. T. Future silvicul. A.. Dolling. 79–90. saman and S.. Forest Ecology and Management 238.. but to harvest it after the first few years of Carpenter.. Classification and Regres- mangium is highly invasive on some sites (Daehler. program of the School of Forestry and Environmental Studies. exotic grass. was funded by NSERC.. Dewees.S. and a startup grant from the University Environmentally sensitive species are used for timber production of Winnipeg. Craven. tural research on environmentally sensitive species may therefore be focused on improving and refining our knowledge of their References microsite preferences under a variety of operational conditions. rather than low rainfall per se. Yale Forest Ecology and Management 112. Guelph. G. graveolens. Clark. intimate mixtures. J. Boshier. tion have wide climatic tolerances. D. Swaine. It seems Panama. staff and students of PRORENA. B..R. Chave.. leaf litter and seed size on seedling demography nutrient status of soils at Rio Hato. Envi- ronmental and biotic controls over aboveground biomass throughout a tropical The last recommendation may be particularly relevant to Rio rain forest. F... D. E. 1980). Arbores de Centro America CATIE.K.B. drainage and nutrient effects in topographic sequences.. Hato. 1). Biological Conservation growth. van den Broek. Ecology 80. C. 84.. 3178–3192. The taxonomic distribution of invasive angiosperm plants: eco- generally degraded condition of the site had most influence over logical insights and comparison to agricultural weeds. Hubbell. 1996. Funding to estab- sonally flooded dark grey soils suggests that this species may be lish and maintain PRORENA is generously provided by The Frank particularly suited to planting on flood prone land... Villa.. Perspectives in Plant Ecology. site restoration (e. C.R. B. and the Smithsonian Tropical Research Institute’s Cen- cyclocarpum and E. Ecosystems 12.. Muller-Landau. R. Each of these end uses demands different Supplementary data associated with this article can be found. 2003. 409–430. mangium achieved consistently high growth rates at Rio Hato. Mullins. 2005. M. 1998.. K. 2005). Dent... D. pyramidale and L. Ecological knowledge of regeneration from We thank the researchers.. and Systematics 6. Berlin. M. Wadsworth International Group. Food and responses to different classes of microsite and a variety of regional Agriculture Organization of the United Nations Excel Spreadsheet. T.. Evolution. Early survival.. 1984. Cordero. Dolling. Classification and regression trees: a powerful yet extension of selection trials to a wider range of local conditions simple tool for ecological data analysis.. saman – receive as little graphic position affects the water regime in a semi deciduous tropical forest in as 600 mm of rain per year in parts of their native range. and planting in micro. I.. 2007). mangium as part of for reforestation. M. Valencia.. University. M.foreco.. M.. Singhakumara. 261–278.. may therefore have the potential to act as a nurse tree for slower 1988. Daehler. cylin..03. Rome. G.P. A. S. Hughes. J. Foster. Loses. Butterfield.. G. 93– may have produced the poor growth performance of native species 105. ulmifolia. Kennedy-Bowdoin. G.. O. nature. R. Forest Ecology and Management 75. amazonia). B.). The size achieved by site generalists resulted from the combined Costa Rica. Hato might be focused on improving soil structure and infiltration Clark. P. S. experiment and for collecting tree growth data.. Early growth of native and exotic trees growth. R.B. L.. (e. (Eds. Between effects of the regional climate and local conditions.S. Important issues to study include the relative importance of soil Arnold.. seemanii) or for fodder. except that cli- and within-site comparisons of structural and physiological characteristics and mate overrode any local environmental effects (Fig. Plant Ecology 179. J. The high growth of S. D.20 A.P. fusca on shallow slopes with coarser textured.F. D. Daws. compatible data sets may also assist in efforts to generalize species FAO. Gunatilleke.. caution should therefore be exercised in its deployment. where poor growth of both locally sensitive and site gen. Friedman. R. ulmifolia. USA.E. sepium and possibly P. New Forests 17.I.R.. 167–180.. Daws. catchments. D.F. Mullins.G. 2662–2675.W.D. Pearson. ulmifolia.. Ontario Institute of Pedology/University of Guelph..E. M. C. 1998. STRI. Yale University.. fertile site and a dry.. (Cordero and Boshier. Field Manual for Describing Soils. R.. Burslem.W... E. Topo- drocarpum. and within regional climates is to be recommended. and forested riparia. Knapp. T. Restoration path- Researchers may also want to investigate whether apparently lim.P. Andrew Park’s research is relevant to ecological..P.. Alternatively. Jones. 117–126. Some species – A. Ashton. ways for rain forest in southwest Sri Lanka: a review of concepts and models...g. H. 1998.. CA. Fabricius. Early species selection for tropical reforestation: a consider- an intimate mixture with O. growth of rain per year with dry seasons extending from 0 to 6 months and foliar nutrients in native Ecuadorian trees planted on degraded volcanic soil.I. Varga. 2001. Aguilar. Variation in tropical forest eralist species makes it hard to justify planting for either timber growth rates: combined effects of functional group composition and resource availability. growing under 1000–3000 mm Davidson. M. Olshen. B. Breiman. for designing and setting up this probably well-drained soils. Beta-diversity in tropical forest trees. G. agricultural and commercial objectives. Pinard. in a semi deciduous tropical forest in Panama. Nunez. Supplementary data and other on-farm uses (e. the Grantham Foundation and the School of Forestry and The environmental stratification of sites recommended here Environmental Studies. and great sion Trees. R. 2004.. 666–669. R. G. 2009. Evans. T. N. 2004).. 2000. 1998). ulmifolia.. Turkabad. 111–121. Griscom. Forest regeneration from pasture Acknowledgements in the dry tropics of Panama: effects of cattle. 161–188. 1999. graveolens. / Forest Ecology and Management 260 (2010) 12–21 horizons on shallow (lower) slopes. 2002.1016/j. it is hard to know whether the dry regional climate or site in Panama. J. T.fao.. 2001. 1998. C. Burslem.. Global Tables in FRA-2000 Summary Report: Overview.D.. Ecology 83.. Leigh. Multivariate regression trees: a new technique for modeling at this site.2010. fusca. 1995.P. Griscom.g. Paton. but A. Science 295. S.H.g. the online version. On-farm trials elsewhere in the region suggest that Condit.grandis and T.. L.. Nichols.A. live fencing Appendix A. Palmer. M. at doi:10.. H.. 335–346.P. Wishnie. A. Meta-analyses of significance. G.. quinata and G.C.. D. especially water availability.P. E. 2007.. De’ath. Restoration Ecology 17. Baker. 1–19. http://www.021. H. G. G. saman in sea. J. therefore. Promoting biodiversity: advances in evaluating native species ter reforestation strategy might be to plant A. breaking up surface hard pans. Y. At Rio Hato.. 2002. these amendments have the potential to improve soil properties. Only the nitrogen fixing exotic.C. Pitman Jr. distribution of tropical rain forest trees. Effects likely. dation. 1999. Evans. M. ronmental sensitivities among tree species in central Panama. Natural Resource Perspectives 26. D. mombin) (Wishnie et al. 36..R. iting site conditions could be ameliorated by planting species in Forest Ecology and Management 154. efforts to improve tree performance at Rio planted on degraded tropical pasture.org/DOCREP/MEETING/003/X9835E/X9835E00. Stone. P. Gunatilleke. A bet. Hernandez. the De’ath. Burslem. 1105–1117. species–environment relationships. G..H.. Planted forests of the wet and dry tropics: their variety.A. ter for Tropical Forest Science. Forest Ecology and Management 196. C. E. D. 367–378. 2008. S. infertile however. gravelolens.. Gagnon.L. 9 pp.B. Butterfield. Forest Ecology and Management 81.R. 2009. Pierpoint.. Unfortunately. J. Braden. a conclusion Levinson Family Foundation at the Peninsula Community Foun- that is supported by expert opinion (Webb et al. Cameron. Forest Ecology and Management 105. E. R. M. M.G. G. Belmont. Edaphic factors and the landscape-scale by mulching. Ashton. Belisle. structure. Molfetta. Plant and Soil 238. Terborgh.F.E. pyramidale. Sandi. that the generally degraded condition and poor of topographic position...P.... Most of the trees for which we have silvicultural informa. 2002.E. .. D. ation of stability. J. 25–36.A.. Rethinking approaches to tree management by farmers.H. 2004... Ecology 81. Ontario..W. D.W. a joint seed in Neotropical forest trees: implications for natural forest management. J.R.B.. Guariguata.. Asner. Ashton. E. 87–99. Park et al. growing native species (Norisada et al. R. and Bates. foliar nutrient content of 14 tree species at a wet. sepium. and S..HTM. D. on November 10th. Accessed environments.

Proceedings of the National Svenning. V.. R. / Forest Ecology and Management 260 (2010) 12–21 21 Haggar. Stallard... D.S. J. Classification and regression trees. M. Journal of Ecology 89. 2007. J. UK. the humid tropics of Costa Rica. 2005.H. 91– S. Development Association Tropical Forestry Paper 15.S. Fredericksen. Silva Fennica 37.B. 2003. using field and laboratory measurements of soil color. J.. R... ITTO Policy Development Series. Tange. R. Lowe. 498–510. Austria. Tree species distributions and local tree species composition in a Scots pine—Norway spruce stand in southern habitat variation in the Amazon: large forest plot in eastern Ecuador. Hitsuma. W.A. Gregoire.. 2001.. Journal of Finland... 2004.L.. Forest People in the Humid Tropics: Past. The hydrological and soil impacts of Healey.. pp. Assessment of morpho. 497–507.. 394–407. through reforestation of abandoned tropical agricultural and pasture lands. Ashton.. Urban. London.V. M. wet tropical forests in Costa Rica. M. Osterlag. C.F. Foster. forestry Systems 61..L.. Bhargava. Habitat associations of trees Vienna.E. 2006.). Magård. R. L. Prediction of soil organic carbon content and challenges. F. S.. L. S.J.A.J. Thomson. The role of species mixtures in plantation forestry. A. 1998... ITTO. 147–157. Forest Ecology 1698–1715. Commonwealth Forestry Institute and the Overseas logical and genetic diversity in Gmelina arborea Roxb.. P. Stanley. Valencia. Ecology 92.. M. M.. T. Forest Ecology and Man. G.. D. Management and Rehabilitation of Restoration Ecology 8..E.. C. 243–249.. H. Forest Ecology and Management 142.. Yavitt. R.. puting. 232. Ecological determinism in plant community structure across a tropical forest Kalinganire.P. Weiblen.L. H.. K. 222– WWF/IUCN. Bioecología tree species used in silvopastoral systems in the humid lowlands of Costa Rica.. Ecology 85.A.. Journal of the Soil Science Norisada.fed.. R. D.. Initial performance and reforestation potential of 24 tropi- 51. Habitat characteriza. Royas.F.. 1997. Mackensen. D. (Eds. Árboles Mader- tropical ecosystems. Kuroda. environmental heterogeneity for productivity and mortality in a tropical tree Silver. A. gophers and rocks.. Río Piedras. Bristow. 2008. 2010. 493– Leopold. Mixed-species plantations: prospects Wills... 2007. Performance of Grevillea robusta in plantations and on farms landscape.f.C.. D.B. Mariscal.. Cedenño. J. tions underestimate the role of edaphic factors controlling the distribution of R Development Core Team.G.. V.. Gara. E. Gilmour. McKenna. Condit.E. S.. Balslev... Park et al.. N. a nurse tree candidate Wishnie.A.. M. D. R... D. J. C. Andrus. L.. 2004. multiple factors in spatial distributions: lilies. Present and Future Hydrological Research for Ecology and Management 176.us/global/iitf/native. Valencia. D.. S. J. Forest Science Ashton.B.B. Professional Geographer 51.. G. Loses. Alfaro. Forest Ecology and Management 233. ence spatial distributions of tropical tree species. T. Forest Ecology and Management 217. Switzerland. J. . Bannister. In: Francis. Forest Service.. No 13. 1999.J. Academy of Sciences of the United States of America 104. P. 1211–1221. Degraded and Secondary Tropical Forests.. 2005.B. (Eds.. Finkeldey. 2000.M.L. y Rodriguez.. S.. Analysis of Ecological Communities. Biomass and nutrient accumulation in pure nization.F. International Institute of Tropical Forestry. 2526–2538. Vartak. Kelty.. Lamb.. Stallard.B. Bruijnzeel. Partitioning the effects of biodiversity and Cambridge. 279–285.. ronmental constraints and opportunities. 380–388. a 20-year venture in income-generating trees and hedgerows in Haiti....S. Scott....V. Condit. Rojas. 2001. 947–959. pp. 2171–2183.D. Montagnini. and mixed plantations of indigenous tree species grown on poor soils in John. A..K. R. M. Hernádez.. Romoleroux. H. Rehabilitation and Restoration of Degraded Forests.. Relation between soil properties and K. Integrated Land and Water Management.. G. Wright. D..J. 2006. 1999. Navarrete.. Vanclay. R.. cal tree species planted across a precipitation gradient in the Republic of Panama.htm Accessed on January 7th.. R: a language and environment for statistical com- Entandophragma. Dolling. M.. 2002.A. Peasants. C. mental relationships of tree species in logging gaps in a Bolivian tropical forest. 195–204. 2007. Kojima. K.J.A. 2003.. S. P. D. Deago. Thomson. Potvin. J. M. Park.. J. The effect of a teak (Tectona grandis) plantation on the forestation in the tropics. Legendre. Acacia mangium. Justiniano.. Harms. T.G. 39–49.. 2004. Legendre. Montagnini. J. agroforesters. J. 1996. L. Agro. Condit. R. Svenning.S. Villa.F. B. Ecology 77. Foster. M. Instituto Nacional de Biodiversidad Hall. J. Reeder. 2003. Healy. P. Attempting restoration of 506.. G. Growth characteristics of some native Weaver.M.. Legendre. Yamanoshita. J. http://www. Hubbell. P. L. 622–651.. 2003. Westman. S. Ecology 85. Grace. Knowles. MJM Software. Kinner. Elsevier Science B.J. under varying environmental conditions in Rwanda. for reforestation on degraded sandy soils in the Malay Peninsula. Hubbell. Plantation forestry in tropical limestone uplands: envi- dam. Amster. J. 2000. The potential for carbon sequestration plantation. 1996. In: McCune.. R Development Core Team. 383–397. Numerical Ecology. and Management 233. Cambridge University Press/UNESCO.A. Society of America 71. Tectona grandis L.J. 99–115. A.. 214–229..H. P. R. Paranjpe.. P.P. 864–869. 2007. Ugalde. T... Natural regeneration and environ.. Soil nutrients influ. D.. P. R. and shrubs in a 50-ha Neotropical forest plot. 1980. In: R Foundation for Statistical Computing. J. M. Sasaki. 2004. H... Lugo. J..W. W.. Gland.. N. Water and establishment of native species in an abandoned pasture in Costa Rica. C. Primary productivity and resource partitioning in model Quírico Jimenéz. Vallejo. F.. 2002. UK.. C..C. A.. ables de Costa Rica: Ecología y Silvicultura. 103.. Nuyim.. R..K. Gotelli. 1999.. Webb. In: Bonell.P.E. Sub-tropical Plantations. B.. Navarro. Forest Ecology and Management 113. Yagi. A Guide to Species Selection for Tropical and Naik.D.M.J.).fs. ITTO Guidelines for the Restoration.. J. K. Smith. Ilvesniemi.. 163–170. A. and anthropologists: Rico. Untangling agement 80.A. International Tropical Timber Orga.. Urich. 903–913. (Eds..J. C. T. USDA Forest Ecology and Management 59. Dent. R. Puerto Murray. Forests. K. Levula.. New Forests 38.... F. Mirabello. Harms.. S..B. Wood. 205–218.. 383–390. Ewel. Sandor.. Forest Ecology and Management 243. E. Engelbrecht.. Foster. E. Masumori. Nichols. Bruijnzeel. Journal of Ecology 96. (INBIO). 2005. Ibarra. J. 2009. Singh. B. Burras..).. de Arboles Nativos y Exóticos de Puerto Rico y las Indias Occidentales. Ecology 78..