June 15, 2007

Floristics of Bamboo-Dominated Stands in Lowland Terra-Firma Forests of Southwestern Amazonia1

By Griscom, Bronson W Daly, Douglas C; Ashton, Mark S

GRISCOM, B. W. (Yale School of Forestry and Environmental Studies, New Haven, CT 06511), D. C. DALY (The New York Botanical Garden, Bronx, NY 10458), AND M. S. ASHTON (Yale School of Forestry and Environmental Studies, New Haven, CT 06511). Floristics of bamboo-dominated stands in lowland terra-firma forests of southwestern Amazonia. J. Torrey Bot. Soc. 134: 108-125. 2007. -We investigated distinctive aspects of the floristic communities associated with bamboo dominance (Guadua sarcocarpa and Guadua weberbaueri) in terra-firma forests of the Tambopata River watershed, southeastern Peru. Data were collected at five sets of paired plots (five bamboo-dominated, five tree-dominated) in which no evidence of prior human disturbance was identified. Three components of the plant community were analyzed: (1) trees (>/= 5 cm diam), (2) tree saplings (/= 1.5 meters). We used three statistical techniques (DCA, TWINSPAN, and ANOVA) to identify distinct assemblages of genera associated with the presence/absence of bamboo, and/or other factors (e.g., soil drainage, geographic location). All three statistical techniques produced similar results: the presence of bamboo is associated with a distinctive floristic community of trees and understory plants, but the tree sapling component of the plant communities was not distinct. Species richness in bamboo-dominated stands, as compared with tree- dominated stands, was 60% lower for trees and 36% lower for understory plants, but not significantly different for tree saplings. Bamboo-associated canopy trees and understory plants were characterized by fast growth rates, tolerance of poorly drained soils, and capacity for vegetative re-sprouting in response to physical damage to stems. These results suggest that differences in both soil and disturbance regimes drive floristic differences between bamboo-dominated stands and adjacent stands without bamboo. Key words: Amazonia, bamboo, DCA, diversity, Guadua sarcocarpa, Guadua weberbaueri, mass loading, plant community, soil, TWINSPAN.

In southwestern Amazonia, 180,000 km^sup 2^ have been classified as bamboo-dominated forest from remote-sensing analysis (Nelson 1994, Nelson et al. 1997). This area supports the largest neotropical bamboo-dominated forest formation. The two most common dominant bamboo species in this formation, Guadua sarcocarpa Londono & P.M. Peterson and Guadua weberbaueri Pilg. (hereafter collectively referred to as "bamboo"), are endemic to northern and western Amazonia (Judziewicz et al. 1999) and are arborescent and climbing, growing to over 20 meters in height when adjacent trees are available for support. Tree basal area is low where bamboo density is high. In southeastern Peru, Griscom and Ashton (2003) reported average tree basal area at 4.6 m2 ha^sup -1^ in plots with high bamboo density (3: 10 stems per 100 m2) as compared with 38.7 m2 ha^sup -1^ in adjacent plots without bamboo. The widespread occurrence of bamboo-dominance on terra-firma sites in this region has perplexed scientists because lush tree-dominated forests are usually widespread on such sites in lowland moist tropical regions (Bailey 1989, Whitmore 1998, Judziewicz et al. 1999).

FIG. 1. The two study sites, "Bahuaja" and "TRC", were located along the Tambopata River in the department of Madre de Dios, southeastern Peru. Stars identify locations of each paired plot site where two adjacent plots (with and without bamboo dominance) are located on terra-firma terraces.

There is controversy in the literature as to whether these bamboo- dominated forests represent a distinct plant community type or not, and to whether observed floristic distinctions there are based upon differences in soil, disturbance, and/or biotic factors. Hill (1999) conducted a classification of plant community types in the lower Tambopata River watershed, southeastern Peru, where our study plots are located (Fig. 1). Hill combined all active agricultural land, secondary forests (forests with evidence of recent anthropogenic disturbance), and bamboo-dominated forests into a single "disturbed" vegetation class. In contrast, Phillips et al. (1994) classified forest types in the same area and combined bamboodominated forest with mature tree-dominated forest in an edaphically-defined "Terra Firme Sandy-Clay" forest type.

Distinct edaphic characteristics have been identified for bamboo- dominated forests of southwestern Amazonia, most notably, more extreme soil moisture conditions (Vidalenc 2000, Griscom 2003); however, it is not known whether the edaphic differences detected in the Tambopata forests are a cause or result of bamboo-dominance, nor whether they are substantial enough to influence floristic patterns.

Guadua sarcocarpa and G. weberbaueri appear to impose a distinct competitive environment on trees in bamboo-dominated forest. Griscom (2003) recorded elevated light transmission beneath bamboo- dominated canopies using hemispherical photography: on average understory light levels (global site factor) were calculated to be 8% higher, and variance in light levels was twice as high, compared with forest plots without bamboo. It is not known if these calculated light level differences influence floristic patterns.

Griscom and Ashton (2006) have also found descriptive and experimental evidence for an autogenic, bamboo-induced disturbance regime in bamboo-dominated forests resulting from chronic mechanical loading of medium to large trees (5-30 cm dbh) by bamboo. Differential response of juvenile trees to mechanical damage has been found in Panamanian forests (Guariguata 1998), although it is not known if the mechanical damage imposed by Guadua spp. in southeastern Peru influences floristic patterns of forest trees.

In order to investigate the extent to which bamboo-dominated stands are floristically distinct, and alternative hypotheses for the cause or causes of floristic distinction, we analyzed three components of the plant community in bamboo-dominated stands (B+) and adjacent stands without bamboo (B-). These were: (1) trees si 5 cm dbh (hereafter referred to as "trees"), (2) tree saplings ^ 1 m height and

We also reviewed the literature to identify ecological traits of plant taxa in each forest type (with and without bamboo dominance), searching for patterns indicating causes of floristic differences between the two forest types. Specifically, we analyze our results with respect to three hypotheses for natural causes of bamboo- dominated plant communities:

1. Allogenic Disturbance

Bamboo-dominated stands represent recently disturbed, early successional stands that will, in the absence of future disturbance events, mature into structural and floristic compositions similar to adjacent tree-dominated stands;

2. Soils

Bamboo-dominated stands are floristically distinct due to edaphic conditions, and soil of bamboo-dominated forests will not support the plant communities occurring on adjacent tree-dominated sites; and

3. Bamboo

Bamboo-dominated stands are floristically distinct due to a unique competitive environment imposed by bamboo. The soils and allogenic disturbance regime of bamboo-dominated stands would support the plant communities found in adjacent stands without bamboo, but bamboo forests persist because bamboo inhibits tree recruitment.

Methods. SITE DESCRIPTION. We inventoried forest stands with and without bamboodominance in "Terra Firme Sandy-Clay" forest type (Phillips et al. 1994) and compared diversity indices and floristic assemblages among plots. We limited our investigation to forests without evidence of anthropogenic disturbance. Signs of past anthropogenic disturbance we looked for but did not encounter in the plots reported here included: charcoal in the A soil horizon, pot shards, tree stumps, logging roads, fences, pasture grass, single- cohort stand structure, and reports of impacts from local informants.

Data reported here were collected at the same locations as our earlier studies (Griscom 2003, Griscom and Ashton 2003). Two study sites were selected within terra-firma tropical moist forests along the Tambopata River, southeastern Peru. The "Bahuaja" site (accessed through Bahuaja Lodge camp site) is located 45 km downstream of the "TRC" site (accessed through the Tambopata Research Center; Fig. 1). The terra firma soils at each of our two study sites have been classified as deep, well-drained Ultisols, with scattered patches of poorly drained Inceptisols (ONERN 1972). Both Bahuaja and TRC soils have predominantly sandy clay loam A-horizons; however, the TRC site plots have significantly higher silt and lower clay in the A- horizon than those of the Bahuaja site (Griscom 2003). Barriers to water drainage were identified at 50-100 cm in depth at three of our five bamboo-dominated plots (B2B+, T1B+, T2B+), and at none of our tree-dominated plots (Griscom, 2003). As compared with the five tree- dominated plots, the soils of the five bamboo-dominated plots had on average: 10% lower CEC, 44% higher available calcium, 16% higher soil moisture during the wet season, and 9% lower soil moisture during the dry season (all differences were significant at P 0.05) was found between plots of the two forest types for other soil variables measured: soil texture, pH, aluminum, and other macronutrients and micronutrients (Griscom 2003). Bahuaja receives approximately 2000 mm rainfall per year (as interpolated from Johnson 1976), with precipitation increasing upriver towards the Andes, due to an orographie effect, and reaching approximately 3500 mm per annum at the TRC site (Pesce 1997). The dry season extends approximately April through September if defined as monthly rainfall: Pounds 200 mm for TRC and

Indicator tree species reported for terra firma forests at the Bahuaja site include: Cedrelinga cateniformis (Ducke) Ducke (Mimosaceae), Bertholletia excelsa Humb. & Bonpl. (Lecythidaceae), Hevea guianensis Aubl. (Euphorbiaceae), and Pourouma minor Benoist (Urticaceae) (CI-Peru 1999). All of these species occurred at both Bahuaja and TRC sites, with the exception of B. excelsa, which occurred only at the Bahuaja Site.

Terra firma forests at Bahuaja and TRC sites are a mosaic of mature bamboo-dominated forests and mature tree-dominated forests. At the Bahuaja site, the dominant bamboo species is Guadua weberbaueri PiIg., while at the TRC site, it is Guadua sarcocarpa Londono & P.M. Peterson.

FOREST INVENTORY. Ten 30 x 30 meter inventory plots were established between June 1999 and August 2001, six at the TRC site and four at the Bahuaja site. Two of the original six plots at Bahuaja site were eliminated from analysis because of prior anthropogenic disturbance. Plots were distributed in pairs, each plot of a pair separated by a minimum of 30 m, and located on either side of a transition zone between a bamboo-dominated forest stand (B+) and a forest stand without bamboo (B-) (Fig. 1). Plot locations were located randomly within the study area using Landsat TM imagery as described by Griscom and Ashton (2003). The B+ plots (n = 5) were defined a priori as having ten or more bamboo culms per 100 m2. A total of 2215 stems were sampled, distributed among B- plots (1398 stems) and B+ plots (817 stems), and among trees (665 stems), saplings (408 stems), and understory plants (1142 stems).

A nested plot design was used to sample different components of the plant community. At each of the ten inventory locations, one 30 x 30 meter plot was established for inventory of trees a 5 cm dbh. For inventory of saplings (trees > l m height and

For most taxa, small trees could be differentiated from large shrubs based on (1) branching morphology, (2) information from field guides (Gentry 1996, Ribeiro et al. 1999), and (3) a matching of tree saplings with canopy trees of the same species. For questionable cases, it was assumed that a species was a shrub if it was never taller than 3 m, and was observed producing flowers and/ or fruit.

Herbarium specimens representing nearly all tree, sapling, and understory plant stems, as well as fertile specimens of the two bamboo species, were collected in August - September of 2000 and 2001. A specimen was collected from each stem inventoried with the exception of common species where visual identification was clear. A minimum of three specimens were collected for all species. Tree climbing equipment was used to collect specimens from larger trees. Extensive notes were taken on bark and trunk slash characteristics (e.g., scent, latex, bark appearance) of each tree stem to aid in later identification. Duplicate herbarium specimens were sent to and deposited at both Museo de Historia Natural, Lima and The New York Botanical Garden (NYBG) where they were independently sorted into morphospecies and identified.

DATA ANALYSIS. Alpha diversity was measured with the following indices: (1) species richness per area (simple count of all species encountered in B+ plots vs. B- plots); (2) species richness per stem (no. stems/no, species); (4) Shannon diversity (H' = - [n-ary summation] p^sub i^ p^sub i^, where p^sub i^, = proportion of individuals found in the ith species); and (5) Shannon evenness (E = H' I InS where S = total number of species) (Magurran 1988). For the variables of species richness per unit area and number of stems per unit area, plot means for B+ plots were compared with those for B- plots ( 5 cm dbh for each species in each forest type following Barbour et al. (1987).

We calculated species richness per stem using equivalent stem sample sizes within each paired plot set because the likelihood of encountering new species decreases as more stems are inventoried, and because stem densities varied substantially between Band B+ stands. To accomplish a balanced sampling, stems from the higher density plots were randomly sub-sampled five times for the number of stems occurring in the lowerdensity adjacent plot and the mean value of the number of species occurring in each of the five trials was calculated.

Beta diversity, the overall similarity in taxonomic composition of inventory plots (

Analysis of beta diversity using stem counts by species was not expected to produce meaningful results since many species were represented by single stems due to very high species diversity in our plots. Therefore, for the three statistical methods above, we used stem count data by genus, and only "common genera" were used in analysis. Common genera within each plant group (trees, saplings, and understory) were defined as those whose stem count was 3% or more of the total stern count for B+ or B- plots. Thus 19 common genera were selected for trees, 22 common genera were selected for saplings, and 14 common genera were selected for understory plants.

PC-ORD was used for calculations of species area curves, TWINSPAN, and DCA (PC-ORD for Windows, V 4.17, MjM Software). Species area curves were generated of Sorensen (Bray-Curtis) type. All other statistical computations were performed using SPLUS software (S-PLUS 2000 Professional Release 3, Mathsoft, Inc.).

LITERATURE REVIEW OF INDICATOR TAXA. Ten genera (seven overstory, three understory) were selected from either end of the first DCA axis as "bamboo-dominated forest indicators" and "tree-dominated forest indicators". We conducted a literature review on these genera for information on the following ecological traits relevant to investigating the three hypotheses discussed in the introduction: canopy position at maturity, growth rate, capacity for vigorous stem re-sprout in response to stem damage, and tolerance for poor soil drainage. This literature review focused on the species within selected genera that occurred in our plots; however, when information was limited for species in our plots, additional information was gathered on congeneric species, or general information on a genus. Genera were characterized as belonging to one or more of the following canopy strata: canopy (> 30 m height), midcanopy (16-30 m height), subcanopy (3-15 m height), and understory (

Results. ALPHA DIVERSITY. Mean species richness per plot was significantly lower in B+ plots compared with B- plots for trees (68% lower, P = 0.01) and understory plants (36% lower, P = 0.04), but no significant difference was found for saplings (Table 1). Mean stem density was significantly lower in B+ plots vs. B- plots for trees (76% lower, P 0.05). For stems/species, Sannon diversity, and Shannon evenness, values are based on all plots combined per forest type and plant type, thus standard deviation and P-values are not available.

A total of 225 tree morphospecies were identified in all plots, 75 of which occurred in bamboo-dominated plots, and 189 of which occurred in tree-dominated plots. A total of 162 sapling morphospecies were identified in all plots, 99 of which occurred in bamboodominated plots, and 109 of which occurred in tree-dominated plots. A total of 46 shrub morphospecies were identified in all plots, 28 of which occurred in bamboo-dominated plots, and 38 of which occurred in tree-dominated plots. A list of all species identified in bamboo-dominated and/or tree-dominated plots is given in Appendix 1.

BETA DIVERSITY. For trees and understory plants, the first DCA axis differentiated plots based on bamboo presence/absence: all B+ plots had higher eigenvalues than B- plots (Figs. 3, 5). In contrast, for sapling data, plots were not differentiated based on bamboo presence-absence for any of the first three axes. Plots did, however, differentiate along the first DCA axis according to geography for sapling data (Bahuaja plots had higher eigenvalues than any of the TRC plots) (Fig. 4). The three plots with barriers to water drainage (B2B+, T1B+, T2B+) were not differentiated from other plots along any of the first three axes for saplings, trees, or understory plants.

For understory plant data, TWINSPAN differentiated plots primarily based on bamboo presence-absence (first branch of dendrogram) in parallel to DCA results. TWINSPAN also further differentiated understory plant plots based on geography with secondary branching of the dendrogram (Fig. 5).

TWINSPAN analysis generated similar yet less discrete patterns for trees and saplings. For mature trees, Bahuaja 2B+ plot (B2B+) was lumped with B- plots in the first branch of the TWINSPAN dendrogram; however, B2B+ was then identified as distinct from all B- plots in the second dendrogram branch. The B2B+ tree plot was distinct from other bamboo-dominated tree plots in the lack of Inga spp., and the presence of Hevea spp. (Fig. 3). TRC sapling plot 3B+ was lumped with Bahuaja plots, but otherwise TWINSPAN also separated sapling plots by geography (Fig. 4).

FIG. 2. Species area curves are presented for trees (>/= 5 cm dbh), saplings (1 m height to 4.9 cm dbh), and understory plants (>/ = 1.5 m height) in bamboo-dominated (B+) vs. bamboo-free (B-) stands in terra-firme lowland forests of southeastern Peru. Trajectories of curves appear distinct for trees and understory plants, but not saplings. Only the curve for B+ understory plants was approaching asymptote.

The non-parametric ANOVA (ScheirerRay-Hare Extension of Kruskal- Wallis Test) confirmed floristic differences between B+ and B- plots for both trees and understory plants: significantly non-random distribution of both tree and understory plant genera between B+ and B- plots was found (P

literature reviewof indicator taxa. All genera identified as indicators of tree-dominated plots (B-) have been characterized as having medium to slow growth rates including species of the canopy, midcanopy, subcanopy, and understory (Chazdon 1991, Korning and Balslev 1994, Gentry 1996, Finegan et al. 1999, Svenning 1999, Flores and Ashton 2000) (Table 3). No information was available for these genera regarding capacity for stem resprout or flood tolerance. Four genera (three of which were understory plants) were palms, and thus lack the capacity for re-sprouting from broken stems.

Genera identified as indicators of bamboodominated plots in the canopy and understory strata were generally characterized by fast growth rates. In contrast, midcanopy/subcanopy taxa of bamboo- dominated plots were generally reported to have medium to slow growth rates (Delamo and Ramos 1993, Korning and Balslev 1994, Oliveirafilho et al. 1994, Gentry 1996, Pennington 1998, Thompson et al. 1998, Yu and Pierce 1998, Finegan et al. 1999, Hiremath 2000) (Table 3). Capacity for vigorous re-sprout from broken stems has been reported for one or more species belonging to each of the three canopy tree genera identified as indicators of bamboodominated stands: Himatanthus Willd. ex Schult., Cordia L., and Inga Mill. (Miller and Kauffman 1998, Pennington 1998, Thompson et al. 1998). We observed aggressive stem resprouting and adventitious rooting in the understory bamboo Olyra latifolia L. that was associated with bamboo-dominated stands. References regarding tolerance of poorly drained soils were found for the bamboo-associated species Inga alba (Sw.) Willd., Inga capitata Desv., and Socratea exorrhiza (Mart.) H. Wendl. (Pennington 1998, Pacheco 2001).

Table 2. The ten species with the highest importance values (IV) are reported for bamboo-dominated (B+) and tree-dominated (B-) plots. Importance values were calculated for all stems > 5 cm dbh based on basal area, stem density, and frequency of each species. No overlap of dominant species occurred between the two forest types.

No information on growth rates was found for the genera Nealchornea Huber, Oenocarpus Mart., Socratea H. Karst., and Aphelandra R. Br. or for understory species of Bactris Jacq. ex Scop., Hyospathe Mart., and Cordia. Thus, growth rates given for these taxa in Table 3 are based on our observations in the field as well as information on reproductive responses to gaps in the case of Cordia nodosa Lam. (Yu and Pierce 1998) and Aphelandra aurantiaca Lindl. (Calvo-Irabien and Islas-Luna 1999). The tree genus casearia Jacq. was not included in Table 3 due to lack of information from the literature, or enough stems in our plots to make inferences about ecological traits.

Discussion. ALPHA AND BETA DIVERSITY. Several measures of diversity reveal floristic differences between bamboo-dominated forests and tree-dominated forests beyond the presence/absence of bamboo. A floristically distinct bamboo-dominated plant community was detected, including trees 2= 5 cm dbh and large understory plants, but not the tree sapling component of the vegetation.

Species richness per area was lower in bamboo-dominated (B+) forests for trees and understory plants, but not for saplings. For understory plants, this plot-level diversity difference (both richness and evenness) may reflect lower understory diversity at the landscape scale as evidenced by higher mean stems/species for B+ plots and a species-area curve for B+ plots approaching asymptote. For trees, sampling was not extensive enough for conclusive results on species richness at the landscape scale. We infer from our results that more sampling will reveal lower tree species richness at the landscape scale in bamboodominated forests, as suggested by the somewhat lower Shannon diversity (//') value found in B+ plots (vs. B- plots), and the B+ species-area curve closer to asymptote than the B- curve.

FIG. 3. Results from TWINSPAN (two-way table on left) and DCA (graph on right) are presented for tree genera (trees > 5 cm dbh) in terra firme lowland forests of southeastern Peru. For the two-way ordered tables, plots are listed vertically along the top. The letters "B" and "T" denote "Bahuaja Site" and "TRC Site" respectively. The numbers on the second line identify the paired plot location, and the positive/negative signs denotes bamboo- dominated (+) vs. control (-) plots. Columns of O's and 1 's on the right identify the branching of the dendrogram for genera listed to the left. The dendrogram for plots is presented diagrammatically at the top. Locations of plots with respect to DCA axis 1 and 2 are identified with circles (closed circles are bamboo-dominated plots, open circles are control plots). Genera are plotted with "+" signs and identified with numbers that correspond with the numbered genus names in the TWINSPAN twoway plots. DCA differentiates control (B-) plots from bamboo-dominated plots (B+). TWINSPAN reveals a similar pattern, although plot B2+ is isolated in the dendrogram from other B+ plots.

Bamboo-dominated stands and tree-dominated stands are also distinct in terms of floristic composition. As with alpha diversity, floristic composition was most clearly distinct for the understory plant community. Understory plots separated according to bamboo presence-absence by both DCA and TWINSPAN analysis. Bamboo- dominated stands were also identified as different, yet not as clearly distinct, in terms of mature tree floristic composition: DCA and TWINSPAN isolated plots without bamboo from bamboo-dominated plots; however, TWINSPAN also lumped a bamboo-dominated plot with treedominated plots in the first branch of the dendrogram. ANOVA results supported a distinct tree and understory plant floristic composition in bamboo-dominated as compared with tree-dominated plots. Also, the importance values of the top ten tree species of the two forest types did not overlap. Sapling floristic composition was not distinctive with respect to bamboo presence-absence, based on results from DCA, TWINSPAN, and AN-OVA. The floristic distinction for trees and understory plants associated with the presence/ absence of bamboo was expressed by the principle component of DCA as a gradient rather than a discrete clustering of plots in genus space. While associations between a variety of plant taxa and the presence or absence of bamboo were apparent, many additional plant taxa occurred in both community types.

Various taxa common in tree-dominated stands including several palms appear to be excluded from bamboo-dominated forests; these include Geonoma diversa (Poit.) Kunth, Hyospathe elegans Mart., Oenocarpus spp., Pourouma minor, Naucleopsis spp., and Iryanthera spp. The few species recorded only in bamboo-dominated plots (e.g., Himatanthus sucuuba (Spruce ex Mull. Arg.) Woodson, Inga alba, Cordia nodosa) occurred at relatively low densities, and more sampling would be required to determine if these could be considered obligate bamboo-dominated forests species in our area. Based on our observations in surrounding forests, most bamboo-associated species, such as Inga alba and Corclia nodosa, are not bamboo-forest obligates; however, some bamboo-associated species may depend upon bamboo-dominated forests for core habitat in metapopulation dynamics (Hanski and Gyllenberg 1993).

FIG. 4. Results from TWINSPAN (two-way table on left) and DCA (graph on right) are presented for sapling genera (trees a 1 m height and

CAUSES OF FLORISTIC PATTERNS. Our results are most consistent with two of the hypotheses presented in the introduction: the floristic composition of bamboo-dominated forests is influenced by (1) competitive effects of bamboo on other plants, particularly chronic loading and light environment (Griscom 2003, Griscom and Ashton 2003, Griscom and Ashton 2006), and (2) hydrologic properties of soils characteristic of bamboo-dominated stands (Griscom 2003).

The floristic patterns we found are less consistent with the hypothesis that floristic differences between bamboo-dominated stands and tree-dominated stands result from recent allogenic disturbance. The only nonanthropogenic source of allogenic disturbance to terra firma forests we are aware of in this region is wind, specifically convective downbursts (Nelson et al. 1994, Garstang et al. 1998) or rare storm events (Foster and Terborgh 1998). These forms of wind disturbance are expected to result in (1) an assemblage of tree taxa occupying the forest canopy that is not floristically distinct from that preceding wind damage as reported by Foster and Terborgh (1998), and (2) an assemblage of tree taxa occupying the regenerating understory that are disproportionately represented by pioneer tree taxa (e.g., Cecropia spp., Jacaranda spp.) as was found in response to artificial gaps created in our study area (Griscom 2003). In contrast, we found that (1) the assemblage of tree taxa occupying the forest canopy was floristically distinct, and (2) the assemblage of tree taxa occupying the forest understory was not floristically distinct. Although recent wind disturbance is not suggested by our results, it is possible that the high light-adapted tree taxa we found occurring in the canopy of bamboo-dominated stands became established after a canopyremoving disturbance decades ago.

FIG. 5. Results from TWINSPAN (two-way table on left) and DCA (graph on right) are presented for understory plant genera (shrubs and herbs SL 1.5 meters height) in terra-firme lowland forests of southeastern Peru. For the two-way ordered tables, plots are listed vertically along the top. The letters "B" and "T" denote "Bahuaja Site" and "TRC Site" respectively. The numbers on the second line identify the paired plot location, and the positive/negative signs denotes bamboo-dominated (+) vs. control (-) plots. Columns of O's and 1 's on the right identify the branching of the dendrogram for genera listed to the left. The dendrogram for plots is presented diagrammatically at the top. Locations of plots with respect to DCA axis 1 and 2 are identified with circles (closed circles are bamboo- dominated plots, open circles are control plots). Genera are plotted with "+" signs and identified with numbers that correspond with the numbered genus names in the TWINSPAN two-way plots. Plots are differentiated primarily on the basis of bamboo presenceabsence and secondarily on the basis of geography (Bahuaja Site vs. TRC Site) in both TWINSPAN and DCA analysis.

The lack of a distinct assemblage of tree sapling genera in bamboo-dominated stands suggests that the unique edaphic characteristics of bamboo-dominated plots do not prohibit colonization by most common tree genera characteristic of tree- dominated stands. In contrast, ecological traits of trees s 5 cm dbh suggest that there may be an edaphic influence on the tree community in bamboodominated forests. Three tree species with the first, second, and sixth highest importance values in bamboo-dominated plots (Inga alba, Socratea exhorri:a, and Inga capitata, respectively) have been described as tolerant of poorly-drained soils (Clark et al. 1995, Pennington 1998, Pacheco 2001). Edaphic constraints may play a role in the floristics of bamboo-dominated forests by shifting the competitive balance in the favor of tree species tolerant of excessive moisture and/or drought or of other characteristics of soils occurring in bamboo-dominated soils, such as reduced soil cation exchange capacity (Griscom 2003).

Table 3. Genera identified as indicators of bamboo-dominated forests (B+) and tree-dominated forests (B-) reviewed for the following ecological traits: canopy position, growth rate, capacity for stem re-sprout, and tolerance of poorly drained soils.

Half of the most important tree species in bamboo-dominated plots are capable of vigorous stem re-sprout in response to stem damage (Table 3), suggesting there may be selection for species that can withstand chronic mass loading and stem breakage. Also, the "top- down" floristic pattern we identified for trees (trees >/= 5 cm dbh were floristically distinct, but saplings were not) is consistent with the evidence that bamboo mass loading has more impact on trees >/= 5 cm dbh than on saplings (Griscom 2003).

The trend of fast growing, relatively shade intolerant tree taxa in the canopy of bamboodominated stands suggests that either (1) trees emerged through a bamboo canopy characterized by relatively high light penetration (Griscom 2003), or (2) trees regenerated in response to a past canopy opening event, whether before or after the establishment of bamboo-dominance. The -30-year cycle of gregarious monocarpic mortality of Guadua spp. offers a repeated phenomenon of largescale canopy opening that may allow a cohort of shade- intolerant tree species to reach the canopy without passing the gauntlet of bamboo mass loading.

Since little is known about the ecophysiological characteristics of most plant species in this forest system, our inferences are based on both direct and indirect evidence from the literature (e.g., from congeneric species) and the field. While we did identify literature sources reporting distinctive traits of bamboo- associated genera (e.g., flood tolerance, vigorous stem re-sprout), we did not identify literature demonstrating lack of these traits in genera identified as indicators of tree-dominated stands. Further research on ecophysiological characteristics of species regarding edaphic constraints, shade tolerance, growth rate, and traits affecting tolerance for mass loading is needed to improve our understanding the bamboo-dominated plant community. A larger study allowing floristic analysis of species-level data is needed to confirm the trends we identified. We predict that many of the bamboo- associated species will be characterized by vigorous stem-sprouting (conferring resilience to bamboo mass loading) and/or tolerance of poor drainage and/or moisturelimited soils (the latter two often being linked).

Two particularly enigmatic floristic patterns we identified for bamboo-dominated forests are (1) the presence of mid- to sub-canopy tree taxa such as Theobroma speciosum Willd. ex Spreng., Neea spp., Cordia nodosa with relatively slow growth rates, and (2) the paucity of palm taxa. The association of mid- to sub-canopy tree taxa with relatively slow growth rates discourages classification of bamboo- dominated stands as entirely an early-serai assemblage, such as expected for a secondary forest stand regenerating from abandoned agricultural fields. If mass-loading by bamboo is indeed a dominant cause of plant mortality for trees, then this functional group should have ecological traits conferring survival advantage under chronic mass loading such as strong wood and strong capacity for re- sprouting from damaged branches/stems.

Palms have traits that may give them a selective advantage in avoiding mass-loading such as shedding leaf petiole "branches" and a lack of true branches on which bamboo can load. They also have traits that may result in selective disadvantage in recovering from bamboo-mass loading (e.g., relatively slow growth rates and a single terminal meristem which, if broken off, results in stem death). Thus, further investigation is necessary to determine whether the absence of palm taxa in bamboo-dominated stands is consistent, inconsistent, or unrelated to our hypothesis that bamboo-mass loading is imposing a selective pressure on tree taxa present in bamboo-dominated stands. Socratea exorrhiza is a noteworthy exception to the apparent exclusion of palms from our bamboo- dominated plots. Socratea exorrhiza may have an advantage over other palms in surviving chronic bamboo mass-loading due to (1) wide aerial root buttresses, and (2) relatively fast growth rate compared with other palms (J. Terborgh, pers. comm.). IMPLICATIONS FOR CONSERVATION AND MANAGEMENT. Evidence from this study suggests that bamboo-dominated forests without evidence of prior anthropogenic disturbance should be treated as a floristically distinct plant community that shares strong floristic affinities with tree- dominated terra firma forests. The distinct bamboo-dominated plant community appears to result from an interaction of variables including the edaphic preferences of various plant taxa and the growth strategy and phenology of Guadua sarcocarpa and Guadua weberbaueri. The -30-year gregarious monocarpic life-cycle of these two species likely results in a dynamic serai forest system which probably drives migration and/or "boom and bust" population cycles of resident plants and animals at the scale of 100 to 10,000 km2 that Nelson et al. (2001) report for gregarious monocarpic events. These large-scale natural phenomena support the need for large- scale protected areas design in order to allow for the metapopulation dynamics that we expect are critical for the maintenance of biological diversity in this system. Of particular concern are the variety of bamboo-obligate fauna which are believed to depend upon mature bamboo-dominated forests for survival, including 19 bird species (Kratter 1997), a primate, Goeldi's Marmoset (Callimico goeldii Thomas; Wright 1985), neotropical bamboo rats of the subfamily Dactylomyinae (Emmons 1981), a poison arrow frog (Dendrobates biolat Morales), at least three ant species, and a yet to be determined number of other invertebrate species, including katydids, butterflies, and dragonflies (D. Davidson, pers. comm.).

Expansion of secondary bamboo stands into formerly tree- dominated forests due to anthropogenic impacts presents a different conservation challenge. Research is necessary to determine if secondary bamboo forests support bamboo-specialist fauna. Given the lower biomass of bamboo-dominated forests (Nelson et al. 2001) and our evidence that mature bamboo-dominated forests support a comparatively reduced plant species diversity, dramatic expansion of bamboo could result in loss of biological diversity and increased carbon emissions. Species such as Inga alba, characterized by fast growth, tolerance of edaphic extremes, and tolerance of stem damage, are good candidates for experimental restoration of tree-dominance in secondary bamboo-dominated sites.

1 This work was supported by funding from The Cullman Fellowship through The New York Botanical Garden, a University Fellowship through Yale Graduate School of Arts and Science, and the RKG Saddleback Institute.

Literature Cited

BAILEY, R. G. 1989. Explanatory supplement to ecoregions map of the continents. Env. Conserv. 16: 307-309.

BARBOUR, M. G., J. H. BURK, AND W. D. PITTS. 1987. Terrestrial Plant Ecology, 2nd ed. The Benjamin/Cummings Publishing Co., Inc., Menlo Park, CA.

BARNES, B. V., D. R. ZAK, S. R. DENTON, AND S. H. SPURR. 1998. Forest Ecology, 4"' ed. John Wiley & Sons, Inc. New York, NY.

CALVO-IRABIEN, L. M. AND A. ISLAS-LUNA. 1999. Predisperal predation of an understory rainforest herb Aphelandra aurantiaca (Acanthaceae) in gaps and mature forest. Am. J. Bot. 86: 1108-1113.

CHAZDON, R. L. 1991. Plant size and form in the understory plant genus Gcononui: Are species variations on a theme? Am. J. Bot. 78: 680-694.

CI-PERU. 1999. Zona Reservada de Tambopata Candamo: (Madre de Dios - Puno) Peru. Conservation International - Peru, Lima, Peru.

CLARK, D. A., D. B. CLARK, R. SANDOVAL, AND M. V. CASTRO. 1995. Edaphic and human effects of landscape-scale distributions of tropical rainforest palms. Ecology 76: 2581-2594.

COSTA, F. R. C., C. SENNA, AND E. M. NAKKAZONO. 2002. Effects of selective logging on populations of two tropical understory herbs in an Amazonian forest. Biotropica 34: 289-296.

DELAMO, R. S. AND P. J. RAMOS. 1993. Use and management of secondary vegetation in a humidtropical area. Agroforestry Systems 21: 27-42.

EMMONS, L. H. 1981. Morphological, ecological, and behavioral adaptations for arboreal browsing in Dactvhmvs dactvlinus (Rodentia, Echimyidae). J. Mammal. 62: "l83 189.

FINEUAN, B., M. CAMACHO, AND N. ZAMORA. 1999. Diameter increment patterns among 106 tree species in a logged and silviculturally treated Costa Rican rain forest. For. Ecol. Manage. 121: 159-176.

FLORES, C. F. AND P. M. S. ASHTON. 2000. Harvesting impact and economic value of Geonoma deverxa, Arecaceae, an understory palm used for roof thatching in the Peruvian Amazon. Econ. Bot. 54: 267- 277.

FOSTER, M. AND J. TERBORGH. 1998. Impact of rare storm events on an Amazonian forest. Biotropica 30: 470-474.

GARSTANO, M., S. WHITE, H. H. SHUOART, AND J. HALVERSON. 1998. Convective cloud downdrafts as the cause of large blowdowns in the Amazon rainforest. Meteor. Atmosph. Physics 67: 199-212.

GAUCH, H. G. AND R. H. WHITTAKER. 1981. Hierarchical classification of community data. J. Ecol. 69: 135-152.

GENTRY, A. H. 1996. A Field Guide to the Families and Genera of Woody Plants of Northwest South America (Colombia, Ecuador, Peru) with Supplementary Notes on Herbaceous Taxa. The University of Chicago Press. Chicago, IL.

GRISCOM, B. W. 2003. The influence of bamboo (Guailua sarcocarpa and Giiaclua weherhaueri) on stand dynamics in lowland terra-firme forests of southeastern Peru. Ph.D. dissertation Yale University, Graduate School of Arts and Sciences, New Haven, CT.

GRISCOM, B. W. AND M. S. ASHTON. 2003. Bamboo control of forest succession: Gutulua xarcocarpa in Southeastern Peru. For. Ecol. Manage. 2003: 445-454.

GRISCOM, B. W. AND P. M. S. ASHTON. 2006. A selfperpetuating bamboo disturbance cycle in a neotropical forest. J. Trop. Ecol. 22: 587-597.

GUARIGUATA, M. R. 1998. Response of forest tree saplings to experimental mechanical damage in lowland Panama. For. Ecol. Manage. 102: 103-111.

HANSKI, I. AND M. GYLLENBERO. 1993. Two general metapopulation models and the core-satellite species hypothesis. Am. Nat. 142: 17- 41.

HILL, M. O. 1979. TWINSPAN A FORTRAN program for arranging multivariate data in an ordered two-way table by classification of the individuals and attributes. Ecology and Systematics Cornell University, Ithaca, NY.

HILL, M. O. AND H. G. GAUCH. 1980. Detrended correspondence analysis: an improved ordination technique. Vegetatio 42: 47-58.

HILL, R. A. 1999. Image segmentation for humid tropical forest classification in Landsat TM data. Int. J. Remote Sensing 20: 1039- 1044.

HIREMATH, A. J. 2000. Photosynthetic nutrient-use efficiency in three fast-growing tropical trees with differing leaf longevities. Tree Phys. 20: 937-944.

JOHNSON, A. M. 1976. The climate of Peru. Bolivia and Ecuador, p. 147-183. In W. Schwerdtfeger [ed.]. Climates of Central and South America. Elsevier Scientific Pub. Co., New York. NY.

JUDZIEWICZ, E. J.. L. G. CLARK, X. LONDONO. AND M. J. STF.RN. 1999. American Bamboos. Smithsonian Institution Press. Washington. D.C.

KORNINU, J. AND H. BALSLEV. 1994. Growth rates and mortality patterns of tropical lowland tree species and the relation to forest structure in Amazonian Ecuador. J. of Trop. Ecol. 10: 151-166.

KRATTER, A. W. 1997. Bamboo specialization by amazonian birds. Biotropica 29: 100-110.

MAGURRAN, A. E. 1988. Ecological Diversity and Its Measurement Princeton University Press, Princeton, NJ.

MILLER, P. M. AND J. B. KALIFFMAN. 1998. seedling and sprout response to slash-and-burn agriculture in a tropical deciduous forest. Biotropica 30: 538-546.

NELSON, B., V. KAPOS, J. ADAMS, W. OLIVEIRA, O. BRAUN, AND I. D. AMARAL. 1994. Forest disturbance by large blowdowns in the Brazilian Amazon. Ecology 75: 853-858.

NELSON, B. W. 1994. Natural forest disturbance and change in the Brazilian Amazon. Remote Sensing Reviews 10: 105-125.

NELSON, B. W.. R. KALLIOLA, AND G. SHEPARD. 1997. Tabocais de Guadua spp. no sudeste amazonico: extensao geografica, mortalidade sincronizada e relacao com incendios florestais, p. 163. In Abstracts of the 38th National Botanical Congress Universidade Regional do Cariri, Sociedade Botanica do Brasil, Crato - Ceara.

NELSON, B. W., A. C. A. OLIVEIRA, G. T. BATISTA, D. VIDALENC, AND M. SiLVEiRA. 2001. Modeling biomass of forests in the southwest Amazon by polarordination of Landsat TM. In M. L. Ribeiro and Y. R. d. S. Souza [eds.], Proceedings, Tenth Brazilian Remote Sensing Symposium. INPE, Foz de Iguacu, Parana, Brazil.

OLIVEIRAFILHO, A. T., E. A. VILELA, M. L. GAVILANES, AND D. A. CARVALHO. 1994. Effect of flooding regime and understory bamboos on the physiognomy and tree species composition of a tropical semideciduous forest in southeastern Brazil. Vegetatio 113: 99-124.

OLIVER, C. D. AND B. C. LARSON. 1996. Forest Stand Dynamics, Updated edition. John Wiley & Sons, Inc., New York, NY.

ONERN. 1972. Inventario, evaluacion e integracion de los recursos naturales de la zona de los Rios Inambari y Madre de Dios. Oficina Nacional de Evaluacion de Recursos Naturales (Director General Ing. Jose Lizarraga Reyes). Republica del Peru, Lima - Peru.

PACHECO, M. A. W. 2001. Effects of flooding and herbivores on variation in recruitment of palms between habitats. J. Ecol. 89: 358- 366.

PENNINOTON, T. D. 1998. Growth and biomass production of Inga species. In T. D. Pennington and E. C. M. Fernandes [eds.], The Genus Inga: Utilization The Royal Botanic Gardens, Kew. PESCE, M. N. 1997. Publication #6 del Tambopata Research Center: Control Meteorologico del Tambopata Research. Rainforest Expeditions Conservation International.

PHILLIPS, O., A. H. GENTRY, C. REYNEL, P. WILKIN, AND C. GALVEZ- DURAND. 1994. Quantitative Ethnobotany and Amazonian Conservation. Conserv. Bio. 8: 225-248.

RIBEIRO, J. E. L. D. S., M. J. G. HOPKINS, A. VlCENTINI, C. A. SOTHERS, M. A. D. S. CoSTA, J. M. D. BRITO, M. A. D. D. SOUZA, L. H. P. MARTINS, L. G. LOHMANN, P. A. C. L. ASSUNCAO, E. D. C. PEREIRA, C. F. D. SILVA, M. R. MESQUITA, AND L. C. PROCOPIO. 1999. Flora da Reserva Ducke: Guia de identificacao das plantas vasculares de uma floresta de terra-firme na Amazonia Central. INPA, Manaus.

SCHLONVOIGT, A. AND M. ScHLONVoiGT. 2002. Initial shoot and root growth patterns of Cordia alliodora in agroforestry systems with perennial crops in the Atlantic lowlands of Costa Rica. J. Sust. Agric. 20: 41-56.

SMITH, D. M., B. C. LARSON, M. J. KELTY, AND P. M. S. ASHTON. 1997. The Practice of Silviculture: Applied Forest Ecology, Ninth Edition. John Wiley & Sons, Inc., New York, NY.

SOKAL, R. R. AND F. J. ROHLF. 1995. Biometry, 3rd ed. W. H. Freeman and Company, New York, NY.

SVENNINO, J. C. 1999. Recruitment of tall arborescent palms in the Yasuni National Park, Amazonian Ecuador: Are large treefall gaps important? J. Trop. Ecol. 15: 355-366.

THOMPSON, J., J. PROCTOR, D. A. SCOTT, P. J. FRASER, R. H. MARKS, R. P. MILLER, AND V. VIANA. 1998. Rain forest on Maraca Island, Roraima, Brazil: artificial gaps and plant response to them. For. Ecol. Manage. 102: 305-321.

TUOMISTO, H. AND K. RuoKOLAiNEN. 1994. Distribution of Pteridophyla and Melastomataceae along an edaphic gradient in an Amazonian rain forest. J. Veg. Sci. 5: 25-34.

VIDALENC, D. 2000. Distribuicao das florestas dominadas pelo bambu Guadua weberbaueri em escala de paisagem no sudeste da Amazonia e fatores edaficos que afetam sua densidade. Masters of Science. Universidade Do Amazonas - UA, Institute Nacional de Pesquisas da Amazonia - INPA, Manaus.

WHITMORE. T. C. 1998. An Introduction to Tropical Rain Forests, second Edition. Oxford University Press, Oxford, UK.

WRIGHT, P. C. 1985. The costs and benefits of nocturnality for Aotus trivirgatus (the night monkey). City University of New York, NY.

YU, D. W. AND N. E. PIERCE. 1998. A castration parasite of an ant- plant mutualism. Proceedings of the Royal Society of London, Series B, Biological 265: 275-282.

Bronson W. Griscom2,3

Yale School of Forestry and Environmental Studies, New Haven, CT 06511

Douglas C. Daly

The New York Botanical Garden, Bronx, NY 10458

Mark S. Ashton

Yale School of Forestry and Environmental Studies, New Haven, CT 06511

2 Many thanks to J. Saavedra, Z. Saavedra, J. Mukhopadhyay, A. Jetmore, and C. Huisa for field work and general expertise in Peru. We thank F. Cornejo, H. Beltram, A. Henderson, J. Janovek, and M. Nee for taxonomic expertise. We thank B. Larson, B. Nelson, C. Peters, J. Terborgh, D. Vogt, K. Vogt, D. Yu, P. Gagnon, and H. Griscom for discussions and advice contributing to this paper. We thank the Museo de Historia Natural (Lima Peru) and The New York Botanical Garden (USA) for use of their herbaria and excellent staff support. For providing logistical support that made this research possible we thank C. Llerena, T. Smith, Greenforce, and everyone at Rainforest Expeditions Inc. We thank the Institute Nacional de Recursos Naturales (INRENA) for granting us permits to access the Tambopata-Candamo National Park and buffer zones in order to conduct our research.

3 Author for correspondence: 255 Franklin St., Harrisonburg, VA 22801; E-mail: [email protected] morphoblue.org

Received for publication November 6, 2005, and and in revised form September 25, 2006.




Copyright Torrey Botanical Society Jan/Feb 2007

(c) 2007 Journal of the Torrey Botanical Society. Provided by ProQuest Information and Learning. All rights Reserved.