Spatial distribution and environmental preferences of the piassaba palm Aphandra natalia (Arecaceae) along the Pastaza and Urituyacu rivers in Peru

Spatial distribution and environmental preferences of the piassaba

palm Aphandra natalia (Arecaceae) along the Pastaza

and Urituyacu rivers in Peru

Thomas Boll a, Jens-Christian Svenning a, Jaana Vormisto a,Signe Normand a, Cesar Grandez b, Henrik Balslev a,*

a Institute of Biological Sciences, University of Aarhus, Building 540, Ny Munkegade, 8000 Aarhus C., Denmarkb Universidad Nacional de la Amazonıa Peruana, Facultad de Ciencias Biologicas, Apdo. 326, Iquitos, Peru

Abstract

Aphandra natalia (Balslev and Henderson) Barfod, an economically important fibre producing palm, is common in rainforest

on low terraces along the Pastaza and Urituyacu rivers in Amazonian Peru. The aim of this study was to investigate the spatial

distribution and environmental preferences of Aphandra in old-growth terrace forest to which it is limited in this region.

Densities of immature and mature individuals were 507 � 212 (S.D.) ha�1 and 19 � 8 ha�1, respectively, in 11 (5 � 500 m)

transects placed in old growth terrace forest near four villages and 739 � 188 ha�1 and 96 � 49 ha�1, respectively, in six

irregular transects placed in what the local villagers considered dense Aphandra stands. We examined environmental and spatial

correlates of Aphandra occurrences using stepwise multiple autologistic regressions. Site, soil moisture, slope inclination, and

topographic position influenced the spatial distribution of Aphandra. Furthermore, the distribution was strongly clumped,

independently of environmental factors, with particularly the concentration of immature individuals around adults pointing to

dispersal limitation as the likely causal mechanism.

Keywords: Dispersal limitation; Ecological sustainability; Fibre plants; Non-timber forest products; Palms; Recruitment; Resource

availability; Tropical rainforest; Extractivism

1. Introduction

For economically important species subject to

extraction from natural stands it is of obvious

* Corresponding author. Tel.: +45 8942 4707/8942 2743;

fax: +45 8942 4747/8613 9326.

E-mail address: [email protected] (H. Balslev).

importance to understand their occurrence, spatial

variation and environmental preferences because these

features are intrinsically related to their sustainable

management. Tropical rainforest species have con-

spicuous spatial variation in occurrence at all scales

including very large ones over thousands of kilometres

(Pitman et al., 2001), over intermediate ones of tens of

kilometres (Tuomisto et al., 2003) to the very small

scales measured in metres (Svenning, 2001a; Valencia

et al., 2004). Competing views of the causes under-

lying this variation range from random walk

hypotheses that explain local variation in occurrence

as being the result of neutral demographic stochas-

ticity and dispersal limitation (Hubbell, 2001) to

deterministic hypotheses that ascribe variation in

occurrences to underlying environmental variation

(e.g. Phillips et al., 2003). Occurrences of Neotropical

rainforest palms are strongly influenced by environ-

mental conditions (Svenning, 2001a), including light

(Svenning, 1999b, 2002), edaphic factors (Clark et al.,

1995, 1999; Svenning, 1999a), topography (Svenning,

1999a; Vormisto et al., 2004a), and soil water

conditions (Kahn and de Castro, 1985; Kahn and de

Granville, 1992), but are also influenced by dispersal

processes (Fragoso, 1997; Svenning and Balslev,

1999; Svenning, 2001b; Charles-Dominique et al.,

2003; Vormisto et al., 2004b). Here we investigate the

spatial distribution and environmental preferences of

the west Amazonian piassaba palm, Aphandra natalia

(Balslev and Henderson) Barfod, which produces stiff

leaf sheath fibres that are extracted and sold on the

local markets for broom production.

A. natalia was discovered relatively recently

(Balslev and Henderson, 1987) in Ecuador and

subsequently several aspects of its ecology, ethnobo-

tany and economic botany in Ecuador have been

described (Pedersen and Balslev, 1990, 1992; Barfod,

1991; Pedersen, 1992, 1996; Barfod and Uhl, 2001;

Pedersen and Skov, 2001; Macıa, 2004). It is the basis

for a blooming peasant based extraction and an artisan

industry of brooms that supplies the entire Ecuador and

parts of Peru with stiff-fibred brooms and brushes.

Apart from producing fibres, Aphandra is used for a

multitude of other purposes, including its edible fruits,

its inflorescence for cattle fodder, its leaves for thatch,

weaving, blowgun darts, stuffing in dart canisters, and

much more. It is well known that extraction of tropical

rainforest products is not always sustainable (e.g. Hall

and Bawa, 1993; Phillips, 1997; Ticktin, 2004) and

indeed, this appears to be true for several palms

(Svenning and Macıa, 2002; Siebert, 2004). Never-

theless extraction of Aphandra fibres appeared sustain-

able in Ecuador where local management strategies

maintained or favoured Aphandra by cutting back

surrounding vegetation, protecting Aphandra when

forest was cleared for agriculture, allowing spontaneous

regeneration in abandoned pastures, and abstaining

from completely defoliating individuals harvested for

fibres (Pedersen, 1992, 1996; Pedersen and Balslev,

1992). In contrast, along Pastaza and Urituyacu rivers in

Peru, where we travelled for the present investigation,

extraction of Aphandra fibres appeared to be highly

destructive. The stems were felled or the entire crown of

leaves was cut to harvest leaves for thatch or fibres, and

local villagers claimed that its local abundance was

declining (personal observation).

Despite its cultural and economic importance little

is known about the general resource availability,

distribution and environmental preferences of Aphan-

dra, and in particular the palm has not been subject of

scientific study in Peru, where it appears to be widely

distributed and of considerable local economic

importance. Because the occurrence of Aphandra is

hardly documented in the literature or in herbaria, we

chose to study it along the Pastaza and Urituyacu

rivers where one of us (CG) had previously observed

its presence. There it occurs in varying abundances in

old growth terrace forest (the only never-flooded

forests in the region) but it is absent in the region’s

other forest types such as forests growing in

permanently wet back-swamps or in temporarily

wet flood plain. Our specific objective was to estimate

the extent to which environmental conditions and

dispersal influence the spatial distribution of Aphan-

dra in old growth terrace forest along the Pastaza and

Urituyacu rivers in Amazonian Peru. Such knowledge

will be crucial for improving the management of this

source of non-timber forest products.

2. Study area and methods

2.1. Study area

Our fieldwork was done in the northern part of the

Peruvian Amazon near Sungachi (0384500S 7682500W)

on the Pastaza River and Velasco (0484400S 7583800W),

Reforma (0483100S 7584200W), and Guineal (0482200S7584900W) on the Urituyacu River (Fig. 1). We chose to

work near these four villages because previous

reconnaissance had shown that fibres were extracted

by their inhabitants. Both Pastaza and Urituyacu flow

through the Pastaza fan, a large wetland area built of

dissected alluvial deposits of Quaternary origin (Neller

Fig. 1. Research sites (A, B) and shape of transects (a, b, c) used to study populations of Aphandra natalia in Amazonian Peru. The small circles

numbered 1–10 show the distribution of the line transects. The positions of cross transects are marked with a cross and numbered I–VI. The shape

of cross-transect II is as in (a), of cross-transects I, III, IV, and V as in (b), and of cross-transect VI as in (c).

et al., 1992; Puhakka et al., 1992) containing volcanic

material (Rasanen et al., 1992). The Pastaza River

originates in the high Andes of Ecuador, whereas the

Urituyacu has its entire drainage basin within the

Pastaza fan. Both rivers discharge into the Maranon

River. Rivers that flow through the Pastaza–Maranon

foreland basin cannot easily be classified in the simple

scheme of ‘‘white water’’, ‘‘clear water’’, and ‘‘black

water’’ rivers as defined by Sioli (1984), because their

waters are of mixed nature (Puhakka et al., 1992). The

topography of the area is flat and lies about 130 m.a.s.l.

The landscape is a mosaic of levees with young forest

with abundant pioneer tree species such as Cecropia,

floodplains with tall forest, back-swamps dominated

by Mauritia flexuosa L.f., and occasional ‘‘islands’’ of

low rarely flooded terraces with tall forest. The climate

is tropical wet with mean annual precipitation at

2900 mm and an average temperature at 26 8C(measured in Iquitos, www.worldclimate.com).

2.2. Study Species

The dioecious A. natalia (Balslev and Henderson)

Barfod is related to the vegetable ivory palms

(Phytelephas spp.) and is distributed in the western

Amazon basin from Ecuador to the Brazilian state of

Acre (Barfod, 1991; Borchsenius et al., 1998; Barfod

and Uhl, 2001). It grows on terra firma and along rivers

on low terraces that may be briefly inundated after

heavy rainfall (Pedersen, 1992; Pedersen and Balslev,

1992). Aphandra is a single-stemmed sub-canopy

palm with a trunk up to 11 m tall (Henderson et al.,

1995) and it has up to 24 leaves that may reach 12.5 m

in length (personal observation). The seeds, which

reach a size of 5 cm � 3.5 cm (Balslev and Hender-

son, 1987), are dispersed over short distances by

squirrels (Sciurus spp.) and agouties (Dasyprocta

spp.), while rivers and humans may disperse them over

larger distances (Pedersen, 1992).

2.3. Field measurements

Field data were collected during July and August

2003 in 17 transects placed with the help of villagers

using two approaches: (1) six transects were located in

major stands of Aphandra within 5 km of the four

villages; in each stand we placed cross transects

reaching from one end of the stand to the other and

perpendicularly from one side of the stand to the other,

supplemented with additional lines when the stand had

an irregular shape (see Fig. 1). Along these lines we

placed 5 m � 5 m subplots for every 10 m. (2) We

placed eleven 500 m line transects with contiguous

5 m � 5 m subplots in forests that were not targeted to

include Aphandra stands but fulfilled the following

criteria: they would be on never or very rarely flooded

ground (local name ‘‘restinga’’ equal to ‘‘low terrace

forest’’), the forest would be undisturbed, not logged,

not secondary, and as close as possible to the villages

which turned out to be within 5 km. In each subplot we

assigned all individuals to one of two size classes:

immature (seedlings and juveniles which showed no

signs of reproduction) and mature (individuals

showing signs of current or past reproduction, and

sterile individuals of similar size, i.e. with leaves

>8 m long). We recorded the following environmental

descriptors for each subplot: (1) soil moisture

classified as dry or inundated/muddy. (2) Canopy

openness quantified by canopy scope (Brown et al.,

2000) into low light [largest visible canopy gap filled

�1 mark on the scope] and high light [largest visible

gap in the canopy filled >1 mark]. For the line transect

subplots we also noted (3) forest phase as whether the

subplot was placed within a gap created by one or

more fallen trees or if it was placed in closed forest, (4)

trails as absent or present, (5) subplot inclination using

a SUUNTO clinometer, and (6) relative elevation, in

meters above the lowest point of the transect.

Inclination and relative elevation had high skewness

and were therefore log (x + 1)-transformed. One line

transect (#7) was excluded when analyzing the spatial

distribution of Aphandra because it contained only

three individuals, but included when calculating

general abundances.

2.4. Statistical analysis

Logistic regression (Trexler and Travis, 1993;

Manel et al., 1999; Hosmer and Lemeshow, 2000) is

widely used for modelling species distributions, but

assumes that observations are independent and hence

ignores the spatial autocorrelation typical of distribu-

tional data (Smith, 1994). Autologistic regression

models, i.e. logistic regression models which include

one or more neighbour variables describing the state of

the response variable in the samples neighbouring

each sample, explicitly accounts for spatial auto-

correlation in the data and hence provides a more

correct approach (Smith, 1994; Wu and Huffer, 1997;

also cf. Svenning, 2001b; Svenning and Skov, 2002).

Hence, we analysed the distribution of Aphandra

(immature, mature, or both stages, recorded as

presence/absence in each subplot) using this approach.

We used the maximum pseudolikelihood method,

where the autologistic model is implemented using

standard logistic regression routines with neighbour

variables being treated as conventional covariates (Wu

and Huffer, 1997).

Explanatory variables were the six environmental

descriptors and site. Site was analysed as a nominal

variable with each village as a category.

Site � inclination and site � relative elevation inter-

actions were also included in the analyses. In the

autologistic regressions presence of individuals of the

same size class in neighbouring subplots was used as

additional explanatory variables. For the cross

transects subplots at distances of 10 and 20 m and

for the line transects subplots of distances of 5, 10, 15,

and 20 m were used for deriving neighbour variables.

In addition, for both transect types, presence of mature

individuals in a given plot was included as an

explanatory variable for the occurrence of immature

individuals in that plot. All neighbour variables were

used as presence/absence data. There was no strong

correlations among the environmental explanatory

variables or between these and site (results not shown).

When building the autologistic regression models

we used a stepwise variable selection procedure (cf.

Trexler and Travis, 1993; Svenning, 1999b; Brito et al.,

1999; Manel et al., 1999) with ‘‘P-enter’’ and ‘‘P-

leave’’ at 0.05 and a log likelihood x2-test to evaluate

the significance of each explanatory variable. In each

step the variable with the lowest P-value was entered.

We note that it is preferable to use the log likelihood

rather than the Wald x2-test, since the latter may show

aberrant behaviour (Trexler and Travis, 1993; Hosmer

and Lemeshow, 2000; Agresti, 2002). The area under

receiver operating characteristic curve (AUC), and the

uncertainty coefficient, U (also called McFadden’s rho

squared) (Hensher and Johnson, 1981) were evaluated

as measures of the explanatory power of each model.

According to Hensher and Johnson (1981) U-values of

0.2–0.4 are extremely good fits. All analyses were

computed in JMP 5.0 (SAS Institute, 2002).

Table 1

Population densities (individuals per ha) of Aphandra natalia

measured in cross-transects inside stands used for extraction and

in line-transects located in old growth terrace forest within 5 km of

each of four villages along the Pastaza and Urituyacu rivers in

Amazonian Peru

Village Sungachi Velasco Reforma Guineal Mean � S.D.

(A) Inside Aphandra stands

Immature 642 980 786 547 739 � 188

Mature 133 140 72 39 96 � 49

Total 775 1120 859 586 835 � 222

(B) Terrace forest in general

Immature 764 580 406 277 507 � 212

Mature 8 22 28 17 19 � 8

Total 772 602 434 294 526 � 207

Table 2

Autologistic regressions of Aphandra (immature, mature, or both stages)

transects located in old-growth terrace forest within 5 km of each of four

Stage Variabley

(A) Inside Aphandra stands

Immature +Immature present in 10 m plot*

Mature Site*

+Mature present in 10 m plot*

Both +Ind. present in 10 m plot***

(B) Terrace forest in general

Immature Site**

�Moisture***

+Relative elevationNS

Site � relative elevation*

+Mature present in plot***

+Immature present in 5 m plot****

+Immature present in 10 m plot*

+Immature present in 15 m plot****

Mature �Inclination*

+Mature present in 5 m plot**

+Mature present in 10 m plot*

+Mature present in 20 m plot**

Both Site*

+Relative elevationNS

Site � relative elevation*

�Moisture***

+Ind. present in 5 m plot****

+Ind. present in 10 m plot***

+Ind. present in 15 m plot****

Environmental and neighbour variables included in final models were chos

the final models are shown.

y The sign indicates if the probability of occurrence increases with incre* P < 0.05.

** P < 0.01.*** P < 0.001.

****P < 0.0001.

3. Results

Aphandra was restricted to slightly elevated

terraces where it mostly occurred in relatively well-

defined stands. Abundances inside these stands were

on average 739 (�188 S.D.) immature and 96 (�49)

mature individuals per hectare with some site

differences (Table 1). In old growth terrace forest

that was not targeted for high Aphandra abundances,

its average abundances were 507 (�212) immature

and 19 (�8) mature individuals per hectare.

The most important environmental predictor of

Aphandra occurrence in the terrace forest outside of

the dense stands was soil moisture (Table 2) with

occurrences (A) in cross-transects inside major stands and (B) line-

villages along the Pastaza and Urituyacu rivers in Amazonian Peru

U AUC

0.011* 0.547

0.061*** 0.681

0.029*** 0.573

0.236**** 0.800

0.104**** 0.729

0.235**** 0.796

en using a stepwise procedure with P-enter and P-leave at 0.05; only

asing values of the predictor (+) or not (�). NS: non significant.

Aphandra preferring drier areas. Furthermore the

analyses suggested that Aphandra preferred relative

elevated and, in the case of mature individuals, flat

parts of the landscape (Table 2). There were site

differences in Aphandra abundance and its relation-

ship to relative elevation both within stands and in the

general terrace forest (Table 2). No environmental

predictors were significant within stands.

Neighbour variables were important predictors of

Aphandra occurrence both within stands and in

general, being particularly strong in the latter case.

Neighbour variables at 5 and 10 m distance were

significant predictors of abundance of immature,

mature and both stages (Table 2). In line transects

neighbour variables were important and distribution of

immature individuals furthermore depended on the

occurrence of mature individuals (Table 2).

Based on the U and AUC values Aphandra

occurrences were less predictable within stands than

in the terrace forests in general (Table 2).

4. Discussion

4.1. Environmental preferences of A. natalia

A. natalia preferred the drier parts of the low

terraces along Pastaza and Urituyacu rivers, with soil

moisture and to a lesser extent relative elevation being

important predictors of its occurrence. Drainage

conditions depend on the soil type and topography

and both are known to influence the occurrence of

palm species and the structure of palm communities in

Amazonia (e.g. Kahn and de Granville, 1992;

Svenning, 1999a; Vormisto et al., 2004a). In central

Amazonia, Kahn and de Castro (1985) used the

hydromorphic condition of the soil to divide palm

communities into three zones on well-drained,

transition and poorly drained soils, respectively.

4.2. Clumped distribution of A. natalia

The many and highly significant neighbour variables

included in the final autologistic regression models

indicate that Aphandra exhibits a highly clumped

distribution pattern that is unaccounted for by the

environmental factors. Tropical trees including palms

often have clumped distributions, apparently reflecting

dispersal limitation (e.g. Condit et al., 2000; Svenning,

2001b; Souza and Martins, 2002; Barot and Gignoux,

2003). Apart from dispersal limitation and environ-

mental conditions historical events and density depen-

dence (Levine and Murrell, 2003) may also cause

clumped distributions. However, the large size of

Aphandra seeds (Balslev and Henderson, 1987), and

the short distances over which its main dispersal agents

(squirrels, Sciurus sp.; agouties, Dasyprocta sp.), are

likely to disperse the seeds make it plausible that it is

dispersal limited (Forget, 1990, 1992; Peres and Baider,

1997; Peres et al., 1997; Silva and Tabarelli, 2001;

Silvius and Fragoso, 2003; Charles-Dominique et al.,

2003). Furthermore, in our data the presence of mature

Aphandra individuals in a subplot was a good predictor

of the presence of immature individuals, supporting the

notion that dispersal limitation is causing the clumping,

(cf. Silva and Tabarelli, 2001 and Svenning, 2001b for

Neotropical rain forest palms). The clumped distribu-

tion of Aphandra may also be enhanced by its dioecism

(Heilbuth et al., 2001). However, it was not possible to

include sex in the analyses because most adult-sized

plants did not exhibit signs of current or past

reproduction.

4.3. Patterns within Aphandra stands versus in the

general terrace forest landscape

Within large Aphandra stands occurrences were

unrelated to the environmental descriptors and exhib-

ited less strong neighbourhood effects. Furthermore,

the U and AUC values were higher for the general

terrace forest transects than for the Aphandra stand

transects (Table 2). These differences may partly reflect

the higher number of explanatory variables for the line

transects, but probably mainly reflects higher variability

in environmental conditions and Aphandra densities

when sampling the old growth terrace forest in general.

5. Conclusion

The occurence of A. natalia depended on hydrol-

ogy, but also exhibited site effects and strong clumping

independent of environmental factors and site. Hence,

Aphandra may be absent or have low densities in

much of the landscape due to dispersal rather than

environmental constraints. In our study area the most

common use of Aphandra was collecting its edible

fruits (personal observation), which in other species has

been shown to reduce recruitment (Hall and Bawa,

1993). Furthermore, harvesting of Aphandra leaves for

thatch or fibres was destructive in the Pastaza–

Urituyacu area, involving the killing of the harvested

individuals (personal observation). According to local

informants Aphandra used to be more abundant in the

study area and they explained this decline in abundance

as a result of over-exploitation (personal observation).

Given the evidence found in the present study for strong

dispersal limitation of Aphandra, we would expect that

if it were exterminated in some areas by over-harvesting

then its recolonization into those areas would be a slow

process. On the other hand the finding that the main

constraint on the occurrence of Aphandra in the terrace

forest habitat is probably dispersal rather than

environmental conditions suggests that planting or

even just seed addition of Aphandra would be likely to

greatly increase the availability of this important source

of non-timber forest products.

Acknowledgements

We thank Guillermo Criollo for assistance in the

field, the villagers in Sungachi, Velasco, Reforma and

Guineal for their hospitality, and Flemming Nørgaard

for making the map (Fig. 1). This study was funded by

a grant from the Danida Research Council to Henrik

Balslev (104.Dan.8-764). The participation of Signe

Normand and Thomas Boll was made possible by

grants from WWF-Novo Nordisk Biodiversity Fund

(project no. 47) and the Faculty of Science, University

of Aarhus. Our work on factors controlling biodi-

versity in the western Amazon is supported by a grant

from the Danish Natural Science Research Council to

Henrik Balslev (21-01-0617). Jaana Vormisto was

supported by a EU Marie Curie Fellowship (EUK2-

CT-2001-50013) and Jens-Christian Svenning by a

Steno postdoctoral stipend from the Danish Natural

Science Research Council (21-01-0415).

References

Agresti, A., 2002. Categorical Data Analysis, second ed. John Wiley

& Sons, New York.

Balslev, H., Henderson, A., 1987. A new Ammandra (Palmae) from

Ecuador. Syst. Bot. 12, 501–504.

Barfod, A.S., 1991. A monographic study of the subfamily Phyte-

lephantoideae (Arecaceae). Opera Bot. 105, 1–73.

Barfod, A.S., Uhl, N.W., 2001. Floral development in Aphandra

(Arecaceae). Am. J. Bot. 88, 185–195.

Barot, S., Gignoux, J., 2003. Neighbourhood analysis in the savanna

palm Borassus aethiopum: interplay of intraspecific competition

and soil patchiness. J. Veg. Sci. 14, 79–88.

Borchsenius, F., Pedersen, H.B., Balslev, H., 1998. Manual to the

Palms of Ecuador. AAU Reports 37.

Brito, J.C., Crespo, E.G., Paulo, O.S., 1999. Modelling wildlife

distributions: logistic multiple regression vs overlap analysis.

Ecography 22, 251–260.

Brown, N., Jennings, S., Wheeler, P., Nabe-Nielsen, J., 2000. An

improved method for the rapid assessment of forest understorey

light environments. J. Appl. Ecol. 37, 1044–1053.

Charles-Dominique, P., Chave, J., Dubois, M.-A., De Granville, J.-

J., Riera, B., Vezzoli, C., 2003. Colonization from of the under-

storey palm Astrocaryum sciophilum in a pristine rain forest of

French Guiana. Global Ecol. Biogeogr. 12, 237–248.

Clark, D.B., Palmer, M.W., Clark, D.A., 1999. Edaphic factors and

the landscape-scale distributions of tropical rain forest trees.

Ecology 80, 2662–2675.

Clark, D.A., Clark, D.B., Sandoval, M.R., Castro, C.M.V., 1995.

Edaphic and human effects on landscape-scale distribution of

tropical rain forest palms. Ecology 76, 2581–2594.

Condit, R., Ashton, P.S., Baker, P., Bunyavejchewin, S., Gunatilleke,

S., Gunatilleke, N., Hubbell, S.P., Foster, R.B., Itoh, A., LaFran-

kie, J.V., Lee, H.S., Losos, E., Manokaran, N., Sukumar, R.,

Yamakura, T., 2000. Spatial patterns in the distribution of

tropical tree species. Science 288, 1414–1418.

Forget, P.M., 1990. Seed-dispersal of Vouacapoua americana (Cae-

salpiniaceae) by caviomorph rodents in French Guiana. J. Trop.

Ecol. 6, 459–468.

Forget, P.M., 1992. Seed removal and seed fate in Gustavia superba

(Lecythidaceae). Biotropica 24, 408–414.

Fragoso, J.M.V., 1997. Tapir-generated seed shadows: scale-depen-

dent patchiness in the Amazon rain forest. J. Ecol. 85, 519–

529.

Hall, P., Bawa, K., 1993. Methods to assess the impact of extraction

of non-timber tropical forest products on plant populations.

Econ. Bot. 47, 234–247.

Heilbuth, J.C., Ilves, K.L., Otto, S.P., 2001. The consequences of

dioecy for seed dispersal: modelling the seed-shadow handicap.

Evolution 55, 880–888.

Henderson, A., Galeano, G., Bernal, R., 1995. Field Guide to the

Palms of the Americas. Princeton University Press, Princeton,

New Jersey.

Hensher, D.A., Johnson, L.W., 1981. Applied Discret-Choise Mod-

elling. Halsted Press, a division of John Wiley & Sons, New

York.

Hosmer Jr., D.W., Lemeshow, S., 2000. Applied Logistic

Regression, second ed. John Wiley & Sons, New York.

Hubbell, S.P., 2001. The Unified Neutral Theory of Biodiversity and

Biogeography. Princeton University Press, Princeton and

Oxford.

Kahn, F., de Castro, A., 1985. The palm community in a forest of

central Amazonia. Brazil. Biotropica 17, 210–216.

Kahn, F., de Granville, J.-J., 1992. Palms in Forest Ecosystems of

Amazonia. Ecological Studies 95. Springer-Verlag, Berlin.

Levine, J.M., Murrell, D.J., 2003. The community-level conse-

quences of seed dispersal patterns. Annu. Rev. Ecol. Evol. Syst.

344, 549–574.

Macıa, M.J., 2004. Multiplicity in palm uses by the Huaorani of

Amazonian Ecuador. Bot. J. Linn. Soc. 144, 149–159.

Manel, S., Dias, J.-M., Ormerod, S.J., 1999. Comparing discrimi-

nant analysis, neural networks and logistic regression for pre-

dicting species distributions: a case study with a Himalayan river

bird. Ecol. Model. 120, 337–347.

Neller, R.J., Salo, J.S., Rasanen, M.E., 1992. On the formation of

blocked valley lakes by channel avulsion in upper Amazon

foreland basins. Z. Geomorph. N. F. 36, 401–411.

Pedersen, H.B., 1992. Uses and management of Aphandra natalia

(Palmae) in Ecuador. Bull. Inst. Fr. Etudes Andines 21, 741–753.

Pedersen, H.B., 1996. Production and harvest of fibres from Aphan-

dra natalia (Palmae) in Ecuador. For. Ecol. Manage. 80, 155–161.

Pedersen, H.B., Balslev, H., 1990. Ecuadorean palms for agrofor-

estry. AAU Reports 23, 1–122.

Pedersen, H.B., Balslev, H., 1992. The economic botany of Ecua-

dorean palms. In: Plotkin, M., Famolare, L. (Eds.), Sustainable

Harvest and Marketing of Rainforest Products. Island Press,

Washington, DC, pp. 173–191.

Pedersen, H.B., Skov, F., 2001. Mapping palm extractivism in

Ecuador using pair-wise comparison and bioclimatic modelling.

Econ. Bot. 55, 63–71.

Peres, C.A., Baider, C., 1997. Seed dispersal, spatial distribution and

population structure of Brazilnut trees (Bertholletia excelsa) in

southeastern Amazonia. J. Trop. Ecol. 13, 595–616.

Peres, C.A., Schiesari, L.C., Dias-Leme, C.L., 1997. Vertebrate

predation of Brazil-nuts (Bertholletia excelsa, Lecythidaceae),

an agouti-dispersed Amazonian seed crop: a test of the escape

hypothesis. J. Trop. Ecol. 13, 69–79.

Phillips, O.L., 1997. The changing ecology of tropical forests.

Biodiversity Conserv. 6, 291–311.

Phillips, O.L., Vargas, P.N., Monteagudo, A.L., Cruz, A.P., Zans,

M.-E.C., Sanchez, W.G., Yli-Halla, M., Rose, S., 2003. Habitat

association among Amazonian tree species: a landscape-scale

approach. J. Ecol. 91, 757–775.

Pitman, N.C.A., Terborgh, J.W., Silman, M.R., Nunez, V.P., Neill,

D.A., Ceron, C.E., Palacios, W.A., Aulestia, M., 2001. Dom-

inance and distribution of tree species in upper Amazonian terra

firme forests. Ecology 82, 2101–2117.

Puhakka, M., Kalliola, R., Rajasilta, M., Salo, J., 1992. River types,

site evolution and successional vegetation patterns in Peruvian

Amazonia. J. Biogeogr. 19, 651–665.

Rasanen, M.E., Neller, R., Salo, J., Jungner, H., 1992. Recent and

ancient fluvial deposition systems in the Amazonian foreland

basin. Peru. Geol. Mag. 129, 293–306.

SAS Institute, 2002. JMP Introductory Guide, Version 5. SAS

Institute Inc, Cary, North Carolina.

Siebert, S.F., 2004. Demographic effects of collecting rattan cane

and their implications for sustainable harvesting. Conserv. Biol.

18, 424–431.

Silva, M.G., Tabarelli, M., 2001. Seed dispersal, plant recruitment

and spatial distribution of Bactris acanthocarpa Martius (Are-

caceae) in a remnant of Atlantic forest in northeast Brazil. Acta

Oecol. 22, 259–268.

Silvius, K.M., Fragoso, J.M.V., 2003. Red-rumped Agouti (Dasy-

procta leporina) home range use in an Amazonian forest:

implications for the aggregated distribution of forest trees.

Biotropica 35, 74–83.

Sioli, H., 1984. The Amazon and its main affluents: hydrography,

morphology of the river courses, and river types, In: Sioli, H.,

(Ed.), The Amazon. Limnology and Landscape Ecology of

Mighty Tropical River and its Basin. Dr. W. Junk, The Hague,

127–165.

Smith, P.A., 1994. Autocorrelation in logistic regression modelling

of species’ distributions. Global Ecol. Biogeogr. Lett. 4, 47–

61.

Souza, A.F., Martins, F.R., 2002. Spatial distribution of an under-

growth palm in fragments of the Brazilian Atlantic Forest. Plant

Ecol. 164, 141–155.

Svenning, J.-C., 1999a. Microhabitat specialization in a species-rich

palm community in Amazonian Ecuador. J. Ecol. 87, 55–

65.

Svenning, J.-C., 1999b. Recruitment of tall arborescent palms in the

Yasunı National Park, Amazonian Ecuador: are large treefall

gaps important? J. Trop. Ecol. 15, 355–366.

Svenning, J.-C., 2001a. On the role of microenvironmental hetero-

geneity in the ecology and diversification of Neotropical rain-

forest palms (Arecaceae). Bot. Rev. 67, 1–53.

Svenning, J.-C., 2001b. Environmental heterogeneity, recruitment

limitation and the mesoscale distribution of palms in a tropical

montane rain forest (Maquipucuna, Ecuador). J. Trop. Ecol. 17,

97–113.

Svenning, J.-C., 2002. Crown illumination limits the population

growth rate of a understorey palm (Geonoma macrostachys,

Arecaceae). Plant Ecol. 159, 185–199.

Svenning, J.-C., Balslev, H., 1999. Microhabitat-dependent recruit-

ment of Iriartea deltoidea (Arecaceae) in Amazonian Ecuador.

Ecotropica 5, 69–74.

Svenning, J.-C., Macıa, M.J., 2002. Harvesting of Geonoma macro-

stachys Mart. leaves for thatch: an exploitation of sustainability.

For. Ecol. Manage. 167, 251–262.

Svenning, J.-C., Skov, F., 2002. Mesoscale distribution of under-

storey plants in temperate forest (Kalø Denmark): the impor-

tance of environment and dispersal. Plant Ecol. 160, 169–

185.

Ticktin, T., 2004. The ecological implications of harvesting non-

timber forest products. J. Appl. Ecol. 41, 11–21.

Trexler, J.C., Travis, J., 1993. Nontraditional regression analyses.

Ecology 74, 1629–1637.

Tuomisto, H., Ruokolainen, K., Aguilar, M., Sarmiento, A., 2003.

Floristic patterns along a 43-km long transect in an Amazonian

rainforest. J. Ecol. 91, 743–756.

Valencia, R., Foster, R.B., Villa, G., Condit, R., Svenning, J-C.,

Hernandez, C., Romoleroux, K., Losos, E., Magard, E., Balslev,

H., 2004. Tree species distribution and local habitat variation in

the Amazon: large forest plot in eastern Ecuador. J. Ecol. 92,

214–229.

Vormisto, J., Tuomisto, H., Oksanen, J., 2004a. Palm distribution

patterns in Amazonian rainforests: What is the role of topo-

graphic variation? J. Veg. Sci. 15, 485–494.

Vormisto, J., Svenning, J.-C., Hall, P., Balslev, H., 2004b. Diversity

and dominance in palm (Arecaceae) communities in terra

firme forests in the western Amazon basin. J. Ecol. 92, 577–

588.

Wu, H., Huffer, F.W., 1997. Modelling the distribution of plant

species using the autologistic regression model. Environ. Ecol.

Stat. 4, 49–64.