Explaining variation in Brazil nut fruit production

Page 1

This article was published in an Elsevier journal. The attached copyis furnished to the author for non-commercial research and

education use, including for instruction at the author’s institution,sharing with colleagues and providing to institution administration.

Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third party

websites are prohibited.

In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further information

regarding Elsevier’s archiving and manuscript policies areencouraged to visit:

http://www.elsevier.com/copyright

Page 2

Author's personal copy

Explaining variation in Brazil nut fruit production

Karen A. Kainer a,b,*, Lucia H.O. Wadt c, Christina L. Staudhammer a

a School of Forest Resources and Conservation, University of Florida, Gainesville, FL 32611, USAb Center for Latin American Studies, Tropical Conservation and Development Program, University of Florida, Gainesville, FL 32611, USA

c Centro de Pesquisa Agroflorestal do Acre (Embrapa Acre), BR-364, Km 14, Rio Branco, Acre 69901-108, Brazil

Received 8 November 2006; received in revised form 17 May 2007; accepted 20 May 2007

Abstract

Brazil nut is widely recognized as the cornerstone of the Amazonian extractive economy. Tight linkages between Brazil nut production,

regional income, and intact mature forests have thrust this species into focus as a key component of Amazonian conservation and income

generation strategies. Nonetheless, a comprehensive synthesis of factors explaining Brazil nut fruit production variation is lacking. We aimed to

address this knowledge gap, asking: (1) What are the rates and annual variation of Bertholletia excelsa fruit production at individual and population

levels? (2) What factors explain B. excelsa production variation, focusing on spatial and temporal variables, diameter at breast height (dbh), crown

attributes, liana loads, and soil attributes? and (3) Does liana cutting affect fruit production?

Our model explained 73% of the variability in fruit production by 140 trees over 5 years. Dbh alone explained >50%, and trees in the middle

diameter range (100 cm � dbh < 150 cm) produced best. Top-producing trees were loosely grouped spatially. Cation exchange capacity and P also

explained production variation. The liana load variable was only weakly significant for explaining production variation, and there were no direct

significant effects of liana cutting. Cutting, however, did improve crown forms—another highly significant model variable. This, coupled with a

weakly significant interaction of liana cutting by year, suggests that liana cutting could improve production over time.

In any given year, approximately 25% of the trees produced 72% of the total population production. Annual variation of fruit production at the

individual level was relatively high, and at the population level was extremely low, with annual production departing from average only in a year of

delayed and reduced rainfall. These results coupled with low synchronicity of fruiting, confirm long-term observations of harvesters that Brazil nut

populations exhibit relatively constant fruit production.

Results suggest that fruit productivity at the individual tree level could be increased, possibly through experimental liana cutting and/or P

amendments. At the population level, Brazil nut collectors could potentially increase efficiency by concentrating collection efforts on consistently

high producers and possible production ‘‘hotspots’’.

# 2007 Published by Elsevier B.V.

Keywords: Amazon; Bertholletia excelsa; Nut production; Liana; Non-timber forest product; NTFP; Nutrients; Seed production; Tropical forest

1. Introduction

Brazil nut (Bertholletia excelsa Humb. and Bonpl.) is widely

recognized as the cornerstone species of the Amazonian

extractive economy, and is the only internationally traded nut

collected almost entirely from natural populations in mature

forests (Clay, 1997). Historically, collection and processing

have been centered in eastern Amazonia where it has been

commercially exploited since the mid-1600s (de Souza, 1963).

Over the last three decades, however, large-scale collection has

shifted westward to the tri-border region of Brazil, Bolivia, and

Peru (Peres et al., 2003) where commercial harvest and

marketing represent a major income source for rural and peri-

urban collectors and processors (Stoian, 2005). Because of this

central importance of Brazil nut to the regional economy,

household income shortfalls across this entire supply chain

were attributed most frequently to insufficient Brazil nut

production in any given locality or year (Stoian, 2005). This

socioeconomic link between Brazil nut production and income

can also have clear consequences for conservation as

documented in Peru, where Brazil nut income alone was

insufficient to thwart deforestation by small producers who

concurrently participated in more environmentally destructive

income-generating activities like cattle ranching and intensive

www.elsevier.com/locate/foreco

Forest Ecology and Management 250 (2007) 244–255

* Corresponding author. Tel.: +1 352 846 0833; fax: +1 352 846 1277.

E-mail address: [email protected] (K.A. Kainer).

0378-1127/$ – see front matter # 2007 Published by Elsevier B.V.

doi:10.1016/j.foreco.2007.05.024

Page 3

Author's personal copy

agriculture (Escobal and Aldana, 2003). Indeed, these tight

linkages between Brazil nut production, regional income, and

intact mature forests have thrust this species into focus as a key

component of Amazonian conservation and income generation

strategies (Clay, 1997; Ortiz, 2002). Nonetheless, a compre-

hensive synthesis of factors explaining Brazil nut fruit

production variation is lacking. Indeed, there is a paucity of

quantitative information on fruit production and sources of

production variation of most tropical non-timber forest product

(NTFP) fruits, nuts, and seeds of commercial interest. This is

surprising given that NTFPs are considered an important

component of developing country economies and tropical

forest conservation strategies (Freese, 1997; Alexiades and

Shanley, 2004). The main aim of our study is to address this

knowledge gap in production variation using Brazil nut as a

model species, while providing scientific insights to the larger

ecological question of fruit production variation.

1.1. Fruit production variation

Variation in fruit and seed production in forest trees has been

attributed to exogenous and endogenous factors. Climate is a

major contributor to crop size variation, leading to a resource

matching hypothesis that annual seed production patterns

‘‘match’’ annual rainfall distribution and/or some key seasonal

temperature variable. Koenig and Knops (2000) reviewed 443

data sets of multiple northern hemisphere species within three

families, and found that annual rainfall and temperature

variation were important, and identified latitudinal patterns

linked to climate. Kelly and Sork (2002) came to similar

conclusions in their global analysis of 570 data sets. Seedfall

variation of 108 tropical species – a forest type poorly

represented in all previous analyses – corroborated these

latitudinal patterns; while production variation was substantial

for these tropical species, it was significantly smaller than for

higher latitudes (Wright et al., 2005).

Nonetheless, most polycarpic woody plants adhere to

alternating supra-annual schedules of low and high production

years (Herrera et al., 1998; Koenig and Knops, 2000; Kelly and

Sork, 2002). This provides strong evidence that rainfall and

temperature, which are more normally distributed and much less

variable than production (Koenig and Knops, 2000), are not the

only determinants of fruit production variation. Disease,

herbivory, or adverse weather such as high winds or storms,

particularly during key phenological events such as pollination or

fruit development, can reduce flowering or seed set (Stephenson,

1981). Liana presence has been shown to reduce fruit production

of trees (Stevens, 1987; Kainer et al., 2006), and proximate

disturbances such as fire also have been implicated (Layne and

Abrahamson, 2004). Controlled experiments in monoculture

orchards (Ponder and Jones, 2001), agroforestry systems (Jones

et al., 1995), and natural forests (Vaughton, 1991) clearly

demonstrate the importance of nutrient and light availability as

explanatory variables of production variation.

Endogenous factors such as life cycle rhythms play a central

role in fruiting. Reproduction is a function of accumulated

resources, and is positively correlated to total tree height

(Layne and Abrahamson, 2004) and diameter (Ares and Brauer,

2004). Successful reproduction in 1 year, however, is often

followed (or ‘‘switched’’) at the expense of vegetative growth

(Harper and White, 1974; Kelly and Sork, 2002).

Fruiting variation has strong implications for fauna

dependent on fruit and seed resources for their caloric and

nutritional values. In general, woody plants produce more

expensive fruits than herbaceous plants (Sutherland, 1986), and

the expected return for this costly expenditure is dispersal.

Scatterhoarding (burial of one or a few seeds at many widely

scattered sites), a dispersal method typically restricted to trees

producing large nutritious nuts (Jansen, 2003), has several

advantages for the plant: facilitation of transportation away

from the source plant, nut burial for seed protection and

improved germination, and caching in microenvironments

more favorable for seedling establishment (Vander Wall, 2001).

In return, these cached food supplies are available to

scatterhoarders for periods of scarcity, such as tropical dry

seasons or temperate zone winters (Jansen, 2003). Cached seeds

not reclaimed by scatterhoarders (e.g., forgotten or due to

predator mortality) are well positioned for germination.

1.2. Management significance of production variation

For tree fruits with high economic value, maximum levels of

fruiting and minimal annual variation can be central to

commercial success. In these cases, plant breeding and

selection programs, resource amendments (nutrients, water),

and other controls to direct resource partitioning to fruits and

minimize production variation are commonly practiced within

domesticated populations (Acquaah, 2002). Very little research,

however, has focused on understanding or manipulating fruit

production of natural populations of species with commercial

and/or subsistence import, such as Brazil nut and other NTFPs

(Ticktin, 2004). For example, most Brazil nut collectors in our

study region cut lianas, stating that this practice increases nut

yield (personal observations, KK and LW), but to our

knowledge, effects of liana cutting have not been quantified.

Understanding demographic parameters, including fruit pro-

duction, also is essential for fostering population persistence,

and such ecological knowledge coupled with management

interventions could enhance fruit production levels for meeting

both economic and demographic goals. Such interventions

advocated include: reducing localized resource competition,

including measures such as liana cutting, tree girdling, and/or

felling; enrichment planting; tending or favoring new recruits;

alternating geographic areas of harvest; setting harvest quotas;

and recruitment monitoring (Peters, 1996; Peres et al., 2003;

Ticktin, 2004).

Using Brazil nut (Bertholletia excelsa Humb. and Bonpl.) as

a model species, we aim to contribute to scientific under-

standing of fruit production variation and provide ecological

information that could contribute to improved management of

natural populations of commercially important NTFPs.

Production variation of tropical species in general is poorly

documented (Herrera et al., 1998; Kelly and Sork, 2002;

although see Wright et al., 2005), and very few studies have

K.A. Kainer et al. / Forest Ecology and Management 250 (2007) 244–255 245

Page 4

Author's personal copy

explored manipulation of natural populations for enhancing

production. We focused on three research questions: (1) What

are the rates and annual variation of B. excelsa fruit production

at individual and population levels? (2) What factors explain B.

excelsa production variation, focusing on spatial and temporal

variables, diameter at breast height (dbh), crown attributes,

liana loads, and soil attributes? and (3) Does liana cutting affect

fruit production?

2. Study species

At maturity, B. excelsa is a very large, emergent tree. Adults

can be long-lived, with three individuals >45 cm dbh radio-

carbon-dated as older than 650 years (Vieira et al., 2005).

Individual trees may or may not be totally deciduous, with floral

buds emerging on axillary spikes at the apex of recent leaf flushes

produced toward the end of the dry season (Mori and Prance,

1990; Maues, 2002). Individuals rely on out-crossing for seed

development (O’Malley et al., 1988), and are pollinated

principally by large bodied bees, especially Euglossinae, capable

of lifting the hood of the zygomorphic flower (Prance, 1976;

Nelson et al., 1985). The hard, large round fruit (10–16 cm)

reaches maturity 14 months after successful pollination, falling

during the rainy season (January and February in our study

region). The 8–26 large (�3.5–5 cm � 2 cm) seeds (or nuts)

remain inside this woody fruit until extraction by humans or other

seed predators/dispersers; agoutis (Dasyprocta spp.) are the

almost exclusive scatterhoarding dispersers. The nut reward is a

nutritious food source packed with 14–17% protein and 65–70%

oil (Prance and Mori, 1979).

3. Study site

Research was conducted in Extractive Reserve (RESEX)

Chico Mendes, a conservation unit 10–118 south of the equator

in the eastern region of the state of Acre, Brazil. The area has

gently undulating topography; a pronounced dry season,

typically from June to August (IMAC, 1991); and the dominant

vegetation is humid, moist tropical forests (Holdridge, 1978).

Average temperature is approximately 25 8C (ZEE, 2000), and

brief intrusions of frigid southern air are common during the dry

season, dropping temperatures to 12 8C. Soils are classified

under the Brazilian system as Argissolos (ZEE, 2000) or red-

yellow podzolics. In our study region, those soils under the

Solimoes geological formation are influenced by volcanic

material, likely ashes, deposited during soil genesis by Andean

winds (Gama et al., 1992). Thus, while aluminium and its

associated acidity dominate other weathered Amazonian

podzolic soils, Solimoes podzolics tend to be relatively new

soils with higher silt content, nutrient stocks, and cation

exchange capacity (Gama et al., 1992).

Research efforts were concentrated in a 420-ha extractivist

landholding (colocacao) in the southeastern portion of the

reserve (Colocacao Rio de Janeiro in Seringal Filipinas) (see

Wadt et al., 2005 for study site map). The resident family

collects Brazil nuts annually, but like others in the region (Ortiz,

2002), would not likely harvest from trees that typically

produce less than 20 fruits. A 2001–2002 inventory of 568 B.

excelsa trees �10 cm dbh in this landholding revealed a

population density of 1.35 trees ha�1 and average diameters of

86.1 � 45.0 cm dbh (Wadt et al., 2005).

4. Methods

Study trees were selected by initially placing all 404

reproductively mature B. excelsa adults (�50 cm dbh) from the

previous inventory into 10 diameter classes, and then randomly

selecting 140 trees to include representatives within each

diameter class and four liana load categories: (1) no lianas in

crown, (2) lianas present, but �25% crown covered, (3) 25–

75% crown covered, and (4) >75% crown covered. Selected

trees were then revisited in the field to exclude individuals with

overlapping crowns or fruit fall areas. Finally, the geospatial

location of the 140 selected trees was determined through GPS

and plotted, confirming fairly representative spatial distribution

of the sample across the study landscape.

4.1. Fruit counts

Production was measured in the last 15 days of February,

after fruit fall, for 5 consecutive years (2002–2006). Fruit

production was quantified as the number of fruits harvested

from the ground by the research team. These count values do

not reflect 100% total counts of fruit production per tree. A

small count error is attributed to the very few fruits remaining in

the crowns at the time of fruit collection. Scatterhoarding

Dasyprocta spp. and other minor seed predators, however, are a

larger source of error since they open and/or remove fallen

fruits prior to the late-February counts. We used two methods to

estimate this count discrepancy.

In September 2004, a period after premature fruit abortion

had occurred and deciduous B. excelsa individuals were

leafless, we surveyed our 140-tree study population. We

counted absolute numbers of fruits on trees in which crowns

were entirely visible (a minority of all trees) and fruits were not

too numerous to impede accurate counts (�100 fruits, N = 19).

When comparing these absolute crown counts with subsequent

on-the-ground harvest from the same trees, we found a strong

correlation (r = 0.89) between crown and ground counts. On

average, we collected only 62% of those fruits counted in the

crown, suggesting a 38% predator removal rate. Since canopy

counts were conducted after the exocarp had hardened, we

assumed little to no macaw predation of immature canopy fruits

as reported by Trevedi et al. (2004). A second method involved

enumerating depressions with traces of the non-woody exocarp

and/or mesocarp on the ground below each tree canopy. These

depressions are created when the heavy fruits fall from the

approximately 50 m tree canopy, and become visible when the

predator removes the fruit from this microsite. Each depression

count was sometimes augmented when additional predator-

opened fruits clearly not associated with the depressions were

encountered. Based on this method applied to the entire study

population during ground counts, predators removed 7.9 and

6.1% of the total 2005 and 2006 production, respectively. Using

K.A. Kainer et al. / Forest Ecology and Management 250 (2007) 244–255246

Page 5

Author's personal copy

similar methods, Zuidema and Boot (2002) estimated that 7%

of all production was not harvested by collectors in Bolivia.

Though far from error-free, we believe that crown counts

more accurately reflect count discrepancies because of the

possibly large error associated with depression counts given

that predators could easily remove a substantial number of

fruits that left no ground trace. Repeated fruit counts that

consider daily or weekly fruit removal by predators are

warranted for highly accurate estimates of absolute fruit

production—a procedure that was not possible in our study.

Nonetheless, for consistency, all subsequent discussion of fruit

production rates reflect fruits counted on the ground without

adjustments for these potential sources of error.

4.2. Tree and soil attributes

Crown attributes (crown form, cross-sectional area, and

canopy position) were assessed for each of the 140 trees. Crown

form categorizations was adapted from Synnott (1979), and

included: (1) complete or irregular circle (perfect or good); (2)

half-crown (tolerable); (3) less than half-crown (poor); and (4)

one or a few branches (very poor). Each crown was originally

evaluated in July 2001, and again in July 2005. Tree crown

cross-sectional area was determined in July 2002 by measuring

maximum crown diameter and a second diameter at right angles

to the maximum axis, followed by application of an ellipsoid

formula. Crown position was scored as: (1) dominant (full

overhead and side light); (2) co-dominant (full overhead light);

(3) intermediate (some overhead or side light); or (4)

suppressed (no direct light).

In October 2002, leaves were collected from a random sample

of 24 study crowns, controlling to include trees at different fruit

production levels, dispersed locations on the 420-ha landscape

and liana cutting treatment (described below). Leaves were dried

at 105 8C for 72 h, ground with a Wiley mill, and passed through

a 1 mm sieve. For P, K, Ca, and Mg, samples were digested in a

2:1 nitroperchloric acid, with concentrations determined using

the metavanadate colorimetry method and absorption spectro-

photometry (P) or atomic absorption spectrophotometry (K, Ca,

and Mg). Nitrogen was analyzed by the Kjeldahl method as

modified by Mills and Jones (1996).

To determine soil properties, a composite of six soil cores

were excavated beneath the crown of each of the 140 trees in

July 2002. Samples were extracted below the litter layer at two

depths: 0–10 and 10–20 cm. Two replicates of each composite

were dried for 4 days at 65 8C, and passed through a 2 mm

stainless steel sieve. Soil pH was measured at a 1:2.5 soil to

water ratio. Extractable P and K were processed using a dilute

double acid extraction (Mehlich-1), with concentrations

determined colorimetrically using the molybdate blue method

for P, and flame emission spectrophotometry for K. Exchange-

able Ca and Mg were measured after extracting cations with a

1N KCl solution, with concentrations determined through

atomic absorption spectrophotometry. To determine total

potential acidity, H+ + Al3+ was extracted with a buffered

solution of calcium acetate at pH 7, and then titrated with 0.1N

NaOH. Cation exchange capacity (CEC) was calculated as the

sum of this potential acidity plus base cations. Oxidizable

organic carbon was determined on soils passed through a 1 mm

screen, ground in a porcelain mortar, and then digested in a

potassium dichromate acid medium with external heat. Organic

matter was estimated by multiplying this value by 1.75. All

soils analyses were conducted at the Soils Laboratory of

Embrapa Acre, Brazil (EMBRAPA, 1997).

4.3. Liana cutting

Lianas were cut from 80 of the 140 trees in July 2002. Each

liana was cut near the ground and at approximately 2 m from

ground level. Since B. excelsa fruits take 14 months to mature,

impacts would only be detectable with 2004 production data.

Thus, fruit and nut production in 2002 and 2003 were clearly not

associated with any liana cutting impacts, while data for 2004,

2005, and 2006 could potentially be affected by liana cutting.

4.4. Data analysis

We modelled fruit production as a function of the liana cutting

treatment, with tree and soil attributes acting as covariates, using

SAS software for all data analysis (Version 9.1). Since the

number of possible predictor variables was large, a principal

components analysis (PCA) was first performed with all

quantitative explanatory variables to illuminate relationships

and show redundancies (PROC PCA) (Appendix A). Two-way

correlations between all variables, including the dependent

variable, were calculated using PROC CORR to verify PCA

results and reveal further relationships between annual fruit

production and possible predictor variables.

A mixed model was constructed first with all quantitative tree

variables (dbh and crown size), class variables (crown position,

crown form class, liana load, and liana cutting treatment), and the

group of soil variables determined from the PCA to have high

explanatory power. Because fruit production data were collected

annually for each of 5 years on 140 trees located in a contiguous

forest area, data were both spatially and temporally correlated. A

mixed effects approach, accounting for the repeated measures

structure of the data ensured correct formation of the error

covariance matrix and proper tests of significance. Year of

measurement was treated as a repeated measurement, but not as a

linear one, as it was known that fruit production by year did not

follow a linear trend. Location (UTM easting and northing) was

included so that spatial location of trees could be incorporated

into the error covariance matrix. Variograms were constructed to

reveal patterns of spatial dependency in the data (PROC

VARIOGRAM), and several candidate models of spatial

covariance (e.g., exponential, power, and Gaussian) were

incorporated and tested in the mixed model (Schabenberger

and Pierce, 2001). Because a previous study showed a quadratic

relationship between dbh and fruit production (Wadt et al., 2005),

dbh2 was included in the model. Using PROC MIXED, model

parameters were found via maximum likelihood, and several

error covariance structures were tested.

Model results were compared using Akaike’s Information

Criteria (AIC, Akaike, 1973), and by visual examination to test

K.A. Kainer et al. / Forest Ecology and Management 250 (2007) 244–255 247

Page 6

Author's personal copy

normality and homoscedastic model residuals. Second- and

third-order interactions, and later covariates, were included in

the first iterations of the model, and those that did not improve

(lower) AIC were dropped sequentially as recommended by

Burnham and Anderson (2002). We also explored model

building by conventional methods of sequentially dropping

non-significant interactions and covariates based on P-values.

In this case, we used a significance level of a = 0.10 rather than

the usual arbitrary 0.05 level (Johnson, 1999; Burnham and

Anderson, 2002) to allow all possible significant effects to

remain for the purpose of building the best explanatory model

(Bancroft, 1968, p. 8). Since we found almost identical results

using this second method, we only report results using the AIC

method. Because a primary purpose of data analysis was to

develop an explanatory model of fruit production with and

without liana cutting treatments, the indicator variable for liana

cutting was considered fundamental. As such, it was retained in

the model without regard to its significance level (Neter et al.,

1996, p. 354).

Using methods detailed in Kelly and Sork (2002), fruit

production variability at the individual level (xCVi) was

determined by calculating the mean of the coefficient of

variation of fruits per tree of each tree over 5 years, and then

calculating the mean of these individual tree coefficients of

variation. Fruit productionvariation at the population level (CVp)

was measured by the coefficient of variation of the mean number

of fruits per tree of each tree over 5 years. Following Snook et al.

(2005), synchronicity of fruiting by the study population (xPCC)

was determined by calculating the Pearson correlation coeffi-

cient of all possible pairs of trees in the sample, and then

calculating the mean of those correlation coefficients.

Finally, we also assessed relevant relationships not

specifically tested in the mixed model. We computed Pearson’s

correlation coefficients for relevant variable pairs, and to

specifically test liana cutting effects on crown form, we

modelled the change in crown form class controlling for initial

crown form class.

5. Results

5.1. Fruit production rates

Without adjusting for the liana cutting treatment, mean

number of B. excelsa fruits produced per tree per year was 66.2

(S.D. = 98.5, N = 681). Of the 140 B. excelsa trees studied,

approximately 40% produced, on average, fewer than

20 fruits year�1. Correspondingly, in any given year, approxi-

mately 25% of the trees (those that produced > 200 fruits

year�1) produced, on average, 72% of the total annual

production.

5.2. Fruit production variation explained

The final mixed model explained 73% of the variability in

number of fruits produced. The analysis incorporated

significant spatial correlation into the error covariance structure

with a power model and included significant random effects for

dbh. Significant fixed effects were detected for year, dbh2,

crown form, liana load, and two soil attributes (Table 1).

5.2.1. Annual variation

The calculated population variability (CVp) was relatively

low at 0.1974. Significant production variation within the study

population was detected by year (P < 0.01) (Table 1), however,

with Scheffe’s test revealing that 2006 was the only year

departing from average (Fig. 1, column 1), producing almost

50% fewer fruits than the preceding 4 years.

The annual individual variation in number of fruits was high

(xCVi = 0.8894), with the widest range represented by one tree

that produced 140 fruits in 1 year and 801 in the subsequent.

There were, however, individual trees that departed from that

high variation. Eleven trees of the 140 studied were consistently

among the best 5 producers in any given year, and 1 tree made

this top-5 list every year. At the other extreme, 5 trees produced

no fruits in any of the 5 years, and 14 others produced no fruits

in 3 of 5 years and very few fruits (�17) in the other 2.

Significant correlations (P < 0.001) revealed that observations

of fruit production in alternate years were more highly

correlated than those of consecutive years, implying that the

number of fruits produced by a tree in the second previous year

was a better predictor of current fruit production than that of the

immediately previous year. Further exploration revealed that

12% of the trees had a strong biannual production pattern over

the 5-year period (years of high production followed by years of

low), while 46% approximated this pattern, deviating from this

Table 1

Results of the final mixed model explaining fruit production variation

Variable DF F-value Pr > F

Year 4 4.38 0.0017

Dbh 1 7.96 0.0054

Dbh2 1 4.59 0.0337

Crown form 3 3.21 0.0225

Liana load 3 2.38 0.0718

Liana cut 1 0.07 0.7970

Year � liana cut 4 2.06 0.0846

CEC 1 7.19 0.0081

P 1 4.19 0.0424

Final model coefficients are available in Appendix B.

Fig. 1. Mean (�standard error) B. excelsa fruit production by year (column 1)

and by 50 cm diameter class (columns 2–4).

K.A. Kainer et al. / Forest Ecology and Management 250 (2007) 244–255248

Page 7

Author's personal copy

sequence in 1 year only. Overall synchronicity of fruiting

(considering all possible pairs of all individuals and all years)

by the study population was very low (xPCC = 0.0565).

5.2.2. Spatial variation

The spatial proximity of trees across the 420-ha landscape was

relevant in explaining production variation. Although initial

variograms of the data did not reveal any obvious spatial pattern,

patterns could easily be masked by treatment effects. Variograms

of the residuals for a mixed model that did not account for spatial

dependencies showed asymptotically increasing variation

between residuals as the distance between observations

increased, which is indicative of significant spatial correlation.

After introducing a spatial covariance matrix for better model fit,

we found a significant relationship between fruit counts and tree

proximity (P < 0.0001), and the 11 most consistently top-

producing trees were grouped spatially (Fig. 2).

5.2.3. Variation by dbh

Fruit production was also explained by dbh (P < 0.01)

(Table 1) in a quadratic manner (Fig. 3). Grouping trees into

three 50-cm diameter classes revealed that those in the middle

range (100 cm � dbh < 150 cm) were consistently the best

producers (Fig. 1, columns 2–4).

5.2.4. Variation by crown form

Crown form was important to explain variation in fruit

production (P < 0.05) (Table 1). Scheffe’s means comparisons

revealed that trees with the best crown forms (complete or

irregular circles) were significantly better fruit producers than

those in all other crown form classes (Fig. 3).

5.2.5. Variation by liana load

The liana load variable may explain some of the production

variance, although statistically this variable was not highly

significant (P < 0.10) (Table 1). Of those trees that produced

the least (0–20 fruits tree�1), the majority had crowns that were

more than 25% covered by lianas; in contrast, those that

produced the most fruits (>200 fruits tree�1) rarely had liana

load levels of that magnitude.

The mixed model suggested an interaction effect of liana

cutting treatment by year, although not highly significant

statistically (P < 0.10) (Table 1). In 2005, those trees from

which lianas were not cut produced more fruits than those that

received the liana cutting treatment. In 2006, this rank changed,

whereby those trees that received the liana cutting treatment

were comparatively superior producers.

Significant changes in crown form over the 2001–2005

period were detected for some trees liberated from lianas, when

initial 2001 crown form condition was considered (P = 0.0003).

Specifically, changes were detected in trees with the poorest

initial crown forms (few branches only) (P < 0.0001) (Fig. 4),

indicating that those trees responded differently to liana cutting

than trees with complete or irregular circles, or half-crowns

(P < 0.01).

5.2.6. Variation by soil attributes

Of the soil variables subjected to a PCA (Table 2), potential

acidity (H+ + Al3+) and CEC had nearly identical loadings—

relationships corroborated with simple correlations. There were

no differences in soil property values between sampling depths

except for exchangeable Ca. Therefore, with the exception of

Ca, all values were averaged by tree and collapsed into one

depth (0–20 cm) for analysis in the mixed model.

CEC and levels of extractable P contributed to explaining

fruit production variation (Table 1) (P < 0.01 and <0.05,

respectively). CEC levels had a positive relationship with fruit

production, while extractable P levels showed a weakly

negative one.

Foliar nutrient concentrations were not included in the

model because they were only assessed on a subsample of 24

study trees. Of those measured, only foliar P at

1.68 � 0.49 g kg�1 (x� S:D:) was significantly correlated

with number of fruits produced in 2002, the year in which

foliar samples were collected (r = 0.414, P = 0.045). Similarly,

Fig. 2. Spatial relationship of study trees. Size of circles corresponds to the

mean number of fruits produced by each individual tree over the study period

(2002–2005), with shaded circles highlighting 11 trees that were consistently

among the best 5 producers for any given year.

Fig. 3. Number of fruits produced by trees characterized by four crown forms.

Data points represent all trees measured over the 5-year study period (N = 681).

K.A. Kainer et al. / Forest Ecology and Management 250 (2007) 244–255 249

Page 8

Author's personal copy

only P foliar levels were correlated with its corresponding (0–

10 cm depth) P soil value (r = 0.478, P < 0.05).

6. Discussion

Our model explained over 70% of the variation in Brazil

nut fruit production over a 5-year period, identifying

multiple factors implicated in regulating B. excelsa fruit

production. This information could contribute to fostering

population persistence and improving commercial yield of

this increasingly managed species, while providing scientific

insights into the larger ecological question of what causes

fruit production variation. To the extent that fruit production

of other NTFPs is regulated by the same factors, our study

may indicate management directions for those species as

well.

6.1. Fruit production rates

Based on our 140-tree data set observed over 5 years, B.

excelsa produced 66 � 98 fruits tree�1 year�1 (x� S:D:). The

only other published study that quantified fruit production

reported much higher production levels and slightly lower

variation at two sites (184 � 122, N = 40, over 5 years;

139 � 104, N = 53, over 3 years) (Zuidema and Boot, 2002).

This study was similar in that it was also conducted in western

Amazonia, in Bolivia; however, it was based on smaller sample

sizes and included only reproductive individuals from which

Brazil nut collectors typically harvest. The variation in

production of this commercially-harvested Bolivian population

would be, by definition, less than our sampled Brazilian

population which included non-producing individuals that

would not be visited during a commercial harvest. Nonetheless,

standard deviations from these three data sets suggest ample

variation in production between individual trees in any given

year, an observation shared by Brazil nut collectors and

researchers (Viana et al., 1998; Ortiz, 2002). Indeed, overall

production of our studied population is attributed to relatively

few trees; 1/4 of the trees produced almost 3/4 of the annual

production. Zuidema (2003) reported similar findings from

Bolivia, noting that the majority of trees in the analyzed

populations produced very few fruits.

6.2. Annual production variation

Annual variation of fruit production at the population level

(CVp = 0.20) was extremely low—far from Kelly’s (1994)

criteria for mast fruiting of >1. It is also substantially lower

than the median value of 1.10 for 108 tropical species, almost

half of which were trees measured over 15 years (Wright et al.,

2005). Indeed, none of their species had a CVp < 0.20. Perhaps

a longer term B. excelsa data set (>the 5 years measured) would

detect greater variation (Kelly, 1994; although see Kelly and

Sork, 2002), but our low CVp coupled with low synchronicity of

fruiting (xPCC = 0.0565) and long-term observations of more

or less continuous production by Brazil nut collectors, confirm

that Brazil nut is not a masting species (sensu Janzen, 1976).

Rather, B. excelsa fruit production variation follows a pattern

predicted by Kelly and Sork (2002) and empirically tested by

Wright et al. (2005): tropical species exhibit more constant fruit

production than temperate species. A departure from average

annual production in our study was only observed in 2006,

which coincided with delayed onset of the rainy season and

prolonged drought just prior to harvest. Based on a 7-year study

of 40 trees, Zuidema (2003) also reported a significant drop in

annual fruit production coinciding with reduced rainfall in an El

Nino year. We found a highly significant, though weak,

correlation (r = 0.10, P = 0.008) between number of fruits

collected and total rainfall during the previous 5 months of the

dry season (May–September), considering all years. Rainfall

data were collected at a meteorological station approximately

200 km from our study site (UFAC, 2006), and we would

anticipate a stronger correlation with more localized rainfall

data.

Fig. 4. Crown form changes following a liana cutting treatment. The first crown

form evaluation was conducted in 2001, and the second in 2005.

Table 2

Mean (�standard deviation) values of soil properties at 0–10 and 10–20 cm

depths from soil samples collected under study tree canopies

Soil property Soil sample depth

0–10 cm 10–20 cm

pH 3.97 (�0.32) 4.01 (�0.28)

Organic matter (%) 2.42 (�0.70) 2.03 (�0.59)

Ca (cmol + kg�1) 0.43 (�0.75) 0.29 (�0.57)

Mg (cmol + kg�1) 0.43 (�0.21) 0.34 (�0.19)

K (cmol + kg�1) 0.18 (�0.15) 0.15 (�0.14)

P (cmol + kg�1) 5.92 (�0.77) 3.59 (�1.55)

H+ + Al3+ (cmol + kg�1) 7.35 (�2.28) 6.98 (�1.81)

CEC (cmol + kg�1) 7.70 (�2.30) 7.28 (�1.79)

Organic carbon (%) 1.38 (�0.40) 1.16 (�0.34)

K.A. Kainer et al. / Forest Ecology and Management 250 (2007) 244–255250

Page 9

Author's personal copy

No single pattern of annual fruit production was detected at

the individual level. Previously, Zuidema (2003) had reported

that Brazil nut individuals with high production in 1 year were

often preceded and followed by years with substantially lower

production. A little over 10% of our study trees exhibited such a

strong biannual pattern, but another 46% (almost half)

approximated it, deviating from this sequence in 1 year.

Resource switching, whereby plants move resources into and

away from reproduction in successive years (Norton and Kelly,

1988), is suggested by these results, but we present no data as to

where resources might be allocated (e.g., growth) in ‘‘off-

production’’ years. We also observed that 11 trees of the 140

studied were consistently among the best 5 producers in any

given year, and 1 tree made this top-5 list every year. Similarly,

Ortiz (2002) reported that Brazil nut collectors in Peru have

observed that highly productive B. excelsa trees typically

produce better in relation to their local conspecifics,

independent of year. These results of consistently good or

poor reproductive performances of individual plants apparently

unrelated to site conditions or tree size (within reproductive

limits) suggest genotypic influences over production (Vander

Wall, 2001).

6.3. Explaining production variation

In addition to rainfall embedded in year, what factors

contributed to the significant temporal and spatial variation in

B. excelsa nut production encountered in our study? The great

variation in annual production within B. excelsa individuals

(CVi = 0.95) indicates that other explanatory variables at the

scale of the individual were important. Dbh was the most

critical variable, explaining >50% of fruit production variation

(partial R2 = 0.565). Trees in the middle diameter range of our

study population (100 cm � dbh < 150 cm) were better pro-

ducers than both smaller and larger trees (Fig. 1, columns 2–4).

Most trees produce their best seed crop in middle age, which

may last from decades to centuries, followed by decline and

senescence (Harper and White, 1974). While B. excelsa

diameter is not perfectly correlated with age (Vieira et al.,

2005), it seems reasonable to use dbh as a proxy for observing

tree development over time. The quadratic nature of the

relationship observed in our study (Fig. 3) contrasts to that of

Zuidema and Boot (2002), who reported a positive linear

relationship between dbh and production, although based on a

smaller and constrained sample. For Swietenia macrophylla,

Snook et al. (2005) also found that fruit production increased

with diameter, although they reported a pattern whereby

production continued to increase with increasing dbh,

apparently throughout tree life.

Crown attributes (area, position, and form) also played a role

in fruit production variation. Crown area was strongly

correlated with dbh (r = 0.53, P < 0.0001), and preliminary

model tests demonstrated that it was a slightly better predictor

of fruit production than diameter (similar to Zuidema, 2003),

although both variables explained fruit production well. We

retained dbh in the mixed model, however, because predictive

differences between dbh and crown area were very small, and to

facilitate data interpretation because dbh is a common field

measurement. Crown position or ability to access light

resources is clearly an important determinant of fruit

production, but since almost all B. excelsa study trees had a

co-dominant or dominant canopy position, this variable fell out

of the model. While Zuidema (2003) states that trees below

canopy level do not seem to produce fruits, we observed six

trees (4%) that were in an intermediate canopy position and

produced fruits. Finally, crown form was an important

explanatory variable in our study (Table 1), such that better

fruit producers consistently had the best crown forms.

A previous analysis of the first 2 years of our data set (prior

to liana cutting), specifically testing liana load and fruit

production, clearly demonstrated an association between liana

presence and reduced Brazil nut fruit counts (Kainer et al.,

2006). Similarly, Stevens (1987) had demonstrated negative

impacts of lianas on Bursera simaruba fecundity. Thus, we

expected positive fruit production effects from liana cutting.

Nonetheless, single variable effects from our current mixed

model did not support this expectation (Table 1). We did find

some evidence that effects of liana cutting on production were

changing, based on the weakly significant interaction effect of

liana cutting by year (P < 0.10). We also observed that liana

cutting improved crown form and/or circumvented mortality of

trees with very poor crown forms (Fig. 4). These observations

confirm that branch regrowth following liana liberation had

begun in some of our study trees, and that crown forms (a highly

significant explanatory variable of B. excelsa fruit production)

of Brazil nut trees liberated of lianas could improve over time.

Both long-term liana consequences, such as branch modifica-

tion (Stevens, 1987), and short-term, such as light interception

(Grauel and Putz, 2004), have been speculated upon for

suppressing host tree fruit production. We conclude that liana

impacts (and liana cutting effects) are complex and may be

attributable to a series of events at multiple levels over time.

Soils and nutrient availability have been cited as important

factors in B. excelsa fruit production (Viana et al., 1998;

Zuidema, 2003); however, these potential effects had not been

tested previously. In analyzing soil chemical characteristics,

only two variables were implicated in fruit production

(Table 1). Not surprisingly, CEC was positively associated

with increased production, and CEC in our study site was

higher compared to other Amazonian podzolics (Lathwell and

Grove, 1986). Base cation levels (Ca, Mg, and K) were

relatively high, and similar to another primary forest in the

region (McGrath et al., 2001), though not significant in our

model. Extractable P was implicated, but this statistical

significance may not be meaningful since it was slightly

negatively correlated with production, and coefficients of

variation for production in years prior to and after soil sampling

were very small (<0.1). Foliar P, however, a better indicator of

plant nutrient status than soil stocks, was fairly strongly

positively correlated with number of fruits produced. These

results are too limited to reveal how P is cycling through the

soil–plant system, but suggest that the relationship between P

and B. excelsa fruit production merits further study. P is often

cited as limiting tropical forest productivity (Jordan, 1985;

K.A. Kainer et al. / Forest Ecology and Management 250 (2007) 244–255 251

Page 10

Author's personal copy

Cuevas and Medina, 1988; McGrath et al., 2001), and is a key

component of reproductive tissues (Marschner, 1995) such as

fruits and nuts. Experimental studies of multiple apple varieties,

revealed that addition of P in 1 year of production significantly

increased flowering and fruit set in the subsequent year (Nielsen

et al., 1990).

We did not directly evaluate predator and pollinator

availability or genotypic variation—factors potentially impor-

tant in B. excelsa fruit production. Discrepancies between

number of fruits collected off the ground and absolute fruit

numbers from crown and soil depression counts reveal strong

interactions with the agouti predator/disperser. These interac-

tions must affect fruit production as measured in our study, and

possibly fruit production variation. Ortiz (2002) speculated that

the principal determinant of low B. excelsa productivity is

pollination deficiency, perhaps due to smoke from increasing

forest fires or lower bee visitation in trees within forest

fragments or isolated in converted pastures. While we did not

study pollinator activity, our study site was located within a

largely unfragmented conservation unit exposed to minimal

smoke. Five trees in our sample exhibited individual

characteristics of good producers (good dbh and crown form,

no to few lianas), but produced no or few fruits (�6) throughout

the 5-year period, implicating at least some genetic influences

(Vander Wall, 2001).

Finally, our data suggest a spatial pattern of production

variation across the 420-ha landscape, with highly productive

trees found in loose groups (Fig. 2). Ortiz (2002) reported the

existence of Brazil nut ‘‘hot spots’’ of greater tree densities, and

bigger and better producers. Reasons for these patterns remain

speculative, but soils and genetic factors influenced by

disperser activity are plausible explanations.

6.4. Management implications

The more or less continuous reproduction pattern of B.

excelsa at the population level documented in our study has

certainly contributed to the high commercial success of Brazil

nut over the decades, and consequently, its importance for

Amazonian conservation. Results from our study suggest

several avenues that could further minimize production

variation and maximize levels of fruiting. Brazil nut collectors

and processors across the Amazon basin depend on B. excelsa

fruit production as part of their cash income and food security,

and interventions suggested from our study are intended to

advance those economic goals, while respecting dynamics of

the broader ecological community and ecosystem.

First, our findings imply that productivity at the individual

tree level could be increased. Our data show that liberating trees

of lianas improves crown forms, and in time, may increase fruit

production. Study results suggest a positive relationship

between foliar P and fruit production. Experimenting with P

amendments may prove fruitful, perhaps testing effectiveness

and practicality of utilizing on-site manure sources. Grazing

animals partition all P into dung (Hedley et al., 1995), and cattle

are increasingly common in extractivist livelihood systems in

many parts of Amazonia.

Second, Brazil nut collectors could potentially augment the

pool of trees from which they collect and perhaps increase

collection efficiency. In our study region, Brazil nut harvest is

based on a trail system originally blazed for collecting rubber.

While Brazil nut collectors are very familiar with the large

reproductive B. excelsa trees on and near these trails, there is

limited preoccupation with systematically seeking and tending

younger trees entering reproductive maturity, which could

increase the harvest pool. Additionally, over time, this

traditional trail system could preferentially concentrate harvest-

ing from the largest trees (>150 cm dbh), whereas we found

that those of intermediate diameters (100–150 cm dbh) are the

most productive. Our study also shows that some trees (11 of

140) were consistently super-producers, while a full 40%

produced fewer than 20 fruits year�1 on average, a quantity that

would not likely be collected given current harvest practices. To

achieve maximum efficiency, harvesters could concentrate

collection efforts on these consistently high producers that also

seem to be spatially aggregated. These production ‘‘hot spots’’

might also be targeted for incorporating new trees into the

harvest regime. Poor producers could be targeted for individual

tree interventions highlighted above.

Differences in ground fruit counts, reported throughout our

study as fruit production, versus our measured crown counts or

fruit depressions, clearly indicate that many fruits are ‘‘lost’’ to

predators, rather than human-harvested. While more intensive

extraction would certainly increase commercial yield, caution

should be exercised to not exceed a theoretically unsustainable

harvest threshold as documented in populations very inten-

sively harvested over decades (Peres et al., 2003). Overharvest

could also adversely affect the broader forest community (such

as agouti dynamics) and ecosystem functions (such as nutrient

cycling) (Ticktin, 2004).

Finally, although we did not find any one pattern of annual

fruit production at the individual tree level, harvesters could

begin tracking production levels of individual trees over time,

revealing patterns that might increase collection efficiency and

predictability of crop yield. Indeed, many Brazil nut harvesters

who have been collecting from the same trees for years, have

developed a mental map of high production trees, and often can

fairly accurately predict average individual tree production

(Wadt et al., 2005). Tapping into this local knowledge and

integrating it with science could enhance our understanding of

B. excelsa autecology and augment Brazil nut productivity. It

may also further our understanding of the complex interactions

between Dasyprocta spp., Homo sapiens, and B. excelsa, which

are poorly documented in the literature. Understanding of

production and production variation coupled with interventions

to achieve management goals can hopefully further advance

Amazonian forest conservation efforts in which Brazil nut is a

keystone element.

Acknowledgements

This research was supported by grants from FINEP/MCT/

CNPq in Brazil, The William and Flora Hewlett Foundation in

the U.S., and the International Science Foundation, Sweden,

K.A. Kainer et al. / Forest Ecology and Management 250 (2007) 244–255252

Page 11

Author's personal copy

through a grant to Dr. Wadt. CNPT/IBAMA in Brazil gave

permission to conduct the research in RESEX Chico Mendes.

We also thank Paulo Rodrigues de Carvalho for his superior

field assistance, Pieter Zuidema and Wendell Cropper for

helpful comments on earlier manuscript drafts, and Paulo Wadt

and Deborah McGrath for nutrient insights. Finally, we are

most grateful to Valderi and Maria Alzenira who graciously

shared their forest home.

References

Acquaah, G., 2002. Horticulture: Principals and Practices, third ed. Prentice

Hall, Upper Saddle River, NJ, USA.

Akaike, H., 1973. Information theory as an extension of the maximum like-

lihood principle. In: Pretrov, B.N., Csaki, F.F. (Eds.), Proceedings of the

Second International Symposium on Information Theory. Akademiai

Kiado, Budapest, Hungary, pp. 267–281.

Alexiades, M.N., Shanley, P. (Eds.), 2004. Productos Forestales, Medios de

Subsistencia y Conservacion: Estudios de Caso Sobre Sistemas de Manejo

de Productos Forestales No Maderables. Volumen 3—America Latina.

Center for International Forestry Research (CIFOR), Bogor, Indonesia.

Ares, A., Brauer, D., 2004. Growth and nut production of black walnut in

relation to site, tree type and stand conditions in south-central United States.

Agroforestry Syst. 63, 83–90.

Bancroft, T.A., 1968. Topics in Intermediate Statistical Methods. Iowa State

University Press, Ames, IA, USA.

Burnham, K.P., Anderson, D.R., 2002. Model Selection and Multi-Model

Inference: A Practical Information-Theoretic Approach, second ed.

Springer-Verlag, New York, NY, USA.

Clay, J.W., 1997. Brazil nuts: the use of a keystone species for conservation and

development. In: Freese, C.H. (Ed.), Harvesting Wild Species: Implications

for Biodiversity Conservation. The Johns Hopkins University Press, Balti-

more, MD, USA, pp. 246–282.

Cuevas, E., Medina, E., 1988. Nutrient dynamics within Amazonian forests. II.

Fine root growth, nutrient availability and leaf litter decomposition. Oeco-

logia 76, 222–235.

de Souza, A.H., 1963. Castanha do Para: estudo botanico, quımico e tecnolo-

gico. Estudos Tecnicos, 23. Ministerio da Agricultura-Servico de Infor-

macao Agrıcola, Rio de Janeiro, RJ, Brazil.

EMBRAPA (Empressa Brasileira de Pesquisa Agropecuaria), 1997. Manual de

Metodos de Analise de Solos. Segunda edicao. Centro Nacional de Pesquisa

de Solos, Rio de Janeiro, RJ, Brazil.

Escobal, J., Aldana, U., 2003. Are nontimber forest products the antidote to

rainforest degradation? Brazil nut extraction in Madre de Dios, Peru. World

Dev. 31, 1873–1887.

Freese, C.T. (Ed.), 1997. Harvesting Wild Species: Implications for Biodi-

versity Conservation. The Johns Hopkins University Press, Baltimore, MD,

USA.Ap

pen

dix

A.

Pri

nci

pa

lco

mp

on

ent

loa

din

gs

an

dp

erce

nt

va

ria

tio

nex

pla

ined

by

qu

an

tita

tiv

etr

ee,

loca

tio

n,

an

dso

ilv

ari

ab

les

Pri

nci

pal

com

ponen

t

NE

Yea

rD

BH

Cro

wn

size

C (0–10

cm)

C (10–20

cm)

K (0–10

cm)

K (10–20

cm)

P (0–10

cm)

P (10–20

cm)

Mg

(0–10

cm)

Mg

(10–20

cm)

Ca

(0–10

cm)

Ca

(10–20

cm)

pH

(0–10

cm)

pH

(10–20

cm)

CE

C

(0–10

cm)

CE

C

(10–20

cm)

H+

+A

l3+

(0–10

cm)

H+

+A

l3+

(10–20

cm)

Var

iati

on

expla

ined

(%)

1�

0.0

50.1

40.0

00.1

10.1

20.2

40.2

6�

0.0

3�

0.0

7�

0.0

9�

0.0

4�

0.2

5�

0.2

3�

0.1

3�

0.0

9�

0.3

1�

0.2

80.3

50.3

50.3

60.3

628

2�

0.1

5�

0.2

70.0

00.2

10.1

4�

0.0

40.0

50.3

70.3

00.1

90.3

60.3

50.4

00.0

40.1

60.0

5�

0.0

10.2

20.2

00.1

70.1

510

30.0

00.2

40.0

00.2

00.2

00.3

80.2

8�

0.0

6�

0.1

70.1

4�

0.0

3�

0.0

3�

0.0

70.5

00.3

70.2

90.3

20.0

7�

0.0

80.0

8�

0.0

59

40.3

00.2

80.0

0�

0.5

3�

0.4

60.2

20.0

30.3

0�

0.0

60.1

5�

0.1

30.2

80.1

10.0

0�

0.0

10.0

40.0

60.1

50.1

00.1

10.1

18

50.5

50.0

60.0

00.0

80.3

40.1

60.0

10.0

20.0

90.5

50.2

2�

0.0

8�

0.0

8�

0.1

3�

0.3

50.0

4�

0.1

2�

0.0

9�

0.0

6�

0.0

9�

0.0

76

6�

0.0

90.4

00.0

0�

0.0

50.0

50.1

50.0

8�

0.2

60.6

0�

0.2

90.1

10.1

10.1

90.1

6�

0.3

8�

0.1

20.1

9�

0.0

2�

0.0

10.0

2�

0.1

05

70.0

00.0

01.0

00.0

00.0

00.0

00.0

00.0

00.0

00.0

00.0

00.0

00.0

00.0

00.0

00.0

00.0

00.0

00.0

00.0

00.0

05

80.4

8�

0.2

80.0

00.1

30.2

60.1

3�

0.1

20.0

80.2

0�

0.3

7�

0.5

60.0

40.1

4�

0.0

50.0

50.1

50.1

00.0

40.1

10.0

30.0

84

90.3

4�

0.1

80.0

0�

0.1

5�

0.2

20.0

4�

0.2

8�

0.3

40.2

9�

0.0

70.3

5�

0.2

3�

0.1

20.1

60.4

3�

0.0

2�

0.1

50.1

6�

0.0

30.2

1�

0.0

84

10

0.2

5�

0.0

70.0

00.0

6�

0.0

3�

0.0

10.2

9�

0.4

9�

0.4

5�

0.2

10.2

50.2

90.3

8�

0.2

3�

0.0

40.0

30.0

40.0

1�

0.0

20.0

80.0

53

11

�0.1

60.0

40.0

0�

0.0

9�

0.0

5�

0.0

70.3

3�

0.3

60.3

50.4

1�

0.3

7�

0.0

30.1

4�

0.3

50.3

30.1

3�

0.1

2�

0.0

90.0

6�

0.0

40.0

13

12

0.1

10.3

00.0

00.1

9�

0.0

50.1

00.2

10.3

90.0

9�

0.3

20.2

0�

0.1

3�

0.0

2�

0.5

00.3

10.0

3�

0.0

5�

0.0

1�

0.2

5�

0.0

6�

0.2

63

13

0.1

50.3

60.0

0�

0.2

80.4

7�

0.3

60.0

10.0

5�

0.0

7�

0.1

0�

0.0

10.1

10.1

40.2

40.3

2�

0.2

7�

0.3

2�

0.1

30.0

7�

0.1

40.0

82

14

0.0

7�

0.3

40.0

0�

0.3

50.0

2�

0.0

30.5

30.1

20.1

3�

0.1

80.2

5�

0.2

0�

0.2

00.0

9�

0.0

30.1

40.1

5�

0.2

50.2

2�

0.2

70.1

92

15

0.2

10.0

30.0

00.4

9�

0.4

8�

0.1

10.2

60.0

40.0

9�

0.0

1�

0.1

10.0

80.0

20.3

7�

0.0

6�

0.0

7�

0.4

1�

0.1

70.0

4�

0.1

80.0

32

16

0.2

40.1

40.0

00.2

2�

0.1

1�

0.5

10.0

5�

0.0

10.0

60.1

30.0

00.0

2�

0.2

2�

0.0

90.0

9�

0.2

80.6

20.0

70.1

30.0

80.1

22

17

�0.0

50.2

00.0

00.1

8�

0.0

60.3

1�

0.4

0�

0.1

1�

0.0

1�

0.0

30.1

70.1

2�

0.0

7�

0.1

80.1

90.0

60.0

7�

0.3

50.3

8�

0.3

40.3

82

18

0.0

20.3

00.0

00.0

3�

0.0

3�

0.3

9�

0.1

00.0

0�

0.0

1�

0.0

80.0

8�

0.2

80.1

0�

0.0

1�

0.1

70.7

2�

0.1

00.1

20.1

60.1

20.1

61

19

�0.0

20.0

00.0

0�

0.0

10.1

1�

0.0

90.0

5�

0.0

90.1

1�

0.1

00.0

20.6

4�

0.6

4�

0.0

7�

0.0

10.2

6�

0.1

70.0

8�

0.0

40.0

9�

0.0

61

20

0.0

00.0

00.0

00.0

00.0

00.0

00.0

00.0

00.1

10.0

00.0

00.0

00.0

00.0

00.0

00.0

00.0

00.0

0�

0.7

00.0

00.7

10

21

0.0

00.0

00.0

00.0

00.0

00.0

00.0

00.0

90.0

00.0

00.0

00.0

00.0

00.0

00.0

00.0

00.0

0�

0.7

10.0

00.7

00.0

00

Appendix B. Final coefficients of the mixed model

explaining fruit production

Variable Base

coefficient

Additional coefficient for

Year Severity

2002 2003 2004 2005 2006 1 2 3 4

Intercept �198.15

Year 20.09 17.55 27.44 17.49 0.00

Dbh 3.0061

Dbh2 �0.0111

Crown form 52.62 31.24 17.98 0.00

Vine load 4.28 5.35 �26.67 0.00

Vine cutting �16.59 23.42 25.97 12.23 36.16 0.00

No vine cutting 0.00 0.00 0.00 0.00 0.00 0.00

P �6.478

CEC 7.617

K.A. Kainer et al. / Forest Ecology and Management 250 (2007) 244–255 253

Page 12

Author's personal copy

Gama, J.R.N.F., Kushaba, T., Ota, T., Amano, Y., 1992. Influencia de material

vulcanico em alguns solos do estado do Acre. Revista Brasileira de Ciencia

do Solo 16, 103–106.

Grauel, W.T., Putz, F.E., 2004. Effects of lianas on growth and regeneration of

Prioria copaifera in Darien, Panama. For. Ecol. Manage. 190, 99–108.

Harper, J.L., White, J., 1974. The demography of plants. Ann. Rev. Ecol. Syst.

5, 419–463.

Hedley, M.J., Mortvedt, J.J., Bolan, N.S., Syers, J.K., 1995. Phosphorus fertility

management in agroecosystems. In: Tiessen, H. (Ed.), SCOPE, 54, Phos-

phorus in the Global Environment: Transfers, Cycles and Management.

John Wiley & Sons, Ltd., Chichester, West Sussex, UK, pp. 59–92.

Herrera, C.M., Jordano, P., Guitian, J., Traveset, A., 1998. Annual variability in

seed production by woody plants and the masting concept: reassessment of

principles and relationship to pollination and seed dispersal. Am. Nat. 152,

576–594.

Holdridge, L.R., 1978. Ecologıa Basada en Zonas de Vida. Instituto Interamer-

icano de Ciencias Agrıcolas, San Jose, Costa Rica.

IMAC (Instituto de Meio Ambiente do Acre), 1991. Atlas geografico ambiental

do Acre. IMAC, Rio Branco, Acre, Brazil.

Jansen, P.A., 2003. Scatterhoarding and tree regeneration: ecology of nut

dispersal in a neotropical rainforest. Thesis. Wageningen University,

Wageningen, the Netherlands.

Janzen, D.H., 1976. Why bamboos wait so long to flower. Ann. Rev. Ecol. Syst.

7, 347–391.

Johnson, D.H., 1999. The insignificance of statistical significance testing. J.

Wildl. Manage. 63, 763–772.

Jones, J.E., Garrett, H.E., Haines, J., 1995. Genetic selection and fertilization

provide increased nut production under walnut-agroforestry management.

Agroforestry Syst. 29, 265–273.

Jordan, C.F., 1985. Nutrient Cycling in Tropical Forest Ecosystems: Principles

and Their Application in Management and Conservation. John Wiley &

Sons, Chichester, West Sussex, UK.

Kainer, K.A., Wadt, L.H.O., Gomes-Silva, D.A.P., Capanu, M., 2006. Liana

loads and their association with Bertholletia excelsa fruit and nut produc-

tion, diameter growth and crown attributes. J. Trop. Ecol. 22, 147–154.

Kelly, D., 1994. The evolutionary ecology of mast seeding. Trends Ecol. Evol.

9, 465–470.

Kelly, D., Sork, V.L., 2002. Mast seeding in perennial plants: why, how, where?

Ann. Rev. Ecol. Syst. 33, 427–447.

Koenig, W.D., Knops, J.M.H., 2000. Patterns of annual seed production by

Northern Hemisphere trees: a global perspective. Am. Nat. 155, 59–69.

Lathwell, D.J., Grove, T.L., 1986. Soil–plant relationships in the tropics. Ann.

Rev. Ecol. Syst. 17, 1–16.

Layne, J.N., Abrahamson, W.G., 2004. Long-term trends in annual reproductive

output of the scrub hickory: factors influencing variation in size of nut crop.

Am. J. Bot. 91, 1378–1386.

Marschner, H., 1995. Mineral Nutrition of Higher Plants, second ed. Academic

Press, San Diego, CA, USA.

Maues, M.M., 2002. Reproductive phenology and pollination of the Brazil nut

tree (Bertholletia excelsa Humb. & Bonpl. Lecythidaceae) in Eastern

Amazonia. In: Kevan, P., Imperatriz Fonseca, V.L. (Eds.), Pollinating Bees:

The Conservation Link Between Agriculture and Nature. Ministry of the

Environment, Brasilia, Brazil, pp. 245–254.

McGrath, D.A., Duryea, M.L., Cropper, W.P., 2001. Soil phosphorus avail-

ability and fine root proliferation in Amazonian agroforests 6 years follow-

ing forest conversion. Agric. Ecosyst. Environ. 83, 271–284.

Mills, H.A., Jones Jr., J.B., 1996. Plant Analysis Handbook II: A Practical

Sampling, Preparation, Analysis, and Interpretation Guide. Micro-Macro

Publishing, Athens, GA, USA.

Mori, S.A., Prance, G.T., 1990. Taxonomy, ecology, and economic botany of the

Brazil nut (Bertholletia excelsa Humb. and Bonpl: Lecythidaceae). Adv.

Econ. Bot. 8, 130–150.

Nelson, B.W., Absy, M.L., Barbosa, E.M., Prance, G.T., 1985. Observations on

flower visitors to Bertholletia excelsa H.B.K. and Couratari tenuicarpa

A.C. Sm. (Lecythidaceae). Acta Amazonica 15, 225–234.

Neter, J., Kutner, M.H., Wasserman, W., Nachtsheim, C.J., 1996. Applied

Linear Statistical Models, fourth ed. WCB/McGraw-Hill, Boston, MA,

USA.

Nielsen, G.H., Hogue, E.J., Parchomchuk, P., 1990. Flowering of apple trees in

the second year is increased by first-year P fertilization. Hortscience 25,

1247–1250.

Norton, D.A., Kelly, D., 1988. Mast seeding over 33 years by Dacrydium

cupressinum Lamb. (rimu) (Podocarpaceae) in New Zealand: the impor-

tance of economies of scale. Funct. Ecol. 2, 399–408.

O’Malley, D.M., Buckley, D.P., Prance, G.T., Bawa, K.S., 1988. Genetics of

Brazil nut (Bertholletia excelsa Humb. & Bonpl.: Lecythidaceae). Theor.

Appl. Genet. 76, 929–932.

Ortiz, E.G., 2002. Brazil nut (Bertholletia excelsa). In: Shanley, P., Laird,

S.A., Guillen, A. (Eds.), Tapping the Green Market: Certification and

Management of Non-Timber Forest Products. Earthscan Publications,

Ltd., London, UK, pp. 61–74.

Peres, C.A., Baider, C., Zuidema, P.A., Wadt, L.H.O., Kainer, K.A., Gomes-

Silva, D.A.P., Salomao, R.P., Simoes, L.L., Francisiosi, E.R.N., Valverde,

F.C., Gribel, R., Shepard Jr., G.H., Kanashiro, M., Coventry, P., Yu, D.W.,

Watkinson, A.R., Freckleton, R.P., 2003. Demographic threats to the

sustainability of Brazil nut exploitation. Science 302, 2112–2114.

Peters, C.M., 1996. The ecology and management of non-timber forest resources.

World Bank Technical Paper 322. The World Bank, Washington, DC, USA.

Ponder, F., Jones, J.E., 2001. Annual applications of N, P, and K interrupt

alternate-year nut crops in black walnut. J. Plant Nutr. 24, 661–670.

Prance, G.T., Mori, S.A., 1979. Lecythidaceae—Part I. The actinomorphic-

flowered New World Lecythidaceae (Asteranthos, Gustavia, Grias, Allan-

toma, & Cariniana). Flora Neotropica Monograph 21. The New York

Botanical Garden, Bronx, NY, USA.

Prance, G.T., 1976. The pollination and androphore structure of some Ama-

zonian Lecythidaceae. Biotropica 8, 235–241.

Schabenberger, O., Pierce, F.J., 2001. Contemporary Statistical Models for the

Plant and Soil Sciences. CRC Press, Boca Raton, FL, USA.

Snook, L.K., Camera-Cabrales, L., Kelty, M.J., 2005. Six years of fruit produc-

tion by mahogany trees (Swietenia macrophylla King): patterns of variation

and implications for sustainability. For. Ecol. Manage. 206, 221–235.

Stephenson, A.G., 1981. Flower and fruit abortion: proximate causes and

ultimate functions. Ann. Rev. Ecol. Syst. 12, 253–279.

Stevens, G.C., 1987. Lianas as structural parasites: the Bursera simaruba

example. Ecology 82, 913–919.

Stoian, D., 2005. Making the best of two worlds: rural and peri-urban livelihood

options sustained by nontimber forest products from the Bolivian Amazon.

World Dev. 33, 1473–1490.

Sutherland, S., 1986. Patterns of fruit-set: what controls fruit-flower ratios in

plants? Evolution 40, 117–128.

Synnott, T.J., 1979. A manual of permanent sample plot procedures for tropical

rainforests. Tropical Forestry Papers No. 14. Commonwealth Forestry

Institute, University of Oxford, Oxford, UK.

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

products: a review. J. Appl. Ecol. 41, 11–21.

Trevedi, M.R., Cornejo, F.H., Watkinson, A.R., 2004. Seed predation on Brazil

nuts (Bertholletia excelsa) by macaws (Psittacidae) in Madre de Dios, Peru.

Biotropica 36, 118–122.

UFAC (Universidade Federal do Acre), 2006. Group of Environmental Services

and Studies, Natural Sciences Department, Federal University of Acre, Rio

Branco, Acre, Brazil.

Vander Wall, S.B., 2001. The evolutionary ecology of nut dispersal. Bot. Rev.

67, 74–117.

Vaughton, G., 1991. Variation between years in pollen and nutrient limitation of

fruit-set in Banksia spinulosa. J. Ecol. 78, 389–400.

Viana, V.M., Mello, R.A., Moraes, L.M., de Mendes, N.T., 1998. Ecologia e

manejo de populacoes de castanha-do-Para em reservas extrativistas Xapurı,

Estado do Acre. In: Gascon, C., Montinho, P. (Eds.), Floresta Amazonica:

Dinamica, Regeneracao e Manejo. Instituto Nacional de Pesquisa da

Amazonia, Manaus, AM, Brazil, pp. 277–292.

Vieira, S., Trumbore, S., Camargo, P.B., Selhorst, D., Chambers, J.Q., Higuchi,

N., 2005. Slow growth rates of Amazonian trees: consequences for carbon

cycling. Proc. Natl. Acad. Sci. USA 102, 18502–18507.

Wadt, L.H.O., Kainer, K.A., Gomes-Silva, D.A.P., 2005. Population structure

and nut yield of a Bertholletia excelsa stand in Southwestern Amazonia.

For. Ecol. Manage. 211, 371–384.

K.A. Kainer et al. / Forest Ecology and Management 250 (2007) 244–255254

Page 13

Author's personal copy

Wright, S.J., Muller-Landau, H.C., Calderon, O., Hernandez, A., 2005. Annual

and spatial variation in seedfall and seedling recruitment in a neotropical

forest. Ecology 86, 848–860.

ZEE (Zoneamento Ecologico-Economico do Acre), 2000. Recursos naturais e

meio ambiente, vol. I. Secretaria de Estado de Ciencia, Tecnologia e Meio

Ambiente, Rio Branco, Acre, Brazil.

Zuidema, P.A., Boot, R.G.A., 2002. Demography of the Brazil nut tree

(Bertholletia excelsa) in the Bolivian Amazon: impact of seed extraction

on recruitment and population dynamics. J. Trop. Ecol. 18, 1–31.

Zuidema, P.A., 2003. Ecology and management of the Brazil nut tree (Berthol-

letia excelsa), PROMAB (Programa Manejo de Bosques de la Amazonia

Boliviana) Scientific Series 6. PROMAB, Riberalta, Bolivia.

K.A. Kainer et al. / Forest Ecology and Management 250 (2007) 244–255 255