|UFDC Home||myUFDC Home | Help|
This item has the following downloads:
NATURAL REGENERATION OF CANOPY TREES IN A TROPICAL DRY FOREST IN
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
O 2007 Bonifacio Mostacedo
To my parents for raising me with love and making me what I am today. To my wife, who brings
great joy into my life. To my children, for the inspiration and strength they gave me throughout
the long process of completing this dissertation.
So many people contributed to this work that I cannot acknowledge them all by name but I
do want to single out a few. First of all, I would like to thank my mentor and principal supervisor
Francis E. "Jack" Putz for asking me difficult questions and pushing me to think and write
logically. He helped me adjust to life in Gainesville and was a good friend throughout my career
as a graduate student. Particularly while I was trying to finish my dissertation, he provided a
great deal of support and encouragement. I hope that we can continue to work together in the
I also want to thank the members of my advisory committee, Kaoru Kitajima, Karen
Kainer, Emilio Bruna, and Colin Chapman, for always being available to provide me guidance.
For much of the past decade, Todd Fredericksen has been a key person in my life. We
spent a great deal of time together in various forests in Bolivia conducting what I believe to be
exciting research and teaching what I hope were useful field courses. No matter how bad the
conditions, he always had a smile on his face and an amusing comment on his lips. He
encouraged me to start my Ph.D., wrote letters of recommendation on my behalf, and helped me
secure financial support for my studies. He continues to provide critical feedback on my work
but, overall, I am most grateful to him for being a good friend.
Numerous University of Florida faculty and staff contributed substantially to my
dissertation proj ect. In particular, Doug Levey helped me design the predator exclosures and
Stephen Mulkey helped me with the irrigation experiment. Among the statisticians who guided
me through the design and analysis phases of this research, I would like to thank Larry Winner,
Andrew Khuri, and Jorge Cassela. I am also very grateful to Ann Wagner, Pamela Williams, and
Paula Maurer from the Botany office for their help.
My Ph.D. studies were supported to a large extent by a fellowship from USAID through
the BOLFOR Proj ect run by Chemonics Intemnational. Thanks go to John Nitler and Ivo
Kralj evic for helping me to secure this funding. They, and several other people at Chemonics,
helped me to deal with culture shock and to otherwise adjust to life in the USA. Part of the
research was funded by Intemnational Foundation for Science (IFS) who gave financial support to
do experiments about seedling dynamic. The final phase of my dissertation research was
supported by a fellowship from the Gordon and Betty Moore Foundation through the Tropical
Conservation and Development Program at the University of Florida.
I want to thank my officemates and friends in Gainesville including Clea Paz, Geoffrey
Blate, Joseph Veldman, Morgan Vamner, Bil Grauel, Skya Murphy, Ana Eleuterio, Paulo Brando,
Camila Pisano, and Christine Lucas. Skya, Ana, Paulo, and Christine provided me helpful
comments on my dissertation. I also want to recognize Claudia Romero for teaching me various
things that made my life easier, helping me with the dreaded dissertation templates, reviewing
the sprouting chapter, and convincing me that I could finish this dissertation on a time schedule
that seemed unattainable.
The Instituto Boliviano de Investigacion Forestal (IBIF), through its director Marielos
Pefia-Claros, provided me with all the necessary logistical support needed to conduct my
fieldwork. Marielos also provided many helpful comments on my dissertation proposal. Many
other people at IBIF were strong supporters of my work, including members of the
administrative (Emma Nufiez, Laly Dominguez, and Karina Munoz) and technical staffs
(Marisol Toledo, Zulma Villegas, Juan Carlos Licona, Alfredo Alarcon, Carlos Pinto, Claudio
Leatio, Vincent Vroomans, Betty Flores and Mayra Maldonado) as well as Lourens Poorter, a
research associate. From Lourens I received many helpful suggestions about my research.
My Hieldwork would not have been possible without help of my student assistants and
"materos." In particular, Marlene Soriano helped me collect a portion of the Hield data for the
first and second chapters of this dissertation. Armando Villca and Turian Palacios helped me
with the Hield research for the third chapter. Alej andra Calderon, Vanessa Sandoval, Monique
Grol, Janeth Mendieta, Carla Gonzalez, and Joaquin Cordero helped in various ways in and on
the way to the field. I want also to thank many other people for Hieldwork assistance including
Juan Carlos Alvarez, Israel Melgar, Daniel Flores, Hugo Justiniano, Rodolfo and Rafael Rivero,
Daniel Alvarez, Alberto Chacon, Antonio Jimenez, Juan Pessoa, Dona "Negrita" Pessoa, and
Juan Alvarez. I also want to thank Miguel Angel Chavez, the driver from IBIF, who was always
willing to volunteer to travel the long bumpy road to INPA.
The INPA PARKET Company was the gracious host of my Hield research and helped in
many different ways in its execution. The company's owner, Paul Roosenboom, company
manager William Pariona, forester Urbano Choque, and many other people working for this
prestigious and FSC-certified forest-products firm contributed to the success of this proj ect.
Last but certainly not least, I want to thank my wife, Ynes Uslar, and children, Gabriela
Ines and Jose Daniel, for the many sacrifices a family makes when a parent takes on a Ph.D.
proj ect. The many nights I spent writing instead of being with them, the long stints of field work,
and the even longer periods when I was in Gainesville and they were back in Bolivia were hard
on all of us but they bore the strain without complaint. I should also thank Ynes for counting
thousands of seeds and typing in even more data, but these contributions pale in comparison with
the rest of what she has done for me and the rest of our family.
TABLE OF CONTENTS
ACKNOWLEDGMENTS .............. ...............4.....
LIST OF TABLES ................ ...............9............ ....
LIST OF FIGURES ................. ...............11................
AB S TRAC T ............._. .......... ..............._ 13...
1 FRUIT PRODUCTION OF TROPICAL DRY FOREST TREES IN BOLIVIA ..................1 5
Introducti on ................. ...............15.................
Methods .............. .. ...... .... .............1
Study Area and Climate............... ...............16
Species Studied...................... .....................1
Experimental Design and Data Collection ................. ...............18........... ...
Data Analysis............... ...............20
Re sults ................ ...............20.................
D discussion ............... ... .......... ... .. ...............22..
Fruiting of Trees in a Tropical Dry Forest .............. ...............22....
Factors Affecting Fruit Production............... ...............2
2 BIOTIC AND ABIOTIC FACTORS AFFECTING TREE SEEDLING DYNAMICS IN
A DRY TROPICAL FOREST............... ...............39.
Introducti on ................. ...............39.................
Methods .............. .. ...... .... .............4
Study Area and Climate............... .. ...............4
Experimental Design and Data Analysis ................. ........... .... .. ........ .......4
Response of regeneration to silvicultural treatment intensity ................. ...............44
Factors affecting seedling establishment and growth............... ...............45.
Re sults................... ............ .............. ........ .......4
Response of Regeneration to Management Intensities ................. ................. ....._48
Factors Affecting Seedling Growth and Establishment .............. .....................4
Seedling growth............... ...............49.
Seedling establishment ................. ...............50.................
D discussion ................ ........... .... ....... .............5
Logging Effects on Seedlings Dynamics .............. ...... ...............51.
Factors Affecting Seedling Establishment and Growth .............. .....................5
3 CONTRIBUTION OF ROOT AND STUMP SPROUTS TO NATURAL
REGENERATION INT A LOGGED TROPICAL DRY FOREST INT BOLIVIA ..................79
Introducti on ................. ...............79.................
M ethods .............. ...............80....
Study A rea ................ .. ........ ....... ...... ...... .............8
Experimental Design and Data Collection .............. ...............81....
Stump sprouts ................. ........... ... .......... .. ... ............8
Comparison of different juvenile types in relation to microsites created by
logging ................. ...............82.......... .....
Data Analysis............... ...............83
R e sults................ .. ......... ...............83.......
Sprout Characterization. .............. ... ...............83.
Juvenile Types and the Effects of Logging ................. ...............86........... ..
Growth of Stump Sprouts ................. ...............86................
Discussion ............... .... .. ........... .. ........... .... .................8
Natural Regeneration and Shade Tolerance: True Seedlings vs. Sprouts .......................88
Allometric Relationships with Stump Sprouting ................. ............ .........._._. ...89
Growth of Stem and Root Sprouts Compared with True Seedlings............... ...............9
LIST OF REFERENCES ................. ...............107...___ ......
BIOGRAPHICAL SKETCH ............ ............ ...............117...
LIST OF TABLES
1-1. Overview of reproductive characteristics of the tree species studied in a tropical dry
forest in Bolivia............... ...............28
1-2. Results of the best model after backward regression steps to remove non-significant
variables (P > 0.1) sequentially from the full multiple regression model including
tree size (DBH), percent liana infestation, crown area, and crown position as
independent variables to explain fruit production. ............. ...............29.....
1-3. Analysis of covariance to determine the liana cutting effect in fruit production of
Caesalpinia pluviosa.................. ...............30.................
2-1. Spatial distributions, crown position, ecological group, geographical range, tree
densities and basal area of timber species in a Bolivian tropical dry forest. ................... ..59
2-2. Means of seedling densities in an unharvested control plot, a plot subj ected to normal
timber harvesting, and a plot subj ected to more intensive harvesting followed by
silvicultural treatments. ........... ........... ...............60.....
2-3. Mean seedling densities (#/m2) for 3 years (2003-2005) of commercial tree species in
microsites created by logging in a Chiquitano dry forest. .......____ ......_ ..............61
2-4. Establishment and mortality rates of seedlings of commercial tree species monitored
over a 3 y period in a control plot and plots subj ected to two harvesting intensities
(N=144 subplots/treatment plot). .............. ...............62....
2-5. Results of repeated measure analysis of variance for a split plot design run for
seedling density for all species combined or six timber tree species analyzed
separately .. ............... ...............63._____......
2-6. Mean relative height growth rates (A 1SE) of seedlings of commercial tree species in
response to bromeliad cover removal evaluated in 4 times.. ............ ......................6
2-7. Mean relative height growth rates (A 1SE) of seedlings of commercial tree species in
irrigated and droughted plots evaluated 4 times. ............. ...............67.....
2-8. Mean relative height growth rates (A 1SE) of seedlings of commercial tree species in
response to mammal exclosure evaluated 4 times during 2006............... ..................6
3-1. Frequency of root and stem sprouting and shade tolerance of commercial and non-
commercial canopy tree species in a tropical dry forest in Bolivia.................. ...............92
Table 3-2. Summary of the best models and their significance from the regression analyses
between stump diameter or stump volume (independent variables) and the heights of
stump and numbers of sprouts per stump (dependent variables). ........._..__.........._.......93
3-3. Mean (a 1SE) densities of true seedlings, stem sprouts, and root sprouts in 10 x 4 m
plots in microsites created during selective logging. ................... ..............9
3-4. Mean (a 1SE) of stump sprout height growth rates (cm/year) by ecological groups........95
3-5. Means (a 1SE) of stem heights by species that sprouted from stems or roots
compared to the heights of seedlings. .............. ...............96....
3-6. Mean (a 1SE) heights of trees I 2 m tall that were root sprouts, stem sprouts, or
seedlings grouped by ecological guild. .............. ...............97....
LIST OF FIGURES
1-1. Locations of the study areas in Bolivia ................. ...............31..............
1-2. Percentage of trees fruiting in 2003 and 2004 (note that none of these species are
dioecious). ............ ...............32.....
1-3. Annual variation in fruit production in three timber species. ........... ......................3
1-4. Comparison of the sizes (DBH) of fruiting and non-fruiting trees in 2003 and 2004
(note that none of these species are dioecious and all the trees were reproductively
m ature). ............. ...............3 4....
1-5. Simple linear regressions between crown area, tree size (DBH) in relation to log
transformed fruit production data for three timber species............... ...............35
1-6. Fruit production of Caesalpinia pluviosa in relation to DBH and crown area for trees
with cut or uncut lianas. .............. ...............36....
1-7. Percentages of trees fruiting in a logged and an unlogged plot. ................ ........._.......37
1-8. Average fruiting intensities (% of crown cover) of timber tree species evaluated in
areas subj ected to normal selective logging and nearby unlogged control areas in a
tropical dry forest. .............. ...............38....
2-1. Monthly rainfall (A) and soil moisture tension measured by Watermark@ soil
sensors (B).. ............ ...............69.....
2-2. Canopy openness in an unlogged control plot, an area subj ected to normal timber
harvesting (4.7 m3/ha harvested), and an area subj ected to intensive harvesting (8.2
m3/ha) 8 months (left-hand bar) and 42 months (right hand bars) after logging............_...70
2-3. Design of the experiment on the effects of ground bromeliads, irrigation, extended
drought, and seed and seedling predator exclosures on seedling establishment. ...............71
2-4. Seedling densities of 11 timber species (A) and 10 species without Acosmium
cardena~sii (B), the most dominant species, in a control plot, a plot subj ected to
normal timber harvesting, and a plot subjected to intensive timber stand
management. ............. ...............72.....
2-5. Temporal changes in seedling recruitment (#/m2/y) and mortality for commercial tree
species in a Chiquitano dry forest in Bolivia. .............. ...............73....
2-6. Seedling density over time in response to (A) bromeliad cover, (B) irrigation or
drought, and (C) mammalian seed predators. ............. ...............74.....
2-7. Mean densities (A 1SE) of seedlings of commercial timber tree species in 2 m2 plOts
(N=40) with bromeliads (filled dots) and without bromeliads (open dots). ....................75
2-8. Mean of seedling densities (A 1SE) in irrigated and droughted experimental plots..........76
2-9. Mean seedling densities (11 SE) in control plots (closed dots) and in plots from
which mammals were excluded (open dots). .............. ...............77....
2-10. Mean relative growth height growth rates (A 1SE) of seedlings of commercial tree
species in response to three experimental treatments: (A) irrigation or drought; (B)
mammalian seed predator exclusion; and, (C) bromeliad cover ................. ................. .78
3-1. Proportions of stumps of commercial timber species that resprouted (number of
stumps noted in parenthesis). ..........__.......__ ...............98..
3-2. The proportions of stumps with live sprouts over time since logging ............... ...............99
3-3. Mean (a 1SE) numbers of sprouts/stump for the most frequent sprouting species.........100
3-4. Probabilities of stump sprouting as a function of stump diameter for the most
frequently sprouting commercial tree species (curves fit by logistic regression). ...........101
3-5. Probabilities of stump sprouting as a function of stump height for the commercial
tree species that most frequently sprouted (curves fit by logistic regression). ................102
3-6. Mean (a 1SE) densities of juveniles < 2 m tall of commercial tree species that were
true seedlings, stem sprouts, and root sprouts............... ...............103
3-7. Percentage of juveniles < 2 m of different origins after logging (for each of 15 dry
forest tree species) ordered by their light requirements ................. ................. ...._104
3-8. Relative growth rates (mean (A 1SE) of stump sprouts measured over the first two
years after logging for commercial tree species in a tropical dry forest in Bolivia
arranged by light requirements. ............. ...............105....
3-9. Mean growth rates of stump sprouts through time for the main commercial tree
species as based on measurements of the tallest sprouts on different stumps 1, 2, and
5 y after creation. ............. ...............106....
Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
NATURAL REGENERATION OF CANOPY TREES IN A TROPICAL DRY FOREST IN
Chair: Francis E. Putz
Major Department: Botany
Fruit production, seedling establishment, and sprouting of canopy trees were studied in a
lowland tropical dry forest in the Department of Santa Cruz, Bolivia. Fruit production by
reproductively mature trees was monitored over a 5 y period to assess variability among species,
trees and years. The effects of tree size, crown area, crown position, and liana infestation on fruit
production were also assessed. In a companion study, I assessed the effects of lianas on fruit
production by Caesalpinia pluviosa with a liana cutting experiment. To determine how logging
disturbances affect seedling recruitment, I monitored seedlings for 3 y in different microsites in
permanent plots in two selectively logged plots and an unlogged control plot. I also
experimentally assessed the effects of bromeliad cover, drought stress, and seed/seedling
predators on seedling recruitment, survival, and growth. Finally, I monitored the emergence,
survival, and growth rates of stump and root sprouts over a range of microsites.
Percentages of trees fruiting and numbers of fruits produced varied among species and
years. In most species, trees that did and did not fruit did not differ in size or crown position, but
in a few cases, the likelihood of fruiting increased with crown area. Contrary to my expectation,
there no effect of liana cutting on Caesalpinia pluviosa fruit production was detected 3 yr after
cutting. Overall, the effect of logging on the proportion of trees fruiting and fruiting intensity
varied among species.
Seedling densities 5 y after selective logging were higher in control than logged plots but
this finding was greatly influenced by the most common species, Acosnzium cardena~sii (43% of
seedlings enumerated). At the microsite level, Acosnzium was found in highest densities in
undisturbed areas while Centrolobium naicrochaete, another common species, was more common
on log extraction paths. Seedling recruitment rates were higher in the unlogged plot and in the
undisturbed portions of the logged forest plots, but seedling mortality rates were also higher in
these areas. Mortality rates of naturally established seedlings varied greatly among species.
Seven of 22 species suffered no mortality during the 2-y monitoring period, whereas relatively
high mortality rates were observed for Caesalpinia (26%/y), Sweetia f~uiticosa (25%/y), and
Machaerium scleroxylon (22%/y).
Results of the experimental study on seedlings suggest that bromeliad competition and
seed/seedling predators greatly affected tree seedling establishment. Soil moisture availability
also affected seedling establishment, but only as an interaction with the bromeliad removal or
predator exclosure treatments. The primary effect of the drought treatment was delayed
germination. Despite these general trends, species varied substantially in their sensitivity to
bromeliads, drought stress, and predators.
Root and stump sprouts constituted about 50% of the individuals <2 m tall of the canopy
tree species studied, but the proportions of sprouts and true seedlings varied among species.
Stump sprouting was common, but the probability of sprouting was not consistently related to
stump diameter or height. Sprout growth rates were consistently high, at least initially, and
sprouting is obviously important to post-logging regeneration in this dry tropical forest.
FRUIT PRODUCTION OF TROPICAL DRY FOREST TREES INT BOLIVIA
Tropical dry forests, which until very recently covered about 30% of Bolivia, are among
the most threatened ecosystems in the world (Dinerstein et al. 1995). Bolivian dry forest are under
siege; 32% has already been cleared (Camacho et al. 2001, Rojas et al. 2003) and most of the
remainder is under intensive pressure for forest products, grazing, and further conversion. This
pressure may be somewhat mitigated in the large portion of the dry forest managed for timber under
the guidelines of Bolivia's 1996 Forestry Law (MDSP 1996, Nittler and Nash 1999). Even when
these guidelines are followed, logging disturbances are larger in area than those that occur naturally.
Given the documented effects of disturbance and habitat modification on reproductive success of
tropical dry forest trees (Fuchs et al. 2003), the sustainability of managed forests even when the
legally required "best management" practices are used, remains uncertain. To assess how forest
management influences the reproduction of commercially valuable tree species, I explore the inter-
tree and inter-annual variation in fruit production in a dry forest in lowland Bolivia.
A wide range of mechanistic explanations have been proposed for the marked variation in
fruit production among years and among individuals in a wide variety of forested ecosystems
(Abrahamson and Layne 2003, Snook et al. 2005). For seasonally dry tropical forests, inter-
annual variation in rainfall, particularly as influenced by El Nifio events, has often been invoked
as the underlying cause of inter-annual variation in tree phenology and fruit production (Bullock
1995, Wright and Calderon 2006). Regarding individual tree characteristics that might influence
reproductive output, several studies have reported a minimum size threshold for tree
reproduction (Afiez 2005, Wright et al. 2005). For trees that have attained reproductive maturity,
several studies have shown a relationship between fruit production and stem diameter (Zuidema
and Boot 2002, Wadt et al. 2005), crown area (Healy et al. 1999, Zuidema 2003), and crown
position (Healy et al. 1999). Cover by lianas has also been shown to reduce fruit production by
Brazil nuts "Bertholletia excels Bonpl." (Zuidema and Boot 2002, Wadt et al. 2005, Kainer et
al. 2006) and other species (Stevens 1987).
Anthropogenic and natural disturbances have been reported to influence canopy tree fruit
production in a number of forests around the tropics but the results have been inconsistent.
Logging increased fruit production by remnant trees in a subtropical humid forest in Meghalaya,
India (Barik et al. 1996), as well as in a tropical montane forest in Costa Rica (Guariguata and
Saenz 2002). In contrast, reproductive output of residual trees in a logged dipterocarp forest in
Indonesia was lower than in an unlogged control area (Curran et al. 1999). Factors responsible
for these contrasting results are not clear but, given the importance of natural regeneration to
sustainable forest management, the issue deserves further exploration.
In this study in a seasonal lowland tropical forest in Bolivia I report on canopy tree fruit
production over a 5-year period. I also examine the relationships between fruit production and
crown features, tree size, and liana infestation. Using one common tree species that is often
heavily liana-laden, I assess the effect of liana cutting on fruit production. Finally, I examine the
effect of selective logging on fruit production by several common canopy tree species.
Study Area and Climate
Research was conducted at INPA Parket, a 30,000-ha tract of privately owned seasonally
dry tropical forest located 30 km northeast of the town of Concepci6n (160 6' 45 S and 610 42'
47" W) and 250 km northeast of the city of Santa Cruz, Bolivia (Figure 1-1). The altitude is
approximately 380 m, mean annual temperature is 24.3 OC, mean annual precipitation is 1150
mm (range 798-1859 mm/y), and there is a dry season that lasts about five months (May-
October) during which most trees are leafless. Extreme annual variation in rainfall is in part
related to the occurrence of El Nifio events, which are typically wet in the study area. Many tree
species in this forest flower at the end of rainy season with another peak in flowering at the
beginning of dry season. Fruiting of the maj ority of species occurs in mid-dry season (see Table
1-1). The forest canopy is 20-25 m tall and dominated by Tabebuia impetiginosa (Mart. Ex DC.)
Standl., Anadenanthera macrocarpa (Benth.) Brenan, Astronium urundeuva (Allemio) Engl.,
and Centrolobium microchaete (Mart. ex Benth.) Lima ex G. P. Lewis (Pariona 2006). The most
abundant species are Acosmium cardena~sii H. S. Irwin & Arroyo (- 38 trees/ha > 20 cm
diameter at breast height, DBH) and Anadenanthera (8 trees/ha > 20 cm DBH). Currently, 21
tree species, including these dominants, are harvested for timber used mostly in the production of
My studies focus on commercially valuable timber tree species, most of which produce
seeds that are wind, gravity, or explosively dispersed (Table 1-1). Overall I consider 31 species
(Table 1-1), of which 15 were used to determine the percentage of fruiting trees, 12 to compare
the sizes of reproductively mature trees that did or did not fruit during a 2-y observation period, 6
to compare fruiting intensity in logged and unlogged areas, and 3 to determine the relationships
between fruit production and DBH, crown area, crown position, and liana infestation. In
addition, one species (Caesalpinia pluviosa DC.) was used to study the effect of liana cutting on
fruit production. Most of the selected species have large or medium-sized fruits with small- to
intermediate- sized seeds.
Of the 3 1 species considered in this study 25 have hermaphroditic flowers and 20 produce
wind-dispersed seeds. Ten species have seeds as their dispersal units (e.g., Amburana cearensis
(Allemao) A.C. Sm.) while in the other 21 species, entire fruits are the dispersal units (e.g.,
Machaerium scleroxylon Tul.). Twenty-four of the 3 1 species monitored for five years produced
at least some fruits annually, but the intensity of fruiting varied substantially. Cecropia concolor
Willd. produced fruits continuously whereas Schinopsis brasiliensis Engl. fruited supra-annually
(sensu Newstrom et al. 1994). Peak fruiting for 23 of 31 species coincided with the dry season
(May September) with most seeds dispersed in the late dry season. Spondia~s mombin L. the
only fleshy-fruited species in the study produced ripe fruit during the rainy season. Most of the
species have only 1 seed per fruit, but Ceiba samnauma (Mart.) K. Schum. and Tabebuia have
more than 90 seeds/fruit. Wind-dispersed seeds or fruits typically have one (Pterogyne nitens
Tul.) or both sides winged (e.g., Platimiscium ulei Harms), whereas Zeyheria tuberculosa (Vell.)
Bureau seeds are entirely surrounded by a wing. Cottony fibers on the seeds of the
Bombacaceae studied (Ceiba and Chorisia speciosa A. St.-Hil.) aid in their dispersal by wind.
Experimental Design and Data Collection
The study was conducted at two different sites at INPA Parket. The first site was within the
Long-Term Silvicultural Research Plots (LTSRP) in Block # 1, 5-6 km south of the IBIF field
station (160 18' 26.8"S, 610 41' 13.5"W). Four large-scales (20 ha) long-term research plots
(LTRSP) were established by the Instituto Boliviano de Investigacion Forestal (IBIF). One of the
plots I used is an unlogged control whereas the other was subj ected to normal logging during
which a mean of 4.3 trees/ha (4.7 m3/ha of commercial timber) were harvested using standard
reduced-impact-logging techniques that include road planning, directional felling, and retention
of 20% of the harvestable trees as seed trees (Mostacedo et al. 2006). In each plot, I censused all
individuals > 20 cm DBH of the 15 most abundant tree species in two 20 x 400 m strips. The
second site is a 1000 x 500 m plot that I established 2 km northeast of the IBIF field station (160
14' 58.5" S, 610 41' 47.4" W). In this plot, I censused individuals > 20 cm DBH of
Anadenanthera and Caesalpinia in twelve 20 x 500 m strips spaced at 100 m intervals. At both
study sites the censused trees were marked, tagged, and mapped. To secure sufficient trees of
Caesalpinia and Zeyheria, I marked additional trees around the LTRSP.
In total, I evaluated 440 trees of 15 species, 116 in the control plot and 147 in the logged
plot of the LTRSPs. An additional 144 trees were monitored at the second site that, a year after
plot establishment, was selectively logged. In 2003, I evaluated only six species from May-
November including Caesalpinia, Anadenanthera, Machaerium scleroxylon, Ceiba, and
Zeyheria. In 2004, every marked tree in the LTSRP plots was phenologically evaluated Hyve
times from May to November. I estimated the percentage of the crown bearing fruits, which I
refer to as a measure of fruiting intensity; 100% fruiting intensity indicates that every terminal
branch bears at least one fruit.
I estimated the number of fruits produced by each tree of three species, Caesalpinia,
Anadenanthera, and Zeyheria, for 5 years (2002-2006). All of these species produce fruits with
valves that are not removed by animals. By counting these undispersed fruit parts under fruiting
trees, I avoided many of the difficulties associated with estimating seed production by species in
which entire fruits are removed by animals. At the beginning of the study I installed Hyve 2x2 m
permanent plots on the ground below the crown of each tree in which I counted and removed all
fruit valves. Subsequent censuses of fallen fruit valves in these plots over the next 5 y were used
as quantitative estimates of fruit production. I also measured the DBH, crown area (based on two
cardinal diameters), crown position (using the 5 categories of Dawkins (1958)), and liana
Liana-infested Caesalpinia trees (N=32) were used in a manipulative experiment on the
effects of lianas on seed production. I measured the DBH and percentage of each tree' s crown
covered by lianas, paired the trees on the basis of liana infestation and DBH, and cut all the
lianas on one tree of each pair, selected at random. Fruit production was measured for 3 y after
treatment using the fruit-valve census method described above.
To determine whether trees of the 12 monitored species that fruited in 2003 and 2004
differed in DBH from those that did not, I conducted Student's-t tests. I tested for simple linear
relationships between the number of fruits (log-transformed) produced per tree and each tree' s
DBH, crown area, crown position, and liana cover separately and then ran multiple regressions
for three monitored species (Anadenanthera, Caesalpinia, and Zeyheria) using the backward
method to avoid co-linearity. Only independent variables that explained significant amounts of
variance (P=<0.05) were included in the models. I ran X2 tests to assess differences between
logged and control plots in the percentage of trees fruiting. To compare maximum fruiting
intensities in the logged and the control plot for the 6 monitored species, I ran Student' s t-tests
after first arcsine transforming the data to achieve normality (Zar 1981). Finally, I compared fruit
production on liana-laden and liana-free trees using analyses of covariance with crown area or
DBH as covariates. For all analyses, I used SPSS Version 12.0 (Field 2000).
The percentage of trees fruiting in any particular year varied among species (Figure 1-2).
Of 14 species monitored during 2004, 7 had >50% of trees in fruit and 4 had >80% of trees in
fruit. For a few species that I also monitored in 2003, the percentage of fruiting trees was lower
in 2003 than in 2004. For example, while only 25% of Caesalpinia trees fruited in 2003, 100%
fruited in 2004. The opposite pattern was observed in Zeyheria; 40% fruited in 2003 and none in
For the 3 species I monitored for 5 years, there was a great deal of inter-annual variation in
fruit production (Figure 1-3). For example, a high proportion of Caesalpinia trees fruited at 2-
year intervals whereas many Zeyheria trees fruited at 3-year intervals. In contrast,
Anadenanthera fruited during each of the first 3 years of the study and not at all in the last 2.
Whether or not a tree fruited was generally not related to its DBH (Figure 1-4). The
exceptions were Caesalpinia in 2004 when the trees that failed to fruit were larger than the trees
that fruited (t=2.18, P=0.03) and Zeyheria in which the fruiting trees were larger in 2004, the
only year in which it fruited (t=4.91, P<0.0001).
For the 3 species in which I monitored fruit production, the number of fruits produced did
not vary with DBH or crown position but increased linearly with crown area (Figure 1-5). The
backward multiple regression of fruit production on tree characteristics revealed that crown area
explained the most variation (Table 1-2). In Anadenanthera, crown area explained 32% of the
variance in fruit production while DBH, crown position, and liana cover together explained only
an additional 10%. In the case of Caesalpinia, crown area explained 23% of the variance in fruit
production and the other three variables only an additional 3%. In Zeyheria, crown area and
crown position together explained only 24% of the variance, to which DBH and liana cover
added an additional 1%.
There was no apparent effect of liana cutting on fruit production by Caesalpinia when the
effects of either DBH or crown area are removed by ANCOVA (Table 1-3, Figure 1-6).
The percentage of trees fruiting in the logged and control plot was similar for 6 of 7
species. The only exception was Centrolobium in which only 27% of trees fruiting in the logged
plot compared to 90% in the control plot (X2=12.9, P<0.0001; Figure 1-7).
The effect of logging on fruit production varied among the 6 tree species studied (Figure 1-
8). Logging apparently stimulated increased fruiting intensity ofAnadenanthera (t=3.40,
P=<0.0001). In contrast, fruiting intensity was higher in the unlogged plot for both Centrolobium
(t=4.75, P=0.0008) and Copaifera (t=3.11, P=0.007). There was no difference in fruiting
intensity between the logged and unlogged plot for Aspidospernza (t=0.68, P=0.5) and
Machaerium scleroxylon (t=0.05, P=0.95).
Fruiting of Trees in a Tropical Dry Forest
In any year, the proportion of reproductively mature trees that fruited generally varied a
great deal among the canopy tree species I studied in lowland Bolivia. Only Anadenanthera trees
produced fruit crops in both 2003 and 2004. In contrast, no trees of2achaerium scleroxylon and
M~ acutifolium fruited in 2003, but many did in 2004. Most of the trees of some species fruited in
at least some years (e.g., Caesalpinia) while in others the percentage of fruiting trees was always
<40%. Perhaps coincidentally, the five species with the greatest proportion of fruiting trees in
2004 were all legumes, which comprised 8 of the 14 species monitored. Few studies report
annual variation in the proportions of fruiting trees but in a similar forest in Lomerio, Bolivia,
only 29% and 36% of reproductively mature trees of 17 commercial tree species fruited during
two years of monitoring (Justiniano and Fredericksen 2000). At the same site, there was a great
deal of variation within species in the proportions of fruiting trees. For example, the proportion
of Copaifera trees fruiting was similar to what I observed in INPA whereas none of the M~
scleroxylon trees in Lomeria fruited during the two years of monitoring. In a tropical dry forest
in Mexico studied by Bullock (1995), only 8-30% of Jacaratia nzexicana A. DC. trees and 0-
50% of Cochlospernaun vitifolium (Willd.) Spreng. trees fruited in any one year. Similarly, in a
study of Swietenia nzacrophylla King. on the Yucatan Peninsula, the proportion of fruiting trees
varied a greatly among years (Snook et al. 2005). The proportion of Hynzenaea courbaril L. trees
fruiting in Costa Rica reportedly varied with water stress (Janzen 1978).
A larger proportion of trees of most species fruited in 2004 than in 2003 or 2005. Such
community-wide synchrony in fruiting has been observed in many forests over the world but has
been particularly well studied in the dipterocarp forests of Southeast Asia where "masting" has
long been known (Appanah 1993, Wich and Van Schaik 2000). Less pronounced is the inter-
annual variation in fruit production on Barro Colorado Island, where mast years are reportedly
related to El Nifio events (Wright and Calderon 2006). In the forest of this study, 2003 was at the
end of an El Nifio event, which was not followed by a strong La Nifia. Nevertheless,
synchronous fruiting in 2004 could have been due to the timing of water stress, as suggested by
The three species I followed for five years showed great inter-annual variation in the
numbers of fruits produced per tree. The two legumes, Anadenanthera and Caesalpinia,
followed the same interannual patterns; both had peak years in 2002 and 2004, but produced few
fruits in 2003 and 2005. In contrast, the 35 Zeyheria trees I monitored produced no fruits in 2004
and 2005, but fruited in 2002, 2003, and 2006. Such interannual variation in fruit production is
common in many species. For example, in a moist tropical forest in Panama, Quararibea
a~sterolepis Pittier, Tetraga;stris pa~namennsi (Engl.) Kuntze, and Trichilia tuberculata (Triana &
Planch.) C. DC. showed great inter-annual variation in fruit production (De Steven and Wright
2002, Snook et al. 2005).
While I expected that the proportion of fruiting trees would increase with tree size, 12 of
the 13 species I monitored in 2003 and 2004 did not display this pattern. In fact, in the case of
Caesalpinia, the fruiting trees were significantly smaller than those that failed to fruit in 2004.
Only Zeyheria displayed the expected pattern. The general failure to find a positive relationship
between tree size and whether or not a tree reproduced runs counter to the results of two other
studies on this topic, one conducted on Barro Colorado Island, Panama (Wright et al. 2005) and
the other at my study site (Afiez 2005). One explanation for the difference between my study and
others in the literature is that whereas they typically used either flowering or fruiting as an
indication of reproduction (Afiez 2005, Wright et al. 2005), I used only fruiting. Nevertheless,
while the probability of fruiting increased with tree size in Copaifera, Sweetia and Machaerium,
it decreased in Anadenanthera and Caesalpinia.
Factors Affecting Fruit Production
Crown area was the best predictor of fruit production in many of the species I studied
whereas DBH and crown position were not. Crown area was also a good predictor of fruit
production in Betholletia excelsa trees in Amazonian Bolivia (Zuidema and Boot 2002), while
DBH was more closely related to fruit production by Swietenia in Mexico (Snook et al. 2005)
and Pinus sylvestris L. in Sweden (Karlsson 2000). Although several studies have shown strong
positive correlations between tree size and fruit production (Karlsson 2000, Zuidema and Boot
2002, Wadt et al. 2005), the relationships revealed in my study were positive but weak. This
difference might be related to the difficulty of making accurate DBH estimates of the trees in my
study site, many of which have irregular-shaped trunks. It is also possible that crown position is
not good predictor of fruit production in forests that are already open-canopied; certainly in my
study site it was difficult to differentiate between dominant and co-dominant trees, and I doubt
that they differ much in terms of light capture.
I expected liana cover to substantially impede fruit production (Stevens 1987), but my
results did not support that expectation. For the 16 pairs of Caesalpinia trees from which I
monitored fruit production for 3 y after cutting their lianas, fruit production was no higher than
in the liana-laden control trees. I can offer a few explanations for this counter-intuitive result. For
one thing, 3 y was perhaps not enough time for the trees to respond to the liberation from their
liana loads. Then there is the problem that fruit production is extremely variable among
individuals and among years in this species and others in my study site. Finally, I wonder
whether the liana leaf phenology at my study site might have something to do with this result
insofar as lianas are typically deciduous at the time of fruiting, which might reduce any
deleterious effect they have on fruit production. Obviously, none of these explanations is very
compelling and further monitoring is warranted.
Although I anticipated that by increasing canopy openness and reducing resource
competition, logging would increase fruit production by remnant trees, my results were
inconsistent at best. Firstly, the percentage of fruiting trees in most of species were similar
between logged and control plots, except for Centrolobium. For some species (e.g., Copaifera,
Aspidosperma, and Sweetia) in which there seemed to be a trend towards increased reproduction
among remnant trees in selectively logged forest, statistical significance was not forthcoming
perhaps due to small sample sizes (N < 20).
Secondly, 3 of the 6 species I monitored starting 1.5 y after selective logging showed no
apparent effect of the treatment on fruit production, 2 of the species fruited less in the logged
than in the control plot, and 1 species showed the opposite pattern. One reason for the observed
decrease in fruit production might have been the effect of lowered tree density on pollinator
effectiveness or offspring quality (Ghazoul and Shaanker 2004, Knight et al. 2005). Other factors
that might have reduced fruit production include fruit abortion due to poor ovule fertilization in
the extreme temperatures and low humid provoked by forest openness (Stephenson 1981, Aizen
and Feinsinger 1994, Dafni and Firmage 2000) or due seed predation during the early stages of
fruit formation (Stephenson 1981). For example, 1 tree/ha > 40 cm DBH of Copaifera remained
after logging of the pre-logging density of 1.7 trees/ha whereas in the unlogged plot, the
population of adult trees was substantially higher (2.7 trees/ha). Similarly, in Southeast Asia,
fruit production was reduced in a logged dipterocarp forest (Curran et al. 1999). In contrast, the
most light-demanding tree species in my study forest, Anadenanthera, produced more fruit in
logged than in unlogged areas. Similarly, Ceiba aesculifolia (Kunth) Britten & Baker f. in a dry
forest in Mexico (Herrerias-Diego et al. 2006) and Quercus costaricensis in montane humid
forest in Costa Rica (Guariguata and Saenz 2002) both increased fruit production in response to
logging. Given the importance of this issue to sustainable timber stand management, more
research is needed on the reproductive responses of trees to logging disturbances. These studies
should integrate flower production, pollination, and fruit production and should be conducted on
species representing a variety of densities and breeding systems.
In the seasonally dry tropical forest I studied in Bolivia, there was a lot of interannual
variation in fruit production among species. Whether or not a reproductively mature tree fruits is
generally not related to it size. The number of fruits produced per tree also did not consistently
change with stem diameter or liana cover, but did increase with crown size. Base on these
findings, I recommend protection of tree crowns during logging to increase fruit production.
Even when I experimentally killed the lianas infesting the crowns of 16 Caesalpinia trees, I did
not observe the expected increase in fruit production. To look at the effect of lianas cutting I
recommend a long-term study to look at crown area recovery and fruit production. Similarly,
after logging, remnant trees of some species increased their fruit production and some species
decreased but most did not change. For most of the species, percentages of trees fruiting were
similar between the control and the logged plot; only in Centrolobium was there a difference,
with more trees fruiting in the control plot. The small and inconsistent effects of logging on fruit
production in my study forest may be explained in part by the relatively low intensity of
harvesting. Obviously, given the substantial inter-tree, inter-specific, and inter-annual variation
in the fruiting of dry forest tree species, studies of more than a few years are needed.
Table 1-1. Overview of reproductive characteristics of the tree species studied in a tropical dry forest in Bolivia.
00 Chonisia speciosa
# Seeds /
(*) Clasification made by Newstron et al. (1994). (Jf) Data extracted from Justiniano and Fredericksen (2000).
Table 1-2. Results of the best model after backward regression steps to remove non-significant
variables (P > 0.1) sequentially from the full multiple regression model including tree
size (DBH), percent liana infestation, crown area, and crown position as independent
variables to explain fruit production. Number of trees (N), standardized regression
coefficients (P), Student-t test values (t), significance levels (P), and coefficients of
multiple determination (R2) are noted.
30 0.560 3.59
50 0.480 3.81
* P 5 0.05; ** P 5 0.01; *** P I 0.001.
Table 1-3. Analysis of covariance to determine the liana cutting effect in fruit production of
Caesalpinia pluviosa. The co-variables considered were crown area and DBH.
Source Crown Area DBH
Mean F P Mean F P
Co-variable 42438.5 2.35 0.14 22090.0 1.18 0.29
Liana cutting 9135.0 0.57 0.48 9673.0 0.52 0.48
Error 18020.7 18747.0
jrIBIF: Field Station
I::::IStudy Area 1
~T1 Study Area 2
SStuldy Area 3
1 0.5 0 1 2 3
sersaa I e-- nm Kilorneters
-** : = -
Figure 1-1. Locations of the study areas in Bolivia. Study Area 1 was where I conducted fruit
production research (Chapter 1); Study Area 2 was used for seedling recruitment
study (Chapter 2); and, Study Area 3 was the site for the sprouting studies (Chapter
3). The map is georeferenced using a Universal Transverse Mecator (UTM) System,
.Percentage of trees fruiting in 2003 and 2004 (note that none of these species are
dioecious). Asterisks indicate years in which a species that was not evaluated.
Abbreviations of species are given in Table 1-1. Numbers in parenthesis indicate the
number of trees evaluated for each species and year.
CAPL ANMA ACCA COCH MASC CESA GAIN CEMI ASRI SWFR CAIA CAPR MAAC ZETU
10002002 2003 2004 2005 2006
e Caesalpinia pluviosa
1- 800 -
80 2002 2003 2004 2005 2006
Zeyheria tuberculosa .
2002 2003 2004 2005 2006
Figure 1-3. Annual variation in fruit production in three timber species. Vertical lines show one
standard error. Note the differences in the y-axis scales.
99 24 3 45
NO YES NO YES
144 36 28 19
NO YES NO YES
1 Year 2003
IIIIII Year 2003
o Year 2004
M 1Year 2004
114 258 7 60
NO YES NO YES
Figure 1-4. Comparison of the sizes (DBH) of fruiting and non-fruiting trees in 2003 and 2004
(note that none of these species are dioecious and all the trees were reproductively
mature). Numbers indicate the sample size for each year and each species. Asterisks
indicate significant size differences between fruiting and non-fruiting trees as
determined with Student-t test. P-value: I 0.05, ** < 0.001, *** I 0.001.
Log(Y) 1 47+0 007(X),
r =0 32, F=12 92,
) 50 100 150 200 250
Spearman's r=-0 25, P=0 18
Spearman's r=-0 135, P=0 35
0 50 100 150 200 250
Log (Y) = 0 91+0 004 (X),
r2=0 17, F=6 53,
Spearman's r=0 012, P=0 95
20 30 40 50 60 70 80
Log (Y) = 1 64+0 004 (X),
r2=0 23, F=14 55,
30 40 50 60 70 80
Dom Cod~om Interm
0 50 100 150 200 250
20 30 40 50 60 70 80
Dom Codom Interm
Crown Area DBH Crown Position
Figure 1-5. Simple linear regressions between crown area, tree size (DBH) in relation to log
transformed fruit production data for three timber species. Spearman correlations
were run to determine the relationships between crown position (ordinal variable) and
fruit production. Crown position numbers refer to dominant (Dom), co-dominant
(Codom), and intermediate exposure (Interm) trees.
Log (Y) =1 86+0 01(X),
12=0 05, F=1 49,
Log (Y)= 1 71+0 013 (X),
r2=0 06, F=3 19,
Log (Y) = 0 59+0 16 (X),
r2=0 08, F=3 07,
ge* o a
O g*O O *
Crown Area (m2)
Figure 1-6. Fruit production of Caesalpinia pluviosa in relation to DBH and crown area for trees
with cut or uncut lianas.
(34Y (139 1 (5) 1(1 21 I12)
Caesalpinia Anadenanthera M~achaerium Aspidosperma Sweetia
Percentages of trees fruiting in a logged and an unlogged plot. Number of trees
evaluated is noted in parenthesis. Differences between logging intensities were tested
using X2 test at 95% of confidence. *** <0.0001; ns non significant.
SAnadenanthera macrocarpa l uu
May04 Jul04 Sep04 Oct04 Nov04
-* Normal Logging
May04 Jul04 Sep04 Oct04 Nov04
May04 Jul04 Sep04 Oct04 Nov04
May04 Jul04 Sep04 Oct04 Nov04
May04 Jul04 Sep04 Oct04 Nov04
May04 Jul04 Sep04 Oct04 Nov04
Figure 1-8. Average fruiting intensities (% of crown cover) of timber tree species evaluated in
areas subj ected to normal selective logging and nearby unlogged control areas in a
tropical dry forest. Vertical lines indicate standard errors.
BIOTIC AND ABIOTIC FACTORS AFFECTING TREE SEEDLING DYNAMICS IN A DRY
Among the many factors that affect tree seedling establishment in tropical forests, seed
predation, pathogen effects on seedlings and light availability figure prominently (Janzen 1971,
Augspurger 1984, Hammond 1995, Huante et al. 1998, Kobe 1999). Given that this view is
mostly supported by studies conducted in moist and wet forests, the relative importance of these
factors might differ in seasonally dry forests (Gerhardt 1994, Holbrook et al. 1995). Given the
widespread mismanagement and destruction of these forests (Steininger et al. 2001, Pacheco
2006), it is increasingly important to increase our knowledge of ecological processes, such as
regeneration, that might lead to their improved management. To further this knowledge, I
monitored seedling populations and conducted experimental studies across a gradient of forest
management intensities in a seasonally dry lowland tropical forest in Bolivia. In the experiments,
I planted seeds of canopy tree species and manipulated moisture availability, seed and seedling
predation, and competition with an abundant understory bromeliad, Pseudananas~ddd~~~ddd~~~dd sagenarius
Seasonally dry tropical forests are characterized by low total rainfall (typically < 1500
mm) and annual periods of drought (i.e., months during which evapotranspiration exceeds
precipitati on usually as sumed to b e months with < 100 mm of precipitati on)(Hol dri dge 1 967).
The rainfall regime of Bolivian, and apparently many other tropical dry forests (Gerhardt 1996a),
is also characterized by huge interannual variation in precipitation, dry season duration, temporal
continuity of rainfall during the rainy and dry seasons, and the starting and ending dates of the
seasons. In addition, paleoecological and archaeological data predict that the dry seasons in
tropical dry forests in South America will be extended in duration and drier in the future (Mayle
et al. 2007). These sources of environmental variation are critical because, although the tree
species characteristic of tropical dry forests are presumably adapted to seasonal drought,
mortality is reportedly concentrated during dry seasons and dry years (Khurana and Singh 2001,
McLaren and McDonald 2003).
Within species and age cohorts, the probability of mortality reportedly decreases with
increasing seedling size, presumably because larger individuals have greater access to soil water
(Khurana and Singh 2001). Seedling growth rates typically increase with increasing illumination
but, due to high temperatures and soil surface drying, seedlings growing in large canopy gaps
may suffer high risks of mortality, even if they grow larger than their shaded counterparts in the
understory. In other words, in a forest with a 6-month rainy season, a newly germinated seedling
has 6 mo to grow large enough to survive the subsequent dry season. Based on this idea, I
conducted an experiment in which I manipulated the duration and intensity of the dry season and
monitored the survival and growth of seedlings of canopy tree species that germinated from
sown seeds. In recognition of the importance of inter-specific competition and seed/seedling
predators to seedling dynamics, I also manipulated these factors in a factorial experiment.
In addition to tolerating drought, tropical tree seedlings must compete for light, water,
and nutrients with plants of various growth forms including lianas (Gerwing 2001, Schnitzer and
Bongers 2002), ferns (George and Bazzaz 1999), and ground bromeliads (Fredericksen et al.
1999, George and Bazzaz 1999). In the seasonally dry forests of the Chiquitania of lowland
Bolivia, competition with a clonal ground bromeliad, Pseudananas~ddd~~~ddd~~~dd sagenarius, seems
particularly intense. This bromeliad covers 25-30% of the ground surface in my study area where
it presumably competes for light and soil resources. Furthermore, its leaves intercept both rain
and falling seeds (Fredericksen et al. 1999), which should affect tree seedling recruitment,
growth, and survival. Finally, by providing cover to small mammals and other animals avoiding
predators, bromeliads may also affect rates of seed and seedling predation. To investigate these
effects, I included a bromeliad removal treatment in my experimental study on the survival and
growth of canopy tree seedlings.
Predators often reduce seed and seedling survival in tropical forests (Janzen 1971, Hulme
1996, Asquith et al. 1997). Given that seed predation rates, at least by mammals, typically
increase with seed size (Moles and Westoby 2003), and given that most tree species in dry
tropical forests have small, wind-dispersed seeds, rates of seed predation may be lower in dry
than in humid forests. On the other hand, given the seasonal scarcities of food in dry forests, seed
and seedling predation might be particularly intense, especially for seedlings that remain leafed
out and succulent during the early dry season. To investigate the importance of seed and seedling
predation, as influenced by seasonality, I experimentally manipulated both soil moisture and
accessibility to predators using the planted seeds of canopy tree species.
In this chapter I present the results of a study in which I monitored the survival and
growth of tree seeds planted in plots in which I experimentally manipulated soil moisture, seed
and seedling predator access, and competition with ground bromeliads. The chapter also includes
the results of 3 y of monitoring of seedling establishment, growth, and survival in a large control
plot and two otherwise similar plots subj ected to different intensities of timber harvesting 7 mo
prior to commencement of my study. I hypothesized that seedling survival and growth are
reduced by bromeliad cover, drought stress, and predators. I also hypothesized that the
establishment, survival, and growth of naturally regenerated seedlings differs among plots
subj ected to different intensities of disturbance resulting from forest management activities.
Study Area and Climate
This study was conducted during 2003-2006 in the 30,000 ha seasonally dry lowland forest
(Holdridge 1967) owned by INPA PARKET, about 30 km from Concepci6n, Bolivia (160 6'
45"S, 610 42' 47"; Figure 1-1). The study area is located in the transition zone between the
forests of the Amazon Basin and those of the Gran Chaco and is locally referred to as Chiquitano
dry forest (Killeen et al. 1998, Ibisch and Merida 2003). Based on 30 years of data collected in
Concepci6n, the annual average temperature in the area is 24.3 OC with a minimum average
temperature of 12.90 C, generally in July, and maximum average temperature of 310C, generally
in November. Over the 30-year monitoring period, the annual average of rainfall was 1 100 mm
but ranged from 700-2000 mm/year, with wet years corresponding with El Nifio events (Coelho
et al. 2002) (Figure 2-1). Both 2005 and 2006 were dry years, with only 980 and 1050 mm of
precipitation, respectively. The rainy season typically runs from October to April, but in 2005,
the rains started in September and rain fell at regular intervals until March of 2006. The 2006
rainy season, in contrast, started in October but then no rain fell in November, with more regular
rains commencing again in December.
The set of Long-Term Silvicultural Research Plot (LTSRPs) established at this site by the
Institute Boliviano de Investigaciones Forestales (IBIF) was used for a portion of this study. The
LTSRP at INPA includes four 20-ha plots that vary in management intensity: unlogged control;
normal logging; improved management; and, intensive management (Mostacedo et al. 2006). For
this study, the improved management plot was not used. In the "normal logging" plot, the
logging company harvested a mean of 4.3 trees/ha (4.7 m3/ha) using their standard reduced-
impact-logging techniques that include road planning, directional felling, and the retention of
20% of the harvestable trees as seed trees. The mean density of trees > 10 cm DBH before
logging was 427 individuals/ha, of which 40% were Acosnzium cardena~sii (after first mention of
a species I will refer to it by its generic name; for a complete list of scientific names see Table 2-
1). The company used these same logging techniques in the "intensive management" plot to
harvest a mean of 8.1 trees/ha (8.2 m3/ha). While harvesting the "intensive management plot,"
the skidder drivers mechanically scarified the soil surface in an average of 0.6 felling gaps/ha
(mean area = 50 m2/ha) where there was no existing regeneration of commercial timber species.
After logging of the intensive management plot, future crop trees (i.e., well-formed trees of
commercial species 10-40 cm DBH) were liberated from liana cover (by slashing the lianas with
a machete; 21 trees/ha) and liberated from competition from nearby non-commercial trees (by
poison girdling; 1.7 trees/ha). These last two treatments for enhancing the growth of future crop
trees were not applied in the normal logging plot. In the unlogged forest plot, the mean density of
trees > 10 cm DBH was 432 individuals/ha, of which 3 8% were Acosnzium (Table 2-1). The total
basal area of trees >10 cm DBH in the unlogged plot averaged only 19.6 m2/ha and 3 8% of the
trees were liana infested (18% severely so).
Canopy openness in the early dry season (May 2003), as measured with a spherical
densiometer (Lemmon 1957) held 1 m above the ground at 140 equally spaced points in the 10
ha permanent plots, averaged 8% in the 20 ha control (unlogged) plot, 13% in the plot subj ected
to normal selective logging in November 2002, and 14% in the plot that was intensively
harvested and silviculturally treated also in November 2002 (Figure 2-2). I remeasured canopy
openness at the same points in March 2006 and found increases in canopy cover in both of the
treated plots, but not in the control plot (Figure 2-2).
Experimental Design and Data Analysis
Response of regeneration to silvicultural treatment intensity
Plants 5-100 cm tall of 22 species of subcanopy, canopy, and emergent tree species (Table
2-2) that originated from seeds or sprouts were monitored in subplots in three of the 20-ha (400 x
500 m) LTSRP plots. In the central 10 ha (400 x 250 m) of each plot, I located 144 pairs of 2 x 1
m subplots separated by 2 m at 25 m intervals (data from the paired plots were subsequently
combined). Each seedling was marked, mapped, and measured for height at regular intervals of
one year for 3 y. The site of each plot pair was categorized in one of the following microsites:
undisturbed forest (includes natural canopy gaps and high stature forest); logging road; and,
logging gap. Undisturbed forests were categorized as those sites in which the structure of the
forest did not change during logging. I initially separated skid trails from primary and secondary
logging roads, but because of small sample sizes, these sites are combined into a category
referred to as log extraction paths. Logging gaps were sites where the canopy was opened during
tree felling and log extraction. Based on the literature (Whitmore 1998, Mostacedo and
Fredericksen 1999, Pinard et al. 1999) and field observations, each tree species was placed in one
of the following ecological groups: light-demanding pioneer; long-lived pioneer; partially shade
tolerant; or, shade tolerant.
I compared seedling abundances among the harvesting treatment plots and microsites using
repeated measures ANOVA (Scheiner and Gurevitch 2001). Measurement dates were considered
to be a within-subj ect factor, while harvesting treatment and microsites were considered as
between-subj ect factors.
Seedling recruitment rates (R) were calculated using the compound interest equation
(McCune and Cottam 1985):
R = (1 +Bx)'lx 1 (2-1)
where Bx is the birth rate in the period x calculated for each year and for each m2 Of
ground area. As recruitment rate I calculated the number of new seedlings of each species in
each harvesting treatment m2/er
Mortality rate (M)1 expressed as percentage per year was calculated as (Primack et al. 1985,
Sheil et al. 1995)
M~ = 1- [1- (No N,)l /No]"' (2-2)
where No is the number of living seedlings at time 0, N, is the number of living seedlings at
time N1, and t is the time period between No and N Mortality rates were calculated by species
and harvesting treatment.
I calculated the relative growth rate (RGR) as
(Ln(H, )- Ln(Ho))
RGR = (2-3)
where, Ho is the seedling at the initial measurement, H; is the height at the second
measurement, to is the initial time, and tl is the time of the second measurement (Hoffmann and
Poorter 2002). I compared RGR of harvesting treatments, microsites and species using repeated
measures ANOVA. For all analyses I used SPSS Version 12.0. For each factor, I also conducted
Bonferoni or LSD post hoc comparisons at the 95% confidence level.
Factors affecting seedling establishment and growth
Tree species used in the following experimental studies were selected on the basis of seed
availability in May-September 2005 when I collected seeds from 3-10 trees per species. Of the
six species used in the experiment, 4 are wind-dispersed, 2 are dispersed by animals, 2 are
characterized as light-demanding, and 4 are considered shade tolerant (Table 2-1). Brief
descriptions of the species follow:
Amburana cearensis (Allemio) A. C. Sm. usually produces one 0.49 g dry weight winged
seed per dehiscent legume. Aspidosperma rigidum Rusby produces elliptical seeds that weigh
0.24 g and are surrounded by a wing. Ceiba samnauma (Mart.) K. Schum. produces 0.17 g seeds
that are covered by fine, cotton-like fibers that aid in wind dispersal. Pterogyne nitens Tul.
produces single-seeded wind-dispersed samaras with 0.16 g seeds. Copaifera chodatiana Hassl.
seeds are red, weigh 0.33 g, are covered with a white aril, and are animal-dispersed. Hymenaea
courbaril L. produces round, large (3.9 g), and animal-dispersed seeds in indehiscent legumes.
The experiment used a split-plot design (Montgomery 2001) in which ground bromeliad
cover, water availability, and seed and seedling predation were experimentally manipulated
(Figure 2-3). The 2 m2 experimental plots were comprised of the two 1 m2 Subplots which were
replicated 40 times in 10 blocks separated by at least 40 m. The experiment commenced in July
2005 and was monitored until December 2006.
The terrestrial bromeliad manipulated in this study is Pseudananas~ddd~~~ddd~~~dd sagenarius, locally
known as "garabata." Rosettes of this clonal species are about 1 m high and cover 1m2 each.
Their colonies typically cover 15-40% of the ground in Chiquitano dry forest in clumps up to
2000 m2. This species is more common in mature than in young stands (Kennard et al. 2002) and
its abundance is reduced by fire (Fredericksen et al. 1999).
The water availability treatment involved either irrigating or withholding water from
plots by adding water or shielding plants from rainfall during what is typically the transition
between the dry and rainy seasons (late September). Rainfall inputs directly into the drought
treatment plots were prevented by covering them with 2.0 x 1.5 m roofs of transparent plastic.
The rain shelters were left for the first 3 mo of what is normally the dry season. The irrigated
plots received natural rainfall plus 10 liters of water every 15 days (=20 mm/mo) starting on 21
September 2005 and finishing in at the end of November when rainfall were more frequent. Soil
water tensions at 0-10 cm depth in one droughted and one irrigated plot in each of four blocks
were monitored hourly with Watermark@ soil moisture sensors attached to data loggers; rainfall
inputs were monitored adj acent to the four irrigated plots (Figure 2-1). Droughted plots
simulated years with prolonged dry seasons whereas the irrigated plots simulated wet years with
short dry seasons.
To exclude seed and seedling predators, half of the plots were surrounded by 60 cm high,
1-mm mesh wire netting fastened to the ground. The exclosures were erected before the seeds
were sown and were maintained for the duration (15 mo) of the experiment. In each treatment
plot, I randomly sowed seeds of all tree species described above (Table 2-1). Each seedling
establishing from a planted seed was marked, mapped, and its height measured five times
between September 2005 and December 2006.
Treatment effects on seedling abundance were tested using repeated measures ANOVA for
a split plot design (Scheiner and Gurevitch 2001) with bromeliads (present or absent) as the main
factor, and irrigation, predator exclosure, and time as subplot factors. Seedling heights (RGR) for
each time period were compared between treatments for each factor using t-test with sequential
Bonferroni corrections (Sankoh et al. 1997). Due to seedling mortality over the study, the sample
sizes varied so I could not run repeated measures ANOVA on the RGR data. I ran the analysis
for all 6 species combined and for each species separately.
Response of Regeneration to Management Intensities
Three years after logging and silvicultural treatment, seedling densities varied with
management intensities and among microsites. Although 22 species are included in Table 2-1,
only data from the most abundant 11 species are presented. Densities were higher in the control
than in the harvested areas, but did not differ between the normal harvesting and the intensive
management plots (Figure 2-4). However, seedling densities without Acosmium, was higher in
normal harvesting and lower in the control plot (Figure 2-4). This pattern was maintained five
years after logging for all species combined, but for nine of the 11 species evaluated, there were
treatment differences and temporal trends (Table 2-2). For example, Copaifera seedling density
increased with increasing management intensity. In contrast, Machaerium acutifolium Vogel
seedling densities did not differ between the two logged plots but were significantly higher in the
normal logging plot than in the control plot. In contrast, seedling densities were lower in the
intensive management plot than in the normal logging or control plot for four Acosmium,
Phyllostylon rhamnnoides (J. Poiss.) Taub., and Machaerium scleroxylon Tul.
Seedling densities varied markedly among microsites. Overall, undisturbed sites had the
highest seedling density (Table 2-3). This pattern was observed at the species level for
Acosmium, Caesalpinia pluviosa,, M scleroxylon and Phyllostylon. In contrast, Centrolobium
microchaete seedling densities were higher on logging roads than in other microsites. The other
nine species had statistically similar seedling densities among microsites (Table 2-3).
Overall, seedling mortality rates were higher in undisturbed areas than in harvested areas
and the mortality rate was higher in 2004 than in 2005 (Figure 2-5). In 2004, 47% and 66% of
the seedlings died in the normal and intensive management plot, respectively. In 2005, mortality
was still higher in undisturbed control plots, but the ratio was reduced (Table 2-4). At species
level, mortality rates were generally higher in undisturbed control plots than in either of the
logged plots. On average, the highest (22-26%/y) mortality rates were for Sweetia f~uticosa
Spreng., Caesalpinia, and M~ scleroxylon. At the other end of the spectrum, Gallesia integrifolia
(Spreng.) Harms and Piptadenia~~ttt~~~~ttt~~~ viridifolia (Kunth) Benth, suffered no mortality (Table 2-4).
Over all species, the recruitment of new seedlings was 0.8 seedlings /m2 /y. Recruitment
was three times higher in undisturbed areas than harvested areas (Figure 2-5) and marginally
higher (0.91 seedlings /m2 /y) in 2004 than in 2005 (0.63 seedlings /m2 /y). Acosnzium had the
highest recruitment rate (0.26 seedlings/m2/y), followed by Phyllostylon (0.13 seedlings/m2/y
and M~ acutifolium (0.08 seedlings/m2/y). Acosnzium and M~ acutifolium recruited mostly in
undisturbed forest, Phyllostylon recruits were common in the normally logged plot, and
Pterogyne and Aspidospernza recruited most new seedlings in the intensive management plot
Factors Affecting Seedling Growth and Establishment
Relative growth rates of seedlings differed among the three logging microsites either in
2004 (F=4.19, P=0.01) or in 2005 (F=10.76, P=<0.0001). In both years, RGRs were higher on
loging roads and logging gaps than in undisturbed microsites. RGRs were higher in 2004 than in
2005 (F=5.64, P=0.018) and were higher for long-lived pioneer species RGRs than partial shade
tolerant and shade tolerant species (F=15.63, P=<0.0001).
RGRs differed among years, species, and microsites following patterns that were not
consistent. For example, the RGRs of Acosnzium, Aspidospernza, Copaifera, and Sweetia were
higher in 2004 than in 2005 (F=7.74, P=0.005). For Aspidospernza and Caesalpinia, RGRs did
not differ among undisturbed sites, logging roads, and logging gaps. Among microsites, RGRs
for M acutifolium were higher in logging roads and logging gaps than in undisturbed microsites
in 2004 but not in 2005. RGRs of Acosnzium and Sweetia were similar among microsites in 2004
whereas in 2005, seedlings grew faster in logging roads and logging gaps than in undisturbed
Monthly RGRs of seedlings in the experimental plots were similar in bromeliad-free and
bromeliad-infested plots at different times (tl=0.14, P1=0.89; t2=2.01, P2=0.05; t3=0.48, P3=0.63;
t4=0.04, P4=0.39). Similarly, RGRs were did not differ in the droughted and irrigated plots
(tl=0.89, P1=0.37; t2=0.87, P2=0.39; t3=0.93, P3=0.35; t4=0.88, P4=0.38) or between the exclosure
and non-exclosure plots (tl=0.01, P1=0.99; t2=0.12, P2=0.90; t3=0.40, P3=0.69; t4=0.36, P4=0.72;
Figure 2-10). At the species level, RGRs were also similar in bromeliad-free and bromeliad-
infested plots, droughted and irrigated plots, and exclosure and non-exclosure plots (Table 2-6,
In general, tree seedling establishment was suppressed by bromeliads, but responses to
bromeliad cover varied among species (Figure 2-6). Based on the six species for which there are
sufficient data, four had similar numbers of seedlings in bromeliad patches and bromeliad-free
areas. For the other two species (Hynzenaea and Pterogyne), seedling abundances were lower in
bromeliad covered plots (Table 2-5, Figure 2-7).
The effects of the drought and irrigation treatments varied over time (Figure 2-6).
Seedlings were initially about 150% more abundant in the irrigated than in the droughted plots
(Figure 2-8), but this difference diminished over time. This pattern was consistent among species
except for Aspidospernza, which had more seedlings in the droughted than in the irrigated plots
Excluding seed and seedling predators generally resulted in higher seedling densities
(Figure 2-6), with some variation among species (Figure 2-9). Anaburana, Copaifera, and
Hymenaea seedlings were more common within the excloures than in the open-access plots
(Table 2-5). In contrast, the exclosure treatment had no effect on seedling densities of
Aspidosperma, Ceiba, and Pterogyne.
There were some significant interactions among factors both overall and at the species
level. In particular, irrigated plots with bromeliads from which predators were excluded had
higher seedling densities than plots that were irrigated and open to seed and seedling predators
At the species level there were significant interactions among treatments only for Ceiba
and Copaifera. For Ceiba in the bromeliad covered plots, irrigation and predator exclosure
increased seedling densities relative to the droughted and open-access plots. For Copaifera in the
bromeliad-infested plots, irrigation and predator exclosure resulted in higher seedling densities.
In the bromeliad-free plots, drought coupled with predator exclosure resulted in higher seedling
densities than drought alone.
Logging Effects on Seedlings Dynamics
Logging usually affects seedling densities due to direct effects and creation of microsites
that differ in environmental conditions, especially light intensities (Beaudet and Messier 2002,
van Rheenen et al. 2004). In most tropical forests, seedling densities of light-demanding species
increase on open microsites. In this study, however, overall seedling densities were higher in
undisturbed forest than in selectively logged areas (Figure 2-2). Similarly, total seedling densities
were higher on undisturbed microsites compared to microsites created by logging. In contrast, at
a site only 40 km from INPA, with a similar forest type and rainfall regime Fredericksen and
Mostacedo (2000) reported that seedlings of commercial timber species increased in abundance
on log landings, roads, and logging gaps.
At the species level, the maj ority of species (64%) had similar seedling densities among
microsites; only three shade-tolerant species had higher seedling densities in undisturbed
microsites. One light-demanding species, Centrolobium, had higher seedling densities on skid
trails and log landings compared to other microsites.
There are several potential explanations for observed differences between the results of this
study and those of Fredericksen and Mostacedo (2000). The forest canopy of Las Trancas where
the latter conducted their work is dominated by Anadenanthera, a light-demanding species,
whereas INPA Parket is dominated by Acosnzium, which is more shade tolerant. Results without
Acosnzium show that harvested plots have higher seedling densities than control plot (Figure 2-
4). The simple difference in dominants might account for some of the overall differences
observed between the sites. Another difference between the two studies is that Anadenanthera
seedlings emerging from seeds dispersed immediately before logging were extremely abundant
at Las Trancas (100-200 individuals/m2; perSonal observation). In contrast, seedlings of
Acosnzium were never as abundant at INPA and those that were present were apparently the
result of several years of recruitment.
Open microsites created by logging may be disadvantageous for seedling recruitment in
INPA Parket because such sites are extremely hot and dry, especially during the dry season. On
the other hand, it is possible that even though there were disturbances associated with fallen
trees, roads, and skid trails, the canopy could still be too closed at INPA to promote abundant
seedling recruitment (Jackson et al. 2002). In addition, the densities of understory plants and
other competitors might remain high in gaps, thus suppressing seedling recruitment in disturbed
areas, especially of shade-intolerant species (Beckage et al. 2000, Schnitzer et al. 2000). On the
log extraction paths, soil compaction might negatively affect seedling establishment and growth.
Finally, several species in this study are shade-tolerant, including the dominant (Acosmium), and
do not require disturbed areas for regeneration.
Seedling recruitment rates were 70% lower in the logged plots than in the unlogged control
plot and recruitment decreased with management intensity. Because of increases in light
availability associated with logging, I expected more recruitment in logged plots than in
unlogged plots. One explanation for this result could be that logging decreased seed-tree
densities, which limited seed inputs (Forget et al. 2001, Makana and Thomas 2004, Grogan and
Galvao 2006). Similarly, in French Guiana, Forget et al. (2001) reported that for all but one
species, recruitment was not favored by logging. The importance of retention of seed trees was
shown by Grogan and Galvao (2006), who reported that mahogany (Swietenia macrophylla) seed
production near gaps was important for recruitment. In addition, even if seed production is not
limiting, logging may create microsites that are not appropriate for seed germination and
seedling growth due to soil compaction (Pinard et al. 1996) or other factors. Site preparation
(e.g., soil scarification) in logging gaps and skid trails can enhance regeneration, as shown by
Pinard et al. (1996) in a dipterocarp forest in Malaysia, but such treatments do not always have
beneficial effects (Heuberger et al. 2002).
Seedling mortality rates were nearly 50% lower in the logged plots than in the unlogged
plot, which was the opposite of what I expected. Several studies report that seedling mortality
usually increases as harvesting intensity increases (Chapman and Chapman 1997, Saenz and
Guariguata 2001), but the logging intensities in my plots were all low. Nevertheless, mortality of
some plant species decreases as light intensity increases. For example, in a tropical rain forest in
Costa Rica, Kobe (1999) found that mortality rates of three of four canopy species monitored
decreased as light levels increased in the understory.
There was considerable interspecific variation in both recruitment and mortality rates. For
example, for three tree species (Acosnzium, Aspidospernza, and Copaifera) that are in the same
shade tolerant guild the percentage of recruitment ranged from 7.5-15.5 %/y. The mortality rates
for the same species ranged from 0.019-0.25 5 individuals/m2/y. Acosnzium, because of its
dominance, drives the overall patterns of recruitment and mortality.
Apparently, while recruitment rates are reduced by disturbance, the seedlings that do
establish grow sufficiently during the rainy season to survive the subsequent dry season. In
contrast, in the relatively closed conditions of the control plot, few of the abundant new recruits
survive the drought stress of their first dry season. Similarly, seedlings of a light-demanding
species (Anadenanthera) established abundantly, but died rapidly in the control plot.
Temporal changes in recruitment and mortality are related to variation in seed production
and rainfall regime (Kitajima and Fenner 2000, Khurana and Singh 2001). In this study, seed
production, seedling recruitment, and seedling mortality rates were all higher during 2003-2004,
a dry year, than during 2004-2005, a relatively wet year with a short dry season.
Within the logged plots, seedling RGRs were typically higher in disturbed microsites, but
there was some variation among species. Over the 2 y monitoring period, seedlings grew faster
on log extraction paths than undisturbed microsites. In a similar study in the same area but using
large transects, seedlings in skid trails and logging gaps grew faster than seedlings in undisturbed
forest (van Andel 2005). Contrary to my expectations, Acosnzium, a shade-tolerant species, had
the highest RGR on log extraction paths while RGRs of Anadenanthera, a light-demanding
species, not vary among microsites. Overall it appears that disturbed microsites are favorable for
seedlings, but this pattern seems to be driven by Acosnzium, the most abundant species. For most
of the species it seems that, in contrast to wetter tropical forests, RGRs of tropical dry forest
seedlings are not governed primarily by light availability and have a different relationship to
disturbance than wet and moist forest tree species (Khurana and Singh 2001).
Factors Affecting Seedling Establishment and Growth
Soil moisture availability is a key factor influencing seedling growth in dry forests
(Khurana and Singh 2001). In my study, seedlings that lost their leaves during the dry season
(e.g., Amburana and Aspidosperma) grew slowly, while the evergreen species (e.g., Copaifera
and Pterogyne) grew faster. I also found that bromeliad cover significantly reduced seedling
Ground bromeliads interfered with the establishment and growth of tree seedlings in this
study, as has been described for another forest in Bolivia (Fredericksen et al. 1999), as well as on
Barro Colorado Island in Panama (Brokaw 1983). As in those other two studies, I also observed
substantial interspecific variation in seedling responses to bromeliads. The observation that the
irrigation treatment eliminated the generally negative effect of bromeliads on seedling survival
and growth suggests that the rain-trapping effect of bromeliad rosettes, coupled with any water
they draw from the soil, is the cause of their deleterious effect. Perhaps the species that were not
sensitive to the presence of bromeliads (Amburana, Ceiba, Copaifera, and Pterogyne) are more
drought-tolerant than the other species.
I expected that experimentally shortening the dry season by irrigation would increase the
establishment and growth of tree seedlings, while exacerbating dry season drought would have
the opposite response. Instead, what I found that the principal impact of the drought treatments
was the timing of seed germination, but not all species were affected. Seeds in the droughted
plots did not germinate until after I removed the roofs and allowed rain to fall on the soil;
seedlings then reached the same densities as in the irrigated plots. A similar study in mahogany
in a dry forest in Guanacaste, Costa Rica reported the same effect of water availability on
seedling density (Gerhardt 1996b). Similarly, RGR were not affected by drought stress.
The tendency for species overall (except Aspidospernza) was higher densities in the
irrigated than in the droughted plots at the time of germination. This tendency suggests that
longer rainy seasons, such as those occurring during El Nifio years, can have dramatic effects on
seedling establishment (Gilbert et al. 2001, Engelbrecht and Kursar 2003). The same pattern of
increased mortality during dry years was revealed by the mortality of naturally established
seedlings during this study (Figure 2-5).
Due to a variety of morphological, physiological, and phenological traits, tree species
differ in the susceptibility of their seedlings to drought stress (Table 2-4; (Reader et al. 1993,
Khurana and Singh 2001, McLaren and McDonald 2003). For example, Anaburana seedlings
reduce their moisture requirements by being dry season deciduous, leaf areas ofAspidospernza
seedlings decline substantially during the dry season, and Acosnzium is noteworthy for its deep
root system (Wright 1991, Engelbrecht and Kursar 2003).
Seed or seedling predation seems to be a very important factor for seedling establishment
in tropical dry forests (Janzen 1971, Hulme 1996). I found that plots with exclosures had higher
seedlings densities than open plots. Exclosures prevented seedling herbivory, but not all types of
seed predation because insects and other small seed predators were not excluded (Grol 2005). In
tropical dry forests, herbivory has the most impact on tree regeneration when seedlings are
emerging, not after roots and stems are lignified (Kitajima and Augspurger 1989, Lucas et al.
2000). Once browsed, established seedlings usually resprout, but only lignified seedlings
survive. In some cases, herbivory damage could be superficial and not dramatically impact the
growth and survival of seedlings (personal observation).
There were some species in which exclosure treatments had an effect on established
seedlings, while for others there were similar. For example, Hymenaea, a species with a large
seed and seedlings larger than Ceiba, had a higher established seedling caused by the exclosure
plots. Several studies support the hypothesis that large seeds are usually most predated (Blate et
al. 1998, Dalling and Hubbell 2002), but see Moles et al. (2003). Likewise, Aspidosperma and
Pterogyne have seeds that probably were predated mostly by small insects (i.e., bruchid beetles)
which were not deterred by the exclosures.
Five years after selective logging and silvicultural treatments, overall tree seedling
densities were higher in the control plot than in the two logged plots. It is important to point out
that this result was heavily influenced by the reduced seedling densities in the logged plots of the
dominant species, the shade-tolerant Acosmium. Other species, which were less common
everywhere (e.g., Centrolobium, Copaifera, and Machaerium cf: acutifolium), actually increased
in response to forest management activities, and many other species were not affected. Within
the logged plots, undisturbed microsites had higher seedling densities except for Centrolobium, a
light-demanding species that sprouts readily from damaged lateral roots, which was more
abundant on log extraction paths.
Seedling recruitment rates were higher in the unlogged control plot and on undisturbed
microsites in the logged plots, but seedling in these areas also suffered higher mortality rates than
seedlings in disturbed areas. The higher turnover rates of seedlings under closed canopy
conditions coupled with their observed lower growth rates suggests that rapid growth under open
conditions during the rainy season is critical for seedling survival during the subsequent dry
season. This result is somewhat surprising given that, even undisturbed forest in my study area,
the canopy is 30-40% open after leaf fall.
Mortality rates of naturally established seedlings varied greatly among species. Seven of
22 species suffered no mortality during the 2-y monitoring period, whereas relatively high
mortality rates were observed for Caesalpinia (26%/y), Sweetia (25%/y), and Machaerium
scleroxylon (22%/y). Mortality rates of these species in the control and logged plots followed
Bromeliad competition and seed/seedling predators greatly reduced seedling recruitment in
this tropical dry forest. Experimentally augmented soil moisture also increased seedling
establishment, but only in interaction with bromeliad removal or predator exclosure.
Experimental droughting delayed germination but did not influence seedling densities once the
treatment terminated and rainfall was allowed into the plots. Despite these general trends, species
varied in their sensitivities to bromeliads, drought stress, and predators.
Finally, seedling growth was not promoted by experimentally lengthening the rainy season.
Nevertheless, the larger seedlings that developed during the extended rainy season were more
likely to survive the subsequent dry season. Given that global change models consistently predict
that the study region will receive less rainfall and suffer longer dry seasons in the future (e.g.,
Mayle et al. 2007), regeneration failures are likely to become more common.
Table 2-1. Spatial distributions, crown position, ecological group, geographical range, tree densities and basal area of timber species
in a Bolivian tropical dry forest.
(#/ha >10 cm
Tabebuia .,~, ruagest..w
Crown position: (EM)
Ecological group: (L) =
emergent, (CA) = canopy, (SC)
light-demanding pioneer, (LL) =
Long-lived pioneer, (PS)
partially shade tolerant, (TS) = shade tolerant.
Geographical range: (R) = restricted, (I) = intermediate, (W) = widespread. Restricted species are those that are found only in one type
of forest and are limited to the Chiquitano dry forest in Bolivia. Intermediate species are those that are found in two or three
forest types but are either restricted to Bolivia or are only rarely found in neighboring countries. Widespread species are found in
several forest types and in other countries.
Species marked with asterisks were used in the seedling survival experiment.
Table 2-2. Means of seedling densities in an unharvested control plot, a plot subjected to normal timber harvesting, and a plot
subj ected to more intensive harvesting followed by silvicultural treatments. Subplots within the main treatment plots are
treated as replicates. Different letters indicate statistical differences (P<0.05) among harvesting treatments as indicated by
repeated measures ANOVAs and LSD post hoc tests.
(0.01)a 0.02 (0.01)b 0.05 (0.01)a
Sweetia fiuticosa 0.03
0.07 2.53 0.080
Table 2-3. Mean seedling densities (#/m2) for 3 years (2003-2005) of commercial tree species in microsites created by logging in a
Chiquitano dry forest. Sample size for each microsite type is noted within brackets; standard errors of the means are noted
in parenthesis. Seedling densities in the different microsites were compared with repeated measures ANOVA with
microsites as the between-subject factor and time as the within-subj ect factor. Different letters indicate different
significance between treatments using pairwise Least Significant Difference (LSD) tests.
Scientific Name Logging Gap  Log Extraction Path  Undisturbed  Mean Square F P
4cosiniuin cardenasii 0.701 (0.103)b 0.389 (0.14(,) 0.945 (0.056)a 24.92 7.88 <0.0001
4nadenanthera inacrocarpa 0.018 (0.015) 0.049 (0.021) 0.050 (0.008) 0.119 1.71 0.18
4spidosperina rigiduin 0.044 (0.024) 0.026 (0.033) 0.056 (0.013) 0.07 0.39 0.68
Caesalpinia pluviosa 0.024 (0.015)b 0.037 (0.020)b 0.073 (0.008)a 0.32 4.71 0.009
Ceiba sainauina 0.000 (0.001) 0.000 (0.001) 0.001 (0.000) 0.0001 0.68 0.50
Centrolobiwn inicrochaete 0.008 (0.005)b 0.022 (0.007)a 0.005 (0.003)b 0.02 2.86 0.05
Chorisia speciosa 0.004 (0.003) 0.002 (0.004) 0.004 (0.002) 0.001 0.23 0.79
Copaifera chodatiana 0.092 (0.045) 0.086 (0.061) 0.146 (0.024) 0.51 0.85 0.43
Gallesia integrifolia 0.000 (0.003)b 0.000 (0.004)b 0.007 (0.002)a 0.009 3.69 0.03
Machaerium acutifolium 0.609 (0.133) 0.668 (0.181) 0.519 (0.072) 2.13 0.40 0.67
Machaerium scleroxvion 0.008 (0.011)b 0.000 (0.015)b 0.048 (0.006)a 0.29 7.89 <0.0001
Phyllostylon rhainnoides 0.007 (0.025) 0.000 (0.035) 0.122 (0.014) 2.203 11.34 <0.0001
Pterogvne nitens 0.000 (0.004) 0.006 (0.005) 0.005 (0.002) 0.003 0.91 0.41
Sweetia fruticosa 0.015 (0.009) 0.011 (0.012) 0.031 (0.005) 0.05 2.24 0.11
Overall Species 1.543 (0.180)b 1.302 (0.245)b 2.019 (0.098)a 52.87 5.44 0.005
Table 2-4. Establishment and mortality rates of seedlings of commercial tree species monitored over a 3 y period in a control plot and
plots subj ected to two harvesting intensities (N=144 subplots/treatment plot).
Scientific Name Mortality Rate (% of Seedlings / y) Number of New Seedlings (#/m2 /y
Normal Intensive Normal Intensive
Undisturbed Logging Management Overall Undisturbed Logging Management Overall
Acosmium cardenasii 17.0 11.7 9.8 12.8 0.489 0.176 0.101 0.255
Anadenanthera macrocarpa 21.6 5.4 18.8 15.2 0.106 0.079 0.003 0.062
Aspidosperma rigidum 14.5 2.5 5.5 7.5 0.007 0.013 0.037 0.019
Caesalpinia pluviosa 35.8 23.2 18.8 25.9 0.117 0.064 0.011 0.064
Ceiba samauma 0.0 0.0 0.0 0.0 0.001 0.001 0.001 0.001
Centrolobium microchaete 0.0 10.7 43.8 18.2 0.000 0.006 0.000 0.002
Chorisia speciosa 0.0 0.0 0.0 0.0 0.004 0.004 0.000 0.003
Copaifera chodatiana 24.1 8.2 14.3 15.5 0.125 0.015 0.050 0.063
Gallesia integrifolia 0.0 0.0 0.0 0.0 0.034 0.000 0.000 0.011
Machaerium c~f acutifolium 16.1 12.4 12.6 13.7 0.003 0.138 0.106 0.082
Machaerium scleroxylon 34.7 30.8 0.0 21.8 0.157 0.012 0.001 0.056
Phyllostylon rhamnoides 19.5 10.0 0.0 9.8 0.381 0.004 0.011 0.132
Pterogyne nitens 43.8 0.0 0.0 14.6 0.000 0.001 0.022 0.008
Sweetia fr~uticosa 41.4 24.8 8.3 24.8 0.020 0.010 0.004 0.011
22.0 12.4 11.2 15.2 1.446
0.523 0.354 0.775
Table 2-5. Results of repeated measure analysis of variance for a split plot design run for
seedling density for all species combined or six timber tree species analyzed
separately. Double asterisks indicate between subj ect factors while the unmarked
variables are within subj ect factors.
Sum of Degrees of Mean
Source Squares Freedom Square F P-value
Time (Ti) 44.01 4 11.00 25.88 <0.0001
Bromeliad (Br)(**) 163.20 1 163.20 9.99 0.005
Time*Bromeliad (Br) 4.95 4 1.24 2.90 0.03
Irrigation (Ir) 27.83 1 27.83 2.74 0.11
Irrigation*Bromeliad 7.16 1 7.16 0.70 0.41
Exclosure (Ex) 79.66 1 79.66 13.01 0.002
Ex*Br 18.71 1 18.71 3.06 0.09
Ti*Ir 13.99 4 3.49 10.81 <0.0001
Ti*Ir*Br 1.92 4 0.48 1.48 0.22
Ti*Ex 7.75 4 1.94 5.76 <0.0001
Ti*Ex*Br 1.43 4 0.36 1.06 0.38
Ir*Ex 11.73 1 11.73 1.31 0.27
Ir*Ex*Br 16.60 1 16.60 1.86 0.19
Ti*Ir*Ex 0.58 4 0.14 0.39 0.81
Ti*Ir*Ex*Br 0.35 4 0.09 0.24 0.92
Time (Ti) 1.74 4 0.44 7.72 <0.0001
Bromeliad (Br)(**) 0.83 1 0.83 0.43 0.53
Time*Bromeliad (Br) 0.13 4 0.03 0.58 0.68
Irrigation (Ir) 3.01 1 3.01 1.87 0.20
Irrigation*Bromeliad 1.09 1 1.09 0.68 0.43
Exclosure (Ex) 4.50 1 4.50 2.59 0.14
Ex*Br 2.52 1 2.52 1.45 0.26
Ti*Ir 1.35 4 0.34 6.03 0.001
Ti*Ir*Br 0.08 4 0.02 0.36 0.84
Ti*Ex 0.55 4 0.14 2.39 0.07
Ti*Ex*Br 0.22 4 0.05 0.95 0.44
Ir*Ex 8.87 1 8.87 3.10 0.10
Ir*Ex*Br 0.09 1 0.09 0.03 0.86
Ti*Ir*Ex 1.03 4 0.26 3.65 0.01
Ti*Ir*Ex*Br 0.08 4 0.02 0.29 0.88
Time (Ti) 1.16 4 0.29 8.59 <0.0001
Bromeliad (Br)(**) 0.59 1 0.59 0.25 0.63
Time*Bromeliad (Br) 0.04 4 0.009 0.28 0.89
Irrigation (Ir) 5.37 1 5.37 1.78 0.23
Irrigation*Bromeliad 1.31 1 1.31 0.43 0.53
Exclosure (Ex) 0.19 1 0.19 0.10 0.76
Ex*Br 2.6 1 2.6 1.38 0.28
Table 2-5. Continued.
Sum of Degrees of Mean
Source Squares Freedom Square F P-value
Ti*Ir 0.32 4 0.08 1.95 0.13
Ti*Ir*Br 0.08 4 0.02 0.47 0.76
Ti*Ex 0.12 4 0.03 1.18 0.34
Ti*Ex*Br 0.13 4 0.03 1.24 0.32
Ir*Ex 1.50 1 1.50 0.87 0.39
Ir*Ex*Br 1.15 1 1.15 0.67 0.44
Ti*Ir*Ex 0.08 4 0.02 0.76 0.56
Ti*Ir*Ex*Br 0.08 4 0.02 0.76 0.56
Time (Ti) 0.53 4 0.13 2.02 0.12
Bromeliad(Br)(**) 7.75 1 7.75 4.65 0.07
Time*Bromeliad (Br) 0.05 4 0.01 0.21 0.93
Irrigation (Ir) 1.00 1 1.00 0.72 0.43
Irrigation*Bromeliad 0.53 1 0.53 0.38 0.56
Exclosure (Ex) 0.001 1 0.001 0.004 0.95
Ex*Br 0.06 1 0.06 0.54 0.49
Ti*Ir 0.30 4 0.08 1.06 0.39
Ti*Ir*Br 0.03 4 0.007 0.10 0.98
Ti*Ex 0.12 4 0.03 0.91 0.47
Ti*Ex*Br 0.05 4 0.01 0.42 0.79
Ir*Ex 1.69 1 1.69 9.28 0.023
Ir*Ex*Br 0.79 1 0.79 4.35 0.082
Ti*Ir*Ex 0.13 4 0.03 0.94 0.46
Ti*Ir*Ex*Br 0.09 4 0.02 0.67 0.62
Time (Ti) 4.1 4 1.03 15.09 <0.0001
Bromeliad (Br)(**) 2.96 1 2.96 1.49 0.24
Time*Bromeliad (Br) 0.16 4 0.04 0.60 0.66
Irrigation (Ir) 1.88 1 1.88 3.64 0.07
Irrigation*Bromeliad 0.06 1 0.06 0.12 0.73
Exclosure (Ex) 3.53 1 3.53 3.18 0.09
Ex*Br 0.03 1 0.03 0.03 0.87
Ti*Ir 0.73 4 0.18 3.59 0.01
Ti*Ir*Br 0.41 4 0.10 2.03 0.10
Ti*Ex 0.53 4 0.13 2.25 0.07
Ti*Ex*Br 0.13 4 0.03 0.56 0.69
Ir*Ex 0.56 1 0.56 0.76 0.39
Ir*Ex*Br 5.13 1 5.13 6.96 0.02
Ti*Ir*Ex 0.12 4 0.03 0.59 0.67
Ti*Ir*Ex*Br 0.31 4 0.08 1.46 0.22
Time (Ti) 1.85 4 0.46 16.09 <0.0001
Bromeliad (Br)(**) 1.63 1 1.63 3.16 0.09
Time*Bromeliad (Br) 0.09 4 0.02 0.83 0.51
Table 2-5. Continued.
Sum of Degrees of Mean
Source Squares Freedom Square F P-value
Irrigation (Ir) 0.37 1 0.37 1.33 0.27
Irrigation*Bromeliad 1.13 1 1.13 0.47 0.51
Exclosure (Ex) 5.28 1 5.28 13.23 0.003
Ex*Br 0.26 1 0.26 0.65 0.44
Ti*Ir 0.30 4 0.07 1.96 0.11
Ti*Ir*Br 0.14 4 0.03 0.90 0.47
Ti*Ex 1.4 4 0.35 8.53 <0.0001
Ti*Ex*Br 0.08 4 0.02 0.49 0.74
Ir*Ex 0.08 1 0.08 0.45 0.52
Ir*Ex*Br 0.28 1 0.28 1.58 0.23
Ti*Ir*Ex 0.42 4 0.10 2.95 0.03
Ti*Ir*Ex*Br 0.18 4 0.05 1.29 0.28
Time (Ti) 1.09 4 0.27 6.68 <0.0001
Bromeliad (Br)(**) 2.12 1 2.12 7.74 0.02
Time*Bromeliad (Br) 0.39 4 0.09 2.36 0.07
Irrigation (Ir) 0.26 1 0.26 1.40 0.25
Irrigation*Bromeliad 0.71 1 0.71 4.03 0.07
Exclosure (Ex) 0.05 1 0.05 0.21 0.65
Ex*Br 0.18 1 0.18 0.75 0.41
Ti*Ir 0.62 4 0.16 4.65 0.003
Ti*Ir*Br 0.08 4 0.02 0.63 0.64
Ti*Ex 0.09 4 0.02 0.62 0.65
Ti*Ex*Br 0.14 4 0.03 0.98 0.43
Ir*Ex 0.13 1 0.13 0.71 0.42
Ir*Ex*Br 0.01 1 0.01 0.06 0.81
Ti*Ir*Ex 0.11 4 0.02 0.45 0.77
Ti*Ir*Ex*Br 0.06 4 0.02 0.45 0.51
Table 2-6. Mean relative height growth rates (+ 1SE) of seedlings of commercial tree species in
response to bromeliad cover removal evaluated in 4 times. Significant differences
between treatments were evaluated using t-tests at 95% of confidence level with
sequential Bonferroni corrections (PB = 0.0125).
Table 2-7. Mean relative height growth rates (+ 1SE) of seedlings of commercial tree species in
irrigated and droughted plots evaluated 4 times. Significant differences between
treatments were determined using t-test at 95% of confidence level with sequential
Bonferroni corrections (PB = 0.0125).
N Mean (SE)
25 -0.006 (0.021)
21 0.044 (0.018)
19 -0.024 (0.012)
15 0.025 (0.008)
7 -0.006 (0.066)
7 0.152 (0.072)
7 0.004 (0.017)
6 0.024 (0.008)
13 0.074 (0.026)
11 0.086 (0.033)
14 -0.029 (0.009)
6 -0.013 (0.009)
35 0.039 (0.011)
37 0.040 (0.013)
31 0.001 (0.008)
18 0.008 (0.006)
9 0.164 (0.092)
15 0.020 (0.023)
12 -0.001 (0.011)
10 0.107 (0.052)
10 0.085 (0.036)
8 0.025 (0.021)
3 -0.036 (0.029)
N Mean (SE)
4 0.015 (0.035)
11 0.084 (0.035)
10 -0.002 (0.005)
8 0.024 (0.013)
Table 2-8. Mean relative height growth rates (+ 1SE) of seedlings of commercial tree species in
response to mammal exclosure evaluated 4 times during 2006. Significant differences
between treatments were evaluated using t-test at 95% of confidence level with
sequential Bonferroni corrections (PB
-20 -,Y C
Fiue21 otl aifl A n olmosuetninmaurdb aemr@si
sesr (B.As hw i stewte de ote riae ol lt.Vria
lie niat 1sadad( =)
80 -I I ... .~. .. ... .. ... ...
0-10 10-20 20-30 30-40 40-50 50-60 > 60
Canopy Openness (%)
Figure 2-2. Canopy openness in an unlogged control plot, an area subjected to normal timber
harvesting (4.7 m3/ha harvested), and an area subj ected to intensive harvesting (8.2
m3/ha) 8 months (left-hand bar) and 42 months (right hand bars) after logging.
Canopy openness measures were made with a spherical densitometer at the end of the
rainy season at 1 m above ground at 144 equally spaced points in each 10 ha plots.
Bromeliads No Bromeliads
Block 1 R R2 R R2W-XNW-NE
Block 2 R1 R2 R1 Ii R2
Bromeliads No Bo
Block 3 R1 R2 R1 R2 WA = WATER
NW = NO WATER
NE = NO EXCLOSURE
SP1 SP2 SP3 SP4 SP5 SP6
.. SP =Species
No Bromeliads Bromeliads
Blc 0 R1 R2 R1 i R2
Figure 2-3. Design of the experiment on the effects of ground bromeliads, irrigation, extended
drought, and seed and seedling predator exclosures on seedling establishment.
2003 2004 2005
(A) AII Species _
0- Normal Harvesting
7- Intensive Management
(B) Without Acosmium
Figure 2-4. Seedling densities of 1 1 timber species (A) and 10 species without Acosmium
cardena~sii (B), the most dominant species, in a control plot, a plot subj ected to
normal timber harvesting, and a plot subj ected to intensive timber stand management.
The forestry treatments were carried out in 2001, 19 mo prior to the first census. Data
are from 4 m2 Subplots distributed regularly in a grid with 25 m spacing through each
10 ha treatment plot. Vertical lines indicate standard errors. Different letters indicate
different significance between treatments using pairwise LSD tests at 95%
r 30 Normal Harvesting
V) E Intensive Harvesting
Figure 2-5. Temporal changes in seedling recruitment (#/m2/y) and mortality for commercial
tree species in a Chiquitano dry forest in Bolivia.
3 -1 -0- No Bromeliad
4 C ` Exlsr
(C) -0- Exclosure
Sep5 Nov5 Jan6 Mar6 May6 Dec6
Figure 2-6. Seedling density over time in response to (A) bromeliad cover, (B) irrigation or
drought, and (C) mammalian seed predators. Arrow indicates the time when I stopped
the irrigation and shielding plants from rainfall. Vertical lines show standard errors of
the means (N=40).
Figure 2-7. Mean densities (+ 1SE) of seedlings of commercial timber tree species in 2 m2 plOts
(N=40) with bromeliads (filled dots) and without bromeliads (open dots). The
number of seeds sown in each 1 m2 plOt is indicated after each species name.
Amburana cearensis (7)
Aspidosperma rigid )
Dec6 Sep5 Nov5Jan6 Mar6 May6
-0 No Bromeliad
I I I I I
Ceiba samauma (8)
I I I I I
Hymenaea courbaril (3)
Sep5 Nov5 Jan6Mar6 May6
Figure 2-8. Mean of seedling densities (+ 1SE) in irrigated and droughted experimental plots.
Note differences in y-axis scales. Arrows indicate the time when I stopped the
irrigation and shielding plants from rainfall.
-0 No irrigation
Dec6 Sep5 Nov5Jan6 Mar6 May6
I I I I I
Sep5 Nov5Jan6 Mar6 May6
0.8 -1 T0.8-
0.6 -1 O 0.6-
0.4 -1 / 0.4-
0.2 -1 ~0.2-
.0.0 -.0 e No exclosure
Ceiba samauma Copaifera chodatiana
0.8 -1 0.8-
0.0 0.0 -
Hymenaea courbaril Pterogyne nitens
0.8 -1 0.8-
0.6 -1 10.6-
0.2 -( T 0.2 i -- -
0.0 -0.0 -
Sep5 Nov5 Jan6Mar6 May6 Dec6 Sep5 Nov5 Jan6 Mar6 May6 Dec6
Figure 2-9. Mean seedling densities (+1 SE) in control plots (closed dots) and in plots from
which mammals were excluded (open dots).
Figure 2-10. Mean relative growth height growth rates (+ 1SE) of seedlings of commercial tree
species in response to three experimental treatments: (A) irrigation or drought; (B)
mammalian seed predator exclusion; and, (C) bromeliad cover. Numbers indicate the
number of seedlings used for each evaluation; numbers in brackets are the seedlings
in irrigated, exclosure, or bromeliad-cover plots while numbers in parenthesis are for
droughted, non-exclosure, and no-bromeliad plots. There were no significant
differences between treatments for each time calculated with t-test at the 95%
confidence level with sequential Bonferroni corrections (repeated-measures ANOVA
was not suitable because sample sizes declined over the study period due to seedling
   
-(45) (87) (84) (54)
 [1 20]  
-(52) (68) (64) (41)
-*- No Exclosure
   
-(100) (130) (127) (77)
-*- No Bromeliad
CONTRIBUTION OF ROOT AND STUMP SPROUTS TO NATURAL REGENERATION IN
A LOGGED TROPICAL DRY FOREST IN BOLIVIA
Securing sufficient natural regeneration of commercial tree species after logging is critical
for sustainable forest management. Most studies of tropical forest regeneration focus on tree
recruitment from seeds and, consequently, regeneration is often viewed as depending on seed
production, seed dispersal, seed viability, and the environmental requirements for seed
germination and seedling establishment (Holl 1999, Dalling and Hubbell 2002, De Steven and
Wright 2002). In tropical dry forests, many tree species produce abundant and well-dispersed
seeds with high viability, but due to seed predation (Janzen 1971), water stress (Gerhardt 1994),
and a multitude of other factors, successful recruitment from seed is rare for many species in
these forests (Mostacedo and Fredericksen 1999). Furthermore, in forests in general and in dry
forests in particular, tree seedlings that do become established often grow more slowly than
sprouts (Miller and Kauffman 1998, Khurana and Singh 2001).
Given the limitations on seed dispersal and germination, as well as on seedling
establishment and survival, successful dry forest regeneration after logging, severe windstorms,
or fires may depend greatly on contributions from stump and root sprouts (Kruger et al. 1997,
Kammesheidt 1998, Miller and Kauffman 1998, Gould et al. 2002, Homma et al. 2003). The
general capacity of dry forest tree species to sprout may be an adaptive response to a history of
fire (e.g., Bond and Midgley 1995) or other disturbances. Whatever the ultimate cause, in a
variety of seasonal tropical forests, logging reportedly stimulates abundant stump sprouting of
felled and broken trees, and root sprouting from superficial roots damaged by heavy equipment
(Kauffman 1991, Kammesheidt 1998, Miller and Kauffman 1998, Kammesheidt 1999, Bell
2001). In the Chiquitano dry forest in Bolivia, although sprouting has been reported few times
following logging (Fredericksen et al. 2000) and fires (Gould et al. 2002, Kennard and Putz
2005), little is known about the overall contributions of sprout-origin plants to forest recovery.
Whereas interspecific comparisons of sprouting ability are numerous for Mediterranean
ecosystems (Bellingham 2000, Pausas 2001), how this ability varies with light requirements and
other ecological attributes is less clear for tropical dry forest species (but see Paciorek et al. 2000
for resprouting across the spectrum of shade tolerance of trees on Barro Colorado Island).
Sprouting is of interest to forest managers and ecologists because sprouts often have more rapid
grow rates than true seedlings (Daniel et al. 1979, Clark and Hallgren 2003).
The purpose of this study was to examine the contribution of sprouts to the natural
regeneration of a tropical dry forest following logging. More specifically, I (1) characterized
stump and root sprouting features of the commercial canopy tree species. I (2) quantified the
effect of logging on relative abundances and growth rates of stump sprouts, root sprouts, and true
seedlings. I (3) related the species-specific probabilities of stump sprouting as a function of
stump diameter and stump height; and I (4) explored how sprouting varies with the ecological
requirements of canopy tree species.
This study was conducted on the property of INPA Parket (hereafter INPA), a 30,000-ha
tract of privately-owned seasonally dry tropical forest 30 km NE of the town of Concepci6n (160
6' 45"S, 610 42' 47"), 250 km northeast of the city of Santa Cruz de la Sierra, Bolivia (Figure 1-
1). The study area is flat to gently sloping, with an altitude of approximately 380 m, mean annual
temperature of 24.3 OC, and mean annual precipitation of 1 100 mm. During the 5-mo dry season
(May-October), most trees are deciduous and many tree species flower and fruit following rain
events in the mid- to late-dry season. The forest canopy is 20-25 m tall with common species
including Acosmium cardena~sii, Tabebuia impetiginosa, Anadenanthera macrocarpa,
Astronium urundeuva, and Centrolobium microchaete (Pinard et al. 1999); after first mention,
species will be referred to by their generic names (for full names see Table 3-1). Currently, 21
tree species, including those mentioned above, are harvested for timber processed mostly into
During the rainy season, canopy openness, as measured 1 m above the ground with a
spherical densiometer, was 8% and 14% in control and logged areas, respectively. During the dry
season, canopy openness triples because many tree species are deciduous. The understory is
dense, partially due to the abundance of lianas, and typically 30-40% of the ground is covered by
the bromeliad, Pseudananas~ddd~~~ddd~~~dd sagenarius.
Experimental Design and Data Collection
I sampled sprouts from the stumps of harvested trees in three areas that varied in time since
logging. The first area (50 ha) was selectively logged by INPA Parket (- 4 trees/ha and 4 m3/ha,
and 10-12 species harvested) 1 y before I began my study. Here I mapped, marked, and
measured the diameters and heights of the stumps and all stump sprouts of the five most
commonly harvested tress (Anadenanthera, Centrolobium, Copaifera chodatiana, Tabebuia, and
Zeyheria tuberculosa; tree densities reported in Table 2-1) and monitored sprout survival and
height growth for one year. The second study area (40 ha) is located in the 20-ha permanent plots
maintained by the Instituto Boliviano de Investigaci6n Forestal (IBIF) for monitoring forest
dynamics after logging. Two years prior to my study, 4-8 trees/ha (5.3-6.4 m3/ha, 14 species)
were logged from these plots. I monitored sprouting as described above. The third area (30 ha)
was logged (2-3 trees/ha, -3 m3/ha, and 5-7 species harvested) 5 y before my study. I measured
the frequency and height of stump sprouts of six tree species (Anadenanthera, Caesalpinia,
Centrolobium, Copaifera, M~ scleroxylon, and Tabebuia) in this area, using a 7 m telescoping
In 2003 I checked for sprouts on the stumps of trees harvested in 1998 (6 species),
2001(10 species), and 2002 (6 species). The 498 stumps evaluated in the three areas were from
trees > 40 cm DBH that were felled with chainsaws 10-90 cm height above ground. Each
species was placed in one of the following four ecological guilds based on field observations and
the literature (Whitmore 1998, Mostacedo and Fredericksen 1999, Poorter et al. 2006): light-
demanding pioneers have light requirements and are short-lived; long-lived pioneers also are
light-demanding but are longer lived; somewhat shade tolerant species establish in the shade but
only mature under moderate to high light intensities; and, shade-tolerant species can establish
and survive in the shade. I counted all sprouts and measured the heights of the two tallest on each
stump (from the point of origin) as well as the height and diameter of each stump dating from the
2001 and 2002 harvests.
Comparison of different juvenile types in relation to microsites created by logging
In two of IBIF's 20-ha experiment plots, I compared the densities and sizes of seedlings
and sprouts < 2 m tall in the following microsites created by an episode of selective logging that
occurred 1.5 years prior to my study: logging gaps (280-330 m2, N = 16); logging roads (N =
16); log landings (N = 8); primary skid trails (N = 16); and, secondary skid trails (N = 16).
Secondary skid trails were those used to extract a single log, while primary skid trails were those
where skidder had extracted > 2 logs. In each microsite, all plants < 2 m tall of 16 canopy tree
species in 10 x 4 m plots were classified as having developed directly from a germinated seed or
sprouted from a root or stem; determination of plant origin often involved excavation, but was
I used logistic regression to determine the probability of a stump sprouting in relation to its
diameter and height for each the Hyve species that sprouted frequently (Anadenanthera,
Centrolobium, Copaifera, Tabebuia, and Zeyheria). Nagelkerke R-square values were used to
determine the percentage of variance explained by each regression and a Hosmer and Lemeshow
X2 gOodness-of-fit tests was used to determine the significance of each relationship (Field 2000).
To determine whether there are relationships between stump diameter and height
(independent variables) with the number and maximum heights of sprouts (response variables), I
used linear regressions or nonlinear regression analyses based on linear, quadratic, cubic, and
inverse models. For each species, the simplest (i.e., fewest parameters) model with a high R2
value was selected in which each parameter had a reasonable biological explanation.
Analyses of variance (ANOVAs) followed by Tukey's post-hoc comparisons were used to
compare densities of seedlings and root or stem sprouts among logged microsites. Absolute
relative annual height growth rates of stump sprouts were calculated for 10 species based on their
height 2 y after logging, and heights after 1, 2, and 5 y after logging for 6 species. Stump sprout
heights were compared among ecological guilds using ANOVAs and Tukey's post-hoc tests. For
each species and ecological guild, an ANOVA and then Tukey's post-hoc comparison were used
to compare mean growth among origin types (i.e., stump sprout, root sprout, or true seedling).
All analyses were carried out with SPSS 12.0 for Windows.
Stump sprouting was common after logging in the dry forest studied; 27 of the 3 1 species
monitored at least occasionally reprouted from stumps; 62% did so frequently (Table 3-1).
Centrolobium, Zeyheria, and Tabebuia were the most frequent stump sprouters (Figure 3-1).
Among the six commercial tree species monitored, the proportion of stumps with living sprouts
decreased with time since logging (Figure 3-2). Overall, for stumps censused 1, 2, and 5 years
after logging the proportion of stumps with live sprouts was 55%, 43%, and 38%, respectively,
but the rate of stump sprout mortality varied by species. In particular, 80% of the Caesalpinia
stumps and 73% of the Centrolobium stumps had live sprouts 5 y after the trees were felled.
Whereas high proportions of Anadenanthera and Copaifera stumps initially sprouted (23 and
13%, respectively), neither species had living stump sprouts in the plot logged 5 y prior to my
Root sprouting was also common after logging in the tropical dry forest of INPA. Of the
31 tree species monitored (Table 3-1), 16 sprouted from lateral roots, 7 species at high
frequencies. Acosmium cardena~sii, Centrolobium, and Casearia gossypiosperma were the most
frequent root sprouters.
Most of the 27 species that frequently sprouted from roots or stumps were shade tolerant
(9) or at least partially shade-tolerant (8). Light-demanding pioneer species sprouted
infrequently, if at all (e.g., Astronium, Piptadenia~~ttt~~~~ttt~~~ viridifolia, Acacia bonariensis, and Schinopsis
brasiliensiss. The most frequent sprouters were the long-lived pioneer species, Centrolobium and
Tabebuia, and the partial shade-tolerant species, Zeyheria (Table 3-1).
Caesalpinia pluviosa had the higher (20.014.4) number of sprouts per stump, followed by
Centrolobium (15.411.47) and Zeyheria (14.211.5). The other four of the seven species
monitored had < 5 sprouts/stump (SE=0.5), with the absolute lowest numbers observed in
Copaifera and Anadenanthera (Figure 3-3)
Considering all sprouted stumps, there was a significant negative linear relationship
between the number of sprouts per stump and stump diameter (Table 3-2). In contrast, when
species were considered separately, the only species that showed a significant (quadratic)
relationship between number of sprouts and stump diameter was Copaifera, and that relationship
was positive (i.e., opposite from the overall trend). For all species considered together, there was
a negative linear relationship between the number of sprouts and stump height, but the
relationship varied among species. The number of Caesalpinia sprouts increased with stump
height whereas in Centrolobium and Tabebuia, the relationship was negative (Table 3-2).
Sprout height growth, based on measures of the tallest stump sprouts 1-5 y after the trees
were cut, generally decreased with stump diameter, but species varied in this relationship. Only 3
of the 7 species for which I have sufficient data showed significant trends: Caesalpinia showed a
positive cubic relationship; in Machaerium the relationship was inverse positive; and, Copaifera
had a quadratic and positive relationship (Table 3-2). Sprout height growth rates decreased with
stump height when all of the species were considered together. In contrast, at the species level,
only Caesalpinia and Machaerium showed significant relationships between sprout growth rates
and stump height, but in the former the relationship was negative and quadratic and the latter,
positive and inverse (Table 3-2).
The probability of stump sprouting as related to stump diameter varied among species
(Figure 3-3). In Copaifera, Anadenanthera, Tabebuia, and Centrolobium, the proportions of
sprouted stumps were approximately 0.17, 0.24, 0.57, and 0.97, respectively, and did not vary
with stump diameter. In Zeyheria, sprouting reached 98% of the stumps 38-40 cm diameter but
decreased to only 40% among stumps 90 cm in diameter (Figure 3-4).
The probability of sprouting in relation to stump height also varied among species Figure
3-5). The proportions of sprouted stumps of Copaifera, Anandenanthera, and Centrolobium did
not vary with stump diameter. In contrast, a Tabebuia stump 10 cm tall was almost certain to
sprout (0.98) whereas this probability declined to 0.13 for a 72 cm tall stump. In Zeyheria, the
probability of sprouting was high (0.91) and did not vary with stump height (Figure 3-5).
Juvenile Types and the Effects of Logging
In the plots censused 1.5 years after logging, 45% of juveniles < 2 m tall of canopy tree
species were root and stem sprouts, not true seedlings (Figure 3-6). At the species level there
was great variation in the proportions of true seedlings (Figure 3-7). All 15 species evaluated
were represented by some sprouts and sprouted at least occasionally from roots whereas only 9
species were represented by stem sprouts. Three species sprouted predominantly from roots
whereas stem sprouting was the predominant mode of regeneration in only one species. Light-
demanding species tended to regenerate more from seeds and root sprouts than from stem sprouts
(F=12. 10, P=<0.0001), while partially shade tolerant and shade tolerant regenerated more from
seeds (F=4.46, P=0.01; F=8.01, P=0.0004; respectively).
Densities of plants < 2 m tall of canopy tree species did not vary among the logging
microsites (F = 1.37, P = 0.24), but microsites differed in the relative contributions of true
seedlings and sprouts (Table 3-3). True seedlings were twice as abundant as root and stem
sprouts combined in logging gaps (F = 9.91, P = 0.0001). In contrast, there was no difference in
plant density by origin in logging roads (F = 0.38, P = 0.68). Densities of plants from root
sprouts and true seedlings were similar in log landings (F = 1.95, P = 0.18). On primary skid
trails most plants <2 m tall were true seedlings, with fewer root sprouts, and almost no stem
sprouts (F = 13.57, P < 0.0001). On secondary skid trails, true seedlings were much more
common than plants of either sprout type (F = 4.6, P = 0.01).
Growth of Stump Sprouts
Based on measures of the tallest sprout per stump, the growth rates of stump sprouts varied
among species by more than an order of magnitude (Figure 3-8). Anadenanthera (197 cm/y),
Centrolobium (195 cm/y) and Zeyheria (185 cm/y) had the highest growth rates, while
Aspidosperma (3.5 cm/y) and Copaifera (25 cm/y) had the lowest. Stump sprouts of light-
demanding pioneer species grew faster than those of shade-tolerant species, but there was a great
deal of within species variation, especially in the growth rates of the latter (Table 3-4). Stump
sprouts of long-lived pioneer species grew at about the same rates as shade tolerant species.
Among individuals of canopy tree species < 2 m tall, root and stem sprouts both grew
faster than seedlings and in 5 of 12 species monitored, root sprouts grew faster than stem sprouts
(Table 3-5). Centrolobium and Chorisia speciosa had the highest root sprout growth rates. I
found no Anadenanthera, Aspidosperma, Casearia arborea, M. scleroxylon, Phyllostylon
rhamnoides, or Piptadenia~~ttt~~~~ttt~~~ stem sprouts (Table 3-5). Among the root sprouters, light-demanding
and long-lived pioneer species grew faster than partially shade tolerant and shade tolerant
species. Among the stem sprouters and true seedlings, shade-tolerant species grew slower than
belonging to other light-syndrome classes (Table 3-6).
Stump sprout heights varied over time and among the six species censused 1, 2, and 5 y
after logging (Figure 3-9). Apparent absolute growth rates (cm/y) increased through the second
year and then height increments stopped except in Centrolobium, which continued to grow at a
rapid rate through the fifth year. Anadenanthera sprouts were apparently growing rapidly
through the second year, but I could find no live stumps in the plot logged 5 y prior to my
Of the 3 1 canopy tree species studied in a dry tropical forest in Bolivia, 27 (87%) have
some capacity to sprout from either roots or stumps. Sprouting is apparently characteristic of
many tropical dry forest tree species and helps them persist in an environment where stress is
severe and disturbances are frequent (Bellingham 2000, Bond and Midgley 2001). As observed
in the USA (Jones and Raynal 1988) and Mexico (Dickinson 1998), root sprouting was promoted
by logging damage to roots in my study species. The high proportion of species that sprouted
from broken and cut stems in INPA may be related to the high frequency with which stems lose
their terminal buds due to herbivore browsing during the dry season when other browse is scarce
as well as to the direct effects of drought stress (Bossard and Rejmanek 1994, Del Tredici 2001,
Natural Regeneration and Shade Tolerance: True Seedlings vs. Sprouts
Scarcity of natural regeneration from seeds is common for most tree species in tropical dry
forest in Bolivia (Mostacedo and Fredericksen 1999). The main reasons for this scarcity appear
to be high seed predation, low seed viability, and high seedling mortality during the dry season.
Sprouting from broken and cut stems, along with root sprouting, appears to be a very important
regeneration mechanisms in tropical dry forests in Brazil (Castellani and Stubblebine 1993),
Jamaica (Bellingham et al. 1994), and Venezuela and Paraguay (Kammesheidt 1999), including
the forest I studied in Bolivia where 45% of the regeneration of canopy trees originated from root
or stem sprouts.
Sprouting is a common mode of tree regeneration in forests around the world. For example
in oak forests in the USA (Clark and Hallgren 2003, Nyland et al. 2006) and in boreal forest in
Russia (Homma et al. 2003), trees reportedly regenerate mainly from sprouts. Sprouting seems to
represent the predominant mode of regeneration in forests frequently subj ected to logging, wind
damage, and fire (Bond and Midgley 2001).
Natural regeneration by sprouting from lateral roots was common in some commercial
species in my study site. In particular, Centrolobium, Tabebuia, Aspidosperma, and M~
scleroxylon regenerated mostly from root sprouts. Centrolobium was previously reported as a
root sprouting species (Fredericksen et al. 2000), but the importance of this mode of regeneration
in the other species has apparently been overlooked. It remains to be seen whether root sprouts
mature into sound trees, and there are reasons to suspect that they will not. First of all, given that
most new stems emerge from damaged stems or roots, sprouts of all sorts seem particularly
prone to butt and root rots. Second, I observed that several Centrolobium root sprouts that were
5-8 cm DBH 5 y after sprouting still had not developed their own root systems. Those that I
excavated emerged from large diameter roots running about 5 cm below the soil surface but had
developed almost no roots of their own and were thus mechanically unstable when pushed
perpendicular to the orientation of the source root. Given the general importance of root
sprouting after fires, logging, and other severe disturbances, such as found in tropical dry forest
in Paraguay and moist semi-deciduous forest in Venezuela (Kammesheidt 1999), root sprout
longevity is an issue that deserves more attention from researchers.
Partially shade-tolerant and shade-tolerant species were more likely to sprout than light-
demanding species. Most of the partially shade-tolerant species in INPA sprouted from either
roots or stems; similar findings were reported for a moist but seasonal tropical forest in Panama
(Paciorek et al. 2000). In contrast, in a moist tropical forest but after slash-and-burn agriculture
of eastern Paraguay, light-demanding species contributed more sprouts than shade-tolerant
species (Kammesheidt 1998). Although some light-demanding species in INPA did not sprout
(13%), others stump sprouted frequently, such as species in the Bombacaceae and
Flacourtiaceae. Furthermore, the light-demanding pioneer species that did sprout grew faster
than sprouts from other ecological groups.
Allometric Relationships with Stump Sprouting
The weak and inconsistent trends in the relationship between either stump diameter or
stump height and the growth rates or number of sprouts per stump means that I have little basis
on which to make firm recommendations for sprout management. Consistent patterns in sprout
responses are also not apparent in the literature. Whereas several studies mention that in the first
years after tree cutting there is a positive relationship between stump diameter and sprout growth
rates (Jobidon 1997), other studies report the opposite (Trani et al. 2005), and generally the
relationship does not remain significant after a few years. These results suggest that factors other
than stump size controls sprouting (McConnaughay et al. 1996, Masri et al. 1998).
The probability of sprouting varied substantially among species but I observed no effect of
stump diameter on the probability of sprouting in four of the five species studied. For example,
Centrolobium had the highest probability of stump sprouting (97%), while Copaifera had the
lowest (17%). In contrast, Zeyheria showed a decreasing probability of stump sprouting with
increasing stump diameter, a pattern also observed in a wetter but still seasonal tropical lowland
forest in Panama (Putz and Brokaw 1989) and in an oak forest in southern Indiana (Weigel and
Peng 2002). The probability of stump sprouting did not vary either with stump height except in
Tabebuia, in which the probability decreased with stump height. These results suggest that
harvesting trees of any tree size will promote same probability of sprouting, except for Zeyheria
in which it is better to cut smaller trees and in Tabebuia in which low stumps are preferred if
sprouting is to be encouraged.
Growth of Stem and Root Sprouts Compared with True Seedlings
One advantage of natural regeneration via sprouting is that sprouts typically grow more
rapidly than true seedlings, at least initially (Gould et al. 2002, Kennard et al. 2002), which was
confirmed by this study. I also observed that small plants of sprout origin typically appeared less
affected by drought than true seedlings (Personal Observation). Nevertheless, in my study forest
as well as in Australia (Enright and Goldblum 1999) and South Africa (Kruger et al. 1997), true
seedlings of some species grew just as fast as sprouts. In several species, especially light-
demanding pioneers (e.g., Cordia, Casearia), height growth rates were similar between true
seedlings and root or stem sprouts.
Due to the high costs and frequent failures of seed and seedling planting, natural
regeneration is critical for the sustainable management of tropical dry forest tree species in
Bolivia. Given that so many tree species sprout prolifically from stumps of all sizes or from
lateral roots, especially after mechanical damage, sprouts need to be considered as a source of
regeneration. The abundance of sprouts and their typically rapid growth rates, when compared
with those of true seedlings, adds to the potential value of sprouts for forest management. In
forests not designated for timber stand management, sprouts deserve at least as much attention
from researchers as seeds and true seedlings. That said, future studies should consider the long-
term fates of sprouts. The observation in this study that the stump sprouts of most species
essentially stopped growing after 2 y needs to be verified and otherwise explored, as do the
factors that cause high rates of mortality of the sprouted stumps of some species. In the case of
root sprouts, which were also abundant in my study area, long-term monitoring is needed to
determine whether they ever grow up to be sound, canopy trees.
Table 3-1. Frequency of root and stem sprouting and shade tolerance of commercial and non-commercial canopy tree species in a
tropical dry forest in Bolivia. Shade tolerance is base on Pinard et al. (1999), and Mostacedo and Fredericksen (1999):
L=Light-demanding pioneer, LL=Long-lived pioneer, PS=Partially shade tolerant, ST=Shade tolerant.
Resprout Type/ Frequency
Commercial Timber Sp~ecies
Amburana cearensis (Allemlio) A.C. Sm.
Anadenanthera macrocarpa (Benth.) Brenan
Aspidosperma rigidum Rusby
Astronium urundeuva (Allemcro) Engl.
Caesalpinia pluviosa DC.
Can'niana ianeirensis R. Knuth
Cedrela fissilis Vell.
Centrolobium microchaete (Mart. ex Benth.) Lima ex G. P. Lewis
Copaifera chodatiana Hassl.
Cordia alliodora (Ruiz & Pay.) Oken
Hymenaea courbaril L.
M2/achaerium scleroxylon Tul.
Phyllostylon rhamnoides (J. Poiss.) Taub.
SPlartymiscium ulei Harms
Schinopsis brasiliensis Engl.
Sweetia fruticosa Spreng.
Tabebuia impetiginosa (Mart. ex DC.) Standl.
Tabebuia \r~rarrit;,idi (Vahl) G. Nicholson
Zeyhena tuberculosa (Vell.) Bureau
Acacia bonariensis Gillies ex Hook. & Arn.
Acosmium cardenasii H.S. Irwin & Arroyo
Aspidosperma cylindrocarpon Milll. Arg.
Capparis pnsca J.F. Macbr.
Caseana gossypiosperma Briq.
Ceiba samauma (Mart.) K. Schum.
Chon'sia speciosa A. St.-Hil.
Enotheca roseorum (Cuatrec.) A. Robyns
Gallesia integrifolia (Spreng.) Harms
M2/achaerium atr urit;,idimu Vogel
Piptadenia 1 n ,, 1,r /.., (Kunth) Benth.
Spondias mombin L.
Table 3-2. Summary of the best models and their significance from the regression analyses between stump diameter or stump
volume (independent variables) and the heights of stump and numbers of sprouts per stump (dependent variables). Tests
were made for species that were sampled with > 8 individuals at alpha = 0.05. Models: L=Linear, Q=Quadratic, C=Cubic,
I=Inverse. Signs mean positive (+) or negative (-) relationship between two variables.
Scintfi Nme# of sprouts Height of sprouts
Model R2 DF F P Model R2 F P
4nadenanthera inacrocarpa L 0.015 36 0.56 0.461 L 0.034 1.27 0.2680
Caesalpinia pluviosa Q(+) 0.267 18 3.27 0.061 C(+) 0.406 6.16 0.0090
Centrolobiwn inicrochaete L 0.005 90 0.41 0.522 L 0.012 1.09 0.3000
Copaifera chodatiana Q(-) 0.390 21 6.71 0.006 Q(-) 0.536 12.11 <0.0001
Machaerium scleroxyvon Q 0.365 5 1.44 0.321 I(+) 0.630 10.24 0.0190
Tabebuia ,,i, rs,.so L 0.026 54 1.45 0.234 L 0.010 0.54 0.4660
Zevheria tuberculosa L 0.001 62 0.05 0.819 I 0.036 2.33 0.1320
All species L(-) 0.020 301 5.61 0.020 L(-) 0.040 3.95 0.0400
4nadenanthera inacrocarpa L 0.01 36 0.49 0.49 L 0.001 0.03 0.87
Caesalpinia pluviosa L(+) 0.4 19 12.90 0.002 Q(-) 0.32 4.24 0.03
Centrolobiwn inicrochaete I(-) 0.36 90 49.96 <0.0001 I 0.02 2.06 0.16
Copaifera chodatiana L 0.03 22 0.66 0.42 L 0.02 0.42 0.52
Machaerium scleroxyvon L 0.006 6 0.04 0.85 I(+) 0.62 10.03 0.02
Tabebuia ,,i, rs,.so I(-) 0.34 59 29.82 <0.0001 L 0.02 0.90 0.35
Zevheria tuberculosa I 0.03 63 2.06 0.16 L 0.008 0.53 0.47
All Species L(-) 0.03 311 8.91 0.003 I(-) 0.16 56.6 <0.0001
Table 3-3. Mean (+ 1SE) densities of true seedlings, stem sprouts, and root sprouts in 10 x 4 m plots in microsites created during
selective logging. Different letters indicate differences between microsites in the densities of plants of different origins
using Tukey post hoc comparisons with 95% of confidence.
Primary Skid Trail
Secondary Skid Trail
Mean of Square
Table 3-4. Mean (+ 1SE) of stump sprout height growth rates (cm/year) by ecological groups.
Different letters indicate significant differences between ecological groups using
Tukey post hoc comparisons with 95% confidence.
Partially Shade Tolerant
47 194.7 (17.4)a
6 84.8 (17.8)"
24 159.5 (12.3)b
4 14.5 (8.8)"
Table 3-5. Means (+ 1SE) of stem heights by species that sprouted from stems or roots compared to the heights of seedlings. Different
letters indicate significant differences in plant origins within species using Tukey post hoc comparisons at the 95%
Table 3-6. Mean (+ 1SE) heights of trees I 2 m tall that were root sprouts, stem sprouts, or seedlings grouped by ecological guild.
Different letters indicate significant differences between ecological groups using Tukey post hoc comparisons at the 95%
CEMI CAPL MASC ZETU TAIM ANMA COCH
Figure 3-1. Proportions of stumps of commercial timber species that resprouted (number of
stumps noted in parenthesis). CEMI = Centrolobium microchaete, CAPL =
Caesalpinia pluviosa, MASC = Machaerium scleroxylon, TAIM = Tabebuia
impetiginosa, ANMA = Anadenanthera macrocarpa, COCH = Copaifera chodatiana.
Centrolobium Caesalpinia M. scleroxylon Tabebuia Anadenanthera Copaifera
Figure 3-2. The proportions of stumps with live sprouts over time since logging. For complete
species names see Table 3-1.
CAPL CEMI ZETU TAIM MASC ANMA COCH
Figure 3-3. Mean (+ 1SE) numbers of sprouts/stump for the most frequent sprouting species.
CEMI = Centrolobium microchaete, CAPL = Caesalpinia pluviosa, MASC =
Machaerium scleroxylon, TAIM = Tabebuia impetiginosa, ANMA = Anadenanthera
macrocarpa, COCH = Copaifera chodatiana.