Citation
Fruit Fall Patterns in an Experimentally Burned Amazonian Forest

Material Information

Title:
Fruit Fall Patterns in an Experimentally Burned Amazonian Forest
Creator:
Carvalho,Eric O
Place of Publication:
[Gainesville, Fla.]
Florida
Publisher:
University of Florida
Publication Date:
Language:
english
Physical Description:
1 online resource (87 p.)

Thesis/Dissertation Information

Degree:
Master's ( M.S.)
Degree Grantor:
University of Florida
Degree Disciplines:
Forest Resources and Conservation
Committee Chair:
Kobziar, Leda Nikola
Committee Members:
Staudhammer, Christina Lynn
Putz, Francis E
Graduation Date:
8/6/2011

Subjects

Subjects / Keywords:
Average linear density ( jstor )
Density ( jstor )
Fires ( jstor )
Forest fires ( jstor )
Forests ( jstor )
Fruits ( jstor )
Lianas ( jstor )
Species ( jstor )
Trees ( jstor )
Tropical forests ( jstor )
Forest Resources and Conservation -- Dissertations, Academic -- UF
amazon -- ecology -- experimental -- fire -- forest -- fruitfall -- transitional -- wildfire
Genre:
bibliography ( marcgt )
theses ( marcgt )
government publication (state, provincial, terriorial, dependent) ( marcgt )
born-digital ( sobekcm )
Electronic Thesis or Dissertation
Forest Resources and Conservation thesis, M.S.

Notes

Abstract:
Wildland fires in Amazonian forests have increased over the last 20 years, being linked to droughts and coupled with anthropogenic land use and ignitions. The aftermath of forest fires is generally a reduction in species richness, live stem density, and tree biomass, as well as increased litter input, canopy openness, and tree mortality rates. Fire-related tree mortality not only influences the future flammability of burned forests but can affect several ecosystem processes that are more elusive to track than the immediate effects of burning. Among these processes there are those related to tree reproduction and forest recovery. The present study addresses fruit-fall patterns in an experimentally burned tropical forest in an attempt to shed some light on the effects of fire on these processes. This study was undertaken within a large scale fire experiment in an Amazonian Transitional forest. The experiment consisted of one experimental unit divided into three 50 ha plots. Treatments consisted of a plot burned two times in five years (B2), a plot burned four times in five years (B4), and a control (never burned B0). Fruit-fall data was gathered in a series of traps (252) systematically placed within the plots. Traps were sampled twice monthly but for the analysis data was scaled to yearly periods. Linear mixed effects models were used for data analysis. In addition to evaluating the sole effect of fire, fruit-fall patterns were also linked to tree and liana density, basal area, diversity, and distance from the forest-agriculture edge. Fruit-fall mass was higher in B4 than in the control at longer distances from the forest edge. Plot B2 and B0 did not differ in this regards. Species richness in fruit-fall was higher in B2 at higher tree and liana densities. Plots B4 and B0 did not differ from the control or from the two times burned plot in this regards. Both results are in contrast with our initial expectation that fruit-fall mass and species richness would decline with the fire treatments. ( en )
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Thesis:
Thesis (M.S.)--University of Florida, 2011.
Local:
Adviser: Kobziar, Leda Nikola.
Statement of Responsibility:
by Eric O Carvalho.

Record Information

Source Institution:
UFRGP
Rights Management:
Copyright Carvalho,Eric O. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Resource Identifier:
779320797 ( OCLC )
Classification:
LD1780 2011 ( lcc )

Downloads

This item has the following downloads:


Full Text

PAGE 1

1 FRUIT FALL PATTERNS IN AN EXPERIMENTALLY BURNED AMAZONIAN FOREST By ERIC OLIVEIRA CARVALHO A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2011

PAGE 2

2 2011 Eric Oliveira Carvalho

PAGE 3

3 To Ant nio Galdino Oliveira and Noan Maia Carvalho

PAGE 4

4 ACKNOWLEDGMENTS I would like to thank first and foremost my parents Cecilia and Wilton Carvalho, for their un yielding and un conditional support and love through all my endeavors. T hanks to my siblings Alex and Livia for their love, friendship and support they give their older brother I thank Noan Carvalho who has given me inspiration, friendship, and love, al ways keeping me on my toes, and pushing me to become a better human being. I thank Paulo Brando, without whom this thesis would not have been possible Thanks to Leda Kobziar, my advisor, for giving me the opportunities to pursue research on themes that I believed relevant to my interests. I thank J ack Putz, for crucial mentoring and numerous insightful suggestions that made it possible to push this work forward. Thanks to Christie Staudhammer for her advice on various statistical analy ses in this project Many thanks to the IPAM field crew of the Tanguru R anch whose hard work makes possible works such as the present thesis. Thanks to More na Maia for sharing with me some of the m ost special moments in my life and for invariably making life a difficult chal lenge I thank Leonardo Pacheco, Pedro Constantino, Ana Carolina Crisostomo, Ane Alencar, Vivian Zeidemann, Anthony Jepson, Lianne Jepson, Lais Guerra, Jesse Kreye, Marina Londres, Wendy F r ancesconi, Melanie Medeiros, Mathew Graham, and Natalie Richardson for their support and friendship. Thanks to Ana Tereza Vital. I also would like to thank all the people who to a greater or lesser extent pa rticipated on our Brazilian music project which has filled my life with joy and eased the difficulties along the way.

PAGE 5

5 the people with whom I grew up and to whom I always return for strong friendship during good times and Valeu Galera do Gantois Thanks to Abel Concei o from the Universidade Estadual de Feira de Santana in Bahia, who initiated a much needed research program o n the fire ec ology of the Chapada Diamantina and support ed my initial research projects in that region.

PAGE 6

6 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 8 LIST OF FIGURES ................................ ................................ ................................ ........ 10 LIST OF ABBREVIATIONS ................................ ................................ ........................... 12 ABSTRACT ................................ ................................ ................................ ................... 13 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 15 2 MATERIALS AND METHODS ................................ ................................ ................ 20 Study Site and Design ................................ ................................ ............................ 20 Experimental Fires ................................ ................................ ................................ .. 21 Floristic Survey ................................ ................................ ................................ ....... 21 Fruit fall ................................ ................................ ................................ ................... 21 Forest Structure and Diversity Analysis ................................ ................................ .. 22 Fruit fall Data Analysis ................................ ................................ ............................ 22 Community Fruit fall Mass Analysis ................................ ................................ 24 Number of Species Reproducing ................................ ................................ ...... 25 3 FI RE AND EDGE EFFECTS ON FOREST STRUCTURE ................................ ...... 29 Tree and Liana Density, Basal Area, and Diversity in Trap centered Plots ............. 29 Distance From the Edge Effects on Tree and Liana Density, Basal Area, and Diversity ................................ ................................ ................................ ............... 30 4 FRUIT FALL MASS RESULTS ................................ ................................ ............... 41 Community wide Result s ................................ ................................ ........................ 41 Fruit fall Mass Patterns as Revealed by Qualitative Analysis ................................ 4 1 Relationship between Density, BA, and Fruit fall Mass ................................ .... 42 Relationship b etween Distance f rom the Edge and Fruit fall Mass .................. 42 Statistical Models of Fruit fall Mass Patterns ................................ .......................... 42 5 NUMBER OF SPECIES REPRODUCING ................................ .............................. 52 Number of Species Reproducing Description and Qualitative Analysis ............... 52 Relationship between Number of Species and Distance, Density, BA, and Diversity ................................ ................................ ................................ ............... 52

PAGE 7

7 Quantitative Analysis of Number of Species in Fruit fall ................................ ......... 53 6 DISCUSSION ................................ ................................ ................................ ......... 61 Fire and Edge Effects on Forest Structure ................................ .............................. 61 Fire Treatments Did Af fect Fruit fall Mass but at Long Distance from the Edge ..... 62 Greater Influx of Fruits into Burned Plots ................................ ......................... 63 Increased Availability of Resources in Burned Forests Leads to Increased Reproductive Output Survivors ................................ ................................ ..... 64 Fire and Edge Did Interact to Influence Fruit fall ................................ .............. 67 Fire Did Affect Fruit fall Species Richness ................................ .............................. 68 Implications for Regeneration of Burned Tropical Forests ................................ ...... 69 7 CONCLUSION ................................ ................................ ................................ ........ 71 APPENDIX: VARIABLE CORRELATIONS, STATISTICAL MODELS, MODEL SELECTION, AND DIAGNOSTICS ................................ ................................ ........ 72 LIST OF REFERENCES ................................ ................................ ............................... 83 BIOGRAPHICAL SKETCH ................................ ................................ ............................ 87

PAGE 8

8 LIST OF TABLES Table page 3 1 ation between tree and liana density in trap centered plots and distance from the forest agriculture edge (m) grouped by treatment and year. ................................ ................................ ... 36 3 2 ation between tree and liana basal area (m 2 ) in trap centered plots and distance from the forest agriculture edge (m) grouped by treatment and year. ................................ ................................ ... 38 3 3 the relation between tree and liana diversity in trap centered plots and distance from the forest agriculture edge (m) grouped by treatment and year. ................................ ................................ ... 40 4 1 Summary of fruit fall mass (g) gro uped by year and treatment. Number in parenthesis following mean is the standard error. ................................ .............. 45 4 2 Number of traps where fruits were found (in parenthesis percent total traps in treatment) and the area sampled (m 2 each trap is 0.5 m 2 ). ............................... 45 4 3 Total fruit fall mass per year and treatment of the 10 most productive species in terms of mass (g) and percentage of the total production per t reatment and year. ................................ ................................ ................................ ................... 45 4 4 Wald test statistics for terms in Global model of fruit fall mass. .......................... 49 4 5 Wald test statistics for sel ected fruit fall mass model. ................................ ........ 49 4 6 Summary of fixed effects in fruit fall mass model. ................................ .............. 49 5 1 Summary of fruit mass (g) by dispe rsal mechanism and percent of total mass by dispersal mechanism per treatment and year. ................................ ............... 55 5 2 Wald tests for terms in model of lowest AIC for number of species in fruit fall. .. 58 5 3 Wald test for terms in the final, most parsimonious, statistical model of number of species in fruit fall. ................................ ................................ ............. 58 5 4 Summary for the model s elected to further analysis of the number of species in fruit fall. ................................ ................................ ................................ ........... 58 A 1 Correlation between fruit fall and tree and liana density by year and tr eatment ................................ ................................ ................................ ........... 75 A 2 Correlation between fruit fall and tree and liana basal area (m 2 ) by year and tr eatment ................................ ................................ ................................ ........... 75

PAGE 9

9 A 3 Correlation between fruit fall and distance from forest agric ulture edge (m ) by year and treatment ................................ ................................ ............................ 75 A 4 Correlation between fruit fall species richness and distance from forest edge by year and treatme nt ................................ ................................ ........................ 80 A 5 Correlation between fruit fall species richness and tree and liana density by year and tr eatment ................................ ................................ ............................ 80 A 6 Correlation between fruit fall species richness and tree and liana basal area by year and tr eatment ................................ ................................ ........................ 80 A 7 Correlation between fruit fall species richness and tree and liana diversit y by year and treatment ................................ ................................ ............................ 80

PAGE 10

10 LIST OF FIGURES Figure page 2 1 Study site location. Tanguru Ranch, State of Mato Grosso, Brazil. .................... 27 2 2 Spatial arrangement of experimental plots, distribution of traps (red squares) and floristic s urvey transects ................................ ................................ ............. 28 3 1 Boxplots of tree and liana densities within 30 m of the traps for each of the five years of th is study grouped by treatment and yea r ................................ ..... 32 3 2 Boxplots of tree and liana basal area (m 2 ) within 30 m of the traps for each of the five years of this study grouped by treatment and year ............................... 33 3 3 for each of the five years of this study grouped by tr eatment and year ............. 34 3 4 Relationship between tree and liana densities in 30 m radius trap centered plots and distance from the forest agri culture edge (m) ................................ ..... 35 3 5 Intercept, slopes (la beled Distance), and confidence intervals for linear relations of tree and liana density in trap centered plots by distance from the fores t agriculture edge (m) ................................ ................................ ................. 36 3 6 Linear relation between basal area (m 2 ) in trap centered plots and distance from the forest agricultur e edge (m). ................................ ................................ 37 3 7 Intercept, slopes (labeled Distance), and confidence intervals for linear relations of tree an d liana basal area (m 2 ) in trap centered plots by distance from the forest agriculture edge (m) ................................ ................................ .. 38 3 8 centered plot s and distance from the forest agr iculture edge (m) ...................... 39 3 9 Intercept, slopes (labeled Distance), and confidence intervals for linear models of tree and liana diversity in trap centered plots by distance from the forest agriculture edge (m) ................................ ................................ ................. 40 4 1 Boxplots of fruit fall mass ................................ ................................ .................. 46 4 2 Fruit fall monthly trends across t he five years of the experiment ........................ 47 4 3 Boxplots of fruit fall mass by treatment and year. ................................ ............... 48 4 4 Least square means for effects of distance from the forest edge x treatment on fruit fall ma ss (g, back transformed values) ................................ ................... 50

PAGE 11

11 4 5 fruit fall ma ss (g back transformed values) ................................ ........................ 50 4 6 year on fruit fall ma ss (g, back transformed values) ................................ .......... 51 5 1 Boxplots of n umber of species in fruit traps ................................ ........................ 56 5 2 Number of species per month, treatment, and year ................................ ........... 57 5 3 Least square means values for the number of species in fruit fall treatment x density interaction. All other effects in the model are at their average values. ... 59 5 4 Lea st square means values for the number of species in fruit fall year x density interaction. All other effects in the model are at their average values. ... 59 5 5 Least square means values for the number of species in fruit fall year x distance interaction. All other effects in the model are at their average values. 60 A 1 Distribution of fruit fall mass (log10(mass+1)) by tree and lian a d ensities in trap centered plots ................................ ................................ .............................. 72 A 2 Distribution of fruit fall mass (log10(mass+1)) by live basal area (m 2 ) in trap centered plots ................................ ................................ ................................ ..... 73 A 3 Distribution of log transformed fruit fall mass (log10(weight+1)) by distance from the edge (m). ................................ ................................ .............................. 74 A 4 Distribution number of species per trap a nd tree and distance from edge ......... 76 A 5 Distribution number of species per trap and tree and liana density in trap centered plots ................................ ................................ ................................ ..... 77 A 6 Distribution number of species per trap and tree and liana ba sal area in trap centered plots ................................ ................................ ................................ .... 78 A 7 Distribution number of species per trap and tree and liana diversity ................................ ........................... 79 A 8 Model residuals by fitted values and explanatory variables. Plot titles indicate variables used. ................................ ................................ ................................ ... 81 A 9 Number of species mo del residuals by fitted values and explanatory variables. Plot titles indicate variables used. ................................ ...................... 82

PAGE 12

12 LIST OF ABBREVIATION S BA Basal area in meters squared CCTF Closed canopy tropical forest dbh diameter at breast height

PAGE 13

13 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science FRUIT FALL PATTERNS IN AN EXPERIMENTALLY BURNED AMAZONIAN FOREST By Eric Ol iveira Carvalho August 2011 Chair: Leda N. Kobziar Major: Forest Resources and Conservation Wildland fires in Amazonian forest s have increased over the last 20 years, being linked to droughts and coupled with anthropogenic l and use and ignitions The af termath of forest fires is generally a reduction in species richness, live stem density, and tree biomass, as well as increased litter input, canopy openness, and tree mortality rates Fire related tree mortality not only influences the future flammability of burned forests but can affect several ecosystem processes that are more elusive to track than the immediate effects of burning. Among these processes there are those related to tree reproduction and forest recovery. The present study a ddresses fruit fa ll patterns in an experimentally burned tropical forest in an attempt to shed some light o n the effects of fire on these processes This study was undertaken within a large scale fire experiment in an Amazonian Transitional forest. The experiment consiste d of one experimental unit divided into three 50 ha plots. T reatments consisted of a plot burned two times in fiv e years (B2), a plot burned four times in five years (B4), and a control (never burned B0) Fruit fall data was gathered in a series of traps ( 252) systematically placed within the plots T rap s were sampled twice monthly but for the analysis data was scaled to yearly periods. Linear

PAGE 14

14 mixed effects models were used for data analysis. In addition to evaluating the sole effect of fire, f ruit fall pat terns were also linked to tree and liana density, basal area, diversity, and distance from the forest agriculture edge. Fruit fall mass was higher in B4 than in the control at longer distances from the forest edge. Plot B2 and B0 did not differ in this re gards. Species richness in fruit fall was higher in B2 at higher tree and liana densities. Plots B4 and B0 did not differ from the control or from the two times burned plot in this regards. Both results are in contrast with our initial expectation that fru it fall mass and species richness would decline with the fire treatments

PAGE 15

15 CHAPTER 1 I NTRODUCTION Over the last two decades, fires in tropi cal closed canopied forest have emerged from a virtually unrecognized phenomenon to a serious threat to biodiversity and resources. While there is growing evidence of the enduring presence of fire in the histories of tropical forests, with potential contributions to the maintenance of biodiversity (Sanford et al. 1985), the severity and frequency with which modern fires burn in tropical forests is alarming. Furthermore, unlike historic tropical forest fires, emergent fire patterns are thought to be mostly a result of human alteration of tropical forests coupled with abundant ignition sources in the form of resource manag ement and extraction fires The alteration of forest cov er (through deforestation) and forest structure (principally through logging) by human land use affects several characteristics of tropical closed canopied forests that formerly conferred some degree of immunity to burning (Uhl and Kauffman 1990, Holdsworth a nd Uhl 1997, Ray et al. 2005 ) The aftermath of forest fires is generally a reduction in species richness, live stem density, and tr ee biomass, as well as increased litter input, canopy openness, and tree mortality rates. These effects and their extent s are dependent on forest location and fire history Fire induced stem mortality, for example, ranges from 8% to 23% in Amazonian transitional forests, while for Central Amazonian forests mortality ra nge from 36% to 64% (reviewed in Barlow and Peres 2006 a ). The severity of tropical forest fires has been attributed to the fact that tropical trees are poor ly equipped to cope with burns In particular, the bark of most species poorly protects their cambi um from lethal fire temperatures (Ulh and Kauffman 1990). The large number of trees killed by fires creates openings in the canopy allowing more

PAGE 16

16 sunlight to reach the forest floor, increasing temperatures and decreasing fuel moisture contents ( Uhl and Kauf fman 1990, Holdsworth and Uhl 1997, Ray et al. 2005 ). Thus, as with logging and other forest damaging practices, fire renders burned forests more prone to subsequent fires (Cochrane and Schulze 1999, Cochrane et al. 1999). Debris f rom dead trees adds to th e fuel loads of burned forests, which are generally higher than those fo und in un burned areas (Uhl and Kauffman 1990) Second and third burns have been reported to be of higher intensity and severity thus further degrading affected forests (Cochrane et al 1999 Cochrane 2003 ). Fire related tree mortality not only influences the future flammability of burned forests but can affect several ecosystem processes that are more elusive to track than the immediate effects of burning. Among these processes there are those related to tree reproduction and forest recovery. With growing concern over increased forest fires in the Amazon Basin, we need to understand the factors that determine the capacity of fire affected forests to recover species di versity and biomas s Tropical forest recovery after burning is certainly a slow process as is with other s evere disturbances (Chazdon 2003 ). In addition the regeneration is likely to be strongly dominated by species capable of sprouting following above ground mortality. Spr outing is an important means by which many species recover after cutting, though cutting followed by burning dramatically decreased the number of sprouts (Uhl et al. 1981). After disturbances of varying intensities, including high and low intensity fires, Kennard et al. (2002) observed that sprouts were the dominant form of regeneration in a transitional forest in the Bolivian Amazon. In contrast, Kauffman (1991) found that 39 69% of the top killed tree s in a burned Amazon forest did not sprout eight mont hs after fire and that 41% of

PAGE 17

17 the 125 species studied la cked the capacity to sprout. In addition, fires can decrease the amounts of viable seeds in soil seed banks by 50 to 94% in low intensity and high intensity burns respectively (Kennard et al. 2002). This realization implies that seed dispersal into burned areas is likely to be an important means by which burned forests can recover. Repeated burns, however, further complicate the picture, because subsequent fires kill sprouts and seedlings, damage see d banks, and kill trees that would otherwise contribute seed s In repeatedly burned Amazonian forests, understory regeneration after fire is dominated by herbaceous species that suppress woody plant regeneration (Barlow and Peres 2008). Fruit production i n burned forests has not been addressed extensively. Fruits and the seeds they contain play a major role in the dynamics of tropical forest plant and animal communities by acting as reproductive resources for the former and food for the later. These functi ons work hand in hand as many tropical plant species produce animal dispersed seeds diets. In some tropical forests, for example, the tree flora may be comprised of up to 95% zoochorous specie s (Peres and Roosmalen 2002). By influencing fruit production, fire can have a profound effect on plant recovery and wildlife populations. Barlow and Peres (200 6b ), demonstrated that fires depressed the basal area of fruiting trees The authors also demons trated that despite a community wide reduction in fruit ing tree basal area not all species presented decreased number of fruiting stems Barlow and fire fruit production though their study did not measure fruit production directly

PAGE 18

18 The study of reproduction in tropical forests is in itself a difficult enterprise in that these forests usually contain numerous tree species that occur at low densities and employ different reproductive s trategies. This implies that reproductive phenomena in tropical forests are spatially and temporally heterogeneous, requiring large areas and long periods of time sampled to provide a glimpse into their dynamics. Furthermore, baseline s to compare disturbed and undisturbed forests are scarce (Barlow and Peres 2006a) Thus it is no surprise that few studies have addressed the consequences of fire for plant reproduction in c losed canopy tropical forests. We present data on the effects of fire on fruit fall pat terns of experimental ly burn plots on the southern edge of the Amazon Basin. Our aim was to assess the effects of repeated burns on temporal patterns of fruit fall. The treatments applied to 50 ha plots consisted of a plot burned twice in five years (B2), a plot burned four times in five years (B4), and a plot never burned (B0, no signs of recent burning). As different burning intervals have been shown to influence fire behavior and tree morality, we assessed differences among these treatments. We first inv estigated whether fruit fall changed over time and among treatment s subsequently relating these patterns to fire induced changes in tree and liana density basal area (BA), and species diversity. Furthermore, because forest fires are often associated with deforestation edges, we evaluate the combined effect of fire and distance from the edge on fruit production. Edges influence both tree mortality patterns and fire behavior (reviewed in Laurance 2006). Higher temperatures, insulation, and wind exposure cre ate edge environments that favor rapid fuel dry down rates. Higher tree mortality rates and litter input on edges insure higher fuel loads in these areas that in association with drier conditions lead to fires of higher

PAGE 19

19 intensity in edges than in forest i nteriors. In addition, invasion of pasture grasses can occur into edge environments with substantia l effects on fire behavior (Balch et al. 2009, Veldman et al. 2010 ). The differences in fire intensity associated with forest edges may have a profound influ ence on tree mortality that in turn can affect fruit production. Our expectations are in conformity with those of Barlow and Peres (2006b), who argue that given the severity of Amazonian forest fires, with substantial reductions in tree density and BA, fru it production is likely to decline in burne d forests when compared to un burned forests

PAGE 20

20 CHAPTER 2 MATERIALS AND METHOD S Study Site and Design This stud y was conducted on Taguru Ranch, a privately owned land holding of ~80,000 ha with transitional for est only 30 km north of the Brazilian savannas ( cerrado ) in the state of Mato Grosso (Figure 2 1; 13 o n the southern edge of the Amazon B a sin, the study site falls within a broad transition zone between the Amazonian forest and ce rrado biomes. These forests share plant species with both the cerrado and the wetter closed canopied forests of central Amazonia. Our study forest, for example, has at least 23 species in common with the cerrado to the south (Balch et al. 2011 ) and 45 spec ies with the wetter forests of the Ducke reserve near Manaus, Brazil (Balch unpublished data) Compared to central Amazonian forests, transitional forests harbor fewer species, are shorter, and support less leaf area. In addition, the study area experience s seasonal droughts from May September, with a mean annual precipitation of 1740 mm for 2004 and 20 05 (Balch et al. 2011 ). agriculture edge (Figure 2 1), one 150 ha (1.5 x 1.0 km) exper imental forest block was located. The study block is surrounded on the three other sides by at least 1000 m of un broken forest, and has not experienced logging or burning in the recent past. The deforested side is currently used for soy cultivation but wa s formerly pasture. The t hree 50 ha (0.5 x 1.0 km) treatment plots established within the ex perimental block (Figure 2 2.A) consisted of a control ( B0; never burned), a twice burned plot (B2, burned twice in five years), and a four times burned plot (B4 bu rned four times in five years) The B4 burn plot was not burned in 2008.

PAGE 21

21 Experimental F ires The first burn of both experimental plots was in September 2004 with s ubsequent burns also near the end of the dry season (August September) Fires were lit as s trips following the north south trails, and each burn plot took 2 3 days to be completed ( Balch et al. 2008 for more details on experimental fires). Floristic S urvey Prior to the first burn every tree with diameter at breast height (DBH) 40 cm in e ach plot was mapped, tag ged, and measured for height and DBH ( stem diameter at 1.30 m ). Trees and lianas 20 40 cm DBH were sampled in belt transects (500 x 20 m) 0, 30, 100, 250, 500, and 750 m from the forest agriculture edge ( i.e., 5.5 ha sampled per 50 ha treatment plot ). Trees and lianas 10 20 cm DBH were sampled within smaller transects (500 x 4 m) nested in the larger o nes (Figure 2 2). All marked trees and lianas were revisited prior to each burn to assess mortality and any new individuals that recru ited into the considered size classes were surveyed. Fruit fall Fruits were collected from a series of 0.5 m 2 traps placed systematically throughout the experimental plots. Traps were located 5 m outside of the forest in the agricultural clearing and 50 apart at 0, 15, 30, 50, 100, 250, 500, and 750 m (north south) from the edge (Figure 2 2.A). A total of 270 traps were deployed in the 150 ha area, 90 per treatment T raps bordering two treatment plots (i.e. traps in B2 but close to B0 and those in B4 but close to B2) were excluded from analysis. In total 18 traps were excluded. These traps were considered to be too close to other treatment plot as they were located less that 10 m from the edge of two treatment plots. Their proximity to two treatment edges makes them prone to be influenced by trees within two different

PAGE 22

22 treatment plots (Figure 2 2.A). In total, therefore, the actual number of traps used for analysis was 252 (90, 81, and 81, for plots B0, B2, and B4 respectively). All t raps were visited at ~tw o week intervals starting after the fire in September 2004 and continuing through September 2009 Fruits and seeds in the trap s were identified to the lowest taxonomic level possible, counted, oven dried and weighed. Forest Structure and Diversity Analysi s The analysis of the effects of fire over forest structure and diversity was conducted descriptively by the interpretation of graphs. The combined effects of fire and distance from the forest edge on tree and liana density, BA, and diversity in trap cente red plots was analyzed through graphs, and correlation analysis, assuming a linear relationship among the variables. For this portion of the analysis data was grouped by year and treatment and correlation values were assigned to each treatment year combina tion with distance from the edge being an explanatory covariate. For this analysis we sought to determine whether the linear relationship between the measures of forest structure and diversity and distance from the edge were affected by repeated fires. Fru it fall Data Analysis Data analyses included both qualitative interpretation of graphs and quantitative assessments with inferential statistics. Analyses were conducted at the community level with all species in the fruit fall pooled Community level ana lyses included fruit fall species richness and fruit fall mass. In the statistical analyses t he unit of observation was the trap and the observation period the year, and fruit/seed dry weight was used as the measure of production Traps were sampled repeat edly; therefore, to account for the repeated measurement structure of the experimental design, mixed modeling

PAGE 23

23 methods were used that specifically accounted for the correlations among measurements taken on the same experimental unit (the trap). We use d dr y weight as the measure of production because fruits were not distinguished from seed s during the processing phase. F or example, if a n intact fruit contained three seeds it would be counted as one but if the same fruit was opened and the seeds exposed it w ould be counted as three. For the purpose of this study we refer to the dry weights of reproductive material (i.e. fruits and seeds), as fruit fall mass. We recognize that t here are issues in utilizing dry weight as the measure of production the most obv ious of which is the fact that a great majority of studies of seed/fruit fall report their results in terms of numbers of seeds Furthermore using dry weight may over emphasize species with large heavy fruits over those with smaller, lighter fruits. Conver sely, using number of seeds as the response variable over emphasize s species that produce large number s of small seeds. All statistical analyses were conducted on a yearly basis because we intended to relate fruit fall and measures of forest structure and diversity As floristic data were acquired on a yearly basis, we scaled the fruit fall observations to yearly periods starting in August when e xperimental fires were conducted. Tree and liana survey data were used to create a serie s of variables that in cluded BA, d ensity and iversity index. The relationship between the se variables and fruit fall were explored graphically and also used as covariates in the analysis of fruit fall Given that the experimental plot s are relatively large unit s in comparison to each trap rather than assigning one value per year per treatment for each explanatory variable, we assigned a unique value for each variable to each trap

PAGE 24

24 To assess local conditions, trees and lianas within 30 m of each trap were used to cal culate structure and diversity variables for each year (Figure 2 2.B). Having tested radii, ranging from 20 m to 70 m we selected a 30 m radius on the basis of the largest correlation between explanatory and response variables. The choice of the 30 m radius plot is also supported by the fact that most fruits tend to disperse close to the parent plants with a rapid decrease in density with increasing distance. For example, Godoy and Jordano (2001) found tha t for the zoochorous Prunus mahaleb in Spain, up to 62% of the seeds arrived within 15 m of the parent tree, while also dispersing seeds to distances up to 316 m. The present study has two major experimental design challenges The first is lack of true r eplication as there is only one plot per treatment and using traps as replicates therefore represents pseudo replication (Hulbert 1984). Nevertheless we used inferential statistics as recommended by Oksanen (2001), who advises that inferential statistics c an assist with the observation of patterns particularly where comparison of time trajectories among treatments is sought. The second deficiency is tha t fruit fall data collection only commenced after the first burn so we must assume that the fruiting patte rns were similar among treatment plots prior to the onset of the experiment. These deficiencies notwithstanding, the study spans a large number of years and was conducted over a large spatial scale. Community Fruit fall Mass Analysis Community level fru it fall mass analysis was primarily intended to elucidate changes in fruit fall patterns among treatments over time. We also wanted to ascertain the relationship between fruit fall mass and forest conditions such as tree and liana d ensity BA, and d istance from the forest agriculture edge. Given our main objectives

PAGE 25

25 we devised a statistical model that contains independent variables and interactions that we believe are appropriate to answer questions posed Thus we initiate d our model selection with the foll owing fixed eff ects: treatment, year, d istance, density, treatment x y ear Treatment x d istance y ear x tree and liana density, year x d istance from the edge, treatment x year x density, and treatment x year x d istance. This model was considered from which several nested models were derived for comparison. We also determined that any model analyzed would at least contain the va riables year, d istance and d ensity. We di d this because previous work on fruit produ ction suggest s that these variables have strong influence s on community reproductive output. In addition we chose to utilize only one measure of tree and liana abundance in models, since there is high correlation between Density and BA (r=0.80 p<0.001). The most parsimonious model w as selected on the basis of Akaike Informat ion Criterion (AIC) As mentioned above, mixed models with a random effect for year and trap were used. The response variable (fruit mass) was log 10 transformed for this portion of the analysis. Number of S pecies Re producing In addition to graphical interpretation, t he number of species in the fruit fall traps was also assessed through the use of linear mixed model s Covariates in this analysis included those utilized in the community analysis (see above) with the addition of i p i Log b p i where p i is the proportional abundance of species i and b is the logarithm base. As in the analysis of fruit fall mass we utilized model selection procedures to arrive at a model that best explains the data.

PAGE 26

26 The global model for sp ecies rich ness of fruit fall contained the follow ing simple effects: treatment, year, density, distance, and d also contain ed the following interactions: treatment x (y ear ; density; distance; diversity), year x (density; distance ; diver s i ty) t reatmen t x year x (density; distance; d iversity). Model selection was conducted by removing model terms until the models contained at least year, density, and d istance as explanatory variables. As with fruit fall mass we selected the model to further analyze the number of species in fruit fall using AIC. Mixed models were also structured with random effects for year and trap All analyses were conducted in R (R Development Core Team 2010), and the following packages were extensively used: gg plot2 (Wickham 2009), lattice (Sarkar 2008 ), nlme (Pinheiro et al. 2010), doBy (Hojsgaard et al. 2010 ), vegan (Oksanen et al. 2011), and lme4 (Bates and Maechler 2010 )

PAGE 27

27 Figure 2 1. Study site l ocation Tanguru Ranch State of Mato Grosso, Brazil, lowe r right image shows the experiment block (black rectangle) surrounded by forest on three sides and edging an agriculture field on the north

PAGE 28

28 Figure 2 2. Spatial arrangement of experimental plots, distribution of traps (red squares) and florist ic survey transects (indicated by small arrows on the left margin ; A ). The green rectangles show areas from which traps were excluded due to proximity of two treatments. B) A sample map of three trap centered plots; red squares represent traps, open red ci rcles the perimeter of each plot, and solid circles trees. A B

PAGE 29

29 CHAPTER 3 FIRE AND EDGE EFFECTS ON FOREST ST RUCTURE Tree and Liana Density, Basal Area and Diversity in Trap centered Plots As reveled by descriptive analysis the data, f ire caused decreases i n tree and liana densities near the traps in both fire treatments especially af ter the 2007 fire, with a more pronounced fall in B2 (Figure 3 1 ). The mean density per trap centered plot in B2 was init ially 24.9 ( standard deviation ( sd ) =9.9), fell to 10.6 ( sd=1 0.6) in 2007 2008, and remained relatively constant at 10.4 (sd=10.2) the following year. Similarly, tree and liana de nsity in B4 steadily fell to 56 % of its 2004 2005 survey by 2008 2009. The mean tree and liana density per trap centered plot in 20 04 2005 in B0 was 24.8 (sd=8.9) but by 2008 2009 was 21.2 (sd=8.7), a decrease to 85.73% of its 2004 2005 survey amount In concert with declines in tree and liana density basal area of trees and lianas also showed a substantial decline in the burned plots a fter the 2007 fire, especially in B 2 (Figure 3 2 ). Both tree densities and BA remained relatively constant after their de cline following the 2007 fire. Lower divers ity of trees and lianas in trap ) was associated with fi re treatments with a more pronounced fall in B2 than B4 (Figure 3 was 0. 37 (sd=0.14) in 2004 2005 and fell to 0.20 (sd=0.19) in 2008 2009 On B4 (sd=0.11) in 2004 2005 ; it decreased with subsequent fires with a pronoun ced decrease in 2007 2008 when it reached 0.29 (sd=0.17) and subsequently fell to 0.27 (sd=0.18 ) in 2008 2009. The control plot (B0) maintained a relatively constant diversity in trap centered plots th r ough out the experiment 0.39 (sd=0.1 3) in 2004 2005 that dropped slightly to 0.35 (sd=0.15) in 2008 2009

PAGE 30

30 Distance From the Edge E ffects on Tree and Liana Density Basal Area and Diversity Tree and liana density in trap centered plots was positively correlated with distance from the forest edge both with all treatments and years pooled and by treatment and year separately (Figure 3 4 ). With treatments pooled distance was weakly correlated with tree and liana density (r=0.42 ). When grouped by treatment and year low correlation values were observed, but increased after 2007 2008 in B2 (Table 3 1 ) In addition fire appear s to have influenced the relation ship between density and distance from the edge as intercepts show a decreasing trend while slopes show an increasing trend with repeated b urns (Figur e 3 4 and Figure 3 5 ). The decline in intercepts indicated that the mean density in trap centered plots at the forest edge decreased with time and fires. The increasing slopes show that fire caused the relationship between density and distance f rom the edge to become steeper. That is density tended to increase with distance at higher rates with passing years and subsequent fires. This effect was more apparent in B2 than B4; while intercepts in B4 decrease d, slopes did not increase as in B2. Tre e and liana basal area in trap centered plots followed the same trend as density, increasing with distance from the forest edge (Figure 3 6 ) Distance was only weakly related to changes in basal area wit h treatments and years pooled (r=0.41 ). With data group ed by treatment and year the correlation between BA and distance from the forest edge varied great ly depending on treatment and year remaining weak, however, in the control and B4 treatments but increasing in B2 after 2007 2008. Furthermore as with densi ty, fire appea r ed to influence the relation ship between basal area and

PAGE 31

31 distance causing a decrease in intercept s both for B2 and B4 and increase in slopes especially in B 2 (Figure 3 7 ). The diversit y of trees and lianas in trap centered plots also was po sitively associated with distance from the forest edge, though this relation was weak (Figure 3 8) For all treatments and years distance from forest edge was weakly correlated with =0 .30 ) In addition with data grouped by year and treatment the co efficient o f determination remained low but tended to increase with time (Table 3 3 ). Of the three treatments B2 had the highest increase in correlation with years, particularly after 2007 2008. The intercepts and slopes of the relation distance fro m the edge showed simil ar tendencies in all treatments, the intercepts declined while slopes increa sed with years (Figure 3 9 ). In the control these tendencies were not apparent as observed by both intercepts and slopes and their confidence in tervals In contrast, B2 and B4 intercepts d eclined after 2007 2009. Only for B2 was the slope increase over time and treatment substantial

PAGE 32

32 Figure 3 1. Boxplots of t ree and liana densi ties within 30 m of the traps for each of the five years of t his study grouped by treatment and year (B0 = control, B2= two times burned, B4 = four times burned) The boxplot represents the 25 th 50 th and 75 th percentile, whiskers are the first and third quartiles +/ outliers.

PAGE 33

33 Figure 3 2. Boxplots of tree and liana basal area (m 2 ) within 30 m of the traps for each of the five years of this study grouped by treatment and year (B0 = control, B2= two times burned, B4 = four times burned). The boxplot represents the 25 th 50 th and 75 th percentile, whiskers are the first and third quartiles +/ 1.5 interquartile range,

PAGE 34

34 Figure 3 3. this study grouped by treatment and year (B0 = control, B2= two times burned, B4 = four times burned). The boxplot represents the 25 th 50 th and 75 th percentile, whiskers are the first and third quartiles +/ 1.5 interquartile range,

PAGE 35

35 Figure 3 4 Relationship between t ree and liana densities in 30 m radius trap centered plots and distance from the forest agriculture edge (m) A) Al l treatments and years pooled (r=0.42 ). B) Grouped by treatment and year.

PAGE 36

36 Table 3 1 rrelation (r) values for the relation between tree and liana density in trap centered plots and distance from the forest agriculture edge (m) grouped by treatment and year. 2004 2005 2005 2006 2006 2007 2007 2008 2008 2009 B0 0.41 0.42 0.49 0.48 0.47 B 2 0.47 0.47 0. 52 0.82 0.77 B4 0.31 0.28 0.32 0.42 0.3 8 Figure 3 5 Intercept, slopes (labeled Distance), and confid ence intervals for linear relations of tree and liana density in trap centered plots by distance from the forest agriculture edge (m) grouped by treatment and year (note that scales differ). Plotted lines represent the 95% confidence intervals with the mean in the middle. B0 B2 B4

PAGE 37

37 Figure 3 6 Linear relation between basal area (m 2 ) in trap centered plots and distance from the forest agricu lture edge (m). A) Al l treatments and years pooled (r=0.41 ). B) Grouped by treatment and year.

PAGE 38

38 Table 3 2 tree and liana basal area (m 2 ) in trap centered plots and distance from the forest agricu lture edge (m) grouped by treatment and year. 2004 2005 2005 2006 2006 2007 2007 2008 2008 2009 B0 0.36 0.37 0.3 8 0.39 0.38 B2 0.48 0.52 0.52 0.7 1 0.69 B4 0.36 0.32 0.34 0.42 0.39 Figure 3 7 Intercept, slopes (labeled Distance), and confid en ce intervals for linear relations of tree and liana basal area (m 2 ) in trap centered plots by distance from the forest agriculture edge (m) grouped by treatment and year (note that scales differ). Plotted lines represent the 95% confidence intervals with t he mean (small dash) in the middle. B0 B2 B4

PAGE 39

39 Figure 3 8 centered plots and distance from the forest agriculture edge (m). A) Al l treatments and years pooled (r=0.30 ). B) Grouped by treat ment and year.

PAGE 40

40 Table 3 3 centered plots and distance from the forest agriculture edge (m) grouped by treatment and year. 2004 2005 2005 2006 2006 2007 2007 200 8 2008 2009 B0 0.18 0.15 0. 1 8 0.29 0.33 B2 0.33 0.40 0.37 0.53 0.56 B4 0.23 0.24 0.28 0.33 0.34 Figure 3 9 Intercept, slopes (labeled Distance), and confidence intervals for linear models of tree and liana divers ity in trap centered plots by d istance from the forest agriculture edge (m) grouped by treatment and year (note that scales differ). Plotted lines represent the 95% confidence intervals with the mean (small dash) in the middle. B0 B2 B4

PAGE 41

41 CHAPTER 4 FRUIT FALL MASS RESU LTS Community w ide Results In total 20 .2 k g dry weight of fruits were collected from the litter traps, 5 .7 k g, 6 .1 Kg and 8 .4 k g, in the control (B0), B 2, and B 4, respectively (for production on a yearly basis per treatment refer to Table 4 1 ). More than 50% of the fruit fall mass pr oduced was usually from a set of only 10 species, of which five are among the species with highest Importance Value Index (IVI) (Table 4 3). It is worth noting that among these 10 species one, Strychnos xinguensis Krukoff., is a zoochorous liana. Fruit fal l Mass Patterns as Revealed by Qualitative Analysis When considered univariately, f ruit fall mass did not appear to differ between the two burned plots and the control (Figure 4 1 .B) but fluctuated substantially among years. The year 2004 2005 had the high est mass an d 2005 2006 the lowest (Figure 4 1 .A). Fruit fall mass also varied considerably intra annually with similar general temporal patterns i n all three plots ( Figure 4 2 ) Monthly produc tion generally peaked January April for all years and treatment s except in 2004 2005, where B2 peaked in September October and B0 and B4 that peake d in November December (Figure 4 2 .A). Annual p eaks coincided with the October April rainy season. When considering the proportion of traps in which fruits were found, a si milar trend towards more fruiting in the rainy season was observed for al l treatment s and years (Figure 4 2 .B). Treatment B2, after the 2007 fire, consistently had lower numbers of traps with fruits than the other two treatments. Annual total fruit fall mass per treatment fluctuate d among years similarly in the three plots e xcept B2 in 2008 2009 (Figures 4 3 .A and B), d uring which its traps

PAGE 42

42 collected only 795 g, which was less than half as much as that collected in B0 and in B4 for that year. By contrast, over the previou s four years, B2 production was: 6 6%, 31%, 30%, and 13% higher than fall mass, for the years of 2004 2005, 05 06, 06 07, and 07 08, respectively. In relation to B4, B2 production for the four previous years was: 87%, 84%, 92%, and 73%, for the years of 2004 2005, 05 06, 06 07, and 07 08 respectively. Relationship between Density, BA, and Fruit fall Mass The mass of fruits trapped increase d with tree and liana density and BA in trap centered plots ( r =0.27, p<0.001 m ass v s density, r =0.24, p<0.001 m ass vs BA, Figu res A 1 and A 2 in Appendix ). The linear relationship s between these variables were weak as attested by the ir correlation ( Table A 1 and A 2 in Appendix ). In addition the correlation between d ensity and fruit fal l mass and BA and fruit fall mass varied with treatments and years. Some years the correlation was not significant while in others the correlation s among variables were stronger. Relationship between Distance from the Edge and Fruit fall Mass Fruit fall m ass in all treatments and years increased from the edge to the interior of the treatment plots ( r =0.19, p<0.001, Figure A 3 ). This relationship was observed bot h with all data pooled (Figure A 3.A ) and whe n grouped by year and treatment (Figure A 3 .B) Nev ertheless, t he correlation coefficients varied substantially by year and treatment ( Table A 3) Statistical M odels of Fruit fall Mass Patterns The global model (the statistical model that encompassed all other models tested) for fruit fall mass ranked 3 6 th among the 45 models tested its high AIC, AICc, and low weight indicate d that it was not an appropriate model to describe fruit fall mass.

PAGE 43

43 Wald tests reve a led that several terms in the global model were significant (p<0.05; including year, treatment, d ensity, distance, year x density, year x distance, treatment x distance, and the triple way interaction year x treatment x distance ; Table 4 4 ) The most parsimonious model contained the following terms: year, t reatment, distance, density, year x density, year x distance, treatment x distance. In contrast with to the global model, all terms in the selected model were significant (p<0.05 T able 4 5 ) The top ranked model had an AIC of 3803.2 AICc = 3805.2 ( Figure A 8 for model diagnostics) which indicates t hat the model has high explanatory power for fruit fall mass. The fixed effects of the model selected to best represent fruit fall mass revealed that fruit fall mass was significantly higher in B4 at larger distances from the forest edge from B0, that is d istance and the treatment B4 interacted significantly showing that on average fruit fall mass increased with distance. Plot B2 and B0 did not show this tendency and fruit fall mass did not differ among these two t reatments (Table 4 6, Figure 4 4 ). Fruit fa ll mass fluctuated among year s consistently showing less frui t mass per trap after 2004 2005 (Table 4 6 ) Fruit fa ll mass per trap in the burned plots did not differ ed from the control Fruit fall mass increased with high er tree and liana densities. The i nteraction between density and year was significant for 2008 2009 and margi nally significant for 2006 2007; f ruit fall mass tended to be higher per unit density in trap centered plots i n those years when compared t o that of 2004 2005 (Figure 4 5 ). In addit ion areas with h igher density showed much higher fruit mass in traps (Figure 4 5 ). Distance and year significantly interact ed in 2005 2006 and 2008 2009 ; during these years fruit fall mass was lower at larger distances from the forest edge. In addition, o ver

PAGE 44

44 all years, there appears to be no statistical difference between fruit fall mass at the forest edge and up to a distance of 100m (Figure 4 6 ). T he results statistical analyse s of fruit fall mass are not in agreement with those revealed by descriptive analyse s. F ruit fall mass was shown to be higher in B4 at further distances into the forest by the statistical model. Treatments B2 and B0 did not differ among themselves

PAGE 45

45 Table 4 1. Summary of fruit fall mass (g) grouped by year and treatment. Numb er in parenthesis following mean is the standard error. Year 2004_2005 2005_2006 2006_2007 2007_2008 2008_2009 total Treatment total M ean T otal mean total mean total mean total mean B0 1438.2 16.0(37.8) 247.1 2.7(5.9) 1053.7 11.7(20.9) 558.1 6.2(11.3 ) 1854.2 20.6(36.0) 5151.4 B2 2387.2 29.5(61.1) 322.7 4.0(7.8) 1375.0 17.0(30.3) 631.1 7.8(17.6) 795.1 9.8(18.0) 5511.1 B4 2737.2 33.8(80.0) 383.0 4.7(9.6) 1493.9 18.4(32.6) 857.4 10.6(19.0) 1865.6 23.0(43.3) 7337.0 yearly total 6562.6 95 2.7 3922.6 2046.6 4514.9 17999.4 Table 4 2. Number of traps where fruits were found (in parenthesis percent total traps in treatment) and the area sampled (m 2 each trap is 0.5 m 2 ). Year 2004_2005 2005_2006 2006_2007 2007_2008 2008_2009 Treatme nt N. Traps Area N. Traps Area N. Traps Area N. Traps Area N. T raps Area B0 87(97) 43.5 84(93) 42 87(97) 43.5 89(99) 44.5 87(97) 43.5 B2 81(100) 40.5 71(88) 35.5 77(95) 38.5 71(88) 35.5 67(83) 33.5 B4 81(100) 40.5 72(89) 36 79(97) 39.5 78(96) 39 76(93) 38 T otal 249(99) 124.5 227(90) 113.5 243(96) 121.5 238(94) 119 230(91) 115 Table 4 3. Total fruit fall mass per year and treatment of the 10 most productive species in terms of mass (g) and percentage of the total production per treatment and year. Year 2004_2005 2005_2006 2006_2007 2007_2008 2008_2009 Total Treatment mass % total mass % total mass % total mass % total mass % total mass % total B0 731.79 50.88 174.87 70.77 891.51 84.61 347.09 62.19 1599.98 86.29 3745.24 72.70 B2 563.81 2 3.62 180.10 55.82 1022.97 74.40 444.72 70.47 534.89 67.27 2746.49 49.84 B4 1024.93 37.44 278.26 72.66 1127.05 75.44 587.67 68.54 1524.99 81.74 4542.90 61.92 Total 2320.54 35.36 633.23 66.47 3041.53 77.54 1379.48 67.40 3659.86 81.06 11034.64 61.31

PAGE 46

46 Figure 4 1 Boxplots of fruit fall mass. A) Yearl y total fruit fall mass (log10(mass +1)) per trap across treatments. B) Fruit fall mass per trap per treatment across all years of the study. The boxplot represents the 25 th 50 th and 75 th per centile, whiskers are the first and third quartiles +/ A B

PAGE 47

47 Figure 4 2. Fruit fall m onthly trends across the five years of the experiment A) M ean monthly fruit fall mas s (g) per treatment and year. B) Proport ion of traps containing fruits per month, year, and treatment. Years start on August (1) and end in July (12), shaded area represents months of the rain season.

PAGE 48

48 Figure 4 3. Boxplots of f ruit fall mass by treatment and year. A) Untransformed mass (g), with extreme values om itted; B) L og (log10(mass +1)) fruit fall mass per trap by treatment. The box plot represents the 25 th 50 th and 75 th percentile, whiskers are the first and third quartiles +/ 1.5 interquartile range and outliers. A B

PAGE 49

49 Table 4 4 Wald test statistics for terms in Global model of fruit fall mass numDF denDF F value p value (Intercept) 1 969 906.8693 <0.001 Y ear 4 969 96.973 <0.001 Distance 1 246 28.7027 <0.001 Density 1 969 51.7514 <0.001 Trea tment 2 246 3.6243 0.0281 year:Distance 4 969 3.2794 0.0111 year:Density 4 969 5.0251 0.0005 Distance:Treatment 2 246 4.114 0.0175 Density:Treatment 2 969 0.8211 0.4402 year:Treatment 8 969 1.2841 0.2479 year:Density:Treatment 8 969 1.0771 0.3767 ye ar:Distance:Treatment 8 969 2.2143 0.0244 Table 4 5 Wald test statistics for selected fruit fall mass model numDF denDF F value p value (Intercept) 1 995 915.4195 <0.001 Y ear 4 995 97.1084 <0.001 Distance 1 246 28.9341 <0.001 Density 1 995 50.55 8 <0.001 Treatment 2 246 3.6354 0.028 year:Distance 4 995 3.2173 0.012 year:Density 4 995 4.8735 <0.001 Distance:Treatment 2 246 4.1237 0.017 Table 4 6 Summary of fixed effects in fruit fall mass model. Value Std.Error DF t value p value (Interc ept) 1.4262392 0.237728 995 5.999461 0.000 2005 2006 1.189299 0.247012 995 4.81474 0.000 2006 2007 0.749551 0.265458 995 2.82361 0.005 2007 2008 0.999954 0.233389 995 4.28449 0.000 2008 2009 0.579747 0.252856 995 2.2928 0.022 Distance 0.000520 6 0.000417 246 1.247855 0.213 Density 0.0245334 0.008719 995 2.813866 0.005 B2 0.2400206 0.147167 246 1.630941 0.104 B4 0.0653731 0.144196 246 0.453362 0.651 2005 2006 x Distance 0.000856 0.000366 995 2.33828 0.020 2006 2007 x Distance 0.000711 0.0 00412 995 1.7244 0.085 2007 2008 x Distance 0.000438 0.000395 995 1.10956 0.268 2008 2009 x Distance 0.001631 0.00046 995 3.54282 0.000 2005 2006 x Density 0.001535 0.009828 995 0.15613 0.876 2006 2007 x Density 0.0208066 0.010983 995 1.894461 0 .059 2007 2008 x Density 0.0176516 0.010409 995 1.695724 0.090 2008 2009 x Density 0.0380271 0.011753 995 3.235652 0.001 Distance x B2 0.0003181 0.000459 246 0.69281 0.489 Distance x B4 0.0012745 0.000457 246 2.787067 0.006

PAGE 50

50 Figure 4 4 Least square means for effects of distance from the forest edge x treatment on fruit fall mass (g, back transformed values). All other effects in the model are at their average values. Figure 4 5 Least square means for effects of tree and liana density x year o n fruit fall mass (g, back transformed values ). All other effects in the model are at their average values.

PAGE 51

51 Figure 4 6 Least square means for effects of distance f rom the forest x year o n fruit fall mass (g, back transformed values ). All other effects in the model are at their average values

PAGE 52

52 CHAPTER 5 NUMBER OF SPECIES RE PRODUCING Number of S pecies Reproducing Description an d Qualitative Analysis Fruits or seeds of 80 plant species were found in the traps: 73 trees and 7 lianas representing 82% of the total number of trees and lianas species found in the floristic inventory (97). Most species trapped are animal dispersed (70. 0%, 56 species), 13.7% (11 species) are anemochoric and 2.5% (2 species) are autochoric (fruits that explode releasing seeds, found in several species in the Euphorbiaceae), 13.7% (11) species had undetermined dispersal mechanisms. Furthermore the proport ion of the fruit fall mass i n different dispersal modes fluctuate d yearly but was not affected by the burn treatments (Table 5 1). The numbers of species found per trap did not appear to differ among treatments (Fi gure 5 1 .A) with medians of 3 species in B0 and B2, and 4 in B4 with years pooled On an annual basis (with treatments pooled) the number of species found per trap in 2004 2005 (median=6), 2006 2007 was the second highest (median=4), and the other years all had median=3 (Figure 5 1 .B). The num ber of species in fruit fall varied monthly and yearly but did not show any divergence among treatments with time. Unlike fruit fall mass, the number of species in fruit was not clearly seasonal (Figure 5 2 .A ) Relationship between Number of Species and Di stance, Density, BA, and Diversity The number of species per trap with all treatments a nd years pooled increased with D istance from the forest edge ( r = 0.25, p <0.001, Figure A 4 ). This relationship was also observed with data grouped by year and treatme nt, tho ugh it remained weak

PAGE 53

53 ( Figure A 4.B and Table A 4 ). The number of species per trap as a function of distance varied by year and treatment and at times yielded high correlations. The number of species falling into individual traps increased w ith tree and liana d ensities ( r = 0.38 p < 0.001 ) and BA ( r = 0.31 p <0.01 ) in trap centered plots (Figure A 5 and A 6 respectively). When the relationship between number of species and tree and liana density were grouped by treatment and year the correlation varied substantially, again, with some years yielding high correlation s (Table A 5 ). T he correlation of tree and liana density with number of species in B2 tended to increase over the years The correlation between number of species and BA also varied by year but was usually lower tha n the correlation between tree and liana d ensity and number of species (Table A 6) I ncreasing diversity per trap centered plot also had an increasing effect on the number of species in fruit fall (r = 0.26 p <0.001 Figure A 7 ) Quantit ative Analysis of Number of Species in Fruit fall Model selection procedures arrived at the following model: year, treatment, d istance, d ensity, y ear x d ensity, year x distance, treatment x d ensity, and treatment x d istance (Table 5 2 ) All terms were si gnificant at p <0.05 except for treatment x d istance ( p =0.099, Wald test). This model had AIC = 5248.5 AICc = 5249.6 and weight = 0.164. The next model in the analysis conta ined the same terms except for treatment x d istance interaction, and all terms wer e significant (p<0.05; Table 5 3 ). This model was not considered statistically different from the model with lowest AIC, AICc, and highest weight ( p =0.09 ). Thus by the principle of parsimony we selected the second model given that there are no differences and the second model contains k= 23 parameters while the first contains k= 25 For this model, AIC = 5249.3 AICc = 5250.2 and weight = 0.1182

PAGE 54

54 The model selected for further inquiry of the data had several interaction terms: treatment x density, year x de nsity, and year x distance. The treatment x density interaction indicated that treatment B2 have more species present in its traps at high tree and liana densities than the control. Treatments B2 and B4 did not have different number of species at any densi ty levels. In addition B4 and B0 did not di ffer in this regards (Figure 5 3 ). The number of species tended to be higher at high levels of tree and liana density for all years with the year of 2004 2005 always having a higher number of species and 2005 2006 consistently having the lowest number of species (Figure 5 4 ) The year x d istance interaction indicated that for all years except 2008 2009 the number of species in fruit fall tended to increase with increasing distance from the edge with a steeper incre ase beyond 100 m from the fore st agriculture edge (Figure 5 5 ). For the year 2008 2009 this interaction indicated that differently from the other years, the number of species tended to decline with increasing distance from the fores edge. Furthermore the y ear 2004 2005 had the highest number of species found and 2005 2006 the lowest. Comparing the qualitative and quantitative analyses of the number of species in fruit fall we observed that some of the conclusions differ. The number of species did fluctuate yearly and this was revealed by both analyses. In addition there was a tendency for the number of species to increase with both increasing distance from the edge and the density of trees and lianas surrounding the traps. The quantitative analysis, however indicated that the number of species in fruit fall in treatment B2 was higher at high tree and liana densities when compared with the control. This result

PAGE 55

55 differs from the pure interpretation of graphs and tables, which, on the contrary, indicates that B 2 had less species falling in its traps than the B0. Table 5 1. Summary of fruit mass (g) by dispersal mechanism and percent of total mass by dispersal mechanism per treatment and year. Year Treatment B0 B2 B4 Dispersal Total mass % of total Total m ass % of total Total mass % of total 2004 2005 a uto 5.33 0.37 3.36 0.14 12.21 0.45 a nemo 201.23 13.99 585.25 24.52 410.22 14.99 zoo 1219.23 84.77 1783.39 74.71 2282.12 83.37 undet 12.45 0.87 15.21 0.64 32.63 1.19 2005 2006 a uto 9.59 3.88 2.93 0.91 2.09 0.55 a nemo 1.83 0.74 4.33 1.34 0.05 0.01 zoo 234.10 94.74 314.32 97.42 378.74 98.90 undet 1.57 0.64 1.07 0.33 2.08 0.54 2006 2007 a uto 8.16 0.77 3.31 0.24 1.92 0.13 a nemo 8.49 0.81 25.25 1.84 27.40 1.83 zoo 1032.06 97. 95 1339.91 97.45 1457.98 97.59 undet 4.97 0.47 6.54 0.48 6.61 0.44 2007 2008 a uto 13.05 2.34 12.56 1.99 8.77 1.02 a nemo 38.39 6.88 111.07 17.60 69.11 8.06 zoo 486.72 87.21 506.30 80.22 753.80 87.92 undet 19.96 3.58 1.18 0.19 25.70 3.0 0 2008 2009 a uto 33.11 1.79 17.58 2.21 5.42 0.29 a nemo 189.25 10.21 122.50 15.41 108.83 5.83 zoo 1561.64 84.22 651.33 81.92 1750.41 93.83 undet 70.22 3.79 3.70 0.47 0.90 0.05 (Dispersal mechanism: auto=autochoric, anemo=anemochoric, zoo= zoochoric, and undet=undetermined.)

PAGE 56

56 Figure 5 1. Boxplots of number of species in fruit traps. A) Number of species per trap per treatment, all years pooled. B) Yearly number of species per trap, treatments pooled. The boxplot represents the 25 th 50 th and 75 th percentile, whiskers are the first and third quartiles +/ A B

PAGE 57

57 Figure 5 2 Number of species per month, treatment, and year A ) Total monthly number of species per treatment during the five year study period ( s haded area indicates rain season ). B) B oxplots of the n umber of species found per trap per year and treatment. The boxplot represents the 25 th 50 th and 75 th percentile, whiskers are the first and third quartiles +/ A B

PAGE 58

58 Table 5 2 Wald tests for terms in model of lowest AIC for number of spe cies in fruit fall. numDF denDF F value p value (Intercept) 1 994 2407.295 <0.0001 Treatment 2 994 3.7494 0.0239 Density 1 994 181.5677 <0.0001 Distance 1 994 9.3542 0.0023 year 4 994 100.5062 <0.0001 Treatment x Density 2 994 4.4186 0.0123 Treatm ent x Distance 2 246 2.339 0.0986 Density:year 4 994 2.6209 0.0336 Distance:year 4 994 6.3576 <0.0001 Table 5 3 Wald test for terms in the final, most parsimonious, statistical model of number of species in fruit fall. numDF denDF F value p value (Intercept) 1 994 2383.22 <0.0001 Treatment 2 994 3.6573 0.0261 Density 1 994 180.29 <0.0001 Distance 1 994 12.7393 0.0004 Y ear 4 994 99.7178 <0.0001 Treatment x Density 2 994 4.3789 0.0128 Density:year 4 994 2.6299 0.0331 Distance:year 4 994 6.3399 <0.0001 Table 5 4 Summary for the model selected to further analysis of the number of species in fruit fall Value Std.Error DF t value p value (Intercept) 3.919191 0.481291 994 8.143086 0 TreatmentB2 0.84614 0.410483 994 2.06132 0.0395 Treatme ntB4 0.16809 0.444622 994 0.37806 0.7055 Density 0.049871 0.019416 994 2.56856 0.0104 Distance 0.002193 0.000581 994 3.774992 0.0002 2005 2006 2.12446 0.440284 994 4.82519 0 2006 2007 0.91692 0.447371 994 2.04956 0.0407 2007 2008 1.60958 0.4340 99 994 3.70787 0.0002 2008 2009 1.50358 0.449903 994 3.342 0.0009 TreatmentB2 x Density 0.044786 0.017191 994 2.60518 0.0093 TreatmentB4 x Density 0.020641 0.018349 994 1.124935 0.2609 Density x 2005 2006 0.02777 0.01764 994 1.57417 0.1158 Densit y x 2006 2007 0.02694 0.018446 994 1.46036 0.1445 Density x 2007 2008 0.01844 0.019521 994 0.94471 0.345 Density x 2008 2009 0.024084 0.020545 994 1.172273 0.2414 Distance x 2005 2006 0.00115 0.000654 994 1.7528 0.0799 Distance x 2006 2007 0.001 06 0.000683 994 1.54505 0.1227 Distance x 2007 2008 0.000511 0.000741 994 0.689233 0.4908 Distance x 2008 2009 0.00277 0.00078 994 3.55421 0.0004

PAGE 59

59 Figure 5 3 Least square means values for the number of species in fruit f all tre atment x density interaction. All other effects in the model are at their average values Figure 5 4 Least square means values for the number of species in fruit fall year x d ensity interaction All other effects in th e model are at their average values

PAGE 60

60 Figure 5 5 Least square means values for the number of species in fruit fall year x d istance interaction. All other effects in the model are at their average values

PAGE 61

61 CHAPTER 6 DISCUSSION Fire and Edge Effects on Forest Structure Burning caused reductions in tree and liana density, BA, and diversity. The two f ire treatments differed in their impacts with generally larger decrease in all variables in B2 than in B4 b y the last yea r of this study. A more pronounced decrease was observed after the 2007 experimental fire in both B2 and B4 This effect might be at tributable to the fact that 2007 was a dry year (Paulo Brando pers. comm.) Droughts increase the risk of unde rstory fires i n the Amazon (Nepstad et al. 2004 ), but also increase fire intensity and severity Thus the high severity of the 2007 fire may be in part due to the dryness of that year. The observed reduction s i n density and BA after repeated burns are in accord with oth er studies reporting on the effects of recurring fires. Balch et al. (2008) noted that short fire return intervals (FRI; as in B4, ~1 year), are associated with decre ased area burned, lower flame heights and due to a lack of fuels. The short FRI in B4 app ears to have buffered trees and lianas from fire induced mortality which resulted in less fire impact on tree and liana d ensity BA and d iversity than i n B2. In contrast, less frequent fires in B2 allowed litter fall and debris of fire killed tree s to ac cumulate b etween the first and second fire. F uel accumulation together with dry weather converged to cause a more dramatic reduction in density, BA, and diversity i n B2 after the 2007 fire Fire and proximity to the forest agriculture edge interact ed to influence mortality. In other words, there was an effect of distance from edge on d ensity, BA, and diversity that was influenced by fire. Burning, particularly in B2, caused decreases i n the intercepts and increases in slopes of the relationship between de nsity, BA, and d iversity

PAGE 62

62 with distance from the edge after repeated fires. The decrease s in intercept s indicate that there was an overall reduction i n the average values of these variables due to fire at the forest agriculture edge. I ncrease in slopes afte r repeated burns indicate that there was there was a differential reduction in density, BA, and diversity closer to the edge than further into the treatment plot s Fire Treatments Did Affect Fruit fall Mass b ut at Long Distance from the Edge Contrary to o ur expectations, fire did cause a decline on fruit fall mass Rather, as evidenced by our results there was no apparent divergence in fruit fall mass by year and treatment. Treatment B4 actually had a higher fruit mass at larger distances from the forest e dge than the control treatment and B2 and B0 did not differ statistically. It is worth considering that when the analysis was run with only the interior traps (i.e. those at least 100 m from any edge) the treatment x distance interaction was significant an d revealed that B4 had a higher fruit mass falling in tis traps than B0. Additionally when the variable distance was removed from the statistical analysis treatment B4 did present higher fruit mass than B0 while B0 and B2 did not differ in this regards Th is emerged despite considerable decrease s in tree and liana densities and basal area s which would be expected to result in fruit production declines (Barlow and Peres 2006a, b ). Even t hough after the 2007 fire tree and liana d ensity and BA decrease d marke dly and to comparable levels as those rep orted by Barlow and Peres (2006b ), still there were no decreases in fruit fall mass during two years after that fire (2008 2009) Barlow and Peres (2006b) observed that fewer trees were in fruit in burned forests wi th numbers further decreasing in twice burned forests. If this response held in our forest, then our results are enigmatic

PAGE 63

63 Greater Influx of Fruits in to Burned Plots One potential explanation for the patterns of fruit fall mass observed in this study is i ncreased use of burned forest by seed dispersing animals that deliver fruits from outside the burned plots. Few studies have addressed the effects of tropical forest fires on wildlife thus we both draw examples from fire studies as well as from logging stu dies. Barlow et al. (2002) observed a decline in overall bird species richness and abundance after fires in Amazon forests ; in particular this effect was more prominent for habitat specialists, rare, and disturbance sensitive species In contrast, t hese au thors observed increases in nectar and seed eaters. Barlow and Peres (2004) studied bird responses one and three years after an understory Amazon fire and reported that substantially more individuals and species were captured three years after the fire. Th ese authors also report ed a marked increase in frugivorous birds, which was correlated with the number of small woody stems (dbh<10cm). T hey also report an increase in abundance of second growth forest bird species which they relate d to the open forest ca nopy that resulted from burning. Barlow and Pe res (2004) relate their finding in burned forest with those of studies of avifaunal responses in logge d forests, concluding that the responses are similar. In another study, Lambert (1992) found significant in creases in the abundance of nectarivores insectivore frugivore birds in selectively logged forests wh en compared to unlogged forest in Borneo In Guiana Thiollay (1997), found overall canopy bird abundance to be unaltered in logged forest, with increase s i n species associated with forest edges, dense second growth, and large gaps. These studies suggest increased use of b urned and logged forests by fru givorous birds, which support the idea that b irds came into the burned plots from undisturbed areas and deli vered seeds Conver sely, if

PAGE 64

64 frugivorous bird abundance increase s following burning or logging it likely do es so because of shif ts in availability of resources utilize d by this foraging guild Increased fru givore abundance may be a reflection of increased f ruit availability in disturbed forests. Barlow and Peres (200 4) for example related the number of fruit and nectar eating bird s to the number of small woody stems and heliconia stems both of which produce resources for these foraging guilds. In addition B arlow and Peres (2006b) report that smaller trees 10 20 and 30 40 cm dbh bore fruit more often than expected in burned forests Increased Availability of Resources in Burned Forests Leads to Increased Reproductive Output Survivors Fire, like other disturba nces, influence s the availability of resources, by opening space that o therwise would be occupied (Bond and Wilgen 1996 ). Surviving individuals enjoy greater resource availability that can be allocated to reproduction. Few studies have dealt with the effec ts of disturbances on tropical forest tree and liana reproductive output in part icular in relation to community wide responses. The shear number of species is daunting (our study found 80 species in fruit fall traps from a total of 97 tree and liana speci es). The few studies that tackled the response of one or a few species to disturbance and may provide a starting point to argue in relation to results observed on the present thesis. In addition, we also draw on information from temperate, sub tropical, an d tropical forest management examples. Logging is comparable to fire insofar as it removes standing biomass (but through a very different selection process ) In any case, surviving individuals enjoy increased availability of resources and display what is

PAGE 65

65 can be both for enhancing growth of remaining trees in a stand as well as to enhance reproductive output (i.e. increase seed production). In experiments with loblolly pine ( Pinus taeda ) Wenger (1954) found tha t released individuals consistently produced more cones and markedly increased three years after the treatment. Karlsson (2000) working with Scots pine ( Pinus sylvestris ) in Sweden report ed a similar trend for released tre es: more cones were produced by re leased trees and he observed a five fold increase in cone production 4 to 5 years after the release In a study of three species of oaks in managed forests Barik et al. (1996) reported that in two, Lithocarpus dealbatus and Schima khasiana acorn producti on increased with increasing levels of disturbance. The third species, Quercus griffthii did not show increases in acorn numbers but produced heavier seeds in disturbed sites. In logged Costa Rican montane forests Guariguata and Saenz (2002) estimated ac orn production at the stand level. They report that the number of acorns of Quercus costaricensis that fell per unit of ground area was highest after the highest cutting level (30% basal area removal). At 20% basal area removal they report that acorn numbe rs were marginally higher than within un logged forest stands. These authors attributed their results in part to an increase in the number of trees fruiting earlier than expected based on their size classes. While these studi es show that individual trees increase fruit production after disturbance s such as logging, it is not possible to discern whether this effect is scalable to the community level In addition some studies report the opposite pattern Ghazoul et al. (1998), studying the Dipterocarpaceae Shorea siamensis found that fruit set was lower per tree than in disturbed sites (disturbed sites had 22 trees dbh>10 cm per ha,

PAGE 66

66 undisturbed 205 trees/ha). These authors attributed the decrease in fruit set to the decrease in S. siamensis density resultin g in increasing distance between individuals. The increased distances between individual trees decreased pollinator visitation causing bees to remain longer periods foraging on fewer tree individuals and consequently not favoring pollen transfer. The effe cts of fire on reproductive output of plants are also reportedly mixed. In fire prone ecosystems fire is reported to stimulate flowering in several species (Whelan 1995, p. 87). The mechanism by which fire stimulates flowering is not understood but may be related to i ncreased post fire resource availability (Whelan 1995). Whelan also suggests that fire damage and pruning might also stimulate flower primordia initiation Though increased flowering has been reported in burned temperate and sub tropical fores t ecosystems it is not possible to ascertain whether tropical forests are stimulated in a similar manor as no study has spec ifically dealt with this issue In neotropical savannas the effects of fire on tree reproduction are also mixed. Hoffmann (1998 ) re ported that only one out of the six species studied showed marked post fire increases in fruit production. H e concluded that frequent burning had a negative effect on sexual reproduction. Because the only studies of fruiting in burned tropical forests are those of Barlow and Peres (2006 a and b) it is not clear what would be reasons for the patterns reported in the present thesis. Barlow and Peres (2004 and 2006b) did observe that smaller stems stems dbh <10 cm fruited more often in burned than in unburned forests. This observation conforms to the above mentioned studies in logged forests and may partially account for the lack of differences in fruit fall mass between treatment B2 and the control and the higher fruit fall mass in B4 than in B0. In ad dition t o increased

PAGE 67

67 fruiting by small individuals, larger individuals may also increase their fruit production after fire s If such patterns of fruit fall mass are a result of increased fruit production by surviving individuals, an interaction between tree densi ty and treatment would be expected but not observed That said, this study was not designed to assess fruit production by surviving individuals Chapman et al. (1994) argu e that while fruit fall traps are a suitable method to conduct comparisons of habitat s and different study areas, data in fruit fall traps cannot be equa ted to fruit production since they measure fruit fall and not fruit production. Thus to take the next step and relate fruit fall mass to changes in density of surrounding individuals migh t not yield results because the content of tr aps poorly relates to fruit production by the surrounding trees. Fire and Edge Did Interact to Influence Fruit fall Our results indicate a significant interaction between treatment and distance from agricultur al edge for B4. This interaction was expected on the basis of that tree mortality due to fire would be comparatively greater closer to forest edges than further into forest patches. Contrary to our expectations, however, fruit fall mass was higher further into the treatment plot in B4 than in the control. As mentioned above B2 and B0 did not differ in this regard, neither did B2 and B4. This interaction might have emerged given that the density of trees and lianas tended to increase with increasing distance from the edge and because in B4 the decline in tree and liana density was not very pronounced as in B2 The decline however might have been just enough to cause an increase in production in that treatment. In B2, on the other hand, the observed decrease i n tree and liana density might have been too great to cause an increase in fruit fall mass to levels higher than those of B0. Despite the significance of the interaction it is difficult to

PAGE 68

68 ascertain causation to fire. Since fruit fall data was only initiat ed after the fire treatments it is not possible to determine whether the observed was already in place before burning activities began. In ad dition, following the fact that fire kill s more trees in the edge than in the forest interior, we would also expect to observe the distance x treatment interaction for B2. Our results indicate that fire did influenc e the relationship between the d ensity of trees and lianas and distance from edge on B2, this influence is noted as a decrease in the model intercept with s ubsequent fires and an increase in regression slopes. The increased regression slopes indicate that a higher proportion of trees and lianas were killed closer to the edge in comparison to areas further into the forest plots. The assumption that fire induce d mortality will cause a decrease in fruit fall mass is not supported rather as observed in this study and corroborated by previous studies reduction in tree density have an opposite effect than that expected. Fire Did Affect Fruit fall Species Richness The number of species in fruit fall traps was higher in B2 at higher tree and liana densities than in the control. Treatments B4 and B0 did not differ as well as B4 and B2 in this regards. These results were un expected as the diversity of trees and lianas declined in both fire treatments. In particular both diversity and density of lianas decreased markedly in B2. One explanation to the observed pattern revealed by the statistical analysis is that in B2 fire might have induced earlier fruiting by some spec ies and/or more continuous fruiting by others. The consequences of this would be that in B2 some species would be fruiting while in the other plots the same species would not be. Note that the larger number of species observed does not entail a greater mas s of fruits falling into the traps. In the analysis of species richness in fruit fall if the mass value that

PAGE 69

69 each species contributed was not determinant of them being considered to be considered to be present or not (it was enough to have fruit mass > 0 g to be included). Implications for Regeneration of Burned Tr opical Forests Assuming that the number of seeds of a given species is related to the mass of its fruits, we can argue that number of seeds reaching the forest floor of the frequently burned tre atment (B4) was greater than the control while for the twice burned forest (B2) burned it did not differ from the unburned control. In addition, t he number of species did not decline (actually increased in B2) due to fire, which suggests that not only the availability of propagules for regeneration may actually increase or r emain unchanged but that the number of species that could potentially regenerate in the post fire environment remains constant or increases Based on the findings of (Norden et al. 2007) in French Guiana, the numbers and identities of the seedlings germinating in a forest site ar e strongly related to variation in seed fall. This relationship suggests (given the assumption that fruit fall mass is related to seed fall quantities) that burne d forest regeneration is not limited by seed availability. It remains to be determined whether fire induced changes in the species composition of the fruit fall. Changes in species composition of seed fall would affect forest regeneration and future compos ition of burned forest s community. I t is also not yet c lear how different species respond to post fire environment s; some species may flourish under the new conditions while others languish The prospects for increased incidence of forest fires in the Amaz on and the positive fire feedbacks under certain fire return intervals may restrict regenera tion in burned tropical forests Thus even if the potential for forest regeneration as indicated by seed input, remains unaltered by fire repeated fires will like ly hamper regeneration both

PAGE 70

70 by killing seedlings and damaging seed banks (Kennard et al. 2002 ). Without protection of burned forest s from repeated fires the prospect s for forest to recover y remain low. In addition repeated burns are likely to lead to a l evel of fruit tree loss that inevitably will cause fruit fall levels to decrease in burned forests. In environments where fire s are common, vegetative reproduction may be favored over sexual reproduction in ensuring plant survival (Hoffmann 1998).

PAGE 71

71 CHAPTER 7 CONCLUSION This thesis reports results on the effects of repeated fire on fruit fall patterns in a transitional forest on the southern edge of the Amazon Basin. Our results indicate that five year after the initiation of the experimental fires fruit fall patterns were altered by fire, but results were in contrast with our expectations. Fruit fall mass showed an increase with increasing distance from the edge in treatment B4 and fruit fall mass did not differ among treatments B2 and B0 (co ntrol) The number of species in fruit fall was greater at higher tree and liana densities in B2. Our study is the first to addre ss the issue of fruit fall in burned tropical forest. The temporal and spatial extent of this experiment ensures that the res ults bear some inferential strength related to the effects of burning o n fruit fall. O ur results indicate that regeneration in burned forests might not be limited by a lack of seed input. This realization warrants further studies on the limitations to plan t regeneration in burned f orests and ways to foster forest recovery. Finally, al though we have shed some light into an important process in the ecolog y of tropical forest m any questions remain unanswered and many more has risen from the results we present ed.

PAGE 72

72 APPENDIX A VARIABLE CORRELATION S, STATIS TICAL MODELS, MODEL SELECTION, AND DIAGN OSTICS Figure A 1. Distribution of fruit fall mass (log10(mass+1)) by tree and liana densities in trap centered plots. A) All treatments and years pooled (R = 0.28, p <0.001; rho=0.3, p<0.001). B) Grouped by treatment and year.

PAGE 73

73 Figure A 2. Distribution of fruit fall mass (log10(mass+1)) by live basal area (m 2 ) in trap centered plots. A) All treatments and years pooled (R = 0.24, p<0.001; rho = 0.27, p<0.001). B) Gro uped by treatment and year.

PAGE 74

74 Figure A 3. Distribution of log transformed fruit fall mass (log10(weight+1)) by distance from the edge (m). A) All years and treatments pooled (R = 0.19, p<0.001; rho = 0.29, p<0.001). B) Grouped by treatment and year.

PAGE 75

75 T able A 1. Correlation between fruit fall and tree and liana density by year and treatment (r ) Year 2004 2005 2005 2006 2006 2007 2007 2008 2008 2009 Treatment r r ho r rho r rho r rho r rho B0 0.26* 0.28** 0.18 0. 2 0.23* 0. 23* 0.27* 0. 28** 0.20 0. 2 B2 0.20 0.18 0.07 0.13 0.29** 0.3** 0.55** 0.52** 0.39** 0.44** B4 0.23* 0.25* 0.35** 0.39** 0.38** 0.42** 0.36** 0.38** 0.28* 0.31** Table A 2. Correlation between fru it fall and tree and liana basal area (m 2 ) by year and treatment (r ) Year 2004 2005 2005 2006 2006 2007 2007 2008 2008 2009 Treatment r r ho r rho r rho r rho r rho B0 0.34** 0.32** 0.18 0. 21 0.36** 0. 4** 0.28** 0. 28** 0.19 0. 2 B2 0.15 0.12 0.06 0.12 0.29** 0.31** 0.45* 0.44** 0.33** 0.37** B4 0.11 0.07 0.24* 0.33** 0.33** 0.39** 0.26* 0.28** 0.22* 0.26* Table A 3. Correlation between fruit fall and distance from f orest agriculture edge (m) b y year and treatment (r ) Year 2004 2005 2005 2006 2006 2007 2007 2008 2008 2009 Treatment r r ho r rho r rho r rho r rho B0 0.27** 0.43** 0 .07 0. 18 0.14 0. 32** 0.01 0. 23* 0.1 0. 19 B2 0.22* 0.26* 0.1 0.25* 0.15 0.32** 0.44* 0.49** 0.12 0.26** B4 0.33** 0.29** 0.23* 0.37** 0.34** 0.47** 0.48** 0.44** 0.12 0.23*

PAGE 76

76 Figure A 4. Distribution number of species per trap and tree and distance fr om edge. A) All treatments and years pooled (R = 0.25, p<0.01; rho = 0.34 p<0.01). B) Grouped by treatment and year.

PAGE 77

77 Figure A 5. Distribution number of species per trap and tree and liana density in trap centered plots. A) All treatments and years poo led (R = 0.38, p<0.01; rho = 0.39 p<0.01). B) Grouped by treatment and year.

PAGE 78

78 Figure A 6. Distribution number of species per trap and tree and liana basal area in trap centered plots. A) All treatments and years pooled (R = 0.31, p<0.01; rho = 0.33 p<0 .01). B) Grouped by treatment and year.

PAGE 79

79 Figure A 7. A) All treatments and years pooled (R = 0.26, p<0.01; rho = 0.26 p<0.01). B) Grouped by treatment and year.

PAGE 80

80 Table A 4. Correlation between fruit fall species richness and distance from fore st edge by year and treatment (r ) Year 2004 2005 2005 2006 2006 2 007 2007 2008 2008 2009 Treatment r rho r rho r rho r rho r rho B0 0.34** 0.48** 0.12 0. 24* 0.08 0. 24* 0.29** 0. 5** 0.005 0. 04 B2 0.23* 0.35** 0.3** 0.43** 0.24* 0.36** 0.66** 0.68** 0.4** 0.4** B4 0.36** 0.43** 0.26* 0.34** 0.39** 0.48** 0.55** 0.52 ** 0.09 0.24* Table A 5. Correlation between fruit fall species richness and tree and liana density by year and treatment (r ) Year 2004 2005 2005 2006 2006 2007 2007 2008 2008 2009 Treatment r rho r rho r rho r rho r rho B0 0.38** 0.38** 0.26* 0. 3* 0.22* 0. 18 0.33** 0. 4** 0.18 0. 14 B2 0.25* 0.28* 0.27* 0.27* 0.35** 0.38** 0.65** 0.61** 0.58** 0.55** B4 0.45** 0.48** 0.32** 0.36** 0.23* 0.16 0.43** 0.45** 0.24* 0.2 6* Table A 6. Correlation between fruit fall species richness and tree and liana bas al area by year and treatment (r ) Year 2004 2005 2005 2006 2006 2007 2007 2008 200 8 2009 Treatment r rho r rho r rho r rho r rho B0 0.35** 0.39** 0.27* 0. 32** 0.36** 0. 33** 0.35** 0. 41** 0.07 0. 05 B2 0.19 0.2 0.19 0.2 0.28* 0.3** 0.56** 0.54** 0.48** 0.45** B4 0.29** 0.34** 0.21 0.27* 0.26* 0.24* 0.34** 0.37** 0.19 0.25* Table A 7. Correlation between fruit fall species richness and tree and liana di versity by year and treatment (r ) Year 2004 2005 2005 2006 2006 2007 2007 2008 2008 2009 Treatm ent r rho r rho r rho r rho r rho B0 0.1 0.05 0.07 0. 06 0.09 0. 12 0.1 0. 1 0.15 0. 13 B2 0.17 0.14 0.22 0.27* 0.39** 0.39** 0.36** 0.42** 0.39** 0.45** B4 0.3** 0.39** 0.1 0.17 0.08 0.09 0.34** 0.34** 0.2 0.21

PAGE 81

81 Figure A 8 Model residuals by fitted v alues and explanatory variables. Plot titles indicate variables used.

PAGE 82

82 Figure A 9 Number of species model residuals by fitted values and explanatory variables. Plot titles indicate variables used.

PAGE 83

83 LIST OF REFERENCES Balch, J. K, D. C. N epstad, P. M. Brando, L. M. Curran, O. Portela, O. Carvalho, P. Lefebvre. 2008. Global Change Biology 14: 2276 2287. Balch, J. K., D. C. Nepstad, and L. M. Curran. 2009. Patterns and process: fire initiated grass invasion at Amazon transitional forest edge s. Pages 481 502 in M.A. Cochrane, editor. Tropical fire ecology: climate change, land use, and ecosystem dynamics. Springer, Berlin, Germany. Balch, J. K., D. C. Nepstad, L. M. Curran, P. M. Brando, O. Portela, P. Guilherme, J. C. Reuning Scherer, and O Carvalho. 2011. Size, species, and fire behavior predict tree and liana mortality from experimental burns in the Brazilian Amazon. Forest Ecology and Management 261: 68 77. Barik, S. K., R. S. Tripathi, H. N. Pandey, and P. Rao. 1996. Tree Regeneration i n a subtropical humid forest: effect of cultural disturbance on seed production, dispersal and germination. Journal of Applied Ecology 33: 1551 1560. Barlow, J., T. Haugaasen, and C. A. Peres. 2002. Effects of ground fires on understory bird assemblages in Amazonian forests. Biological Conservation 105: 157 169. Barlow, J. and C. A. Peres. 2004. Avifaunal responses to single and recurrent wildfires in Amazonian forests. Ecological Applications 14: 1358 1373. Barlow, J. and C. A. Peres. 2006 a Consequences o f cryptic and recurrent fire disturbances for ecosystem structure and biodiversity in Amazonian forests. Pages 225 240 in W.F.Laurance and C.A. Peres, editors. Emerging threats to tropical forests. The University of Chicago Press, Chicago, Illinois, USA. Barlow, J. and C. A. Peres. 2006 b Effects of single and recurrent wildfires on fruit production and large vertebrates abundance in a central Amazonian forest. Biodiversity and Conservation 15: 985 1012. Barlow, J. and C. A. Peres. 2008. Fire mediated die back and compositional cascade in a Amazonian forest. Philosophical Transactions of the Royal Society B 363: 1787 1794. Bates, D., and M. Maechler. 2010. lme4: Linear mixed effects models using S4 classes. R package version 0.999375 37. http://CRAN.R proje ct. Org/package=lme4. Bond, W. J. and B. W. van Wilgen. 1996. Fire and plants. Chapman and Hall, London, UK. Chapman, C. A., R. Wrangham, and L. J. Chapman. 1994. Indices of habitat wide fruit abundance in tropical forest. Biotropica 26: 160 171.

PAGE 84

84 Chazdon, R. L. 2003. Tropical forest recovery: legacies of human impact and natural disturbances. Perspectives in Plant Ecology, Evolution and Systematics 6: 51 71. Cochrane, M. A. 2003. Fire science for rainforests. Nature 421: 913 919 Cochrane, M. A., A. Alencar, M. D. Schulze, C. M. Souza, D. C. Nepstad, P. Lefebvre, E. A. Davidson. 1999. Positive feedbacks in the fire dynamics of closed canopy tropical forests. Science 284: 1832 1835. Ghazoul, J., K. A. Liston, and T. J. B. Boyle. 1998. Disturbance induced densi ty dependent seed set in Shorea siamensis (Dipterocarpaceae), a tropical forest tree. Journal of Ecology 86: 462 473. Godoy, J. A. and P. Jordano. 2001. Seed dispersal by animals: exact identification of source trees with endocarp DNA microsatellites. Mole cular Ecology 10: 2275 2283. Greene, D. F., C. Messier, H. Asselin, and M. Fortin. 2002.The effect of light availability and basal area on cone production in Abies balsamea and Picea gluaca Canadian Journal of Botany 80: 370 377. Guariguata, M. R. and G. P. Saenz. 2002. Post logging acorn production and oak regeneration in a tropical montane forest, Costa Rica. Forest Ecology and Management 167: 285 293. Hoffmann, W. A. 1998. Post burn reproduction of woody plants in a neotropical savanna: the relative imp ortance of sexual and vegetative reproduction. Journal of Applied Ecology 35: 422 433. Hojsgaard, S., U. Halekoh, J. Robinson Cox, K. Wright, and A. A. Leidi. 2010. doBy: Groupwise summary statistics, general linear contrasts, LSMEANS (least squares means ), and other utilities. R package version 4.1.2. URL http://CRAN.R project.org/package=doBy. Holdsworth, A. R. and C. Uhl. 1997. Fire in Amazonian selectively logged rain forest and the potential for fire reduction. Ecological Applications 7: 713 725. Hulb ert, S. H. 1984 Pseudoreplication and the design of ecological field experiments. Ecological Monographs 54: 187 211. Karlsson, C. 2000. Seed production of Pinus sylvestris after release cutting. Canadian Journal of Forest Research 30: 982 989. Kauffman, J. B. 1990. Survival by sprouting following fire in tropical forests of the eastern Amazon. Biotropica 23: 219 224.

PAGE 85

85 Kennard, D. K., K. Gould, F. E. Putz, T. S. Fredericksen, and F. Morales. 2002. Effect of disturbance intensity on regeneration mechanisms in a tropical dry forest. Forest Ecology and Management 162: 197 208. Lambert, F. R. 1992. The consequences of selectively logging for Bornean lowland forest birds. Philosophical Transactions of the Royal Society B 335: 443 457. Laurance, W. F. 2006. Fragment s and fire: alarming synergisms among forest disturbance, local climate change, and burning in the Amazon. Pages 87 103 in W. F.Laurance and C. A. Peres, editors. Emerging threats to tropical forests. The University of Chicago Press, Chicago, Illinois, U SA. Solymos, M.H.H. Stevens, and H. Wagner. 2011. vegan: Community ecology package. R package version 1.17 6. URL http://CRAN.R project.org/package=vegan. Oksanen, L. 2001. Logic of experiments in ecology: is pseudoreplication a pseudoissue? Oikos 94: 27 38. Norden,N., J. Chave, A. Caubere, P. Chatelet, N. Ferroni, P. Forget, and C. Thebaud. 2007. Is temporal variation of seedling communities determined by environment or by seed arrival? A test in a neotropical forest. Journal of Ecology 95: 507 516. Peres, C. A., and M. van Roosmalen. 2002. Primate frugivory in two species rich neotropical forests: implications for the demography of large seeded plants in overhunted areas. P ages 407 421 in D. J. Levey, W. R. Silva, and M. Galetti, editors. Seed dispersal and frugivory: ecology, evolution, and conservation. CABI publishing, New York, New York, USA. Pinheiro, J., D. Bates, S. Debroy, D. Sarkar, and the R Development Core Team. 2010. nlme: Linear and nonlinear mixed effects models. R package version 3.1 97. R Development Core Team. 2010. R: A language and environment for statistical computing. R foundation for statistical computing, Vienna, Austria. URL http://www.R project.org/ Ray, D., D. Nepstad, and P. Moutinho. 2005. Micrometeorological and canopy controls of fire susceptibility in a forested Amazon landscape. Ecological Applications 15: 1664 1678. Sanford, R. L., J. Saldarriaga, K. Clark, C. Uhl, and R. Herrera. 1985. Amaz on rainforest fires. Science 227: 53 55. Sarkar, D. 2008. Lattice: Multivariate data visualization with R. Springer, New York, USA.

PAGE 86

86 Thiollay, J. 1997. Disturbance, selective logging, and bird diversity: a neotropical forest study. Biodiversity and Conserva tion 6: 1155 1173. Uhl, C., K. Clark, H. Clark, and P. Murphy. 1981. Early plant succession after cutting and burning in the upper Rio Negro region of the Amazon basin. Journal of Ecology 69: 631 649. Uhl, C. and J.B. Kauffman. 1990. Deforestation, fire su sceptibility, and potential tree responses to fire in the eastern Amazon. Ecology 71: 437 449. Veldman, J. W., B. Mostacedo, M. Pena Claros, and F. E. Putz. 2009. Selective logging and fire as drivers of alien grass invasion in a Bolivian tropical dry forest. Forest Ecology and Management 258: 1643 1649. Wenger, K. F. 1954. The stimulation of loblolly pine seed trees by preharvest release. : 115 118. Whelan, R. J. 1995. The ecology of fire. Cambridge University Press, Cambridge, UK. Wickham, H. 2009. ggplot2: Elegant graphics for data analysis. Springer, New York, USA.

PAGE 87

87 BIOGRAPHICAL SKETCH Eric Oliveira Carvalho was born in Athens, Ohio and was raised in Salvador Bahia, Brazil. Early in life he experienced a diversity of tropical ecosystems raging from the rain forests of the Atlantic coast to the dry lands to the north and interior of t he state. His interest in science was influenced by the scientific endeavor s of his father on the field of geology and meteorite science In 2000 Eric ingressed in the Universidade Federal de Vicosa in Minas Gerais, Brazil, to pursue a degree in Forest Engineering remaining in that institution until 2004. While at UFV he was expose d to research through the program for sc ientific initiation of the CNPq; he investigated the insects visiting flowers of Bignoniaceae in a forest fragment. In 2003 Eric was accepted as an exchange student between the UFV and the University of Florida and in 2004 transferred to UFL School of Forest Resources and Conservation from which he graduat ed in 2006 with a major in Natural Resource Conservation and minor in Wildlife Ecology and Conservatio n. During his undergraduate studies Eric worked on projects i n the Atlantic Forest of Northeastern Brazil, on b ee s and forest ecology. Upon graduation he accepted a position of Fore ster with the US Forest Service a t the Hiawatha National Forest o s aspects of forest management including timber marking, inventory, and timber sale preparation. It was during that time that his interest in fire ecology and management incr eased by working in iconic fire dependant ecosystems such as Jack pine forests, tr aining in fire management activities degree in Forest Ecology and Management with a concentration in fire ecology.