1 FACTORS THAT DETERMINE PATTERNS OF SEEDLING RECRUITMENT IN AN AFROTROPICAL FOREST By CONNIE JANE CLARK A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORI DA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2009
2 2009 Connie Jane Clark
3 To my parents who encouraged my curiosity, To my husband with whom I explore, To our son Â– our greatest adventure begins
4 ACKNOWLEDGMENTS I thank the government of C ongo (particularly the Ministry of Forestry Economy and the Ministry of Scientific Resear ch), the Wildlife Conservation Society (WCS), and Congolaise Industrielle des Bois (CIB) for th eir collaboration and support. In particular, I thank B. Curran, J. Mokoko, P. Elkan, P. Telfer, H. Thomas, O. Desmet, D. Paget, J.-M. Mevellec, L. Vander Walt, P. Kama, J-C Dengui and P. Ngouemb. The large scale nature of this work would ha ve been impossible wit hout logistical support from J. Beck, M. Gately, C. Prevost, A. Niamazoc k, C. Assobam, and R. Aleba. I owe a debt of gratitude for the tireless work of my field team. Special thanks to team leaders: J. Poulsen, V. Medjibe, O. Mbani, Y. Nganaga, G. Modouka, and F. Etono. Thanks also to U.Sabo, I. Loungoumba, Ekoume, Simba, Mbe, Iyena, B. Kimbembe, C. Makoumbou, P. Ipete, M. Moke, E. Elenga, I. Loungouba, F. Adouma, G. Abeya, J. Lamba, R. Bokoba F. Iyenguet B. Modzoke, and the Bomassa guides. Thanks to the villager s of Kabo for scouring the forests for seeds and assuring the successful completion of my seed addition experiments. Botanical work conducted for this project was co mpleted by D. J. Harris, A. Wortley and J. M. Moutsembot. V. Medjibe deserves special th anks for his assistance with vegetation plots. R. Mylavarapu provided significant assistance with the soil anal ysis and interpretation. The Levey, Holt and SOB labs provided me with feedb ack and discussion at various stages of the study. John Poulsen and Ricco Holdo a ssisted with statistical analysis. Financial support was generously provided by an SNRE alumni fellowship, EPA STAR fellowship (#91630801-0), NSF dissertation impr ovement grant (#00074232 ), the Madelyn Lockhart Dissertation Fellowship and an Am erican Association of University Women Dissertation Fellowship. Fiel d work was supported by two USFWS Great Ape Conservation
5 Grants to C.J. Clark and J.R. Poulsen, the W ildlife Conservation Society, and several donors who generously support WCS research in norther n Congo (ITTO, CARPE, USFWS, LCAOF, BCTF, and others). Chapter I of this dissertation was initiated as part of the Quantitative Methods and Ecological Inference course at the University of Florida. I tha nk the following course participants for assistance with the literature search, data extraction, and feedback regarding data analysis: D. Blondel, N. Brennan, H. Klug, J. Martin, M. McCoy, M. Mota and N. Seavy. Numerous colleagues provided key data from published and unpublished studies, which often required them to resurrect retired datasets. This chapter was publishe d 2007 by The University of Chicago. I thank my committee members, Drs. Doug Levey, Bob Holt, Ben Bolker, Kaorou Kitajima, and Scott Robinson for ch allenging me to be a better scie ntist every step of the way. Doug taught me the art of experimentation. B ob provided me with a cu shy lab space and access to a parade of great minds. Ben spent hours wo rking through complicated data sets. Kaoru taught me to see the forest floor from a different perspective and helped me stay abreast of the literature. Finally, thank you Scott Robinson for st epping up to the plate to help push this work over the finish line. It has been a pleasure working with you all. Thanks also to my friends and family for being patient with the holidays, births and birthdays we missed as we pursued our studies over the years. From now on, we will be there. Colette St. Mary, Todd Palmer, Rico Holdo, and Dan and Hilary Zarin provided encouragement, distraction, and assured we held it together when thi ngs got a little overwhelming toward the end of this dissertation.
6 Most importantly, I am grateful for the suppor t, backstopping and pa tience of my husband and closest colleague, John Poulsen. Personally, I owe him my sanity for knowing exactly when to propose we reduce our stress with a long run or a video and a bottle of wine. Professionally, he has been my most dedicated field assistant, the R master behind the statistics, and the sound board for nearly every idea included in this doc ument. Without his support, both personal and professional, this entire dissertati on would have been impossible.
7 TABLE OF CONTENTS page ACKNOWLEDGMENTS ............................................................................................................... 4LIST OF TABLES ................................................................................................................ ...........9LIST OF FIGURES ............................................................................................................... ........10ABSTRACT ...................................................................................................................... .............11 CHAPTER 1 ARE PLANT POPULATIONS SEED LI MITED? A CRITIQUE AND METAANALYSIS OF SEED ADDITION EXPERIMENTS ..........................................................13Abstract ...................................................................................................................... .............13Introduction .................................................................................................................. ...........14Methods ....................................................................................................................... ...........19Database ...................................................................................................................... ....19Meta-Analysis ................................................................................................................. .21Definition of Effect Sizes and Weighting Factors. ..........................................................21Absolute Response Effect Size ........................................................................................23Summary Analyses. .........................................................................................................24Results ....................................................................................................................... ..............26Seed Limitation in Undisturbed Plots ..............................................................................26Seed Limitation in Disturbed Plots ..................................................................................27Effect of Disturbance .......................................................................................................28Absolute Response Effect Size ........................................................................................29Discussion .................................................................................................................... ...........29The relative Importance of Seed vs. Establishment Limitation ......................................30Plant and Site Characteristics ..........................................................................................31Limitations and Proposed Improvement s of Seed Addition Experiments ......................332 A NARROW NICHE FOR NEUTRAL PROCESSES IN THE RECRUITMENT OF AFROTROPICAL TREE SEEDLINGS ................................................................................45Abstract ...................................................................................................................... .............45Introduction .................................................................................................................. ...........45Methods ....................................................................................................................... ...........48Overview ...................................................................................................................... ...48Site Delineation and Characterization of Seed Rain .......................................................48Seed Sowing Experiments ...............................................................................................50Quantification of Seed and Establishment Limitation .....................................................51Realized limitation: ..................................................................................................51Fundamental limitation: ...........................................................................................53Results ....................................................................................................................... ..............55
8 Discussion .................................................................................................................... ...........573 TERRESTRIAL MAMMALS, MORE TH AN ENVIRONMENTAL FILTERING OR NEGATIVE DENSITY-DEPENDENCE, DRIVE PATTERNS OF TROPICAL SEEDLING RECRUITMENT ...............................................................................................65Abstract ...................................................................................................................... .............65Introduction .................................................................................................................. ...........65Methods ....................................................................................................................... ...........69Study Area .................................................................................................................... ...69Site selection ............................................................................................................70Experimental design .................................................................................................70Environmental Variables .................................................................................................71Light availability ......................................................................................................71Soil sampling and analysis .......................................................................................72Ecological Variables ........................................................................................................73Herbivory and seed predation ..................................................................................73Density and distance effects Â– evaluating Janzen-Connell ......................................73Data Analysis ................................................................................................................. .........74Relative Importance of Mechanisms that Limit Seedling Emergence and Survival ..............74Results ....................................................................................................................... ..............75Environmental Factors .....................................................................................................76Seed Predation and Herbivory .........................................................................................76Densityand Distance Dependenc e (Janzen-Connell Effects) ........................................77Discussion .................................................................................................................... ...........77Niche Partitioning and Environmental Filters .................................................................78Densityand Distance -Depende nce (Janzen-Connell Effects) .......................................80Vertebrate Seed Predation and Herbivory .......................................................................81Conclusion .................................................................................................................... ..........83APPENDIX A SELECTION OF THE EFFECT SIZE FO R SEED LIMITATION EXPERIMENTS ..........95Conceptual Approaches to Effect Sizes ..................................................................................95Parameter Estimation .......................................................................................................... ....95Elasticity or Sensitivity ..................................................................................................... ......97Limitation .................................................................................................................... ...........97Empirical Estimates of Seed Limitation Using Two Treatments ...........................................98Comparison of Effect Sizes ..................................................................................................10 0B SUPPLEMENTARY MATER IAL FOR CHAPTER 2 .......................................................106C SUPPLEMENTARY MATERIAL FOR CHAPTER 3. ......................................................119LIST OF REFERENCES ............................................................................................................ .124BIOGRAPHICAL SKETCH .......................................................................................................138
9 LIST OF TABLES Table page 1-1 Comparison of seed limitation (per seed response) by grouping variables .......................371-2 Comparison of seed limitation (absol ute response) by grouping variables .......................402-1 Ecological characteristics of focal tree species selected for use in this study ...................592-2 Ambient seed rain density and seed add ition levels used for each species in each subplot (N = 63) .............................................................................................................. ...602-3 Results from generalized linear mixed model (GLMM) analyses .....................................613-1 Ecological characteristics of focal tree species selected for use in this study ...................853-2 Number of 1-hectare sites (N=30) in which adult indivi duals >10 cm dbh of our focal species co-exist .............................................................................................................. ....863-3 Summary of GLMM analys is identifying factors ..............................................................873-4 Summary of GLMM analysis identifying the factors ........................................................88B-1 Species specific results from generalized linear mixed model (GLMM) analyses .........107B-2 Parameter values from the density dependent .................................................................109C-1 Complete results of GLMM .............................................................................................119
10 LIST OF FIGURES Figure page 1-1 Seed limitation effect sizes for all grouping variables .......................................................431-2 Seed limitation in relationship to seed mass, logevity, and seed bank. .............................442-1 Per seed recruitment effect size E an d total recruitment for all species.. ..........................622-2 Fit of the two candidate recruitment function models to seed augmentation data pooled for all five species ..................................................................................................6 32-3 Results of limitation analys is for all species combined .....................................................643-1 Map of 30 site locations in the northern Republic of Congo .............................................893-2 Site establishment and delineation .....................................................................................903-3 Experimental design........................................................................................................ ...913-4 Graph depicting the variation in percent transmitted diffuse light ....................................923-5 Principal component analysis of soil variables in 63 stations ............................................933-6 Seedling recruitment and survival at three months and two years. ...................................94A-1 Depiction of relative limitation .......................................................................................10 4A-2 Depiction of Absolute limitation. ....................................................................................105B-1 Study site selection ...................................................................................................... ....111B-2 Site delineation, mapping and seed trap set up ................................................................112B-3 Experimental design........................................................................................................ .113B-4 Graphical representation of seed lim itation based on the Beverton-Holt (1957) ............114B-5 Realized seed-establishment limitation for each of 5 species. .........................................115B-6 Fit of the final two of four candida te recruitment function models to seed augmentation data ............................................................................................................1 16B-7 Results of the analyses of limita tion analysis for each of 5 species ................................117
11 Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy FACTORS THAT DETERMINE PATTERNS OF SEEDLING RECRUITMENT IN AN AFROTROPICAL FOREST By Connie Jane Clark August 2009 Chair: Doug Levey Major: Interdisciplinary Ecology Tropical forests account for nearly 50% of a ll known species. Very little is understood about the processes that maintain or promote su ch diversity. Theoretical models suggest that processes limiting recruitment of new individuals into populations may be key to maintaining species diversity. By keeping population numbe rs of more competitive species in check, recruitment limitation should allo w greater numbers of species to co-exist. Two opposing hypotheses have been proposed to explain how re cruitment limitation might influence tropical tree diversity; the seed limitation hypothesis and the establishment limitation hypothesis. These hypotheses are generally treated as mutually exclusive, and evidence of either is used to bolster competing theories of community composition th at are tightly associated with each. In chapter 1, I develop a framework that views seed and establishment limitation as processes that occur at opposite ends of a continuum. I adopt this framework in a meta-analysis to assess the relative strength of seed and establishment limitation across a range of plant systems. I find that most species are seed limited, though the effects of seed addition are typically small. Establishment, on average, proves to be a stronger limiting force for most plants. I provide recommendations to improve experimental approaches used to examine the relative strength of thes e two processes. Chapter 2 applies these recommendations to a large
12 scale experiment designed to teas e apart the roles of seed and establishment limitation for five randomly-chosen tree species in an Afrotropical forest. I conclude that though seed limitation is relatively weak, it can balance the exclusion pro cess of competition and niche partitioning at very high levels of seed arriva l. Yet, niche-based, post disp ersal processes more importantly limit seedling recruitment than s eed arrival. Chapter 3 delves into the mechanisms responsible for post dispersal seed and seedling mortality. I evaluate the strength and relative importance light availability, soil fertility, competition, densityand distance-dependence, seed predation and herbivory at two stages of seedling recruitment. I conclude that seedling recruitment in the Congo Basin is most strongly dictated by generalis t vertebrate seed predators and herbivores, with relatively weaker abioti c environmental filtering and dens ity-dependence playing secondary roles.
13 CHAPTER 1 ARE PLANT POPULATIONS SEED LIMITED? A CRITIQUE AND META-ANALYSIS OF SEED ADDITION EXPERIMENTS Abstract We examine the relative importance of pro cesses that underlie plant population abundance and distribution. Two opposing views dominate the fi eld. One posits that the ability to establish at a site is determined by the availability of suitable microsites (Â“establishment limitationÂ”), while the second asserts that recruitment is limited by the availabil ity of seeds (Â“seed limitationÂ”). An underlying problem is that es tablishment and seed limitation are typically viewed as mutually exclusive. We conducted a meta-analysis of seed addition experiments to assess the relative strength of establishment a nd seed limitation to seedling recruitment. We asked: (1) To what degree are populations seed and establishment limited? (2) Under what conditions (e.g., habitats and life history traits) are species more or less limited by each? (3) How can seed addition studies be better desi gned to enhance our understanding of plant recruitment? We also examined if previous re sults based upon the cruder summary technique of Â“vote-countingÂ” were upheld when quantitative es timates of seed limitation were considered. We found that in keeping with previous studies, most species are seed limited. However, the effects of seed addition are typically small, and most added seeds fail to recruit to the seedling stage. As a result, establishment limitation is stronger than seed limitation. Seed limitation was greater for large-seeded species, species in distur bed microsites, and for species with relatively short-lived seedbanks. Most s eed-addition experiment s cannot assess the relationship between number of seeds added and number of subsequent recruits. This shortcoming can be overcome by increasing the number and range of seed addition treatments.
14 Introduction Identifying mechanisms that determine the a bundance and distribution of plant and animal populations is a central challe nge of ecology (Coomes and Grubb 2003; Levine and Rees 2002; Osenberg et al. 2002; Tilman 1997; Turnbull et al. 2005). The failure of a species to recruit at a given site can result from pro cesses that occur at practically any life history stage and include propagule production and transportation, competition, predation and herbivory. Despite this range of disparate processes and stages, several of the best-known models of species coexistence are focused on propagule availability in sp ace or time (Coomes and Grubb 2003; Hurtt and Pacala 1995; Pacala and Levin 1997; Sale 1982; T ilman 1994). These models are bolstered by empirical studies across diverse sy stems, demonstrating that early life history events (e.g., during the transition from seed to seed ling, or larva to juvenile fish) can be bottlenecks for recruitment (Chambers and Macmahon 1994; Doherty 2002; Fenner 2000; Persson and Greenberg 1990). Indeed, there is growing consensus that processes underlying mortality at ea rly stages in the life cycle may disproportionately influence the st ructure, dynamics, and species composition of communities. This consensus is particularly evident in studies of plant communities. Two processes thought to limit plant r ecruitment at early stages in th e plant life cycle are seed and establishment limitation. Seed limited populations have fewer individuals than possible because seeds fail to arrive at saturating densities to all po tential recruitment sites (Eriks son and Ehrlen 1992; Nathan and Muller-Landau 2000; Svenning and Wri ght 2005; Turnbull et al. 2000). Seed limitation can be partitioned into two processes that restrict the ab ility of seeds to reach recruitment sites: (1) source limitation -not enough seeds are produced to satura te potential recruitment sites even if the seeds could reach all sites, and (2) dispersal limitation Â– not enough seeds reach all
15 recruitment sites, even though e nough are produced to saturate si tes (Clark et al. 1998b; Schupp et al. 2002). Establishment limitation (also called microsite limitati on) occurs when plant population size is constrained by the number and quality of available sites for establishment, not by the number of seeds (Clark et al. 1998b; Nathan and Muller-Landau 2000). Establishment limitation can be partitioned into several processes that o ccur between seed deposition and recruitment into the adult population (Nathan a nd Muller-Landau 2000). In this paper we focus on seedling recruitment, as it represents a key stage of es tablishment limitation. Specifically, we examine the time between seed arrival at the soil surface and the census of seedlings after the first season of growth. Establishment limitation is thus dete rmined by factors that constrain the recruitment of new individuals into the seed ling population, regardless of the numbe r of seeds that arrive at a site. Seed and establishment limitation are an alogous to supply limitation and post-settlement mortality, as developed in the literature on reef fish ecology (Doherty 2002; Osenberg et al. 2002; Schmitt et al. 1999). Because both seed and establishment limitation can limit plant recruitment, both are likely to influence the abundance and distribution of sp ecies (Dalling et al. 2002; Hubbell et al. 1999; Juenger and Bergelson 2000; Levine and Rees 2002; Zobel et al. 2000) At issue is their relative importance. At stake are competing theori es of community composition (Coomes and Grubb 2003; Turnbull et al. 2005). If establishmen t limitation dominates, then the abundance and distribution of a species is readily framed as an issue of compet itive ability, rege neration niches, and the relative abundance and quality of micr osites (Grubb 1977; Muller-Landau et al. 2002; Pearson et al. 2002; Turnbull et al 2000). If seed limitation domi nates, then the abundance and distribution of a species is better viewed in the context of a lottery system, where few sites are
16 Â“wonÂ” by the best possible competitor and most are Â“won by defaultÂ” Â– recruits are drawn at random from the seeds that happen to arrive at a site (Cornell a nd Lawton 1992; Hubbell 2001; Sale 1982). Thus, empirical studies on the re lative importance of seed and establishment limitation can guide theoretical models of community dynamics. The most direct means of testing the rela tive importance of seed and establishment limitation is to conduct seed addition experiment s (Muller-Landau et al. 2002; Turnbull et al. 2000; Turnbull et al. 2005). Seeds are added to pl ots and the numbers of seedlings that emerge are compared to control plots in which no seeds have been added. If no increase in seedling density is observed following s eed addition, one can conclude th at recruitment opportunities for that species are not seed limited. Instead, the number of microsites availabl e or the suitability of those sites for seedlings limits recruitment, a nd establishment limitation is more important for that species. If, on the other hand, an increase in seedling density is observed following seed addition, one can conclude that lim itations on species presence or abundance are at least partially attributable to seed availability (although its impor tance relative to factor s that limit recruitment at later life histories cannot be evaluated w ithout longer-term study). Such experiments, by decreasing the extent of seed limitation and is olating the emergence and early post-emergence stages of establishment limitation, offer a conser vative estimate of the strength of establishment limitation relative to seed limitation in plan t populations. The re lative importance of establishment limitation would be expected to incr ease if additional mortality at later life history stages were included. The difficulty in interpreting results of seed a ddition experiments lies in situations in which seed limitation is detected (i.e., one finds a sta tistically detectable increase in seedlings following seed addition). In large part, interpretation depends upon how the e xperiment was framed. If the
17 underlying goal was to determine why a given species does not occur at a particular site, then even a single seedling demonstrates seed lim itation. This goal is common among seed addition studies based on small plots; the response they do cument is local and the number of seedlings largely irrelevant (assuming enoug h survive to establish a populat ion). If, on the other hand, the underlying goal was to determine factors lim iting population size or density, the number of seedlings becomes key to disentangling the re lative strengths of s eed and establishment limitation. In this scenario, detection of seed limitation is largely irrele vant -attention should focus on the magnitude of response rather than its presence or absence. The magnitude of seed limitation is rarely considered in seed addition studies. For example, a review of seed addition experiment s concluded that as many as 50% of all plant populations are seed limited (Turnbull et al. 2000) However, seed limitation was depicted dichotomously Â– either seed availability limite d plant population size (i.e ., there was a significant effect of seed augmentation) or it did not (i.e., the resulting P -value was > 0.05). A central theme of this paper is that seed limitation is a continuous variable, pote ntially varying widely among species, habitats, life forms, plant ch aracteristics, and seed sizes. Using P -values to infer seed limitation, not only dichotomizes this continuous scale, but also confus es a statistical view of significance with the mo re appropriate biological view of ove rall impact e.g., Osenberg et al. (1997). Indeed, the use of eff ect sizes can give very diffe rent results than the use of P -values derived from null hypothesis test s (Osenberg et al. 2002). Viewing seed limitation as a continuous vari able provides a framewo rk for evaluating the relative strength of seed and es tablishment limitation. In particular, one can view seed and establishment limitation as being inversely re lated, occupying opposite ends of a gradient (Muller-Landau et al. 2002). By quantifyi ng the position of a plant population along this
18 gradient, one can judge the rela tive strength of seed vs. establ ishment limitation and determine the magnitude of each. For example, if 100 seeds are added to a plot and result in 100 emerged seedlings, the proportion of emerged seedlings to sown seeds is one, and the population is strongly seed limited. If, on the other hand, 100 seeds are added and no new seedlings emerge, the proportion of additional emerged seedlings to added seeds is zero, and the population is strongly establishment limited, w ith no evidence of seed limitation. More typical and revealing are situations in which the proportion of added s eeds that emerge is intermediate, indicating that populations are simultaneously limited by two f actors but probably to different degrees. We present a meta-analysis of seed augmenta tion experiments, with the goal of teasing apart the relative strengths of seed and estab lishment limitation for seedling recruitment. Because our focus is on factors that limit populati on size and density, we develop an effect size measure based on per seed return (i.e., change in seedling density/density of augmented seeds). We also use a second effect size to examine th e absolute extent by wh ich plant populations and species distributions are seed li mited (i.e., the change in seedli ng density without correction for augmentation level). We then examine variati on among studies in effect sizes to determine differences among them in the magnitude of seed and establishment limita tion. We have three objectives: (1) To examine the degree to whic h plant populations are seed and establishment limited. (2) To determine under what conditions we might expect plant species to be most seed limited. Specifically, we test how life form, ha bitat, dispersal mode, plant characteristics, reproductive characteristics, seed bank persiste nce and density, and species origin (native or exotic), influence the degree of seed and esta blishment limitation. We evaluate the suggestion that seed limitation is more common in early successional habitats (T urnbull et al. 2000) by examining studies in which seeds were sown into disturbed and undist urbed plots, and we
19 determine whether the positive relationship betw een seed limitation and seed size observed by Moles and Westoby (2002) is maintained when the ma gnitude of effect is considered. (3) To use our results to inform future studies based on seed addition. Specifically, our examination of this literature revealed shortcomings of common ex perimental designs that greatly limit the interpretation of seed addition experiments. Thus, we conclude by suggesting improvements for the design of future studies. Methods Database We searched for published studies in which seed s had been experimentally added to plots, regardless of why they had been added. We used a recent review of seed augmentation experiments (Turnbull et al. 2000) as our main source of references, but also searched Web of Science (ISI 2004) for all papers published by the su mmer of 2004 that cited this review or included the keywords: Â“seed sowingÂ”, Â“seed limitationÂ”, Â“seed augmentationÂ”, Â“germinationÂ”, Â“seed introductionÂ”, or Â“seedling recruitmentÂ”. When necessary, we contacted authors for information. In the process, we learned of several unpublished studies, which we included with permission. Many studies were not included in our analysis because they failed to meet one or more of the following criteria: (1) Experiments were cond ucted in natural or semi-natural settings (e.g., not in greenhouses). (2) Estimates of seedli ng emergence/early post-emergence establishment for a single plant species for both treatment (see ds added) and control plots (no seeds added) were available. The only exceptions were studies that introduced seeds of species absent from the study site. We included these studies lacking tr ue control plots if the au thor explicitly stated that the species was not presen t in nearby sites. In these cases, seedling emergence under ambient conditions (control) was assumed to be zero. (3) Sample sizes, replication, means, and
20 variance were appropriately repor ted (i.e., no pseudo-replication) or were made available by authors. When studies monitored plots for >1 yr, we rest ricted our analyses to the end of the first growing season. The sole exception to this ru le was the inclusion of Edwards and Crawley (1999), which quantified seedling de nsity after 15 mo (450 d). Thus our effect sizes apply only to first-season seedlings. The period of tim e between seed sowing and first-season seedling censuses varied among studies (ranging from 14 to 450 days, with a mean of 292 days); we assumed that investigators censused first-season seedlings at the most appropriate time for each species. From all studies that met our criteria, we extracted: (1) mean de nsity and variance of recruited seedlings in treatment and control plots, (2) number of replicate plots, (3) number of seeds added in each plot, and (4) grouping variables thought to influence the degree of seed limitation. Grouping variables included charac teristics of the study site (habitat and geographical zone), characteristics of the focal sp ecies (plant life form maximum plant height, plant longevity, seed mass, aver age fecundity, dispersal mechanism, presence/absence of seed bank, seed bank density, seed bank longevity, time to first flowering, seedli ng growth rate), and characteristics of the experimental treatments (removal of vegetation, ster ilization of soil, or turnover of the soil; Table 1-1). These envi ronmental and plant characteristics potentially influence seedling emergence and, presumably, the strength of seed limita tion (Turnbull et al. 2000). If studies did not provide information on gr ouping variables, we gathered these data from outside references whenever possible (Grime et al. 1988; Moles and Westoby 2004; Royal Botanic Gardens 2002; Thomps on 1987; USDA NRCS 2004).
21 Meta-Analysis Meta-analysis involves two key steps. First, results of each study are used to calculate a biologically relevant effect size, often a measure of the disparit y of responses between a control and treatment group (Osenberg et al. 1999). Second, effect sizes are statistically summarized to estimate a weighted average for the sample of stud ies (average effect size) and to test hypotheses (Gurevitch et al. 1992). Definition of Effect Sizes and Weighting Factors. Although we considered several potential measur es of effect size for seed limitation, we chose the metric that most closely matched our question of interest and the design of the seed sowing experiments (Online Appendix A provide s a theoretical discussion and empirical evaluation of alternative effect size metrics). Our metric of seed limitation, the per seed response, was the difference between seedling densities in treatment and control plots, standardized by the number of seed s added to treatment plots: i i cont i iA R R E, exp, (1-1) where Ei is the effect size, Rexp,i is the average density of recr uits (seedlings) in experimental plots, Rcont,i is the average density of r ecruits in control plots, and Ai is the density of seeds added to treatment (seed augmentation) plots in the ith study. Ai varied by more than an order of magnitude among studies and necess itated the standardization in E qn. 1. Our effect size can be interpreted as the number of recruits obtained per sown seed. In theory, E should vary between 0 and 1, unless density effects are so strong that total recruitment is reduced by the addition of more seeds (i.e., if there is overcompensation) Because recruit densities are estimated and background seed rain is an uncontro lled variable, estimated effect sizes also could be <0 or >1 due to sampling error.
22 Meta-analysis combines effect sizes obtained from a collection of studies, giving greater weight to studies with higher precision. In general, n i i i n i iw E w E1 1 (1-2) where E is the average effect size, and wi is the weight associated with the ith effect size. Parametric approaches use weights that are invers ely related to the variance in effect size for a given study (Rosenberg et al. 2000 ). In our dataset, however, many studies had small numbers of replicate plots and sown seeds, which often resulted in no emergence ( Rexp = Rcont = 0), a variance of zero, and a weight of infinity. For this reason, using the inverse of va riance as the weight was impractical and likely not a good refl ection of precision. Therefore, we used a weighted, resampling procedure (with replacemen t) in MetaWin 2.0 (Rosenberg et al. 2000). Weights were based on the total number of s eeds added to augmentation plots (across all replicates), which we assumed was approximately proportional to the precision of the estimated effect sizes Â– i.e., we assumed that effect size s were better estimated when more seeds were added and therefore the number of potential recr uits was greater. Because we did not take a parametric approach based on true variance esti mates, we could not partition within and amongstudy sources of variation. Thus, we used: iX x x i iN w1 (1-3) where Ni,x is the number of seeds added to the xth replicate of the augm ented treatment, and Xi is the number of replicates of study i
23 Some plant species were used in more than one study, or were added at more than one seed density within a single study. To prevent spec ies that were used in multiple studies from carrying more weight in the calcul ation of average effect sizes, we first estimated the effect size from each study and then averaged these effect sizes for each species using Equations 2 and 3 and the resampling procedure in MetaWin 2.0 (R osenberg et al. 2000). We then derived a pooled weighting term, reflecting that the averaged effect size was based on several studies (all with different levels of a ugmentation and replication): ji jK i X x x i j j jN K w11 2) / 1 (, (1-4) where Kj is the number of effect sizes being pooled for species j x i jN, is the number of seeds added to the xth replicate of the ith study for species j and Xj,i is the number of replicates in the ith study for species j When sample sizes are equal, Eqn 4 reduces to KjXjNj, or total augmentation across all studies for species j which is comparable to the weighting term given in Eqn 3. Absolute Response Effect Size Although Ei (Equation 1) is the most appropriate effect size given the available data (Online Appendix A), we also calculate the absolu te response of the popul ation to seed addition to better examine the degree by which seed and establishment limitation limit the absolute extent of plant populations. The abso lute response measures the abso lute change in recruitment (seedling density) between the augmented (e xperimental) and control treatments: i cont i i absR R E, exp, (1-8) where Rexp,i and Rcont,i are the average densities of recruits in the experimental and control plots, respectively, in the ith study (see Online Appendix A for further discussion).
24 Summary Analyses. Studies were conducted under di sturbed or undisturbed conditions, and sometimes in both disturbed and undisturbed c onditions. Therefore, we distinguis h between three types of effects: ED, EU, and E Disturbed conditions were created by re moving vegetation or litter, turning the soil, or physically manipulating the plot in some other way. We took two approaches to analyzing these data. First, we examined pa tterns of seed limitation separately for seed augmentations done in disturbed an d undisturbed settings. This yiel ded effect sizes in disturbed plots ( ED, i: i.e., seed limitation in disturbed plots for species i ), and effect sizes in undisturbed plots ( EU, i). We then explored the relationships between the magnitude of seed limitation (i.e., using either ED, i or EU, i) and the grouping variables (e.g., grow th form or seed mass for species i ), by examining the heterogeneity of effect sizes using Q statistics, which are essentially weighted sums of squares following a2distribution. The corresponding P -value indicates whether the variance among effect sizes is grea ter than expected by chance. We ighted effect sizes and biascorrected 95% confidence intervals of seed lim itation (Eqn 1) were estimated for categorical grouping variables using resampling methods with 10,000 iterations in MetaWin 2.0 (Rosenberg et al. 2000). For continuous gr ouping variables, we conducted we ighted linear regressions to determine whether plant characteri stics explain variation in the effect size of seed limitation. After examining plots of residuals, we used a lo garithmic transformation on several of the plant characteristics to satisfy the assumptions of norma lity and homoscedasticit y. We then regressed the plant characteristic against our estimate of effect size ( ED, i or EU, i), using randomization tests with 10,000 iterations to conduct sign ificance tests. Regressions were done using the R language (R Development Core Team 2005).
25 Because the species and studies in the dist urbed and undisturbed datasets differed, it is problematic to infer the effect of disturbance on seed limitation by comparing the distributions of ED and EU. Instead, we took a second approach to dire ctly evaluate the effect of disturbance on seed limitation. We used only the studies in wh ich seeds of a single species were sown in both undisturbed and disturbed plots, and defined the effect of disturbance on seed limitation for species i as: Ei = ED, i Â– EU, i (1-5) where ED, i and EU, i were calculated with Equation 1. Note that E will be negative if seed limitation is more severe in undisturbed plots and positive if seed limitation is more severe in disturbed plots. If seed limitati on is independent of disturbance regime (i.e., equal in disturbed and undisturbed plots), then E = 0. Thus, E can be small (close to zero) for a particular species even when seed limitati on is strong (but comparable in magnitude) in the disturbed and undisturbed plots. In such a cas e, other characteristics of the plot or the species (but not disturbance, per se) determine th e magnitude of seed limitation. To derive an appropria te weighting term for E we assumed that: Var( Ei) = Var( ED ,i) + Var( EU ,i) = c/ ND ,i + c/ NU ,i = c( ND ,i + NU ,i)/( ND ,i NU ,i), (1-6) where c is a scaling term that relates the number of seeds sown and the resulting variances, and ND,i and NU,i are the total number of seeds sown in the disturbed and undisturbed augmentation treatments for study i Because parametric weighting factor s should be inversely related to the variances, we defined a weighting term that was inversely proportional to the presumed variance (Eqn. 6): i U i D i U i D iN N N N w, , (1-7) where w ,i is the weighting given to Ei
26 As described above, species that occurred more than once in the dataset were combined into a single pooled effect size, and the pooled weighting term describe d above (equation 3) was calculated. A cumulative disturbe d effect size was calcu lated in MetaWin 2.0, and differences in seed limitation among grouping variables were ev aluated. Disturbed pl ots were considered significantly more (or less) seed limited than undi sturbed plots if the 95% confidence intervals on E did not overlap zero. Results Studies reported in forty-th ree publications met all crite ria for inclusion, yielding 798 effect sizes based on 159 species in 49 families. The most common reasons for exclusion of a study were lack of control plots or presence of multiple species of seeds sown together in treatment plots. Other studies were excluded because establishmen t or survival was not recorded until after the first growing season. Seed Limitation in Undisturbed Plots In most undisturbed environments tested, plant species were seed limited -adding seeds to a plot generally resulted in more seedlings th an in plots where no seeds were added (Figure 11A). However, the average effect of seed li mitation was small, with only 15 out of 100 seeds emerging as seedlings ( UE = 0.15, 95% CI = 0.111 Â– 0.195). Assuming an inverse relationship between seed and establishment limitation, this ef fect size indicates that establishment limitation, calculated as 1 Â– E more strongly limits emergence of s eedlings than does seed limitation (0.85 vs. 0.15). Effect sizes of seed limitation for all species were relatively low: 68% of species had E < 0.25, 20% had 0.25 < E < 0.50, and only 12% had E > 0.50 (Online Appendix B). Of the habitat and life history characteristics that we examined, only s eed origin, seed size, seed presence (seed augmentation versus seed in troduction), and the average seed bank longevity explained a significant portion of variation in seed limitation e ffect size (Table 1-1, Figure 1-
27 1A). Exotic species were more seed limited than native species ( P = 0.019; Figure 1-1A). Similarly, native species that were introduced in to an area where they naturally occurred, but were not present during the study, were more seed limited than nativ e species whose seed densities were augmented by the researcher (T able 1-1, Figure 1-1A). There was an inverse relationship between the longevi ty of the seedbank and the strength of seed limitation ( P = 0.012): species with seed banks of short duration we re more seed limited than those with seed banks of longer duration (Table 1-1). The streng th of seed limitation was also significantly different among groups of species with different seed masses ( P = 0.002; Figure 1-1A): species with larger seeds were more seed limited. Regres sion analysis identified a marginally significant positive relationship between seed mass and the degree of seed limitation, with a 3% increase in seed limitation per mg increase in seed mass ( P = 0.051). The species with the largest seed (200 mg) was approximately 20 times more seed limite d than the species with the smallest seed (0.02 mg). To examine the possibility that study design in fluenced the magnitude of seed limitation, we also regressed EU against duration of the study and seed augmentation density Â– but see (Osenberg et al. 1999). Neither of these factors signifi cantly influenced EU (Table 1-1). Seed Limitation in Disturbed Plots Plant species were also significantly seed lim ited in disturbed plots, with an effect size comparable to that obtained in the undisturbed plots ( DE = 0.14, 95% CI = 0.102 Â– 0.180), indicating that 14 out of every 100 seeds sown in disturbed plots t ypically emerged as seedlings. This effect represents the average number of s eedlings that emerged per seed added in disturbed plots, but not the direct effect of disturbance on seed limitation. Similar to the analysis with undisturbed plots, the small eff ect size suggests that establis hment limitation more strongly limits emergence of seedlings than does seed limitation (0.86 vs. 0.14). Effect sizes for all
28 species were relatively low (81%: E < 0.25; 11%: 0.25 < E < 0.50; 8%: E > 0.50; Figure 1-1B, Online Appendix B). We found no significant differences in the eff ect size of seed limita tion among most of the plant and habitat grouping variables that we investigated for studies using disturbed plots (Table 1-1; Figure 1-1B). Again, the strength of seed limitation wa s significantly different among groups of species with different seed masses ( P = <0.001; Figure 1-1B). This difference was driven by the difference between seeds weighing less than 0.2 mg compar ed to those weighing between 1 and 2 mg (Figure 1-1B); however, wh en seed size was regressed against seed limitation, the slope of the linear regression wa s not significantly different from zero ( P = 0.304). Neither the level of seed augmentation nor the duration of the study significantly influenced the effect size in disturbed plots ( P = 0.069 and P = 0.838, respectively). This result is the opposite of what would be expected based on de nsity-dependent seed survival (Poulsen et al. 2007) and suggests that another factor was corr elated with sowing density (e.g., seed mass). Effect of Disturbance Disturbance had a significant positive effect on seedling emergence (i.e., disturbance increased the magnitude of seed limitation). A di rect comparison of seed limitation in disturbed and undisturbed plots from the same studies (N = 75 studies with disturbed and undisturbed treatments) revealed that disturbance resulted in ~10 more emerged seedlings per 100 seeds sown, relative to undisturbed plots ( E = 0.10, 95% CI = 0.073 Â– 0.132). We found no significant differences in the eff ect size of seed limita tion among most of the plant and habitat grouping variable s that we investigated, with the exception of species with different seed masses (Table 1-1; Figure 1-1C). The strength of seed limitation was significantly different among groups of species with different seed masses ( P = 0.011; Figure 1-1C).
29 Absolute Response Effect Size The absolute population size of emerged seedlings increased with seed augmentation. The average increase in population size was greater in disturbed environments ( D absE,= 1412.7, 95% CI = 768.246 Â– 2704.661) than undisturbed environments ( U absE,= 658.2, 95% CI = 348.629 Â– 1177.679). Disturbance also had a positive effect on seedling emergence ( absE = 1172.1, 95% CI = 667.834 Â– 2045.675). The absolute response was influenced by the seed augmentation density, with the number of emerged seedlings increasing significantly with the density of seed augmentation in disturbed plots (D absE,; Table 1-2). Similarly, the increase in the number of emerged seedlings with increased seed augmentation density was margin ally significant for both undisturbed plots U absE, and the effect of disturbance (absE ; Table 1-2). It is likely, theref ore, that the absolute response is largely influenced by the study design (degr ee of augmentation) and not the biology of the system. Whereas the per seed response ( E ) varied significantly with several grouping variables (Table 1-1), the absolute seed response only va ried significantly with gr ouping variables related to the study design (seed augmentation dens ity and duration of study; Table 1-2). Discussion Recent reviews on seed limitation have employed a Â“vote countingÂ” approach, in which plant populations were deemed either seed lim ited or not seed limited (Moles and Westoby 2002; Turnbull et al. 2000). They concluded that most plant species were seed limited. Using an approach based on magnitude of response, we also found that plants were generally seed limited, as indicated by an average effect size greater than zero. Howeve r, a statistically detectible pattern of seed limitation is not necessarily biologically important The average effect sizes for
30 seed limitation were usually small, averagi ng ~0.14-0.15 with ~90% of all species showing effect sizes <0.50. These results show that only a small fraction of augm ented seeds recruit to the seedling stage and suggest that establishmen t processes play a major role in determining plant population density, at leas t at the seedling stage. Our analysis focused on the earliest phase of recruitment Â– from seed to seedling. Because our analysis did not include factors that cause pl ant mortality during later life history stages, we likely overestimated the relative importance of seed limitation and underestimated the role of establishment limitation to recruitment. Why were the effects of seed addition experi ments so small? We offer two non-mutually exclusive explanations. Most obvi ously, plant populations may be more establishment than seed limited, with an establishment bottleneck occurring between when seeds were deposited and when seedlings were censused during the first gr owing season. Alternatively, the small effect sizes may be artifacts of experimental design, generated when seeds were added to plots at biologically irrelevant densities. We first discuss several fact ors likely to constrain seedling establishment following seed arrival, particularly seed predation and char acteristics of the plant and microsite. We conclude by discussing the experimental design of augmentation studies and providing suggestions fo r future studies. The relative Importance of Seed vs. Establishment Limitation Based on the huge discrepancy between the number of seeds produced by most plant species and of the number of seedlings that resu lt, it has long been argued that the stage between seed production and seedling emergence is a crit ical bottleneck for plant populations (Harper 1977). Our low effect sizes for seed augmenta tion experiments support this observation Â– in practically all studies, added seeds rarely surv ived to the seedling stage. Thus, the seed-toseedling bottleneck is best explai ned by processes that affect seed survival after seed arrival,
31 contrary to the conclusions of many studies (D alling et al. 2002; Eriksson and Ehrlen 1992; Szentesi and Jermy 2003; Turnbull et al. 2000; Zobel et al. 2000). Although seeds must arrive at a site before a plant can estab lish there, our results underscore the need to further examine factors that underlie post-di spersal seed mortality. Post-dispersal factors that might constrain seedling establishment include seed viability and senescence, abiotic (e.g., light water, nutrients, soil struct ure) and biotic (e.g., seed and seedling predators, pathogens, comp etitors) factors that affect germination, seed survival and seedling survival. For species included in th is study, germination rates under lab conditions average 90% (Royal Botanic Gardens 2002) but the effect sizes we calculated for these same species indicate that, on average, ~85% of all seeds in the field failed to emerge as seedlings. Thus, the low proportion (~15%) of seeds that em erged as seedlings cannot easily be explained by low seed viability or senescence. Rather the consistently low proportion of seedling emergence (and early post-emergence survival) ob served across studies must be explained by (1) high rates of post-dispersal seed predation, pa thogen attack, or seedling herbivory or (2) characteristics of the plant or microsite th at decrease the probability of germination. Plant and Site Characteristics In theory, some plant and site characteristics should be disproportionately important in determining the probability that a seed successful ly recruits to the seedling stage, thereby influencing the relative importance of seed and establishment limitation. However, our results indicate that the effect of any given grouping vari able on seed limitation is weak at best (plant height, seedling growth rate, age of first fl owering, longevity, and fecundity). Of the 16 variables, we explored, only dist urbance, seed size, seed origin, seed presence, and seed bank longevity affected seed limita tion by our meta-analysis.
32 Seed limitation has been presumed to be mo re common in disturbed (early successional) than undisturbed habitats (Turnbull et al. 2000). Our results substa ntiate this view. Removal of adults and imposition of soil disturbance generally increased seed limitati on (i.e., the number of seedlings that recruited from added seeds). Ho wever, the magnitude of the increase was small, suggesting that although disturban ce creates recruitment sites, most seeds still fail to emerge. We conclude that microsite charac teristics unrelated to the distur bance event itself more strongly influence the number of seedlings that emerge at a site than does seed availability. Our results provide support for the hypothe sis of a tradeoff between seed size and colonization ability (Coomes and Grubb 2003; Dalli ng et al. 1998; Levine and Rees 2002; Moles and Westoby 2002; Turnbull et al. 1999). In partic ular, we found that seed limitation was more severe for species with large seed s, a result supported by a previous review that did not consider seed limitation as a continuous variable (Mol es and Westoby 2002; Svenning and Wright 2005). Again, the magnitude of seed limitation was low, even for large-seeded species. Small-seeded species may be less seed limited than large-seed ed species because of the inverse relationship between the number and mass of seeds produced (Dalling et al. 1998; Smith and Fretwell 1974). By producing smaller seeds, a species can achieve greater fecundity and presumably reach more recruitment sites, thereby decr easing seed limitation (Hener y and Westoby 2001; Jakobsson and Eriksson 2000; Moles and Westoby 2002; Smith and Fretwell 1974). Smallseeded species also are more commonly associated w ith persistence in th e soil (Bakker et al 1996; Bekker et al. 1998; Eriksson 1995; Thompson 1987). Long-lived seed banks should further decrease the magnitude of seed limitation because viable s eeds in the soil accumulate over time and can recruit into the population even in the absence of a productive seed year for that species. Our results suggest that seed limitation decreases with longe vity of the seedbank.
33 The greater per-seed recruitment of large-s eeded species revealed by our meta-analysis contrasts with studies that have failed to find a strong relatio nship between seed mass and the number or percentage of seedlings that emerge at a given site (Andresen and Levey 2004; Chen et al. 2002; Moles et al. 2004; Moles and Westoby 2004; Svenni ng and Wright 2005). Other studies have demonstrated that large-seeded species are relatively common as young seedlings, even though they are relatively uncommon in the seed bank (Dalling et al. 1998; Turnbull et al. 2005). Two general types of mechanism may expl ain this post-dispersal advantage of large seeds over small seeds -i.e., why establishmen t limitation is more severe for small-seeded species. First, large-seeded species may have a competitive advantage over small-seeded species (Levine and Rees 2002; Rees and Westoby 1997; Tilman 1994; Turnbull et al. 1999). This advantage appears to decrease as seedlings ma ture (Dalling et al. 2002; Svenning and Wright 2005; Turnbull et al. 2005). Sec ond, large-seeded species may expe rience a wider range of sites suitable for germination and establishment; their greater reserves make them better able to withstand herbivory, drought, shad e, and burial, wherever they are placed (Dalling et al. 2002; Moles and Westoby 2004; Pearson et al. 2002; Turnbull et al. 2005; Westoby et al. 1996). Limitations and Proposed Improvement s of Seed Addition Experiments The small effect of seed augmentation on recruitment may be attributed to multiple processes, including both density -independent and density-depende nt mortality. However, most seed addition experiments are not designed to de cipher these processes an d tend to overlook key assumptions Â– but see Shaw and Antonovics (1986) For example, if th e recruitment function (the relationship between seed input and seedli ng emergence) is non-linear, the outcome of a seed addition experiment (and our estimate of seed limitation) will depend on the number of seeds added by the researcher. Imagine a system in which there is density dependent emergence and either 25 or 1000 seeds were added to a s ite. In both cases, one might find that more
34 seedlings emerge in seed addition plots than in c ontrol plots. However, in the experiment with 25 added seeds, the per seed recruitment rate co uld be higher than in the study with 1000 added seeds because the overall effect of density would be less. Thus, the investigator might falsely conclude that the system with 25 added seeds was more seed limited than the system with 1000 added seeds. Our standardization (Eqn 1) does no t address this issue. Indeed, because most studies include only two treatments (with and without seed augmentation), fitting non-linear recruitment functions and estimating propagule limitation was not possible Â– as done by Schmitt et al. (1999) for reef fi sh. Our only option was to fit a linear model. A companion study by Poulsen et al. (2007) explores the few (18) studies (consisting of predominately short-lived species) in which multiple augmentation levels were used. Consistent with our results, they found that plants were seed limited (seed augmentation led to higher recruitment), and that mortality losses were high In addition, they found that saturation of a system with seeds led to a much greater increa se in recruitment than removal of mortality sources because ambient seed densities were so low that removing mortality affected only a few seeds. In most cases, a linear recruitment func tion fit as well as non-line ar functions, suggesting only a minor role of density-depe ndence and supporting ou r use of a linear model for effect size. Thus, our two studies suggest that the addition of seeds to a system can increase recruitment, but the increase in recruitment will be relatively small (due to high mortality during this critical life stage) unless very large numbers of seeds are added. These results call to question the assumption th at seed augmentation saturates recruitment sites (only 14 of 36 datasets examined by P oulsen et al. exhibited saturation recruitment functions) Â– especially when background seed de nsity is typically not known. Therefore the
35 small effect sizes observed in our study are not likely caused by researcher s saturating potential recruitment sites with seeds. The design and interpretation of seed augmentation studies would be facilitated by knowing the number of seeds deposited by natural seed rain or already present in the seed bank. If the number of seeds added to an experimental plot is not substantially great er than that in seed rain or in the seed bank, it may be difficu lt to detect a difference between control and experimental plots (even when populations are strongly seed li mited) because of the noise induced by the naturally occurring variation in seed supply. For example, adding 25 seeds to plots with a natural seed density averaging 5 seeds/plot will be more likely to detect seed limitation than adding the same number to plots with a natural density averaging 50 seeds/plot because the background variation in recruitm ent should increase with seed rain. In short, interpreting results of seed a ddition experiments can be misleading without knowing the recruitment function and ambient s eed density, which are necessary to place a particular study plot on the recruitment functi on (Poulsen et al. 2007). We urge a two-step process in designing seed addition experiments. First, measure the ambient seed rain and the pre-existing seed bank. Then, add se eds at different densities in di fferent plots, with the goal of spanning the spectrum of natural densities as well as higher densities to fa cilitate description of the recruitment function (Poulsen et al. 2007; Schmitt et al. 1999). Because limitations of experi mental design are an inherent part of the seed addition literature, they apply to our meta -analysis also. The questions that we can answer with our metaanalysis are constrained by the experimental design of the original studies, which may or may not have been conducted to study seed limitation.
36 In summary, we suggest that post-dispersal mo rtality is very high, and that establishment limitation therefore requires more focused study. We emphasize that many processes can cause post-dispersal seed mortality and each coul d influence long-term population abundance. Empirical tests are needed to de termine which of the post-dispersa l processes of seed mortality manifest themselves at the population level. Idea lly, such studies should include field assays to examine the extent to which recruitment into pop ulations is seedor microsite-limited, followed by studies to examine the specific mechanisms of mortality when little or no seed limitation is observed. Similarly, when seed limitation is deemed important for a population, we need to understand the mechanisms that explain the speciesÂ’ inability to occur in suitable sites. Uniting the strength of seed limitation with its underlying mechanisms will make it possible to predict the degree of seed limitation.
37 Table 1-1. Comparison of seed limitation (p er seed response) by grouping variables for A) undisturbed treatments (EU), B) dist urbed treatments (ED), and C) the effect of disturbance on seed limitation (E). The test statistics give Q statistics for categorical variables and R2 statistics for continuous variables (s ee Methods). All randomiza tion tests were two-sided tests, and therefore a P-value of 0.025 i ndicates statistical significance at the 5% level. The code after the grouping variable indicates where data for the gr ouping variable were collected : O (original paper), E (external sources), and OE (data extracted from original papers when possi ble, but often supplemente d by external sources). Undisturbed (EU) Disturbed (ED) Disturbance (E) Grouping variable Test statistic ( Q or R2) df P Test statistic ( Q or R2) df P Test statistic ( Q or R2) df P Description Categorical grouping variables Habitat O 10.149 2 0.176 12.350 2 0.146 1.986 1 0.392 Habitat in which the study was conducted (intertidal, grassland, forest). Geographic zone O 0.342 1 0.538 2.530 1 0.413 1.088 1 0.514 Zone where study was conducted (midlatitudes or subtropical). Dispersal mode OE 1.415 4 0.727 2.746 3 0.788 5.269 2 0.246 Typical method of seed dispersal (unassisted, ballistic, wind, water, and animal). Plant life form O 1.354 3 0.687 9.365 3 0.407 3.629 3 0.683 Tree, perennial herb, annual herb, and perennial grass. Existence of seed bank OE 0.815 1 0.245 0.642 1 0.645 5.438 1 0.075 Seed bank defined as seeds remaining in the ground for longer than one year.
38 Table 1-1. Continued. Plant longevity O 0.012 1 0.914 6.081 1 0.191 1.609 1 0.426 Perennial or annual. Plant longevity E 0.845 3 0.729 3.555 2 0.560 5.840 2 0.219 Number of years the plant lives (<1, 1-10, 11-100, >100 yrs). Seed origin O 10.604 1 0.005 3.093 1 0.359 0.079 1 0.868 Native or exotic. Seed presence O 7.777 1 0.007 0.986 1 0.610 0.167 1 0.807 Seed addition refers to the sowing of seeds in microhabitats where the species is known to be present. Introduction is the sowing of seeds of a species that may be native, but is not present at the time of sowing. Seed size category OE 9.915 5 0.044 66.407 5 0.002 33.926 5 0.010 Seed mass categories: 0 0.20, 0.21 0.50, 0.51 1.0, 1.01 2.00, 2.01 10.0, > 10.01 mg Continuous grouping variables Seed augmentation density (seeds m-2) O -5.754x 10-3 126 0.687 2.403 x 10-2 92 0.074 2.150 x 102 74 0.092 Density of seeds sown in treatment plots. Duration of study (days) O -4.348 x 10-4 126 0.05 6.617 x 10-5 92 0.838 -1.782 x 10-5 74 0.446 Number of days for which seeds were monitored for recruitment in studies.
39 Table 1-1. Continued. Maximum plant height (m) OE 2.110 x 10-2 119 0.304 -1.473 x 10-2 86 0.503 1.312 x 103 70 0.962 Seedling growth rate (day-1) E -7.384 x 10-1 36 0.262 -1.565 x 10-1 21 0.607 -2.667 x 10-3 19 0.990 Average first flowering (yr) E -1.815 x 10-2 22 0.606 -2.287 x 10-2 16 0.844 -9.346x 103 15 0.811 Age at which plants typically flower for the first time. Average fecundity (seeds plant-1) E 4.178 x 10-3 74 0.760 -2.409 x 10-2 62 0.276 -1.391 x 10-2 54 0.424 Average seed bank density (seeds m-2) E -2.138 x 10-2 73 0.309 -5.563 x 10-5 50 0.999 3.011 x 104 43 0.988 Average seed bank longevity (yrs) E -1.160 x 10-2 38 0.091 -6.682 x 10-3 23 0.672 -1.586 x 10-3 21 0.890 Number of years that seeds remain viable in the seedbank. Seed mass (mg) OE 2.477 x 10-2 120 0.056 1.71 x 102 84 0.305 1.157x 10-2 70 0.362
40 Table 1-2. Comparison of seed limitation (abs olute response) by grouping vari ables for A) undisturbed treatments (Eabs, U), B) disturbed treatments (Eabs, D), and C) the effect of disturbance on seed limitation (Eabs). The te st statistics for grouping variables are Q for categorical variables and R2 for continuous variables (see Met hods). All randomization tests were twosided tests, and therefore a P-value of 0.025 indicates statistical significance at th e 5% level. The code after the grouping variable indicates where data for the gr ouping variable were collected : O (original paper), E (external sources), and OE (data extracted from original papers when possi ble, but often supplemente d by external sources. Undisturbed (Eabs U) Disturbed ( Eabs D) Disturbance (Eabs) Grouping variable Test statistic ( Q or R2) df P Test statistic ( Q or R2) df P Test statistic ( Q or R2) df P Description Categorical grouping variables Habitat O 1.314 2 0.393 3.467 2 0.197 23.883 1 0.050 Habitat in which the study was conducted (intertidal, grassland, forest). Geographic zone O 1.761 1 0.085 4.593 1 0.109 1.559 1 0.121 Zone where study was conducted (midlatitudes or subtropical). Dispersal mode OE 6.195 4 0.621 4.832 3 0.532 2.734 2 0.473 Typical method of seed dispersal (unassisted, ballistic, wind, water, and animal). Plant life form O 11.927 3 0.285 11.698 3 0.324 8.092 3 0.226 Tree, perennial herb, annual herb, and perennial grass. Existence of seed bank OE 15.727 1 0.058 2.955 1 0.355 0.529 1 0.653 Seed bank defined as seeds remaining in the ground for longer than one year. Plant longevity O 6.422 1 0.209 9.423 1 0.162 6.728 1 0.120 Perennial or annual. Plant longevity E 82.347 3 0.088 11.448 2 0.195 6.170 2 0.158 Number of years the plant lives (<1, 1-10, 11-100, >100 yrs). Seed origin O 3.053 1 0.498 2.803 1 0.159 2.175 1 0.108 Native or exotic.
41 Table 1-2. Continued. Undisturbed (Eabs U) Disturbed ( Eabs D) Disturbance (Eabs) Grouping variable Test statistic ( Q or R2) df P Test statistic ( Q or R2) df P Test statistic ( Q or R2) df P Description Seed presence O 0.378 1 <0.962 0.711 1 0.561 0.508 1 0.469 Seed addition refers to the sowing of seeds in microhabitats where the species is known to be present. Introduction is the sowing of seeds of a species that may be native, but is not present at the time of sowing. Seed size category OE 19.526 5 0.362 10.299 5 0.641 7.043 5 0.546 Seed mass categories: 0 0.20, 0.21 0.50, 0.51 1.0, 1.01 2.00, 2.01 10.0, > 10.01 mg Continuous grouping variables Seed augmentation density (seeds m-2) O 175.485 126 0.039 494.252 92 0.012 412.711 74 0.038 Density of seeds sown in treatment plots. Duration of study (days) O -3.777 126 0.005 -8.870 92 0.036 -6.148 74 0.033 Number of days for which seeds were monitored for recruitment in studies. Maximum plant height (m) OE -124.799 119 0.301 -482.732 86 0.092 -427.478 70 0.126 Seedling growth rate (day-1) E -5196.236 36 0.076 -452.965 21 0.808 -197.931 19 0.917 Average first flowering (yr) E -209.715 22 0.414 -457.751 16 0.776 -497.532 15 0.379 Age at which plants typically flower for the first time.
42 Table 1-2. Continued. Undisturbed (Eabs U) Disturbed ( Eabs D) Disturbance (Eabs) Grouping variable Test statistic ( Q or R2) df P Test statistic ( Q or R2) df P Test statistic ( Q or R2) df P Description Average fecundity (seeds plant-1) E 85.744 74 0.403 225.028 62 0.388 193.522 54 0.416 Average seed bank density (seeds m-2) E 60.187 73 0.618 514.754 5 0.232 281.927 43 0.390 Average seed bank longevity (yrs) E -34.677 38 0.201 -71.397 23 0.45 -27.086 21 0.816 Number of years that seeds remain viable in the seedbank. Seed mass (mg) OE 9.221 120 0.909 -206.275 84 0.34 -150.727 70 0.343
43 Figure 1-1. Seed limitation for A) undisturbed plots ( UE) B) disturbed plots ( DE), and C) the effect of di sturbance on seed limitation ( E ), for each of the categorical grouping variables examined. Randomization tests were used to examine differences in effect size within the grouping variable clusters on the y-axis; significant hete rogeneity (P<0.05) among factors within a grouping variable (e.g., tree, pere nnial herb, annual herb, and perennial grass within life fo rm) is indicated by '*' and suggests significant differences in the effect size of seed limitation among the f actors. Bars give bias-corrected bootstrapped 95% confidence intervals. The number of species included in the analyses for each of the factors within a grouping variable is lis ted to the right of the c onfidence interval bars. Native and exo tic species have significantly differe nt effects sizes despite having overlapping conf idence intervals. This is because si gnificance tests were carried out with randomization tests, while confidence intervals were estimated with re sampling (see Methods).
44 Figure 1-2. Seed limitation for A) undisturbed plots ( UE) and B) disturbed plots ( DE), and C) the effect of disturbance on seed limitation ( E ), in relationship to seed mass, seed bank longevity, average seed bank density. The fitted lines represent weighted regressions to determine whether plant char acteristics explain variation in the effect size of seed limitation. Note that seed mass and average seed bank density are on a log10 scale. The P-value indi cating the significance of the regression is labeled in the upper right-hand corner of each plot.
45 CHAPTER 2 A NARROW NICHE FOR NEUTRAL PROCESSES IN THE RECRUITMENT OF AFROTROPICAL TREE SEEDLINGS Abstract Two processes, seed limitation and establishment limitation, have been proposed to explain how recruitment limitation might influence trop ical tree diversity. Seed and establishment limitation are generally treated as mutually exclusive, and evidence of either is used to bolster competing theories of community composition that are tightly associated with each. Both processes limit plant populations, but probably to di fferent degrees. The chal lenge is to evaluate their relative importance. Here we present resu lts of a large scale experiment designed to tease apart the relative strength of seed and establis hment limitation for five randomly-chosen tree species in an Afrotropical forest. We find that po st-dispersal differences in a speciesÂ’ ability to establish and survive more strongl y controls seedling recruitment than does the availability of seed. We also demonstrate that density-indepe ndent mortality constrains population size more than density-dependent mortalit y, unless seed densities reach de nsities similar to those found under canopies of reproductive a dults, at which point density-dependent mechanisms strongly limit seedling recruitment. We conclude that the process of seed limitation is relatively weak but can help plant populations overcome the exclusio n process of competition and niche partitioning by (1) getting propagules to open recruitment sites and (2) numerically overwhelming high densitydependent mortality at very high levels of seed arrival. Introduction Tropical rainforests harbor much of the wo rldÂ’s biotic diversity and provide globally critical ecosystem services, such as carbon sequestration, provision of pharmaceutical products and the regulation of water cycles. Most tr opical forests are thre atened by logging and
46 agricultural conversion. Conserving tropical forests and the ecologica l services they provide will demand an improved understanding of the proce sses that maintain high species diversity. Recruitment limitation, the failure of a species to successfully reproduc e and establish at a site, is likely to play a central role in mainta ining diverse community assemblages such as those of tropical forests (Chesson 1985; Hu rtt and Pacala 1995; Tilman 1994). By preventing superior competitors from becoming dominant, recruitmen t limitation slows the exclusion of inferior competitors from the community and allows gr eater numbers of species to co-exist. Two processes that act on the seed and seed ling stage of tropical tree recruitment are posited to drive recruitment limita tion and potentially foster tree diversity: lack of seeds (seed limitation) and lack of suitable micro-sites for recruitment (establishment limitation) (Clark et al. 1998b; Cornell and Lawton 1992; Hubbell 1986; Mulle r-Landau et al. 2002; Nathan and MullerLandau 2000; Schupp 1988). Strong seed lim itation indicates species abundances and distributions are restricted by seed dispersal. U nder this scenario, sites are Â“wonÂ” by default Â– or chance events of arrival Â– with li ttle regard to species characteris tics or competitive ability. By contrast, if species do not recruit randomly among sites, strong establishment limitation indicates species abundance and distributi on are restricted by characterist ics of the sites. Functional differences among species in how they respond to environmental heterogeneity structures community dynamics, with the most competitiv e species winning limited recruitment sites (Grubb 1977; Muller-Landau et al. 2002; Turnbull et al. 2000). Fueled by evidence in support of each (Dalling and Hubbell 2002; Duivenvoorden et al. 2002; Hubbell et al. 1999; Norden et al. 2009; Svenning and Wright 2005; Tuomis to et al. 2003; Uriarte et al. 2004) these contrasting paradigms to explain plant community diversity have jostled for predominance for nearly a decade.
47 It is likely however, that both seed and esta blishment limitation operate in natural systems and that community theory can be reconciled by viewing them as two processes that lie at opposite ends of a continuum (Clark et al. 2007; Poulsen et al. 2007) The challenge then, is to quantify their relative importance. Seed addition experiments provide an elegant tool for testing the relative importance of seed and establishmen t limitation. By experimentally adding seeds to plots (thus decreasing or removing the magnitude of seed limitation) and comparing seedling recruitment to control plots in which no seeds ar e added, one can directly quantify the degree to which populations are limited by seed availabi lity relative to competition and niche-based processes of establishment limitation. When seedling growth and mortality in experimental plots is followed over time, seed addition experiment s allow us to further isolate the specific mechanisms of establishment limitation. In particular, decoupling density-independent and density-dependent mechanisms of establishm ent limitation is important for understanding tropical tree diversity because some of the best supported diversity models invoke negative densityand frequency-dependent mortality (C hesson and Warner 1981; Condit et al. 1994; Connell 1971; Harms et al. 2000; Hubbell et al. 1 990; Janzen 1970). In each of these models, negative density-dependence is pr edicted to constrain locally abundant species, which opens space for otherwise less successful species, thus promoting co-existence by creating a rarespecies advantage. We conducted a large scale seed addition e xperiment in the northern Republic of Congo (Brazzaville) to determine to what degree tropical tree recruitment at the seedling stage is driven by seed and establishment limitation. We then examined the strength of seed limitation relative to densitydependent and density-independent mechanisms of establishment limitation. Most previous studies that examine recruitment limitation in tropical tree communities have been
48 carried out in Neotropical envir onments. Our study is the first to examine these issues in an Afrotropical ecosystem and thus will contribute to an understandi ng of the generality of several proposed mechanisms of tree species coex istence in tropical forest systems. Methods Overview We established 30 1-ha permanent plots in wh ich all trees were id entified and mapped and in which natural rates of seed ra in were recorded for one year. We then added seeds of five randomly selected species to 63 experimental stati ons. At each station, s eeds of all five species were augmented at seven dens ities ranging from 0 Â– 2000 times those observed in the ambient seed rain. By adding seeds at multiple densities and quantifying the relationship between seedling recruitment and conspecific tree density, we were able to examine the strength of seed limitation relative to densitydependent and de nsity-independent mechanisms of establishment limitation. We then monitored seedling recruitm ent every three months for two years. To evaluate the relative importance of seed and establishment limitati on as processes that constrain seedling population size, we a dopted a framework based on the concepts of realized and fundamental limitation (Muller-Landau et al. 2002) Realized limitation di rectly quantifies the degree to which seed and establishment limitatio n together determine recruitment, given actual establishment conditions following s eed addition. Fundamental limitation estimates the degree to which seed and establishment limitation separately prevent seedlings from achieving their maximum population size in the absence of constraints imposed by the other. Site Delineation and Characterization of Seed Rain Prior to seed augmentation we used satellite images to identify forest areas that contained dense mixed, terra firma forests in and around Nouabal-Ndoki National Park. From these
49 potential study areas, we randomly selected th e 30 study sites, spanning an area of over 3000 km2 (Appendix B, Figure B-1). The sites were separated by at least 2km to promote independence of samples. At each site, we delineated a 100 x 100 m (1-ha) plot and marked, mapped and identified all trees >10 cm diameter-at-breast-height (dbh). For each tree, we collected three voucher specimens for species ve rification, measured dbh, estimated height, and recorded its speciesÂ’ rege neration niche as defined by (Hawthorne 1995). To determine biologically relevant seed de nsities for seed additi on experiments and the pool of species from which to draw our focal specie s, we quantified the rate and diversity of seed rain in each site for one year prior to beginning the seed addition experiments. Rates of natural seed rain were quantified by capturing fruits and seeds in seed traps (21 per site, N=630; Appendix B, Figure B-2). Seed traps consisted of 1 x 1 m wooden frames with a canvas center elevated approximately 0.75 meters from the grou nd to avoid seed predation. All fruits and seeds falling into traps were collected, counted, and identified to species at two week intervals. We continued this seed rain study for an addition al year following initiation of the seed addition experiment, resulting in 2 years of seed rain data. To facilitate a broader generalization of resu lts from the seed addition experiment, we replicated across five randomly selected tree sp ecies. Species were chosen from a list of all naturally occurring tree species, with the constraint that species ha d to have least five seeds in the seed traps during the first ye ar of trap monitoring (N=277 speci es). Constraining the list in this way allowed us to collect sufficient numbers of seeds to conduct the experiment, while not biasing selection towards any partic ular species characteristic. Fo cal species varied substantially in terms of regeneration niche, dispersal mode, seed size, and relative abundan ce (Table 2-1).
50 Seed Sowing Experiments We established 63 seed additi on Â“stationsÂ” in random locatio ns in 21 of our 30 plots (3 stations per plot). We subdivided each seed addition station into 35 0.5 x 0.5 m Â“quadratsÂ” and sowed seeds of the five focal species at seven di fferent densities. Densities were multiples (0, 25, 50, 100, 200, 500, and 2000) of the mean natural seed rain density of each species observed in the seed traps the previous year. We added one seed per 0.25 m2 when the required augmentation treatment based on natural seed rain densities was less than one (Table 2-2). The highest augmentation level exceeded the greatest a nnual seed rain density in any single seed trap over two years of trap monitoring (Table 2-2) but is likely a reasonable re presentation of seed densities directly under large fruiting trees. Augmenting seeds over a wide range of densities allowed us to precisely describe the recruitment function and assure site saturation. Therefore, we could quantify the relative importance of seed versus establishment limitation over essentially any level of seed rain and coul d identify the biological importan ce of various levels of seed supply compared to other factors (density depe ndent and density independent mortality) in natural communities (Poulsen et al. 2007). Seed collection and sowing were conducted in th e second year of the study at the height of the fruiting season for each species. We empl oyed families of indigenous Mbenzl to widely search the forest for seeds (t otal search area ~ 500,000 hectares). The Mbenzl are huntergatherer, semi-nomadic forest people who have an intimate knowledge of the forest, including locations of even rare tree species. Approxima tely 40,000 seeds were collected, cleaned of pulp, and sown into seed addition plots. Following seed addition, seedling emergence and mortality were monitored every three months for the first two years of growth. We numbered each seedling and recorded height and number of leaves at each observation period.
51 Quantification of Seed and Establishment Limitation Realized limitation: As outlined above, the failure of a species to recruit at less than maximum density at any given location can be the result of either failure of seeds to arrive in the first place or the lack of suitable conditions for establishment upon arrival. In nature, one would expect a continuum of site quality and seed arrival rates that jointl y influence the relative strengths of seed and establishment limitation. Our seed addition expe riment quantifies how many sites are colonized by seedlings when the effect of seed limitation is decreased but all other limitations remain in the system. In effect, we quantified realized seed and establishment limitation for seedling emergence and survival to the second year of growth (adapted from Muller-Landau et al. 2002; Nathan and Muller-Landau 2000 who similarly qu antified realized seed and establishment limitation for adult trees). We can decouple the strength of rea lized seed and establishment limitation at each of our seed augmentation levels using a per seed recruitment effect size (Clark et al. 2007) as a measure of the relati ve strength of each process. The effect size is the difference between seedling densities in treatment and control quadrats (0 augmentation level), for the ith species (i=1Â…5) and the jth (j=1Â…7) augmentation level, standardized by the numbe r of seeds added to treatment (seed addition) quadrats: ij AUG ij CONT ij EXP ijS X X E, (2-1) where XEXP,ij is the number of seedlings in experimental quadrats, XCONT,ij is the number of seedlings in control quadrats, and SAUG,ij is the number of seeds added to treatment quadrats. This quantity is a measure of the number of new recruits, per seed added. We averaged effect sizes of different quadrats within the same plot to estimate a single plot level effect size for each species and seed augmentation combination. This e ffect size is expected to vary between 0 and 1,
52 on average, if density-dependence is weak Â– alth ough by chance control pl ots could contain more seedlings than seed addition plots at the end of the experiment (in which case E would be negative). If density-dependence is negative and sufficiently strong to le ad to overcompensation, E again could be negative. If density-dependen ce is positive and strong, E could exceed unity. In the end, none of these situa tions occurred over the course of this study; E always varied between 0 and 1. With this metric of effect si ze, realized seed and establishment limitation are inversely related, occupying opposite ends of a grad ient (Clark et al. 2007 ; Muller-Landau et al. 2002). Realized seed limitation is calculated by our effect size E. Realized establishment limitation is equated as 1E. By determin ing the position of a plant population along this gradient, we can compare the relative strength s of seed and establishment limitation and how they are influenced by specific character istics of the site and/or species. Because seeds for all species were added at eq uivalent levels relativ e to species specific patterns of seed rain, and were a dded to randomly selected sites, all species and sites should have equal recruitment probabilities under neutral models of commun ity diversity. Significant species-by-plot interactions will therefore provi de evidence in support of niche-based models. To examine the degree to which th e strength of realized seed and establishment limitation vary as a function of seed addition, plot, and species, we fitted and evaluated generalized linear mixed models (GLMMs) for E, with a binary error di stribution, a logit-link, a nd treating species and plot as a crossed random interact ion effect. We also fit GLMMs to the number of seedlings per seed addition plot using a lognormal Poisson erro r distribution to account for over-dispersion and a log-link, with a crossed random effect of speci es and plot. Bayesian inference with Markov Chain Monte Carlo (MCMC) simulation was used to estimate posterior distributions of model parameters and test for significance. For both mo dels, we used weak, normally distributed priors
53 for fixed and random effects and uniform priors on the precisions of the variance components. We fit our models using the software WinBUG S v. 1.4.1 (Spiegelhalter et al. 2003). For each model, we achieved convergence after 50,000 iter ations (the Â“burn-in Â”) and based summary statistics on an additional 25,000 iterations. We ran three chains to m onitor convergence based on variance components of multiple sequences and assessed convergence by visual inspection and with Gelman-Rubin statistics from the R c ontributed package, coda (Plummer et al. 2005). For point estimates, we extracted the means of the posterior distributions and we derived 95% credible intervals based on the observed percentiles from the MCMC replicates. Fundamental limitation: An alternative approach to examining the re lative importance of seed and establishment limitation to seedling recruitment is to estimat e the degree to which each process would be limiting in the absence of the other a measure of fundamental seed and establishment limitation sensu Nathan and Muller-Landau (2000). We de fine fundamental seed limitation as the difference between the number of seedlings that w ould recruit to a site if the seed supply were limitless, but post dispersal constraints to recruitment (e.g. density dependent and independent mortality) are absent (Poulsen et al. 2007), relative to the numbe r of seedlings that recruit under ambient conditions. An advantage of estimating s eed and establishment limitation in this way is that it allows us to explain why seedlings ar e not achieving their Â“funda mentalÂ” optimum (i.e., their potential maximum population size) and offers a degree of pr edictive ability regarding how species and communities should behave as the lo cal environment changes (McGill et al. 2006). Using the Â“fundamental limitationÂ” framework to examine the relative importance of seed and establishment limitation to seedling recruitment, and decomposing establishment limitation into density dependent and inde pendent components, assumes that seedling survival is densitydependent and recruitment has an upper limit (i .e., the relationship between seed addition and
54 seedling recruitment must be nonlinear). To estimate fundamental limitation, we first examined whether the assumption of non-lin earity between seed augmenta tion and seedling recruitment was reasonable. To do so, we fit two simple models of seedling recruitment based on the asymptotic Beverton-Holt function, (2-2) where is the number of recruits (seedlings) that em erge from seed input, In this study, seed input was comprised of the seeds sowed in the experiment and the natural seed rain, where was a multiple of is the proportion of seeds that recruit in the absence of density effects and is th e maximum number of seedlings. The second model, which is a special case of the first, is a linear model derived by letting Thus, the first model includes seed limitation, density independence and density dependence, whereas the second model excludes density dependence. By comparing these two models we test whether density influences seedli ng recruitment. We also tested the two other possible nested models (seed limitation only and no density independent mortality), but because neither of these models fit the data well and we exclude them from further analysis. We fitted two models to seedling recruitment data for each species separately and all species combined, at 3 and 24 mont hs after seed addition. We estimated model parameters in using Metropolis Hastings optimization and derived 95% confidence intervals by sampling parameter values directly from the algorithm. We used AkaikeÂ’s information criterion (AIC) to compare the Beverton-Holt and linear models, with a difference of 4 AIC points signifying that one model fit better than anothe r (Burnham and Anderson 2002). We then estimated fundamental limitation using the parameter estimates obtained from the Beverton-Holt model (Poulsen et al. 2007). Th e amount of limitation imposed by a single
55 process is calculated by subt racting the number of seedlings that occurred under ambient conditions from the number that would have occu rred if the limiting process were removed. Therefore, seed limitation, is the difference between the number of seedlings that would emerge if seed supply were limitless and the number of recruits under ambient conditions ( ). Limitation from density-indepen dence and density-dependence can be found by setting and respectively, a nd comparing these results to seedling recruitment under. Thus, limitation from de nsity-independent, or density-dependent losses, are calculated as: and Establishment limitation is represented by the removal of both density-independent and density-dependent loss and can be found by setting and so that Removal of establishment limitation ensures that all dispersed seeds survive to recruit Â– that seed input alone determines local abundance. Results Averaged over all species, our results suggest that realized establ ishment limitation more strongly limits tropical tree recrui tment than does realized seed limitation. From approximately 40,000 sowed seeds, 6389 (17 %) seedlings recrui ted after 3 months and 2303 (6.1%) recruited after 24 months. Per seed recruitment effect sizes were initially low (E = 0.21 0.28 at 3 months) and decreased with time (E = 0.09 0.20 at 24 months). Fundamental limitation demonstrated that even under optimal conditions, seed limitation rapidly decreases in strength and becomes less important than establishmen t limitation at low seed densities commonly observed in natural systems (< 5 times the mean seed rain density which corresponds to < 1 seed/m2). The position of a given species along th e establishment seed limitation continuum
56 and the total number of seedlings that recruite d varied among sites and species (species-specific responses available in supplementary material). The strength of realized establishment limitati on was strong at all s eed addition levels, but increased in strength with both seed augmentati on level and time, indicating that negative density dependence at least partially regul ated plant population size at the seedling stage (Figure 2-1). Across species, seed augmentation level negativ ely influenced the probability of seedling recruitment (per seed effect size, E) following th ree months and 2 years of growth (Table 2-3). However, at the scale of our hectare plots, we identified no significan t relationship between seedling recruitment and the density of conspeci fic adults, indicating obs erved density-dependent mortality was a function of dir ect and/or indirect influences of seeds and seedlings on one another, not the influence of conspecific adul t density (see Appendix B, Figures B-5 and B-6, Tables B-1 and B-2 for sp ecies-specific results). Similarly, we evaluated eviden ce of density-dependence by co mparing models of seedling recruitment. The model that included density-d ependent mortality provi ded marginally better fits to the data than the model excluding it. Fund amental limitation analysis suggested that in the absence of all other forms of limitation, density-dependent mechanisms can importantly limit seedling establishment (Figure 2-1; Appendix B, Figure B-7), although this effect was less important than density-independent mortality until seed densities achieve levels similar to those observed naturally only under parent tree canopies (Clark et al. 2004). Despite the observed negative effects of seed addition (density dependence) observed in this study, the addition of seeds at extremely high seed densities (e.g. 2000-times background density, a level that surpass those observed in seed traps) resulted in a significant increase in total
57 numbers of recruited seedlings, though th is pattern may be driven, in part, by Myrianthus arboreus (Figure 2-1b; Table 2-3, Appendix B, Figure B-5). Discussion Several specific insights to trop ical forest dynamics emerge from this study. First, at local scales (1-ha plots), niche based processes associated with postdispersal seedling establishment more strongly influenced seedling recruitment than did seed availabi lity. The significant species-by-plot effect reflects post-dispersal, specie s-specific differences in ability to establish at a site given equal seed arrival. When coupled w ith the strong effects of realized establishment relative to seed limitation, our results provide evidence in supp ort of niche-based models of community diversity. Second, seed limitation plays a prominent role in species recruitment at two points assuring a species ge ts to a site and overwhelming density-dependence when seed densities are especially high. At low to modera te rates of seed arriva l, population size is controlled by strong density-independent and rela tively weaker density-dependent mechanisms of establishment limitation. Third, negative de nsitydependent mortality fails to dominate density-independent mortality until extremely high seed densities, suggesting that negative density-dependent mortality importantly regul ates seedling population size only when seed supply reaches densities that mirror or surpass t hose observed under parent trees. The cross-over value at which this occurs is similar among speci es. However, contrary to many theories put forth to explain patterns of species diversity (Chesson 2000; Ch esson and Warner 1981; Connell 1971; Janzen 1970; Wright 2002), we found no evidence that these effects are directly influenced by plot level densities of consp ecific adults. We discuss additi onal support of this conclusion with respect to distance to conspecific adul ts in chapter 3. Finally, even though densityindependent factors killed seeds and seedlings at low and moderate densities and densitydependent factors killed seeds and seedlings at extremely high densities, some individuals
58 nonetheless escape these processes and survivew ith total number of seedlings increasing with the addition of seed. Detailed studies of the mechanisms causing post dispersal seedling mortality were beyond the scope of this study. However, contrary to current theories that posit tropi cal forests are strongly structured by density-d ependence and seed limitation, our results suggest densityindependent factors and post-dispersal niche partitioning or competition more importantly prevent seedling populations from reaching their ma ximum densities. Future research aimed at understanding tropical forest dive rsity should direct increased a ttention toward quantifying the complex range of factors responsible for post di spersal limitations to s eedling establishment. Once mechanisms are clearly defined, the field of biodiversity science ca n move from one that describes pattern to a predictive science armed with the information required to mitigate and resolve threats to tropical forest diversity.
59 Table 2-1. Ecological characterist ics of focal tree species sel ected for use in this study. Regeneration guilds follow Hawthor n, 1995. N.P.L.D = Non-pioneer, light demanding. Animal dispersal categories in clude: P= primate (arboreal primates, chimpanzees and gorillas), B= bird, E= elepha nt. Seed sizes are averaged lengths (L) measured from 100 seeds of each species. Average conspecific density was estimated from 30 1-ha plots in which adults >10c m dbh of all tree speci es were measured, mapped, and identified to species. Species Regeneration Guild Dispersal mode Average seed size (cm) Average conspecific density/ha (> 10cm dbh) Pancovia laurentii Shade bearer Animal (P) L = 1.1 2.37 Staudti kamerunensis N.P.L.D. Animal (P,B) L = 1.9 0.57 Manilkara mabokeensis Shade bearer Animal (P) L = 1.4 1.67 Myrianthus arboreus Shade bearer Animal (P,B,E) L = 2.1 3.96 Entandophrag ma utile N.P.L.D. Wind L = 0.8 1.17
60 Table 2-2. Ambient seed rain density and seed addition levels used fo r each species in each subplot (N = 63). Seed addition de nsities for each treatment level were de termined as a function of magn itude (e.g. 25X) greater than the seed rain density, estimated from seed traps (N=630) set up with in each forest plot Pala = Pancovia laurentii, Stka =Staudtia kamerunensis, Mama =Manilkara mabokeensis, Myar=Myriant hus arboreus, Enut =Entandophragma utile Seed rain (seeds/ m2) Seed augmentation (seeds/ 0.25 m2) Species Mean density 1 year Mean combined density 2 years Range 2 years 25 X 50 X 100 X 200 X 500 X 2000 X Total Pala 0.1 0.20 0-39 1 (4/m2) 2 (8/m2) 3 (12/m2) 5 (20/m2) 13 (52/m2) 50 (200/m2) 4662 Stka 0.11 0.26 0-16 1 (4/m2) 2 (8/m2) 3 (12/m2) 6 (24/m2) 14 (56/m2) 55 (220/m2) 5103 Mama 0.15 0.58 0-94 1 (4/m2) 2 (8/m2) 4 (16/m2) 8 (32/m2) 19 (76/m2) 75 (300/m2) 6867 Myar 0.4 0.53 0-114 3 (12/m2) 5 (20/m2) 10 (40/m2) 20 (80/m2) 50 (200/m2) 200 (800/m2) 18144 Enut 0.1 0.1 0-17 1 (4/m2) 2 (8/m2) 3 (12/m2) 5 (20/m2) 13 (52/m2) 50 (200/m2) 4662
61 Table 2-3. Results from generalized linear mi xed model (GLMM) analyses of effect size E and total number of seedlings as a function of seed addition level and conspeci fic tree density, at (a) three months a nd (b) two years after seed addition. CI are 2.5% and upper =97.5% credible intervals. *Indicates significance (critical intervals do not overlap zero). A. Results of GLMM for all species combined fo llowing 3 months of growth (seed to seedling transition) Response Predictors Mean effect SD Lower CI Median Upper CI E (3 mo) Conspecific 0.11690.3579-0.6605 0.12370.8109 Level -0.18680.04488-0.2747 -0.1864-0.09821 Plot x Species random effect* 3.3260.2622.859 3.313.884 Individual random effect* 1.4730.057011.365 1.4711.588 No. of Seedlings (3 mo) Conspecific 0.25610.1755-0.09232 0.25280.6066 Level* 1.0340.034770.9652 1.0341.101 Plot x Species random effect* 1.2070.04311.125 1.2061.294 Individual random effect* 1.860.16021.571 1.852.198 B. Results of GLMM for all species two years after seed addition E (2 yrs) Conspecific 0.21980.4999-0.7559 0.2341.215 Level* -0.22080.0585-0.3355 -0.2211-0.1054 Plot x Species random effect* 5.0690.42744.308 5.0415.979 Individual random effect* 1.7040.086181.541 1.7021.878 No. of Seedlings (2 yrs) Conspecific 0.19870.2989-0.3664 0.19330.782 Level* 0.94650.044960.8596 0.94621.036 Plot x Species random effect* 1.3810.065061.258 1.381.512 Individual random effect* 2.9480.26912.47 2.9313.521
62 Figure 2-1. (A) Per seed recruitment effect size, E, (realized seed and establishment limitation). E varies from 0-1 with 1 re presenting complete realized seed limitation and 0 representing complete establishment limitation. These relatively low effect sizes (E < 0.5) indicate this natural forest system is more st rongly establishment than seed limited. (B) Total number of seedlings averaged over 5 species ( Pancovia laurentii, Staudtia kamerunensis Manilkara mabokeensis, Myrianthus arboreus, Entandophragma utile) as a function of seed augmentation level, fo r the first three months (black) and two years (gray) of seedling growth. Weak seed limitation observed in 1A results in a gradual, but significant increase in total seedling numbers at very high s eed densities (Table 2-3).
63 Figure 2-2. Fit of the two ca ndidate recruitment function mode ls to seed augmentation data pooled for all five species at (A) 3 mont hs after sowing and (B) two years after sowing. The dashed line represents the no density-dependent-limitation model (fitting P 0 and S amb) and the solid line represents the seed-limitation, densityindependent-limitation, and densitydependent-limitation model (fitting P 0, R max, and S amb). Level of seed augmentation is a multiplying factor of ambient densities observed in seed traps for each species during the first year of this project. For all species, the full Beverton-Holt model provides an improved fit to the linear model (see Appendix B, Table B-2), providing evidence of density dependence.
64 Figure 2-3: Results of limitation analysis for a ll species combined at (A) three months and (B) 2 years following seed augmentation. The blue arrow represents the crossover point at which establishment limitation more strongly limits recruitment than seed limitation. The gray arrow represents the point at wh ich density-dependence more strongly limits recruitment than density-independent mechanisms of mortality. The importance of fundamental seed limitation exceeds that of fundamental establishment limitation only at very low seed densities ; with crossover points occu rring at 5.2 (3 months) and 4.2 (two years) times ambient seed conditions These values are well within the range of natural seed rain densities observed acr oss this study site (Table 2-2). Densityindependent mechanisms of seedling mo rtality more strongly contribute to establishment limitation than do density-dependent mechanisms of mortality until seed densities reach approximately 236 (3 months) and 217 (2 years) times the average ambient seed densities. Because these values roughly mimic those often identified directly under parent trees but exceed the observed seed rain densities for most species, we suggest density-independe nt factors limit seedling recruitment at most Â“naturalÂ” seed densities but dens ity dependent mechanisms likely control seedling population size at the very hi gh seed densities observed under fruiting canopies.
65 CHAPTER 3 TERRESTRIAL MAMMALS, MORE TH AN ENVIRONMENTAL FILTERING OR NEGATIVE DENSITY-DEPENDENCE, DRIVE PATTERNS OF TROPICAL SEEDLING RECRUITMENT Abstract Quantifying mechanisms responsible for postdispersal seed and seedling mortality is critical to understanding tropical forest diversity. Factors posited to constrain successful emergence and survival of seedlings include light availability, soil fertility, competition, density dependence, seed predation and herbi vory. To examine their importance in explaining patterns of tropical seedling recruitmen t in an Afro-tropical forest, we conducted seed addition experiments for five randomly selected tree spec ies in each of 30 heterogeneous study sites. We evaluated the strength and relativ e importance of these mechanisms at two stages: the seed-toseedling transition, and seedling survival to the second year of growth. We conclude that seedling recruitment in the Congo Basin is most st rongly dictated by generalist vertebrate seed predators and herbivores, with abiotic environmental filtering and density-dependence playing secondary roles. Our study also provides support fo r niche-based theories of tropical tree species coexistence, with species e xhibiting highly variable respon ses to naturally occurring environmental characteristics among sites. Cont rary to predictions of the Janzen-Connell hypothesis, seed and seedling r ecruitment were not related to the distance or density of conspecific adult trees. Introduction A central question in community ecology is : What processes control local species diversity? This questio n has been particularly compelling for tropical tree communities, where hundreds of species can co-exist in a single hectare (De Oliveira and Mori 1999; Valencia et al. 1994). Though many mechanisms have been proposed to explain such high diversity of trees
66 (Givnish 1999; Hubbell 2001; Tilman and Pacala 1993; Wright 2002), only two enjoy substantial empirical support: (1) niche differentiation associ ated with micro-topography and deterministic tradeoffs among species (Chase and Leibold 200 3) such that differe nt species perform differentially at different points along environmental microor m acrogradients; and (2) density and frequency-dependent mechanisms that lead to higher survival of locally rare species (Wright 2002). Niche differentiation occu rs when functional differen ces among species lead to differences in their competitive rankings across heterogeneous environments, with trade-offs usually determining where a particular species does best. Negative density-dependence occurs when nearby conspecifics reduce individual recr uitment probabilities, thus facilitating coexistence by opening space for otherwise less successful or less common species. Support for the importance of niche differen tiation has emerged from studies emphasizing differences among species differences that can o ccur at various life history stages (Ashton 1993; Clark and McLachlan 2003; J ohn et al. 2007; Kobe 1999; Mont gomery and Chazdon 2002; Potts et al. 2002; Svenning 2001; Wright 2002). Altho ugh habitat specialization among tree species can theoretically operate at every stage, the paucity of resources on which adult trees can specialize suggests that ni che partitioning, if it occurs, is most likely at early life-history stages (Grubb 1977). Yet, most studies exploring habita t specialization have ta rgeted adult trees along environmental gradients of light, soil water and nutrient availa bility (Aiba et al. 2004; Brokaw and Busing 2000; Canham 1989; Cannon and Leighton 2004; Clark et al. 1998a ; Davies et al. 1998; Denslow 1980; Gunatilleke et al. 2006; Harm s et al. 2001; Plotki n et al. 2000; Svenning 1999; Tateno and Takeda 2003; Valencia et al. 2004; Webb and Peart 2000). Few studies have explicitly examined the degree by which niche-partitioning and environmental filtering influences the distribution and abundance of se edlings (Comita et al. 2007; Webb and Peart
67 2000). Seedling establishment is a crucial filter in population persistence, and niche differences at a small spatial scale (e.g., in response to mi cro-topographic features), which are not important in mature adults, could be essential determinants of seed germination and seedling survival and growth. If niche partitioning o ccurs at the seedling stage, se edlings should differentiate by specializing in particular combinations of li ght, soil, water and nutrients, beneath and across canopy openings and within the forest understory (Baillie et al. 1987; Bloor and Grubb 2003; Russo et al. 2008). Studies supporting a central role of density and frequency-dependent mechanisms also emphasize the importance (albeit in a different way) of the early life histor y stages -seed arrival and seedling recruitment -for sp ecies co-existence. One of the leading hypotheses, the JanzenConnell hypothesis (Connell 1971; Janzen 1970), posits that seed s dispersed farther away from parent plants have higher survival rates th an those dispersed under parent plants, where conspecific seed density is greatest, because such seeds are able to escape host-specific pests and predators Â– creating a form of spatially-mediat ed negative densityand frequency-dependence. This spatially structured mortality should lead to rare-species advantage because the space or resources freed by density-dependent deaths are then exploited by less-common species. Numerous studies have documented suppor t for the Janzen-Conne ll Hypothesis (JCH) by demonstrating disproportionate seed and seed ling mortality from in sects, pathogens, or vertebrates near parent trees (Augspurger 1984; Clark and Cl ark 1984; Gilbert et al. 2001; Hammond and Brown 1998; Packer and Clay 20 00; Webb and Peart 1999). Furthermore, pervasive negative densityand frequency-dependen ce at early life history stages is sometimes correlated with increased sp ecies richness of seedlings (Harms et al. 2000).
68 Taken together these findings suggest that multiple mechanisms of post-dispersal seed and seedling mortality underlie which sp ecies will recruit in a given lo cation. Understanding tropical forest diversity requires quantifying and teas ing apart the relative importance of these mechanisms, a task most directly acco mplished through large-scale experiments. The successful establishment of a tree from the seedling stage necessitates overcoming two consecutive filters (1) seedling emergence (the transition from seed to seedling) and (2) seed ling survival. Factors posited to constrain the successful emergence and surv ival of seedlings at early life history stages include light availability (M ontgomery and Chazdon 2002; Nicotra et al. 1999), soil fertility (Fine et al. 2004; Hall et al. 2003; Palmiotto et al. 2004), competition (Paine et al. 2008), density dependence (Harms et al. 2000; HilleRisLambers et al. 2002), seed predation and herbivory (Jones et al. 2008; Paine and Beck 2007; Rao et al. 2001). It is likel y that each of these factors differentially influences seedling emergence and survival and that they vary and co-vary in complex ways, both spatially and temporally. Ho wever, the relative role s of these factors and the degree to which species-specif ic responses to them result in predictable patterns at the seedling stage remain untested. We examined mechanisms that explain patterns of tr opical seedling recruitment in an Afrotropical forest. To do so, we conducted seed ad dition experiments for 5 randomly selected tree species across 30 heterogeneous stu dy sites in mature forest of th e Republic of Congo. We relate seedling establishment and survival following seed addition to (1) abioti c variables posited to influence seedling recruitment (light availabilit y, soil structure and fert ility) and (2) ecological mechanisms proposed to limit s eed and seedling recruitment ( seed predation, herbivory and density dependence), with a particular emphasis on the Janzen-Connell predic tion that patterns of recruitment should be negatively influenced by de nsity of and distance to conspecific adults. We
69 evaluate the strength and relativ e importance of each of these m echanisms at two stages: the seed-to-seedling transition, and seedling survival to the second year of growth. Methods Study Area This study was conducted in the north of the Republic of Congo (Brazzaville), in Nouabal-Ndoki National Park (NNNP) and the Ka bo forestry concession (Figure 3-1). The Republic of Congo is known for its relatively intact forest system, rich in flora and fauna. The region is characterized as tropical lowland forest with highly weathered sandstone, quartzite, and schist bedrock, overlain in plac es by ancient basin alluvial de posits that have formed welldeveloped soils (Lanfranchi and Schwartz 1991). The relief of the site is generally flat, with altitude varying between approximately 350 and 400 meters. The climate is dominated by a pronounced dry season, typically be ginning at the end of November and extending through early March. The mean annual rainfall is 1700 mm and highly seasonal. Minimum and maximum average temperatures range between 21.1 21.9 C and 26.5-26.8 C, respectively (unpublished data, Bomassa Research Station). The region is charac terized by 7 distinct vegetation types (Harris 2002), with mixed species terra firma forest occupying 70% of the area (Laporte 2002). The forests of NNNP have never been commercially logged, although huntergather human populations have inhabited th e region for approximately 40,000 years and iron smelting sites, which can seriously degrade forest habitats, have been found in the region that date as early as 800 BC (Lanfranchi et al. 1998 ; Zangato 1999). The Kabo concession was selectively logged (< 2 trees/ha; CIB management plan 2006) approximately 30 years ago, and is exploited for non-timber forest products by a population of approximately 3000 people. Combined, the NNNP and Kabo concessions provid e a contiguous yet heterogeneous landscape
70 with which to evaluate how differences in biotic and abiotic conditions in fluence patterns of seed and seedling recruitment. Site selection We used satellite images to identify forest areas within the NNNP and Kabo concessions that contained dense terra firma forests. From these potential study areas, we used the geographic survey design component of the Di stance 4.1 software (Thomas et al. 2006) to randomly select 30 sites, sp anning an area of > 3000 km2. Sites were separated by at least 2 km to promote independence (Figure 3-1). At each site, we delineated a 100 x 100 m (1-ha) plot (Figure 3-2) and marked, mapped and identified al l trees >10 cm diameter-a t-breast-height (dbh). For each tree (N = 11,360), we collected three voucher specimens for species verification, recorded dbh, estimated height, and the species Â’ regeneration niche as described in Hawthorn (1995). Experimental design To evaluate the relative importance of mechan isms that influence seedling recruitment at (1) the transition from seed to seedling and (2) seedling establ ishment to the second year of growth, we randomly established 63 Â“stationsÂ” in 21 of our 30 mapped study sites. Into each station, we sowed seeds of five randomly-selected, tree species ( Pancovia laurentii, Staudti kamerunensis, Manilkara mabokeensis, Myrianthus arboreus, Entandophragma utile ; Figure 33). Species were chosen from a list of all natura lly occurring tree species that recorded at least five seeds in the first year of a concurrent seed rain study (N=277 speci es; Chapter 2 of this dissertation). Constraining the list in this way allowed us to collect sufficient numbers of seeds to conduct the experiment, while not biasing selection towards any particular species characteristic. These focal species varied in te rms of regeneration niche, dispersal mode, seed size, and relative abundance (Table 3-1), and adult individuals of all spec ies co-exist across the
71 study site (Table 3-2). Random species selection facili tates the generalization of our results to the broader tree community, though we recognize the small sample size requires we do so with caution. Each station was divided into 60, 0.5 x 0.5 m qua drats with 0.5 m separating each quadrat to provide access by field crews (N = 180 quadr ats per plot and 1780 qua drats total). Each quadrat received one species of seed in one of se ven different densities. Seeds were scattered on the soil surface. By augmenting seeds at multi ple densities we were able to experimentally evaluate the role of density re lative to other factors (light, soil, seed and seedling predation, herbivory, etc.) that likely influence seedling recruitment and mortality. Seed augmentation densities varied by species, each a multiple of the natural seed rain density (0, 25, 50, 100, 200, 500, and 2000 times observed seed rain over the prev ious year; see Chapter 2 for details). Following seed addition, seedling emergence and mortality were monitored every three months for two years. We numbered ea ch seedling and recorded height condition, and number of leaves at each observation period. Environmental Variables Light availability We took hemispherical photographs at the cen ter of each seed addition station using a Nikon Coolpix 5000 camera with a Nikon Fisheye Convertor FC-E8 lens. To avoid overexposure by direct sunlight, phot ographs were taken 30 cm above the ground, early in the morning (6:00-8:00am), late in the afternoon (1600-17:30h), or on over cast days (Montgomery and Chazdon 2002). Photographs were analyzed using the Gap Light Analyzer (GLA Version 2.0; Frazer et al. 1999). We re lated seedling emergence and es tablishment to estimates of transmitted diffuse light (Figure 3-4) which vari ed significantly among stations and plots (F = 3.72; Df = 20,42; p = > 0.000).
72 Soil sampling and analysis To determine how differences in soil com position and nutrient availability may affect seedling recruitment and mortality, soil sample s were collected at three randomly-selected locations from each station, using a soil probe (size: 2.85 cm x 83 cm) at 15 cm depth. Samples were weighed (Â“wetÂ” mass), then air dried and weighed again (Â“dryÂ” mass) prior to shipping. Sub-samples were pooled as a composite of soil sample for each station. Soil analysis included soil characteristics (% sand, clay, and silt) and nutrient avai lability analysis (N,P,K, Al, Ca, Mg, Mn) and pH. All analyses were conducted by the IFAS Extension Soil Testing Laboratory University of Florida, USA. We extracted available cations and P using the Mehlich III extractant solution (Tran and Si mard 1993). Elemental analysis for the cations and P was done on the Mehlich-III extracts by using inductively coupled pl asma (ICP) spectroscopy (EPA Method 200.7). We extracted N as NH4 andNO3 -. Nitrogen was estimated colorimetrically (EPA Method 353.2) using a Technicon II Auto-Analyzer. The Kjeldahl method was used for the determination of Total N in soil samples (Hesse 1971). Soil pH was measured in an AdamsEvans Buffer solution made up of one volume of soil diluted in 2 volumes of water. Subsoil samples were analyzed for soil texture, using a hydrometer method (Day 1965; Sheldrick and Wang 1993). We used Principal Components Analysis (PCA) to identify major tre nds in the soil data among our 63 stations and to reduce the number of variables describing soil factors for inclusion in further statistical analysis. The first PC axis explained 26.7% of the to tal variance in soil data and was strongly correlated with soil texture (f ractions of clay, sand, a nd silt), total N, and exchangeable cations (Figure 3-5); these parameters are strongly associated with soil fertility (Laurance et al. 1999). The second PC axis e xplained an additional 17.5% of the variance and was most strongly correlated with pH and phos phorous. PC axis 3 explained an additional
73 13.5% of the variance and was most strongly correl ated with Fe and aluminum. Combined, these three axes explained 57.5% of the variance in soil conditions. Hen ceforth, we refer to the first, second, and third PCA axes as soil PC1, PC2 an d PC3, respectively. Significant differences among plots were identified for PC1 (F = 33.49; DF = 18,38 ; p = > 0.000), PC2 (F = 11.65; DF = 18,38 ; p = > 0.000), and PC3 (F = 18.73; DF = 18,38 ; p = > 0.000), Ecological Variables Herbivory and seed predation To quantify the role of s eed predation and herbivory as potential post-dispersal mechanisms limiting seedling establishment, we conducted seed addition experiments with caged treatments for three of the five tree species ( Entandophragma angolense,Manilkara mabokeensis, and Myrianthus arboreus). We were limited to three species because of the logistical constraints of constr ucting and carrying cages to remote forest sites. By conducting seed addition experiments with both caged and uncaged treatment s at each seed addition level and site combination, we are able to disenta ngle seed mortality resulting from vertebrate predation and herbivory from seed and seedling mo rtality caused by specif ic characteristics of the micro-site. By replicating each seed addition level with a caged treatment, we were also able to disentangle the degree to which vertebrate s eed predation and herbivor y vary with density. Cages constructed for this experi ment excluded vertebrate seed predators and herbivores. They do not allow us to directly examine the degree to which soil pathogens or invertebrate pests on seeds and seedlings may limit seedling recruitment. Density and distance effects Â– evaluating Janzen-Connell The effect of density on seedling recruitment was evaluated at two time steps for each species. First, we examined how seed density influenced the probability of transitioning from seed to seedling by calculating the proportion of seedlings that recr uited into each quadrat as a
74 function of seed augmentation density. We then examined how the density of emerged seedlings influenced the probability of surviving to the second year of growth by calculating the proportion of seedlings that survived to the end of the experiment as a function of the maximum number of seedlings that emerged within the same quadrat. These values were used as response variables for statistical analyses (see below). To unde rstand the degree to which the Janzen-Connell spacing mechanism might influence seed and seed ling recruitment, we measured the distance between each seed addition stati on and the nearest conspecific a dult of each of the five focal species. We also measured the density of adu lts of the same conspecifics within each one hectare plot. Data Analysis Relative Importance of Mechanisms that Limit Seedling Emergence and Survival To examine the relative importance of envi ronmental variables (light and soil) and ecological mechanisms (density, seed preda tion and herbivory, and Janzen-Connell effects Â– distance to and density of conspecific adults) to seedling recruitment and survival, we fitted and evaluated generalized linear mixe d models (GLMMs) to (1) the proportion of seedlings that emerged as a function of the number of seeds adde d to a given quadrat an d (2) the proportion of seedlings that survived as a f unction of the maximum number of seedlings emerged. Each seed addition quadrat was treated as a sampling unit. Our full set of variables for these models included: diffuse light transmission, soilPC1, so ilPC2, soilPC3, seed augmentation level, sitelevel conspecific tree density, and distance to ne arest conspecific adult. To make parameter estimates comparable across explanatory variable s, we standardized a ll continuous explanatory variables by subtracting the mean and dividing by the standard deviati on to yield a Z-score (Gelman and Hill 2007). We ran th ree sets of models. First we examined the effect of the covariates on all species by anal yzing the data of all five spec ies together and excluding the
75 effect of caging. Second, we examined the eff ect of caging by running the same models for the three species for which the caging treatment was applied and included the effect size for caging as an additional variable. Finally, we examined the effect of the envi ronmental and ecological covariates on each species taken alone. We fit all models with a binomial error distribution and a logit-link. For species-level analys es we included plot as a random effect. For analyses of all five species together, we applied a speci es-by-site random effect. We used Laplace approximation (lme4 package) for maximum likeli hood estimation of the pa rameters and tested the statistical significance of fixed effects w ith Wald Z-statistics (Bolker et al. 2009). All statistical analyses were performed in R 2.7.2 (R Development Core Team 2005). Results Across all species, densities and caging tr eatments, a total of 10,399 (22.3%) seedlings emerged and survived to three months of gr owth following the addition of 46,620 sown seeds. Of these, 3,355 (7.2% of all seeds and 32.3 % of all emerged seedlings) su rvived the first two years of the study. Stated differently, 36,221 (77.7 %) of seeds sown into pl ots died within three months of seed sowing, and 7,044 (67.7%) of the s eedlings that survived through the seed to seedling transition had died by th e end of the second year. Analysis of all species together with a GL MM resulted in a large effect of the plot-byspecies interaction effect relative to other factors at both life hist ory stages (Table 3-3 and 3-4; Appendix C, Table C-1). The large variance am ong species and its dependence on plot identity suggests that niche partitioning ex plains patterns of s eedling emergence and survival. Hence, we focus our examination of the specific mechanis ms driving seedling emergence and survival on a species-by-species basis.
76 Environmental Factors Probability of seedling emergence was signifi cantly influenced by light availability and soil characteristics; all fi ve species exhibited signif icant responses to one or both of these factors. Species responded differently to environmental variables, exhibiting neutral, positive and negative responses to light and soilPC1 and neutra l or negative responses to soilPC2 and soilPC3 (Table 3-3). Overall, soil characteristics, part icularly those associated with PC2 (Mean = -1.57) and PC3 (Mean = -1.41), exhibited stronger effect s on seed to seedling transition probabilities than did light availability (Mean = -0.15; Table 3-3). Environmental factors had little effect on seedling survival two years after seed augmentation. Only Entandophragma utile demonstrated significantly improved chances of survival with increased light availability. Pancovia laurentii exhibited decreased recruitment success in response to soilPC3. Seed Predation and Herbivory Vertebrate seed predation str ongly limited seedling emergence: all three species exhibited increased emergence with caging (Table 3-3). Indeed, the large effect sizes observed for the caging effect relative to other variables (Table 3-3) suggest seed predation more strongly influenced the successful transition from seeds to seedlings than any other factor. Similarly, vertebrate herbivory (effect of caging at 2 years) significantly d ecreased seedling survival for 2 of 3 species. Overall seedling survival in caged plots was 1.97 times higher for Myrianthus arboreus and 1.58 times higher for Entandophragma utile than in uncaged plots. Although not statistically significant in our models, likely due to relatively small numbers of surviving seedlings, caged plots for Manilkara mabokeensis had 3.52 times more seedling than uncaged plots. Indeed the effect of caging more strongly influenced seedling surviv al probabilities than
77 any other variable for all species, suggesting seedling herbivory str ongly limits seedling population size in this system. Densityand Distance Depende nce (Janzen-Connell Effects) Higher densities of added seeds resulted in significantly lower seedling emergence probabilities for 3 of 5 species, and increased em ergence for one species, indicating that densitydependent mortality significantly reduces seedling recruitment at the seed to seedling transition for some species (Tables 3-3 and 3-4; Figures 33, 3-4,3-5 and 3-6). Di stance to conspecific adult only significantly explained mortality at th e seed to seedling transition for one of four species. Furthermore, only Pancovia laurentii exhibited mortality associated with distance to and density of conspecific adults Contrary to hypotheses that suggest to influence species diversity density-dependent mortality should be str ongest in common species, Myrianthus arboreus the most common species in cluded in this study (Table 3-1), demonstrated weakly positive (rather than negative) density-dependence at this stage of recruitment. Patterns of seedling survival to the second year of growth were not explained by seedling density, adult conspecific density or distance to conspecific adults for any of the five species, offering no evidence that Janzen-Connell effects importantly limit seedling survival two years following seed augmentation. Discussion Our study used a large-scale field experiment to evaluate the relative importance of abiotic resources (soil and light), s eed predation, herbivory and de nsity dependence to seedling emergence and survival. We did so most direct ly through detailed examination of three species for which exclosure experiments were conducted. Based on the large standardized effect sizes of caging relative to other variables for all three of these species, we suggest that vertebrate seed predators and herbivores more strongly determine patterns of seedling recruitment than other
78 factors. We did not find strong support for th e notion that hostspecific predators cause disproportionate mortality near conspecific ad ults (as implied by the original Janzen-Connell models) but rather suggest the mechanism of fr equency dependent mortality often observed in tropical forests might be explained by patterns of predation by genera list herbivores. These results do not suggest that environmental fact ors are unimportant; indeed, they also support niche-based theories of tropical tree species co existence, with speciesÂ’ differences exhibiting highly variable responses to na turally occurring environmental characteristics among sites. We conclude that seedling recruitm ent in the Congo Basin is most strongly dictated by generalist vertebrate predators, coupled with a relative ly weaker influence of abiotic environmental filtering and negative density-dependence. Niche Partitioning and Environmental Filters Our results support the notion th at species co-existence is at least partially caused by species-specific habitat specialization to differe nt abiotic conditions (B altzer et al. 2005; Baraloto et al. 2005; Cavender-B ares et al. 2004; Coomes and Grubb 2000; Harper 1977; John et al. 2007; Vargas-Rodriguez et al. 2005). We identified differential seedling emergence and survival among sites and species, suggesting seedling recruitment depends in a species-specific manner on the characteristics of the habitat into which seeds arrive. Niche partitioning with respect to edaphic conditions and light availability are well docu mented in other tropical regions (Aiba and Nakashizuka 2007; Engelbrecht and Kurs ar 2003; Harms et al. 2001; John et al. 2007; Kitajima 1994; Paoli et al 2006; Queenborough et al. 2007; Sv enning et al. 2004). Our study demonstrates that, in general, the environmental f ilters of soil and light act more strongly on the transition of plants from the seed to seedling stage than on seedling surv ival probabilities once a seed has passed through the establishment stage. This result is somewhat surprising because the seed to seedling transition stag e of recruitment is strongly dependent on seed reserves for
79 nutrients and water (Kitajima and Fenner 2000), s uggesting specific micro-site characteristics should more strongly influence seedlings only after a plant has exhausted its seed reserves (post germination seedling survival). Any important li mitations to seedling emergence that occurred between the seed to seedling tr ansition must thus be explaine d by either (1) the absence of appropriate physiological cues to stimulate ge rmination (e.g. appropriate light cues) or (2) characteristics of the site that result in seed and seedling mortality (e .g. desiccation, toxic levels of metals, or soil pathogens). Overall, soil characteristics, particularly those associated with soil PC2 and soil PC3, exhibited stronger effects on seed-t o-seedling transition probabilities than did light availability, as evidenced by the small effects si zes of light relative to soil ch aracteristics for most species at both the seed to seedling transition and seedling su rvival stages of plant development. Because previous studies have demonstrated that light av ailability is a strongly limiting resource for most tree species at the seedling stag e (Montgomery and Chazdon 2002) we expected to see a stronger influence of light relative to soil characteristics at this early stage of seed ling recruitment. Light availability significantly increased seedling emerge nce and survival for only two of five and one of five species, respectively, and the effect sizes for light were sm all relative to other factors for all species. Furthermore, the observed species-sp ecific responses to light availability sometimes failed to match those expected based on species regeneration-niches (Table 3-1), suggesting the light requirements and selective pressures exerte d on these species likely change as species pass from one life history stage to the next (Com ita et al. 2007; Werner and Gilliam 1984). Soil characteristics tended to negatively in fluence seedling emergence. One species, Staudtia kamerunensis was only able to emerge in soils with more clay than sand. Seedling emergence for 2 of 5 species was negatively infl uenced by high acidity (PCA2) and elevated
80 concentrations of Fe and Al (PCA3). Soil acid ity has been shown to cau se significant seedling mortality in previous studies (N orden et al. 2007). Acid soils can present several interrelated problems for plants, including toxicity of al uminum, and (under reducing conditions) iron (von Uexkll and Mutert 1995; Xu et al. 1991), whic h likely explains why species that were negatively affected by acidity (P CA2) also tended to be nega tively affected by Fe and Al (PCA3). The risk of aluminum toxicity is t hought to be increased in highly weathered tropical soils (Gillman 1991; von Uexkll and Mutert 1995). Soil characteristics had little effect on seedling survival, with only Pancovia laurentii showing lower survival associated with higher levels of Fe and Al. This could equally be ex plained by symptoms of aluminum toxicity, which reduce root development and makes plants sens itive to drought stress and reduces access to nutrients in the subsoil (Rowell 1988). Densityand Distance -Depende nce (Janzen-Connell Effects) Density-dependent mortality (as determined by seed density) influenced seedling recruitment during the transition from seed to seedlings for three of five species, as suggested by other studies (Clark and Cl ark 1984; Hammond and Brown 1998; Wright 2002), although these effects were weak relative to ot her factors in two of the three species. Seedling emergence, however, was unrelated to distance to and density of conspecific adults for all but one species, and we found no significant increas e in the strength of seed pred ation in relation to distance to conspecific adult, providing little support for Janzen-Connell effects at the seedling emergence stage. Seed density also failed to influence seed ling survival to the second year of growth, and conspecific adult tree density significantly increa sed (not decreased) the probability of seedling survival for two of five species. Distancedependent mortality had no influence on seedling survival. Furthermore, our results do not s upport hypotheses that density -dependent mortality disfavors the recruitment of co mmon species relative to rare species in a manner predicted to
81 promote species co-existence (Chesson 2000; Che sson and Warner 1981; Janzen 1970). On the contrary, the stronger effects of density-dependen t mortality in rare species observed in our study may be precisely what keeps them rare (Hubbe ll 2001; Klironomos 2002). Taken together, our results suggest the importance of density de pendence and, more spec ifically Janzen-Connell effects, as mechanisms that determine patterns of tropical tree recruitm ent and diversity may be inapplicable to Afrotropical forests. Vertebrate Seed Predation and Herbivory Results of our vertebrate excl usion experiments demonstrated that vertebrate herbivores and seed predators more strongly influence seedli ng recruitment and survival than abiotic factors and density-dependent mortality. At our site, the forest has re tained its full complement of rodents and large herbivores, and it is perhaps no t surprising that they dictate, to some degree, patterns of seedling recruitment. Because the main herbivores in our system do not appear to exhibit distance-dependent fora ging behavior at the plot le vel (see above), and distancedependent seedling mortality shoul d predominately occur when pred ators and herbivores are host specialized natural enemies, we suspect the vert ebrate herbivores and pr edators responsible for the high seed and seedling mortality observed in this study are likely polyphagous, generalist vertebrates (see Poulsen et al. in prep for a species list and average densities of vertebrate predators and herbivores at this site). Previous studies have demonstr ated that terrestrial mammals affect both the abundance and spatial distribution of seeds and seedlings through seed predation and herbivory (Augspurger and Kitajima 1992; Crawley 1988; Curran et al. 1999 ; Curran and Webb 2000; DeMattia et al. 2006; Fine et al. 2004; Grogan and Ga lvao 2006; Janzen 1970; Nathan and Casagrandi 2004; Rey and Alcantara 2000; Terborgh et al. 2008; Terborgh and Wright 1994; Vallejo -Marin et al. 2006). However, results regarding the degree to which seed and seedling predators limit plant
82 population sizes are mixed (Andersen 1989; Br own and Heske 1990; Brown and Human 1997; Crawley 2000; Davidson 1993; Louda 1989; Louda and Potvin 1995; Maron and Simms 2001). Seed predators and seedling herbivores shoul d only importantly influence tree species recruitment when they reduce seed and seedli ng densities below the level at which densitydependent mortality occurs (Crawley 1988; Hu lme 1996; Schupp 1990) in other words, when numbers of consumed seeds and seedlings surp ass those otherwise doomed for mortality through density-dependent thinning. In this study, we demonstrate that indeed, seed predation and herbivory were sufficient in magnitude to decr ease seedling recruitmen t beyond mortality levels imposed by density-dependent thi nning Â– at both low and high seed densities. Although the mechanism by which they influence recruitment is not consistent with those proposed by the Janzen-Connell hypothesis Â– as usually assumed in tropical forest systems Â– vertebrate seed predators and herbivores are disp roportionately important in de termining plant population size. In addition, for the three species in cluded in our vertebrate exclosur e experiments, the strength of seed predation and herbivory observed were re lated to regional (not plot) scale relative abundance of the species in this forest system (mean effect of seed predation: Myar = 0.615, Mama = 0.286, Enut = 0.088; mean effect of herbivory: Myar = 0.350, Mama = 0.168, Enut = 0.129; Table 3-1); with more comm on species demonstrating greater effects of herbivory than less common species. If herbivores and seed predators disproportionately attack the most common (or more competitive) species, poorer competitors (less common species) could be maintained in the system. Though the limited samp le size employed in this study prevents us from drawing too strong of a conclusion from th ese trends, we suggest that species-specific mechanisms to recover from and/or avoid seed and seedling predation may serve as an important
83 factors explaining frequency-depe ndence in this forest system even in the absence of strong distanceand density-dependence. Conclusion For decades ecologists have sought to uncove r the causes of high sp ecies diversity in tropical forests, evaluating niche-partitioning, de nsity-dependence and the effects of herbivores and seed predators on plant survival. We have de monstrated that none of these factors act alone; they differ in relative importance with planteating predators playi ng a disproportionately important role. Though herbivores may not opera te in the species-specific manner proposed by the original Janzen-Con nell hypothesis, we conclude they st ill have potentia l to play an important role in maintaining tropical tree di versity. We suggest that one conduit by which herbivores could influence plan t diversity is by severely limiting recruitment at the seed and seedling stage, perhaps altering competitive inte ractions among species. In other words, herbivory could counter the unde rlying niche-based differences observed in this study and prevent the competitive exclusion of less common (or less competitive) species from the system. These results could have important implicati ons for the conservation and management of tropical forests. On their cu rrent trajectory, tropical forests are losing many of their mediumand large-bodied herbivores. As generalist mammalian herbivores are lost through overhunting the relative importance of other pr ocesses would be expected to sh ift (Poulsen et al. in prep). Because the effects of density dependence we re similar in both our caged and uncaged treatments, we would not expect its importance to dramatically change in the absence of herbivores (but see Clark et al in prep. for fu rther discussion of this issue). However, nichebased mechanisms may become increasingly importa nt. Thus, deterministic processes such as competition for light and suitable micro-sites ma y proceed to reduce local diversity through the exclusion of inferior competitors. Alterna tively, large bodied mammals could be replaced by
84 smaller bodied seed predators, resulting in increased seed pr edation and no change in the strength of niche part itioning. Further experimental rese arch including a greater number of species is needed to determine (1) if verteb rate seed predation and herbivory is frequency dependent across a greater range of species a nd (2) how the potential shift from top-down control over plant populations vi a predation and herbivory to bottom-up control via resource competition might influence tropical forest diversity.
85 Table 3-1. Ecological characteri stics of focal tree species selected for use in this study. Regeneration guilds follow Hawtho rn, 1995. N.P.L.D = Non-pioneer, light demanding. Animal dispersal categories include: P= primate (arboreal primates, chimpanzees and gorillas), B= bird, E= elephant. Seed sizes are averaged lengths (L) measured from 100 seeds of each species. Average conspecific density was estimated from 30 1-ha plots in which adults >10cm db h of all tree species were measured, mapped, and identified to species. Species Regeneration Guild Dispersal mode Average seed size Average conspecific density/ha (> 10cm dbh) Pancovia laurentii Sapindaceae Shade bearer Animal (P) L = 1 cm 2.37 Staudti kamerunensis Myristicaceae N.P.L.D. Animal (P,B) L = 1.9 cm 0.57 Manilkara mabokeensis Sapotaceae Shade bearer Animal (P) L = 1.4 cm 1.67 Myrianthus arboreus Urticaceae Shade bearer Animal (P,B,E) L = 2.1 cm 3.96 Entandophragma utile Meliaceae N.P.L.D. Wind L = 0.8 cm 1.17
86 Table 3-2. Number of 1-hectare sites (N=30) in which adult individuals >10 cm dbh of our focal species co-exist. All comb inations of the five random ly selected species were observed to co-exist in a minimum of 5 (16%) of our study sites. Pala = Pancovia laurentii, Stka =Staudtia kamerune nsis, Mama =Manilkara mabokeensis, Myar=Myrianthus arboreus, Enut =Entandophragma utile. Mama Pala Stka Enut Myar Mama 0 Pala 11 0 Stka 5 7 0 Enut 5 15 5 0 Myar 7 17 5 11 0
87 Table 3-3. Summary of GLMM analys is identifying factors that most significantl y influence seed to s eedling transition probabil ities (a) excluding and (b) including caging eff ects as explanatory variables in the m odel. Numbers represent standardized parameter estimates to facilitate direct comparison of all continuous explanator y variables. represent statistical significance at p = 0.05. Full model results are available in Appendix C. Level Light Soil 1 Soil 2 Soil 3 Conspecific Caging Distance to conspecific Plot x Sp A. All_1 0.024 0.013* -0.393* 0.15* -0.189* -0.269* N/A -0.094 1.259 Mama_1 -0.312* -0.327* 0.655 -0.387 -0.208 1.373* N/A -0.156 0.476 Myar_1 0.175* 0.025* 0.440* -0.358* -0.356* 0.158 N/A 0.007 0.963 Enut_1 -0.229* 0.006 -0.801 -0.309 -0.248 -0.261 N/A 0.228 1.190 Stka_1 -0.207 -0.526* -21.69* -6.934* -6.129* -15.22* N/A -0.134 0.000 Pala_1 -0.174* 0.058* 0.53 0.135 -0.151 -0.828 N/A -0.269* 0.825 B. All_1 0.044* 0.022* 0.17* -0.207* -0.442* -0.001 0.529* -0.044 1.093 Mama_1 -0.207* -0.313* 0.389 -0.269 -0.331* 0.746* 0.884* -0.293* 0.434 Myar_1 0.122* 0.024* 0.426* -0.285* -0.273* 0.161 0.435* -0.028 0.874 Enut_1 -0.225* -0.004 -0.43 -0.13 -0.111 -0.592 0.235* 0.133 1.026
88 Table 3-4. Summary of GLMM analys is identifying the factors that most significan tly influence seedling survival to two years o f growth given a seedling emerges (a) excl uding and (b) including caging effects as explanatory variable s in the model. Numbers represent standardized parameter es timates to facilitate direct comparison of all continuous explanatory variables. represent statistical significance at p = 0.05. Full model results are available in Appendix C. Level Light Soil 1 Soil 2 Soil 3 Conspecific Caging Distance to conspecifics Plot x Sp A All_8 -0.051 -0.01 -0.063 -0.2* -0.139* -0.125 N/A 0.059 0.589 Mama_8 -0.097 0.047 0.149 1.476 -0.285 0.417 N/A -0.216 0.000 Myar_8 -0.027 0.005 0.197 0.163 0.049 0.250* N/A 0.095 0.416 Enut_8 -0.098 0.123* -0.165 -0.133 -0.3 0.109 N/A 0.257 0.000 Stka_8 0.016 -0.031 1.995 0.495 -0.279 2.854 N/A -0.211 0.000 Pala_8 -0.027 -0.026 0.359 0.208 -0.55* 0.78* N/A 0.042 0.000 B All_8 -0.048 0.040* -0.111 0.104 -0.099 -0.036 0.326* 0.1 0.328 Mama_8 -0.028* 0.155 -0.848 0.084 -0.001 -0.047 0.846 0.102 0.000 Myar_8 -0.064 0.031* 0.212 0.089 0.038 0.202* 0.285* 0.065 0.282 Enut_8 -0.036 0.061* -0.202 -0.201* -0.182 -0.071 0.382* 0.180 0.000
89 Figure 3-1. Map of 30 site locations in the northern Republic of Congo. We used satellite images to identify forest areas that contained dense mixed, terra firma forests in and around Nouabal-Ndoki National Park, Repub lic of Congo. From these potential study areas, we used the geographic surv ey design component of the Distance 4.1 software to randomly select 30 pl ots spanning an area of over 3000 km2. The sites were separated by at least 2km to pr omote independence of samples.
90 Figure 3-2. Site establishment and delineation. Example of one 100 x 100 m (1-ha) plot (N=30). Within each plot we mapped and identified all trees >10 cm diameter-at-breast-height (dbh).
91 Figure 3-3. Experimental design. We established 63 seed addition stations in stratified random locations across 21 of our 30 plots. Each seed addition station was subdivided 60 0.5 x 0.5 m quadrats. Into these quadrats, we sowed seeds of the five focal species at seven different densities, randomly altering the position of treatments for each station. Densities for each augmentation level were multiples (0, 25, 50, 100, 200, 500, and 2000) of the natural seed rain density of each species. Letters and numbers within quadrats respectively represent the species (Pala= Pancovia laurentii, Stka =Staudtia kamerunensis, Mama = Manilkara mabokeensis, Myar =Myrianthus arboreus, and Enut =Entandophragma utile) and number of seeds added We monitored seedlings within each quadrat at three month intervals for two years.
92 Figure 3-4 Graph depicting the variation in percent transmitted diffuse light observed within and among the 63 experimental stations ne sted within 21 plots (Mean = 10.07; Range = 3.03 Â– 19.38; SD = 4.06).
93 Figure 3-5. Principal component analysis of soil variables in 63 stations The first PC axis explained 26.7% of the total vari ance in soil data and was strongly correlated with so il texture (fractions of clay, sand, and si lt), total N, and exchangeable cations (Figure 3-2); these parameters are strongly associated with so il fertility. The second PC axis explained an additional 17.5% of the variance and was most strongly co rrelated with pH. PC axis 3 explaine d an additional 13.5% of the variance and was most strongly correlated with Fe and aluminum.
94 Figure 3-6. (A) Proportion of seedlin gs that established to three months of growth as a function of augmentation level for cage d and uncaged experimental quadrats (B) Proporti on of seedlings that survived 2 years of growth, given the seedling successfully recruited to seedling emergence, as a function of seed augmentation density.
95 APPENDIX A SELECTION OF THE EFFECT SIZE FO R SEED LIMITATION EXPERIMENTS There are several plausible effect sizes to m easure seed limitation. The selection of an appropriate effect size depends upon the type of data available as well as the functional form of the recruitment function and th e question being addressed (Ose nberg et al. 1999). Here we examine several effect size metrics, their underl ying assumptions and the situations in which they would be most appropriate. We also pres ent empirical estimators of these effect size metrics, and compare estimates to the theoretical va lues using a subset of our data taken from the few studies that quantified recru itment over a range of seed augmen tation studies (i.e., with four or more levels of seed augmentation instead of the more typical two le vels of Â“ControlÂ” and Â“AugmentationÂ”). From this analysis, we conc lude that a linear effect size measure (which measures the number of new recruits per added s eed) is the best effect size to summarize the currently available literature because: 1) most studies used only two augmentation levels, precluding the use of a non-linear effect size m easure, 2) the majority of studies using > 4 augmentation levels yielded an approximately li near recruitment function (Poulsen et al. 2007), and 3) the linear effect size measure matched th e theoretical prediction in more than twice the number of situations compared to other possible e ffect sizes. Because this choice is arguable, we also use an effect that is unadj usted by augmentation level (i.e., th e total number of new recruits without division by augmentati on). Below we develop our rationale in more detail. Conceptual Approaches to Effect Sizes Parameter Estimation Ideally one would like to determine the func tional form of seedling recruitment by fitting hypothetical models of seedling recruitment to s eed limitation data (number of seedlings that recruit with different densities of seeds), and interpreting the parameters of the recruitment
96 function biologically (Osenberg et al. 1997; Osenberg et al. 1999). For example, Poulsen et al. (2007) have suggested the non-li near Beverton-Holt recruitment function, commonly applied to fishes (e.g., Schmitt et al. 1999), as a lik ely model for seedling recruitment: max 0 01 R S P S P R (A-1) where R is the density of recruits (seedlings) that emerge from an input density of S seeds (consisting of augmented, A and naturally occurring, Samb, seeds: i.e., S = A + Samb), P0 is the proportion of seeds that recruit in the absence of density effects (i .e., the slope of the recruitment function at S = 0), and Rmax is the maximum density of recruits (i.e., the asymptote). Seed limitation (by any definition; see below) can then be evaluated at any seed density along the curve (e.g., at Samb, which indicates ambi ent seed density). Fitting the functional form requires multiple au gmentation levels. However, only 9 of the 43 papers (representing only 18 of 163 species and 37 of the 835 effect sizes) that met our criteria for inclusion also used four or more seed densities. Therefore, parameter estimation using non-linear recruitment functions is not a feasib le approach if the goal is to examine seed limitation across the majority of the published st udies. This shortcoming of the available literature requires we take an a pproach that can be applied with only two augmentation levels but that still reflects biological processes, at least approximately, even if th e recruitment function is non-linear. We outline two general approaches that can be used when the recruitment function is unknown. We evaluate them by reference to the Beverton-Holt recruitment function, a nonlinear function that provides a good general description of the avai lable studies of seed limitation (Poulsen et al. 2007). Next we discuss two wa ys of conceptualizing limitation, examine the potential effect sizes stemming from these concep tual definitions, and then determine the most
97 appropriate effect size for our me ta-analysis given the question being asked and the design of the experiments being summarized. Elasticity or Sensitivity Seed limitation can be defined as the ch ange in recruitment produced by a small perturbation to seed density (e.g., Schmitt et al 1999 and Poulsen et al 2007; Figure A-1). Thus, we can conceptualize seed limitation in terms of Â“sensitivityÂ” (i.e., dR/dS : the derivative of the recruitment function with respect to seed density) or Â“elasticityÂ” (i.e., d ln R / d ln S = ( S / R )( dR / dS )). For the Beverton-Hol t recruitment function: 2 max 0 01 R S P P S R, (A-2) and max 01 ln ln R S P S S R (A-3) Sensitivity expresses the effect of seed a ugmentation on an absolute scale (change in recruitment per seed), whereas elasticity gives the effect on a relative scale (the proportionate change in recruitment for a pr oportionate change in seeds). Graphically, these definitions correspond to the slope of the recruitment function (on an absolute or loglog scale) at a given seed density. The most appropr iate density to evaluate the sl ope is the ambient seed density (Samb). If Samb = 0, then el asticity is undefined (because both R and S = 0 and a proportional change cannot be defined). Limitation The effect of a putatively limiting factor ca n also be assessed by a comparison of the ambient state of the syst em with that achieved after the lim itation factor has been eliminated
98 (Schmitt et al. 1999). In the context of seed limitation, we can compare the recruitment when seeds are not limited (i.e., supplied in excess: Rmax) with recruitment at the ambient seed density ( Ramb). This difference can be measured on an absolute scale (i.e., Â“Absolute LimitationÂ”: Rmax Â– Ramb, Figure A-1a) or on a relative s cale (i.e., Â“Relative LimitationÂ”, Rmax Â– Ramb or ln( Rmax Â– Ramb) = ln( Rmax) Â– ln( Ramb): Figure A-1a, A-1b). Empirical Estimates of Seed Limitation Using Two Treatments Because most empirical studies use only two seed densities (e.g., Control and Augmented), we must select an effect size th at requires only two densities and is therefore linear on some scale. We present four general candidates and di scuss their relationship to the above conceptual definitions of seed limitation (all variables defined in the main text): 1. Absolute Response: i cont iR R, exp, measures the absolute chan ge in recruitment (seedling density) between the augmented (experimental) and control treatments: Rexp,i and Rcont,i are the average densities of s eedlings in the experimental a nd control plots, respectively, in the ith study. This measure approximates th e conceptual definition of Â“Absolute LimitationÂ” if the augmentation level is sufficientl y high to saturate the system and eliminate seed limitation (i.e., if max exp,R Ri ). 2. Relative Response: i cont iR R, exp, (which can be log-transf ormed without a qualitative change in meaning) measures the relative ch ange in recruitment. It approximates Â“Relative LimitationÂ” if the augmentation level is sufficien tly high to saturate the system and eliminate seed limitation. This eff ect size measure will be problematic if0, i contR. 3. Per Seed Response: i i cont iA R R, exp, measures the absolute change in recruitment per seed. It approximates the conceptual measur e of Â“sensitivityÂ” if the recruitment function is linear, or if iA is small relative to the non-linearity. 4. Relative per Seed Response: i cont i cont i i cont iR S A R R, exp, (where i contS, is the seed density in the control treatment and presumably equal to Samb) measures the relative effect of a proportionate change in seed density recruitment. It approximates the conceptual
99 measure of Â“elasticityÂ” if the recruitment function is linear on a log-scale, or if iA is small relative to the non-linearity. These effects sizes, although linked to conceptu al definitions of seed limitation (see above definitions), also have potential shortcomings. The Absolute and Relative Responses can lead to problems comparing studies that used very differe nt augmentation densities: e.g., all else being equal, a larger effect size will result from the addition of 1000 seeds vs. 100 seeds. Such was the case in our meta-analysis, where the densities of sowed seeds varied by more than an order of magnitude among studies. By standardizing the Absolute Response by the density of seeds sowed, the Per Seed Response gives a measure of Â“return on i nvestmentÂ” (recruits / seed). However, if the recruitment func tion is non-linear, this metric w ill give smaller effect sizes under higher augmentation levels, even if all else is equal. The Relative Response and Relative per Seed Response can lead to problems when there is no recruitment in the control (which was the case in many of our studies). No matter which measure of effect size is us ed, it should match th e question being asked and the design of experiments being summarized, and should be interpre ted in light of how limitation is defined. If the augmentation was small (relative to any non-linearity), then the Per Seed Response or Relative Per Seed Response can be interpreted as the marginal return per seed (i.e., sensitivity or elastic ity). In this case, the Relative Response and Absolute Response can not be clearly interpreted because their magnitudes are greatly in fluenced by the degree of augmentation and not the biology of the system (see Osenberg et al. 1999). In contrast, if the augmentation was large and eliminated seed limitation, then either of the Per Seed Responses would be poor choices for an effect size becau se their magnitudes decline with augmentation density (and fail to match any of our definitions of seed limitation). When augmentation saturates the systems, then either Relative Response or Absolute Response are better choices and
100 can be interpreted as Â“LimitationÂ” (Osenberg an d Mittelbach 1996). Of c ourse, the challenge is that with only two augmentation levels, one cannot know where a system lies along the recruitment function (i.e., whet her the augmentation range occu rred over a relatively linear portion of the function or if th e maximum augmentation level satu rated the system): see Figure A-2a. Comparison of Effect Sizes Relative per Seed Response cannot be calculated from the available literature because ambient seed density is rarely reported in seed augmentation st udies. Thus, to help determine which of the other effect sizes ( Relative Response, Absolute Response or Per Seed Response ) best approximates seed limitation in our collect ion of studies, we comp ared each of them to theoretical expectations based on the few studies that had four or more augmentation levels and thus allowed us to fit the non-lin ear Beverton-Holt recruitment f unction (see Poulsen et al., in review). We calculated effect sizes for the 37 studies (consisting of 18 species, many of which were sown under different conditions) reported in Poulsen et al. We fi rst calculated the three different effect sizes for each study using data (d ensity of seeds sown and density of recruits) from the control treatment and the treatment with the greatest number of sown seeds. We then compared each effect size to its corresponding theoretical effect size by assuming the true recruitment relationship was described by the Beverton-Holt function with the parameters estimated by Poulsen et al.: Absolute Limitation : i amb iR R, max, Relative Limitation : i amb iR R, max, Sensitivity : amb i iS S R
101 where i serves as an index for the ith study, other terms are defined as above, and sensitivity is the slope of the recruitment function evalua ted at the ambient seed density. If augmentation levels are small relative to the non-linearity of th e recruitment function, then the Per Seed Response should match Sensitivity closely, but the other effect sizes should not perform well: i.e., Absolute Response Absolute Limitation and Relative Response Relative Limitation If augmentation leads to satu ration of the system, then the Per Seed Response should not equal its theore tical value (i.e., Sensitivity ), but the Relative and Absolute Responses should equal their theoretical values (i.e., Relative Limitation and Absolute Limitation respectively), indicating that they would be be tter choices of effect sizes. Thus, by comparing the observed and theoretical expectations, we can determin e which metric applies most often and how it should be interpreted in light of the recruitment function. The Per Seed Response matched (within 30%) its theoreti cal counterpart mo re often (15/37 comparisons) than did the Absolute Response (12/37) or the Relative Response (4/37); in 6 cases, none of the effects matched. In some cases, the poor fit resulted from the presence of zeroes or the inability to estimate Rmax (e.g., in some cases the best estimate of Rmax was precluding the estimation of Limitation at all: i.e., there was no asymptote). Given its poor performance, we do not consider the Relative Response further (adding a constant to d eal with zeroes did not help its performance). Most interesting, the Per Seed Response and the Absolute Response performed in opposite ways and their performance depended on the qua litative shape of the recruitment function. When the recruitment function was demonstrab ly non-linear (i.e., a Beverton-Holt function was a better fit to the data than a lin ear model: see Poulsen et al. 2007), the Absolute Response did well (11/14 matches) and the Per Seed Response did poorly (1/14). However, when the function
102 was not demonstrab ly non-linear, the Absolute Response did poorly (1/22 matches) and the Per Seed Response performed best (22/22 matches). In this dataset, recruitment functions that were approximately linear (n=22) were more common th an demonstrably non-lin ear ones (n=14); in 1 case, we could not evaluate the shape of the function. We predicted this result based upon the exp ected match/mismatch between the empirical estimates and their theoretical counterparts and the conditions under which they should apply. Our results further highlight the importance of selecting effect size metrics by matching effect size metrics to characteristics of the system a nd a model of the systemÂ’s response (Downing et al. 1999). No effect size metric will match all ques tions or study systems. Indeed, this is the key problem in our application: which metric work s best, how should it be interpreted, and how might we discern the studies to which the e ffect size metric should be applied (and more importantly, not be applied)? Because we cannot examin e the recruitment function for most of our studies (because they have only two augmenta tion levels), we do not know if the function is relatively linear over the augmentation range or if the highest augmenta tion level is near the asymptotic recruitment value. Knowing this wo uld help us differentiate between studies in which Per Seed Response (or Absolute Response ) is most suitable and reveal how the effect size should be best interpreted (i.e., as Sensitivity or Absolute Limitation ). Instead, we seek a general approach that we can apply to all studies (b ecause we lack specif ic knowledge about most studies). Our analyses (using Poulsen et al .Â’s data set) suggest that the Per Seed Response matches theoretical expectations mo re often than other effect size options Furthermore, it is expected to work best when the non-linearity is relatively small. Because the majority of studies (22/36) failed to detect a non-linearit y in the recruitment function, we have chosen to use the Per Seed
103 Response as our primary response variab le in our meta-analysis. We note, however, that this metric will not behave well in some cases (e.g., wh ere the augmentation lead s to saturation). In these cases, which cannot be identif ied given the available data, the Per Seed Response will underestimate seed limitation as defined by Sensitivity and a more appropriat e variable would be the Absolute Response which corresponds to the concept of Absolute Limitation when augmentation saturates the system. This ambiguity is an unfortunate consequence of the types of studies that are available in the seed limitation literature. We remain hopeful that our analysis will lead to more useful empirical studies that can facilitate future an alyses derived from estimation of the recruitment function.
104 Figure A-1. (A) The density of emerged seedlings or recruits, R versus the number of seeds, S assuming a Beverton-Holt recruitment function. The dotted line represents the slope at S = Samb, where Samb is the number of seeds occu rring naturally without seed augmentation. The arrow demonstrates the difference between the maximum seedling emergence, Rmax, and seedling emergence at ambient conditions, Rmax. (B) Same as above, but on a log scale: Log10( R ) versus Log10( S ).
105 Figure A-2. (A) A Beverton-Holt recruitment function with two levels of seed augmentation. The first augmentation is small and the Per Seed Response or Relative Per Seed Response can be interpreted as the marginal re turn per seed (i.e., sensitivity or elasticity). The second augmentation is la rge and saturates the system. Therefore either the Relative Response or Absolute Response would be better choices for quantifying seed limitation. In the meta-ana lysis dataset, most experimental studies augmented seeds at a small level relative to the saturation point of the recruitment function. (B) As augmentation level increas es from zero to large values the Per Seed Response starts at the theoretical value corresponding to sensitivity (See Figure A-1a) and declines to zero. In othe r words, the slope of the Â“S ensitivityÂ” line in SM Figure A-2a becomes flatter as seed augmentation increases (i.e., moves farther out along the recruitment function). As a result, Per Seed Response best estimates sensitivity when augmentation is small relative to the non-lin earity in the recruitm ent function. (C) As augmentation increases from zero, the Absolute Response (difference in recruitment between the augmented and control treatment s) increases from zero to a maximum. This maximum corresponds to Absolute Limitation Thus, the Absolute Response is best when augmentation saturates the system and should be interp reted in the context of Absolute Limitation
106 APPENDIX B SUPPLEMENTARY MATERIAL FOR CHAPTER 2
107 Table B-1. Species specific results from generalized lin ear mixed model (GLMM) analyses of effect size E and number of seedlings as a function of treatment level and co nspecific adult (dbh >10 cm) tree density three months and tw o years after seed addition. CI are 2.5% and upper = 97.5% credible intervals. *Â†Indicates significance at 3 mont hs and 2 years respectively, as defined by CIÂ’s that do not overlap with 0. A. 3 months of growth (seed to seedling transition B. 2 years after seed addition Species Response Predictors Mean Effect SD Lower CI Median Upper CI Mean Effect SD Lower CI Median Upper CI PALA E Conspecific -0.5150.293-1.099-0.5160.057 -0.3460.490-1.242-0.3630.686 Level*Â† -0.1870.031-0.248-0.187-0.125 -0.3210.041-0.402-0.321-0.239 Random effect of plot*Â† 1.2850.2200.9341.2571.790 2.1100.3571.5442.0652.936 Seedlings Conspecific* -0.2460.119-0.488-0.244-0.014 -0.2530.319-0.912-0.2460.358 Level*Â† 0.8450.0660.7150.8450.975 0.7340.0880.5650.7330.910 Random effect of plot*Â† 0.3700.1470.1170.3660.680 1.3130.2680.8801.2821.922 Random effect of individual*Â† 1.1340.0750.9951.1311.288 1.3630.1211.1401.3581.617 STKA E Conspecific -0.1731.029-2.192-0.1872.002 0.2391.260-2.3840.2112.815 Level 0.0590.054-0.0480.0590.165 -0.0220.051-0.122-0.0220.079 Random effect ofplot Â† 4.4040.7803.1744.3026.211 5.1771.0543.5725.0217.662 Seedlings Conspecific -0.1540.720-1.666-0.1381.196 0.4300.815-1.2130.4452.063 Level*Â† 0.8420.1290.5880.8431.099 0.9400.1160.7150.9401.169 Random effect ofplot Â† 3.3760.6662.3193.2904.917 3.8850.8322.5943.7755.831 Random effect ofindividual Â† 1.6520.1981.3021.6392.078 1.3630.1701.0641.3521.728 MAMA E Conspecific 0.2220.884-1.4610.2081.973 0.2761.407-2.4360.2772.930 LevelÂ† -0.0510.044-0.136-0.0510.035 -0.3280.086-0.495-0.328-0.159 Random effect of plot*Â† 4.0570.7452.8933.9595.779 6.4971.4064.3596.2919.819
108 Table B-1. Continued. Seedlings Conspecific 0.1750.584-0.9280.1591.374 0.0961.047-1.9650.0742.255 Level*Â† 0.9330.0920.7520.9331.114 0.6090.1380.3370.6090.882 Random effect of plot*Â† 2.5710.5291.7332.5053.797 4.7271.1183.0364.5647.395 Random effect of individual*Â† 1.2710.1281.0381.2651.540 1.3480.2270.9531.3331.847 MYAR E Conspecific -0.1700.347-0.858-0.1760.522 0.0250.876-1.9820.0961.659 Level*Â† 0.1820.0170.1490.1820.215 0.0900.0300.0320.0900.151 Random effect of plot*Â† 1.5540.2541.1501.5232.141 3.4470.6122.4913.3664.883 Seedlings Conspecific -0.2060.236-0.665-0.2140.279 -0.1650.335-0.850-0.1610.537 Level*Â† 1.4680.0611.3501.4671.590 1.2560.0791.1061.2541.416 Random effect of plot*Â† 1.0450.1850.7461.0231.466 1.3690.3110.8761.3292.086 Random effect of individual*Â† 1.0900.0680.9641.0871.229 1.2210.1011.0341.2171.432 ENUT E Conspecific -0.1530.520 -1.211-0.1480.850 -0.3620.916-2.257-0.3521.353 Level*Â† -0.2660.037-0.337-0.266-0.194 -0.2860.052-0.387-0.286-0.184 Random effect of plot*Â† 2.4220.4221.7532.3703.409 3.8270.7232.7033.7285.520 Seedlings Conspecific -0.0680.320-0.670-0.0820.594 -0.1740.709-1.547-0.1981.280 Level*Â† 0.8310.0650.7050.8300.960 0.8240.0990.6350.8221.023 Random effect of plot*Â† 1.4790.2901.0141.4452.144 2.9270.6011.9762.8524.315 Random effect of individual*Â† 0.9780.0860.8200.9751.156 1.2680.1481.0001.2601.581
109 Table B-2. Parameter values from the de nsity dependent (Beverton-Ho lt function with seed, density-independent, and densitydependent limitation) and density independe nt (linear) models. Models were run wi th ALEV. Note that when parameter results of the density dependent (DD) model between 3mo and 2 yrs are compare d, the density independent (DI) parameter usually differs from 0 at 3 months but not at 2 years. This s uggests that the effect of seed addition begins to disappear by 2 years. Species Time Model DI SE L95 U95 Overdispersion SE L95 U95 rmax SE L95 U95 plot SE L95 U95 ALL 3 mo. DI 0.002 0.000 0.002 0.002 3.205 0.162 2.867 3.543 . 0. 319 0.120 0.068 0.569 ALL 3 mo. DD 0.021 0.003 0.014 0.028 2.549 0.132 2.273 2.826 5.101 0.148 4.791 5.410 3.508 1.247 0.898 6.117 ALL 2 yrs. DI 0.001 0.000 0.001 0.002 4.093 0. 294 3.477 4.709 1.984 0.665 0.593 3.374 ALL 2 yrs. DD 0.019 0.005 0.009 0.030 3.448 0.252 2.920 3.976 4.325 0.191 3.926 4.724 7.009 2.488 1.802 12.216 Pala 3 mo. DI 0.001 0.000 0.001 0.002 1. 436 0.179 1.061 1.811 . 0.182 Pala 3 mo. DD 0.013 0.004 0.005 0.021 0.981 0.132 0.704 1.257 3.993 0.246 3.478 4.507 1.794 0.828 0.061 3.527 Pala 2 yrs. DI 0.001 0.000 0.001 0.001 2. 714 0.432 1.809 3.619 . 0.664 Pala 2 yrs. DD 0.015 0.007 0.000 0.030 1.931 0.331 1.239 2.623 3.405 0.364 2.644 4.166 3.728 1.825 -0.092 7.547 Stka 3 mo. DI 0.001 0.000 0.001 0.002 4. 149 0.930 2.203 6.096 . 6.113 Stka 3 mo. DD 0.006 0.005 -0.004 0. 016 3.636 0.841 1.875 5.396 3.779 0.750 2.210 5.349 9.247 4.159 0.542 17.952 Stka 2 yrs. DI 0.001 0.000 0.001 0.002 2. 252 0.509 1.187 3.317 . 11.750 Stka 2 yrs. DD 0.009 0.005 -0.001 0. 018 1.784 0.434 0.875 2.692 4.085 0.524 2.989 5.181 17.770 7.688 1.680 33.862 Mama 3 mo. DI 0.001 0.000 0.001 0.002 2. 022 0.379 1.228 2.816 . 5.314 Mama 3 mo. DD 0.008 0.003 0.001 0.015 1.626 0.324 0.949 2.303 4.363 0.462 3.395 5.330 9.392 3.903 1.222 17.562 Mama 2 yrs. DI 0.001 0.000 0.000 0.001 3. 662 1.232 1.083 6.240 . 14.628 Mama 2 yrs. DD 0.009 0.006 -0.004 0. 022 2.706 0.947 0.724 4.689 2.977 0.611 1.698 4.256 23.584 11.309 -0.085 47.254 Myar 3 mo. DI 0.002 0.000 0.002 0.002 1. 046 0.112 0.811 1.281 . 0.745 Myar 3 mo. DD 0.014 0.002 0.010 0.018 0.484 0.063 0.353 0.616 6.002 0.200 5.583 6.420 1.305 0.501 0.257 2.354 Myar 2 yrs. DI 0.002 0.000 0.002 0.002 1. 652 0.247 1.135 2.169 . 1.834 Myar 2 yrs. DD 0.008 0.002 0.003 0. 012 1.405 0.214 0.957 1.853 5.823 0.573 4.625 7.022 3.914 1.647 0.466 7.362 Enut 3 mo. DI 0.001 0.000 0.001 0.001 1. 102 0.174 0.737 1.466 . 2.061 Enut 3 mo. DD 0.023 0.007 0.008 0.037 0.554 0.108 0.328 0.779 4.051 0.247 3.534 4.567 7.336 2.817 1.441 13.231
110 Table B-2. Continued. Species Time Model DI SE L95 U95 Overdispersion SE L95 U95 rmax SE L95 U95 plot SE L95 U95 Enut 2 yrs. DI 0.001 0.000 0.001 0.002 2. 261 0.485 1.246 3.277 . 6.377 Enut 2 yrs. DD 0.012 0.006 -0.001 0. 024 1.517 0.375 0.732 2.301 3.584 0.391 2.765 4.403 12.126 5.085 1.484 22.769
111 Figure B-1. Study site selection. We used satellite images to iden tify forest areas that contained dense mixed, terra firma forests in and around Nouabal-Ndoki National Park, Republic of Congo. From these potential st udy areas, we used the geographic survey design component of the Distance 4.1 softwa re to randomly select 30 plots spanning an area of over 3000 km2. The sites were separated by at least 2km to promote independence of samples.
112 Figure B-2. Site delineation, mapping and seed trap set up. Example of one 100 x 100 m (1-ha) plot (N=30). Within each plot we mapped and identified all trees >10 cm diameterat-breast-height (dbh). We qua ntified the rate and diversity of seed rain at each site for one year prior to the beginning of seed addition experiments, then for a second year following seed addition. Rates of natu ral seed rain were quantified by capturing fruits and seeds in seed traps (21 per site N=630). Seed traps consisted of 1 x 1 m wooden frames with a canvas center elevat ed approximately 75 cm above the ground. All fruits and seeds were colle cted, counted, and identified at two week intervals. In total we collected 431,770 mature seeds and 51,541 mature fruits from of 428 species.
113 Figure B-3: Experimental design. We established 63 seed addition stations in stratified random locations across 21 of our 30 plots. Each seed addition station was subdivided 60 0.5 x 0.5 m quadrats. Into these quadrats, we sowed seeds of the five focal species at seven different densities, randomly altering the position of treatments for each station. Densities for each augmentation level were multiples (0, 25, 50, 100, 200, 500, and 2000) of the natural seed rain density of each species. Letters and numbers within quadrats respectively represent the species (Pala=Pancovia laurentii, Stka =Staudtia kamerunensis, Mama= Manilkara mabokeensi s, Myar =Myrianthus arboreus, and Enut =Entandophragma utile) and number of seeds added. Dashed lines indicate caged treatments for each species; wh ich are discussed in Chapter 3.
114 Figure B-4. Graphical representation of seed lim itation based on the Beverton-Holt (1957) recruitment function that relates seedling density (i.e., recruitment, R ) to seed density ( S ). The Â“ambientÂ” curve represents an obs erved relationship betw een the density of emerged seedlings (i.e., recruits) and initial s eed density. When the input of seeds is the only limiting factor, the recruitment f unction is linear (Â“Seed limitation onlyÂ”) with a slope of 1, indicating the no post dispersal mortalit y and perfect viability of seeds. S amb is the ambient seed density in the sy stem that results from seed rain and the seed bank and yields a seedling density of R amb. Removing limitation due to density-independent mortality resu lts in a recruit density of R DI (Â“No Density independent limitationÂ”). Similarly, removi ng limitation due to density-dependent mortality results in a seedling density max of R DD (Â“No Density dependent limitationÂ”). Removing seed limitation results in the saturation density of recruits ( R max). Seed limitation is the differen ce between Â“AmbientÂ” and Â“No seed limitationÂ”. Modified max 0 max 0 from Schmitt et al.(1999).
115 Figure B-5. (A) Per seed recruitment effect si ze E (realized seed-estab lishment limitation). E varies from 0-1 with 1 representing perfect seed limitation and 0 representing perfect establishment limitation. These relatively low effect sizes indicate this forest system is more strongly establishment than seed limited. (B) Total number of seedlings for each of 5 species ( Pancovia laurentii, Staudtia kamerunensis Manilkara mabokeensis, Myrianthus arboreus, Entandophragma utile) as a function of seed augmentation level, for the first three m onths and two years of seedling growth. Weak seed limitation observed in A results in a gradual, but significant increase in total seedling numbers at very hi gh seed densities (Table 3-3).
116 Figure B-6. Fit of the final two of four candidate recruitment function models to seed augmentation data (level of seed augmen tation density as a function of ambient densities observed in seed tr aps during the first year of this project) for the five species included in this study. The dashed line represents the no density-dependentlimitation model (fitting P 0 and S amb) and the solid line represents the seedlimitation, density-independent-limitation, a nd density-dependent -limitation model (fitting P 0, R max, and S amb). For all species, the full Beverton-Holt model provides an improved fit to the lin ear model (see Table B-2).
117 Figure B-7. Results of the analyses of limita tion analysis for each of 5 species (Pancovia laurentii, Staudtia kamerunensis Manilk ara mabokeensis, Myrianthus arboreus Entandophragma utile,), at (A) three mont hs and (B) two years following seed augmentation. The blue arrow indicates the crossover point at which establishment limitation more strongly limits recruitment than seed limitation. The gray arrow represents the point at which density-dependence more strongly limits recruitment than density-independent mechanisms of mortality. For all species, establishment limitation becomes a stronger source of recruitment limitation than does seed limitation at very low seed input levels (4-6 times mean ambient seed densities). The role of seed limitation in limiting seedlings from achieving their maximum potential densities declines sharply after seed additi on levels below one seed per meter 2 (0.160.98 seeds/m2) For all species, density-independent mechanisms of seedling mortality more strongly prevents seed lings from achieving maximum population densities than do density-dependent mechan isms until seed availability reaches high addition levels (222-765 times mean ambient seed rain densities). For four of five species, the point at whic h density-dependence more strongly limits seedling recruitment than either seed limitation or density-independent factors occurs at seed densities within the range obs erved in seed traps.
118 Figure B-7. Continued
119 APPENDIX C SUPPLEMENTARY MATERIAL FOR CHAPTER 3. Table C-1. Complete results of GLMM for (A) all species at T1 (seed to seedling transition) and T8 (seedling survival) without caging and (B) with caging A. All species without caging effect Species Time EstimateSE Z p significance All 1 (Intercept) -2.2360.216-10.3380.000 *** Lev.z 0.0240.0171.4320.152 ns Trans_Diffuse.z 0.0130.0062.1020.036 soil1.z -0.3930.049-8.0290.000 *** soil2.z 0.1500.0423.5860.000 *** soil3.z -0.1890.038-4.9530.000 *** Consp.z -0.2690.039-6.8820.000 *** DistConsp.z -0.0940.023-4.0460.000 *** Plot Sp.nocage 1.259 Mama 1 (Intercept) -0.2650.843-0.3140.754 ns Lev.z -0.3120.076-4.1250.000 *** Trans_Diffuse.z -0.3270.106-3.0810.002 soil1.z 0.6550.4271.5350.125 ns soil2.z -0.3870.229-1.6910.091 ns soil3.z -0.2080.179-1.1600.246 ns Consp.z 1.3730.3843.5760.000 *** DistConsp.z -0.1560.221-0.7040.482 ns Plot 0.476 Myar 1 (Intercept) -2.7450.322-8.5190.000 *** Lev.z 0.1750.0237.5960.000 *** Trans_Diffuse.z 0.0250.0083.0570.002 soil1.z 0.4400.1552.8340.005 soil2.z -0.3580.089-4.0110.000 *** soil3.z -0.3560.085-4.1740.000 *** Consp.z 0.1580.1750.9050.366 ns DistConsp.z 0.0070.0310.2150.829 ns Plot 0.963 Enut 1 (Intercept) -1.3200.509-2.5940.009 Lev.z -0.2290.045-5.0950.000 *** Trans_Diffuse.z -0.0060.025-0.2560.798 ns soil1.z -0.8010.489-1.6370.102 ns soil2.z -0.3090.199-1.5570.120 ns soil3.z -0.2480.253-0.9820.326 ns
120 Table C-1. Continued. Species Time EstimateSE Z p significance Consp.z -0.2611.718-0.1520.879 ns DistConsp.z 0.2280.1171.9490.051 ns Plot Groups NameVarianceStd.Dev. Stka 1 (Intercept) 2.1731.7631.2320.218 ns Lev.z -0.2070.123-1.6840.092 ns Trans_Diffuse.z -0.5260.143-3.6710.000 *** soil1.z -21.6935.539-3.9170.000 *** soil2.z -6.9341.827-3.7950.000 *** soil3.z -6.1291.222-5.0160.000 *** Consp.z -15.2203.811-3.9930.000 *** DistConsp.z -0.1340.293-0.4590.646 ns Plot 0.000 Pala 1 (Intercept) -1.2490.330-3.7820.000 *** Lev.z -0.1740.035-5.0290.000 *** Trans_Diffuse.z 0.0580.0163.6770.000 *** soil1.z 0.5300.3301.6030.109 ns soil2.z 0.1350.1680.8070.419 ns soil3.z -0.1510.186-0.8130.416 ns Consp.z -0.8280.393-2.1060.035 DistConsp.z -0.2690.065-4.1630.000 *** Plot 0.825 All 8 (Intercept) -1.1820.181-6.5470.000 *** Lev.z -0.0510.028-1.8560.063 ns Trans_Diffuse.z 0.0100.0130.7860.432 ns soil1.z -0.0630.083-0.7570.449 ns soil2.z 0.2000.0712.8060.005 soil3.z -0.1390.062-2.2530.024 Consp.z -0.1250.062-2.0000.045 DistConsp.z 0.0590.0441.3470.178 ns Plot Sp.nocage 0.589 Mama 8 (Intercept) -3.0651.751-1.7510.080 ns Lev.z -0.0970.163-0.5940.553 ns Trans_Diffuse.z 0.0470.1870.2540.800 ns soil1.z 0.1491.1110.1340.893 ns soil2.z 1.4760.8551.7280.084 ns soil3.z -0.2850.268-1.0630.288 ns Consp.z 0.4170.5670.7350.462 ns DistConsp.z -0.2160.537-0.4010.688 ns Plot 0.000
121 Table C-1. Continued. Species Time EstimateSE Z p significance Myar 8 (Intercept) -1.3720.274-4.9990.000 *** Lev.z -0.0720.041-1.7340.083 ns Trans_Diffuse.z 0.0050.0180.2810.779 ns soil1.z 0.1970.1631.2050.228 ns soil2.z 0.1630.1001.6250.104 ns soil3.z 0.0490.1180.4170.677 ns Consp.z 0.2500.1162.1490.032 DistConsp.z 0.0950.0581.6270.104 ns Plot 0.416 Enut 8 (Intercept) -1.9880.396-5.0170.000 *** Lev.z -0.0980.067-1.4570.145 ns Trans_Diffuse.z 0.1230.0422.9600.003 soil1.z -0.1650.262-0.6270.531 ns soil2.z -0.1330.164-0.8100.418 ns soil3.z -0.3000.241-1.2430.214 ns Consp.z 0.1091.1820.0920.927 Ns DistConsp.z 0.2570.1511.7010.089 Ns Plot 0.000 Stka 8 (Intercept) -0.9801.382-0.7100.478 ns Lev.z 0.0160.1130.1440.885 ns Trans_Diffuse.z -0.0310.100-0.3060.760 ns soil1.z 1.9951.1401.7500.080 ns soil2.z 0.4950.8580.5780.564 ns soil3.z -0.2790.408-0.6850.493 ns Consp.z 2.8541.7651.6170.106 ns DistConsp.z -0.2110.253-0.8370.403 ns Plot 0.000 Pala 8 (Intercept) -1.3530.312-4.3350.000 *** Lev.z -0.0270.053-0.5020.616 ns Trans_Diffuse.z -0.0260.022-1.1960.232 ns soil1.z 0.3590.3820.9400.347 ns soil2.z 0.2080.1971.0580.290 ns soil3.z -0.5000.085-5.8720.000 *** Consp.z 0.7800.1734.5190.000 *** DistConsp.z 0.0420.1060.3990.690 ns Plot 0.000
122 Table C-1. Continued. B. All species with effect of caging Species Time Estimate SE Z p significance All 1 (Intercept) -2.3410.224-10.4320.000 *** Lev.z 0.0440.0162.8100.005 Trans_Diffuse.z 0.0220.0063.3810.001 ** soil1.z 0.1700.0822.0870.037 soil2.z -0.2070.050-4.1070.000 *** soil3.z -0.4420.045-9.9020.000 *** Consp.z -0.0010.068-0.0160.987 ns factor(Cage)Y 0.5290.0408.1340.000 *** DistConsp.z -0.0440.024-1.8300.067 ns Plot Spcage 1.093 Mama 1 (Intercept) -1.8430.596-3.0900.002 Lev.z -0.2080.062-3.3660.001 ** Trans_Diffuse.z -0.1330.059-2.2370.025 soil1.z 0.3890.3401.1440.253 ns soil2.z -0.2690.185-1.4520.146 ns soil3.z -0.3310.147-2.2470.025 Consp.z 0.7460.2573.4020.001 ** factor(Cage)Y 0.8840.1655.1080.000 *** DistConsp.z -0.2930.141-2.0750.038 Plot 0.434 Myar 1 (Intercept) -2.4520.287-8.5520.000 *** Lev.z 0.1220.0186.6440.000 *** Trans_Diffuse.z 0.0240.0073.2610.001 ** soil1.z 0.4260.1353.1600.002 soil2.z -0.2850.077-3.6940.000 *** soil3.z -0.2730.071-3.8500.000 *** Consp.z 0.1610.1561.0290.303 ns factor(Cage)Y 0.4350.0467.3360.000 *** DistConsp.z -0.0280.027-1.0440.296 ns Plot 0.874 Enut 1 (Intercept) -1.2100.429-2.8210.005 Lev.z -0.2250.036-6.1690.000 *** Trans_Diffuse.z -0.0040.020-0.2170.828 ns soil1.z -0.4300.392-1.0970.273 ns soil2.z -0.1300.165-0.7890.430 ns soil3.z -0.1110.202-0.5470.584 ns Consp.z -0.5921.445-0.4100.682 ns factor(Cage)Y 0.4300.1092.1100.035 DistConsp.z 0.1330.0931.4240.155 ns Plot 1.026
123 Table C-1. Continued Species Time Estimate SE Z p significance All 8 (Intercept) -1.4800.161-9.1740.000 *** Lev.z -0.0480.026-1.8280.068 ns Trans_Diffuse.z 0.0400.0113.6790.000 *** soil1.z -0.1110.082-1.3480.178 ns soil2.z 0.1040.0591.7760.076 ns soil3.z -0.0990.060-1.6460.100 ns Consp.z -0.0360.059-0.6160.538 ns factor(Cage)Y 0.3260.0714.6140.000 *** DistConsp.z 0.1000.0402.5080.012 Plot Spcage 0.328 Mama 8 (Intercept) -2.7931.054-2.6490.008 Lev.z -0.0280.111-0.2520.801 ns Trans_Diffuse.z 0.1550.1021.5240.128 ns soil1.z -0.8480.721-1.5910.112 ns soil2.z 0.0840.3920.2130.831 ns soil3.z -0.0010.127-0.0100.992 ns Consp.z -0.0470.261-0.1810.857 ns factor(Cage)Y 0.8460.3211.7040.088 ns DistConsp.z 0.1020.2940.3470.729 ns Plot 0.000 Myar 8 (Intercept) -1.5690.206-7.6140.000 *** Lev.z -0.0640.032-2.0090.045 Trans_Diffuse.z 0.0310.0132.3490.019 soil1.z 0.2120.1231.7300.084 ns soil2.z 0.0890.0731.2220.222 ns soil3.z 0.0380.0860.4480.654 ns Consp.z 0.2020.0842.4080.016 factor(Cage)Y 0.2850.0823.4600.001 ** DistConsp.z 0.0650.0461.4190.156 ns Plot 0.282 Enut 8 (Intercept) -1.6560.288-5.7410.000 *** Lev.z -0.0360.052-0.6890.491 ns Trans_Diffuse.z 0.0610.0272.2280.026 soil1.z -0.2020.174-1.1610.245 ns soil2.z -0.2010.103-1.9570.050 soil3.z -0.1820.157-1.1570.247 ns Consp.z -0.0710.659-0.1070.914 ns factor(Cage)Y 0.3820.1592.4100.016 DistConsp.z 0.1800.1061.6890.091 ns Plot 0.000
124 LIST OF REFERENCES Aiba, M., and T. Nakashizuka. 2007. Variation in juvenile survival a nd. related physiological traits among dipterocarp species co-existing in a Bornean forest. Journal of Vegetation Science 18:379-388. Aiba, S., K. Kitayama, and M. Takyu. 2004. Hab itat associations w ith topography and canopy structure of tree species in a tropical montane forest on Mount Kinabalu, Borneo. Plant Ecology 174:147-161. Andersen, A. N. 1989. How important is seed pred ation to recruitment in stable-populations of long-lived perennials? Oecologia 81:310-315. Andresen, E., and D. J. Levey. 2004. Effects of dung and seed size on secondary dispersal, seed predation, and seedling esta blishment of rain forest trees. Oecologia 139:45-54. Ashton, P. S. 1993. Species richness in plant communities, Pages 4-22 in P. L. Fiedler, and S. K. Jain, eds. Conservation Biology. New York, Chapman and Hall. Augspurger, C. K. 1984. Seedling survival of trop ical tree species inte ractions of dispersal distance, light-gaps, an d pathogens. Ecology 65:1705-1712. Augspurger, C. K., and K. Kitajima. 1992. Experi mental studies of seedling recruitment from contrasting seed distri butions. Ecology 73:1270-1284. Baillie, I. C., P. S. Ashton, M. N. Court, J. A. R. Anderson, E. A. Fitzpatrick, and J. Tinsley. 1987. Site characteristics and the distribution of tree species in mixed Dipterocarp forests on tertiary sediments in central Sarawak, Malaysia. Journal of Tropical Ecology 3:201220. Bakker, J. P., P. Poschlod, R. J. Strykstra, R. M. Bekker, and K. Thompson. 1996, Seed banks and seed dispersal: Important t opics in restoration ecology:461-490. Baltzer, J. L., S. C. Thomas, R. Nilus, and D. Burslem. 2005. Edaphic specialization in tropical trees: Physiological correlates and responses to reciprocal transplantation. Ecology 86:3063-3077. Baraloto, C., D. E. Goldberg, and D. Bonal. 2005. Performance trade-offs among tropical tree seedlings in contrasting mi crohabitats. Ecology 86:2461-2472. Bekker, R. M., J. P. Bakker, U. Grandin, R. Kalamees, P. Milberg, P. Poschlod, K. Thompson et al. 1998. Seed size, shape and ve rtical distribution in the soil: indicators of seed longevity. Functional Ecology 12:834-842. Bloor, J. M. G., and P. J. Grubb. 2003. Growth a nd mortality in high and low light: trends among 15 shade-tolerant tropical rain forest tree species. Journa l of Ecology 91:77-85.
125 Bolker, B. M., M. E. Brooks, C. J. Clark, S. W. Geange, J. R. Poulsen, M. H. H. Stevens, and J. S. S. White. 2009. Generalized linear mixed models: a prac tical guide for ecology and evolution. Trends in Ecology & Evolution 24:127-135. Brokaw, N., and R. T. Busing. 2000. Niche versus chance and tree diversity in forest gaps. Trends in Ecology & Evolution 15:183-188. Brown, J. H., and E. J. Heske. 1990. Control of a desert-grassland transition by a keystone rodent guild Science 250:1705-1707. Brown, M. J. F., and K. G. Human. 1997. Effects of harvester ants on plant species distribution and abundance in a serpentine grassland. Oecologia 112:237-243. Burnham, K. P., and D. R. Anderson. 2002, M odel selection and multimodel inference: a practical information-theoretic appr oach. New York, Springer-Verlag. Canham, C. D. 1989. Different responses to ga ps among shade-tolerant tree species. Ecology 70:548-550. Cannon, C. H., and M. Leighton. 2004. Tree species distributions across fi ve habitats in a Bornean rain forest. Journal of Vegetation Science 15:257-266. Cavender-Bares, J., K. Kitajima, and F. A. Bazzaz. 2004. Multiple trait associ ations in relation to habitat differentiation among 17 Floridian oa k species. Ecological Monographs 74:635662. Chambers, J. C., and J. A. Macmahon. 1994. A day in the life of a seed movements and fates of seeds and their implications for natural a nd managed systems. Annual Review of Ecology and Systematics 25:263-292. Chase, J. M., and M. A. Leibold. 2003, Ecologi cal Niches: Linking Classical and Contemporary Approaches. London, University of Chicago Press. Chen, Z. H., J. F. Peng, D. M. Zhang, and J. G. Zhao. 2002. Seed gemination and storage of woody species in the lower subtropical forest. Acta Botani ca Sinica 44:1469-1476. Chesson, P. 2000. Mechanisms of maintenance of species diversity. Annual Review of Ecology and Systematics 31:343-+. Chesson, P. L., and R. R. Warner. 1981. Enviro nmental variability promotes coexistence in lottery competitive systems. American Naturalist 117:923-943. Chesson, P. L. a. R. R. W. 1985. Coexistence medi ated by recruitment fluc tuations: a field guide to the storage effect American Natura list 125:769-787.
126 Clark, C. J., J. R. Poulsen, E. F. Connor, and V. T. Parker. 2004. Fruiting trees as dispersal foci in a semi-deciduous tropical forest. Oecologia 139:66-75. Clark, C. J., J. R. Poulsen, D. J. Levey, and C. W. Osenberg. 2007. Are plant populations seed limited? A critique and meta-analysis of seed addition experiments. American Naturalist 170:128-142. Clark, D. A., and D. B. Clark. 1984. Spacing dynami cs of a tropical rain-f orest tree: Evaluation of the Janzen Connell model. American Naturalist 124:769-788. Clark, D. B., D. A. Clark, and J. M. Rea d. 1998a. Edaphic variation and the mesoscale distribution of tree species in a neotropical rain forest. Journal of Ecology 86:101-112. Clark, J. S., E. Macklin, and L. Wood. 1998b. Stages and spatial sc ales of recruitment limitation in southern Appalachian forest s. Ecological Monographs 68:213-235. Clark, J. S., and J. S. McLachlan. 2003. Stab ility of forest biodi versity. Nature 423:635-638. Comita, L. S., R. Condit, and S. P. Hubbell. 2007 Developmental changes in habitat associations of tropical trees. Jour nal of Ecology 95:482-492. Condit, R., S. P. Hubbell, and R. B. Foster. 1994. Density dependence in two understory tree species in a neotropical forest. Ecology 75:671-680. Connell, J. H. 1971. On the role of natural enemies in preventing competitive exclusion in some marine animals and rain forest trees, Pages 298-312 in P. J. D. Boer, and G. R. Gradwell, eds. Dynamics of populations. Wageningen, Center for Agricultural Publication and Documentation. Coomes, D. A., and P. J. Grubb. 2000. Impacts of root competition in forests and woodlands: A theoretical framework and review of e xperiments. Ecological Monographs 70:171-207. Â—. 2003. Colonization, tolerance, competition and s eed-size variation within functional groups. Trends in Ecology & Evolution 18:283-291. Cornell, H. V., and J. H. Lawton. 1992. Species in teractions, local and regional processes, and limits to the richness of ecological communitie s a theoretical perspective. Journal of Animal Ecology 61:1-12. Crawley, M. J. 1988. Herbivores and plant population dynamics in A. J. Davy, M. J. Hutchings, and A. R. Watkinson, eds. Plant populat ion ecology. Oxford, Blackwell Scientific Publications. Â—. 2000. Seed predators and plant population dynamics, Pages 167-182 in M. Fenner, ed. Seeds: The ecology of regeneration in plant co mmunities. Wallingford, CAB International.
127 Curran, L. M., I. Caniago, G. D. Paoli, D. Astianti, M. Kusneti, M. Leighton, C. E. Nirarita et al. 1999. Impact of El Nino and logging on canopy tree recruitment in Borneo. Science 286:2184-2188. Curran, L. M., and C. O. Webb. 2000. Experimental tests of the spatiotemporal scale of seed predation in mast-fruiting Dipterocar paceae. Ecological Monographs 70:129-148. Dalling, J. W., and S. P. Hubbell. 2002. Seed size, growth rate and gap microsite conditions as determinants of recruitment success for pi oneer species. Journal of Ecology 90:557-568. Dalling, J. W., S. P. Hubbell, and K. Silvera. 1998. Seed dispersal, s eedling establishment and gap partitioning among tropical pioneer trees. Journal of Ecology 86:674-689. Dalling, J. W., H. C. Muller-Landau, S. J. Wright, and S. P. Hubbell. 2002. Role of dispersal in the recruitment limitation of neotropical pi oneer species. Journal of Ecology 90:714-727. Davidson, D. W. 1993. The effects of herbivory and granivory on terrestrial plant succession. Oikos 68:23-35. Davies, S. J., P. A. Palmiotto, P. S. Ashton, H. S. Lee, and J. V. Lafrankie. 1998. Comparative ecology of 11 sympatric species of Macaranga in Borneo: tree distribution in relation to horizontal and vertical resource hete rogeneity. Journal of Ecology 86:662-673. Day, P. R. 1965. Particle fractionation a nd particle-size analysis, Pages 545-566 in C. A. Black, and e. al., eds. Methods of Soil Analysis Part 1. Madison WI. American Society of Agronomy. De Oliveira, A. A., and S. A. Mori. 1999. A centr al Amazonian terra firme forest. I. High tree species richness on poor soils. Biodiv ersity and Conservation 8:1219-1244. DeMattia, E. A., B. J. Rathcke, L. M. Curran, R. Aguilar, and O. Vargas. 2006. Effects of small rodent and large mammal excl usion on seedling recruitment in Costa Rica. Biotropica 38:196-202. Denslow, J. S. 1980. Gap partitioning among tropi cal rain forest trees. Biotropica 12:47-55. Doherty, P. J. 2002. Variable replenishment and the dynamics of reef fish populations, Pages 327-355 in P. F. Sale, ed. Coral reef fishes : dynamics and diversity in a complex ecosystem. San Diego, CA., Academic Press. Downing, J. A., C. W. Osenberg, and O. Sa rnelle. 1999. Meta-analysis of marine nutrientenrichment experiments: Variation in th e magnitude of nutrient limitation. Ecology 80:1157-1167. Duivenvoorden, J. F., J. C. Svenning, and S. J. Wright. 2002. Beta-diversity in tropical forests Response. Science 297.
128 Edwards, G. R., and M. J. Crawley. 1999. Rodent seed predation and seedling recruitment in mesic grassland. Oecologia 118:288-296. Engelbrecht, B. M. J., and T. A. Kursar. 2003. Comparative drought-resistan ce of seedlings of 28 species of co-occurring tropical wo ody plants. Oecologia 136:383-393. Eriksson, O. 1995. Seedling recruitmen t in deciduous forest herbsthe effects of litter, soil chemistry and seed bank. Flora 190:65-70. Eriksson, O., and J. Ehrlen. 1992. Seed and mi crosite limitation of recruitment in plant populations. Oecologia 91:360-364. Fenner, M. 2000, Seeds: the ecology of regene ration in plant commun ities. Wallingford, CAB International. Fine, P. V. A., I. Mesones, and P. D. Cole y. 2004. Herbivores promote habitat specialization by trees in amazonian fore sts. Science 305:663-665. Gelman, A., and J. Hill. 2007, Data analysis us ing regression and multilevel/hierarchical models. Cambridge, Cambridge University Press. Gilbert, G. S., K. E. Harms, D. N. Hamill, a nd S. P. Hubbell. 2001. Effects of seedling size, El Nino drought, seedling density, and distance to nearest conspecific adult on 6-year survival of Ocotea whitei seedlings in Panama. Oecologia 127:509-516. Gillman, G. P. 1991. The chemical properties of ac id soils with emphasis on soils of the humid tropics, Pages 3-14 In Plant-Soil Interacti on at Low pH. Dordrecht, The Netherlands, Kluwer Academic Publishers. Givnish, T. J. 1999. On the causes of gradients in tropical tree divers ity. Journal of Ecology 87:193-210. Grime, J. P., J. G. Hodgson, and R. Hunt. 1988, Comparative plant ecology: a functional approach to common British plants. London, Unwin Hyman. Grogan, J., and J. Galvao. 2006. Factors limiting pos t-logging seedling regeneration by big-leaf mahogany (Swietenia macrophylla) in southeaste rn Amazonia, Brazil, and implications for sustainable management. Biotropica 38:219-228. Grubb, P. J. 1977. Maintenance of species rich ness in plant communities importance of regeneration niche. Biological Reviews of the Cambridge Philosophical Society 52:107145. Gunatilleke, C. V. S., I. Gunatilleke, S. Esufali, K. E. Harms, P. M. S. Ashton, D. Burslem, and P. S. Ashton. 2006. Species-habitat associati ons in a Sri Lankan dipterocarp forest. Journal of Tropical Ecology 22:371-384.
129 Gurevitch, J., L. L. Morrow, A. Wallace, and J. S. Walsh. 1992. A meta-analysis of competition in field experiments. Am erican Naturalist 140:539-572. Hall, J. S., P. M. S. Ashton, and G. P. Berlyn. 2003. Seedling performance of four sympatric Entandrophragma species (Meliaceae) under simula ted fertility and moisture regimes of a Central African rain forest. J ournal of Tropical Ecology 19:55-66. Hammond, D. S., and V. K. Brown. 1998. Di sturbance, phenology and life-histroy characteristics: factors influencing distance /density-dependent attack on tropical seeds and seedlings, Pages 51-78 in D. M. Newbery, H. H. T. Prins, and N. D. Brown, eds. Dynamics of tropical communities. Oxford, Blackwell. Harms, K. E., R. Condit, S. P. Hubbell, and R. B. Foster. 2001. Habitat as sociations of trees and shrubs in a 50-ha neotropical fore st plot. Journal of Ecology 89:947-959. Harms, K. E., S. J. Wright, O. Calderon, A. Hernandez, and E. A. Herre. 2000. Pervasive density-dependent recruitment enhances seedli ng diversity in a trop ical forest. Nature 404:493-495. Harper, J. L. 1977, Population Biology of Plants. London, Academic Press. Harris, D. J. 2002. The vascular plants of the Dzanga-Sangha Reserve, Central African Republic. Scripta Botanica Belgica:1-274. Hawthorne, W. D. 1995, Ecological profiles of Ghanaian forest tr ees: Tropical Forestry Papers, v. 29. Oxford, Oxford Forestry Institute. Henery, M. L., and M. Westoby. 2001. Seed mass and s eed nutrient content as predictors of seed output variation between species. Oikos 92:479-490. Hesse, P. R. 1971, A textbook of soil chemical analysis. London, John Murry. HilleRisLambers, J., J. S. Clark, and B. Beck age. 2002. Density-dependent mortality and the latitudinal gradient in species diversity. Nature 417:732-735. Hubbell, S. P. 2001, The Unified Neutral Theory of biodiversity and Biogeography. Princeton, NJ, Princeton University Press. Hubbell, S. P., R. Condit, and R. B. Foster. 1990 Presence and absence of density dependence in a neotropical tree community. Philosophical Tran sactions of the Royal Society of London Series B-Biological Sciences 330:269-281. Hubbell, S. P., R. B. Foster, S. T. O'Brien, K. E. Harms, R. Condit, B. Wechsler, S. J. Wright et al. 1999. Light-gap disturbances, recruitm ent limitation, and tree diversity in a neotropical forest. Science 283:554-557. Hubbell, S. P. a. F., R.B. 1986. Biology, chance and history and the structur e of tropical rain forest tree communities
130 Pages 314-329 in J. a. C. Diamond, T.J. ed. Commun ity Ecology. New York, Harper and Row. Hulme, P. E. 1996. Herbivory, plant regeneration and species coexistence. Journal of Ecology 84:609-615. Hurtt, G. C., and S. W. Pacala. 1995. The conseq uences of recruitment limitationreconciling chance, history and competitive difference s between plants. Journal of Theoretical Biology 176:1-12. ISI, T. 2004. Web of science. Jakobsson, A., and O. Eriksson. 2000. A comparativ e study of seed number seed size, seedling size and recruitment in grassland plants. Oikos 88:494-502. Janzen, D. H. 1970. Herbivores and number of tree species in tropical forests. American Naturalist 104:501-&. John, R., J. W. Dalling, K. E. Harms, J. B. Yavitt R. F. Stallard, M. Mirabello, S. P. Hubbell et al. 2007. Soil nutrients influence spatial distri butions of tropical tree species. Proceedings of the National Academy of Sciences of the United States of America 104:864-869. Jones, E. R., L. M. Curran, D. D. Wright, and A. L. Mack. 2008. Differential effects of mammalian seed predators on the regeneration of five Papua New Guinean tree species and implications for sapling recruitm ent. Journal of Tropical Ecology 24:259-267. Juenger, T., and J. Bergelson. 2000. Factors lim iting rosette recruitmen t in scarlet gilia, Ipomopsis aggregata: seed and dist urbance limitation. Oecologia 123:358-363. Kitajima, K. 1994. Relative importance of phots ynthetic traits and allocation patterns as correlates of seedling shade tolerance of 13 tropical trees. Oecologia 98:419-428. Kitajima, K., and M. Fenner. 2000. Ecol ogy of seedling regeneration, Pages 331-359 in M. Fenner, ed. Seeds: the Ecology of Regenera tion in Plant Communities. Wallingford, CAB International. Klironomos, J. N. 2002. Feedback with soil biota c ontributes to plant rari ty and invasiveness in communities. Nature 417:67-70. Kobe, R. K. 1999. Light gradient partitioning am ong tropical tree specie s through differential seedling mortality and growth. Ecology 80:187-201. Lanfranchi, R., J. Ndanga, and H. Zana. 1998. Ne w carbon 14C datings of iron metallurgy in the Central African dense forest, Pages 41-50 in e. a. Eves H.E, ed. Resource use in the Trinational Sangha river region of Equatorial Africa: histories, knowledge forms, and institutions. New Haven, CT Yale School of Forestry & Environmental Studies.
131 Lanfranchi, R., and D. Schwartz. 1991. Les rema niements des sols pendant le Quaternaire suprieur au Congo. II : Evolution des paysag es dans la rgion du Mayombe. Cahiers ORSTOM, srie Pdologie. Laurance, W. F., P. M. Fearnsid e, S. G. Laurance, P. Delamoni ca, T. E. Lovejoy, J. Rankin-de Merona, J. Q. Chambers et al. 1999. Rela tionship between soils and Amazon forest biomass: A landscape-scale study. Fore st Ecology and Management 118:127-138. Levine, J. M., and M. Rees. 2002. Coexiste nce and relative abunda nce in annual plant assemblages: The roles of competition and colonization. American Naturalist 160:452467. Louda, S. M. 1989. Predation in the dynami cs of seed regeneration, Pages 25-59 in M. Leck, V. T. Parker, and R. Simpson, eds. Ecology of soil seed banks. New York, Academic Press. Louda, S. M., and M. A. Potvin. 1995. Eff ect of inflorescence-feeding insects on the demography and lifetime fitness of a native plant. Ecology 76:229-245. Maron, J. L., and E. L. Simms. 2001. Rodent-lim ited establishment of bush lupine: field experiments on the cumulative effect of granivory. Journal of Ecology 89:578-588. McGill, B. J., B. J. Enquist, E. Weiher, and M. Westoby. 2006. Rebuilding community ecology from functional traits:178-185. Moles, A. T., D. S. Falster, M. R. Leishman, and M. Westoby. 2004. Small-seeded species produce more seeds per square metre of canopy per year, but not per individual per lifetime. Journal of Ecology 92:384-396. Moles, A. T., and M. Westoby. 2002. Seed additi on experiments are more likely to increase recruitment in larger-see ded species. Oikos 99:241-248. Â—. 2004. Seedling survival and seed size: a synt hesis of the literatu re. Journal of Ecology 92:372-383. Montgomery, R. A., and R. L. Chazdon. 2002. Li ght gradient partitioning by tropical tree seedlings in the absence of canopy gaps. Oecologia 131:165-174. Muller-Landau, H. C., S. J. Wright, O. Calder on, S. P. Hubbell, and R. B. Foster. 2002. Assessing recruitment limitation: concepts, methods and case-studies from a tropical forest Pages 35-54 in W. R. S. D. J. Levey, and M. Galetti, ed. Seed Dispersal and Frugivory: Ecology, Evolution and Conserva tion. Wallingford, CAB International. Nathan, R., and R. Casagrandi. 2004. A simple m echanistic model of seed dispersal, predation and plant establishment: Janzen-Connell and beyond. Journal of Ecology 92:733-746. Nathan, R., and H. C. Muller-Landau. 2000. Spatial patt erns of seed dispersal, their determinants and consequences for recruitment. Trends in Ecology & Evolution 15:278-285.
132 Nicotra, A. B., R. L. Chazdon, and S. V. B. Iriarte. 1999. Spatial heterogeneity of light and woody seedling regeneration in tropi cal wet forests. Ecology 80:1908-1926. Norden, N., J. Chave, P. Belbenoit, A. Caubere, P. Chatelet, P. M. Forget, B. Riera et al. 2009. Interspecific variation in seed ling responses to seed limitati on and habitat conditions for 14 Neotropical woody species. Jo urnal of Ecology 97:186-197. Norden, N., J. Chave, A. Caubere, P. Chatelet N. Ferroni, P. M. Forget, and C. Thebaud. 2007. Is temporal variation of seedling commun ities determined by environment or by seed arrival? A test in a neotropical forest. Journal of Ecology 95:507-516. Osenberg, C. W., and G. G. Mittelbach. 1996. The relative importance of resource limitation and predator limitation in food chans, Pages 134-148 in G. A. Polis, and K. O. Winemiller, eds. Food webs: integration of patterns and dynamics. New York, Chapman and Hall. Osenberg, C. W., O. Sarnelle, and S. D. Cooper. 1997. Effect size in ecol ogical experiments: The application of biological mode ls in meta-analysis. Amer ican Naturalist 150:798-812. Osenberg, C. W., O. Sarnelle, S. D. Coope r, and R. D. Holt. 1999. Resolving ecological questions through meta-analysis: Goals, metrics, and models. Ecology 80:1105-1117. Osenberg, C. W., C. M. St Mary, R. J. Schmitt, S. J. Holbrook, P. Chesson, and B. Byrne. 2002. Rethinking ecological inference: density de pendence in reef fishes. Ecology Letters 5:715-721. Pacala, S. W., and S. A. Levin. 1997. Biologicall y generated spatial patte rn and the coexistence of competing species, Pages 204-232 in D. Tilman, and P. Kareiv a, eds. Spatial ecology: the role of space in population dynamics and interspecific interactions. Princeton, NJ, Princeton Universtiy Press. Packer, A., and K. Clay. 2000. Soil pathogens and spatial patterns of seedling mortality in a temperate tree. Nature 404:278-281. Paine, C. E. T., and H. Beck. 2007. Seed predatio n by neotropical rain forest mammals increases diversity in seedling recr uitment. Ecology 88:3076-3087. Paine, C. E. T., K. E. Harms, S. A. Schn itzer, and W. P. Carson. 2008. Weak competition among tropical tree seedlings: Implications for sp ecies coexistence. Biotropica 40:432-440. Palmiotto, P. A., S. J. Davies, K. A. Vogt, M. S. Ashton, D. J. Vogt, and P. S. Ashton. 2004. Soil-related habitat specialization in dipter ocarp rain forest tree species in Borneo. Journal of Ecology 92:609-623. Paoli, G. D., L. M. Curran, and D. R. Zak. 2006. So il nutrients and beta di versity in the Bornean Dipterocarpaceae: evidence for niche partitioning by tropical rain forest trees. Journal of Ecology 94:157-170.
133 Pearson, T. R. H., D. Burslem, C. E. Mullin s, and J. W. Dalling. 2002. Germination ecology of neotropical pioneers: Interac ting effects of environmenta l conditions and seed size. Ecology 83:2798-2807. Persson, L., and L. A. Greenberg. 1990. Juveni le competitive bottlenecks-the perch (Percafluviatilis)-roach (Rutilus-rutilu s) interaction. Ecology 71:44-56. Plotkin, J. B., M. D. Potts, N. Leslie, N. Manokaran, J. LaFrankie, and P. S. Ashton. 2000. Species-area curves, spatial aggregation, and ha bitat specialization in tropical forests. Journal of Theoretical Biology 207:81-99. Plummer, M., N. G. Best, M. K. Cowles, a nd S. K. Vines. 2005.coda: output analysis and diagnostics for MCMC.R package, Version 0.9-5. Potts, M. D., P. S. Ashton, L. S. Kaufman, and J. B. Plotkin. 2002. Habita t patterns in tropical rain forests: A comparis on of 105 plots in Northwes t Borneo. Ecology 83:2782-2797. Poulsen, J. R., C. W. Osenberg, C. J. Clark, D. J. Levey, and B. M. Bolker. 2007. Plants as reef fish: Fitting the functional form of seedli ng recruitment. American Naturalist 170:167183. Queenborough, S. A., D. Burslem, N. C. Garw ood, and R. Valencia. 2007. Habitat niche partitioning by 16 species of Myristicaceae in Amazonian Ecuador. Plant Ecology 192:193-207. R Development Core Team. 2005.R: a language and environment for statistical computing, version 2.2.2.R Foundation for Statistical Computing, Vienna. Rao, M., J. Terborgh, and P. Nunez. 2001. Increased he rbivory in forest isol ates: Implications for plant community structure and com position. Conservation Biology 15:624-633. Rees, M., and M. Westoby. 1997. Game-theoretical evolution of seed mass in multi-species ecological models. Oikos 78:116-126. Rey, P. J., and J. M. Alcantara. 2000. Recru itment dynamics of a fleshy-fruited plant (Olea europaea): connecting patterns of seed dispersal to seedling establishment. Journal of Ecology 88:622-633. Rosenberg, M. S., D. C. Adams, and J. Gurev itch. 2000.MetaWin: statistical software for metanalysis, version 2.0.Sinauer, Sunderland, MA. Rowell, D. L. 1988. Soil Acidity and alkalinity, Pages 844-898 in A. Wild, ed. Soil Conditions and Plant Growth. UK., Longman Group. Royal Botanic Gardens. 2002. Elect ronic plant informaton centre.
134 Russo, S. E., P. Brown, S. Tan, and S. J. Davies 2008. Interspecific dem ographic trade-offs and soil-related habitat a ssociations of tree sp ecies along resource gradients. Journal of Ecology 96:192-203. Sale, P. F. 1982. Stock-recruit relationships and regional coexistence in a lottery competitive systema simulation study. Am erican Naturalist 120:139-159. Schmitt, R. J., S. J. Holbrook, and C. W. Os enberg. 1999. Quantifying the effects of multiple processes on local abundance: a cohort approach for ope n populations. Ecology Letters 2:294-303. Schupp, E. W. 1988. Factors affection post-dispersal s eed survival in a tropi cal forest. Oecologia 76:525-530. Â—. 1990. Annual variation in seedfall, postdispersal predation and recruitment of a neotropical tree. Ecology 71:504-515. Schupp, E. W., T. Milleron, and S. E. Russo. 2002. Dispersal limitation and the origin and maintenance of species-rich tropical forests, Pages 19-34 in D. J. Levey, W. R. Silva, and M. Galetti, eds. Seed dispersal and fr ugivory: ecology, evolut ion and conservation. Wallingford, CAB. Shaw, R. G., and J. Antonovics. 1986. Density-d ependence in Salvia lyrata, a herbaceous perennialthe effects of expe rimental alteration of seed densities Journal of Ecology 74:797-813. Sheldrick, B. H., and C. Wang. 1993 Particle size distribution in M. R. Carter, ed. Soil sampling and methods of analysis Canadian Society of Soil Science. Smith, C. C., and S. D. Fretwell. 1974. Optimal balance between size an d number of offspring. American Natura list 108:499-506. Spiegelhalter, D. J., Thoomas, N. G. Best and D. Lunn. 2003.WinBUGS version 1.4.Imperial College and MRC Biostatistics Unit: UK. Svenning, J. C. 1999. Microhabita t specialization in a specie s-rich palm community in Amazonian Ecuador. Journal of Ecology 87:55-65. Â—. 2001. On the role of microenvironmental heteroge neity in the ecology a nd diversification of neotropical rain-forest palms (A recaceae). Botanical Review 67:1-53. Svenning, J. C., D. A. Kinner, R. F. Stallard, B. M. J. Engelbrecht, and S. J. Wright. 2004. Ecological determinism in plant community st ructure across a tropical forest landscape. Ecology 85:2526-2538. Svenning, J. C., and S. J. Wright. 2005. Seed limitation in a Panamanian forest. Journal of Ecology 93:853-862.
135 Szentesi, A., and T. Jermy. 2003. Pre-dispersal se ed predation and seed limitation in an annual legume. Basic and Applied Ecology 4:207-218. Tateno, R., and H. Takeda. 2003. Forest structure and tree species distribution in relation to topography-mediated heterogeneity of soil ni trogen and light at the forest floor. Ecological Research 18:559-571. Terborgh, J., G. Nunez-Iturri, N. C. A. Pitman, F. H. C. Valverde, P. Alvarez, V. Swamy, E. G. Pringle et al. 2008. Tree recruitment in an empty forest. Ecology 89:1757-1768. Terborgh, J., and S. J. Wright. 1994. Effects of mammalian herbivores on plant recruitment in two neotropical forests. Ecology 75:1829-1833. Thomas, L., J. L. Laake, S. Strindberg, F. F. C. Marques, S. T. Buckland, D. L. Borchers, D. R. Anderson et al. 2006.Distance 5.0. Release 2. Research Unit for Wildlife Population Assessment.University of St. Andrews, St. Andrews, UK. Thompson, K. 1987. Seeds and seed banks. New Phytol ogist 106:23-34. Tilman, D. 1994. Competition and biodiversity in spatially struct ured habitats. Ecology 75:2-16. Â—. 1997. Community invasibility, recruitment li mitation, and grassland biodiversity. Ecology 78:81-92. Tilman, D., and S. Pacala. 1993. The maintenance of species richness in plant communities, Pages 13-25 in R. E. Ricklefs, and D. Schluter, eds. Species diversity in ecological communities. Chicago, University of Chicago Press. Tran, T. S., and R. R. Simard. 1993. Pages 43-50 in M. R. Carter, ed. Soil Sampling and Methods of Analysis Boca Raton, FL., Lewis. Tuomisto, H., K. Ruokolainen, and M. Yli-Halla 2003. Dispersal, environment, and floristic variation of western Amazonian forests. Science 299:241-244. Turnbull, L. A., M. J. Crawley, and M. R ees. 2000. Are plant populations seed-limited? A review of seed sowing experiments. Oikos 88:225-238. Turnbull, L. A., L. Manley, and M. Rees. 2005. Niches, rather than neutrality, structure a grassland pioneer guild. Proceedings of the Royal Society B-Biological Sciences 272:1357-1364. Turnbull, L. A., M. Rees, and M. J. Crawley. 1999. Seed mass and the competition/colonization trade-off: a sowing experiment Journal of Ecology 87:899-912.
136 Uriarte, M., R. Condit, C. D. Canham, and S. P. Hubbell. 2004. A spa tially explicit model of sapling growth in a tropical forest: does th e identity of neighbour s matter? Journal of Ecology 92:348-360. USDA NRCS. 2004. The PLANTS database, versi on 3.5., National Plant Data Center, Baton Rouge. Valencia, R., H. Balslev, and G. P. Y. Mi no. 1994. High tree alpha-div ersity in Amazonian Ecuador Biodiversity and Conservation 3:21-28. Valencia, R., R. B. Foster, G. Villa, R. Condit, J. C. Svenning, C. Hernandez, K. Romoleroux et al. 2004. Tree species distributions and local habitat variation in the Amazon: large forest plot in eastern Ecuador. Journal of Ecology 92:214-229. Vallejo-Marin, M., C. A. Dominguez, and R. Dirzo. 2006. Simulated seed predation reveals a variety of germination responses of neotropical rain forest species. American Journal of Botany 93:369-376. Vargas-Rodriguez, Y. L., J. A. Vazquez-Ga rcia, and G. B. Will iamson. 2005. Environmental correlates of tree and seedling-sa pling distributions in a Mexican tropical dry forest. Plant Ecology 180:117-134. von Uexkll, H. R., and E. Mutert. 1995. Global extent, development and economic impact of acid soils. Plant Soil 171:1-15. Webb, C. O., and D. R. Peart. 1999. Seedling de nsity dependence promotes coexistence of Bornean rain forest trees. Ecology 80:2006-2017. Â—. 2000. Habitat associations of tr ees and seedlings in a Bornean rain forest. Journal of Ecology 88:464-478. Werner, E. E., and J. F. Gilliam. 1984. The ontog enetic niche and species interactions in size structured populations. Annual Review of Ecology and Systematics 15:393-425. Westoby, M., M. Leishman, and J. Lord. 1996. Compar ative ecology of seed size and dispersal. Philosophical Transactions of the Royal Soci ety of London Series BBiological Sciences 351:1309-1317. Wright, S. J. 2002. Plant diversity in tropical forests: a review of mechanisms of species coexistence. Oeco logia 130:1-14. Xu, S., J. G. J. B. Harsh, and J. S. Boyle. 1991. Solid phase control of aluminum activity in an artificial plant growth medium contai ning hydroxy-Al-montmorillonite, Pages 25-34 in R. J. Wright, ed. Plant-Soil Interaction at Low pH. Dordrecht, The Netherlands,, Dordrecht, The Netherlands,. Zangato, M. E. 1999, African archaeo logy: Socits Prhistoriques et Mgalithes dans le NordOuest de la Rpublique Centrafr icaine, v. 46. Cambridge, UK, BAR.
137 Zobel, M., M. Otsus, J. Liira, M. Moora, and T. Mols. 2000. Is small-scale species richness limited by seed availability or micr osite availability ? Ecology 81:3274-3282.
138 BIOGRAPHICAL SKETCH Connie Jane Clark was born in Boise, Ida ho. Her love of biological organisms was developed at an early age, due in part to th e summers she and her six siblings spent chasing snakes through the cornfields of Nebraska a nd moose through the mountains of Wyoming. Two lessons stuck with her from this period: her sist er would go ashen white and faint when presented with a snake and hold on before shooting a shotg un while on a horseÂ’s back. Later, she and her family moved to Kennewick, Washington, where life was more about sports than nature, but she still graduated at the top of her class in 1989. Three years into a pre-med program at Willamette University, Connie spent a semester in Kenya studying wildlife manage ment and ecology with the School for Field Studies. Bedridden from an unfortunate case of malaria, Connie woke from a hallucinatory dream to realize th at she was meant to be an ecol ogist not a medical doctor. She graduated from Willamette University with a bachelorÂ’s degree in biology and psychology, rather than a pre-med degree. Upon gradua tion, Connie spent three years working as an Integrated Aqua Culture Extension Agent for the United States Peace Corps. Later, she spent 18 months working as a field assistant for the Dja Reserve Hornbill project in the south of Cameroon. Here she realized that she was remarkab ly gifted at identifying seeds that had been passed through an animalÂ’s body. Looking for a ny domain where she could apply this unique skill, she embarked upon and completed a masterÂ’s degree in ecology and systematics, which she gained from San Francisco State University in 2000. Upon completing her masterÂ’s degree, Connie worked for two years as the co-director fo r the Wildlife Conservation Society program in Lac Tele Community Reserve, Republic of Congo. Realizing that she missed the world of seeds and fruits, she left to get her Ph.D. at the Univ ersity of Florida. Later, while simultaneously conducting her dissertation work, Connie served as the Research Director for the Wildlife Conservation Society Buffer Zone Project su rrounding Nouabal-Ndoki National Park, Republic
139 of Congo. She gained a Ph.D. in interdisciplinary ecology from the University of Florida in 2009. Upon completion of this degree, Connie, her husband John Poulsen, and son moved to Falmouth Massachusetts where they joined the sc ientific staff at the Woods Hole Research Center.