Edge-effects on tree regeneration in the Colombian Andes

LIST OF TABLES
Chapter 1
Table 1. Description of study edges. Fragment size in hectares, elevation in meters
above sea level, absence of cattle disturbance, edge structure, edge
orientation, percentage of slope in study plots…..……………....……….…. 63
Table 2. Effectiveness of discriminant function analysis (DFA) to predict transect
membership to edge and interior groups (PTM) and to determine the depth of
edge influence (DEI) of abiotic edge-effects………………………………. 64
Table 3. Number of light-demanding and shade-tolerant tree species in different
life stages per site. Species are grouped based on their increased in
abundance towards edge (E) or interior (I) transects, or showed no pattern
(NP) as a function of distance from edge. The species are displayed in
ordination plots and had more than 50% of their variation explained by the
first two axes……………………………………………………………….. 65
Table 4. Multiple regression models for predicting species composition of A.
seedlings in wet season, B. seedlings in dry season, C. juveniles in wet
season, and D. juveniles in dry season, based on abiotic and biotic factors
as a function of distance from the forest edge. Abiotic: abiotic
environment in wet or dry season. DEI: depth of edge influence of the
abiotic environment in wet or dry season. Adults: adult composition.
Seedlings: seedling composition in wet or dry season. -: variable not
included in a particular model…………………………………………….. 66

Chapter 2
Table 1. Description of study tree species by habit, habitat, type of cotyledon,
seed diameter and regeneration habit (pers.obs.; Vargas 2002)…………...132
Table 2. Nonparametric analyses using the Wilcoxon (Gehan) statistics to test
the effect of habitat (edge/interior) and protection treatments over time on:
A. seed survival of large-seeded species, and B. seed germination and
C. seedling survival of large and small-seeded species, at Cairo and Aldea.
Overall habitat and protection effects were tested. Pairwise comparisons
tested the effect of habitat for each protection treatment: T – Total exclusion
plots, M – Mammal exclusion plots, and U - Unprotected plots, and among
protection treatments: T vs. M, M vs. U, T vs. U. ……………………….. 133
Table 3. Analysis of variance with repeated measures by site to test for the effect
of habitat over time on: A. Seed removal by ants (Miconia, Saurauia), and
B. Index of herbivory (Calophyllum, Brosimum, Miconia), and to test for

ix

the effect of habitat, protection, and their interaction on C. Seedling growth
rates of Calophyllum, Brosimum, Tapirira, and b) Miconia and Saurauia…135
Table 4. Summary table of edge-mediated effects of the abiotic environment in the
absence (AH) and presence of herbivores (H), of edge-mediated effects of
seed predation and seedling herbivory by invertebrates (I) and smallmammals
(M), and of seed predation and seedling herbivory independent
of habitat, on seed and seedling performance of large and small seeded
species at Cairo and Aldea.……………………………………….………... 138

Chapter 3
Table 1. Description of study edges. Fragment size in hectares, elevation in meters
above sea level, absence of cattle disturbance, edge structure, edge
orientation, percentage of slope in study plots. ………………………... 192
Table 2. Description of study tree species. Habitat: OGF – Old growth forest; SF –
Secondary forest species. Light requirement for regeneration: ST - shadetolerant;
LD - light demanding. Relative leaf thickness was based on
observations of mature leaves. Other characteristics refer to aspects that
could affect susceptibility to herbivory. Source of species identification
and plant characteristics are in Vargas (2002). …………………………… 193
Table 3. Means, standard errors (SE) and number of focal seedlings (n) for
percentage of herbivory, light availability (% PPFD), and number of
conspecific seedlings in neighborhoods of different areas (1 m², 9 m², 25 m²,
49 m² 100 m²) for eight tree species. Number of focal seedlings varied
among species due to differential mortality before the end of the herbivory
monitoring, and varied within species among neighborhoods due to
seedling elimination in analyses (see methods).………………………....... 194
Table 4. Best models for predicting percent herbivory in eight species of trees,
based on light availability (% PPFD), number of conspecific seedlings in
neighborhoods of different areas, and distance from the forest edge……... 196

x

LIST OF FIGURES

Chapter 1
Figure 1. Conceptual model underlying this study. The presence of edges may
generate abiotic edge-effects that are hypothesized to differ seasonally (1).
Abiotic edge-effects may affect forest structure and tree composition in
different plant stages (seedlings, juveniles, adults) (2). Abiotic edge-effects
may prevail over biological edge-effects (i.e., adults as sources of seedlings
and juveniles, and seedlings as sources of juveniles) in explaining patterns
of tree composition in early stages of development (3)……………….…….. 68
Figure 2. Diagram of the sampling design within a 50 x 50 m plot. Ten transects
were sampled per plot. Within each transect, I collected seedling data in 15
quadrats of 1 x 1m, juvenile data in 10 quadrats of 2 x 2 m, adult data in one
quadrat of 5 x 30 m, and abiotic and vertical foliage density data in 10 points.
In each transect, 10 hemispherical photographs were taken (only one shown).
Diagram is not to scale.……………….…………………………………….. 70
Figure 3. Plots from PCA’s for the first two axes for abiotic variables in wet and
dry season in five sites. Black dots represent distance transects and adjacent
numbers represent distance from the forest edge in meters. AT – Air
temperature, ST – Soil temperature, LG – light (%PPFD), SH – Soil
humidity, AH – Air humidity. Circles represent edge and interior groups
based on Discriminant Function Analysis (see Table 2)….….……………. 72
Figure 4. Plots of RDA’s for the relationship of forest structure and the abiotic
environment in five sites in wet and dry season. P values represent the
significance of this relationship for the first axis (F) and all axes combined
(A) based on Monte Carlo permutations. Black dots represent transects and
adjacent numbers represent distances from the forest edge. Abiotic variables
shown had a significant contribution to variation in forest structure based on
Monte Carlo permutation tests (P < 0.05). In italic letters: AT – Air
temperature, ST – Soil temperature, LG – light (%PPFD), SH – Soil
humidity, AH – Air humidity. Forest structure variables shown had ≥50%
of their variation explained by the first two axes. SA – Seedling abundance,
SR – Seedling richness, JR – Juvenile richness, JA – Juvenile abundance,
UN – Foliage density in the understory, MD – Foliage density in the
midstory.…………………………………………………………………… 74
Figure 5. Plots of RDA’s for the relationship of seedling composition and the abiotic
environment in five sites in wet and dry season. P values represent the
significance of this relationship for the first axis (F) and all axes combined

xi

(A) based on Montecarlo permutations. Black dots represent transects and
adjacent numbers represent distances from the forest edge. Abiotic variables
shown had a significant contribution to variation in seedling composition
based on Monte Carlo permutation tests (P < 0.05). In italic letters: AT – Air
temperature, SH – Soil humidity, AH – Air humidity. Species shown had
≥50% of their variation explained by the first two axes. Name abbreviations
are explained in Appendix 2....…..…………………..………………………. 76
Figure 6. Plots of RDA’s for the relationship of juvenile composition and the
abiotic environment in five sites in wet season. P values represent the
significance of this relationship for the first axis (F) and all axes combined
(A) based on Montecarlo permutations. Black dots represent transects and
adjacent numbers represent distances from the forest edge. Abiotic variables
shown had a significant contribution to variation in juvenile composition
based on Monte Carlo permutation tests (P < 0.05). In italic letters: ST –
Soil temperature, AH – Air humidity, SH – Soil humidity. Species shown
had ≥50% of their variation explained by the first two axes. Name
abbreviations are explained in Appendix 2...…..……………….…………... 78
Figure 7. Plots of RDA’s for the relationship of adult composition and the abiotic
environment in five sites in wet and dry season. P values represent the
significance of this relationship for the first axis (F) and all axes combined
(A) based on Montecarlo permutations. Black dots represent transects and
adjacent numbers represent distances from the forest edge. Abiotic variables
shown had a significant contribution to variation in adult composition based
on Monte Carlo permutation tests (P < 0.05). In italic letters: SH – Soil
humidity. Species shown had ≥50% of their variation explained by the first
two axes. Name abbreviations are explained in Appendix 2……………..…. 80

Chapter 2
Figure 1. Mean and standard errors of abiotic variables at edge and interior habitats
in Cairo (white background) and Aldea (black background). Bars without
lines represent wet season, and squared bars represent dry season. E/I: The
habitat treatment E –Edge and I – Interior, S: The season treatment. Light and
air temperature measured only in wet season (one-way analysis of variance).
Soil humidity and soil temperature measured in wet and dry season (one way
repeated measures analysis of variance). * P < 0.05, ** P < 0.01, *** P <
0.001.……………………………………………………………….……….. 140
Figure 2. Curves of the cumulative proportions of seed survival over time of
Calophyllum brasiliense, Brosimum utile and Tapirira myriantha at Cairo
and Aldea. E - Edge habitat, I - Interior habitat, U - Unprotected plots, M –
Mammal exclusion, T – Total exclusion plots. Error bars are standard errors.
……………………………………………………………………………….. 142

xii

Figure 3. Curves of the mean number of seeds removed by ants on Miconia notabilis
and Saurauia brachybotrys by habitat at Cairo and Aldea. C – Cairo, A –
Aldea, E – Edge, I – Interior. Error bars are standard errors.………………. 144
Figure 4. Percentage of sources of seed mortality in Calophyllum brasiliense,
Brosimum utile and Tapirira myriantha by protection treatment at Cairo and
Aldea. UN – Unprotected plots, MAM –Mammal exclusion plots, TOT –
Total exclusion plots. “Others” refers to sources of mortality due to seed
dessication, fungal infection or unknown source of mortality…………….. 146
Figure 5. Curves of the cumulative percentage of germinated seeds of Calophyllum
brasiliense, Brosimum utile and Tapirira myriantha at Cairo and Aldea .
E - Edge habitat, I - Interior habitat, U - Unprotected plots, M – Mammal
exclusion plots, T – Total exclusion plots. Error bars are standard errors. … 148
Figure 6. Curves of seedling growth rates per month of Calophyllum brasiliense,
Brosimum utile, Tapirira myriantha, Miconia notabilis and Saurauia
brachybotrys at Cairo and Aldea. Error bars are standard errors. Empty
circles are interior habitats and black circles are edge habitats. ……………. 150
Figure 7. Curves of leaf production per month of Calophyllum brasiliense, Brosimum
utile, Tapirira myriantha, Miconia notabilis and Saurauia brachybotrys at
Cairo and Aldea. Empty circles are interior habitats and black circles are
edge habitats. Error bars are standard errors. Significant values for Kruskal-
Wallis non-parametric test per month. * - 0.05 < P < 0.01, ** - 0.01 < P
< 0.001, *** - P < 0.001. …………………………………………………... 152
Figure 8. Curves of the cumulative index of herbivory over time on seedlings of
Calophyllum brasiliense, Brosimum utile and Miconia notabilis by habitat
at Cairo and Aldea. In Brosimum, a decrease in the cumulative mean leaf
herbivory damage in July was a consequence of an increase in total leaf areas
rather than a reduction in leaf damaged areas; the absolute proportion of
damage was the same or higher than the previous survey..........……………… 155

Chapter 3
Figure 1. Conceptual model underlying this study. The following hypotheses were
tested: 1) Herbivory varies with distance from the forest edge (Arrow 1),
light availability (Arrow 2), and seedling density (Arrow 3), and 2) Light
and seedling density vary with distance from the forest edge (Arrow 4 and 5
respectively). First, I predict greater seedling herbivory at forest edges, if
herbivory increases with high levels of light and light is greater at forest
edges (Arrows 2 and 4, positive sign). Second, if seedling herbivory increases
with seedling density (Arrow 3, positive sign), I predict greater herbivory

xiii

at edges of light-demanding species if they have greater densities at forest
edges (Arrow 5, positive sign), and greater herbivory at forest interiors of
shade-tolerant species if they have greater densities at forest interiors
(Arrow 5, negative sign). Light and plant density may interact influencing
edge-effects on herbivory (Dotted arrow).…………………………………... 197
Figure 2. Diagram of the sampling design within a fragment. Within a 50 x 50 m
plot, ten distance transects were sampled. Within each transect, I monitored
herbivory on eight focal seedlings (stars) in each of two species. I mapped all
conspecific seedlings per species within the plot (black dots) and calculated
the number of conspecific seedlings per focal plant in neighborhoods of
different sizes (continuous squares). In each transect, 10 hemispherical
photographs were taken (broken square; only one shown). Location of focal
and conspecific seedlings is hypothetical and diagram is not to scale.……… 199
Figure 3. Frequency distribution of percentage of seedling herbivory after a one year
period in eight species of trees. Breaks in X axes separate seedlings
individuals with zero herbivory and herbivory up to 1%, from individuals
with more than 1% herbivory in 5% intervals……………………………..…. 201
Figure 4. Scatter plots of percent herbivory (arcsin transformed) as a function of
light availability (% PPFD) for eight species in four sites. ……………………203
Figure 5. Scatter plots of percent herbivory (arcsin transformed) as a function of
conspecific seedling density in neighborhoods at which the highest R² and
the lowest probability were obtained based on multiple regressions (Table 4),
for eight species in four sites………………………………………………….. 205
Figure 6. Scatter plots of percent herbivory (arcsin transformed) as a function of
distance from the forest edge (m) for eight species in four sites……………… 207

xiv

LIST OF APPENDICES

Chapter 1

Appendix 1. Mean and standard deviations of abiotic variables in five sites in wet
and dry season per distance transect. ST – soil temperature (º C), SH- soil
humidity (0-10 scale), AT– air temperature (º C), AH – air humidity (%),
LG – light (% PPFD). In parenthesis standard deviations.………...……………82
Appendix 2. Species specific responses in different life stages (numbers) per site,
to their increase in abundance towards edge or interior transects based on
their regeneration habit (RH). A – Agreement: Light-demanding species (L)
increased towards edge transects and shade-tolerant species (S) towards
interior transects. N – No Agreement: species increased in abundance with
opposite pattern than in A. NP – No pattern: species with no pattern of
abundance increase as a function of distance from the forest edge………….. 85

Chapter 3
Appendix 1. Number of seedlings originally marked for herbivory studies (A) and
fate of those seedlings (B)……………………………………………………209
Appendix 2. Pearson correlation coefficients among seedling herbivory, distance
from the forest edge (m), light availability (%PPFD), and number of
conspecific seedlings in neighborhoods of different areas in eight species
of trees……………………………………………………………………….. 211

xv

GENERAL INTRODUCTION

Some of the most pervasive effects of contemporary human impact on the persistence
of species are changes in land use and land cover, and the associated effects of habitat
fragmentation (Davies et al. 2001). Consequences of fragmentation are not only
reduction of total forest cover and isolation of remaining forest patches, but also the
opening up of remnant forests to external influences of the surrounding matrix (edgeeffects)
(Haila et al., 1993). The harsh contrast between the matrix and forest edges
results in changes in light and wind regimes, which affect the physical environment
(abiotic edge-effects) (Saunders et al. 1991). Biotic edge-effects result from the direct
influence of abiotic edge-effects on vegetation structure and relative abundance,
distribution, and behavior of species, and indirectly, through changes in plant – animal
interactions (Murcia 1995). Consequently, edge-effects may degrade remaining habitats
by altering the structure and function of ecosystems (Saunders et al. 1991; Haila et al.
1993; Murcia 1995; Laurance 1997). Because in highly fragmented landscapes edge
habitats become the dominant feature of forest remnants, understanding edge-effects is
important for the conservation of biodiversity and management of fragmented ecosystems
(Laurance 1997).

Although edge-effects are one of the most important consequences of habitat
fragmentation affecting population dynamics, species interactions, and communities, a
general theory of edge impacts is not yet available (Laurance et al. 1997; Davies et al.
2001). The study of edge-effects is thus important to understand local dynamics within
fragments, but this knowledge may not contribute to general ecological principles since
edge-effects often are site-specific (Murcia 1995; Davies et al. 2001). Site differences

1

likely reflect differences in biotas among remnants within a fragmented landscape,
because remnants often represent only subsets of once heterogeneous original continuous
forest (Laurance et al. 1999). Moreover, modulators of the intensity and depth of edge
influence (DEI) of edge-effects vary due to edge orientation, structure, age, matrix type,
history of disturbance, among other factors (Murcia 1995). Understanding edge-effects is
essential as they are critical in the short-term for species persistence and often precede
other fragmentation effects (Davies et al. 2001). In addition, understanding edge-effects
is key to uncovering ecological principles for edge dynamics.

General “edge principles” might not have emerged yet as most research has been
limited to describing patterns of single variables as a function of distance from the edge,
and describing the ecology of edges themselves (Murcia 1995; Fagan et al. 1999). Few
studies have formulated and tested mechanistic hypotheses regarding edge-effects, which
are key to developing a better understanding of how edge-effects work and providing
generalizations about edge-effects (Murcia 1995; Fagan et al. 1999). In addition, little
consensus exists on the impact of edge-effects due to improper and/or dissimilar
sampling designs, failure to take into account modulators of the intensity of edge-effects,
and oversimplification of edge-effects (i.e., only monotonic responses, no variable
interactions, ignoring the scale) (Murcia 1995; Bierregard et al. 1997; Davies et al.
2001). Moreover, the knowledge of fragmentation effects is largely based on temperate
and tropical lowland ecosystems. This provides a very limited view of how
fragmentation and particularly edge-effects impact the habitats left since some
ecosystems may be more vulnerable to fragmentation than others (Bierregaard et al.
1997).

2

The goals of the research reported here are to contribute to the theoretical
foundations of edge-effects and to provide a basis for conservation practices by first,
examining correlations between edge-related patterns and mechanisms of tree
regeneration through a hierarchical approach (i.e., species, communities) and second,
adding information on fragmentation effects on a largely unexplored ecosystem, Andean
forest, which is one of the most threatened and richest ecosystems worldwide (Olson &
Dinerstein 1998).

In the first chapter, I use a community-level approach to examine edge-related
patterns of the abiotic environment and their link to forest structure and tree species
composition during a both wet and a dry season in five edge sites. I also evaluate
whether abiotic factors prevail over biotic factors in explaining edge-related responses on
tree species composition. In chapters two and three, I employ a species-level approach to
examine the edge-mediated effects of abiotic and biotic factors on regeneration processes
over time to establish some of the mechanisms that may explain the community patterns
observed in chapter one. Specifically, in chapter two I experimentally test whether the
abiotic environment and/or seed predation and seedling herbivory exhibit edge-mediated
effects on regeneration processes in five species of trees in two edge sites. In chapter
three, I empirically examine if light and/or conspecific seedling density function as edgemediated
effects on insect seedling herbivory in eight species of trees in four edge sites.

Chapter 1. Edge-effects on vegetation structure and tree species composition

Chapter one takes a multivariate approach to document and interpret edge-related
patterns of the abiotic environment in different seasons and to link them to community

3