Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Using aerial imagery to assess tropical forest cover surrounding restoration sites in Costa Rica
(USC Thesis Other)
Using aerial imagery to assess tropical forest cover surrounding restoration sites in Costa Rica
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
Using Aerial Imagery to Assessing Tropical Forest Cover Surrounding Restoration
Sites in Costa Rica
by
Jorge Amar
A Thesis Presented to the
Faculty of the Dornsife College of Letters, Arts and Sciences
University of Southern California
In Partial Fulfillment of the
Requirements for the Degree
Master of Science
(Geographic Information Science and Technology)
December 2020
Copyright © 2020 by Jorge Amar
ii
Dedication
To my mother,
who has given us her all,
and inspires us to do the same.
iii
Acknowledgements
I thank my committee chair, Dr. Bernstein, for her unfaltering guidance and support through this
process, and my committee members, Dr. Loyola and Dr. Marx, for their help in creating a sound
thesis document. Thank you to Alex for keeping me on task during the whole program, as well as
offering moral support. I want to thank all of the faculty and staff at the University of Southern
California’s Spatial Sciences Institute for their role in shaping a positive and meaningful
experience in earning my M.S. in GIST.
iv
Table of Contents
Dedication ....................................................................................................................................... ii
Acknowledgements ........................................................................................................................ iii
List of Figures ................................................................................................................................ vi
List of Tables ................................................................................................................................ vii
List of Abbreviations ................................................................................................................... viii
Abstract .......................................................................................................................................... ix
Chapter 1 Introduction .................................................................................................................... 1
1.1. Tropical Forests – A History in Costa Rica ........................................................................2
1.2. Study Area ..........................................................................................................................4
1.2.1. Site Description ..........................................................................................................4
1.2.2. Restoration Sites ........................................................................................................5
1.2.3. Planting Styles ...........................................................................................................6
1.3. Current Research .................................................................................................................9
1.4. Objective .............................................................................................................................9
1.5. Thesis Organization ............................................................................................................9
Chapter 2 Background and Literature Review.............................................................................. 11
2.1. Understanding Landscape-Level Processes in Restoration Ecology ................................11
2.1.1. Restoration Ecology .................................................................................................11
2.1.2. Landscape Ecology ..................................................................................................12
2.1.3. Habitat Cover Effects Observed in Restoration Efforts ...........................................14
2.2. Topographic Variables and Regression Analysis .............................................................16
2.2.1. Influence of Topographic Variables on Restoration Outcomes ...............................16
2.3. Assessing Change in Forest Cover using Aerial Imagery ................................................18
2.3.1. Acquisition of Aerial Imagery .................................................................................18
v
2.3.2. Quantifying Forest Cover through Aerial Imagery ..................................................19
Chapter 3 Methodology ................................................................................................................ 22
3.1. Data Sources and Processing ............................................................................................22
3.2. Workflow and Data Analysis ............................................................................................26
3.2.1. Data Preparation.......................................................................................................27
3.2.2. Forest Cover Change................................................................................................28
3.2.3. Multiple Ring Buffer................................................................................................29
3.3. Regression Analysis ..........................................................................................................32
Chapter 4 Results .......................................................................................................................... 40
4.1. Forest Cover Change.........................................................................................................40
4.2. Multiple Ring Buffer.........................................................................................................43
4.3. Regression Analysis ..........................................................................................................47
Chapter 5 Discussion and Conclusion .......................................................................................... 50
5.1. Forest Change Cover.........................................................................................................50
5.2. Multiple Ring Buffer.........................................................................................................52
5.3. Regression Analysis ..........................................................................................................53
5.4. Conclusion ........................................................................................................................56
Bibliography ................................................................................................................................. 59
Appendix I .................................................................................................................................... 63
vi
List of Figures
Figure 1 Study Area Extent............................................................................................................. 5
Figure 2 Outcomes of three widely used restoration planting treatments ..................................... 7
Figure 3 Experimental design of the treatment plots ..................................................................... 8
Figure 4 Aerial view of treatment site ........................................................................................ 22
Figure 5 Methodology Workflow ................................................................................................ 24
Figure 6 Extent of the 2005 and 2014 forest cover raster ............................................................ 25
Figure 7 Illustration of how forest cover change was computed .................................................. 26
Figure 8 Input settings used in the Multiple Ring Buffer Analysis ............................................. 27
Figure 9 Example of the Multiple Ring buffer output .................................................................. 28
Figure 10 ModelBuilder workflow used to calculate forest cover within each ring ................... 29
Figure 11 Methodology used in the regression analysis ............................................................... 30
Figure 12 Elevation profile of the study region ............................................................................ 31
Figure 13 Slope profile of the study region ................................................................................. 32
Figure 14 Equation used to convert aspect to linear variable ...................................................... 33
Figure 15 Aspect profile as a linear measurement ....................................................................... 34
Figure 16 Regression analysis buffer region ................................................................................ 35
Figure 17 Extent of regression analysis ....................................................................................... 36
Figure 18 Output from the forest cover change analysis ............................................................. 38
Figure 19 3D model of the observed forest cover gains and losses ............................................. 39
Figure 20 Output from the multiple rings buffer analysis ........................................................... 41
Figure 21 Percent gain and loss by treatment site ........................................................................ 43
vii
List of Tables
Table 1 Data description ............................................................................................................... 21
Table 2 A list of the thirteen treatment sites and treatment plots ................................................. 23
Table 3 Summary of the forest cover change for the entire study region ..................................... 37
Table 4 Multiple ring buffer forest cover changes by distance .................................................... 41
Table 5 Average cover distribution by site ................................................................................... 43
Table 6 Forest cover gain and topographic variables by site ........................................................ 43
Table 7 Coefficient summary ........................................................................................................ 43
Table 8 R square values ................................................................................................................ 43
viii
List of Abbreviations
AOI Area of Interest
ASL Above Sea Level
DEM Digital Elevation Model
GIS Geographic information system
LCBS Las Cruces Biological Station
LiDAR Light Detection and Range
UAV Unmanned Aerial Vehicle
ix
Abstract
Tropical landscapes in Costa Rica have increasingly become targets of restoration efforts after
deforestation depleted 90% of the region’s forests by the end of the 20th century. Research has
shown that the environment surrounding a restoration site influences outcomes in fragmented
landscapes, particularly as to the amount of forest cover surrounding restoration areas. However,
the degree of influence that forest cover has on restoration sites and the long-term effects have
historically been understudied due to the difficulty in assessing forest cover in remote regions
through conventional field methods. As a result, there is a need for more time and cost-effective
ways of evaluating and understanding forest cover change within the context of restoration
efforts in remote areas.
Geographic Information Systems (GIS) and remote sensing technologies have been
utilized by researchers to understand better the relationships between abiotic and biotic factors in
ecosystems. This study analyzed forest cover changes from 2005 to 2014 using high-resolution
remote imagery to understand how forest cover changed surrounding 13 restoration sites near
Las Cruces Biological Station (LCBS). The forest cover analysis revealed that the study region
experienced a 9% net increase in forest cover over nine years. Similarly, all except one of the
restoration sites had a net increase in forest cover within 200 meters. Topographic variables were
extracted from a 5-meter DEM to understand their influence on the changes in forest cover. We
hypothesized that elevation, slope, aspect, and distance to restoration site would have a strong
and positive correlation with whether areas surrounding the restoration sites reforested from
2005 to 2014. A regression analysis revealed that topographic factors do not solely explain the
variations in forest cover gain between sites; However, aspect, elevation, and distance to the
restoration sites center had a significant impact on forest cover gain in the study sites.
1
Chapter 1 Introduction
Tropical forests are the most biologically diverse ecosystems on Earth and provide vital
ecosystem services, such as carbon sequestration and water filtration (Busch and Ferretti-Gallon
2017). Deforestation in tropical regions mostly results from anthropogenic land-use changes such
as agriculture and logging (Gibson et al. 2011). Decades of research demonstrate how the
perverse degradation of landscapes has resulted in changes in the global carbon cycle and loss in
biodiversity (Vitousek et al. 1997; Foley et al. 2005). As a result, tropical landscapes have
increasingly become targets of restoration efforts worldwide due to the adverse effects observed
from deforestation, agriculture, and fragmentation, to name a few (Sader and Joyce 1988; Holl
and Kappelle 1999).
Ecological restoration is the science of rehabilitating degraded habitats to a semblance of
their historical state, restoring ecosystem services, and improving biodiversity (Bell et al. 1997).
Restoration ecology is complex and interdisciplinary – drawing on concepts from landscape
ecology, biology, geography, and geology. Some of the research in the rehabilitation of tropical
forests focuses on improving the methodologies behind active restoration practices, but studies
are often limited in scope and quantity (Bell et al. 1997; de Souza et al. 2013).
Geographic Information Systems (GIS) have been employed in landscape-level studies to
assess forest cover changes relating to ecological restoration efforts. Research in restoration
ecology has benefitted from GIS by allowing users to better understand landscape-level elements
and their impact on restoration outcomes. Research studies have also demonstrated instances
where the restoration's success is conditional to the landscape context, specifically variables such
as habitat connectivity, the amount of surrounding forest cover, and the degree of fragmentation
(Bell et al. 1997; Naveh 1994; de Souza et al. 2013). As a result, it is critical to understand the
2
influence of surrounding forest cover in ecology, so practitioners can better implement effective
restoration strategies that consider how the surrounding environment is contributing to the
success of active restoration efforts (de Souza et al. 2013).
A large-scale tropical forest restoration project was established in 2005 at Las Cruces
Biological Station (LCBS) in Costa Rica. The study aimed sought to understand the efficacy of
different tree-planting strategies in tropical regions. Since there is good evidence that the
outcomes of restoration efforts depend largely on the landscape context – such as the positive
influence of high habitat cover on restoration effectiveness – this study will supplement the
ongoing research in Costa Rica by quantifying the surrounding forest cover near 13 restoration
sites. By providing baseline data on forest cover changes since the start of the project, future
research can evaluate restoration success against the landscape context presented in this study.
Additionally, this study will investigate the effect that elevation, slope, aspect, and distance to
the site's center have on forest cover gain surrounding the research sites between 2005 and 2014
periods using regression analysis.
1.1. Tropical Forests – A History in Costa Rica
Tropical forests are the most biodiverse region on Earth. They are regarded for their
essential roles as terrestrial carbon sinks, sequestering carbon dioxide from the atmosphere, and
storing it in the vegetation and soil (Pan et al. 2011). Tropical forests once covered 96 to 99% of
the land in Costa Rica, but after an increase in agriculture and pasture grazing, deforestation rates
skyrocketed in the late 20th century (Leopold et al. 2000; Keenan et al. 2015). It is estimated that
90% of the original forests were lost during this period. Following the destruction of the timber-
producing forests, farmers were left with no choice but to abandon their now nutrient-poor
pastures (Leopold et al. 2001). However, forest cover in Costa Rica had increased from 2,564-
3
kilo hectares in 1990 to 2756 kilo hectares in 2015, owing largely to local and international
initiatives to reforest cleared areas (Algeet-Abarquero et al. 2015). Additionally, reforestation
rates were higher than deforestation rates between 1990 and 2015 (Algeet-Abarquero et al. 2015;
Keenen et al. 2015). As a result, many conservation efforts in the tropics aim to foster the
recovery of secondary forests through restoration practices.
Tropical forests are defined as closed-canopy forests that exist between 28 degrees north
and south of the equator and are regarded highly for their abundant levels of biomass and
biodiversity (Park 2002). Secondary forests in the tropics mainly result from human impacts such
as the abandonment of cleared forest lands, typically areas previously used for agriculture
(Brown and Lugo 1990). In contrast, primary forests are forests with no visible evidence of
human disturbance and now comprise a smaller area of the tropics. However, the regeneration of
old-growth forests is not possible (Chazdon 2017). As a result, much attention has shifted
towards the recovery and maintenance of secondary forests, as they now comprise more than half
of the tropical forests worldwide (Chazdon 2016).
Due to their significance in the global carbon cycle, much attention has been placed on
the recovery of aboveground biomass in tropical regions. The fostering of secondary forests is
corroborated by studies that have shown the resiliency and productivity in tropical secondary
forests (Poorter et al. 2016). However, the natural regeneration of tropical forest systems is
impeded by low seed dispersal, predation, poor seed germination, and low survival rates of
seedlings (Holl et al. 2001), calling for active restoration strategies that accelerate the natural
recovery process.
Initiatives to combat climate changes through tropical forests restoration have enacted
international policies such as the Reduced Emissions from Deforestation and Land Degradation
4
(REDD+), which incentivizes the reforestation through the monetization of ecosystem services,
such as payment for carbon sequestration (Daniels et al. 2010). In 1996, Costa Rica instituted a
Payment for Environmental Services (PES) program called Pago por Servicios Ambientales to
incentivize and compensate landowners for providing ecosystem services through their forested
lands (Daniels et al. 2010). As a result, reforestation efforts have grown tremendously due to
policies enacted by growing environmental degradation concerns.
Consequently, there has been an increasing need for viable restoration strategies that
accelerate the rate of recovery in areas that were previously used for pasture regions into more
productive landscapes. Additionally, since most of the tropical forests are now comprised of
regenerating forests, there is a need for understanding the underlying elements influencing the
restoration outcomes, particularly in abandoned pastures (Chazdon 2017). Accordingly,
restoration research in Costa Rica focuses on gaining a comprehensive understanding of tropical
ecosystems to implement more efficient restoration strategies.
1.2. Study Area
1.2.1. Site Description
The study will examine data collected near Las Cruces Biological Station, which is in
southern Costa Rica (Figure 1; (LCBT; 8 ̊ 47' 7" N; 82 ̊ 57' 32" W). The ~ 326-hectare (ha)
reserve was once an area primarily used for agriculture and grazing before its acquisition in 1962
by the Organization of Tropical Studies (OTP) and repurposed for botany, conservation,
reforestation research (Holl et al. 2017). The reserve also serves other functions, such as for
research and education about tropical systems and a tourist destination.
The region still maintains remnant fragments of old-growth forest (~200 ha) with no
history of logging, burning, agriculture, or other disturbances. Approximately 50 ha are
5
composed of secondary forests, which are forests that have regrown from disturbances from a
long enough time to where the effects of logging, fire, grazing, or agriculture are no longer
apparent. The region is classified as a tropical premontane forest, existing at an elevation range
of 1100-1430 meters above sea level (asl) and averages 4 meters (~157 inches) of rainfall
annually.
Figure 1 Study area surrounding the 13 restoration sites at Las Cruces Biological station. There
are 39 treatments and three treatments at each site.
1.2.2. Restoration Sites
A total of 13 restoration sites were established near Las Cruces Biological Station in
southern Costa Rica between 2004 and 2005 (Zahawi et al. 2013) to understand the efficacy of
three restoration planting strategies in a tropical premontane rain forest zone (Holdridge et al.
6
1971; Holl et al. 2011). LCBS is a highly fragmented landscape of forest patches and areas
previously used for various types of agriculture (Zahawi et al. 2013). Specifically, the sites were
chosen by Zahawi and colleagues (2013) and consist of abandoned pastures that were once used
for agriculture for over 18 years. The sites were cleared and burned before the start of the study,
but not after. Before clearing, the sites were dominated by exotic, or non-native, grass species
(Zahawi et al. 2013).
The ongoing study in the LCBS seeks to determine the efficacy and long-term effects of
these three planting styles to determine which are better suited for tropical forest restoration
(Zahawi et al. 2013). The treatments are 50 x 50 meters and are a minimum of 5 meters apart
from each other. Elevation ranges from 1,060 to 1430 meters above sea level. The thirteen sites
are separated by a minimum, and a maximum distance of 0.7 and 8 kilometers, respectively, and
have different measures of slope ranging between 5-35 degrees. The aspect ranges between each
site as well (Zahawi et al. 2013). Additionally, the sites are spread over regions with varying
surrounding forest cover. The varying topographic profiles of each site allow us to compare the
relative importance of topography in forest cover gain over time.
1.2.3. Planting Styles
The goal of assessing different planting strategies is essential when attempting to
accelerate and influence the rate of recovery through active restoration practices. Each
restoration site consists of three treatments, which include plantation, nucleation, and natural
regeneration (Figure 2). Plantation restoration treatments are designed to cover the entire target
region, rows of varying plant styles. As a result, plantation strategies are more expensive to
implement and can result in homogenous landscapes (Holl et al. 2011; Zahawit et al. 2013).
7
Nucleation, or island, treatments refer to planting done in separate clusters rather than
rows. Research suggests that the nucleation model strongly mirrors the natural succession of
forests. As a result, studies have looked at applying nucleation treatments rather than plantation
or natural regeneration. Additionally, previous studies show that nucleation treatments (Figure 2)
were associated with higher restoration success, having higher seedling survivability and species
density. However, the effect that nucleation treatments have on the surrounding landscape cover
has not been extensively studied. Opting for island-style treatments have garnered attention in
restoration ecology because it is more cost-effective, especially when rehabilitating larger
landscapes (Lindell et al. 2012) since they require fewer plantings than plantation-style
treatments.
Figure 2. The observed outcomes of the three planting strategies. Passive restoration outcomes
vary. Applied nucleation, or island treatments, result in more heterogeneous cover. Plantation
style treatments can result in monocultures with the outcome varying greatly from natural
succession outcomes. Source Holl et al. 2011
8
Natural regeneration has variable outcomes and, in some cases, without interference,
regions left to recover naturally remain in a suspended state dominated by woody vegetation
(Figure 2; Holl et al. 2011). The restoration study sought to evaluate the efficacy of these
planting strategies in a tropical landscape.
Figure 3 (A)Experimental design of each treatment plot. Treatments are a minimum of 5-meters
apart and were randomized. Capitalized letters within island and plantation plots represent tree
species seedlings (T = Terminalia amazonia, V = Vochysia guatemalensis, E = Erythrina
poeppigiana, I = Inga edulis). Black squares in the plantation treatment represent locations of
seed traps that were used in a previous study. (B) Photo of the experimental design. Images were
taken from Cole, Holl, and Zahawi 2010.
9
1.3. Current Research
The initial study by Zahawi et al. 2013 investigated the potential of three different
planting styles for active restoration practices and found that the island strategy was a viable
option for returning heterogeneity and closely match the outcomes of natural succession.
However, they explain that continued monitoring should be conducted to understand the long-
term effects and outcomes better.
Since the establishment of the restoration project near LCBS, additional studies have
investigated the long-term effects that different tree-planting styles had on seed recruitment (Holl
et al. 2017), seed establishment (Reid, Holl, and Zahawi 2015), and bird recruitment (Reid et al.
2014). Additionally, Holl et al. 2017 assessed the effects that the surrounding forest cover had on
seed recruitment within the three treatments, and they found that forest cover did not have a
strong correlation with the establishment of trees or the amount of seed rain found at each site.
However, no studies have evaluated how the surrounding forest cover surrounding the
restoration project in Costa Rica has changed between 2005 and 2014.
1.4. Objective
The goals of this study are to (1) quantify and compare the changes in tropical forest
cover between 2005 and 2014 surrounding thirteen restoration sites in southern Costa Rica using
hand-digitized aerial imagery; (2) compare the overall changes between 2005 to 2014 in forest
cover density for the entire study region; (3) assess the relationship between forest cover gain,
distance from the center of the site, elevation, slope, and aspect using regression analysis.
1.5. Thesis Organization
The following section will provide information supporting the importance of
understanding and incorporating the landscape-level processes and the effects they have on
10
restoration outcomes. Information on the type of remote sensing technology used in this study
will also be discussed. Chapter 3 outlines the process used to complete this project, such as data
collection, processing, and analysis. Chapter 4 presents the results of the analysis. Chapter 5
discusses the findings, implications, and ways to improve future studies.
11
Chapter 2 Background and Literature Review
This chapter reviews literature highlighting the influence of surrounding habitat cover in the
context of restoration ecology. The goal of this chapter is to provide background information on
landscape ecology as it relates to restoration studies and the methodologies used for quantifying
forest cover. This study aims to supplement the restoration work that is ongoing in these field
sites in Costa Rica and understanding how the surrounding landscape has changed from the start
of the project. The following literature has demonstrated the different uses of GIS and aerial
imagery to quantify forest cover and the connection to restoration ecology. This chapter presents
background information on the importance of landscape-level factors in restoration ecology,
aerial imagery to classy forest cover, and other studies that pertain to the thesis objective.
2.1. Understanding Landscape-Level Processes in Restoration Ecology
2.1.1. Restoration Ecology
Restoration ecology is the discipline involving the recovery of degraded, damaged, or
destroyed ecosystems (Aronson 2005). Historically, conservation efforts have focused on
preserving areas with little disturbance; however, efforts have now shifted towards the active and
passive restoration of degraded ecosystems due to extensive anthropogenic land cover changes
(Holl and Aide 2011). In landscapes that have been degraded due to agriculture production,
ecological restoration attempts to improve the functionality of the land and return a semblance of
the former ecosystem. More importantly, active restoration practices enhance biodiversity and
ecosystem services at the landscape level (Aronson 2005).
Tropical regions are primarily known for their biodiversity hotspots and richness of
endemic species that are not found elsewhere (Mittermeier et al. 1998). However, biodiversity
12
hotspots are often found in developing countries, like Costa Rica, where the restoration of
degraded ecosystems are seldom supported without evident socio-economic advantages
(Aronson 2005). Some ecosystems in the tropics recover rapidly on their own, while others
require humans to facilitate the restoration process actively. As a result, it is the responsibility of
land managers to make informed decisions on whether to take an active or passive approach, all
the while considering factors such as land-use history, surrounding landscape density, and the
natural ecosystem resiliency (Holl and Aide 2011).
Restoration practices vary by degree of human interference but can mostly be classified
as active or passive restoration. Active restoration practices involve practitioners in
implementing management techniques, such as planting seeds and removing competition of non-
native species. Passive restoration involves no interference except for the removal of the
disturbance, such as logging or grazing (Rakan, Reid, and Holl 2014). Frequently employed
active restoration practices in tropical forest recovery include plantation and island reforestation
(Holl et al. 2011). Plantation forestry involves the planting of monocultures, usual rows of a fast-
growing species, to kickstart the succession towards native forests. Island forestry is the planting
of trees in patches, rather than rows, which is less costly and time-consuming than the latter
(Holl et al. 2011).
2.1.2. Landscape Ecology
Landscape ecology studies how spatial processes interact with the abiotic and biotic
components of an ecosystem. Advances in remote sensing, geographic information systems, and
aerial imagery allow landscape ecologists to understand better the effect of spatial heterogeneity,
which is the different distribution of species in an area and the effect on ecosystem processes
(Brudvig 2009, 2011).
13
Understanding the spatial relationship between landscape and restoration ecology is
essential for practitioners to develop optimal strategies to rehabilitate and restore degraded
habitats, as various studies have shown the influence that landscape-level processes have on
restoration outcomes (Bell et al. 1997; de Souza 2013). Landscape-level processes refer to the
composition (density) and configuration (connectivity and heterogeneity) of a landscape (de
Souza et al. 2013). Instead of solely focusing on elements such as planting style (island,
plantation) and the plant species used, future studies can also consider surrounding landscape
cover, elevation, and aspect and how they affect restoration outcomes. Large-scale restoration
projects in remote regions often utilize different remote sensing and GIS technologies to quantify
and assess changes in forest cover, since field measurements at this scope are impractical, time-
consuming, and expensive (Chen et al. 1998; Boutin and Hebert 2002; Ruiz-Jaen 2005).
Motivations to track changes in forest cover, particularly relating to restoration ecology,
arise from the literature supporting the positive influence of surrounding landscape cover and the
increasing need to optimize restoration efforts in degraded lands. For example, De Souza et al.
(2013) conducted a metanalysis on restoration projects found that very few studies (54 total)
within the past fifteen years had utilized a landscape approach, the majority of which had
occurred in the most recent years (2009-2011). Landscape approaches refer to studies that
incorporated habitat cover, connectivity, and isolation variables in their investigations. In these
studies, 84% successfully demonstrated the role that the landscape had on the outcome of
restoration (de Souza 2013).
The authors also found the landscape context to have a positive influence on restoration
effectiveness, specifically when neighboring patches with high habitat density were in proximity
to restored areas. The metanalysis demonstrated that landscape-level factors are as important as
14
site-specific factors in the outcome of the restoration and that future studies should incorporate
landscape elements (de Souza 2013). Including landscape factors, like habitat cover, can help
ecologists set more specific restoration outcomes based on the context of the surrounding
landscape of the restoration site.
2.1.3. Habitat Cover Effects Observed in Restoration Efforts
Many studies assess the forest cover influence on wetland restoration projects using
digital orthophotography and GIS (Alsfeld et al. 2010). For example, one study used concentric
ring buffers to quantify the elements (e.g., streets, forest, developments, freshwater bodies)
surrounding the center of 20 previously restored wetland communities. It was found that distance
to the nearest forest was the most crucial variable contributing to the vegetation richness and
percent cover wetlands, attributing the vegetation richness to spillover effects from the
surrounding landscape. Another wetland study (Houlahan et al. 2006) found that surrounding
forest cover was a significant variable contributing to the species richness found within the
wetland restoration site. Both studies suggest that future restoration endeavors consider the
surrounding forest cover in their projects, as proximity and percent cover show a positive
influence on measures of restoration success, such as vegetation and species richness (Houlahan
et al. 2006; Alsfeld et al. 2010). Similarly, both studies exemplify the use of aerial imagery and
GIS in restoration efforts.
The interaction between restoration treatment sites and the surrounding forest cover,
specifically on the dynamic between forest cover and observed bird communities, has also been
studied in restoration ecology. Reid et al. (2014) studied the landscape-bird community dynamic
on the same thirteen restoration site at Las Cruces Biological Reserve. There was an observed
interaction between the local restoration efforts and the landscape context, affecting the
15
composition of the bird communities observed at the restoration sites. Specifically, areas with
higher forest cover had a higher representation of bird communities from the surrounding
landscapes at the restoration sites. The implication of these findings suggests that restoration
projects near regions with high forest cover can expect bird visitation from communities that are
representative of the reference habitat, an essential concept in restoration ecology as specific bird
species act as propagules for seed dispersion (Reid et al. 2014). In this context, GIS and aerial
imagery were used to examine the effects of surrounding forest cover on biotic factors such as
bird communities, and the implications demonstrate how spatial analysis can lead to more
informed decision making relating to restoration practices.
Reid et al. (2015) conducted a subsequent study based on the previous findings by Reid et
al. (2014) that examined the effect of forest cover on seed rain establishment for the same
restoration project at Las Cruces Biological Reserve. Since high forest cover had shown to be
positively correlated with the presence of bird communities in restoration, it was expected that
seed rain – the falling of wind-dispersed seeds —would be similar in restoration sites that of
surrounding reference forest (Reid et al. 2014). Contrastingly, Reid et al. (2015) did not observe
a relationship between seed rain and forest cover at 100- and 500-meter buffers around the
restoration sites, suggesting that surrounding forest cover is not a significant factor for seed rain
establishment in restoration sites. Nonetheless, they suggest that the effects of habitat cover on
restoration sites should be observed over time, as the composition of the surrounding forest and
restoration sites will continue to change.
Active restoration practices have been shown to assist the rate of regeneration in
deforested landscapes, and the effect that habitat cover has on seed recruitment has also been
studied (Holl et al. 2017). By using forest cover as a landscape variable, it was hypothesized that
16
higher seed recruitment –the establishment of seeds in a region—would be observed in plots with
higher areas of surrounding forest cover, due to the higher availability of seed dispersers in these
regions, as demonstrated in previous studies (Munro et al. 2007; Reid et al. 2014). Contrastingly,
they found no strong evidence for surrounding forest cover effects on seed recruitment. They
hypothesized a more substantial landscape effect would be detected if individual tree species
were used, rather than a total area forest cover, since other studies have shown that distance to
parent trees affected the dispersal of seeds (de la Peña‐Domene, Minor and Howe 2016).
Nonetheless, they explain that given the extent of the study, incorporating specific tree species is
impractical at larger scales.
Although forest cover has been used as a variable to predict outcomes, forest cover
changes at each site have not been directly studied against the topographic variations in the
underlying region. The thirteen restoration sites were placed in various regions near LCBS, each
with distinct elevations and with varying degrees of surrounding forest cover. It is essential to
consider how restoration outcomes can be explained by the variations in the context of their
location.
2.2. Topographic Variables and Regression Analysis
2.2.1. Influence of Topographic Variables on Restoration Outcomes
Other studies have sought to identify and evaluate the biophysical variables that affect
forest recovery in tropical regions. Variables such as elevation and aspect have shown to have a
relationship on forest recovery. Forest recovery, particularly natural reforestation, is more
commonly observed at higher elevations with steep slopes as these regions are more isolated and
less affected disturbances such as agriculture due to unsuitability (Thomlinson 1996). Aspect is
similarly thought to affect forest recovery, as regions respond differently to varying amounts of
17
sunlight. For example, in the northern hemisphere, south-facing slopes receive more sunlight and
consequently less favorable for tree growth (Maren et al. 2015).
For areas undergoing restoration and reforestation, it is crucial to understand what factors
influence the observed changes in vegetation cover. Regression analyses are used in studies to
understand the relationship between different variables. For example, given the topographic
profile of a region, can we see a relationship between these variables and the changes in
vegetation growth. Crk et al. (2009) investigated the relationship between forest recovery and
landscape variables using logistic regression. It has been observed that forest recovery is more
likely observed at higher elevations and steeper slopes (Thomlinson et al. 1996). Crk et al.
(2009) sought to identify the landscape-level factors that determine forest recovery in regions in
Puerto Rico previously used for agriculture. Their study used Landsat imagery and the
topographic variables elevation, slope, and aspect. Slope and aspect were derived from the
Digital Elevation Model (DEM). Of the studied variables, they found that slope and aspect were
the most important predictors of forest recovery, and, overall, the model was useful at predicting
the spatial pattern of forest recovery for use in land use planning and recovery studies (Crk et al.
2009).
The findings of Crk et al. suggest that slope and aspect could be strong predictors of the
observed recovery observed at the restoration sites in Costa Rica. Likewise, this study will
examine how the topographic variables, slope percent rise, aspect, and elevation influenced the
gain in forest cover surrounding the thirteen restoration sites by Zahawi and colleagues (2013).
Can we attribute the variation in forest cover gain in and surrounding the restoration sites to the
variations in topography?
18
The implications of this study would allow restoration ecologists to understand better the
variables influencing restoration outcomes in Costa Rica as well as help make better-informed
decisions regarding sites to prioritize. For example, if this study found that reforestation rates in
Costa Rica are inherently greater at higher elevations, then ecologists could prioritize
implementing active restoration strategies at lower elevations and allow natural regeneration to
occur in other areas. Likewise, if the variations in slope and aspect between the sites can explain
the disparity in forest gain outcomes, then restoration ecologists would be able to anticipate
better restoration outcomes as well as implement strategies that consider the landscape.
2.3. Assessing Change in Forest Cover using Aerial Imagery
2.3.1. Acquisition of Aerial Imagery
Remote sensing (RS) has often been used to monitor land cover and land-use changes, in
particular, those resulting from human activities such as deforestation and forest regeneration
(Read, Denslow, and Guzman 2001). RS technologies allow users to collect information about
the earth using cameras, satellites, or sonar systems (Read, Denslow, and Guzman 2001). RS
offers a more practical approach to assess forest cover, especially in large-scale projects in
remote regions where in-situ field checking methods are more like to be challenging. RS through
aerial imagery is the process of acquiring obtained through aircraft such as helicopters and fixed-
wing vehicles. More recently, unmanned aerial vehicles (UAV’s) have also proved to be a viable
method for obtaining high-quality imagery for use in forest cover studies (Zahawi et al. 2015).
This study will rely on remote sensing technologies, specifically aerial imagery, to quantify the
changes in forest cover in a remote area of Costa Rica.
19
2.3.2. Quantifying Forest Cover through Aerial Imagery
Various studies have utilized aerial imagery to quantify forest cover (Nowak et al. 1996;
Walton et al. 2008). Simply, the interpretation of aerial imagery in forest cover studies involves
detecting the presence or absence of forest cover from aerial photographs through a GIS and is
made in through a variety of methods. Aerial photographs need to be interpreted by someone
who can discern tree canopies. Typically, leaf-on imagery is interpreted, although skilled
interpreters can infer canopy from tree branches in leaf-off imagery (Walton et al. 2008). The
resolution needed to interpret tree cover, specifically digital images, is generally 1 meter,
although high-resolution imagery is larger in size and more time-consuming to process (Walton
et al. 2008).
One example is demonstrated in a study by Nowak et al. (1996) in which they quantify
urban tree cover in the United States (U.S.) using aerial imagery, which they regard as a cost-
efficient remote sensing method to analyze cover. The method involved scanning aerial imagery
quantify tree cover in the urbanized cities across the U.S., in which cover estimates were hand-
digitized by a photo interpreter using GIS. Nowak et al. (1996) explain that although scanning
aerial images is the most precise and detailed method of analyzing forest cover, it is labor-
intensive and conditional on the skill of the photo interpreter. The study was successful in
quantifying coverage, and it was also discovered that urban tree cover was primarily affected by
the surrounding natural environment. Additionally, the authors note that GIS and tree cover data
can be used to assess landscape-level features – such as forest fragmentation, patch sizes, and
connectivity – because it provides a baseline for assessing forest cover change as well as reveal
patterns in the landscape (Nowak et al. 1996).
Aerial imagery is digitized by creating polygon or raster files that signify forest cover
regions that correspond with the underlying aerial image using a GIS. Digitizing requires one to
20
trace georeferenced imagery to create raster, line, or polygon layers to create digital data, which
can then be used for spatial analysis. Spatial analysis tools can then be applied to the raster or
polygon representing forest cover using GIS. However, the accuracy of the features representing
forest cover depends on the image resolution (Pelz and Dickinson 2014). Although automated
methods exist to digitize aerial imagery, the hand-digitization of smaller regions can result in
more accurate raster layers when done by users who are familiar with the area, as was done in
this study (Cunningham 2006). Similarly, an adequate measure of forest cover quantity can be
obtained from aerial images, but more specific distribution measures of vegetation types and
classes are much more difficult to assess (Walton et al. 2008). Therefore, most forest cover
studies utilizing aerial imagery focus on structure and quantity analyses (Walton et al. 2008).
Monitoring forest cover through aerial imagery offers a low-cost method of assessing
spatial patterns in forest cover through GIS. Workflows can be quickly established, making
forest monitoring through aerial imagery a reliant, repeatable, and appealing methodology for
disciplines like restoration ecology. Zahawi et al. (2015) captured aerial images using a UAV in
order to extract essential monitoring parameters – including canopy height biomass and canopy
structure, to name a few— used in restoration to assess the progress of tropical forest recovery in
Costa Rica. The goal of the study was to compare the accuracy of UAV results to those of
traditional field-checking methods. Field-checking methods in remote regions are limited by
funds and require more time, making temporal monitoring at large spatial scales unreliable
through traditional approaches (Melo et al. 2013; Zahwai et al. 2015). The study UAV-obtained
aerial images and used Ecosynth methods to develop a high- resolution 3D model of the study
area. The Ecosynth Project consists of open source tools that help create 3D models of
ecosystems using images obtained from UAV flyovers. The study found that aerial imagery and
21
Ecosynth produced results comparable to field measures, particularly for measuring above-
ground biomass and percent openness parameters. The findings demonstrated the viability of
aerial imagery, GIS, and drones for assessing forest structure in large-scale restoration studies.
Reid et al. 2018 used high-resolution aerial photographs (10-meter resolution) to quantify
the persistence of secondary forests in southern Costa Rica between 1947-2014. The persistence
of secondary forests refers to the maximum age (in years) that a secondary forest reaches before
it is converted to other land types (Reid et al. 2017). The study examined six potential predictors
of secondary forest persistence which included, distance to the nearest road, distance to the
nearest river, mean elevation, slope, patch area, and distance to the nearest protected area. The
study found that patch size and distance to the nearest river were strong predictors of forest
persistence. For example, secondary forests at a 200-meter distance from the river were more 1.5
times more likely to be cleared than patches that were directly adjacent to rivers.
In the same way, forest patches of 14 hectares were half as likely to be cleared than
patches that were 0.1 hectares. Slope and elevation were not reliable predictors of forest
persistence. The study demonstrates the importance of evaluating landscape-level elements to
understand the context of forest cover changes. Landscape-level elements are essential in the
context of forest cover change. Identifying the variables that influence restoration outcomes will
help restoration practitioners implement strategies that incorporate their influence in the
decision-making process.
22
Chapter 3 Methodology
This chapter describes the proposed methods to assess the changes in forest cover for an on-
going restoration project in southern Costa Rica, specifically surrounding thirteen treatment sites
in Las Cruces Biological Reserve. The research methodology is based on a statistical analysis of
data derived from aerial imagery and a digital elevation model. This study will attempt to
understand the relationship between forest cover change, elevation, slope, aspect, and distance
using regression analysis. Researchers at LCBS provided all data, and the information obtained
from this study will be used in future studies to help better understand the landscape-level
processes restoration treatments. Geographic analyses were performed using ArcGIS Pro 2.6.
3.1. Data Sources and Processing
The study will utilize four datasets (Table 1) consisting of two forest cover layers (TIFF),
one treatment site layer, and a 5-meter DEM of the study region. The forest cover layers were
obtained from high-resolution orthorectified aerial photographs of 2005 and 2014 and had a
three-meter resolution. Aerial images were then hand-digitized by persons familiar with the
landscape at LCBS. The digitization of aerial imagery included primary and secondary forests,
live fences, individual trees, and hedgerows of all sizes as tree cover; all other areas were
classified as no cover (Reid et al. 2014). The treatment site layer consists of thirty-nine polygons
measuring ~50 x 50 meters and separated into thirteen sites. Each site contains a control, island,
and plantation treatment (Figure 3). All the data were re-projected to the WGS 1984 UTM Zone
17N coordinate system using the Project Raster tool. An example of treatment site OM is shown
in Figure 4 with the three different treatment plots (control, island, and plantation). The forest
cover growth between the two years is also shown in Figure 4.
23
Table 1 Data Description
Layer Date
Collected
Contents Spatial
Resolution
Source &
Format
Projection
Costa Rica
Forest Cover
2005 Raster file from
digitized aerial
imagery of the same
year
3m Raster (TIFF)
provided by
Organization
for Tropical
Studies
WGS 1984
UTM Zone
17N
Costa Rica
Forest Cover
2014 Raster file from
digitized aerial
imagery of the same
year
3m Raster (TIFF)
provided by
Organization
for Tropical
Studies
WGS 1984
UTM Zone
17N
Treatment Plot
Locations
2005 Polygon files
outlining the
treatment plot
locations
Polygon Shapefile
provided by
Organization
for Tropical
Studies
WGS 1984
UTM Zone
17N
DEM Study
Area
2013 Digital Elevation
Model of Southern
Costa Rica
5m Raster (TIFF)
provided by the
Organization
for Tropical
Studies
WGS 1984
UTM Zone
17N
24
Figure 4 An example of one of the treatment site locations (Site OM). There are 13 sites
scattered through various regions within the study area boundary. Each site has three treatments
(control, island, and plantation) with varying setups.
25
Table 2. A list of the thirteen restoration sites with three treatments at each site. The area refers
to the plot size of each treatment (~50 x 50 meters).
Site Treatment Area (sq.m)
AC
P 2343.0
C 1990.5
I 1769.8
BB
P 2466.0
I 1891.6
C 2302.9
EC
P 2384.8
I 2379.7
C 2216.8
GN
P 2303.5
I 2241.0
C 2182.4
HB
P 2312.2
I 2279.3
C 2216.9
JG
P 2501.7
I 2021.4
C 2219.3
LL
P 2414.4
I 2340.9
C 2248.8
MM
P 1881.4
I 2109.0
C 1887.5
OM
P 2368.5
I 2266.3
C 2245.9
RS
P 2152.9
C 1477.3
I 1717.8
SC
P 2378.4
I 2110.7
C 2148.9
SG
P 2488.1
I 2115.4
C 2018.1
SP
P 2452.9
I 2272.0
C 2380.7
26
3.2. Workflow and Data Analysis
The next section outlines the workflow of the data analysis portion of the project (Figure
4). The process is divided down into four main steps. The first section involves preparing the
data preparation, which involves the creation of the study area and raster projections. Next in the
workflow is the analysis of forest cover change using the Raster Calculator tool. A multiple ring
buffer analysis will also be conducted to provide baseline data for each of the thirty-nine
treatment sites. Finally, a regression analysis will be performed to discover any relationships
between the changes observed and topographic variables derived from a 5-meter digital elevation
model.
27
Figure 5 The workflow is summarized into four sections. The database symbol represents the
datasets used in the study. The purple box represents the change in forest cover layer. Green
boxes represent input and outputs of the workflow. Yellow boxes represent an analysis step.
3.2.1. Data Preparation
This study focuses on assessing forest cover change in the areas in and surrounding Las
Cruces Biological Reserve. The confines of the study area were created by finding the
overlapping regions from 2005 and 2014 forest cover layers using GIS. A polyline feature was
created to outline the extent of each forest cover layer and was then joined using the Union tool.
The polygon layer representing the overlapping regions was then exported using the Export
Features tool resulting in the boundary of the study area (shown in Figure 6). The 2014 forest
28
cover layer had a more considerable extent and was clipped to overlap with the 2005 forest cover
layer. The northeast 2014 cover layer shows a highly forested region in Costa Rica.
Figure 6 The map shows the extent of the 2005 and 2014 forest cover layers. The study area is
the region where both forest cover layers overlap. The 2014 forest cover layer had a much larger
extent than the 2005 layer and was clipped using the study boundary layer.
3.2.2. Forest Cover Change
To calculate the change in cover, the forest cover layers were clipped to the same study
area using Extract by Mask tool. The Raster Calculator tool was then used to assess the changes
in forest cover for nine years. Before the raster calculator could be run, both raster files had to be
resampled from their initial values of 0 (no cover) and 1(cover). The 2014 cover raster was
resampled so that areas of no cover were represented by the number 2, and areas with forest
29
cover were represented by the number 3. Similarly, the 2005 raster was resampled so that regions
with no cover were represented by the number 5, and areas with forest cover were represented by
the number 6. This was done to ensure that when these two rasters were added, they would
produce four unique values describing the possible change outcomes. For example, if these
rasters remained binary (0,1), then adding them would result in three possible outcomes (-1, 0,
and 1), with 0 representing areas of no change, which does not allow us to discern whether these
regions were cover and remained cover, or if they were no cover and remained no cover.
In adding the 2014 cover layer to the 2005 layer, we can see which areas (cells)
experienced gain, loss, or no change in forest cover. Figure 7 illustrates the logic behind the
Raster Calculator tool. An additional raster cell reclassification will be performed to separate the
regions that experienced no change to quantify which areas remained cover and which remained
no cover.
Figure 7 Illustration of how forest cover change was assessed using the Raster Calculator tool.
3.2.3. Multiple Ring Buffer
A multiple-ring buffer analysis was conducted to calculate changes in forest cover for
each of the thirty-nine treatments, or areas of interest (AOI), at various ring intervals from the
30
plot and supplement the ongoing research at these restoration sites. For each of ~ 50 x 50-meter
treatment plots, 11 concentric rings were placed at 50- to 1000- meter intervals, as shown in
Figure 8. From 50- to 200- meters, the rings were created at 50-meter intervals. Ring buffers
from 200 – 1000 meters from the plot were placed at 100-meter intervals.
Figure 8. The input setting used in the Multiple Ring Buffer Tool
The rings were created using the Multiple Ring Buffer Analysis tool (Figure 8). The treatment
plots (39 AOI’s) were used as the input features. The Dissolve Option was set to non-
overlapping rings so that the output would result in individual rings that did not cover the area of
the input feature. For example, the 50-meter ring covers the distance from the edge of the input
polygon and 50-meters outward. The 100-meter ring covers the area from 50- to 100-meters and
31
does not include the smallest ring. Figure 9 shows the multiple ring buffer configuration output.
Additionally, this study is only interested in assessing forest cover surrounding the treatment site,
or the input polygon, and, therefore, the area inside the input buffer was excluded.
Although the multiple rings will overlap for each of the treatment plots, this analysis is
interested in extracting forest cover for each of the treatment sites so they can be studied
individually in the future.
Figure 9 Multiple Ring buffer at site OM for the control treatment. This shows one site with
three treatments. The multiple ring buffer will be created for each treatment (39 multiple-ring
buffers).
Once the rings were created, the Intersect tool was used to find the forest cover change
regions within each circle. To do this, the cover change raster was first converted to a polygon
32
using the Raster to Polygon tool, since the Intersect tool only works on Feature Layer files. To
quantify the cover change within each ring, a workflow was created to quantify changes for each
ring at each of the sites, since there was a total of 429 rings between all sites.
Figure 10. The ModelBuilder workflow is used to quantify the cover changes within each ring
buffer. This process was automated since it would be too time-consuming to analyze each ring
manually.
The model in Figure 10 shows how the forest cover changes for each ring interval was
quantified. For each of the treatment plots, or AOI’s, multiple ring buffers were created, as
shown in Figure 8. After the buffers were created, the Intersect tool was used to find the polygon
areas of forest cover within each ring for each of the treatment plots. The Intersect tool works by
intersecting the ring polygon with the forest cover layer polygon, extracting the regions where
they both overlap.
3.3. Regression Analysis
This study also investigated the relationship between forest cover gain, elevation, aspect,
slope, and distance in a 200-meter area surrounding each of the thirteen restoration sites. First, a
33
30 x 30-meter grid was created using the Create Fishnet tool. Grid label points were also created
to derive the study variables within each cell. The grid was used as a container from which to
extract the variables percent gain, mean slope rise, mean aspect, and mean elevation values from
multiple raster datasets (Figure 11). The grid was created using the Create Fishnet tool, and the
cell width and height was set to 30-meters. The geometry type was set to polygon, and the extent
was set to that of the study region using the study area boundary layer as the input feature. Figure
14 shows the 30 x 30 fishnet grid at one of the research sites (OM).
Figure 11 The study extent for the regression analysis. Each point corresponds to a 30 x 30 -
meter grid. This was done for each of the 13 sites in the study.
The 5-meter DEM contains the elevation data, which was clipped to the extent of the
study region (Figure 12). The Zonal Statistics as a Table tool was then used to calculate the mean
34
elevation within each grid cell. Each grid cell contains a unique identified which will be used to
group multiple variables at each grid cell location.
Figure 12. The elevation profile for the study region in meters. The map shows the different
elevations of where the treatments are found.
The percent rise of the surface, or slope, was calculated using the Slope tool using DEM
as the input feature (Figure 13). The percent rise is the inclination of the slope calculated as
percent values, which range from 0 to infinity. A flat surface is represented by a value of 0, while
35
a 45-degree surface would have a rise in the slope of 100%. A high percent slope value
represents a more vertical surface. The output of the Slope tool is shown in Figure 13. Once
calculated, the mean slope was obtained using the Zonal Statistics as a Table tool. Each grid has
a corresponding grid label, which was used to extract the variables within each grid cell.
Figure 13. The output of the Slope tool. The red points represent the location of the treatment
plots. Each site has varying slope values.
36
The aspect was derived using the Aspect tool using the 5-meterDEM as the input feature.
Running the aspect tool provides an output in degrees, which is a circular measurement. All
variables must be linear to perform a linear regression analysis. The Raster Calculator tool was
used to transform aspects into a linear variable. The aspect output was first converted to radians,
and then the cosine function was used. The aspect values (in degrees) were first converted to
radians and then divided by 180, as shown in the equation below:
1° =
𝑝𝑖
180
𝑟𝑎𝑑𝑖𝑎𝑛𝑠
Next, applying the cosine function to the radian values generated a variable between
1(north) and -1 (south). In contrast, if one wanted to discover how east or west a surface faces,
they would apply the sine function to the radian value. This study is only interested in the effect
that north and south-facing slopes since, in the northern hemisphere, south-facing slopes are
typically warmer, drier, and less conducible to vegetation growth. The equation used in the
Raster calculator is shown in Figure 14.
𝐶𝑜𝑠 (
𝑚𝑎𝑡 ℎ. 𝑝𝑖 ∗ Aspect_5meters
180
)
Figure 14 Equation used in Raster Calculator to transform aspect from degrees into a linear
variable.
The output of aspect transformation is shown in Figure 15. The values range from
negative one to positive one corresponding to south-facing and north-facing aspects,
respectively.
37
Figure 15. The result of converting aspect into a linear measurement.
The final variable used in the regression is the distance from the site, specifically the
three treatment plots' geometric center. Before calculating the center of the study sites, a 200-
meter buffer was created surrounding each treatment plot using the Buffer tool. The rings were
dissolved to create a single polygon, and the geometric center of the polygon was calculated
using the Calculate Field Geometry tool. The geometric center of two restoration sites is shown
in Figure 16. This study site took the geometric center of the study region as the region
38
influencing the surrounding environment. The distance to the center was calculated rather than
the shortest distance to the nearest site since each site had varying treatment configurations.
Additionally, for the regression analysis, it is important not to sample the same area twice. For
this reason, the 200-meter buffer distance was chosen since this is the minimum distance
between site overlap.
Figure 16 The map shows the buffer regions used in the regression analysis. For each of the
sample points taken within the treatment buffer, the distance to the geometric center was
recorded and used in the regression analysis.
39
Figure 17 The map shows the 13 sites that were sampled in the regression analysis
The treatment site study area within the boundary of the study area is shown in Figure 17. Each
treatment site area is independent of each other, and the point samples do not overlap into other
treatment buffer boundaries. The sites vary in shape because of the differences in treatment plot
configuration found at each site. This was another justification for using the geometric center of
the study site to assess the relationship between forest cover gain and distance to the treatment
study area. Lastly, the Zonal Statistics as a Table tool as a table was used to calculate the mean
elevation, mean aspect, and mean slope within each 30-meter grid cell. Lastly, the distance from
the center of each grid cell to the study site's geometric center was calculated using the Generate
Near Table tool.
40
Chapter 4 Results
Chapter 4 describes the thesis results from GIS and statistical analysis. Section 4.1 presents the
cover change analysis findings by treatment type, site, and general study area. This analysis
provides baseline data for the ongoing and consequent projects near Las Cruces Biological
Station as well as demonstrates how the study region has changed over nine years. Section 4.3
shows the findings from the multiple ring buffer analysis for the treatment plots. Section 4.3
outlines the results from the multiple linear regression, which will investigate the relationships
between percent cover gain, elevation, aspect, slope, and distance surrounding the thirteen
treatment sites.
4.1. Forest Cover Change
The forest cover change analysis showed revealed a ~15% increase in cover between
2005 and 2014 for a study area spanning approximately 20,400 hectares. Additionally, the
analysis showed a 5% decrease in forest cover. Regions that experienced no change over the nine
years comprised ~ 80% of the study area. Overall, there was a net increase of about 9.7%
between the two time periods. Figure 13 illustrates the forest cover change for nine years. Table
3 lists the quantified categories of changes.
Table 3 Quantified forest cover change between 2005-2014. The rows cover, and no cover
represent regions that experienced no change.
Change Area (km) Hectares Percent %
No Cover 103864 10386.4 50.8
Gain 30086 3008.6 14.7
Loss 10296 1029.6 5.0
Cover 60095 6009.5 29.4
Total 204,342 20,434.2 100.0
41
Figure 18 Changes in forest cover between 2005 and 2014. Areas in blue have experienced a
gain in cover. We can see that horizontal growth in vegetation on the edges of larger forest
patches.
42
Figure 19 shows a 3D model of the study area based on the 5-meter digital elevation model. The
3D model can show the variations in the topography of the study area better than a flat map. In
the map, areas in red experienced forest cover loss and areas in blue experienced forest cover
gain Forest cover loss and gain are shown at a 3-meter raster resolution.
Figure 19 A 3D model of the study extent shows the treatment sites within the context of
the topography. The 5-meter DEM was used to generate the surface. The surface relief was
exaggerated by a factor of 2 to show the variation in landscape better. Areas in yellow are the
location of the treatment sites. Blue regions are regions that experienced a gain in forest cover,
and the red areas are those that suffered a loss in forest cover.
43
4.2. Multiple Ring Buffer
The multiple-ring buffer analysis was conducted for 39 treatment plots across 13 different
sites. Figure 11 shows the multiple ring buffer for the control plot at one of the 13 sites. The
three treatment plots at each site in all cases overlapped the ring buffer areas. The goal of this
was to quantify each treatment site individually so they could be studied independently from
each other in future studies. The multiple-ring analysis found that areas closer to the center of
the site experienced higher forest cover increases. This was expected since the regions within 50
meters contain the other treatment plots where there was active reforestation.
Regions at a 100-meter distance from the site experienced the second-highest mean
percent increase in forest cover. Although we generally observe a higher mean percent increase
in regions closer to the treatment plots, this trend does not consider how regions further from the
center covered a much greater area. For example, the ring buffer at a 900-meter distance covers a
circular area with likely higher variations in topography and landscape, particularly at opposite
ends of the buffer. Also, rings at 1000-meters from the plots covered 58,000 square meters, while
the 50-meter distance ring covered an area of 18,000 square meters (Table 4).
44
Figure 20 One example of the ring buffer analysis was conducted. This was performed for 39
treatment plots across 13 sites. The ring buffers overlapped due
Table 4 Multiple Ring Buffer distances with mean percent changes at each buffer distance across
39 treatment sites.
Distance Gain (%) Loss (%) No Cover (%) Cover (%) Total Area
50 25.8 4.8 30.4 38.9 17595.0
100 17.6 6.4 29.2 46.8 33480.8
150 14.7 7.1 33.5 44.8 49073.7
200 13.9 6.6 36.0 43.6 64757.4
300 13.4 7.1 41.2 38.4 176593.8
400 12.9 7.5 45.3 34.3 239375.0
500 13.4 7.5 43.8 35.4 302159.9
600 14.5 6.4 45.4 33.7 363677.3
700 14.1 6.3 46.7 32.9 421216.9
800 13.5 6.0 48.3 32.2 479144.0
900 13.2 5.9 48.5 32.4 537056.7
1000 12.4 6.0 49.6 32.0 578025.5
45
The average percent gain in cover across the sites ranged between 5% and 25%, with the
highest percent increase in the total area seen at Site OM (Figure 12). The lowest increase in
forest cover was observed at site LL with a mean increase of 5% across the three plots. The
highest percent loss was observed at site SC with an average of 12%. The lowest percent loss in
cover was recorded at site MM with a mean loss of ~2%. For 12 sites, the mean percent gain was
always higher than the mean percent loss, except for site SP, which had a ~9% loss and an ~8%
gain.
In the ring buffer analysis, we do not see apparent trends in forest cover loss and gain
concerning the percent forest cover already present. For example, the two treatment sites with the
highest mean percent loss in cover also experienced relatively high cover gains within nine
years. Sites BB and SC (Table 5) experienced the most significant mean percent loss but
similarly experienced high forest gain levels. The same can be said for areas that experienced the
highest mean percent gain across the nine years.
46
Figure 21 The percent gain and loss by each site.
Table 5 Cover distribution by site showing mean percentages of the three treatments.
Site % Cover % Gain % Loss % No Cover
AC 44.65 9.5 7.67 38.18
BB 23.98 20.98 11.08 43.97
EC 39.77 9.81 3.04 47.38
GN 38.61 19.81 8.35 33.22
HB 18.55 11.96 4.17 65.33
JG 64.15 12.32 4.8 18.73
LL 40.72 5.28 3.33 50.66
MM 75.41 11.38 2.43 10.77
OM 20.4 24.58 7.2 47.81
RS 34.65 18.58 6.86 39.91
SC 25.75 17.99 12.33 43.92
SG 13.58 24.32 3.07 59.04
SP 42.15 7.69 9.88 40.27
47
4.3. Regression Analysis
A multiple linear regression was conducted using JMP software. A multiple linear regression is
used to the relationship between a response variable and explanatory variables. The response, the
gain in forest cover, must be a continuous variable, and the explanatory variables can either be
continuous or categorical. The explanatory variables mean slope, mean aspect, mean elevation,
and distance to the geometric center was used as continuous explanatory variables within a 30 x
30-meter grid cell. A total of 1807 observations were sampled between 13 sites, 139 observations
per site. The response variable, forest cover gain, was calculated as the percent gain in forest
cover within a 30 x 30-meter grid. The multiple linear regression explored the relationship
between predictor and response variables within a 200-meter distance around the restoration
sites. Table 6 shows the variations in the forest cover gain and topographic variables between the
thirteen sites.
Table 6 The variations in percent forest cover gain, elevation, aspect, and slope between the
thirteen sites within a 200-meter buffer. The negative values for aspect represent south-facing
slopes with -1 being the most south-facing. Positive values for aspects represent more north-
facing slopes with values closer to 1 facing the most north.
Site Gain %
Elevation (m) Aspect Slope %
Min Max Min Max Min Max
AC 13.1 1277.11 1458.72 -0.98 0.85 2.11 135.41
BB 29.3 1212.85 1316.19 -0.97 0.92 9.31 91.64
EC 18.9 1151.49 1178.90 -0.92 0.96 0.18 45.32
GN 21.4 1142.14 1194.11 -0.86 1.00 1.44 63.27
HB 16.3 1087.97 1118.73 -0.99 0.93 3.08 42.51
JG 5.8 1145.58 1208.47 -0.52 0.99 4.66 57.90
LL 2.8 1131.27 1159.10 -0.90 0.96 1.21 43.80
MM 11.0 1041.75 1141.26 -0.79 0.98 1.21 56.61
OM 36.0 1109.60 1148.60 -0.97 0.77 0.74 60.86
RS 23.7 1165.01 1241.42 -0.96 0.96 1.47 67.04
SC 24.4 1083.48 1158.47 -0.91 0.93 6.14 75.98
SG 30.0 1090.88 1145.83 -0.95 0.85 2.18 61.27
SP 10.7 1266.08 1354.87 -0.98 0.97 3.60 82.07
48
The model began with the four predictor variables, and then a backward stepwise
regression approach was used to eliminate variables with no significant effects on the response
(P > 0.05). Of the four variables, only elevation, aspect, and distance were kept in the final
model; the mean slope showed no significance and was removed (P >0.05). Distance to the site’s
center, mean elevation and mean aspect all had a statistically significant impact on the percent
gain in forest cover (Table 6). The coefficient summary shows that for every single unit of
change in coefficients, there is a minimal, although significant, change in forest cover gain. For
the distance variable, moving away from the center of the site tends to result in a decrease in
forest cover gain, suggesting that forest cover gain is higher at distances closer to the center of
the site. Likewise, for elevation, an increase in elevation results in a decrease in forest cover
gain. Also, as the aspect becomes more north-facing, the percent gain decreases.
Table 7 Coefficients Summary
Variables Coefficient Std Error t Ratio Prob>|t|
Intercept 0.6554085 0.083712 7.83 <.0001*
Distance -0.000899 0.000131 -6.87 <.0001*
Elevation Mean -0.000308 7.012e-5 -4.39 <.0001*
Slope Mean 0.0002194 0.000361 0.61 0.5436
Aspect Mean -0.030849 0.011133 -2.77 0.0056*
The results of the regression show that the coefficient of determination, or the adjusted R Square,
had a value of 0.036, indicating that the independent variables explain approximately 3% of the
variability in percent forest gain.
Table 8 R Square values
R Square 0.038453
R Square Adjusted 0.036318
Root Mean Square Error 0.260242
Mean of Response 0.17746
Observations 1807
49
The model did not satisfy all the assumptions of a multiple linear regression needed to
validate whether the data were appropriate for the statistic. The Durbin Watson statistic test,
which is used to look for autocorrelation in the residuals, had a low value (d = 0.4868),
indicating a strong positive autocorrelation between variables (a value close to 2 suggests no
autocorrelation). Additionally, the significant p-value of 0.001 associated with the Durbin
Watson test allows us to reject the null hypothesis and further support that there is a first-order
positive autocorrelation. Another assumption was that of independent observations. The model
met this assumption as all the observations in the data sampling were independent of each other,
and no location was sampled twice. Additionally, there was normality in the distribution of the
data among the variables tested. The model also looked for linearity, analyzing whether certain
variables had a positive or negative linear effect on the amount of forest cover gained in an area.
50
Chapter 5 Discussion and Conclusion
This chapter discusses the findings, methodology, and implications of this study. The study had
three main goals:
1. Quantify forest cover change for two time periods and assess the overall changes.
2. Analyze and establish baseline data of cover change surrounding the treatment plots.
3. Understand the relationship that slope, elevation, aspect, and distance from the site have
on forest cover change using regression analysis.
5.1. Forest Change Cover
The study was successful in quantifying the changes in forest cover between the two time
periods. Conducting forest cover analysis with aerial imagery is more advantageous when
working with larger-scale studies as field checking methods are much more difficult. In
restoration studies, general forest cover metrics can be extracted, such as rate of change and
forest structure. For example, we can study how the cover surrounding restoration regions is
changing and understand the effect, if any, that it has restoration outcomes.
The overall changes suggest that this region, like the rest of Costa Rica, experienced
increases in forested areas between 2005 and 2014. Algeet-Abarquero et al. (2015) found that f.
Still, the study region is dominated by the unforested land cover (~56% no forest cover). In
comparison to the entire state of Costa Rica, which in 2010 was composed of 51% forest cover,
the study region near LCBS has less forest cover (approximately 44%) since 2014. However, this
study only analyzed the changes between two time periods and did not consider the year-to-year
variations in forest cover. A more accurate rate of reforestation can be attained if multiple years
were analyzed instead. As high-resolution aerial imagery becomes more easily accessible, future
analysis can continue to quantify forest cover change over more extended periods and at multiple
51
intervals. Forest cover analyses provide useful information on the distribution of cover and serve
to estimate ecosystem services, such as carbon sequestration.
Quantifying forest cover through field methods is impractical when working at larger
scales. Even at the local scale, such as evaluating the forest cover at each restoration site, in situ
methods are difficult to repeat on an annual basis. High-resolution imagery can produce sensible
estimates of forest cover and allow us to track changes throughout many years. This
consideration is especially important when considering that forest cover studies need to be
monitored across large temporal scales to evaluate long-term effects better.
When analyzing forest cover change through digitized imagery, it is essential to maintain
the same resolution across several different years to reduce error and improve accuracy in the
analysis. Although, conducting forest cover analysis of consecutive years through is likely to be
more difficult when working with aerial imagery, as it might be challenging to quantify slight
pixel variations. Likewise, conducting an accurate temporal analysis of forest cover change is
dependent on the scale. For example, assessing forest cover change through digital imagery
would be more accurate when working at larger scales since one is more likely to generalize
cover when working with a reduced resolution in the digitization process.
Additionally, when using the whole pixel classification of “cover” and “no cover” used in
the study, some areas are probably misrepresented, especially in more heterogeneous regions.
Other methodologies exist, however, that address this problem. Subpixel classification allows
one to estimate the percent canopy for each pixel as a number between 0 and 100 (Zhu 1994).
Forest cover change with subpixel classification is still possible and would allow for more
classifications of cover. Likewise, this methodology would better represent actual forest cover
and detect changes at a finer scale.
52
Other methods to quantify and study forest cover changes, such as using LIDAR, can
significantly improve forest cover detection (Walton, Nowak, and Greenfield 2008, although
they can be costly and inaccessible in some regions (Zahawi et al. 2015). Unlike LIDAR
technology, estimating cover change through aerial imagery is relatively inexpensive and
accurate (Walton, Nowak, and Greenfield 2008). Additionally, LIDAR can accurately measure
canopy height. Horizontal cover growth studies, such as done in this paper, do not take into
consideration the height of the canopy. Future studies can incorporate canopy height as a
variable to explain the variation in restoration outcomes.
5.2. Multiple Ring Buffer
The multiple ring buffer analysis provides useful baseline data for future studies
investigating the effects of forest cover change. Establishing a rate of forest cover increase for
the surrounding areas is essential to understanding the long-term impacts of restoration sites on
outer forest cover. For these nine years, we see a mean 15% increase in forest cover area for all
sites and a mean 6% loss across all sites.
The multiple ring buffer analysis revealed that the mean percent gain was always more
significant than the mean percent loss across all sites, except for site SP, which had a ~9% loss
and an ~8% gain. An individual linear regression on this site did not attribute the changes in loss
to any of the variables used in this study. This analysis suggests that SP is experiencing a higher
loss in forest cover compared to other regions. Future studies could investigate the variables
contributing to a higher loss in cover compared to forest cover gain. The mean percent gain
shows that a significant amount of forest cover was gained at distances up to 100 meters from the
site. Subsequent studies should consider this critical distance as forest cover change does not
change drastically at distances beyond 100 meters.
53
The multiple ring buffer analysis provided each a treatment site with a 9-year forest cover
analysis of the surrounding environment (see Appendix I). A future study could analyze multiple
years of aerial imagery to see the yearly increase in forest cover for the same region. Similarly, it
would be beneficial to examine the forest cover increase per year since the project's inception to
get a better timeline of the reforestation rate. This study only assessed the changes between two
time periods, which reduces the assumptions we can draw from the data.
5.3. Regression Analysis
The analysis indicated that slope, aspect, elevation, and distance to restoration had a
significant effect on forest cover gain; however, they were not able to explain the variability in
forest cover gain observed across the thirteen restoration sites. For the linear regression model,
the significance of the P-value indicates that we can reject the null hypothesis that the variables
did not affect forest cover gain. The low R-squared value suggests that the topographic variables
studied are not reliable predictors of whether an area would become forested.
The thirteen restoration sites are separated by a minimum and maximum distance of 0.7
and 8 kilometers, respectively (Zahawi et al. 2013). Similarly, they note that there each site
varies in elevation, ranging from a low of 1,060 meters to a high of 1,430 meters above sea level
(Zahawi et al. 2013). Likewise, each site has different measures of slope ranging between 5-35
degrees.
The study investigated whether the differences in topography contributed to any forest
cover gain at each site. Based on previous research, this study was expected to discover
relationships between the increase in forest cover gain and certain topographic variables. We
discovered that there was no strong relationship between the topographic variables. Had we
found a relationship between individual variables and increases in forest cover, future studies
54
evaluating the success of the treatment plots at each site could have attributed the outcomes to
these variations in topography derived variables. In contrast, the regression analysis did not
reveal a strong correlation between increases in forest cover gain and the surrounding
environment. The findings of this study, however, are useful to the ongoing research at these
sites because we can now deduce that any variations in restoration outcomes at each site are
likely not attributed to the variations in the topographic variables in this study.
Although this study was significant in scale, there was likely not enough variation in
topographic elements that could result in different rates in forest cover gain. According to the
First Law of Geography, things near each other are more related than things further apart. The
literature shows that higher elevations are typically associated with higher recovery rates because
these regions are more isolated and less affected by disturbances, such as agriculture, due to
unsuitability. For elevation, likely, there was not much disparity between minimum and
maximum elevation values between the study sites, as there was only about ~ 400-meter
difference in elevation. Similarly, the restoration sites were areas that were previously used for
agriculture, suggesting that these regions inherently shared similar elevation and slope profiles.
Additionally, this study assessed the influence that distance to the restoration site had
changes in forest cover in the surrounding environment. We expected to find that a closer
distance to the geometric center of the study would result in a higher percentage of forest cover
gain. This assumption was made under the belief that active restoration practices can act as a
catalyst for surrounding forest cover growth. This study was able to show a more significant
percentage gain in cover in areas closer to the site's primary area of influence, which in this study
was taken as the geometric center. One explanation could be that the surrounding forest cover for
each of the sites has varying levels of development. For example, some regions are surrounded
55
by significant roads, while others are surrounded entirely by forest cover. The distance to the
geometric center was used to assess how the center of the site influenced the external
environment. As a recommendation for future studies, the distance variable should be calculated
as the shortest distance to the nearest treatment plot, rather than the distance to the geometric
center. Measuring the distance of the sample points to the geometric center likely introduced
error.
Likewise, we expected to find a positive correlation between northern facing slopes and
gain in forest cover since these regions are more conducible to tree growth in the Northern
hemisphere. The study found no strong correlations between north-facing slopes and increases in
forest cover. Unlike temperate regions that experience more seasonality due to their distance
from the equator, Costa Rica experiences less seasonality as tropical areas are characterized by
two seasons: summer and winter. Less seasonality and the relative position of the sun in tropical
regions can explain why aspect does not have a strong influence on vegetation growth.
Also, despite the variation in aspect, the sun hits tropical regions overhead much more
than in the temperate areas in the northern hemisphere. Therefore, solar radiation does not likely
vary by aspect. The findings of this study are consistent with this as they show that aspect did not
influence the changes in forest cover observed. Future studies in the study region can assume that
aspect has very little influence on forest cover in the tropics. The findings of this study suggest
that more complex factors are at play regarding forest cover changes, particularly at restoration
sites in areas previously used for agriculture.
Other considerations for subsequent studies should also assess the biodiversity in the
surrounding regions through more complex datasets. For example, how do reforestation rates
56
compare when looking at regions near the old growth forest against secondary forests. Old-
growth forests are less disturbed regions and contain much lower levels of human disturbance.
Reid et al. 2018 used aerial photography to evaluate the persistence of secondary forest
cover in Costa Rica and found that larger forest fragments and proximity to rivers were strong
predictors of whether forest regions persisted over 54 years. If proximity to rivers is a strong
predictor of forest persistence, then we might find forest cover gain to be more strongly
correlated with nearer distances to the river. Correspondingly, we might forest cover loss to be
more strongly correlated with treatment sites being a further distance from rivers.
There were many other essential variables this study did not consider. For example, some
sites were surrounded by roads and developed lands, areas where forest cover will not change
regardless of the influence that the restoration site. Other studies can consider how distance to
the nearest road or building affects the changes in forest cover observed at each site. Based on
the low adjusted R-square value, there are many unexplored variables that can be contributing to
the variation in forest cover gain at each study site.
5.4. Conclusion
This study aimed to evaluate the changes in forest cover surrounding restoration sites at
Las Cruces Biological Reserve. Ongoing research at this facility is relevant because it holds
some of Costa Rica's last remaining old-growth forest. This study's scope was to evaluate
landscape-level changes, which is difficult because accessibility to finer datasets, such as
LIDAR, is expensive in remote regions. Aerial imagery provides an inexpensive alternative to
assess forest cover changes but provides minimal information on the context of these changes.
The findings of this study provide useful information for this study. The restoration
treatment sites were initiated in regions with varying degrees of aspect, slope, and elevation;
57
however, this study did could not strongly attribute the observed changes to these variables.
Although the study did not reveal any correlation between landscape-level elements, quantifying
forest cover, and creating baseline data, future studies can still incorporate these elements as it
can be essential to the overall understanding of forest cover changes.
Future studies can evaluate performance metrics against baseline data to uncover whether
initial forest cover had any long-term effects on the restoration outcomes. Although the data
suggest a weak correlation between topographic variables, the study limited the explanation of
forest cover changes to topographic variables. Subsequent studies can use other variables, such
as distance to the nearest river, as other studies have shown the persistence of forest cover near
riparian regions (Reid et al. 2018).
Although the multiple-ring study did not see any trend of increasing forest cover in
regions with high forest cover, this study did not consider the patch size instead of focusing on
the percent cover within a region. Previous studies have demonstrated how patch size strongly
affects the recovery rate (Holl et al. 2017). A consequent analysis could assess the degree of
connectivity in forest cover and see whether the patch size correlates with the changes observed.
The measurement of landscape patterns and structure is becoming more easily calculated
with advances in data collection and software. Land managers can assess and monitor landscape
patterns and the effects they have on the ecological processes. FRAGSTATS is a program that is
used to quantify landscape structures from remotely sense data and can assess landscape-level
elements such as size, shape, connectivity, and diversity. The data obtained from FRAGSTATS
is also used in correlation analysis in large scale studies (Kupfer 2012). Future studies can assess
the landscape metrics and their influence on restoration outcomes, particularly in regions
surrounding restoration sites. For example, landscape metrics regarding connectivity can be
58
measured and used in future studies to assess their effect on the gain in forest cover for areas
surrounding restoration sites, helping landscape managers understand the importance of
landscape connectivity within the context of ecological restoration efforts.
The development of effective restoration practices will require researchers to understand
further the elements contributing to forest regeneration. Through active restoration practices,
land managers can alter the ecological trajectory of pasture lands into those closely resembling
natural succession. Understanding the landscape context to evaluate the efficacy of restoration
strategies is essential to understanding how regeneration behaves in shifting landscapes. This
study attempted to uncover whether the changes in the surrounding landscape could be attributed
to variations in topographic variables surrounding the restoration sites. The implications of this
study suggest that slope, elevation, aspect, and distance to restoration sites are not reliable
predictors on whether areas reforested. Also, we can suggest that the variations in restoration
outcomes, specifically relating to forest cover growth, are not like influenced by the variations in
topographic variables at each of the restoration sites.
59
Bibliography
Alsfeld, Amy J., Jacob L. Bowman, and Amy Deller‐Jacobs. 2010. "The Influence of Landscape
Composition on the Biotic Community of Constructed Depressional Wetlands." Restoration
Ecology 18: 370-378.
Bell, Susan S., Mark S. Fonseca, and Little B. Motten. 1997. "Linking Restoration and
Landscape Ecology." Restoration Ecology 5 (4): 318-323.
Bortoleto, Ludmila Araujo, Cassio Jose Montagnani Figueira, John B. Dunning Jr, John Rodgers,
and Alexandre Marco da Silva. 2016. "Suitability Index for Restoration in Landscapes: An
Alternative Proposal for Restoration Projects." Ecological Indicators 60: 724-735.
Boutin, Stan and Daryll Hebert. 2002. "Landscape Ecology and Forest Management: Developing
an Effective Partnership." Ecological Applications 12 (2): 390-397.
Brudvig, Lars A. 2011. "The Restoration of Biodiversity: Where has Research been and Where
does it Need to Go?" American Journal of Botany 98 (3): 549-558.
Brudvig, Lars A., Ellen I. Damschen, Joshua J. Tewksbury, Nick M. Haddad, and Douglas J.
Levey. 2009. "Landscape Connectivity Promotes Plant Biodiversity Spillover into Non-
Target Habitats." Proceedings of the National Academy of Sciences 106 (23): 9328-9332.
Calvo-Alvarado, J., B. McLennan, A. Sánchez-Azofeifa, and Theresa Garvin. 2009.
"Deforestation and Forest Restoration in Guanacaste, Costa Rica: Putting Conservation
Policies in Context." Forest Ecology and Management 258 (6): 931-940.
Chazdon, Robin L. 2008. "Beyond Deforestation: Restoring Forests and Ecosystem Services on
Degraded Lands." Science 320 (5882): 1458-1460.
Clewell, Andre and John P. Rieger. 1997. "What Practitioners Need from Restoration
Ecologists." Restoration Ecology 5 (4): 350-354.
Cole, Rebecca J., Karen D. Holl, and Rakan A. Zahawi. 2010. "Seed Rain Under Tree Islands
Planted to Restore Degraded Lands in a Tropical Agricultural Landscape." Ecological
Applications 20 (5): 1255-1269.
Crk, Tanja, María Uriarte, Fabio Corsi, and Dan Flynn. 2009. "Forest Recovery in a Tropical
Landscape: What is the Relative Importance of Biophysical, Socioeconomic, and Landscape
Variables?" Landscape Ecology 24 (5): 629-642.
Crouzeilles, Renato, Michael Curran, Mariana S. Ferreira, David B. Lindenmayer, Carlos EV
Grelle, and José M. Rey Benayas. 2016. "A Global Meta-Analysis on the Ecological Drivers
of Forest Restoration Success." Nature Communications 7 (1): 1-8.
60
Cunningham, Mary Ann. 2006. "Accuracy Assessment of Digitized and Classified Land Cover
Data for Wildlife Habitat." Landscape and Urban Planning 78 (3): 217-228.
De la Peña‐Domene, Marinés, Emily S. Minor, and Henry F. Howe. 2016a. "Restored
Connectivity Facilitates Recruitment by an Endemic Large‐seeded Tree in a Fragmented
Tropical Landscape." Ecology 97 (9): 2511-2517.
———. 2016b. "Restored Connectivity Facilitates Recruitment by an Endemic Large‐seeded
Tree in a Fragmented Tropical Landscape." Ecology 97 (9): 2511-2517.
De Souza Leite, Melina, Leandro Reverberi Tambosi, Isabella Romitelli, and Jean Paul Metzger.
2013. "Landscape Ecology Perspective in Restoration Projects for Biodiversity
Conservation: A Review." Natureza & Conservao 11: 108-118.
Elmqvist, Thomas, H. Setälä, S. N. Handel, S. Van Der Ploeg, J. Aronson, James Nelson
Blignaut, Erik Gomez-Baggethun, D. J. Nowak, J. Kronenberg, and R. De Groot. 2015.
"Benefits of Restoring Ecosystem Services in Urban Areas." Current Opinion in
Environmental Sustainability 14: 101-108.
Fink, Rebecca D., Catherine A. Lindell, Emily B. Morrison, Rakan A. Zahawi, and Karen D.
Holl. 2009. "Patch Size and Tree Species Influence the Number and Duration of Bird Visits
in Forest Restoration Plots in Southern Costa Rica." Restoration Ecology 17 (4): 479-486.
Holl, Karen D. and Maarten Kappelle. 1999. "Tropical Forest Recovery and Restoration." Trends
in Ecology & Evolution 14 (10): 378-379.
Holl, Karen D., Michael E. Loik, Eleanor HV Lin, and Ivan A. Samuels. 2000. "Tropical
Montane Forest Restoration in Costa Rica: Overcoming Barriers to Dispersal and
Establishment." Restoration Ecology 8 (4): 339-349.
Holl, Karen D., John Leighton Reid, José Miguel Chaves‐Fallas, Federico Oviedo‐Brenes, and
Rakan A. Zahawi. 2017. "Local Tropical Forest Restoration Strategies Affect Tree
Recruitment More Strongly than does Landscape Forest Cover." Journal of Applied Ecology
54 (4): 1091-1099.
Houlahan, Jeff E., Paul A. Keddy, Kristina Makkay, and C. Scott Findlay. 2006. "The Effects of
Adjacent Land use on Wetland Species Richness and Community Composition." Wetlands
26 (1): 79-96.
Keenan, Rodney J., Gregory A. Reams, Frédéric Achard, Joberto V. de Freitas, Alan Grainger,
and Erik Lindquist. 2015. "Dynamics of Global Forest Area: Results from the FAO Global
Forest Resources Assessment 2015." Forest Ecology and Management 352: 9-20.
Kupfer, John A. 2012. "Landscape ecology and biogeography: rethinking landscape metrics in a
post-FRAGSTATS landscape." Progress in physical geography 36, no. 3 (2012): 400-420.
61
Leopold, A. Carl, R. Andrus, A. Finkeldey, and D. Knowles. 2001. "Attempting Restoration of
Wet Tropical Forests in Costa Rica." Forest Ecology and Management 142 (1-3): 243-249.
Lindell, Catherine A., Rebecca J. Cole, Karen D. Holl, and Rakan A. Zahawi. 2012. "Migratory
Bird Species in Young Tropical Forest Restoration Sites: Effects of Vegetation Height,
Planting Design, and Season." Bird Conservation International 22 (1): 94-105.
Måren, Inger Elisabeth, Sikha Karki, Chanda Prajapati, Ram Kailash Yadav, and Bharat Babu
Shrestha. 2015. "Facing North Or South: Does Slope Aspect Impact Forest Stand
Characteristics and Soil Properties in a Semiarid Trans-Himalayan Valley?" Journal of Arid
Environments 121: 112-123.
Munro, Nicola T., David B. Lindenmayer, and Joern Fischer. 2007. "Faunal Response to
Revegetation in Agricultural Areas of Australia: A Review." Ecological Management &
Restoration 8 (3): 199-207.
Nowak, David J., Rowan A. Rowntree, E. Gregory McPherson, Susan M. Sisinni, Esther R.
Kerkmann, and Jack C. Stevens. 1996. "Measuring and Analyzing Urban Tree Cover."
Landscape and Urban Planning 36 (1): 49-57.
Palmer, Margaret A., Joy B. Zedler, and Donald A. Falk. 2016. "Ecological Theory and
Restoration Ecology." In Foundations of Restoration Ecology, 3-26: Springer.
Park, Chris C. 2002. Tropical Rainforests Routledge.
Pelz, Kristen A. and Yvette L. Dickinson. 2014. "Monitoring Forest Cover Spatial Patterns with
Aerial Imagery: A Tutorial." Technical Brief CFRI-TB-1401, Colorado Forest Restoration
Institute, Colorado State University, Fort Collins, Colorado.
Peng, Shaolin, Ting Zhou, Liyin Liang, and Wentao Ren. 2012. "Landscape Pattern Dynamics
and Mechanisms during Vegetation Restoration: A Multiscale, Hierarchical Patch Dynamics
Approach." Restoration Ecology 20 (1): 95-102.
Rappaport, Danielle I., Leandro R. Tambosi, and Jean P. Metzger. 2015. "A Landscape Triage
Approach: Combining Spatial and Temporal Dynamics to Prioritize Restoration and
Conservation." Journal of Applied Ecology 52 (3): 590-601.
Reid, J. Leighton, Karen D. Holl, and Rakan A. Zahawi. 2015. "Seed Dispersal Limitations Shift
Over Time in Tropical Forest Restoration." Ecological Applications 25 (4): 1072-1082.
Reid, J. Leighton, Chase D. Mendenhall, J. Abel Rosales, Rakan A. Zahawi, and Karen D. Holl.
2014. "Landscape Context Mediates Avian Habitat Choice in Tropical Forest Restoration."
PLoS One 9 (3).
Ruiz‐Jaen, Maria C. and T. Mitchell Aide. 2005. "Restoration Success: How is it being
Measured?" Restoration Ecology 13 (3): 569-577.
62
Sader, Steven A. and Armond T. Joyce. 1988. "Deforestation Rates and Trends in Costa Rica,
1940 to 1983." Biotropica: 11-19.
Stouffer, Philip C. and Richard O. Bierregaard Jr. 1995. "Use of Amazonian Forest Fragments by
Understory Insectivorous Birds." Ecology 76 (8): 2429-2445.
Tambosi, Leandro R., Alexandre C. Martensen, Milton C. Ribeiro, and Jean P. Metzger. 2014.
"A Framework to Optimize Biodiversity Restoration Efforts Based on Habitat Amount and
Landscape Connectivity." Restoration Ecology 22 (2): 169-177.
Thomlinson, John R., Mayra I. Serrano, Tania del M. Lopez, T. Mitchell Aide, and Jess K.
Zimmerman. 1996. "Land-use Dynamics in a Post-Agricultural Puerto Rican Landscape
(1936-1988)." Biotropica: 525-536.
Walton, Jeffrey T., David J. Nowak, and Eric J. Greenfield. 2008. "Assessing Urban Forest
Canopy Cover using Airborne Or Satellite Imagery."
Wijdeven, Sander MJ and Mirjam E. Kuzee. 2000. "Seed Availability as a Limiting Factor in
Forest Recovery Processes in Costa Rica." Restoration Ecology 8 (4): 414-424.
Young, Truman P. 2000. "Restoration Ecology and Conservation Biology." Biological
Conservation 92 (1): 73-83.
Zahawi, Rakan A., John L. Reid, and Karen D. Holl. 2014 "Hidden costs of passive restoration."
Restoration Ecology 22, no. 3: 284-287.
Zahawi, Rakan A., Jonathan P. Dandois, Karen D. Holl, Dana Nadwodny, J. Leighton Reid, and
Erle C. Ellis. 2015. "Using Lightweight Unmanned Aerial Vehicles to Monitor Tropical
Forest Recovery." Biological Conservation 186: 287-295.
Zahawi, Rakan A., Federico Oviedo-Brenes, and Chris J. Peterson. 2017. "A Degradation Debt?
Large-Scale Shifts in Community Composition and Loss of Biomass in a Tropical Forest
Fragment After 40 Years of Isolation." PloS One 12 (8).
63
Appendix I
Ring Buffer Data
Site Treatment Distance Cover Cover Gain Gain Loss Loss NoCover No Cover Total Area
(m) (m) % m % m % m % m
AC C 50 7771.1 46.2 1929.4 11.5 420.2 2.5 6691.6 39.8 16812.3
100 19470.7 59.9 6111.2 18.8 469.5 1.4 6457.9 19.9 32509.3
150 35069.2 72.7 3231.2 6.7 2207.7 4.6 7698.4 16.0 48206.4
200 42404.4 66.4 2202.6 3.4 6768.3 10.6 12528.3 19.6 63903.6
300 88152.2 50.4 5186.0 3.0 14005.6 8.0 67554.4 38.6 174898.3
400 74401.8 31.3 11177.7 4.7 18154.0 7.6 133953.3 56.4 237686.7
500 94862.2 31.6 23640.0 7.9 33861.6 11.3 148111.3 49.3 300475.1
600 164846.2 45.4 38417.0 10.6 25130.9 6.9 134869.5 37.1 363263.5
700 182553.2 42.8 54582.6 12.8 35450.8 8.3 153465.2 36.0 426051.8
800 205578.4 42.1 56174.8 11.5 35979.0 7.4 191108.0 39.1 488840.2
900 211549.1 40.0 52428.0 9.9 42718.2 8.1 222571.7 42.1 529267.0
1000 118965.0 26.6 32977.8 7.4 47876.8 10.7 248005.6 55.4 447825.1
AC I 50 8611.9 52.8 3727.4 22.9 429.5 2.6 3529.3 21.7 16298.2
100 20360.6 63.6 4661.3 14.6 1155.8 3.6 5817.6 18.2 31995.2
150 30139.3 63.2 2920.7 6.1 4635.3 9.7 9996.9 21.0 47692.2
200 38892.4 61.4 1532.0 2.4 5351.3 8.4 17613.5 27.8 63389.2
300 76607.9 44.1 6956.9 4.0 15422.4 8.9 74882.1 43.1 173869.3
400 75144.3 31.8 11979.7 5.1 17931.1 7.6 131602.3 55.6 236657.3
500 94535.9 31.6 27920.0 9.3 30532.3 10.2 146457.1 48.9 299445.3
600 165756.7 45.8 55595.0 15.3 24848.5 6.9 116033.1 32.0 362233.2
700 181809.8 42.8 35375.2 8.3 43421.3 10.2 164414.8 38.7 425021.1
800 202572.7 41.5 59410.4 12.2 39422.3 8.1 186403.7 38.2 487809.0
900 221670.5 40.3 46582.4 8.5 49915.0 9.1 232426.2 42.2 550594.1
1000 121127.0 28.5 38214.9 9.0 32346.0 7.6 232876.3 54.9 424564.2
AC P 50 7478.9 42.6 5474.4 31.2 234.9 1.3 4351.5 24.8 17539.8
100 16465.7 49.5 2568.1 7.7 3716.9 11.2 10486.1 31.5 33236.7
150 25006.8 51.1 1367.4 2.8 4377.1 8.9 18182.4 37.2 48933.7
200 33363.0 51.6 2069.5 3.2 4141.3 6.4 25056.9 38.8 64630.7
300 71111.4 40.3 6916.7 3.9 14016.1 7.9 84308.0 47.8 176352.2
400 85211.9 35.6 11456.4 4.8 17562.4 7.3 124909.3 52.2 239140.1
500 111960.0 37.1 32658.5 10.8 33597.4 11.1 123712.1 41.0 301928.0
600 173352.9 47.5 51315.6 14.1 25639.2 7.0 114408.0 31.4 364715.8
700 172266.0 40.3 37161.1 8.7 37633.8 8.8 180442.7 42.2 427503.6
800 178521.3 36.4 46381.5 9.5 48074.6 9.8 217314.1 44.3 490291.5
900 210755.4 38.1 56953.7 10.3 49071.4 8.9 236298.9 42.7 553079.5
1000 171085.0 34.4 46190.5 9.3 35525.9 7.1 244244.3 49.1 497045.7
BB C 50 2626.6 14.8 8762.4 49.3 1223.1 6.9 5159.9 29.0 17772.0
100 7031.5 21.0 9109.6 27.2 4328.0 12.9 12999.8 38.8 33469.0
150 10151.4 20.6 12498.6 25.4 8063.7 16.4 18452.2 37.5 49166.0
200 16117.5 24.8 12302.2 19.0 7514.3 11.6 28929.0 44.6 64863.0
300 27292.1 15.4 32785.9 18.5 22384.3 12.7 94354.6 53.4 176816.9
400 62973.5 26.3 46117.5 19.2 29106.0 12.1 101407.8 42.3 239604.9
500 99607.3 32.9 51622.1 17.1 36176.7 12.0 114986.7 38.0 302392.8
600 84004.5 23.0 57802.7 15.8 42372.1 11.6 181001.6 49.6 365180.8
700 79446.0 18.6 72845.1 17.0 45274.3 10.6 230403.4 53.8 427968.7
800 115181.6 23.5 95070.5 19.4 44678.0 9.1 235826.6 48.1 490756.7
900 186970.2 33.8 89445.5 16.2 51892.9 9.4 225236.1 40.7 553544.7
1000 211511.4 34.3 94662.0 15.4 61070.4 9.9 249088.7 40.4 616332.5
BB I 50 4467.6 27.0 6345.0 38.3 1431.1 8.6 4330.3 26.1 16574.0
100 7126.0 22.1 7715.8 23.9 4550.2 14.1 12878.8 39.9 32270.9
150 11504.8 24.0 8471.0 17.7 5356.4 11.2 22635.6 47.2 47967.8
200 10151.6 15.9 10635.5 16.7 8016.4 12.6 34861.2 54.8 63664.7
300 32571.6 18.7 36812.0 21.1 21468.9 12.3 83567.4 47.9 174419.9
400 50729.4 21.4 43664.4 18.4 27635.5 11.7 115178.1 48.6 237207.4
500 103836.3 34.6 51914.4 17.3 39482.3 13.2 104762.0 34.9 299995.0
600 94222.7 26.0 57301.0 15.8 37388.2 10.3 173870.6 47.9 362782.5
700 79395.2 18.7 73522.2 17.3 49058.4 11.5 223594.1 52.5 425570.0
800 110005.0 22.5 79509.4 16.3 55526.5 11.4 243316.6 49.8 488357.5
900 167235.2 30.3 92342.6 16.8 46215.2 8.4 245352.1 44.5 551145.0
1000 196604.5 32.0 91055.2 14.8 57718.9 9.4 268553.8 43.7 613932.4
BB P 50 2950.3 16.6 5015.4 28.2 2252.2 12.7 7579.5 42.6 17797.4
100 3330.6 9.9 11971.4 35.7 2298.7 6.9 15893.7 47.5 33494.4
64
Site Treatment Distance Cover Cover Gain Gain Loss Loss NoCover No Cover Total Area
(m) (m) % m % m % m % m
150 13547.5 27.5 13313.3 27.1 4784.6 9.7 17546.1 35.7 49191.5
200 9267.5 14.3 16763.0 25.8 7405.7 11.4 31452.3 48.5 64888.5
300 42711.6 24.1 35692.9 20.2 21891.1 12.4 76572.4 43.3 176868.1
400 56034.9 23.4 42050.9 17.5 34383.4 14.3 107187.1 44.7 239656.3
500 97905.4 32.4 46226.1 15.3 36421.2 12.0 121891.7 40.3 302444.4
600 76764.3 21.0 70525.9 19.3 45786.8 12.5 172155.6 47.1 365232.6
700 78927.1 18.4 75708.0 17.7 41804.3 9.8 231581.3 54.1 428020.7
800 112318.7 22.9 97635.4 19.9 39720.1 8.1 241134.8 49.1 490808.9
900 210881.7 38.1 98027.6 17.7 57981.5 10.5 186706.3 33.7 553597.1
1000 199292.4 32.3 103886.4 16.9 53666.8 8.7 259539.4 42.1 616385.1
EC C 50 9835.0 56.8 4637.9 26.8 289.2 1.7 2540.2 14.7 17302.3
100 21956.9 66.5 5566.8 16.9 426.9 1.3 5048.8 15.3 32999.4
150 28959.4 59.5 9648.0 19.8 579.8 1.2 9509.2 19.5 48696.4
200 44395.8 68.9 6442.7 10.0 1228.7 1.9 12326.3 19.1 64393.5
300 96207.0 54.7 13526.0 7.7 3478.4 2.0 62666.7 35.6 175878.1
400 81086.3 34.0 7251.9 3.0 5026.6 2.1 145301.5 60.9 238666.4
500 96931.2 32.2 9661.1 3.2 5776.9 1.9 189085.5 62.7 301454.6
600 100127.4 27.5 12898.5 3.5 12515.4 3.4 238701.5 65.5 364242.9
700 116855.3 27.4 16939.7 4.0 24279.7 5.7 268956.4 63.0 427031.1
800 135661.1 27.7 23570.6 4.8 15599.0 3.2 314988.7 64.3 489819.4
900 120527.7 21.8 42407.8 7.7 26037.5 4.7 363634.7 65.8 552607.6
1000 128214.4 20.8 51535.0 8.4 35969.5 5.8 399676.7 64.9 615395.7
EC I 50 6186.7 35.1 5625.0 31.9 360.6 2.0 5448.0 30.9 17620.4
100 19418.5 58.3 9784.8 29.4 484.7 1.5 3629.4 10.9 33317.4
150 39982.5 81.6 7507.3 15.3 246.1 0.5 1278.5 2.6 49014.4
200 49986.0 77.2 3923.8 6.1 926.2 1.4 9875.5 15.3 64711.4
300 85369.7 48.4 11039.3 6.3 4247.6 2.4 75857.2 43.0 176513.8
400 86558.0 36.2 7589.0 3.2 4867.2 2.0 140287.7 58.6 239301.9
500 100406.5 33.2 11758.4 3.9 6272.1 2.1 183652.8 60.8 302089.9
600 100004.8 27.4 14343.1 3.9 7413.2 2.0 243116.8 66.6 364878.0
700 107588.6 25.2 19207.3 4.5 27837.9 6.5 273032.2 63.8 427665.9
800 122653.1 25.0 28640.1 5.8 17867.9 3.6 321292.9 65.5 490454.0
900 122101.4 22.1 46489.5 8.4 28628.2 5.2 356023.0 64.4 553242.1
1000 124925.7 20.3 49512.8 8.0 39680.5 6.4 401911.0 65.2 616030.0
EC P 50 9402.4 53.3 5948.9 33.7 248.5 1.4 2027.6 11.5 17627.4
100 23144.0 69.5 2556.7 7.7 896.1 2.7 6727.6 20.2 33324.4
150 25697.5 52.4 5710.3 11.6 1181.7 2.4 16431.9 33.5 49021.5
200 24772.5 38.3 6958.8 10.8 1277.7 2.0 31709.6 49.0 64718.6
300 64690.9 36.6 11016.8 6.2 3844.6 2.2 96975.9 54.9 176528.2
400 79457.7 33.2 14600.8 6.1 5328.0 2.2 139930.0 58.5 239316.5
500 92401.4 30.6 11093.1 3.7 18600.3 6.2 180009.9 59.6 302104.8
600 105973.3 29.0 16698.6 4.6 15233.6 4.2 226987.5 62.2 364893.0
700 120256.9 28.1 17793.3 4.2 13227.2 3.1 276403.8 64.6 427681.2
800 111685.7 22.8 26187.5 5.3 19199.2 3.9 333397.0 68.0 490469.5
900 132464.7 24.0 43546.5 7.9 23590.3 4.3 353331.1 63.9 552932.6
1000 149523.7 26.3 49508.5 8.7 24317.7 4.3 344412.0 60.7 567762.0
GN C 50 8710.4 50.5 3721.2 21.6 1572.5 9.1 3236.2 18.8 17240.3
100 15952.7 48.4 8693.4 26.4 2724.3 8.3 5566.9 16.9 32937.3
150 20445.1 42.0 10577.8 21.7 6180.7 12.7 11430.7 23.5 48634.3
200 26103.8 40.6 12455.8 19.4 5987.0 9.3 19784.8 30.8 64331.3
300 63855.3 36.3 39178.7 22.3 14438.8 8.2 58280.7 33.2 175753.5
400 66307.0 27.8 45716.3 19.2 17706.5 7.4 108811.6 45.6 238541.5
500 108916.9 36.1 65902.3 21.9 19280.0 6.4 107230.1 35.6 301329.4
600 126522.9 34.7 83223.7 22.9 29458.1 8.1 124912.7 34.3 364117.4
700 119167.4 27.9 92776.2 21.7 33944.4 8.0 181017.4 42.4 426905.3
800 178245.2 36.4 84621.3 17.3 39009.7 8.0 187817.0 38.4 489693.2
900 225818.5 40.9 80742.2 14.6 42509.2 7.7 203411.4 36.8 552481.3
1000 260419.0 42.3 75551.2 12.3 46735.7 7.6 232563.1 37.8 615269.0
GN I 50 9136.6 52.6 1801.2 10.4 1968.1 11.3 4452.4 25.6 17358.3
100 18408.8 55.7 7583.5 22.9 3037.1 9.2 4025.9 12.2 33055.2
150 18286.7 37.5 11099.5 22.8 5955.2 12.2 13410.9 27.5 48752.2
200 20702.1 32.1 15554.0 24.1 5781.0 9.0 22412.2 34.8 64449.3
300 72516.8 41.2 37189.6 21.1 12787.8 7.3 53495.1 30.4 175989.4
400 78111.3 32.7 45067.6 18.9 14926.9 6.3 100671.6 42.2 238777.4
500 108560.8 36.0 65940.1 21.9 20856.2 6.9 106208.3 35.2 301565.4
600 130389.1 35.8 78126.2 21.4 30788.6 8.5 125049.4 34.3 364353.3
700 124830.5 29.2 101374.4 23.7 31661.4 7.4 169275.0 39.6 427141.3
65
Site Treatment Distance Cover Cover Gain Gain Loss Loss NoCover No Cover Total Area
(m) (m) % m % m % m % m
800 187946.4 38.4 75844.7 15.5 41023.3 8.4 185114.9 37.8 489929.3
900 235208.7 42.6 76390.1 13.8 42156.1 7.6 198962.3 36.0 552717.2
1000 243354.2 39.5 79785.3 13.0 47603.1 7.7 244762.5 39.8 615505.1
GN P 50 4660.6 26.6 6982.1 39.8 1342.7 7.7 4558.1 26.0 17543.5
100 18508.9 55.7 4278.3 12.9 2881.0 8.7 7572.0 22.8 33240.3
150 23983.6 49.0 8344.8 17.1 3873.0 7.9 12735.7 26.0 48937.1
200 28926.7 44.8 11342.4 17.5 6657.8 10.3 17707.0 27.4 64634.0
300 48004.1 27.2 39793.6 22.6 17440.9 9.9 71119.7 40.3 176358.3
400 64398.9 26.9 64694.7 27.1 19061.7 8.0 90990.4 38.0 239145.6
500 117031.6 38.8 60598.9 20.1 19845.9 6.6 104456.3 34.6 301932.9
600 128560.2 35.2 79908.1 21.9 31376.1 8.6 124875.7 34.2 364720.1
700 126791.2 29.7 80321.9 18.8 36472.5 8.5 183921.8 43.0 427507.4
800 180979.5 36.9 78859.9 16.1 33041.8 6.7 197413.4 40.3 490294.6
900 207876.9 37.6 88214.1 15.9 42887.7 7.8 214103.2 38.7 553081.9
1000 273157.3 44.4 79106.5 12.8 45899.3 7.5 217706.0 35.3 615869.0
HB C 50 2148.9 12.4 4045.0 23.4 456.5 2.6 10657.1 61.6 17307.3
100 4594.3 13.9 7114.2 21.6 1098.9 3.3 20196.9 61.2 33004.3
150 6259.9 12.9 6065.8 12.5 2645.9 5.4 33729.6 69.3 48701.2
200 9281.7 14.4 5269.0 8.2 3057.0 4.7 46790.5 72.7 64398.2
300 26167.3 14.9 27452.7 15.6 8486.3 4.8 113780.8 64.7 175887.0
400 51198.6 21.5 34275.6 14.4 12649.7 5.3 140550.9 58.9 238674.8
500 86850.2 28.8 35063.2 11.6 15427.0 5.1 164122.1 54.4 301462.5
600 98822.3 27.1 43668.1 12.0 16245.7 4.5 205514.1 56.4 364250.2
700 92359.5 21.6 37921.3 8.9 12788.4 3.0 283968.7 66.5 427037.9
800 107676.0 22.0 40699.7 8.3 17557.0 3.6 323893.0 66.1 489825.6
900 110036.2 19.9 42029.5 7.6 21113.8 3.8 379433.9 68.7 552613.4
1000 120104.0 19.5 38977.3 6.3 25851.0 4.2 430468.7 69.9 615401.0
HB I 50 2017.8 11.6 4521.6 26.0 282.2 1.6 10588.8 60.8 17410.4
100 5532.7 16.7 3875.7 11.7 1157.0 3.5 22542.1 68.1 33107.5
150 4757.1 9.7 5108.1 10.5 1795.6 3.7 37143.7 76.1 48804.5
200 9233.9 14.3 6426.9 10.0 3480.2 5.4 45360.6 70.3 64501.6
300 23078.5 13.1 27773.4 15.8 8667.0 4.9 116575.4 66.2 176094.3
400 50276.2 21.0 29079.4 12.2 9536.1 4.0 149990.8 62.8 238882.6
500 75369.1 25.0 30891.9 10.2 17426.9 5.8 177982.9 59.0 301670.8
600 93720.0 25.7 40967.6 11.2 18093.1 5.0 211678.4 58.1 364459.1
700 91022.9 21.3 45470.7 10.6 13458.1 3.1 277295.6 64.9 427247.3
800 107236.4 21.9 41727.1 8.5 15370.1 3.1 325702.0 66.5 490035.5
900 121409.9 22.0 46060.3 8.3 22906.6 4.1 362447.0 65.6 552823.8
1000 107771.8 17.5 38776.0 6.3 27069.5 4.4 441994.7 71.8 615611.9
HB P 50 1128.4 6.5 2125.3 12.2 634.5 3.6 13579.5 77.7 17467.7
100 2683.1 8.1 6080.7 18.3 867.0 2.6 23534.0 71.0 33164.8
150 7806.1 16.0 5647.8 11.6 2605.7 5.3 32802.2 67.1 48861.8
200 7514.1 11.6 8046.2 12.5 3565.9 5.5 45432.6 70.4 64558.8
300 34065.6 19.3 22280.4 12.6 8310.7 4.7 111551.9 63.3 176208.6
400 51876.3 21.7 38471.0 16.1 11378.6 4.8 137270.7 57.4 238996.7
500 80665.2 26.7 37865.4 12.5 16458.9 5.5 166795.2 55.3 301784.7
600 93582.4 25.7 43172.2 11.8 17053.8 4.7 210764.4 57.8 364572.8
700 99084.5 23.2 37456.2 8.8 12431.6 2.9 278388.4 65.1 427360.8
800 105398.9 21.5 42105.2 8.6 15521.7 3.2 327123.2 66.7 490148.9
900 113614.4 20.5 44187.7 8.0 19583.4 3.5 375551.4 67.9 552937.0
1000 112124.5 18.2 35775.2 5.8 27936.9 4.5 439888.3 71.4 615724.9
JG C 50 17370.7 99.5 78.4 0.4 6.9 0.0 0.0 0.0 17456.0
100 29321.7 88.4 967.5 2.9 1380.8 4.2 1483.1 4.5 33153.1
150 37027.4 75.8 3577.3 7.3 1837.5 3.8 6408.0 13.1 48850.1
200 45260.0 70.1 4966.3 7.7 1973.1 3.1 12347.8 19.1 64547.2
300 115307.1 65.4 17467.2 9.9 11676.4 6.6 31734.6 18.0 176185.4
400 138387.9 57.9 33151.5 13.9 16212.7 6.8 51221.4 21.4 238973.6
500 167972.0 55.7 55509.9 18.4 16536.3 5.5 61743.5 20.5 301761.7
600 187905.3 51.5 65899.8 18.1 18923.1 5.2 91821.7 25.2 364549.9
700 220499.3 51.6 76619.6 17.9 20389.6 4.8 109829.5 25.7 427338.0
800 277544.9 56.6 79025.2 16.1 23219.8 4.7 110336.3 22.5 490126.2
900 341233.8 61.7 82849.0 15.0 25208.4 4.6 103623.3 18.7 552914.4
1000 366168.1 59.5 91090.3 14.8 31899.3 5.2 126544.7 20.6 615702.4
JG I 50 15893.1 94.3 616.5 3.7 114.0 0.7 237.3 1.4 16860.9
100 24388.7 74.9 1829.3 5.6 934.0 2.9 5405.9 16.6 32557.9
150 33969.1 70.4 3638.2 7.5 2239.0 4.6 8408.5 17.4 48254.9
200 46357.4 72.5 7274.0 11.4 2496.9 3.9 7823.6 12.2 63951.9
66
Site Treatment Distance Cover Cover Gain Gain Loss Loss NoCover No Cover Total Area
(m) (m) % m % m % m % m
300 114284.9 65.3 16639.3 9.5 8385.3 4.8 35685.1 20.4 174994.6
400 155784.3 65.5 23480.2 9.9 15245.6 6.4 43272.4 18.2 237782.5
500 174365.9 58.0 52401.8 17.4 14417.1 4.8 59385.6 19.8 300570.4
600 197345.7 54.3 64949.1 17.9 20312.5 5.6 80750.9 22.2 363358.3
700 215557.6 50.6 78901.8 18.5 24172.2 5.7 107514.5 25.2 426146.1
800 255691.1 52.3 85160.8 17.4 24101.3 4.9 123980.8 25.4 488934.0
900 343116.9 62.2 82718.7 15.0 24853.6 4.5 101032.7 18.3 551721.9
1000 356136.1 58.0 89013.4 14.5 28517.9 4.6 140842.2 22.9 614509.7
JG P 50 15204.0 85.1 179.3 1.0 1434.9 8.0 1042.2 5.8 17860.4
100 26100.1 77.8 3249.0 9.7 1420.5 4.2 2787.9 8.3 33557.4
150 31399.4 63.7 4664.3 9.5 1145.6 2.3 12045.0 24.5 49254.4
200 41318.8 63.6 5735.6 8.8 4912.6 7.6 12984.4 20.0 64951.4
300 103269.1 58.3 20471.5 11.6 9367.7 5.3 43885.3 24.8 176993.7
400 131441.7 54.8 41054.5 17.1 15808.0 6.6 51477.5 21.5 239781.6
500 173458.5 57.3 49939.1 16.5 18161.4 6.0 61010.7 20.2 302569.6
600 185566.3 50.8 53460.4 14.6 18406.6 5.0 107924.2 29.5 365357.5
700 226776.2 53.0 65586.5 15.3 22265.4 5.2 113517.4 26.5 428145.4
800 272379.4 55.5 84031.4 17.1 25642.2 5.2 108880.2 22.2 490933.3
900 327672.7 59.2 94957.0 17.1 23075.8 4.2 108015.8 19.5 553721.3
1000 358704.0 58.2 89612.0 14.5 32740.4 5.3 135452.7 22.0 616509.1
LL C 50 7849.4 45.1 512.3 2.9 412.1 2.4 8644.8 49.6 17418.7
100 22589.5 68.2 630.7 1.9 244.5 0.7 9650.9 29.1 33115.7
150 32198.1 66.0 778.1 1.6 694.0 1.4 15142.3 31.0 48812.6
200 41351.4 64.1 1673.6 2.6 892.1 1.4 20592.6 31.9 64509.6
300 86296.9 49.0 9284.3 5.3 5664.1 3.2 74864.6 42.5 176109.9
400 110575.2 46.3 12422.1 5.2 16389.0 6.9 99511.5 41.7 238897.7
500 124421.1 41.2 19519.1 6.5 15233.7 5.0 142511.6 47.2 301685.5
600 113732.0 31.2 18430.7 5.1 9608.2 2.6 222702.4 61.1 364473.3
700 161307.6 37.8 31265.3 7.3 10080.4 2.4 224607.8 52.6 427261.0
800 157685.6 32.2 37438.0 7.6 15661.9 3.2 279263.4 57.0 490048.9
900 97457.6 17.6 41868.8 7.6 17659.3 3.2 395851.0 71.6 552836.7
1000 103145.9 16.8 54736.4 8.9 28804.7 4.7 428937.3 69.7 615624.3
LL I 50 6337.3 34.0 337.0 1.8 254.0 1.4 11702.5 62.8 18630.8
100 16258.4 47.5 1232.3 3.6 806.4 2.4 15900.6 46.5 34197.6
150 26403.8 52.9 981.1 2.0 984.3 2.0 21502.1 43.1 49871.3
200 40219.1 61.3 2508.4 3.8 1278.7 2.0 21551.5 32.9 65557.7
300 104657.3 58.7 8470.3 4.8 13170.2 7.4 51894.5 29.1 178192.3
400 111264.0 46.2 8688.3 3.6 15418.9 6.4 105603.6 43.8 240974.7
500 111208.3 36.6 18163.7 6.0 10042.5 3.3 164347.4 54.1 303761.8
600 123536.5 33.7 19829.3 5.4 9254.1 2.5 213927.9 58.4 366547.8
700 194297.5 45.3 28561.3 6.7 8487.4 2.0 197988.9 46.1 429335.2
800 140749.3 28.6 38929.6 7.9 12921.6 2.6 299522.5 60.9 492123.0
900 88923.9 16.0 42649.0 7.7 20963.4 3.8 402374.6 72.5 554911.0
1000 118821.1 19.2 56954.4 9.2 31873.1 5.2 410050.2 66.4 617698.9
LL P 50 5209.1 28.9 431.1 2.4 295.0 1.6 12119.4 67.1 18054.6
100 16880.8 50.3 815.5 2.4 743.6 2.2 15139.4 45.1 33579.4
150 26746.6 54.3 1549.0 3.1 922.6 1.9 20045.3 40.7 49263.6
200 39596.0 61.0 2635.4 4.1 1734.8 2.7 20990.9 32.3 64957.2
300 86035.7 48.6 10351.4 5.8 12565.2 7.1 68050.3 38.4 177002.7
400 99291.2 41.4 7416.4 3.1 15002.5 6.3 118080.1 49.2 239790.2
500 120240.0 39.7 14517.1 4.8 11383.7 3.8 156436.4 51.7 302577.2
600 121631.8 33.3 23666.2 6.5 9318.3 2.6 210748.2 57.7 365364.5
700 190751.9 44.6 30149.8 7.0 8608.1 2.0 198642.2 46.4 428152.0
800 158258.4 32.2 45548.5 9.3 14281.5 2.9 272851.2 55.6 490939.6
900 96039.8 17.3 40967.7 7.4 24225.4 4.4 392494.5 70.9 553727.3
1000 116534.3 18.9 57389.6 9.3 28784.8 4.7 413806.0 67.1 616514.8
MM C 50 10330.4 62.4 6235.1 37.6 0.0 0.0 0.0 0.0 16565.6
100 26277.7 81.4 5984.9 18.6 0.0 0.0 0.0 0.0 32262.6
150 41806.6 87.2 6153.0 12.8 0.0 0.0 0.0 0.0 47959.6
200 61121.7 96.0 2534.9 4.0 0.0 0.0 0.0 0.0 63656.7
300 165614.9 95.0 7749.2 4.4 170.3 0.1 870.0 0.5 174404.3
400 212184.2 89.5 17594.0 7.4 1780.6 0.8 5633.7 2.4 237192.5
500 240592.5 80.2 33471.9 11.2 8284.5 2.8 17631.7 5.9 299980.6
600 285239.8 78.6 46671.8 12.9 2232.1 0.6 28624.9 7.9 362768.7
700 346441.1 81.4 32997.6 7.8 2987.5 0.7 43130.5 10.1 425556.8
800 374682.6 76.7 38900.2 8.0 11955.0 2.4 62807.1 12.9 488344.9
900 405745.2 73.6 43914.2 8.0 8843.1 1.6 92630.5 16.8 551133.0
67
Site Treatment Distance Cover Cover Gain Gain Loss Loss NoCover No Cover Total Area
(m) (m) % m % m % m % m
1000 422994.4 68.9 55629.0 9.1 19405.6 3.2 115892.0 18.9 613921.0
MM I 50 13400.4 78.6 3554.9 20.9 88.1 0.5 0.0 0.0 17043.5
100 29580.7 90.3 7.9 0.0 3151.8 9.6 0.0 0.0 32740.4
150 41369.7 85.4 1113.4 2.3 4230.7 8.7 1723.5 3.6 48437.4
200 49603.1 77.3 4725.5 7.4 1246.2 1.9 8559.5 13.3 64134.4
300 123842.2 70.6 17161.0 9.8 2224.5 1.3 32132.0 18.3 175359.6
400 144227.9 60.6 32218.2 13.5 8029.4 3.4 53672.0 22.5 238147.6
500 192500.7 64.0 32552.5 10.8 8069.5 2.7 67812.8 22.5 300935.5
600 233747.6 64.3 51573.7 14.2 8270.0 2.3 70132.1 19.3 363723.4
700 271401.7 63.6 72239.7 16.9 11969.0 2.8 70900.9 16.6 426511.2
800 292902.2 59.9 86621.9 17.7 11669.8 2.4 98105.2 20.1 489299.1
900 391814.0 71.0 74094.0 13.4 11042.9 2.0 75136.1 13.6 552087.0
1000 470948.8 76.6 49962.0 8.1 20080.1 3.3 73883.9 12.0 614874.8
MM P 50 13570.3 81.8 3014.3 18.2 0.0 0.0 0.0 0.0 16584.6
100 31989.1 99.1 0.0 0.0 292.5 0.9 0.0 0.0 32281.6
150 37798.2 78.8 2942.9 6.1 5581.5 11.6 1656.0 3.5 47978.6
200 48913.5 76.8 4654.8 7.3 2577.0 4.0 7530.2 11.8 63675.7
300 129425.1 74.2 20550.7 11.8 1652.6 0.9 22813.9 13.1 174442.3
400 160203.3 67.5 27758.6 11.7 5981.8 2.5 43286.8 18.2 237230.5
500 187550.4 62.5 35474.3 11.8 9445.6 3.1 67548.2 22.5 300018.6
600 235545.9 64.9 46804.7 12.9 7409.2 2.0 73046.9 20.1 362806.7
700 281696.3 66.2 66626.5 15.7 10578.2 2.5 66693.8 15.7 425594.7
800 309543.7 63.4 76186.0 15.6 10793.5 2.2 91859.7 18.8 488382.9
900 392844.3 71.3 73403.7 13.3 11392.1 2.1 73531.0 13.3 551171.0
1000 461394.7 75.2 53844.0 8.8 16100.9 2.6 82619.4 13.5 613959.0
OM C 50 1399.8 8.1 10133.9 58.4 767.4 4.4 5054.2 29.1 17355.3
100 7046.7 21.3 9409.2 28.5 1566.7 4.7 15029.7 45.5 33052.3
150 10321.7 21.2 11222.1 23.0 4290.6 8.8 22915.0 47.0 48749.4
200 16275.6 25.3 17917.7 27.8 4568.4 7.1 25684.7 39.9 64446.4
300 27892.3 15.8 36586.0 20.8 8768.8 5.0 102736.6 58.4 175983.8
400 32891.4 13.8 56978.0 23.9 18315.0 7.7 130587.6 54.7 238771.9
500 70124.5 23.3 56343.1 18.7 20393.7 6.8 154698.7 51.3 301560.0
600 88125.7 24.2 63765.1 17.5 31163.8 8.6 181293.5 49.8 364348.0
700 97073.9 22.7 83570.6 19.6 37974.1 8.9 208517.5 48.8 427136.1
800 76787.1 15.7 84101.9 17.2 44077.7 9.0 284957.5 58.2 489924.2
900 122728.6 22.2 105299.6 19.1 52068.9 9.4 272615.2 49.3 552712.3
1000 181884.1 29.6 100362.2 16.3 42264.9 6.9 290989.0 47.3 615500.2
OM I 50 1482.6 8.5 10449.8 60.1 211.4 1.2 5250.3 30.2 17394.0
100 6661.7 20.1 11331.9 34.2 3340.1 10.1 11757.3 35.5 33091.0
150 12880.0 26.4 12983.1 26.6 3773.5 7.7 19151.4 39.3 48788.0
200 12841.7 19.9 16137.9 25.0 2131.8 3.3 33373.6 51.8 64485.0
300 27930.4 15.9 39351.3 22.4 10822.6 6.1 97956.5 55.6 176060.8
400 26571.8 11.1 54992.0 23.0 19402.1 8.1 137882.9 57.7 238848.8
500 62725.5 20.8 61234.1 20.3 26111.3 8.7 151565.8 50.2 301636.7
600 76570.8 21.0 65904.2 18.1 27452.4 7.5 194497.3 53.4 364424.6
700 105756.8 24.8 84555.5 19.8 31728.0 7.4 205172.2 48.0 427212.5
800 95557.5 19.5 91269.6 18.6 41665.9 8.5 261507.4 53.4 490000.4
900 112296.2 20.3 95835.2 17.3 50790.2 9.2 293866.8 53.2 552788.4
1000 159125.5 25.8 110500.5 18.0 47212.7 7.7 298737.4 48.5 615576.1
OM P 50 3491.8 19.9 8992.6 51.1 879.8 5.0 4221.4 24.0 17585.6
100 8590.2 25.8 9585.6 28.8 1928.9 5.8 13177.9 39.6 33282.5
150 12214.0 24.9 12345.7 25.2 4059.2 8.3 20360.6 41.6 48979.4
200 17581.4 27.2 17035.2 26.3 3373.1 5.2 26686.8 41.3 64676.4
300 22140.0 12.5 43573.7 24.7 9297.3 5.3 101432.6 57.5 176443.6
400 30303.7 12.7 57970.6 24.2 17809.6 7.4 133147.5 55.7 239231.4
500 69755.2 23.1 57721.3 19.1 22692.6 7.5 151850.1 50.3 302019.2
600 81343.9 22.3 63781.5 17.5 26621.6 7.3 193060.0 52.9 364807.0
700 96885.9 22.7 86698.2 20.3 40466.0 9.5 203544.7 47.6 427594.7
800 94644.7 19.3 88617.5 18.1 39158.9 8.0 267961.4 54.6 490382.5
900 108896.3 19.7 99467.6 18.0 51969.3 9.4 292837.1 52.9 553170.3
1000 168274.4 27.3 107728.4 17.5 48272.9 7.8 291682.2 47.4 615958.0
RS C 50 8644.4 55.6 6040.8 38.8 99.0 0.6 765.5 4.9 15549.8
100 20257.3 64.8 2380.8 7.6 854.8 2.7 7753.8 24.8 31246.7
150 26660.7 56.8 5226.4 11.1 2627.7 5.6 12428.8 26.5 46943.6
200 30394.8 48.5 12152.9 19.4 6755.3 10.8 13337.5 21.3 62640.5
300 59836.1 34.7 31345.4 18.2 21404.7 12.4 59785.5 34.7 172371.7
400 76029.6 32.3 41466.9 17.6 14203.6 6.0 103459.3 44.0 235159.3
500 93073.7 31.2 44459.9 14.9 22875.1 7.7 137538.3 46.2 297947.0
68
Site Treatment Distance Cover Cover Gain Gain Loss Loss NoCover No Cover Total Area
(m) (m) % m % m % m % m
600 102783.3 28.5 55032.8 15.3 25010.5 6.9 177908.0 49.3 360734.6
700 103135.6 24.4 73261.0 17.3 45886.3 10.8 201239.3 47.5 423522.2
800 114950.8 23.6 80757.2 16.6 33217.9 6.8 257384.0 52.9 486309.9
900 109494.3 19.9 99991.7 18.2 42926.7 7.8 296684.9 54.0 549097.5
1000 87343.3 14.3 122186.7 20.0 33841.1 5.5 368513.9 60.2 611885.0
RS I 50 17959.7 62.8 8136.0 28.5 783.9 2.7 1707.4 6.0 28587.1
100 32118.4 61.8 11574.5 22.3 1762.3 3.4 6495.0 12.5 51950.2
150 26229.9 41.2 16934.5 26.6 5728.6 9.0 14731.3 23.2 63624.3
200 26696.1 33.9 15421.7 19.6 10316.3 13.1 26381.1 33.5 78815.1
300 69681.2 34.1 33235.8 16.3 15936.5 7.8 85420.2 41.8 204273.6
400 88123.2 33.0 33650.6 12.6 15990.0 6.0 129039.4 48.4 266803.2
500 103630.5 31.5 43768.4 13.3 24154.6 7.3 157918.1 47.9 329471.6
600 107338.1 27.4 65668.2 16.7 35962.8 9.2 183222.1 46.7 392191.1
700 98778.5 21.7 77175.5 17.0 36865.0 8.1 242130.2 53.2 454949.2
800 106931.2 20.7 91140.5 17.6 38450.7 7.4 281195.2 54.3 517717.6
900 103868.4 17.9 112978.8 19.5 41400.6 7.1 322239.4 55.5 580487.2
1000 90865.9 14.1 116837.3 18.2 31372.0 4.9 404186.8 62.8 643262.1
RS P 50 10478.6 61.2 5522.3 32.2 281.6 1.6 852.3 5.0 17134.8
100 17971.2 54.7 10198.9 31.1 1297.4 4.0 3364.3 10.2 32831.8
150 21856.7 45.0 10572.2 21.8 2906.5 6.0 13193.4 27.2 48528.7
200 26745.5 41.6 15374.0 23.9 5081.3 7.9 17025.0 26.5 64225.8
300 58576.5 33.4 26309.8 15.0 14744.5 8.4 75911.7 43.2 175542.5
400 68849.5 28.9 31832.7 13.4 14023.1 5.9 123625.2 51.9 238330.5
500 100992.2 33.5 29941.0 9.9 18053.2 6.0 152132.0 50.5 301118.5
600 126996.7 34.9 40527.2 11.1 28756.4 7.9 167626.2 46.1 363906.5
700 106480.6 25.0 67115.6 15.7 30841.4 7.2 222256.8 52.1 426694.4
800 96546.5 19.7 83513.1 17.1 40056.6 8.2 269366.2 55.0 489482.4
900 118664.0 21.5 102329.1 18.5 40837.0 7.4 290440.3 52.6 552270.4
1000 80484.9 13.1 98489.5 16.0 40434.8 6.6 395649.1 64.3 615058.2
SC C 50 5900.0 34.4 7342.1 42.8 1122.6 6.5 2805.7 16.3 17170.4
100 14138.9 43.0 7982.5 24.3 4776.7 14.5 5969.4 18.2 32867.5
150 18184.4 37.4 11645.2 24.0 9409.6 19.4 9325.4 19.2 48564.5
200 19772.7 30.8 14446.2 22.5 9801.6 15.3 20241.1 31.5 64261.6
300 61981.7 35.3 27563.0 15.7 21441.5 12.2 64627.9 36.8 175614.1
400 43797.4 18.4 23511.1 9.9 57236.3 24.0 113857.4 47.8 238402.2
500 43582.8 14.5 52576.5 17.5 49583.5 16.5 155447.5 51.6 301190.3
600 61964.4 17.0 75048.6 20.6 19384.9 5.3 207580.7 57.0 363978.5
700 95420.7 22.4 62654.0 14.7 28843.7 6.8 239848.1 56.2 426766.5
800 97914.3 20.0 73867.1 15.1 37675.3 7.7 280098.0 57.2 489554.7
900 103321.0 18.7 88607.6 16.0 36629.8 6.6 323784.5 58.6 552342.8
1000 97830.7 15.9 102077.3 16.6 47065.3 7.7 368157.5 59.9 615130.8
SC I 50 5691.8 33.3 4615.7 27.0 1663.6 9.7 5120.3 30.0 17091.4
100 15467.1 47.2 9316.4 28.4 3508.4 10.7 4496.5 13.7 32788.4
150 17137.8 35.3 9048.4 18.7 10635.7 21.9 11663.5 24.1 48485.4
200 23031.3 35.9 14606.0 22.8 8867.0 13.8 17678.1 27.5 64182.4
300 49741.6 28.3 24078.1 13.7 34752.8 19.8 66883.2 38.1 175455.8
400 48675.0 20.4 28427.6 11.9 55275.2 23.2 105866.0 44.4 238243.8
500 52864.1 17.6 41996.8 14.0 36308.3 12.1 169862.6 56.4 301031.8
600 52608.3 14.5 68876.2 18.9 22164.0 6.1 220171.4 60.5 363819.8
700 97256.9 22.8 72241.8 16.9 24350.0 5.7 232759.0 54.6 426607.8
800 106580.1 21.8 71180.2 14.5 38239.1 7.8 273396.4 55.9 489395.8
900 89900.3 16.3 85767.5 15.5 36128.7 6.5 340387.3 61.6 552183.8
1000 93374.5 15.2 93365.7 15.2 43706.3 7.1 384525.2 62.5 614971.7
SC P 50 6511.8 36.9 2525.1 14.3 4721.9 26.8 3880.8 22.0 17639.7
100 15458.4 46.4 6495.7 19.5 5516.8 16.5 5865.8 17.6 33336.7
150 21152.9 43.1 9968.9 20.3 4181.0 8.5 13730.9 28.0 49033.7
200 17959.9 27.7 12308.7 19.0 10502.1 16.2 23960.1 37.0 64730.7
300 41197.2 23.3 22570.2 12.8 47007.2 26.6 65778.0 37.3 176552.5
400 58239.2 24.3 33228.6 13.9 44829.2 18.7 103043.6 43.1 239340.6
500 51201.8 16.9 40378.4 13.4 30713.9 10.2 179834.6 59.5 302128.7
600 64904.8 17.8 56738.3 15.5 24959.3 6.8 218314.4 59.8 364916.8
700 92015.1 21.5 75252.7 17.6 25056.3 5.9 235380.7 55.0 427704.8
800 104542.1 21.3 69767.0 14.2 36712.1 7.5 279471.9 57.0 490493.0
900 88420.6 16.0 83641.0 15.1 34322.5 6.2 346896.9 62.7 553281.1
1000 95511.8 15.5 92525.9 15.0 42659.9 6.9 385371.4 62.6 616069.0
SG C 50 638.3 3.8 4814.2 28.6 136.3 0.8 11270.3 66.8 16859.1
100 1166.5 3.6 10571.5 32.5 392.9 1.2 20425.1 62.7 32556.0
150 3619.4 7.5 13122.8 27.2 674.4 1.4 30836.3 63.9 48252.9
69
Site Treatment Distance Cover Cover Gain Gain Loss Loss NoCover No Cover Total Area
(m) (m) % m % m % m % m
200 11446.0 17.9 18139.0 28.4 910.8 1.4 33454.1 52.3 63949.9
300 19784.1 11.3 35572.6 20.3 4025.4 2.3 115608.4 66.1 174990.5
400 18009.9 7.6 43243.0 18.2 5244.7 2.2 171280.6 72.0 237778.2
500 48463.2 16.1 66821.0 22.2 11705.4 3.9 173576.3 57.7 300565.9
600 62664.9 17.2 97929.1 27.0 11411.0 3.1 191348.6 52.7 363353.6
700 63592.6 14.9 94411.7 22.2 15960.8 3.7 252176.1 59.2 426141.2
800 80766.7 16.5 112913.4 23.1 21529.4 4.4 273719.4 56.0 488928.9
900 127937.1 23.2 124726.1 22.6 32041.1 5.8 267012.4 48.4 551716.7
1000 148547.3 24.2 125718.5 20.5 37107.7 6.0 303130.7 49.3 614504.2
SG I 50 1634.1 9.6 7699.3 45.1 625.5 3.7 7113.0 41.7 17072.0
100 1986.8 6.1 11059.5 33.8 334.2 1.0 19388.4 59.2 32768.9
150 1376.4 2.8 10414.3 21.5 205.9 0.4 36469.3 75.2 48465.8
200 4526.4 7.1 14408.4 22.5 1895.9 3.0 43332.1 67.5 64162.8
300 25223.8 14.4 44654.9 25.5 5021.1 2.9 100516.6 57.3 175416.3
400 20231.7 8.5 54424.1 22.8 3256.0 1.4 160292.3 67.3 238204.1
500 25612.3 8.5 79030.8 26.3 7356.0 2.4 188992.8 62.8 300991.9
600 58360.9 16.0 90306.2 24.8 12549.8 3.4 202562.8 55.7 363779.7
700 65273.1 15.3 91758.6 21.5 16833.6 3.9 252702.1 59.2 426567.4
800 99605.2 20.4 110687.1 22.6 29964.9 6.1 249097.9 50.9 489355.1
900 95697.7 17.3 125762.4 22.8 26846.7 4.9 303836.1 55.0 552142.9
1000 140444.0 22.8 130289.5 21.2 39733.2 6.5 304463.9 49.5 614930.6
SG P 50 0.0 0.0 2961.9 16.6 0.0 0.0 14879.2 83.4 17841.1
100 2556.7 7.6 10903.3 32.5 309.9 0.9 19768.1 58.9 33538.1
150 6590.4 13.4 14326.2 29.1 676.2 1.4 27642.3 56.1 49235.1
200 10048.0 15.5 14184.1 21.8 700.1 1.1 40000.0 61.6 64932.1
300 15812.2 8.9 36022.6 20.4 2278.3 1.3 122842.0 69.4 176955.1
400 33466.4 14.0 38076.6 15.9 9014.8 3.8 159185.4 66.4 239743.1
500 53995.5 17.8 62643.7 20.7 11867.9 3.9 174024.1 57.5 302531.1
600 59882.9 16.4 94546.2 25.9 12427.8 3.4 198462.2 54.3 365319.1
700 66665.4 15.6 105670.3 24.7 15810.6 3.7 239960.8 56.1 428107.1
800 86360.4 17.6 106573.0 21.7 21097.4 4.3 276864.2 56.4 490895.1
900 125849.4 22.7 120443.9 21.8 33097.0 6.0 274292.9 49.5 553683.2
1000 164785.3 26.7 132484.7 21.5 29115.1 4.7 290085.8 47.1 616470.9
SP C 50 187.5 1.1 2637.3 15.0 1029.6 5.8 13762.3 78.1 17616.8
100 10440.6 31.3 3281.7 9.9 7678.2 23.0 11913.3 35.8 33313.9
150 17999.7 36.7 3760.4 7.7 7958.7 16.2 19292.1 39.4 49010.9
200 30368.2 46.9 4497.1 6.9 5008.7 7.7 24834.0 38.4 64708.0
300 85621.4 48.5 14537.1 8.2 13299.2 7.5 63049.4 35.7 176507.0
400 124506.8 52.0 20641.1 8.6 23325.8 9.7 70821.6 29.6 239295.3
500 140356.2 46.5 22728.3 7.5 37661.4 12.5 101337.6 33.5 302083.5
600 148789.3 40.8 29402.6 8.1 45903.5 12.6 140776.3 38.6 364871.7
700 161558.9 43.5 21922.3 5.9 38535.8 10.4 149728.6 40.3 371745.6
800 150320.3 41.8 16135.2 4.5 27296.3 7.6 165576.0 46.1 359327.7
900 132265.4 36.5 19842.8 5.5 14075.2 3.9 195699.9 54.1 361883.3
1000 98180.1 40.6 8938.4 3.7 10700.1 4.4 124297.3 51.3 242115.8
SP I 50 7867.1 45.2 5679.1 32.7 2876.7 16.5 968.3 5.6 17391.1
100 18184.8 55.0 1462.5 4.4 4745.4 14.3 8695.4 26.3 33088.1
150 24605.5 50.4 2094.8 4.3 2492.2 5.1 19592.6 40.2 48785.1
200 33868.4 52.5 2774.9 4.3 3557.8 5.5 24281.0 37.7 64482.2
300 100724.3 57.2 11925.0 6.8 11105.2 6.3 52300.7 29.7 176055.2
400 127870.0 53.5 14111.3 5.9 28494.7 11.9 68367.2 28.6 238843.3
500 144313.4 47.8 18873.3 6.3 50523.5 16.8 87921.1 29.1 301631.3
600 126229.4 38.5 22541.4 6.9 49134.6 15.0 129546.6 39.6 327452.1
700 118347.7 37.3 30742.1 9.7 30859.7 9.7 137449.6 43.3 317399.1
800 119063.8 36.5 18080.2 5.5 17174.1 5.3 171840.9 52.7 326159.0
900 119883.3 35.6 12467.9 3.7 11998.9 3.6 192027.9 57.1 336378.0
1000 128858.6 37.0 23056.6 6.6 15716.6 4.5 180795.5 51.9 348427.2
SP P 50 3999.5 22.5 3543.8 19.9 2195.4 12.4 8032.5 45.2 17771.3
100 12555.7 37.5 2750.8 8.2 6494.6 19.4 11667.1 34.9 33468.3
150 24142.8 49.1 1632.8 3.3 4971.2 10.1 18418.5 37.5 49165.3
200 32146.9 49.6 2830.8 4.4 5671.2 8.7 24213.5 37.3 64862.3
300 94318.9 53.3 14509.0 8.2 10937.6 6.2 57050.1 32.3 176815.6
400 127010.6 53.0 16121.5 6.7 20289.4 8.5 76182.2 31.8 239603.7
500 143132.5 47.3 22372.8 7.4 53845.7 17.8 83040.8 27.5 302391.8
600 136852.2 38.8 31027.9 8.8 44882.1 12.7 139904.5 39.7 352666.7
700 137240.8 40.4 21616.2 6.4 32127.4 9.5 148570.8 43.8 339555.3
800 140390.6 41.0 16794.8 4.9 24233.4 7.1 160603.4 47.0 342022.2
900 121511.5 34.7 15845.7 4.5 12514.0 3.6 200312.2 57.2 350183.4
70
1000 109797.7 37.2 16917.3 5.7 10865.1 3.7 157585.1 53.4 295165.2
Abstract (if available)
Abstract
Tropical landscapes in Costa Rica have increasingly become targets of restoration efforts after deforestation depleted 90% of the region’s forests by the end of the 20th century. Research has shown that the environment surrounding a restoration site influences outcomes in fragmented landscapes, particularly as to the amount of forest cover surrounding restoration areas. However, the degree of influence that forest cover has on restoration sites and the long-term effects have historically been understudied due to the difficulty in assessing forest cover in remote regions through conventional field methods. As a result, there is a need for more time and cost-effective ways of evaluating and understanding forest cover change within the context of restoration efforts in remote areas. ❧ Geographic Information Systems (GIS) and remote sensing technologies have been utilized by researchers to understand better the relationships between abiotic and biotic factors in ecosystems. This study analyzed forest cover changes from 2005 to 2014 using high-resolution remote imagery to understand how forest cover changed surrounding 13 restoration sites near Las Cruces Biological Station (LCBS). The forest cover analysis revealed that the study region experienced a 9% net increase in forest cover over nine years. Similarly, all except one of the restoration sites had a net increase in forest cover within 200 meters. Topographic variables were extracted from a 5-meter DEM to understand their influence on the changes in forest cover. We hypothesized that elevation, slope, aspect, and distance to restoration site would have a strong and positive correlation with whether areas surrounding the restoration sites reforested from 2005 to 2014. A regression analysis revealed that topographic factors do not solely explain the variations in forest cover gain between sites
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Assessing woody plant encroachment in Marin County, California, 1952-2018
PDF
Classifying the alpines: developing a methodology to track environmental changes in the alpines utilizing remote sensing and ecozone vegetation patterns
PDF
Wetland mapping and restoration decision making using remote sensing and spatial analysis: a case study at the Kawainui Marsh
PDF
GeoBAT: crowdsourcing dynamic perception of safety data through the integration of mobile GIS and ecological momentary assessments
PDF
Spatial analysis of human activities and wildfires in the Willamette National Forest
PDF
Geographic object based image analysis for utility scale photovoltaic site suitability studies
PDF
Urban areas and avian diversity: using citizen collected data to explore green spaces
PDF
Improving wetland determination utilizing unmanned aerial systems
PDF
Spatio-temporal analysis of Western snowy plover nesting at Vandenberg Air Force Base
PDF
Using GIS to explore the tradeoffs in hydrographic survey planning: an investigation of sampling, interpolation, and local geomorphology
PDF
An exploration of the spatiotemporal distribution of snow crab (Chionoecetes opilio) in the eastern Bering Sea: 1982 – 2018
PDF
Exploring remote sensing and geographic information systems technologies to understand vegetation changes in response to land management practices at Finke Gorge National Park, Australia Between ...
PDF
Using GIS to perform a risk assessment for air-transmitted bioterrorism within San Diego County
PDF
Quantifying changes in monarch butterfly habitat in California
PDF
The role of precision in spatial narratives: using a modified discourse quality index to measure the quality of deliberative spatial data
PDF
Risk assessment to wildlife from Ohio on-shore wind farm development: a landscape model approach
PDF
Using GIS to predict human movement patterns in complex humanitarian emergencies: a test case of the Syrian Conflict
PDF
Beyond visual line of sight commercial unmanned aircraft operations: site suitability for landing zone locations
PDF
Filling in the gaps: 3D mapping Arizona's Basin and Range aquifer in the Prescott Active Management Area
PDF
Utilizing advanced spatial collection and monitoring technologies: surveying topographical datasets with unmanned aerial systems
Asset Metadata
Creator
Amar, Jorge L.
(author)
Core Title
Using aerial imagery to assess tropical forest cover surrounding restoration sites in Costa Rica
School
College of Letters, Arts and Sciences
Degree
Master of Science
Degree Program
Geographic Information Science and Technology
Publication Date
09/23/2020
Defense Date
08/26/2020
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
aerial imagery,Ecology,forest cover,OAI-PMH Harvest,Restoration,tropical
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Bernstein, Jennifer (
committee chair
), Loyola, Laura (
committee member
), Marx, Andrew (
committee member
)
Creator Email
jamar@usc.edu,jamjamar@gmail.com
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c89-378124
Unique identifier
UC11666086
Identifier
etd-AmarJorgeL-9015.pdf (filename),usctheses-c89-378124 (legacy record id)
Legacy Identifier
etd-AmarJorgeL-9015.pdf
Dmrecord
378124
Document Type
Thesis
Rights
Amar, Jorge L.
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
aerial imagery
forest cover