Download DEFORESTATION PATTERNS AND HUMMINGBIRD DIvERSITy IN

Institutionen för naturgeografi
Deforestation patterns and
hummingbird diversity in the
Amazon rainforest
Felicia Labor
Examensarbete grundnivå
Geografi, 15 hp
GG 191
2016
Förord
Denna uppsats utgör Felicia Labors examensarbete i Geografi på grundnivå vid Institutionen
för naturgeografi, Stockholms universitet. Examensarbetet omfattar 15 högskolepoäng (ca 10
veckors heltidsstudier).
Handledare har varit Simon Jakobsson, Institutionen för naturgeografi, Stockholms
universitet. Examinator för examensarbetet har varit Stefano Manzoni, Institutionen för
naturgeografi, Stockholms universitet.
Författaren är ensam ansvarig för uppsatsens innehåll.
Stockholm, den 8 februari 2017
Steffen Holzkämper
Chefstudierektor
Abstract
In recent decades expanding land-use change has caused extensive deforestation of the tropical rainforest
inducing large-scale transformation of the landscape patterns across the South American continent. Landscape
change is a modification process of the natural forest cover into fragments which generate various ecological
impacts. Habitat loss is identified to be a major threat to biodiversity, as it exposes species to the risk of
extinction. This study investigates 80 locations within tropical rainforest biomes to examine the landscape change
which has occurred from 1993 – 2014. The intention is to identify the impacts of landscape fragmentation on
hummingbird species diversity by spatial landscape analysis in GIS and regression modeling. The analysis found
that there is no relationship between deforestation and reduction of hummingbird diversity. The results indicate
that hummingbird species are not particularly sensitive to landscape change as they have high resilience in regard
to forest fragmentation. A potential threshold value of deforestation degree could be identified, up to which
hummingbird species richness increased, but locations subjected to over 40% fragmentation were estimated to
have lower hummingbird diversity. However, by using the spatial explicit biological data, the analysis indicate
that an extinction debt may exist in the landscape, and that future extinctions may be expected to occur in the
following decades as consequence of deforestation. Other factors may be as important determining variables for
species richness: the spatial scale of the study, the habitat connectivity, hummingbird generalist tendencies.
Conclusively, identification of the key factors of deforestation impacts on species diversity is essential for future
efficiency in conservation planning and sustainability of the tropical rainforest biodiversity.
Keywords: Landscape fragmentation, deforestation, extinction debt, hummingbirds, biodiversity
database, spatial landscape analysis, the Amazon rainforest, land-cover change.
Contents
1. Introduction.............................................................................................................................................................1
2. Background ..............................................................................................................................................................2
2.1
Impacts of deforestation and landscape fragmentation on species diversity ...........................2
2.2
Hummingbird taxa.................................................................................................................................4
3. Methodology ...........................................................................................................................................................6
3.1
Study area ...............................................................................................................................................6
3.2
Selection of data.....................................................................................................................................7
3.2.1 Hummingbird data................................................................................................................................7
3.2.2 Geographic data.....................................................................................................................................7
3.3
Data analysis ..........................................................................................................................................7
3.3.1
Forest data ..............................................................................................................................................7
3.3.2 Hummingbird data processing...........................................................................................................9
3.3.3 Hummingbird diversity in relation to forest cover ................................................................... 10
4. Results..................................................................................................................................................................... 11
5. Discussion .............................................................................................................................................................. 16
5.1
Hummingbird diversity in relation to deforestation ................................................................. 16
5.2
Method & data quality ...................................................................................................................... 19
5.3
Future work......................................................................................................................................... 20
6. Conclusion............................................................................................................................................................. 21
7. References ............................................................................................................................................................. 22
8. Appendix ............................................................................................................................................................... 27
1. Introduction
The Amazon territory extends across nine national borders located on the South American continent and
encompasses the largest tropical rainforest area in the world. The Amazon region consists primarily of tropical
and subtropical moist broadleaf rainforest biomes characterized by low seasonal variations, enclosing the
Amazon basin. Historical environmental aspects of the Amazon have led to large regional variations in forest
types and significant species diversity. The Amazon rainforest is estimated to include approximately 50.000
terrestrial vascular plant species (Hubbell et al., 2008). The Amazon rainforest is one of the most biodiverse
regions on Earth and holds a third of the world’s avifauna, containing 1.500 avian species and two endemic avian
families, one of them hummingbirds (Trochilidae) (BirdLife International, 2015).
During recent decades expanding rainforest deforestation caused by anthropogenic exploitation has generated
a large-scale geographical transformation process across the Amazon region. The deforestation process is
causing an extensive forest canopy disturbance, resulting in 20% loss of the original rainforest canopy (Abdala,
2015). Tropical deforestation mainly occurs because of expanding infrastructure, mining, logging, pastoral
farming and extensive agriculture (Charity et al., 2016). Deforestation generates global scale impacts as it
accounts for approximately 20% of annual greenhouse gas emissions into the atmosphere (Flato et al., 2013). On
a regional scale, deforestation changes the landscape structure and create a mosaic landscape pattern containing
rainforest patches surrounded by a matrix of altered landscapes.
Research in landscape ecology agrees that forest disturbances have vast forms of ecological effects and generate
stresses to the ecosystem dynamics of the rainforest region (Peres et al., 2010). Forest-cover disturbance and
landscape fragmentation affects the ecosystem functions and generate ecological consequences in terms of
habitat deterioration, decreased biomass and alteration of the nutrient cycles (Haddad et al., 2015). Landscape
fragmentation means that continuous natural vegetation cover becomes reduced into smaller fragments, a
process which induces that a species' habitat is degraded or lost due to isolation, edge effects or because the
habitat is not large enough (Mitchell et al., 2015). Habitat degradation means that the habitat remains, but it no
longer obtains sufficient quality for a population of a species to ensure viability within it (Fischer & Lindenmayer,
2007). Forest fragmentation and habitat degradation are considered as some of the major threats to biodiversity
worldwide as they generate population decline and expose species to the risk of extinction (Brook et al., 2008).
Species’ response to landscape change occurs with a certain time-delay, which indicates that the extinction of
species takes place gradually. The time-delay of species response to landscape alteration is known as extinction
debt, meaning that species could be expected to become extinct in the future as a cause of current or historical
habitat loss (Kuussaari et al., 2009). However, the time-delay of extinction allows the opportunity for a
conservation window indicating that protection and recovery of the continuous habitat can be implemented
(Wearn et al., 2012). Additionally, the time-delay can function as a potential source of error in extinction
forecasts (Halley et al., 2014).
Habitat loss has a varied impact on different taxonomic groups in the Amazon rainforest as they react on forest
disturbance on different time scales. The relaxation time i.e. the response time to a disturbance in a system is
for birds estimated to be one decade (Halley et al., 2016). Other factors that make it difficult to identify the
relaxation time is that species in the Amazon rainforest have irregular distribution patterns and varied habitat
range, which means that the relaxation time between different species varies on both spatial and temporal scale
(Brooks et al., 1999).
Previous studies have successfully used spatial biological data from the Global Biodiversity Information Facility
to examine spatial species distribution and biodiversity decline on a large-scale (e.g. Beck et al., 2014; Feeley &
Silman, 2011; Newbold et al., 2014). As research in landscape ecology continues, it is acknowledged that
understanding the parameters which determine a species relaxation time and how fast extinction of species
occurs is considered as key for protection of the tropical rainforest biodiversity (Halley & Iwasa, 2011).
1
Progress in the understanding of rainforest disturbance and landscape fragmentation, in terms of avifaunal
species richness, could refine conservation techniques and hence reverse the vulnerability of the rainforest
biodiversity to prevent endangered species from becoming extinct. Furthermore, this progress is urgent to avoid
incorrect estimates of species diversity in rainforest habitats and to gain better insight into the time scale of
species extinction.
The main objective of this study is, therefore, to conduct a spatial analysis of landscape change due to
anthropogenic influence through deforestation of the tropical rainforest biome. The aim is to investigate
correlations between land-cover change and forest fragmentation in relation to present-day species richness of
hummingbirds.
By using freely available geographical and biological data a spatial landscape analysis was performed in a
Geographical Information System. The species richness was investigated in 80 study locations within tropical
rainforest biomes and further analyzed in relation to the degree of deforestation, comparing past and present
forest cover in the landscapes. The intention was to detect the impact extent of deforestation on hummingbird
diversity for future efficiency in conservation planning and sustainability of the tropical avifauna. The aim of the
study is to answer two research questions:
(1) How does species richness of hummingbirds differ between locations of closed-canopy rainforest and
rainforest locations exposed to deforestation and landscape fragmentation?
(2) What is the potential of using spatially explicit biological data in large-scale analysis of species
richness in relation to changing landscape patterns?
2. Background
2.1 Impacts of deforestation and landscape fragmentation on species diversity
Landscape ecological studies of species richness in modified landscapes commonly focus on how the changing
landscape patterns in tropical forest impacts species richness. Studies of this relationship indicates a general
diversity decline in relation to increasing habitat disturbance along with fluctuating response to deforestation
and landscape fragmentation between various taxonomic groups (Gardner et al., 2009). Extinction risk of species
on a regional scale and the taxonomic variation depend on a number of factors such as the size of the population
and number of offspring because it determines the potential for proliferation for the species to further maintain
viability (O'Grady et al., 2004).
General spatial biodiversity patterns are commonly identified by the species-area relationship (SAR) which
indicate that the number of observed species will increase with the size of the sample area (O'Dwyer & Green,
2010). The ecosystem resilience i.e. the capacity of a system to recover after being subjected to disturbance or
perturbation and yet sustain its essential functions is an important factor in terms of landscape alteration (Walker
et al., 2004). Some of the main features which are identified to regulate the impact degree of deforestation in
terms of resilience are the size and shape of the remaining landscape patches (Antongiovanni & Metzger, 2005).
Habitat connectivity is identified as another key factor in terms of species resilience to landscape fragmentation
and refers to the degree of forest patch isolation and to what extent movement between significant habitat
patches is possible (Awade & Metzger, 2008). In accordance, one study demonstrates that habitat connectivity
is the main cause generating a negative impact on nectarivorous birds, rather than the size of a remaining forest
patch (Martensen et al., 2008).
However, evidence suggests that the size of the remaining forest patches play a major part in extinction rates as
larger forest patches are found to include higher species diversity than smaller patches (Marini, 2001). An
investigation extending over a period of 20 years demonstrated that the extinction rates were reduced in
conjunction with increased fragment size (Stouffer et al., 2009).
Furthermore, the investigation indicated that more than 30% of the containing species became extinct within
habitats of approximately 1 hectare, in comparison to approximately 5% in forest patches of about 100 hectares.
2
Several studies have identified niche breadth i.e. the range of resources utilized by a species as a determining
factor in terms of ecological resilience (Swihart et al., 2006), and thus also decide a species geographical range
(Slatyer et al., 2013). Geographically separated populations of the same species often occupy different niches in
their respective communities. Species that occupy a limited niche breadth are called specialists and species which
occupy a wide niche are in turn called generalists (Boulangeat et al., 2012). Generalist species have a tendency
to successfully adapt to environments modified by land-use change and establish in new habitats (Ducatez et al.,
2015). Specialists are considered to experience a higher risk of extinction than generalist species, as they are
more susceptible to landscape modification and habitat loss (Büchi & Vuilleumier, 2014).
Furthermore, physiological traits of a species define its dispersal ability, referring to the capability of individuals
of a species to move between different sites and habitats and is considered as an important aspect in relation to
if a species could occupy a forest fragment at all (Lees & Peres, 2009). The probability that a species dispersal
ability is enough to cross habitat boundaries is determined by the distance, the time course of the movement
and the landscape structure of the moving path (Fahrig, 2007).
Dispersal ability is documented to vary among avian species and is therefore viewed as a vital factor in regard to
resilience to landscape fragmentation and habitat loss (Moore et al., 2008). Additionally, research suggests that
isolated forest fragments should primarily contain avian species with high dispersal abilities or species with a
high tolerance towards a deforested matrix (Lees & Peres, 2008). Another study showed similar evidence of the
generalist taxa being relatively unaffected by deforestation and fragmentation degree and found indications that
the diversity of generalist species instead were found to increase in areas exposed to deforestation (Pardini et
al., 2010).
How landscape fragmentation and the lack of habitat connectivity affect avian species in tropical rainforests are
also determined by the species dependency on a certain forest ecosystem, species which relies on the dense
closed-canopy forest are identified to be more vulnerable to fragmentation (dos Anjos et al., 2011). Reduction
of species richness has been found to have a strong correlation with landscape changes, as a natural variation of
species richness occurs dependent on the quantity of biomass and vegetation type (Schulze et al., 2004).
Additionally, avian species richness is documented to decline according to a landscape disturbance gradient, with
higher species diversity in the mature closed-canopy forest in comparison to secondary forest, with a further
decline in relation to forest conversion intensity in plantations, cropland and pasture (Philpott et al., 2008). The
level of endemic species abundance has also been demonstrated to decline according to the same forest
conversion gradient (Waltert et al., 2011).
However, another study suggests that pasture dominated landscapes are more efficient than landscapes
dominated by crops in terms of conservation, hence the fragmentation patterns vary among these in such way
that pastures comprise larger fragments than cropland; thus not affecting species richness as negatively
(Carvalho et al., 2009).
Landscape fragmentation generates smaller forest areas which also limit the habitat range and lead to higher
competition for resources along with isolation. The process of landscape fragmentation could lead to
overexploitation of resources and reduction of breeding and nesting opportunities that could strongly impact
the viability of a species and the ecosystem dynamics over time (Saunders et al., 1991).
Compilation of the results from 37 previous studies of avian diversity in relation to habitat disturbance in tropical
rainforests has showed that approximately an equal number of cases demonstrated that species diversity
increased or decreased in relation to degree of landscape modification, and also found that research conducted
in large-scale sampling areas had a tendency to have a higher avian diversity, and small-scale spatial studies in
areas less than 25 hectares generated much lower diversity (Hill & Hamer, 2004).
However, several studies conclude significant evidence for this correlation pattern; that the determining factor
in studies of biodiversity in relation to habitat disturbance is the spatial scale at which the studies are performed,
thus generating fluctuating results of the species richness (Dumbrell et al., 2008). Another study suggests that
extinction rates increase tenfold, according to a decrease of the fragmented area by 1000 (Ferraz et al., 2003).
3
Other studies have examined forest fragmentation in terms of threshold values of deforestation rate and found
indications that habitats in areas deforested between 60 – 70 % to host a significantly lower number of species
than regions with a larger proportion of continuous closed-canopy forest (Ochoa‐Quintero et al., 2015).
Many deforestation studies address the aspect of deforestation and forest fragmentation in terms of edge effect
as a negative consequence that hence could influence the dynamics of forest microclimate, carbon storage and
tree mortality (Laurance et al., 2011). Most species of understory birds in the Amazon rainforest are identified
to be edge-averse and are generally negatively affected by the land-use change in terms of forest fragmentation
(Powell et al., 2015). Indications suggest that the impact degree of edge effect on the avifauna depend on the
size of the remaining forest patches because it determines the edge quantity, along with the species dependency
on the closed-canopy primary forest (Lees & Peres, 2006). However, some studies suggest that forest edges may
also favor some avian species, especially nectarivorous species (Rodewald & Brittingham, 2004). According to
other studies related to forest edge impacts on avian species, evidence suggests that hummingbirds are not as
susceptible to forest fragmentation as other avian families and remain rather unaffected by landscape
fragmentation (Stouffer et al., 2006; Kennedy et al., 2010).
The reason that hummingbirds are resilient to landscape change is estimated to be that they have a habitat
preference for forest ecotones and therefore tend to have a high abundance at forest edges (Banks-Leite et al.,
2010), presumably due to increased light at the forest edges favoring flowering and thus the access of nectar.
However, forest edges also tend to have a higher abundance of predators, generating an increased risk for bird
species at ecotone habitats to be subjected to predation (Gillies & St Clair, 2010).
Regional hummingbird abundance has been identified to have a strong correlation with spatial fluctuations of
resources and flowering patterns along with temperature variation (Pansarin & Pedro, 2016). Vegetation
dynamics and changing the structure in rainforest ecosystems are closely linked to climatic feedback in terms of
fluxes of energy and moisture (Bonan et al., 2003). Deforestation generates changing photosynthesis rates,
alteration of the carbon cycle and vapor exchange to the atmosphere causing a large-scale impact on a global
scale (Law et al., 2002). Modification of the atmospheric moisture could result in reduced precipitation and
fluctuating humidity, which in turn could affect the seasonality and climate on a large-scale (Bala et al., 2007).
Hydrological fluctuations in the Amazon basin will control the rainfall pattern, and thus impacts the occurrence
of leafing, flowering, and fruiting and hence control the timing and spatial distribution of food resources for avian
species (Beja et al., 2010). Decreased humidity and rainfall could, in turn, generate prolonged droughts increasing
the risk of forest fires, a form of disturbance not included in rainforest ecosystems resilience (Laurance &
Williamson, 2001).
2.2
Hummingbird taxa
The Hummingbird (Trochilidae) family is the second largest avian family on Earth. Hummingbirds origin from the
lowlands of South America and historical analysis show that hummingbird species have invaded the Andean
mountain region and other primary landmasses more than 30 times creating complex distribution patterns, that
lead to the geographical species distribution of today (McGuire et al., 2007).
The avian family includes approximately 361 species of hummingbirds, whereof 274 species are found in South
America. Hummingbird species richness is documented to decrease with altitudinal zonation and fluctuating
vegetation, hence reduce in relation to decreased abundance of terrestrial herbs, shrubs, and vines at higher
elevation (Buzato et al., 2000). However, this general pattern doesn’t account for the remarkable species richness
of the montane region of the Andes, where the elevation instead has generated varied topographical features
which have promoted high speciation rates (Rahbek et al., 2007).
Hummingbirds have developed a specific plant-mutualism during co-evolution for 22 million years and are
specialist nectarivorous species providing an essential process in the functioning of the ecosystems as pollinators
(Dalsgaard et al., 2011). However, research has identified hummingbird pollination patterns which indicate that
they also could be considered as generalists since they have a tendency to pollinate non-ornithophilous species
of flowers (Rodrigues & Araujo, 2011).
4
Another study examined hummingbird foraging of 62 plant species of primary epiphytes but also included some
shrubs, lianas, and trees and found that hummingbirds foraged 95% of the investigated species, but only
pollinated about 68% (Rocca-de-Andrade, 2006). Approximately 7.000 species of flowers belonging to 65 plant
families are estimated to depend on pollination from one or several species of hummingbirds (Abrahamczyk &
Renner, 2015).
Hummingbird habitat selection occurs according to foraging determined by the preference of floral
morphological traits, identified to be plant species with diurnal anthesis, red colored inclined flowers and tubular
shape (Rodrigues & Rodrigues, 2014). Another factor determining hummingbird foraging is possible nectar
accessibility without landing as they are highly aerial birds with adapted wing stroke mechanism, keeping them
flight hovering constantly (Bleiweiss, 2009).
Possibly the most important hummingbird habitat preferences are large volumes of nectar available, as they are
known to have some of the highest metabolic rates among vertebrates and ingest nectar which almost only
contains pure sugars (Chen & Welch, 2014). Hummingbirds are territorial birds and exhibit aggressive defense
patterns of the resources within their selected habitat, which are estimated to be about 1.6 hectares
(Abrahamczyk & Kessler, 2010).
According to The IUCN Red List of Threatened Species™, 56 species of hummingbirds are threatened to the risk
of extinction in the next century, among those species ten are critically endangered indicating an estimated 50%
risk of those species to become extinct within ten years and two species already becoming so (IUCN, 2016),
highlighting the urgent need for operative conservation efforts.
5
3. Methodology
3.1 Study area
Figure 1. Distribution of the selected study sites in the Amazon rainforest region.
The study region is located on the northern half of the South American continent and includes the Amazon
rainforest and encircling locations within terrestrial biomes of evergreen tropical and subtropical moist broadleaf
forest. The biomes of the study region are characterized by humid tropical climate and annual temperatures
ranging from 21 - 30°C with an average temperature of 25°C (Christopherson & Birkeland, 2015). The Amazon
drainage basin has considerable precipitation levels ranging from 180 - 400 cm annually, with a wet climate all
year (Costa & Foley, 1998). The carbon absorption capacity of the Amazon rainforest vegetation is estimated to
be approximately 1.5 billion tons annually (Lewis et al., 2011).
The Amazon region’s landscape development has characterized the ecological evolution of the rainforest which
is influenced by tectonic uplift and the formation of the Andes, which has formed large scale elevation differences
ranging from lowland <500 m to the high mountain peaks that rise to almost 7000 m above sea level. The forest
types of the Amazon are partly generated by denudation processes of the Andean Mountains, hence producing
95% of the sediment and nutrient deposit throughout the region (Volkoff et al., 2012). The sediment deposit
generates different soil characteristics which have generated a forest type gradient across Amazon basin of
primary and secondary forest types (McClain & Naiman, 2008).
The Amazon lowlands have been covered by closed-canopy moist forest during at least the last two glacial and
interglacial cycles, which have enabled the development of the region’s vast biodiversity (Colinvaux et al., 2000).
6
The Amazon rainforest region is considered to comprise one of the most valuable geographic systems on the
planet, as it is regarded as a biodiversity hotspot with a high level of endemism (Myers et al., 2000) and possesses
vital ecosystem services on a local and global scale (Foley et al., 2007). The Amazon rainforest thereby has an
inherent high conservation value combined with high deforestation rates and extensive landscape fragmentation
making it a relevant study region in terms of identification of extinction debt.
3.2
Selection of data
3.2.1 Hummingbird data
The taxonomic delimitation of the study constitutes of Neotropical hummingbird species from the Trochilidae
family. Birds are the most well documented taxonomic group in the Amazon region, and hummingbirds are
considered to be one of the most recorded avian groups in the Neotropics, with few new species discovered
since 1950’s (Rahbek & Graves, 2000). Therefore, hummingbirds are relevant to select because there is a large
amount of data to accurately use in the analysis of species richness according to the method used in this study.
The record of hummingbird species distribution was collected using The Global Biodiversity Information Facility
(GBIF) database. All geographical hummingbird data used in this study were collected from January 2006 –
September 2016 and based on human observations, machine observations or preserved specimens. The
hummingbird data is geo-referenced by coordinates within each of the selected study sites and compromises
70202 observations of totally 203 different species of hummingbirds.
3.2.2 Geographic data
The geographical boundary of the study was retrieved from a dataset consisting of Terrestrial Ecoregions of the
World (TEOW) classified into 14 biomes. The data that were selected from the ecoregions were reclassified into
a new layer that consisted of terrestrial biomes of evergreen tropical and subtropical moist broadleaf forest in
South America. The TEOW data was obtained from The Nature Conservancy (Olson et al., 2001).
The land cover information datasets used in this study consist of tree canopy cover from year 2000 and global
forest cover loss from between 2000 – 2014. The land cover dataset is originally derived from the time-series
analysis of 654,178 Landsat 7 ETM+ Landsat 8 LDCM images, characterizing global forest extent and forest cover
change from 2000 through 2014 presented in 500 meters spatial resolution (Hansen et al., 2015). The historical
land cover data that was used in the analysis were from 1992 – 1993 presented with 1 km spatial resolution
(Hansen et al., 2003).
3.3
Data analysis
3.3.1
Forest data
The spatial landscape analysis and data processing were executed in the Geographical Information System
(ArcMap 10.4.1” ESRI 2016) and Microsoft Excel 2013. The basis for the study consists of 80 study sites selected
within terrestrial biomes of evergreen tropical and subtropical moist broadleaf forest in South America. Primarily
the 80 study sites were selected according to a great geographical range to accurately represent the tropical
rainforest biomes. All of the sampled study areas consisted of circular landscape patches with 20 km in diameter,
with an area of 314.16 km2 (31416 hectares).
The potential study areas were selected to overlap with regions of available hummingbird distribution data. The
selected areas were delimited by circular polygons converted to a new layer, this was done with the raster
geoprocessing tool, a function applied to select a new delimited data extent only containing the data within each
study sites.
The spatial landscapes analysis was performed by remote sensing vegetation analysis of two geographic datasets
of forest cover from the year 2000 and 1993 along with forest cover loss from 2000 – 2014.
7
The forest cover data from the year 2000 was designated by tree canopy density, defined as canopy closure for
all vegetation taller than 5m in height and further reclassified into landscape areas of forest > 50% density and
forest with < 50% density within each study location. The dataset of forest cover loss 2000 – 2014 was classified
into regions of forest loss, referring to locations subjected to stand-replacement disturbance, or a change from
a forest to a non-forest state, and locations unexposed to forest loss classified into no forest loss. Furthermore,
to identify the total quantity of deforestation within each study area The Raster Calculator tool was applied to
get new Map Algebra values.
Figure 2. The data processing steps of the landscape analysis of the deforestation degree within the
study locations. A: Forest density <50% from 2000 (marked in white). B: Forest loss occurrence from
2000 – 2014 (marked in red). C: The union from the A and B datasets combined.
The dataset variables representing < 50% forest density from the 2000 forest cover raster (A) was combined with
the forest loss values from the forest loss 2000 – 2014 dataset raster (B) to create the union of the two datasets
(C). This was done to achieve the present forest cover of 2014 and detect the total quantity of deforestation that
has occurred until then within each study area. The union (C) was acquired by examining each pixel of the raster
datasets (A) and (B) individually and calculating the result through the or-operation with the above-described
conditions, as seen in equation 1 (Figure 2).
𝐴 ∪ 𝐵=𝐶
(Eq. 1)
Furthermore, the union data (C) was examined by applying the Zonal Histogram function in ArcGIS. This function
creates a table of the frequency distribution of cell values from a raster. The zones used to define the raster areas
were polygons encompassing the 80 circular study locations.
The Zonal histogram function calculated the pixel quantity of continuous closed-canopy forest (marked in black)
and the pixel quantity of deforested land (marked in red) within each study location. Finally, the percentage of
the land surface containing closed-canopy forest and deforested areas were then calculated.
By calculation of pixel quantity representing the fragmented landscape in relation to the continuous closedcanopy forest, the study areas were divided into two categories. The study areas were divided in half, and 40 of
the sample areas were categorized as primarily dominated by continuous closed-canopy forest cover where less
than <13 % deforestation has occurred since the year 2000, representing a low deforestation degree. The other
half of the sample areas were designated to represent landscape patches exposed to a higher deforestation
degree with more than >13 % of the total 314.16 km2 being deforested since the year 2000.
The land cover data from 1992 – 1993 was presented in 13 land cover classes consisting of different forest types,
grassland, shrubland, cropland, and bare ground. The land cover dataset from 1993 included another
classification than the land cover dataset from 2000. This classification difference in the datasets could lead to
potential errors when comparing the data. The raster data from 1993 was included in the study nevertheless
since it was the most accurate historical land cover data available to examine land use change. To eliminate the
above-described problem the 1993 raster was reclassified to better fit the requested land cover features.
8
The first class included forest types and closed shrub values representing forested areas, and open land-cover
types and bare ground in the second class representing deforested areas.
The reclassification enables comparable results from the raster data values from 1993, 2000 and 2014. The Zonal
Histogram function was further applied to quantify the pixel of the two categories of forested and deforested
areas.
An error of spatial factors encountered in the analysis, depending on the current data quality, was the spatial
sampling resolution. The geographical raster data of the land-cover of 1993 has 1 km spatial resolution, and the
raster data of forest cover from 2000 and forest loss from 2014 has 500 m spatial resolution. The resolution issue
resulted in that some of the study locations obtained negative change in deforestation (Δ deforestation)
percentage values, which would then indicate that extensive reforestation occurred in many locations, which can
be assumed to have emerged due to differences in resolution and data values.
To eliminate this source of error of intractable data, the study locations which obtained negative percentage
values of Δ deforestation, therefore, were excluded from the regression analysis of deforestation change from
1993 – 2014.
3.3.2
Hummingbird data processing
The georeferenced hummingbird data were analyzed within the 80 study areas, providing data of the total
quantity of hummingbirds individuals recorded in each location along with the number of species.
Standardization of the data of hummingbird species richness was implemented to eliminate a potential source
of error that comes from the data quantity available for each study location.
The logarithmical regression approach could be used because a logarithmic relation between species richness
and data quantity is expected. As the data quantity grows the possibility of finding more species should grow as
well. However, the more species that are found the more acquired data is necessary to find additional species
until the maximum possible species are found. This means that the approximation function should have a high
derivative for low x-values, a low derivative for high x-values and eventually converge when x goes to infinity.
These are all true for a logarithmic function. Without standardizing the data, the study sites with a high quantity
of data acquisition would be biased and consequently show greater species richness when analyzing the
deforestation and fragmentation impact on species richness.
The data was standardized in two different ways: by using the natural logarithm as standard and using the
function achieved by the logarithmic regression as described in equation 2 and 3, respectively.
𝑦
𝑦𝑠1 =
ln 𝑥
(Eq. 2)
where y stands for species richness, x is the data quantity and y s1 is the standardized species richness.
𝑦𝑠2 =
𝑦+𝑐
−𝑘
(Eq. 3)
𝑦 = 𝑘 ln 𝑥 − 𝑐
(Eq. 4)
ln 𝑥
where y stands for species richness, x is the data quantity, ys2 is the standardized species richness and c and k are
the variables describing the approximation function achieved from the logarithmic regression (Eq. 4) between
species richness and the number of individuals.
The variable k is subtracted from the standard value y s2 to offset and center the standard around the
approximation. Consequently, study sites with lower species richness than approximated are presented as
negative numbers, study sites with greater species richness as positive values and study sites with species
richness equal to the approximation as zero values. Additionally, the magnitude of the new standardized value
shows the deviation from the approximation.
9
By using the standardized value rather than the raw data value to describe each study site’s value of species
richness, the logarithmic relation between richness and data quantity is accounted for.
Study sites where the number of species found was greater than the logarithmic approximation function will
yield a higher standardized value than the study sites where the number of species was lower than approximated.
Consequently, a study site might yield a higher value of species richness than another even though the actual
amount of species would actually be lesser.
The logarithmic function achieved with the regression analysis of species richness including all the 80 study
locations in relation to available data quantity, is used in the standardization where equation 3 is used, instead
of the natural logarithm. The advantage of using the approximation achieved through the logarithmic regression
is that each study point yields a value of richness based on the deviation towards the approximation function
instead of the deviation towards the natural logarithm. The approximation achieved was used to standardize the
species richness with equation 3 of the study locations.
3.3.3 Hummingbird diversity in relation to forest cover
The resulting data from the spatial analysis of species richness in relation to deforestation degree were further
examined by statistical regression analysis. Primarily the raw data of the hummingbird species richness
observations calculated from the 80 study locations containing either a high degree of closed-canopy rainforest
or high deforestation degree, was examined in three different regression models, in relation to available
hummingbird data quantity retrieved from The Global Biodiversity Information Facility database.
Furthermore, another regression analysis was conducted of the standardized hummingbird species richness
achieved from equation 2, in relation to the current amount of closed-canopy forest and deforestation in year
2014 to investigate a potential correlation between species richness in relation to the degree of deforestation.
Additionally, the resulting standardized species richness calculated by using equation 3 was further used to
conduct another regression analysis of hummingbird species richness in relation to the percentage of the land
surface subjected to deforestation change from 1993 – 2014 (Δ deforestation), and thus shows the increase in
deforestation during this time, regardless of the amount of deforestation that has occurred at each location
before 1993. A polynomial regression model has used because it had the highest coefficient of determination.
Finally, a qualitative analysis was conducted investigating the specific landscape conditions of the four locations
that obtained the highest hummingbird species richness and their surrounding matrix, according to the
standardized data achieved by equation 2. This was done to enable a closer analysis of the specific landscape
conditions in each location according to the total deforestation that has occurred until 2014, including land
deforested before 1993 as well, to investigate other features that may determine the species richness of
hummingbirds.
10
4. Results
Species richness of the study areas with low deforestation degree
60
Species richness (raw data)
50
40
30
20
y = 6,8211ln(x) - 12,056
R² = 0,8131
10
0
0
500
1000
1500
2000
Hummingbird data quantity
2500
3000
3500
Figure 3. The regression model shows the calculated raw data species richness on the y-axis and the
available hummingbird data quantity (GBIF) on the x-axis. Each of the points represents data collected
from one study area. The regression demonstrates a logarithmic correlation between species richness
and the available data quantity of areas with low forest deforestation degree.
According to the regression analysis, there is a non-linear relationship between the variables of species richness
and data quantity. The approximation achieved from the regression demonstrates that it exists a positive
logarithmic correlation between the species richness and data quantity at study sites of low deforestation degree
(Figure 3). The regression shows that the coefficient of determination is high and that approximately 81 % of the
hummingbird species richness variation could be explained by the data quantity available.
11
Species richness of the study areas with high deforestation degree
40
Species richness (raw data)
35
30
25
20
15
y = 5,88ln(x) - 8,2524
R² = 0,6691
10
5
0
0
200
400
600
800
1000
1200
Hummingbird data quantity
1400
1600
1800
Figure 4. The regression model shows the raw data of hummingbird species richness presented on the
y-axis and the available hummingbird data quantity (GBIF) on the x-axis. Data from each study location
represent one point in the model. The model demonstrates a logarithmic correlation between species
richness and the available data quantity of areas with high deforestation degree.
The second regression model shows that it also exists a positive logarithmic correlation between hummingbird
species richness in each location, in relation to the hummingbird data quantity available (Figure 4). The R2 value
achieved in the regression model indicates that 66 % of the variance in the calculated amount of species in each
location is explained by the available data quantity. This shows that it is a slightly lower correlation to the
available amount of hummingbird data in study locations of higher deforestation degree in relation to locations
with lower deforestation degree.
The following regression analysis shows the result of the first two regression models combined, as it includes the
data of all study locations in the same model (Figure 5). The result shows that there is a logarithmic correlation
between hummingbird species richness and quantity of available hummingbird data. The coefficient of
determination indicates that the variance in a number of hummingbird species in each study site could be
explained by the data quantity available by approximately 76 %.
12
Species richness of all study areas in relation to data quantity
60
Species richness (raw data)
50
40
30
y = 6,5155ln(x) - 10,79
R² = 0,7682
20
10
0
0
500
1000
1500
2000
2500
Hummingbird data quantity
3000
3500
4000
Figure 5. The regression model shows raw data of the hummingbird species richness presented on the
y-axis correlated to the available hummingbird data quantity (GBIF) on the x-axis. The data from each
location is presented as one point in the regression. The model demonstrates a logarithmic correlation
between species richness and the available data quantity of all study areas included in the study.
Species richness of the study areas
correlated to degree of deforestation
8
y = -0,487x + 3,9491
R² = 0,0044
Species richness ys1
7
6
5
4
3
2
1
0
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
Deforestation 2014 (%)
Figure 6. The regression model demonstrates the correlation between the standardized species richness
(ys1), presented on the y-axis in relation to the degree of deforestation in all study areas on the x-axis.
The data from each study location represents one point in the regression model.
Following regression model demonstrates the calculation of the standardized species richness y s1, achieved by
applying the equation 2, presented on the y-axis in relation to forest deforestation degree on the x-axis. Each
point in the model shows the relation between percentage forest loss and standardized species richness.
13
The approximation achieved from the regression indicates that there is no correlation between species richness
and degree of deforestation. The R2 value suggests that the coefficient of determination a possible correlation is
0.44 %.
Δ Deforestation 10 - 70% since 1993 - 2014
3
y = -7,6944x2 + 6,4814x - 1,0322
R² = 0,0563
2,5
Species richness ys2
2
1,5
1
0,5
0
10%
-0,5
20%
30%
40%
50%
60%
70%
-1
-1,5
-2
Δ Deforestation between 1993 - 2014 (%)
Figure 7. The polynomial regression model shows the standardized hummingbird species richness (ys2)
on the y-axis. In relation to the deforestation change between 10 – 70% which has occurred from 1993 –
2014 as Δ deforestation on the x-axis. Each point in the model represents data from one study location.
The regression analysis of deforestation from 1993 – 2014 includes only the study locations with an increased Δ
deforestation value between 10 – 70 %, the increased percentage of the deforestation that took place within
those years. Because some of the study locations obtained negative percentage values of Δ deforestation, they
were excluded (listed in appendix I) (Figure 7). The result of the regression model demonstrates that the
hummingbird species richness of the locations included, at first increases slightly between deforestation degrees
of 10 – 40%, and then decreases when reaching deforestation degrees above 40%. This regression thus
demonstrates that the species diversity of hummingbirds at first appear to increase as a result of deforestation,
and then decrease at a threshold of 40% deforestation within a location.
14
Figure 8. Location of the four circular study areas with the highest standardized species richness (ys2)
and the surrounding landscape matrix, classified into three land-cover categories as closed-canopy forest
(marked in black), deforestation before 2000 (marked in white) and deforestation after 2000 (marked in
red). Satellite data (Hansen et al., 2013).
The four circular landscape areas are assigned with letters from A – D, where A represents the area with highest
standardized hummingbird diversity, with a decreasing species richness towards D. The result of qualitative
analysis demonstrates that the four locations obtaining the highest species richness in the study are located
within two countries, Peru and Colombia (Figure 8). The surrounding landscape matrix 2014 of the four areas,
further indicate that the landscape structure encircling location A – C consist of a large proportion of closedcanopy rainforest. Location A had a low deforestation degrees within the study area of 2.42%. Location B just
slightly higher deforestation degree within the study area of 4.28%. Location C had a deforestation degree of
3.05%. Showing that the tree locations of highest hummingbird species richness have not been exposed to high
deforestation rates. Location D, on the other hand, has quite a high deforestation degree of 28.73 % within the
study area.
15
5. Discussion
5.1 Hummingbird diversity in relation to deforestation
Interestingly, the overall result from the landscape analysis shows evidence that species richness does not have
any correlation with the degree of deforestation and forest fragmentation in the landscape. In contrast, the result
of this study indicates that it generally does not seem to affect the species diversity of hummingbirds much at
all. Accordingly, it could be considered as good news, as it suggests that hummingbirds have a high resilience to
the ongoing deforestation process in the Amazon rainforest region. Earlier studies suggest that bird species
relaxation times often extend over a 10 year period (Halley et al., 2016). However, this study examines the
deforestation process that has occurred since 1993, a time period of approximately 22 years. In relation to this,
it could be expected to find hummingbird species decline in areas exposed to deforestation. Nevertheless, little
evidence of hummingbird extinctions from the impact of forest fragmentation has been obtained in this research.
First of all, the two initial regression models demonstrate that there exists a strong correlation between the
species quantity found in each study location in regard to how many observations of hummingbirds that were
made in the first place (the GBIF database). The regression models demonstrate that approximately 81 % and 66
% of the species richness variation could be explained by the data quantity for the low and high deforested areas,
respectively. However, the data could also be interpreted to indicate that it is generally higher species occurrence
in locations exposed to low fragmentation degrees. This conclusion is derived from the difference of maximum
species found in a location exposed to high or low fragmentation respectively. The first model representing
locations of low deforestation shows that the maximum number of species found were 55 in relation to 2073
data points. However, the maximum species richness in locations of high deforestation degree were 35 species,
even though they were found in locations with 547 data points. This could thus be interpreted as that locations
of high fragmentation degree contain more species. It may also depend on that it is probably easier to locate and
observe hummingbirds in the less dense rainforest.
Altogether, 203 hummingbird species were included in the study, out of the approximately 274 hummingbird
species known to be found in South America. This high availability of data is noteworthy as it indicates that the
potential of using biological data from the GBIF database is great, and could be considered reliable since a lot of
species is represented in the data. On the other hand, what can be determined from the two regression models
is the fact that there is evidently a higher abundance of hummingbirds in terms of quantity in locations with
lower deforestation degree, in relation to study areas subjected to high deforestation degree. The maximum
number of hummingbird observed within one of the highly deforested study location were 1654 individuals, in
comparison with a maximum 3429 hummingbirds observed in a location of low deforestation degree. This
variation thus leads to that there is more than twice as much data available in study locations with low
deforestation degree. If the significant differences of available raw data, in terms of both quantity and number
of species found, in reality, depend on whether one has been searching for hummingbirds at different extent at
locations of high fragmentation degree, or if it simply were not as many species to find there, remains uncertain.
Although, this spatial species distribution pattern could be expected since forest fragmentation induces
reduction of the forest patch sizes and hence also geographical habitat range, in turn suggesting that the number
of observed species would be lower in regions exposed to landscape fragmentation according to the speciesarea relationship (O'Dwyer & Green, 2010).
However, to obtain a clearer picture of the variables determining hummingbird diversity, a closer qualitative
analysis of the four areas which obtained the highest species richness was conducted. The result provides
information about the geographical landscape conditions of the four locations along with the surrounding matrix.
What could be stated from the analysis of the species-richest study locations is that they, to a great extent, are
surrounded by a landscape largely compromised by closed-canopy rainforest. According to the nature of the
encircling landscape, this would suggest that the habitat amount and connectivity is especially high in the first
three areas (A – C, Figure 8) with the highest species richness. This result indicates that the character of the
landscape structure surrounding a forest habitat is important for the number of hummingbird species occupying
it. This analysis would cohere in accordance with previously identified patterns of habitat connectivity as a key
factor in terms of species richness of nectarivorous birds (Martensen et al., 2008).
16
Despite these indications, approximately a third of the study area of location D has been exposed to
deforestation along with a relatively large proportion of its matrix. Looking at the subsequent study areas with
high hummingbird species richness (Appendix I), they have equivalent high deforestation degrees ranging
between approximately 20 – 40 %.
The regression model of the Δ deforestation from 1993 – 2014 demonstrates that increasing fragmentation rates
with >10 % change initially seems to favor hummingbird species richness, in relation to habitats subjected to >40
% fragmentation change, which in turn indicates to have a negative impact as the hummingbird species diversity
declines thereafter. This indication, if still vague, argues for the existence of a threshold value of deforestation
degree in terms of a potential optimum fragmentation rate. If more data would exist representing higher Δ
deforestation degrees, the regression would hypothetically decline further along with higher deforestation rates.
These interpretations indicate that it appears that hummingbirds possibly benefit from a certain degree of
fragmentation, but at a threshold value of especially high deforestation degree, fragmentation would instead
have a negative impact.
The threshold value estimated in this study correlates with the qualitative analysis of the study areas with highest
species richness, and should tentatively be around 40%. Since many of the locations with high species richness
have been exposed to approximately 20 – 40 % deforestation, although seemingly resilient with continuingly
high species richness, strengthens the reliability of the data showing that hummingbirds seem to benefit of
deforestation at first, but that species diversity declines after a certain threshold value (40%). Previous studies
have also highlighted deforestation thresholds values as an interesting parameter and managed to show that
species richness declines in the area first when they become exposed to a deforestation degrees of 60 - 70%
(Ochoa‐Quintero et al., 2015). In the context of the suggestion by Ochoa-Quintero et al. (2015) the threshold
found in this study represents deforestations values in terms of changing deforestation degree from 1993 – 2014.
This means that the degree of exposure in each location could be higher, as deforestation may have occurred
before 1993, which opens the possibility that the threshold of 40% also could be greater.
Most avian species found in the rainforest is generally known to be negatively affected by deforestation,
regardless the degree, in terms of abundance and mostly evade forest edges (Powell et al., 2015).However,
studies of avian diversity have identified nectarivorous species to be significantly less susceptible to landscape
change in relation to other taxa (Stouffer et al., 2006; Kennedy et al., 2010).One explanation for this is that
hummingbirds are found to be favored by forest fragmentation, as they have a habitat preference for forest
ecotones (Banks-Leite et al., 2010). In accordance with the result of this study, previous research has found
evidence indicating that hummingbird, in fact, is favored by forest fragmentation (Rodewald & Brittingham,
2004).
The specific characteristics of hummingbirds seem to have significant importance for their low susceptibility to
modified landscapes. Initially, hummingbirds are considered to be specialist nectarivorous and plays an
important part for ecosystem functionalities as pollinators (Dalsgaard et al., 2011). However, the apparent
variation of hummingbird abundance between locations of high versus low fragmentation degree may be caused
by the degree of specialization among hummingbird species. Study findings suggest that some hummingbird
species have developed generalist tendencies over time (Rodrigues & Araujo, 2011), and are identified to forage
a wide variety of plant species divergent from their general flower preferences (Rocca-de-Andrade, 2006;
Rodrigues & Rodrigues, 2014). Extending generalist behavior of hummingbirds would, in turn, mean that
hummingbirds are developing a greater adaptability for landscape change, as they are less limited to a narrow
range of resources. This would also mean that the potential geographical range of hummingbirds would thus also
be enlarged (Slatyer et al., 2013).
Land-use change is identified to enable extensive benefits for generalist species, as they are less susceptible to
habitat modification (Büchi & Vuilleumier, 2014) and has high resilience in habitats subjected to disturbances
(Ducatez et al., 2015) and also identified to proliferate in habitats subjected to deforestation (Pardini et al., 2010).
These factors regarding generalist adaptabilities suggest that the niche breadth may be an essential variable
which has led to hummingbird capability to successfully occupy fragmented forest habitats.
17
Earlier studies have also identified niche breadth as a vital factor in regard to ecological resilience (Swihart et al.,
2006) and hummingbirds are known to effectively recognize floristic patterns and detect resource fluctuation
(Cotton, 2007) as they could take advantage of in adaptation to habitats subjected to landscape change.
Forest fragmentation is a process which is found to reduce the habitat size and lead to higher competition for
resources within the ecosystem (Saunders et al., 1991) a process earlier observed to generate behavioral change
in terms of foraging as a hummingbird species natural niche is forced to change or extend in order to satisfy the
high demand of nectar for hummingbird motoric functionalities (Chen & Welch, 2014). Presumably, as it takes
large quantities of nectar to keep a hummingbird in hovering flight to travel far distances, it would be estimated
that their dispersal ability is rather limited depending on the matrix in between forest habitats. As for this, habitat
connectivity would be closely linked to the dependency of the vegetation type comprising the matrix. A natural
diversity decline has been identified to exist in relation to change of the vegetation structure (Schulze et al.,
2004), as it becomes more homogeneous and plant species decreases, consequently, it simultaneously occurs a
decline of other species that depend on them. This pattern is significantly distinct in comparison to mature
closed-canopy forest and landscapes under anthropogenic influence.
However, apart from the landscape structure, the size of the continuous forest patches is vital. As it determines
the extent of forest edges and could be affecting the habitat connectivity as well. Several studies suggest that
extinction rates decrease in accordance to the size of a forest patch (Marini, 2001; Stouffer et al., 2009), inspiring
new research idea as described in the future work section.
Land-use change and altered vegetation type thus affect species dependent on a specific ecosystem and is proved
to locally influence species with narrow niche breadth the most. Furthermore, the vegetation features are of
importance in consideration to large-scale impacts of deforestation as well. Deforestation worldwide is known
to cause approximately 20% of the annual greenhouse gas emissions (Flato et al, 2013), at the same time the
Amazon rainforest alone is estimated to absorb 1.5 tons carbon annually (Lewis et al., 2011). The Amazon,
therefore, holds an important function as a carbon sink, which hence could reverse stresses of climate change.
Reduction of biomass generates further complex impacts on ecosystems. Atmospheric fluxes of energy,
moisture, and carbon are closely linked to the vegetation dynamics (Bonan et al., 2003); alteration of the biomass
hence affect the dynamics of the atmospheric fluxes. In a long-term perspective, changing landscape patterns
may result in extending large-scale impact from deforestation if the precipitation levels and humidity fluctuate
in the region as it can modify the seasonality of the region (Bala et al., 2007). Seasonal variation and precipitation
change will, in turn, affect the spatial distribution of food resources for hummingbirds as the distribution of
flowering and nectar depends on hydrological patterns (Beja et al., 2010). Regional hummingbird abundance is
found to be closely linked to fluctuating resources, access to nectar and temperature variations (Pansarin &
Pedro, 2016). In relations to this spatial correlation, it could be determined that the dynamics of hummingbird
diversity will be affected by fluctuations of temperature and humidity in the long run, although indirectly, as a
consequence of deforestation as it influences large-scale climate change. Increased climate change, in turn, leads
to extensive environmental impact worldwide.
18
5.2
Method & data quality
Figure 9. Showing the raw data of hummingbird distribution available from the GBIF database based
on individuals observed, collected from January 2006 – September 2016.
One of the main complications that arose in the choice of method was data quality issues, caused by the
landscape condition of the areas examined together with high data occurrence in selective parts of the Amazon
region. To investigate the impact of deforestation degree in terms of species richness in a sufficient way, equally
available data would be needed spanning over areas within all deforestation degrees. Consequently, the
statistical problem that occurs in this study is that 33 study areas compromises of quite low deforestation degrees
<10 % and 33 study areas with mediate deforestation degree of approximately 10 – 28% and only 14 of the study
locations are exposed to >30 % deforestation. Therefore, it is probably difficult to find statistical correlations,
since the various amount of data is available for different sample categories and generates data biased to
locations of low deforestation degree. Additionally, the approximation function achieved will be over-fitted for
low deforestation degree and contain great variance for high deforestation degrees.
The delimitation of the study areas at 13 % deforestation, with 40 areas that have either higher or lower
deforestation degrees than 13%, is questionable. The division was implemented because 13% deforestation is
the median value of the study areas. This may cause misinterpretation of the correlation between species
richness and data quantity as areas with closely adjacent deforestation degrees are considered separately (Figure
3 and 4). However, two separate regression analysis were conducted with the study areas delimited at 30%
deforestation instead. The first regression was created with 66 of the study areas exposed to <30% deforestation
and the second with 13 of the study areas exposed to >30% deforestation. The regressions showed a very similar
result as the 13% delimitation regressions, therefore the problem with a 13% delimitation seemed less
significant. Mainly because all study areas were analyzed together in all the other regression analysis in this study.
Nevertheless, despite the uneven distribution of available data the selection of study locations has been
determined according to a wide geographical range, in order to get a representative selection of the Amazon
region. The distribution of hummingbird data argues for the selected study boundary, to include encircling
tropical rainforest outside the Amazon borders (Figure 9).
19
The large variety of available data from different regions in the Amazon ought to have various technical and
territorial explanations. Primarily it may be explained by different land use factors and land-owning rights, which
in turn affects the accessibility of an area. The map of data distribution indicates that a high amount of data is
located on the west coast of the Amazon and quite an absence of data in the central Amazon, in contrast,
provides evidence that research is carried out selectively in concentrated areas. The geographical difference of
data may be due to topographical factors that are generating impassable terrain. The existence of adjacent
infrastructure also affects the availability of a location that enables the possibility for research. Regions where
forestry or mining is actively conducted or where the land is used for farming or ranching will unlikely have a
scientific priority or permitted access to unauthorized scientists. The data distribution could also occur due to
cultural aspects of land ownership in respect of indigenous peoples.
The Global Biodiversity Information Facility (GBIF) is an international open data infrastructure containing data
published by scientific institutions worldwide. The use of secondary data requires a fundamental trust to
academics contributing with scientific biological data to the GBIF. Since the biological data sampled in the studies
may have been conducted in various ways and with different premises that could cause deficiencies in relation
to whether the data represents the real biological distribution patterns or not.
Selection bias could easily occur due to the fact that the quantity of recorded hummingbirds and the geographical
distribution is naturally inconsistent throughout the Amazon (Buzato et al., 2000), with the Andean mountain
range being identified as one of the most diverse avifaunal regions on earth (Rahbek et al., 2007). This spatial
pattern of hummingbird occurrence is significantly visible in consideration to the data availability within regions
of terrestrial biomes of evergreen tropical and subtropical moist broadleaf forest.
Finally, the spatial scale of this study was selected with the intention to enable generalization of large-scale
deforestation impacts on hummingbird species richness across the tropical rainforest region, therefore each of
the study locations consists of 31416 hectares of land. Earlier investigations of deforestation impacts on
biodiversity have concluded that the spatial scale of the study determine how species richness responds to
landscape changes in an area (Dumbrell et al., 2008; Ferraz et al., 2003). Commonly small-scale studies
investigating areas less than 25 hectares conclude that the species richness decline in relation to high
deforestation rates (Hill & Hamer, 2004). In regard to the scale of this earlier research, this would suggest that
the selected spatial scale of the sample areas examined in this study may be, simply stated, far too large to detect
any spatial patterns of species richness decline. Furthermore, the extent of a hummingbird’s habitat range is
estimated to be approximately 1.6 hectares (Abrahamczyk & Kessler, 2010) in relation to this, study areas of
314116 hectares include many neighboring habitats and spatial distribution patterns would, therefore, be
complex to detect.
5.3
Future work
Primarily an interesting aspect in future research would be to conduct an examination and select study areas of
varying size and see if this will lead to different results, as many studies have found study area size to be highly
determining for the resulting species richness in a landscape (Dumbrell et al., 2008; Ferraz et al., 2003; Hill &
Hamer, 2004). The extent of forest edges is suggested to be one of the main factors in terms of impact on diversity
from deforestation along with the size and shape of the remaining forest patch (Marini, 2001; Stouffer et al.,
2009). In relation to this, a suggestion for future work would be to implement edge detection methods. Edge
detection compromises techniques to process images and highlight the contrast and brightness to clarify the
features in the image, and further apply algorithms hence detect the discontinuities in the image – representing
edges. By implementation of a forest edge detection method in this way, it would enable to quantify the exact
scope of forest edges in a landscape. Additionally, this would allow calculating the sizes of remaining forest
patches in a landscape subjected to forest fragmentation, to further analyze the impact of these parameters in
regard to species richness.
20
This study could conclude that the surrounding matrix is important as it reduces or decreases the habitat
connectivity. However, several earlier studies suggest that it exists a diversity and endemism gradient dependent
on vegetation types (Philpott et al. 2008; Waltert et al., 2011) along with an identified difference between
pastures and cropland, because pastures comprises larger fragments and hence does not have equally negative
effect on diversity (Carvalho et al., 2009). In this context, to map the specific vegetation characteristics of
landscape fragmentation patterns and matrices surrounding forest patches to detect if such difference is to be
found would further refine the knowledge of habitat fragmentation impacts. Identification of vegetation types
could be done by remote sensing analysis using vegetation indices, methods used to detect vegetation growth
and density hence indicating the vegetation type. A multivariate analysis of the hummingbirds’ species richness
using the degree of forest edges and vegetation types, together with other parameters that could affect the
hummingbird species richness, would be a suitable and interesting approach for future work.
Some of the result provided in this study could thus efficiently become more reliable as the data availability
expands, and this could eliminate a potential source of error from the result caused by over-fitted data and
inaccuracy of the biodiversity patterns of reality. This highlight the potential for future research conducted using
biological data, as more scientists and institutions contribute to databases like the GBIF.
6. Conclusion
The overall finding of this large-scale spatial analysis is that the species richness of hummingbirds does not differ
between locations of closed-canopy rainforest and locations subjected to forest fragmentation, to any particular
extent. No correlation could be found to exist between hummingbird diversity and deforestation degree, hence
the study has managed to conclude that hummingbird species are not specifically sensitive to landscape change.
The study has identified a threshold value in terms of fragmentation degree at 40%, suggesting that deforestation
thus favors hummingbird species richness as first, as they benefit from forest edges. However, habitats exposed
to >40% deforestation thus seem to have a negative impact on hummingbird species richness instead. The
explanation of hummingbirds’ high resilience in terms of habitat disturbance and landscape fragmentation
caused by deforestation is partly their expanding niche breadth as fluctuating foraging behavior has been
identified. Hummingbirds tend to exploit the resources available in the environment, inducing a high adaptability
to new habitats. The locations where the highest species richness was found also had high habitat connectivity,
to a large extent surrounded by a matrix of the closed-canopy rainforest. Aforementioned factors are identified
to be determining drivers in terms of hummingbird resilience of habitat disturbance. Deforestation cause
atmospheric influence which in turn generate climatic fluctuations that could affect the precipitation, humidity,
and seasonality in the rainforest. This impact can lead to changing spatial patterns of flowers and nectar and thus
also affect the distribution of hummingbird species in the long run. However, this study may also indicate that
an extinction debt may have been identified in the landscape. Hummingbirds’ relaxation time commonly
comprise one decade simultaneously severe habitat disturbance has occurred in the last two decades, hence
future extinctions may be expected to occur in the following decades as a consequence of the ongoing
deforestation process in the Amazon region. Additionally, using spatially explicit biological data from GBIF in
large-scale analysis of species richness has great potential in relation to the reliability in accordance with the high
representation of species in the data.
21
7. References
Abdala, G. C. (2015) The Brazilian Amazon: challenges facing an effective policy to curb deforestation 1.st EDITION
Brasilia, Brazil WWF Living Amazon Initiative and WWF Brazil 2015
Abrahamczyk, S., & Kessler, M. (2010). Hummingbird diversity, food niche characters, and assemblage
composition along a latitudinal precipitation gradient in the Bolivian lowlands. Journal of Ornithology, 151(3),
615-625.
Antongiovanni, M., & Metzger, J. P. (2005). Influence of matrix habitats on the occurrence of insectivorous bird
species in Amazonian forest fragments. Biological Conservation, 122(3), 441-451.
Awade, M., & Metzger, J. P. (2008). Using gap‐crossing capacity to evaluate functional connectivity of two Atlantic
rainforest birds and their response to fragmentation. Austral Ecology, 33(7), 863-871.
Bala, G., Caldeira, K., Wickett, M., Phillips, T. J., Lobell, D. B., Delire, C., & Mirin, A. (2007). Combined climate and
carbon-cycle effects of large-scale deforestation. Proceedings of the National Academy of Sciences, 104(16),
6550-6555.
Banks‐Leite, C., Ewers, R. M., & Metzger, J. P. (2010). Edge effects as the principal cause of area effects on birds
in fragmented secondary forest. Oikos, 119(6), 918-926.
Beck, J., Böller, M., Erhardt, A., & Schwanghart, W. (2014). Spatial bias in the GBIF database and its effect on
modeling species' geographic distributions. Ecological Informatics, 19, 10-15.
Beja, P., Santos, C. D., Santana, J., Pereira, M. J., Marques, J. T., Queiroz, H. L., & Palmeirim, J. M. (2010). Seasonal
patterns of spatial variation in understory bird assemblages across a mosaic of flooded and unflooded Amazonian
forests. Biodiversity and Conservation, 19(1), 129-152.
BirdLife International (2015) Bird species distribution maps of the world. BirdLife International, Cambridge, UK
and NatureServe, Arlington, USA.
Bleiweiss, R. (2009). The tail end of hummingbird evolution: parallel flight system development in living and
ancient birds. Biological Journal of the Linnean Society, 97(3), 467-493.
Bonan, G. B., Levis, S., Sitch, S., Vertenstein, M., & Oleson, K. W. (2003). A dynamic global vegetation model for
use with climate models: concepts and description of simulated vegetation dynamics. Global Change
Biology, 9(11), 1543-1566.
Boulangeat, I., Lavergne, S., Van Es, J., Garraud, L., & Thuiller, W. (2012). Niche breadth, rarity and ecological
characteristics within a regional flora spanning large environmental gradients. Journal of Biogeography, 39(1),
204-214.
Brook, B. W., Sodhi, N. S., & Bradshaw, C. J. (2008). Synergies among extinction drivers under global change.
Trends in ecology & evolution, 23(8), 453-460.
Brooks, T. M., Pimm, S. L., & Oyugi, J. O. (1999). Time lag between deforestation and bird extinction in tropical
forest fragments. Conservation Biology, 13(5), 1140-1150.
Buzato, S., Sazima, M., & Sazima, I. (2000). Hummingbird‐Pollinated Floras at Three Atlantic Forest Sites1.
Biotropica, 32(4b), 824-841.
Büchi, L., & Vuilleumier, S. (2014). Coexistence of specialist and generalist species is shaped by dispersal and
environmental factors. The American Naturalist, 183(5), 612-624.
Carvalho, F. M., De Marco, P., & Ferreira, L. G. (2009). The Cerrado into-pieces: Habitat fragmentation as a
function of landscape use in the savannas of central Brazil. Biological Conservation, 142(7), 1392-1403.
22
Charity, S., Dudley, N., Oliveira, D. and Stolton. S. (2016). Living Amazon Report 2016: A regional approach to
conservation in the Amazon. WWF Living Amazon Initiative, Brasília and Quito.
Chen, C. C. W., & Welch, K. C. (2014). Hummingbirds can fuel expensive hovering flight completely with either
exogenous glucose or fructose. Functional ecology, 28(3), 589-600.
Christopherson, W. R., & Birkeland, G. (2015). Geosystems: An Introduction to Physical Geography, 9/E. Pearson
Education Limited.
Colinvaux, P. A., De Oliveira, P. E., & Bush, M. B. (2000). Amazonian and Neotropical plant communities on glacial
time-scales: the failure of the aridity and refuge hypotheses. Quaternary Science Reviews, 19(1), 141-169.
Costa, M. H., & Foley, J. A. (1998). A comparison of precipitation datasets for the Amazon basin. Geophysical
Research Letters, 25(2), 155-158.
Dalsgaard, B., Magård, E., Fjeldså, J., González, A. M. M., Rahbek, C., Olesen, J. M., ... & Lara, C. (2011).
Specialization in plant-hummingbird networks is associated with species richness, contemporary precipitation
and quaternary climate-change velocity. PLoS One, 6(10), e25891.
DeFries, R. S., M. Hansen, J. R. G. Townshend and R. Sohlberg. 2003. LBA Regional Land Cover from AVHRR, 8km, 1984 (DeFries et al.). Available on-line [http://www.daac.ornl.gov] from the ORNL Distributed Active Archive
Center, Oak Ridge National Laboratory, Oak Ridge, Tennessee, U.S.A. doi:10.3334/ORNLDAAC/680
dos Anjos, L., Collins, C. D., Holt, R. D., Volpato, G. H., Mendonça, L. B., Lopes, E. V., ... & Carvalho, J. (2011). Bird
species abundance–occupancy patterns and sensitivity to forest fragmentation: Implications for conservation in
the Brazilian Atlantic forest. Biological Conservation, 144(9), 2213-2222.
Ducatez, S., Clavel, J., & Lefebvre, L. (2015). Ecological generalism and behavioural innovation in birds: technical
intelligence or the simple incorporation of new foods? Journal of Animal Ecology, 84(1), 79-89.
Dumbrell, A. J., Clark, E. J., Frost, G. A., Randell, T. E., Pitchford, J. W., & Hill, J. K. (2008). Changes in species
diversity following habitat disturbance are dependent on spatial scale: theoretical and empirical
evidence. Journal of Applied Ecology, 45(5), 1531-1539.
Fahrig, L. (2007). Non‐optimal animal movement in human‐altered landscapes. Functional Ecology, 21(6), 10031015.
Feeley, K. J., & Silman, M. R. (2011). The data void in modeling current and future distributions of tropical
species. Global Change Biology, 17(1), 626-630.
Ferraz, G., Russell, G. J., Stouffer, P. C., Bierregaard, R. O., Pimm, S. L., & Lovejoy, T. E. (2003). Rates of species
loss from Amazonian forest fragments. Proceedings of the National Academy of Sciences, 100(24), 14069-14073.
Fischer, J., & Lindenmayer, D. B. (2007). Landscape modification and habitat fragmentation: a synthesis. Global
Ecology and Biogeography, 16(3), 265-280.
Flato, G., Marotzke, J., Abiodun, B., Braconnot, P., Chou, S. C., Collins, W. J., ... & Forest, C. (2013). Evaluation of
Climate Models. In: Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the
Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Climate Change 2013, 5, 741-866.
Foley, J. A., Asner, G. P., Costa, M. H., Coe, M. T., DeFries, R., Gibbs, H. K., ... & Snyder, P. (2007). Amazonia
revealed: forest degradation and loss of ecosystem goods and services in the Amazon Basin. Frontiers in Ecology
and the Environment, 5(1), 25-32.
Gardner, T. A., Barlow, J., Chazdon, R., Ewers, R. M., Harvey, C. A., Peres, C. A., & Sodhi, N. S. (2009). Prospects
for tropical forest biodiversity in a human‐modified world. Ecology letters, 12(6), 561-582.
Gillies, C. S., & St Clair, C. C. (2010). Functional responses in habitat selection by tropical birds moving through
fragmented forest. Journal of Applied Ecology, 47(1), 182-190.
23
Haddad, N. M., Brudvig, L. A., Clobert, J., Davies, K. F., Gonzalez, A., Holt, R. D., ... & Cook, W. M. (2015). Habitat
fragmentation and its lasting impact on Earth’s ecosystems. Science Advances, 1(2), e1500052.
Halley, J. M., & Iwasa, Y. (2011). Neutral theory as a predictor of avifaunal extinctions after habitat
loss. Proceedings of the National Academy of Sciences, 108(6), 2316-2321.
Halley, J. M., Sgardeli, V., & Triantis, K. A. (2014). Extinction debt and the species–area relationship: a neutral
perspective. Global ecology and biogeography, 23(1), 113-123.
Halley J.M., Monokrousos N., Mazaris A.D., Newmark W.D. & Vokou D. (2016). Dynamics of extinction debt across
five taxonomic groups. Nature Communications 7, Article number: 12283.
Hansen, M., R. S. DeFries, J. R. G. Townshend, and R. Sohlberg. 2003. LBA Regional Land Cover from AVHRR, 1km, 1992-1993 (Hansen et al.). Available on-line [http://www.daac.ornl.gov] from the Oak Ridge National
Laboratory Distributed Active Archive Center, Oak Ridge, Tennessee, U.S.A. doi:10.3334/ORNLDAAC/678
Hansen, M. C., P. V. Potapov, R. Moore, M. Hancher, S. A. Turubanova, A. Tyukavina, D. Thau, S. V. Stehman, S. J.
Goetz, T. R. Loveland, A. Kommareddy, A. Egorov, L. Chini, C. O. Justice, and J. R. G. Townshend. 2013. “HighResolution Global Maps of 21st-Century Forest Cover Change.” Science 342 (15 November): 850–53. Data
available on-line from: http://earthenginepartners.appspot.com/science-2013-global-forest
Hill, J. K., & Hamer, K. C. (2004). Determining impacts of habitat modification on diversity of tropical forest fauna:
the importance of spatial scale. Journal of Applied Ecology, 41(4), 744-754.
Hubbell, S. P., He, F., Condit, R., Borda-de-Água, L., Kellner, J., & ter Steege, H. (2008). How many tree species are
there in the Amazon and how many of them will go extinct?. Proceedings of the National Academy of Sciences,
105(Supplement 1), 11498-11504.
IUCN 2016. The IUCN Red List of Threatened Species. Version 2016-3. <http://www.iucnredlist.org>. Downloaded
on 07 December 2016.
Kennedy, C. M., Marra, P. P., Fagan, W. F., & Neel, M. C. (2010). Landscape matrix and species traits mediate
responses of Neotropical resident birds to forest fragmentation in Jamaica. Ecological Monographs, 80(4), 651669.
Kuussaari, M., Bommarco, R., Heikkinen, R. K., Helm, A., Krauss, J., Lindborg, R., ... & Stefanescu, C. (2009).
Extinction debt: a challenge for biodiversity conservation. Trends in ecology & evolution, 24(10), 564-571.
Laurance, W. F., Camargo, J. L., Luizão, R. C., Laurance, S. G., Pimm, S. L., Bruna, E. M., ... & Van Houtan, K. S.
(2011). The fate of Amazonian forest fragments: a 32-year investigation. Biological Conservation, 144(1), 56-67.
Laurance, W. F., & Williamson, G. B. (2001). Positive feedbacks among forest fragmentation, drought, and climate
change in the Amazon. Conservation Biology, 15(6), 1529-1535.
Law, B. E., Falge, E., Gu, L. V., Baldocchi, D. D., Bakwin, P., Berbigier, P., ... & Goldstein, A. (2002). Environmental
controls over carbon dioxide and water vapor exchange of terrestrial vegetation. Agricultural and Forest
Meteorology, 113(1), 97-120.
Lees, A. C., & Peres, C. A. (2006). Rapid avifaunal collapse along the Amazonian deforestation frontier. Biological
Conservation, 133(2), 198-211.
Lees, A. C., & Peres, C. A. (2008). Avian life‐history determinants of local extinction risk in a hyper‐fragmented
Neotropical forest landscape. Animal Conservation, 11(2), 128-137.
Lees, A. C., & Peres, C. A. (2009). Gap‐crossing movements predict species occupancy in Amazonian forest
fragments. Oikos, 118(2), 280-290.
Lewis, S. L., Brando, P. M., Phillips, O. L., van der Heijden, G. M., & Nepstad, D. (2011). The 2010 amazon drought.
Science, 331(6017), 554-554.
24
Marini, M. A. (2001). Effects of forest fragmentation on birds of the cerrado region, Brazil. Bird Conservation
International, 11(01), 13-25.
Martensen, A. C., Pimentel, R. G., & Metzger, J. P. (2008). Relative effects of fragment size and connectivity on
bird community in the Atlantic Rain Forest: implications for conservation. Biological Conservation, 141(9),
2184-2192.
Myers, N., Mittermeier, R. A., Mittermeier, C. G., Da Fonseca, G. A., & Kent, J. (2000). Biodiversity hotspots for
conservation priorities. Nature, 403(6772), 853-858.
McGuire, J. A., Witt, C. C., Altshuler, D. L., & Remsen, J. V. (2007). Phylogenetic systematics and biogeography
of hummingbirds: Bayesian and maximum likelihood analyses of partitioned data and selection of an
appropriate partitioning strategy. Systematic biology, 56(5), 837-856.
McClain, M. E., & Naiman, R. J. (2008). Andean influences on the biogeochemistry and ecology of the Amazon
River. BioScience, 58(4), 325-338.
Mitchell, M. G., Suarez-Castro, A. F., Martinez-Harms, M., Maron, M., McAlpine, C., Gaston, K. J., ... & Rhodes, J.
R. (2015). Reframing landscape fragmentation's effects on ecosystem services. Trends in ecology & evolution,
30(4), 190-198.
Moore, R. P., Robinson, W. D., Lovette, I. J., & Robinson, T. R. (2008). Experimental evidence for extreme dispersal
limitation in tropical forest birds. Ecology letters, 11(9), 960-968.
Newbold, T., Hudson, L. N., Phillips, H. R., Hill, S. L., Contu, S., Lysenko, I., ... & De Palma, A. (2014). A global model
of the response of tropical and sub-tropical forest biodiversity to anthropogenic pressures. Proceedings of the
Royal Society of London B: Biological Sciences, 281(1792), 20141371.
Ochoa‐Quintero, J. M., Gardner, T. A., Rosa, I., Barros Ferraz, S. F., & Sutherland, W. J. (2015). Thresholds of
species loss in Amazonian deforestation frontier landscapes. Conservation Biology, 29(2), 440-451.
O'Dwyer, J. P., & Green, J. L. (2010). Field theory for biogeography: a spatially explicit model for predicting
patterns of biodiversity. Ecology letters, 13(1), 87-95.
O'Grady, J. J., Reed, D. H., Brook, B. W., & Frankham, R. (2004). What are the best correlates of predicted
extinction risk?. Biological Conservation, 118(4), 513-520.
Olson, D. M., Dinerstein, E., Wikramanayake, E. D., Burgess, N. D., Powell, G. V. N., Underwood, E. C., D'Amico, J.
A., Itoua, I., Strand, H. E., Morrison, J. C., Loucks, C. J., Allnutt, T. F., Ricketts, T. H., Kura, Y., Lamoreux, J. F.,
Wettengel, W. W., Hedao, P., Kassem, K. R. 2001. Terrestrial ecoregions of the world: a new map of life on Earth.
Bioscience 51(11):933-938.
Pansarin, E. R., & Pedro, S. R. D. M. (2016). Reproductive biology of a hummingbird‐pollinated Billbergia: light
influence on pollinator behaviour and specificity in a Brazilian semi‐deciduous forest. Plant Biology, 18(6), 920927.
Pardini, R., de Arruda Bueno, A., Gardner, T. A., Prado, P. I., & Metzger, J. P. (2010). Beyond the fragmentation
threshold hypothesis: regime shifts in biodiversity across fragmented landscapes. PLoS One, 5(10), e13666.
Peres, C. A., Gardner, T. A., Barlow, J., Zuanon, J., Michalski, F., Lees, A. C., ... & Feeley, K. J. (2010). Biodiversity
conservation in human-modified Amazonian forest landscapes. Biological Conservation, 143(10), 2314-2327.
Philpott, S. M., Arendt, W. J., Armbrecht, I., Bichier, P., Diestch, T. V., Gordon, C., ... & TEJEDA‐CRUZ, C. E. S. A. R.
(2008). Biodiversity loss in Latin American coffee landscapes: review of the evidence on ants, birds, and trees.
Conservation Biology, 22(5), 1093-1105.
Powell, L. L., Zurita, G., Wolfe, J. D., Johnson, E. I., & Stouffer, P. C. (2015). Changes in Habitat Use at Rain Forest
Edges Through Succession: a Case Study of Understory Birds in the Brazilian Amazon. Biotropica, 47(6), 723-732.
25
Rahbek, C., Gotelli, N. J., Colwell, R. K., Entsminger, G. L., Rangel, T. F. L., & Graves, G. R. (2007). Predicting
continental-scale patterns of bird species richness with spatially explicit models. Proceedings of the Royal Society
of London B: Biological Sciences, 274(1607), 165-174.
Rahbek, C., & Graves, G. R. (2000). Detection of macro-ecological patterns in South American hummingbirds is
affected by spatial scale. Proceedings of the Royal Society of London B: Biological Sciences, 267(1459), 2259-2265.
Rocca-de-Andrade, M. A. (2006). Recurso floral para aves em uma comunidade de Mata Atlântica de encosta:
sazonalidade e distribuição vertical. Dissertation, Universidade Estadual de Campinas.
Rodewald, P. G., & Brittingham, M. C. (2004). Stopover habitats of landbirds during fall: use of edge-dominated
and early-successional forests. The Auk, 121(4), 1040-1055.
Rodrigues, L. C., & Araujo, A. C. (2011). The hummingbird community and their floral resources in an urban forest
remnant in Brazil. Brazilian Journal of Biology, 71(3), 611-622.
Rodrigues, L. D. C., & Rodrigues, M. (2014). Flowers visited by hummingbirds in the open habitats of the
southeastern brazilian mountaintops: species composition and seasonality. Brazilian Journal of Biology, 74(3),
659-676.
Saunders, D. A., Hobbs, R. J., & Margules, C. R. (1991). Biological consequences of ecosystem fragmentation: a
review. Conservation biology, 5(1), 18-32.
Slatyer, R. A., Hirst, M., & Sexton, J. P. (2013). Niche breadth predicts geographical range size: a general ecological
pattern. Ecology letters, 16(8), 1104-1114.
Stouffer, P. C., Bierregaard, R. O., Strong, C., & Lovejoy, T. E. (2006). Long‐Term Landscape Change and Bird
Abundance in Amazonian Rainforest Fragments. Conservation Biology, 20(4), 1212-1223.
Stouffer, P. C., Strong, C., & Naka, L. N. (2009). Twenty years of understorey bird extinctions from Amazonian
rain forest fragments: consistent trends and landscape‐mediated dynamics. Diversity and Distributions, 15(1),
88-97.
Swihart, R. K., Lusk, J. J., Duchamp, J. E., Rizkalla, C. E., & Moore, J. E. (2006). The roles of landscape context, niche
breadth, and range boundaries in predicting species responses to habitat alteration. Diversity and Distributions,
12(3), 277-287.
Volkoff, B., Mello, F. F. D. C., Maia, S. M. F., & Cerri, C. E. P. (2012). Landscape and soil regionalization in southern
Brazilian Amazon and contiguous areas: methodology and relevance for ecological studies. Scientia Agricola,
69(3), 217-225.
Walker, B., Holling, C. S., Carpenter, S. R., & Kinzig, A. (2004). Resilience, adaptability and transformability in
social--ecological systems. Ecology and society, 9(2), 5.
Waltert, M., Bobo, K. S., Kaupa, S., Montoya, M. L., Nsanyi, M. S., & Fermon, H. (2011). Assessing conservation
values: biodiversity and endemicity in tropical land use systems. PLoS One, 6(1), e16238.
Wearn, O. R., Reuman, D. C., & Ewers, R. M. (2012). Extinction debt and windows of conservation opportunity in
the Brazilian Amazon. Science, 337(6091), 228-232.
26
8. Appendix
Appendix I - Records of raw data
Study
sites
Deforestation
of forest
cover (%)
1993
Deforestation
of forest
cover (%)
2014
Δ Deforestation
change (%)
1993 - 2014
Bird data
quantity
available
Species
richness
(raw data)
Standardized
species richness
(Eq. 3)
Standardized
species
richness
(Eq. 2)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
98,88%
18,95%
-79,93%
299
33
1,16
5,78
98,33%
4,28%
-94,05%
927
46
1,79
6,73
95,75%
81,70%
-14,05%
973
35
0,13
5,08
87,15%
3,19%
-83,96%
284
31
0,88
5,48
57,95%
0,72%
-57,23%
21
11
0,64
3,61
59,53%
16,75%
-42,78%
63
21
1,15
5,06
95,25%
3,39%
-91,86%
119
22
0,34
4,60
94,07%
2,64%
-91,43%
103
20
0,12
4,31
99,15%
6,48%
-92,68%
3429
39
-0,39
4,79
30,86%
12,12%
-18,75%
252
22
-0,58
3,97
0,00%
8,42%
8,42%
866
25
-1,22
3,69
0,00%
3,95%
3,95%
21
11
0,64
3,61
11,60%
5,18%
-6,43%
67
20
0,80
4,75
1,49%
9,24%
7,75%
23
8
-0,52
2,55
14,62%
6,84%
-7,78%
1615
41
0,49
5,55
33,76%
16,11%
-17,64%
1654
29
-1,14
3,91
1,53%
6,35%
4,81%
72
15
-0,48
3,50
54,68%
3,05%
-51,63%
2408
52
1,54
6,67
39,82%
73,23%
33,41%
547
35
0,74
5,55
0,00%
0,05%
0,05%
10
6
0,77
2,60
44,66%
27,58%
-17,08%
278
32
1,08
5,68
42,94%
2,42%
-40,53%
2073
55
2,09
7,20
0,00%
6,84%
6,84%
603
33
0,32
5,15
0,00%
4,26%
4,26%
455
23
-0,99
3,75
0,29%
6,16%
5,87%
188
20
-0,63
3,81
22,32%
9,05%
-13,27%
203
24
0,03
4,51
34,86%
9,94%
-24,92%
107
24
0,92
5,13
52,68%
13,82%
-38,86%
62
14
-0,50
3,39
1,69%
8,82%
7,13%
2226
28
-1,48
3,63
0,00%
4,51%
4,51%
12
8
1,04
3,21
34,93%
1,85%
-33,08%
38
3
-2,72
0,82
8,46%
5,51%
-2,95%
8
3
0,11
1,44
5,68%
38,71%
33,03%
16
8
0,26
2,88
5,57%
0,89%
-4,68%
17
8
0,11
2,82
0,56%
12,82%
12,25%
11
4
-0,34
1,66
0,84%
0,09%
-0,75%
74
10
-1,68
2,32
0,28%
3,13%
2,85%
121
20
-0,09
4,17
1,56%
4,72%
3,16%
141
20
-0,29
4,04
27
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
5,59%
0,86%
-4,73%
213
23
-0,21
4,29
No data
11,87%
No data
344
21
-1,07
3,59
No data
25,04%
No data
82
21
0,69
4,76
6,15%
24,95%
18,81%
51
22
1,82
5,59
0,00%
41,33%
41,33%
9
4
0,21
1,82
0,00%
21,73%
21,73%
314
24
-0,46
4,17
11,42%
25,37%
13,95%
198
19
-0,88
3,59
0,00%
57,81%
57,81%
117
18
-0,46
3,77
0,34%
34,01%
33,66%
198
18
-1,07
3,40
0,00%
14,29%
14,29%
263
21
-0,81
3,76
3,65%
22,90%
19,25%
58
11
-1,14
2,70
0,00%
9,98%
9,98%
27
8
-0,81
2,42
0,00%
13,16%
13,16%
108
20
0,06
4,27
0,00%
11,71%
11,71%
166
20
-0,49
3,91
0,00%
13,91%
13,91%
132
19
-0,41
3,89
0,00%
14,08%
14,08%
11
4
-0,34
1,66
0,56%
42,19%
41,63%
74
28
2,49
6,50
0,00%
14,43%
14,43%
125
25
0,89
5,17
0,00%
17,20%
17,20%
68
16
-0,16
3,79
0,00%
28,73%
28,73%
72
28
2,55
6,54
0,00%
18,65%
18,65%
34
7
-1,47
1,98
0,00%
27,13%
27,13%
13
8
0,81
3,11
0,00%
48,75%
48,75%
48
11
-0,88
2,84
0,00%
18,49%
18,49%
178
18
-0,95
3,47
0,00%
5,71%
5,71%
38
20
1,94
5,49
0,00%
19,76%
19,76%
28
8
-0,87
2,40
0,28%
44,17%
43,89%
25
11
0,25
3,41
0,00%
56,47%
56,47%
51
15
0,04
3,81
0,00%
27,70%
27,70%
12
7
0,64
2,81
0,00%
11,28%
11,28%
22
10
0,21
3,23
0,00%
13,57%
13,57%
129
13
-1,62
2,67
0,00%
9,95%
9,95%
69
14
-0,66
3,30
75,89%
73,47%
-2,42%
107
6
-2,92
1,28
15,86%
52,99%
37,12%
23
6
-1,16
1,91
7,33%
36,86%
29,53%
14
7
0,22
2,65
0,00%
9,69%
9,69%
15
6
-0,31
2,21
1,21%
27,41%
26,20%
31
11
-0,17
3,20
0,00%
15,00%
15,00%
40
14
0,20
3,79
1,33%
28,21%
26,88%
107
17
-0,56
3,63
2,82%
20,54%
17,73%
41
20
1,77
5,38
88,58%
11,10%
-77,48%
26
15
1,40
4,60
0,00%
60,83%
60,83%
21
12
0,97
3,94
• Light sections represent locations of present low deforestation degree and dark sections represent locations of present
high deforestation degree.
28