Download Predicting Distribution, Habitat Suitability and the Potential Loss of

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Document related concepts

Latitudinal gradients in species diversity wikipedia, lookup

Wildlife corridor wikipedia, lookup

Molecular ecology wikipedia, lookup

Extinction debt wikipedia, lookup

Ecological fitting wikipedia, lookup

Island restoration wikipedia, lookup

Bifrenaria wikipedia, lookup

Restoration ecology wikipedia, lookup

Theoretical ecology wikipedia, lookup

Occupancy–abundance relationship wikipedia, lookup

Assisted colonization wikipedia, lookup

Biogeography wikipedia, lookup

Source–sink dynamics wikipedia, lookup

Biodiversity action plan wikipedia, lookup

Mission blue butterfly habitat conservation wikipedia, lookup

Habitat destruction wikipedia, lookup

Biological Dynamics of Forest Fragments Project wikipedia, lookup

Reconciliation ecology wikipedia, lookup

Habitat wikipedia, lookup

Habitat conservation wikipedia, lookup

Transcript
Grand Valley State University
[email protected]
Masters Theses
Graduate Research and Creative Practice
4-2006
Predicting Distribution, Habitat Suitability and the
Potential Loss of Habitat for Sitta formosa, The
Beautiful Nuthatch
Jill C. Witt
Grand Valley State University
Follow this and additional works at: http://scholarworks.gvsu.edu/theses
Part of the Biology Commons
Recommended Citation
Witt, Jill C., "Predicting Distribution, Habitat Suitability and the Potential Loss of Habitat for Sitta formosa, The Beautiful Nuthatch"
(2006). Masters Theses. 637.
http://scholarworks.gvsu.edu/theses/637
This Thesis is brought to you for free and open access by the Graduate Research and Creative Practice at Scholar[email protected] It has been accepted
for inclusion in Masters Theses by an authorized administrator of [email protected] For more information, please contact
[email protected]
PREDICTING DISTRIBUTION, HABITAT SUITABILITY AND THE POTENTIAL
LOSS OF HABITAT FOR Sittaformosa, THE BEAUTIFUL NUTHATCH
A thesis submitted in partial fulfillment of
the requirements for the degree of
Master of Science
By
Jill C. Witt
To
Biology Department
Grand Valley State University
Allendale, Michigan
April 2006
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
111
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Copyright by
Jill Christine Witt
2005
IV
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
ACKNOWLEDGEMENTS
I am grateful to the following people for their support at various stages of this
research: S. Menon, M. Lombardo, L. Thomasma, J. Peters, and K. Thompson of GVSU,
J. Tordoff of BirdLife International, S. Swan and S. O’Reilly of Fauna and Flora
International, Dr. Mohammed Irfan-Ullah, Ashoka Trust for Research in Ecology and the
Environment (ATREE), Bangalore, and A. Hirzel, University of Bern. Additionally, I
would like to thank GVSU’s Summer Student Scholar’s program for funding portions of
this research.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
ABSTRACT
The beautiful nuthatch, Sitta formosa, occurs in high-altitudc evergreen and semi­
evergreen forests throughout the south and southeastern extent of the Himalayan
Mountains. Populations of S. formosa are small, declining and severely fragmented as a
result of habitat degradation and fragmentation, and therefore it is considered vulnerable
by the World Conservation Union and is included in the 2004 Red List o f Threatened
Species. I used ecological niche factor analysis to model the potential distribution and
predict suitable habitat for S. formosa. By using 59 presence locations, collected from
museum specimens and biodiversity surveys, and together with topographic and climate
variables, I found S. formosa to be linked to much greater than average rainfall and
greater than average slopes throughout the study area. S. formosa presence points were
highly correlated with mature forests (frequency = 0.86), consisting of evergreen
broadleaf, deciduous broadleaf and mixed forests. Core habitat (habitat suitability index >
80) was predicted for 918,000 km^ within south and southeast Asia, yet current cover
type maps indicate that only 57% of this area remains forested. Potentially, as much as
270,000 km^ of this historically highly suitable habitat has been converted to croplands.
At a coarse scale, S. formosa populations that are threatened by agriculture and timber
extraction are potentially most vulnerable to habitat loss, fragmentation and population
isolation; however, a closer look at the biological and ecological needs of this species is
necessary for effective management. The habitat suitability maps and models derived
VI
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
here with ecological niche factor analysis can be useful tools for identifying areas for
future research, management and conservation of S. formosa.
VII
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
TABLE OF CONTENTS
LIST OF TABLES...........................................................................................
ix
LIST OF FIGURES.........................................................................................
x
INTRODUCTION...........................................................................................
1
BACKGROUND INFORMATION...............................................................
The species and its habitat..................................................................
Predictive species modeling...............................................................
4
4
7
METHODS.......................................................................................................
Study area, species data and environmental variables.....................
Predicting areas of suitable habitat with ENFA................................
Land cover type analysis.....................................................................
13
13
17
18
RESULTS.........................................................................................................
Ecological niche factor analysis..........................................................
Habitat suitability map.........................................................................
Cover type and corresponding habitat suitability..............................
19
19
22
25
DISCUSSION...................................................................................................
Predictive modeling, species distribution and habitat requirements.
Habitat and cover type analysis..........................................................
Caveats of predictive modeling..........................................................
26
26
30
33
LITERATURE CITED...................................................................................
37
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
LIST OF TABLES
TABLE
PAGE
1.
Sources of Sitta formosa presence point locations.....................
15
2.
Environmental variables used in ENFA model of
Sitta formosa habitat suitability.................................................
16
Variables used in ENFA model development and their
respective validation scores....................................... ................
19
4.
Marginality and specialization scores in ENFA........................
21
5.
Total area and relative proportions of forested and cropland
landscapes in areas predicted suitable for S. formosa...............
25
3.
IX
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
LIST OF FIGURES
FIGURE
PAGE
1.
ENFA marginality and specialization...............................................
II
2.
Sitta formosa distribution..................................................................
14
3.
Area-adjusted frequency of validation points in ENFA analysis..
23
4.
Sitta formosa habitat suitability map................................................
24
5.
Land cover analysis of Sitta formosa potential distribution
32
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
INTRODUCTION
Forest fragmentation and loss of forested habitats constitute some of the greatest
threats to faunal biodiversity and are primary contributors to species extinctions in many
developing countries. Specifically, it has been repeatedly documented that deforestation
and habitat fragmentation adversely affect local forest bird communities (Bierregaard and
Lovejoy 1989, Stouffer and Bierrregaard 1995, Turner 1996). The beautiful nuthatch,
Sitta formosa, is a bird species that occurs throughout the eastern and southeastern extent
of the Himalayan Mountains, including India, Bhutan, Myanmar, China, Thailand, Laos
and Vietnam (Birdlife International 2003). Little is known of S. form osa’s habitat
requirements other than an apparent link with the oldest and largest trees in montane
forests (Birdlife International 2003). S. formosa has been documented at elevations
ranging from 350 to 2,400 meters (Grimmett et al. 1998), and has been found in
association with broadleaf evergreen and semi-evergreen forests (Smythies 1949, Robson
2002, Tordoff et al. 2002). S. formosa is extremely local within its extensive range,
suggesting that it may have specialized habitat needs, and it may be restricted seasonally
or locally to certain altitudinal zones and/or forest types (Collar et al. 1994, Birdlife
International 2003). To date, no consistent habitat association has been identified;
however, it is thought that S. formosa has a small, declining and severely fragmented
population as a result of habitat loss and fragmentation (Birdlife 2003). For this reason it
is classified as vulnerable (facing a high risk of extinction in the wild in the medium-term
future) in Threatened Birds o f Asia: The BirdLife International Red Data Book (2003),
and is included in the World Checklist of Threatened Birds (Collar et al. 1994) and the
World Conservation Union’s Red List of Threatened Species (lUCN 2003).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Information on species distribution, habitat requirements, and key areas for
protection is critical for species conservation. In the last few decades, advances in
computer technology and the development of geographic information systems (GIS) have
allowed for increasingly powerful tools for mapping, spatial and statistical analyses, and
species and habitat modeling. These multivariate, spatially-explicit models combine
species occurrence data with biotic and abiotic ecological and environmental variables to
identify suitable habitat and predict the potential distribution of a species. Such models
have been used to predict species occurrences, identify regional population patterns,
predict species invasions and identify important areas for conservation (Corsi et al. 1999,
Osborne et al. 2001, Peterson and Vieglais 2001).
Ecological niche factor analysis (ENFA) is one such spatially explicit model that
builds on Hutchinson’s (1957) definition of an ecological niche as a hyper-volume in
multidimensional space of ecological variables within which a species can maintain a
viable population (Hirzel et al. 2002). Ecological niche factor analysis differs from many
multivariate species distribution models in that analysis requires presence-only data as
opposed to presence and absence data (Hirzel et al. 2002). As in the case of S. formosa,
where little is known and information on the species’ distribution comes only from
museum specimens and biodiversity surveys, data are considered to be presenee-only,
and absence data is unavailable. ENFA compares the species distribution to a set of
environmental (e.g. climate, topography) variables (ENVs) for the study area, and using a
factor analysis, similar to principal component analysis, determines the suitability of
habitat for the species over the entire study area (Hirzel et al. 2002).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The objectives of this study were 1) to predict potential areas of suitable habitat
for S. formosa using Ecological Niche Factor Analysis, 2) to determine land cover types
within areas of predicted suitable habitat, and 3) to explore the significance of the
methods and the results for developing conservation strategies and priorities.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
BACKGROUND INFORMATION
The species and its habitat
At present, little is known of S. formosa's distribution or habitat requirements. S.
formosa has been found in a wide range of elevations throughout the eastern and
southeastern extent of the Himalayan and Annamite mountain ranges (Grimmett et al.
1998). It has been documented at altitudes ranging from 350 to 2,400 meters (Birdlife
International 2003). Robson (2002), in Birds o f Thailand, suggests an association with
broadleaved evergreen forests and a vertical distribution of 1,800-2,285 meters, whereas
King and Dickenson (1975) simply state its residency as forests above 1,000 meters. The
Pictoral Guide to Birds o f India (All 1983) states that S. formosa is found in the
Northeast Hill States in deep, wet semi-evergreen and evergreen forests with a vertical
distribution of 330-2,400 meters. Ali and Ripley (1973) suggest a summer distribution
between 1,500 and 2,100 meters and in winter between 330 and 2,000 meters, affecting
deep forests. Grimmett et al. (1998) suggests some altitudinal movement downslope in
the non-breeding season, with distributions at 1,500-2,400 meters in summer and 3502,200 meters in winter.
Throughout the literature, S. formosa is noted to be seen most often in evergreen
and semi-evergreen forests. Collar et al. (1994) and Harrap and Quinn (1995) speculated
that S. formosa may have highly specialized habitat needs. Individuals have been found
in association with broadleaved evergreen forests (Robson 2002), in deep, wet semi­
evergreen and evergreen forests (Ali and Ripley 1973), and in typical old-growth
evergreen forests devoid of large conifers (Davidson 1998, as summarized in Birdlife
2003). It is frequently observed foraging in mixed feeding flocks (Birdlife International
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
2003, Tordoff et al. 2002) and in particularly large trees draped in moss, lichens, orchids
or other epiphytes (Harrap and Quinn 1996, Hopkins 1989). Tordoff et al. (2002)
documented S. formosa in upper and lower montane evergreen forests. Both forests were
relatively undisturbed and contained a high density of mature Fokienia hodginsii (Dunn),
a cypress-like conifer. Individuals were regularly eneountered in eanopies of F. hodginsii
at Nakai-Nam Theun, Laos (Tobias 1997, Thewlis et al. 1998, Tordoff et al. 2002),
where the species may be partially or locally ecologically reliant on this rare tree species.
In a survey of three montane forests in north-east Vietnam, S. formosa was seen foraging
in moss- and lichen-draped branches in three separate locations in areas of primary or old
secondary evergreen forests (Vogel et al. 2003). Smythies (1949) suggests that its
favored habitat is dense evergreen broadleaved forests; however, his expedition sighted
one in Myanmar in “open country with scattered trees.” This may imply that it ean
tolerate a certain degree of habitat disturbance (Birdlife International 2003). Sitta fomosa
is extremely local within its extensive range, suggesting that it may have specialized
habitat needs (Collar et a l, 1994). Birdlife (2003) suggests that S. formosa might be
seasonally or locally restricted to certain altitudinal zones or forest types; however, no
consistent association has been identified. S. formosa is presumably vulnerable to
destruction and degradation of forests (Collar et al. 1994, Harrap and Quinn 1996).
Forest loss, degradation and isolation are recurrent themes throughout the entire
range of distribution occupied by S. formosa (Birdlife International 2003) as a result of
timber extraction and exploitation, and slash and bum cultivation. Specifically, F.
hodginsii, a commercially valuable tree speeies, is being extraeted from areas inhabited
by S. formosa. F. hodginsii is a shade- intolerant species that is well adapted to mild
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
climates with abundant rainfall and occurs naturally on humid soils in high mountain
areas, slopes or flats (Earle 1997). Its distribution extends through China, Laos and
Vietnam, where it is becoming scarce throughout its range. It is threatened by agriculture
and timber extraction and is considered “near threatened” in the lUCN Red List of
Threatened Species (lUCN 2003). While the association between S. formosa and F.
hodginsii remains unclear, it is clear that F. hodginsii is a montane species utilized by S.
formosa and it is being exploited throughout its range (Osborn 2004, Thewlis et al.
1998). Thewlis et al. (1998), Birdlife International (2003), and Tordoff et al. (2002)
recommend further studies investigating the ecological association or lack thereof
between S. formosa and F. hodginsii. Additionally, Osborn (2004) in a desk study for
Fauna and Flora International’s Hoang Lien Son Project in Vietnam suggests that if an
association does indeed exist between these two species, at least locally, conservation of
both species together could provide a useful mechanism for protection.
Sitta formosa currently receives no protection in China, India, Laos or Vietnam,
although it is legally protected in Thailand and appears on a list of protected species in
Myanmar (Birdlife International 2003). Birdlife International (2003) states that further
research is necessary for the protection of S. formosa. In particular, information on
population size estimates and status in key protected areas is necessary to determine
whether protection of these sites will be sufficient for maintaining viable populations.
Additionally, they suggest that further research is needed to clarify habitat requirements
for this species, including association or lack thereof with F. hodginsii.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Predictive species modeling
The mapping of species’ distributions is fundamentally important for
understanding biodiversity patterns and improving understanding of the appropriateness
of habitat areas for individual species (Bailey and Hogg 1986, Fa and Morales 1993,
Miller 1994, Teuton et al. 2000). Geographic Information Systems (GIS) are powerful,
computer-based tools for organizing, accessing, displaying, analyzing and modeling
spatial information such as species distribution maps. Data are stored in different layers
within GIS in vector and raster format. For example, vegetation cover types ean be stored
as a network of grid cells, with each cell representing a single value of cover type. These
layers ean be combined and used to create descriptive, predictive and prescriptive models
for management and decision making, which can then be displayed as formal
representations of spatial information.
Multivariate, spatially explicit models have been used to identify and predict
species’ distributions and habitat suitability, and therefore they are an important tool for
speeies and biodiversity conservation, management and planning. These models combine
speeies oeeurrenee data with biotic and abiotic ecological and environmental variables to
create a model of species’ requirements and predict potential distributions. Logistic
regression, for example, has been widely used to model species’ distributions and to
predict species occurrences as well as favorable habitat (Osborne et al. 2001, Mladenoff
et al. 1995, Carroll et al. 1999). Aspinall and Veitch (1993) used a Bayesian probability
method to predict curlew occurrence in Scotland. Mahalanobis distance statistic has been
used to model spatial patterns for identifying regional patterns in gray wolf distribution
(Corsi et al. 1999) and to predict habitat use potential for black bears (Clark et al. 1993).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Another computer-based modeling approach, genetic algorithm for rule-set prediction
(GARP), uses a genetic algorithm to develop a set of rules that describe the relationship
between species’ data and environmental variables (Stoekwell and Noble 1992,
Stockwell 1999, Stoekwell and Peters 1999). These rules can then be applied to other
cells in a study area to predict a species’ distribution. GARP has been used to predict bird
community composition, species invasions and speeies distribution using breeding bird
surveys (Feria and Peterson 2002, Peterson 2001, Peterson and Vieglais 2001).
Guisan and Zimmerman (2000) described the process of formulating a conceptual
model. They suggested that the formulation of an ecological model should be based on an
underlying ecological concept, such as a species’ realized niche. The model should be
formulated in a statistical way. The model should then pass through a series of
calibrations and validations to adjust the parameters and constants to improve agreement
between the distribution data and environmental variables.
Ecological niche factor analysis (Hirzel et al. 2002) builds on Hutchinson’s
(1957) definition of an ecological niche as a hyper-volume in the multidimensional space
of ecological variables within which a species can maintain a viable population.
Ecological niche factor analysis, as explained by Hirze/ et al. (2002), differs from many
multivariate species distribution models in that analysis requires presence-only data as
opposed to presence and absence data. Accurate absence data are often difficult to obtain
because either the species could not be detected even though it was present, or the habitat
is suitable but the speeies is absent for historical reasons. ENFA compares the
distribution of localities where a species was observed to a reference set of ecological and
geographical variables describing the entire study area. A factor analysis, similar to
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
principal component analysis, is used to extract factors that give weights to each
independent ecologieal or geographical variable (independent axes). In ENFA, the first
axis aeeounts for the marginality of the species. Marginality refers to the difference
between the mean of the study area and the mean of the species for a particular
environmental variable. For example, a species of tree may grow only at the highest
elevations in a mountainous area. It is considered a more marginal species. Flirzel et al.
(2002) formally defines marginality as the absolute difference between the global
(reference or study area) mean (mo) and species mean (ms) divided by 1.96 standard
deviations {Og) of the global distribution (Figure 1). Division by cTq is needed to remove
any bias introduced by the variance of the global distribution and to ensure that
marginality values will most often be between zero and one (a value may exceed one).
Values close to zero mean that there is no differenee in the mean of species habitat and
the mean of the available habitat. A value closer to one indieates that the species lives in
a very particular habitat relative to that of the study area (Hirtzel et al. 2002).
Aecording to Hirzel et al. (2002), the remaining axes generated in ENFA result in
a linear combination of the ecological variables that maximize the variance of the species
distribution as compared to that of the study area. These axes represent the species’
specialization. Specialization is the ratio between the range of values for a partieular
environmental variable in the study area eompared with the species’ range of tolerànee
(Figure I). For example, a species of fish may live in a very narrow range of temperatures
compared to the entire range of temperatures available in a lake. Thus, this species is
specialized and has a very narrow range of tolerance. Specialization is expressed in
ENFA as the ratio of the standard deviation of the global distribution ( (7^) to that of the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
species ( a^). The specialization values of the study area range from one to infinity. The
inverse of the specialization value is the tolerance value, ranging from zero to one. A
species’ tolerance value closer to zero indicates a mueh thinner niche than that closer to
one (Patthey 2003). Spécifié values of marginality and specialization are dependant on
the mean and variance of the global set (study area) chosen as a reference. A speeies may
appear extremely marginal or speeialized when the seale of a eontinent, region or country
is used as a reference set as opposed to a small subset of those areas.
10
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Figure 1 (Hirzel et al. 2002): ENFA Marginality and Specialization.
Value of ecogeographical variable
The distribution of the focal species (black bars) for any ecogeographical variable may
differ from that of the whole set of cells (gray bars) with respect to its mean, thus
allowing marginality to be defined. It may also differ with respect to standard deviation,
thus allowing specialization to be defined._______________________________________
Hirtzel et al. (2002) described the process of calculating and building the species’
habitat suitability map. The habitat suitability (HS) for each cell in the study area for the
focal species is calculated from its value relative to the species distribution for all niche
factors selected. The overall suitability index of a cell is then computed from a
combination of all scores for each factor. Marginality and specialization are weighted
equally; however, since specialization is made up of several factors, weighting is
apportioned among all factors proportionally to their eigenvalue (Hirzel et al. 2002).
Repeating this for each cell produces a habitat suitability map, with suitability values of
11
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
zero to one. A cross-validation analysis, such as jackknife cross-validation analysis
(Fielding and Bell 1997), can be used to determine the threshold between suitable and
unsuitable habitat. Applications of ENFA include Patthey (2003), who used ENFA to
characterize and map suitable habitat for red deer populations in western Switzerland. He
found that current red deer distribution is adequately modeled using mainly land-use
variables measured at home range scales. Reutter et al. (2003) used museum specimens
only to model habitat suitability for three species of alpine mouse. Using the marginality
and specialization for each species, the authors were able to directly compare the niches
of all three species. Reutter et al. (2003) suggested that although an adequate sampling
design is the best way to collect data for predictive modeling, these are often time- and
money-consuming processes. Additionally, Bibby (1995) suggested that predictive
modeling could be an important tool when trying to identify sites that might be important
for birds, especially in areas that are difficult to access for political reasons or due to
geographic inaeeessibility.
12
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
METHODS
Study area, species data and environmental variables
Since little is known of S. form osa’s habitat requirements and the current
distribution is speculative based on current and historical sightings, I chose a study area
extent that was much larger than the currently known distribution of this species. As S.
formosa is thought to be a montane species, ranging in elevation from 350-2,400 meters,
the study area was extended to include the Himalayan mountain chain to the west of its
currently known distribution throughout India and Nepal. Additionally, in order to
account for a potential artifact in the lack of species observations due to inaccessibility of
the region, the study area was expanded to include much of Burma and China. The study
area encompasses an area of over 14 million (km^) extending from 6°N by 43°N latitude
to 69°E by 122°E longitude and includes much of the Himalayan mountain chain as well
as the foothills to the south and southeast (Figure 2).
I compiled species data from the Birdlife International species account database,
museum collections and biodiversity surveys (Table 1). Using geographic coordinates, I
converted a total of 59 presence location points for S. formosa to a boolean raster map for
use as a species presence map in ENFA analysis with each data point occupying a 1 km
by 1 km cell in the study area. The 59 S. formosa presence locations were composed of
sightings ranging from the late 1800’s to the present; therefore environmental variables
chosen for use in predictive modeling were limited to those which remain reasonably
static over time. For instance, present forest cover may differ significantly from that of
the late 1800’s as a result of direct human intervention (logging, etc.). However, the type
13
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
of forest cover likely to be found in an area is determined by various underlying
environmental (climate and topography) variables.
ri
I
i
f*.*
r
“f i t
V*
m
Cambodia
500
K ilom eters
Figure 2. Sitta formosa distribution.
S. formosa data points (n=59) locations (depicted with red symbols) throughout south and
southeast Asia.
14
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 1. Sources of Sitta formosa presence point locations.
Data source
BirdLife
International,
Species account
database
H
38
Wildlife
Conservation
Society (WCS),
Lao Program
Fauna and Flora
International (FFI),
Vietnam Program
Type
Observations
Study skins
Observations
Reference
BirdLife International (2001) Threatened
Birds of Asia: The BirdLife International
Red Data Book. Cambridge, UK: BirdLife
International.
http://www.rdb.or.id/detailbird.php?id=197
Michael Hedemark, WCS, Lao Program
Personal communication
Observations
Steven Swan, FFI, Vietnam Program,
Hoang Lien Mountains Project, Personal
communication
Bombay Natural
History Society
12
Observations
Study skins
Zafar ul-Islam, Personal communication
Peabody Museum
of Natural History,
Ornithology
Collection
2
Study skins
Peabody Museaum of Natural History,
Yale University, New Haven, CT; Online
collections database,
http://george.peabody.yale.edu/om/_____
Environmental variables consisted of topographic and climatic data (Table 2) with
a 1 km by 1 km resolution. Topographic variables of slope, elevation and aspect were
obtained from the United States Geological Survey HYDRO Ik database (U. S.
Geological Survey 2002). I converted aspect layers into sine and cosine derived easting
and northing layers to circumvent problems inherent with aspect circularity. Bioclimatic
variables of precipitation and temperature were obtained from the WorldClim
interpolated global terrestrial climate surface (Hijmans et al. 2004). Slope, elevation and
precipitation variables were divided by 100 to decrease any biases introduced by their
large values in ENFA. Using Idrisi Kilimanjaro GIS software (Eastman 2004), all
environmental variables were converted to geographic coordinate system with a 0.01
degree (one kilometer) resolution for seamless overlay with the S. formosa presence map.
15
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CD
■D
O
Q.
C
gQ .
■D
CD
C/)
C/)
Table 2. Environmental variables used in ENFA model of Sitta formosa habitat suitability.
Variable
Description
8
(O '
ELEVATION 100
SLOPE_100
3.
3"
CD
CD
■D
NORTH_ASPECT
Sine transformed
EASTASPECT
O
Q.
C
a
o
3
"O
Elevation in meters
divided by 100
Slope in percent
divided by 100
Cosine transformed
PRECIP_WARM_100
o
CD
Q.
PRECIP_COLD_100
■CDD
Mean monthly
precipitation (mm) of
warmest quarter;
Values divided by 100.
Mean monthly
precipitation (mm) of
coldest quarter:
Values divided by 100.
Range of
Values Used
in Layer
Actual
Range
Mean
Standard
Deviation
0 -8 7 .5 2
0 - 8,752m
1,532
1,690
0 -6 4 .2
-1 to +1
(South to
North)
-1 to +1
(West to
East)
0-64.2%
3.3
4.8
-1 to +1
0.028
0.699
-1 to +1
-0.021
0.710
0.01-60.99
106,099mm
327.8
306.6
0 -5 1 .8 6
05,186mm
59.3
147.4
http://lpdaac.usgs.gov/gtopo30/hydro/
http://lpdaac.usgs.gov/gtopo30/hydro/
Values in data layer multiplied by 100.
http://lpdaac.usgs.gov/gtopo30/hydro/
http://lpdaac.usgs.gov/gtopo30/hydro/
M E A N T E M PW A R M
Mean temperature of
warmest quarter.
T * 10
-147.0 to
356.0
-14.7 to
35.6
22.0
8.9
MEAN_TEMP_COLD
Mean temperature of
coldest quarter.
°C * 10
-376.0 to
275.0
-37.6 to
27.5
4.0
14.1
C/)
C/)
Source
16
Hijmans, R.J., S.E. Cameron, J.L. Parra,
P.G. Jones and A. Jarvis, 2004. The
WorldClim interpolated global terrestrial
climate surfaces. Version 1.3.
http://biogeo.berkeley.edu
Hijmans, R.J., S.E. Cameron, J.L. Parra,
P.G. Jones and A. Jarvis, 2004. The
WorldClim interpolated global terrestrial
climate surfaces. Version 1.3.
http://biogeo.berkeley.edu
Hijmans, R.J., S.E. Cameron, J.L. Parra,
P.G. Jones and A. Jarvis, 2004. The
WorldClim interpolated global terrestrial
climate surfaces. Version 1.3.
http://biogeo.berkeley.edu
Hijmans, R.J., S.E. Cameron, J.L. Parra,
P.G. Jones and A. Jarvis, 2004. The
WorldClim interpolated global terrestrial
climate surfaces. Version 1.3.
http://biogeo.berkeley.edu
Predicting areas of suitable habitat with ENFA
I used Biomappev (Hirzel et al. 2004) as the GIS and statistical tool kit designed
to run the ENFA algorithms and to build the habitat suitability (HS) model and maps.
Three combinations of ENVs were overlayed with the S. formosa presence map in
Biomapper. 1) topographic, 2) topographic, temperature and precipitation, and 3)
topographic and precipitation (Table 2). A factor analysis was used to extract factors that
give weights to each ENY. The first axis accounted for the marginality of S. formosa.
The remaining factors represented S. formosa specialization.
Factors for use in calculating habitat suitability maps were selected by using
McArthur’s broken stick method (Hirzel et al. 2002). Habitat suitability for each cell in
the study area is calculated by comparing its value relative to the species value for all
factors selected. Each grid cell is assigned a habitat suitability index ranging from zero to
100 with a value of 100 being considered to have the highest suitability, whereas a value
of zero is considered to be unsuitable.
The predictive power of the habitat suitability map was evaluated with an areaadjusted frequency cross-validation process built into the Biomapper software. Species
occurrence points were randomly separated into six equal partitions. Five of these
partitions were used to build the model, and the last independent partition was used to test
the effectiveness of the model. In other words, if most of the test points of known
nuthatch locations were outside the predicted model we can say it is not a very good
model for predicting nuthatch occurrence. This process was then repeated five times.
Each new model was generated with a different random set of species occurrences and
tested with the remaining points and so on. The models were then reclassified into four
17
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
equal-sized bins covering habitat suitability values of 0-25, 25-50, 50-75 and 75-100,
with each containing some proportion of the validation points. An area-adjusted
frequency (F) for each bin was calculated by dividing the proportion of validation points
in each bin by the proportion of area in each bin (Boyce et al. 2002). A good model will
predict F < 1 for all low- quality habitat and F > 1 for high-quality habitat with a positive
correlation between area-adjusted frequency and habitat suitability. However, if F = 1 for
all habitat suitability bins, then the HS model is completely random. A Spearman’s rank
correlation was computed for habitat suitability bins and area-adjusted frequencies. The
overall quality of the initial habitat suitability map created in Biomapper was determined
by averaging the Spearman’s rank correlation across model validation subsets.
Land cover type analysis
I conducted a cross tabulation and area analysis, in Idrisi Kilimanjaro, of land
cover type with the habitat suitability map produced in Biomapper in order to quantify
the current land cover types within predicted areas of suitable habitat. For the land cover
data I used the IGBP-DIS global 1 km land cover data set thematic map (Belward 1996)
derived from 1km AVHRR data spanning the time period from April 1992 - March 1993
downloaded from USGS EROS data center (U. S. Geological Survey 1998).
Additionally, I used the cross tabulation analysis, in Idrisi Kilimanjaro, to assess current
cover type conditions associated with each of the 59 S. formosa data points.
18
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
RESULTS
Ecological niche factor analysis
Three combinations of ecological and geographical variables were used in the
ENFA modeling of S. formosa habitat suitability (Table 3). ENFA model C, which
incorporated elevation, slope, aspect and rainfall variables, outperformed other models in
the area-adjusted frequency cross-validation (Spearman’s Rs = 0.97) and was therefore
the model selected for further analyses.
Table 3. Variables used in ENFA model development and their respective validation
scores.
Mean
Standard
ENFA Model Variables
Spearman’s Rs
Deviation
ELEVATION 100
ENFA_A
0.73
0.3
SLOPE too
NORTH ASPECT
EASTASPECT
ENFA B
ELEVATION 100
SLOPE 100
NORTH ASPECT
EAST ASPECT
PRECIP WARM 100
PRECIP COLD 100
MEAN TEMP WARM
MEAN_TEMP COLD
0.87
0.24
ENFA C
ELEVATION 100
SLOPE 100
NORTH ASPECT
EAST ASPECT
PRECIP WARM 100
PRECIP COLD 100
0.97
0.03
The six environmental variables used in ENFA Model C were summarized in
Biomapper into uncorrelated factors representing marginality and specialization. High
marginality values (those closer to one and above) indicate a species’ tendency to inhabit
19
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
extreme conditions, and low values (close to zero) indicate a tendency to inhabit
conditions average to the study area. The model had an overall marginality value of
1.685, indicating that S. formosa has a tendency to inhabit extreme conditions within this
study area extent. In other words, considering that such a wide range of environmental
conditions exists in my study area (e. g. climates ranging from desert to humid tropics, or
elevations ranging from sea level to the peak of Mt. Everest), S. formosa occupies
conditions that are far different from the average. Tolerance values are the inverse of
specialization. Low tolerance values (close to zero) indicate that a species has a narrow
niche relative to the conditions in the study area. Values approaching one indicate that the
species utilizes the entire range of values within the study area. The tolerance value for S.
formosa was 0.3, indicating a somewhat narrow niche.
Environmental variables are sorted by decreasing order of absolute value of
coefficient of the marginality factor (Table 4). Positive values on the marginality factor
(factor 1) indicate that a species prefers locations with higher values on the corresponding
ENY than the mean of the study area. The high marginality coefficient (0.946) for
precipitation in the warmest quarter (PRECIP WARM IOO) indicates that S. formosa is
linked primarily to areas with precipitation amounts much greater than those average
throughout the study area. Additionally, S. formosa is found in areas with greater than
averages slopes (0.313). Aspect and elevation varied little from the average of the study
area.
20
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 4. Marginality and specialization scores in ENFA.
Environmental variables (ENVs) are sorted by decreasing order of absolute value of
coefficient on the marginality factor (factor 1). Positive values on the marginality factor
indicate that S. formosa utilizes locations with higher values on the corresponding ENVs
than the mean of the study area, with negative values indicating the utilization of lower
values. Sign of the coefficient has no meaning on the specialization factors (factors 26).The amount of specialization accounted for with each factor is indicated in
parentheses in each column heading.
Factor specialization
explained (%)
1
(1.8)
2
(83.2)
3
(9.9)
4
(2.3)
0.023
5
(1.8)
-0.044
6
(1.0)
0.308
PRECIP WARM too
0.946
0.009
0.068
SLOPEIOO
0.313
-0.001
-0.129
-0.091
0.355
-0.845
EAST ASPECT
0.080
-0.003
-0.020
0.181
-0.902
-0.234
ELEVATION 100
-0.021
0.105
0.986
-0.014
-0.162
0.362
NORTH ASPECT
-0.008
-0.018
0.059
0.979
0.175
-0.040
PRECIP COLD 100
-.006
0.994
0.058
0.025
-0.032
0.061
21
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Habitat suitability map
The first two factors (Table 4) were retained in Biomapper (accounting for 92.5%
of the total variance) to compute the habitat suitability map (Hirzel et al. 2002). These
two factors accounted for 100% of S. formosa marginality and 85.1% of the
specialization. The area-adjusted frequency cross-validation curve was exponential in
shape with a large proportion of validation points falling into the highest categories of
habitat suitability (75-100); there were no validation points recorded in unsuitable habitat
(Figure 3). Suitable habitat was subsequently divided into five categories; category 0 is
considered unsuitable (HS < 50%), category 1-4 (HS > 50%) represents progressively
more suitable habitat, with a value of 1 signifying marginal habitat and a value of 4
corresponding to core habitat (Figure 4). Ninety-five percent of presence points
corresponded to habitat predicted suitable for S. formosa: n = 36 (HS >80, categories 3
and 4 combined), n = 10 (HS 65-79, category 2), n = 10 (HS 50-64, category 1).
22
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1 0 .7 0
5 0 -7 4
7 5 -1 0 0
Habitat suitability score
Figure 3. Area-adjusted frequency of validation points in ENFA analysis.
Area-adjusted frequency of validation points for habitat suitability scores in crossvalidation evaluation of ENFA model of S. formosa potential distribution. A value of F=1
(red line) for all habitat suitability bins would indicate no difference exists between highan d In w -n iialitv b a h ita t
23
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Unsuitable
500
K ilo m e te r
Figure 4. Sitta formosa habitat suitability.
Habitat suitability as predicted by ENFA for study area (a). Predicted habitat
suitability encompassing the known distribution of S. formosa (area bounded by
dashed line) (b) overlaid with presence points (red symbol).
24
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Cover type and corresponding habitat suitability
Suitable habitat (HS >50) was predicted for an area encompassing greater than 3
million km^, or 21% of the southeast Asian study area (Tahle 4). Within this area of
suitable habitat, only 30.5% (918,000 km^) would be considered to contain core areas
(HS >80) in which to find S. formosa habitat. S. formosa observations corresponded to
evergreen broadleaf (n = 12), deciduous broadleaf (n = 27) and mixed forests (n = 12) for
51/59 (86%) observations; eight observations corresponded to shrublands, woody
savannahs or croplands. Forested landscapes presently make up only 39.7% of the area
that would be considered suitable (HS >50), with 17.4% (5.2 x 10^ km^) corresponding
to core habitat (HS >80). In contrast, cropland accounted for 1.14 x 10^ km^, or 38% of
habitat predicted to be suitable (HS >50). At HS >80, cropland accounted for 2.7 x 10^
km^, or 29.4% of the area that is considered to be core habitat for S. formosa.
Table 5. Total area (km^) and relative proportions (in parentheses) of forested and
cropland landscapes in areas predicted suitable for S. formosa.
HS Value
Area (VoŸ
< 50
1.11x10^(78.6)
>50
3.01 X 10^ (21.3)
>80_______9.18 X 10^ (6.5)
Forested (%)*
Cropland (%)*
1.19x10^(39.5)
5.23x10^(57.0)
1.14x10^(37.8)
2.70x 10^(29.4)
t Area and proportions reported relative to study area.
* Area and proportions reported relative to area included in HS ^ 0 and HS >80.
25
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
DISCUSSION
Predictive modeling, species distribution and habitat requirements
By using only 59 observation points and basic topographic and climate variables I
was able to predict the geographic range of what may have been historically suitable
habitat as well as the area that may currently contain suitable habitat for S. formosa
throughout south and southeast Asia. For a species such as the beautiful nuthatch, where
little is known of its life history or habitat requirements, knowledge of a species’ historic,
current and potential distribution is paramount for conservation, especially when
confronted with habitat loss and fragmentation.
Historic and current habitat loss and fragmentation is a concern for at risk species
in developing countries (Novaeek and Cleland 2001). Using environmental predictor
variables that did not directly include vegetation allowed us to assess the current versus
historically available habitat. Core habitat (HS >80), as predicted for S. formosa in
ENFA, consisted of an area of 918,000 km^. Of this, 57% (522,000 km^) remains forested
and 24% (270,000 km^) has been converted to croplands. It is not known to what extent
deforestation and the conversion of land to agriculture has affected S. formosa
populations (Figure 5). However, if indeed the 270,000 km^ was historically a forested
ecosystem, it is likely that this landscape history could have had a significant effect on S.
form osa’s abundance and distribution. Schrott et al. (2005) suggest that the amount of
remaining habitat or degree of fragmentation may not be a sufficient measurement for
assessing long-term viability or extinction risk of a species if historically the rate of
landscape change occurred faster than a species’ demographic response time. By using
spatially structured demographic models, they concluded that songbirds are likely to
26
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
exhibit lag responses to habitat loss in rapidly changing landscapes. Castelletta et al.
(2000) give a poignant example of the effects of landscape history on forest bird species
of primary and secondary growth forests in Singapore. They found that 61 species went
extinct following a period of rapid rainforest clearance. The majority of these species
were insectivorous birds, with canopy feeding birds being especially hard hit. Present
rates of deforestation in south and southeast Asia average 0.8% per year (FAO 2003),
though local rates may potentially be much higher. Timber extraction and agriculture
generally occur at lower elevations in south and southeast Asia, yet logging and shifting
cultivation are still considered to be the primary threats to S. formosa populations (Collar
et al. 1994, Harrap and Quinn 1996). Birdlife International (2003) summarizes what is
known of forest destruction in areas where S. formosa is known to exist. Slash-and-bum
cultivation and shifting agriculture are recurrent themes in India, Bhutan, Laos and
Myanmar. The northeastern states of India, in particular, have experienced considerable
loss of forests, especially in areas near roads and including areas surrounding Namdapha
National Park where S. formosa is currently known to exist. Logging and timber
extraction are cited as primary threats in Vietnam and Laos (Thewlis et al. 1998, Osborn
2004); the highly valuable Fokienia hodginsii is harvested for lumber in prime areas
where S. formosa has recently been identified. Birdlife International (2003) does suggest
that although S. formosa is often found at higher elevations than where current logging
and agricultural activities are taking place, it is conceivable that deforested lowlands
could be a potential factor in impeding species dispersal and could lead to population
isolation and long-term decreases in population viability.
27
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Predictive models are used to estimate the geographic extent of a species’
fundamental niche. However, despite the greater than 3 million km^ of area predicted
suitable for S. formosa, the current known range (Figure 2) covers a relatively small
portion of that area. Our model incorporated only topographic and climate variables, and
although topography and climate are spatially correlated with vegetation, they do not take
into account other factors, such as species dispersal, interspecific competition, predation
and available habitat, that determine a species realized niche. This model, using only
abiotic variables, over-predicted suitable habitat, albeit these are areas of low suitability
or marginal habitat (HS = 50-64%), in areas of central and eastern China where only one
other species of Sittidae, Sitta europaea, the wood nuthatch, is known to exist. Seoane et
al. (2004) concluded that models including only topographic and climate variables were
adequate at predicting breeding bird distributions in Spain at a coarse scale. Additionally,
Dettki et. al. (2003) produced three separate habitat suitability maps with ENFA to model
moose (Alces alces) distribution in Sweden. Each ENFA model analyzed the following
ENVs: 1) vegetation indices, 2) topographic variables and 3) a combination of vegetation
indices and topographic variables. They found that of the combination of vegetation
indices and topographic variables determining moose habitat preference, nine were
geographical variables and only the ninth most important variable was an index of
vegetation. Thus, the focus of this predictive modeling was to, using the limited number
of S. formosa observations (n=59), determine on a coarse scale the geographic extent that
would encompass the fundamental distribution of this species.
With ENFA, S. formosa was found to be a marginal species (i.e. the species mean
on the combinations of all variables was far different than that of the study area)
28
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
throughout the south and southeast Asia study area (M = 1.685). S. formosa had a
tolerance value of 0.3, indicating a fairly narrow niche within our study area. Marginality
and specialization/tolerance values calculated in ENFA are inherently affected by
geographical extent and therefore would be affected by a change in ranges of values in
the study area. S. formosa would likely appear less marginal and more tolerant if our
study area consisted solely of the Himalayan mountain chain and its southeastern
foothills.
ENFA marginality and specialization/tolerance values could be a useful tool when
comparing the niches of related or sympatric species. Several species of nuthatch do
occur in overlapping or adjoining distributions and/or habitats (altitudinal zonation,
foraging or nesting habitats) as S. formosa: Sitta nagaensis, Sitta castanea, Sitta
himalayensis, Sitta frontalis, and Sitta magna (Matthysen 1998). Although at present
ENFA has not been used to compare species of songbirds or other avians, Reutter et al.
(2003) did use ENFA to model habitat suitability for three sympatric Apodemus species,
endemic alpine rodents of Switzerland. They used presence data derived solely from
museum specimens to directly compare the niches of these three species. Two species of
Apodemus were found to be generalists, while a third, Apodemus apicola, was found to
have a very specialized habitat selection. Additionally, it has been proposed that S.
formosa may be partially or locally ecologically reliant on Fokienia hodginsii, a
commercially valuable cypress-like conifer that is considered near-threatened in the
lUCN Red List of Threatened Species (lUCN 2003). At the time of this analysis, little
information was available on the distribution of F. hodginsii. A recent publication,
however, is now available on the habitat and distribution of F. hodginsii in Vietnam
29
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(Osborn 2004). A similar ENFA analyses could be used to compare the niches of
overlapping Sitta species or to identify whether a correlation exists between S. formosa
and this important tree species.
Habitat and cover type analysis
S. formosa presence locations when overlaid with land cover maps corresponded
primarily to broadleaf evergreen, broadleaf deciduous and mixed forest (Figure 5). This
finding is consistent with Davidson’s observation (Birdlife International 2003) that S.
formosa was found in “typical old-growth evergreen forests devoid of large conifers.” Ali
and Ripley (1973) and Robson (2002) suggest that S. formosa has been found in
association with broadleaved evergreen and wet semi-evergreen forests. O f particular
interest, however, are the 27 observations in deciduous broadleaf forests, implying that S.
formosa is not constrained to evergreen and semi-evergreen forests. Though Smythies
(1949) suggests that the favored habitat of S. formosa is dense evergreen broadleaved
forests, he observed a single bird in Myanmar (Burma) in “open country with scattered
trees.” This also is consistent with our findings, as two observations corresponded with
closed shrubland and woody savannahs which, as suggested by Birdlife International
(2003), may imply that S. formosa can tolerate a certain degree of habitat disturbance.
ENFA predicted rainfall and slope to be the most important topographic variables
determining S. formosa habitat suitability. S. formosa was found in conjunction with
rainfall amounts 95% greater than the average of the study area in the warmest quarter of
the year in south and southeast Asia. In addition, it was linked to slopes that were over
30% greater than the average conditions. However, its range of elevation was similar to
30
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
those found throughout the study area elevation, and aspect and elevation were not
considered to be strong factors in determining habitat suitability. This apparent link to
greater than average slopes, tied in with the lack of correlation with elevation and aspect,
may be due to the fact that often those forests that are the least degraded are found in
areas where access is limited due to the steep grade of the slopes (Kiimaird et al. 2003).
In the case of S. formosa, this may imply that the only remaining suitable habitat for this
bird may be in areas that are relatively inaccessible or that are already receiving some
sort of protection from degradation.
31
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
I Mixed Forest
IDeciduous Broadleaf Forest
I Evergreen Broadleaf Forest
IEvergreen Needleleaf Forest
ICroplands
Figure 5. Land cover analysis of Sitta formosa potential distribution.
Forested vegetation types and distribution for a) HS >80 and c) HS >50. Cropland and
forested vegetation types for b) HS >80 and d) HS >50. e) S. formosa (presence point
indicated with black symbol) proximity to cropland cover type. Extent indicated by
rectangular dashed box in d.
32
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Though there appears to be a correlation between S. formosa observations and
forested landscapes, especially at such a coarse scale, an inference of causality should not
be made without taking a closer look. Seoane et al. (2004) argue that although coarse
scale modeling can predict correlations between bird distribution and climate and
topographic ENVs, the addition of vegetation structure and/or vegetation landscape
variables may be a better predictor of causality at a more local (fine) scale. In order to
ascertain key habitat requirements for S. formosa, future iterations of predictive
distribution modeling should include a vegetation component (i.e. land cover type,
canopy closure, area, perimeter, etc.). Caution, however, should be taken when using S.
formosa museum records as species presence locations in predictive modeling, as
present-day land cover/land use derived data may not be accurate for these historic
specimens.
Caveats of predictive modeling
Knowledge of historic and present distribution of a species is important when
trying to ascertain the potential distribution of a species and make predictions for future
management planning. Museum records can date back for more than a century, as is true
of S. formosa presence locations, and are often the only resource available for
determining a species’ historic distribution and predicting its potential distribution. They
are, however, not always the most reliable. Segurado and Araujo (2004) conducted an
evaluation of several methods for predicting the probability of species occurrence and
errors associated with these predictive models. They argued that model performance was
dependent on the types of ENVs and distribution of the species being modeled. In
33
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
particular, they suggested that the spatial and environmental distribution of a species can
have a substantial effect on model performance, and that when limited records of
presence alone are available, ENFA is a robust method for modeling species
distributions. Quality of ENVs and a species distribution on those ENVs may also affect
a model’s sensitivity and performance. Imprecise locations of species occurrence may
influence model outcome when used at too fine of a resolution or with changes in scale. It
should be noted that this initial prediction of S. formosa species distribution and habitat
suitability was done at a coarse scale, and future iterations should take into consideration
the accuracy and reliability of historic as well as more recent S. formosa presence
locations.
Errors of omission and commission are also inherent in predictive distribution
modeling (Fielding and Bell 1998, Peterson 2001). First, an error of omission is the
failure of a model to include all ecological conditions and extents in which the species is
able to maintain a population. S. formosa was found in areas considered to be unsuitable
(HS < 50) by our ENFA model for 5% of initial presence observations. Pulliam (2000), in
a review of applications of Hutchinson’s ecological niche concept, discussed a variety of
factors that influence the observed relationship between species distributions and the
availability of suitable habitat. In particular, he suggested that a variety of conditions
exist (i. e., limits to dispersal, source-sink, metapopulations, competitive interaction) in
which a species may be found, and even be common, in habitat predicted to be unsuitable
when modeling a species ecological niche. Pullium concluded that rigorous determination
of habitat suitability should be conducted under field conditions in order to verify the
accuracy of model predictions. Second, an error of commission occurs when areas
34
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
predicted to be suitable are not actually inhabited by the species. Peterson (2001)
suggests two reasons for this error of commission: 1) combinations of ecological
conditions are not actually suitable, or 2) conditions are suitable, but historical factors,
colonization or dispersal ability, predation or interspecific competition have led to
suitable habitat being uninhabited by the species. Any one or several of these conditions
could apply to S. formosa throughout south and southeast Asia, and it is still very
possible that future sampling efforts in areas where resource assessment is just beginning
to take place will identify additional regions presently occupied by this species.
If historic as well as current habitat degradation events have the potential to lead
to a decline in a species’ population viability, then now is the time to consider a more
rigorous management plan for S. formosa. Birdlife (2003) provides a brief summary of
currently protected areas (in specific areas of India, Bhutan, Laos, Thailand, Vietnam and
potential areas of Myanmar) and gives suggestions for improving the protection within
those areas as well as broadening them to include adjacent areas of forest. They also
strongly suggest that additional surveys and population estimates are necessary
throughout its range. The predictive distribution maps presented here illustrate areas of
potentially viable habitat that may already or could potentially be home to this elusive
nuthatch. Predictive distribution models inherently contain prediction errors and biases;
however, when little information is available for a species of concern and time is of the
utmost importance, this method can be a valuable tool to help focus research efforts.
This ENFA model provided a preliminary analysis of S. formosa habitat requirements
and potential distribution, as well as an assessment of loss of historic habitat. These
35
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
results should be coupled with rigorous field studies of S. formosa ecology, life history
and behavior for the effective conservation of S. formosa.
36
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
LITERATURE CITED
Ali, S. 1983. A Pictoral Guide to the Birds o f the Indian Subcontinent. Oxford
University Press. 177pp.
Ali, S. and S. D. Ripley. 1973. Handbook o f the Birds o f India and Pakistan. New Delhi:
Oxford University Press.
Aspinall, R. and N. Veitch. 1993. Habitat mapping from satellite imagery and wildlife
survey data using a Bayesian modeling procedure in a GIS. Photogrammetric
Engineering and Remote Sensing 59: 537-543.
Bailey, R.G. and H.C. Hogg. 1986. A world ecoregions map for resource reporting.
Environmental Conservation, Vol. 13, No. 3, pp. 195-202.
Belward, A. S., ed. 1996. The IGBP-DIS global 1 km land cover data set (DISCover)
Proposal and implementation plans. IGBP-DIS Working Paper No. 13.
International Geosphere-Biosphere Programme Data and Information Services,
Toulouse, France. Downloaded at http://edcsnsI7.er.usgs.gov/glcc/.
Bibby, C. 1995. A global view of priorities for bird conservation: a summary. Ibis, 137.
Bierregaard, R. O., Jr. and T. E. Lovejoy. 1989. Effects of forest fragmentation on
Amazonian understory bird communities. Acta Amazonica 19: 215-241.
BirdLife International 2001. Threatened Birds o f Asia: The BirdLife International Red
Data Book. Cambridge, UK. http://www.rdb.or.id/detailbird.pbp?id=197
BirdLife International 2003. Threatened Birds o f Asia: The BirdLife International Red
Data Book. Cambridge, UK. bttp://www.rdb.or.id/detailbird.pbp?id=197
Boyce, M. S., P. R. Vernier, S. E. Nielsen and F. K. A. Scbmiegelow. 2002. Evaluating
resource selection functions. Ecological Modelling 157: 281-300.
Carroll, C., W. J. Zielinski and R. F. Noss. 1999. Using presence-absenee data to build
and test spatial habitat models for the fisher in the Klamath Region, U.S.A.
Conservation Biology 13: 1344-1359.
Castelletta, M., N. S. Sodbi and R. Subaraj. 2000. Heavy extinctions of forest avifauna in
Singapore: lessons for biodiversity conservation in Southeast Asia. Conservation
Biology 14: 1870-1880.
Clark, J. D., J. E. Dunn and K. G. Smith. 1993. A multivariate model of female black
bear habitat use for a geographic information system. Journal o f Wildlife
Management 51: 519-526.
37
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Collar, N. J., M. J. Crosby and A. J. Stattersfield. 1994. Birds to Watch 2: The World List
o f Threatened Birds. Cambridge, UK: BirdLife International (BirdLife
Conservation Series 4).
Corsi F., E. Duprè and L. Boitani. 1999. A large-scale model of wolf distribution in Italy
for conservation planning. Conservation Biology 13: 150-159.
Dettki, H., R. Lofstrand and L. Edenius. 2003. Modeling habitat suitability for moose in
coastal northern Sweden: empirical vs. process-oriented approaches. AMBIO
32(8): 549-556.
Earle, C. J. 1997. Fokienia hodginsii. The Gymnosperm Database, Department of
Botany, Rheinische Friedrich-Wilhelms-Universitat Bonn, Germany.
http://www.botanik.uni-bonn.de/conifers. Downloaded Mareh 15, 2004.
Eastman, R. 2004. IDRISI Kilimanjaro version 14.02. Clark Labs, Worcester, MA.
Fa, J.E. and L. M. Morales. 1993. Patrones de diversidad de mamiferos de México. In
T.P. Ramamoorthy, R. Bye, A. Lot, and J. Fa, editors, Diversidad Biologiea de
México. Origenes y Distribucion. Instituto de Biologia, UN AM. pp. 315-354.
FAO. 2003. State of Forestry in Asia and the Pacific-2003. Status, Changes and Trends.
Asia-Pacific Forestry Commission.
Feria, T. P. and A. T. Peterson. 2002. Using point occurrence data and inferential
algorithms to predict local communities of birds. Diversity and Distributions 8:
49-56.
Fielding, A. H. and J. F. Bell. 1997. A review of methods for the assessment of prediction
errors in conservation presence/absence models. Environmental Conservation 24:
38-49.
Grimmett, R., C. Inskipp and T. Inskipp. 1998 Birds o f the Indian subcontinent. London:
A. & C. Black/Christopher Helm. As summarized in Birdlife International, 2001.
Guisan, A. and N. E. Zimmermann. 2000. Predictive habitat distribution models in
ecology. Ecological Modeling 135: 147-186.
Harrap, S. and D. Quinn. 1995. Chickadees, Tits, Nuthatches and Treecreepers. Princeton
University Press. 464pp.
Hijmans, R.J., S.E. Cameron, J.L. Parra, P.G. Jones and A. Jarvis, 2004. The WorldClim
interpolated global terrestrial climate surfaces. Version 1.3.
http://biogeo.berkeley.edu.
38
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Hirzel, A., J. Hausser and N. Perrin. 2004. Biomapper 3.1. Lab of Conservation Biology,
Department of Ecology and Evolution, University of Lausanne.
http://www.imil.ch/biomapper.
Hirzel A.H., J. Hausser, D. Chessel andN. Perrin. 2002. Ecological-niche factor analysis:
How to compute habitat-suitability maps without absence data? Ecology 83:
2027-2036.
Hopkin, P. J. 1989. Beautiful Nuthatch (Sitta formosa)'. a species new to Thailand.
Natural History Bulletin o f the Siam Society 37(1): 105-107.
Hutchinson, G. E. 1957. Concluding remarks. Cold Spring Harbour Symposium on
Quantitative Biology 22: 415-427.
lUCN 2003. 2003 lUCN Red List o f Threatened Species. Gland, Switzerland and
Cambridge, U. K.
King, B. F. & E. C. Dickinson. 1975. A Field Guide to the Birds of South-east Asia.
Houghton Mifflin, Boston.
Kinnaird, F.M., E. W. Sanderson, T. G. O’Brien, H. T. Wibisono, and G. Woolmer. 2003.
Deforestation trends in a tropical landscape and implications for endangered large
mammals. Conservation Biology 17(1): 245-257.
Lenton, S. M., J. E. Fa and J. P. D. Val. 2000. A simple non-parametric GIS model for
predicting species distribution: endemic birds in Bioko Island, West Africa.
Biodiversity and Conservation 9: 869-885.
Matthysen, E. 1998. The Nuthatches. Academic Press. San Diego, CA. 315pp.
Miller, R.I. (ed.) 1994. Mapping the Diversity o f Nature. Chapman and Hall. London,
U.K.
Mladenoff, D.J., T.A. Sickley, R.G. Haight and A.P. Wydeven. 1995. A regional
landscape analysis and prediction of favorable gray wolf habitat in the Northern
Great Lakes Region. Conservation Biology 9(2): 279-294.
Novaeek, M. J. and E. E. Cleland. 2001. The current biodiversity extinction event:
Scenarios for mitigation and recovery. Proceedings of the National Academy of
Sciences. 98(10): 5466-5470.
Osbom, T. 2004. Preparation and implementation o f a strategy fo r the management o f
Fokienia hodsinsii in Vietnam by 2008. A desk study for the Hoang Lien Son
Project, Fauna and Flora International Vietnam.
39
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Osborne, P. E., J. C. Alonso and R. G. Bryant. 2001. Modeling landseape-seale habitat
use using GIS and remote sensing: a case study with great bustards. Journal o f
Applied Ecology 38: 458-471.
Patthey, P. 2003. Habitat and corridor selection of an expanding red deer (Cervus
elaphus) population. Ph.D. dissertation. Institute of Ecology, University of
Lausanne.
Peterson, A. T. 2001. Predicting species’ geographical distributions based on ecological
niche modeling. The Condor 103: 599-605.
Peterson, A. T. and D. A. Vieglais. 2001. Predicting species invasions using ecological
niche modeling: new approaches from bioinformatics attack a pressing problem.
Bioscience 51: 363-371.
Pulliam, H. R. 2000. On the relationship between niche and distribution. Ecology Letters
3: 349-361.
Reutter, B. A., V. Heifer, A. H. Hirzel, and P. Vogel. 2003. Modelling habitat-suitability
using museum collections: an example with three sympatric Apodemus species
from the Alps. Journal o f Biogeography 30: 581-590.
Robson, C. 2002. Birds o f Thailand. New Holland and Princeton University Press. 272pp.
Sehrott, G. R., K. A. With and A. W. King. 2005. On the importance of landscape history
for assessing extinction risk. Ecological Applications 15(2): 493-505.
Segurado, P. and M. B. Araujo. 2004. An evaluation of methods for modeling species
distributions. Journal o f Biogeography 31: 1555-1568.
Seoane, J., J. Bustamante and R. Diaz-Delgado. 2004. Competing roles for landscape,
vegetation, topography and climate in predictive models of bird distribution.
Ecological Modelling 171: 209-222
Smythies, B. E. 1949. A reconnaissance of the N ’Mai Hka drainage, northern Burma. Ibis
91: 627-648.
Stoekwell, D.R.B. 1999. Genetic Algorithms II. 123-144pp. in A.H. Fielding (ed.).
Machine Learning Methods for Ecological Applications. Kluwer Academic
Publishers, Boston.
Stoekwell, D.R.B. and I.R. Noble. 1992. Induction of sets of rules from animal
distribution data: a robust and informative method of data analysis. Mathematics
and Computers in Simulation 33: 385-390.
40
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Stoekwell, D.R.B. and D. Peters. 1999. The GARP Modeling System; problems and
solutions to automated spatial prediction. International Journal o f Geographical
Information Science 13(2): 143-158.
Stouffer, P. C. and R. O. Bieregaard, Jr. 1995. Use of Amazonian forest fragments by
understory insectivorous birds. Ecology 76(8): 2429-2445.
Tbewlis, R. M., R. J. Timmins, T. D. Evans and J. W. Duckworth. 1998. The
conservation status of birds in Laos. Bird Conservation International 8 (SuppL):
1-159.
Tobias, J. 1997. Environmental and social action plan for the Nakai-Nam Tbeun
catchment and eorridor areas. Report of the wildlife survey. Vientiane: Wildlife
Conservation Society.
Tordoff, A. W., Le Manb Hung, Nguyen Quang Truong and S. R. Swan. 2002. A rapid
field survey of Van Ban District, Lao Cai Province, Vietnam. Unpublished report
to the BirdLife International Vietnam Programme.
Turner, I. M. 1996. Speeies loss in fragments of tropical rain forests: a review of the
evidence. Journal o f Applied Ecology 33: 200-209.
U.S. Geologieal Survey. 1998. Global Land Cover Cbaracteristies Data Base, version 2.0.
Website : http ://edcsns 17.er.usgs.gov/glee/.
U.S. Geological Survey. 2002. Hydro Ik Elevation Derivative Data Base. Website:
http ://ededaae .usgs.gov/gtopo3 0/bydro/.
Vogel, C.J., P.R. Sweet, M.H. Le, and M.M. Hurley. 2003. Ornithological records from
Ha Giang province, northeast Vietnam, during March-June 2000. Forktail 19: 2130.
41
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.