GIS 5306 GIS in Environmental Systems

GIS 5306 GIS in Environmental Systems
Week 2: Some more Theory, then Rudimentary
Species Distribution Modeling: Habitat Suitability
Models and Logistic Regression
Last Time
• Intro and Logistics
• Main Questions
– How are organisms distributed on the surface of the Earth, and why?
– As Earth’s environment changes (climate, land use, geomorphologic, geologic,
biologic), how will the distribution of organisms change?
• Species Distribution Maps: Basic unit of Biogeography
Dot maps
Outline maps
Contour maps
• Physical Factors (Abiotic Environment) of species distributions
• Ecological Foundations of species distributions
Population growth – intraspecies interactions
Interspecies interactions
Dispersal, extinction not dealt with
First: Project Options
1. Either a) the class or b) each student (or
student team) will develop a research
question (or questions).
2. Each student will argue for a taxon or group
of taxa to study, including the study area
(Florida gets highest priority?).
3. When 1&2 are decided, Each student will
then learn, evaluate, and teach one or more
methods for studying the taxa to answer the
research question to the class.
Option b
• When the methods have been examined, the
class will decide the best methods and data
required to answer the original research
• Then the whole class will conduct various
aspects of the analysis (comparison of
multiple methods?).
• Then we will write the paper, and submit it to
an appropriate journal.
Pre-project exploration
• Each student (or team of 2 depending on the
final size of the class) will become familiar
with one or two of the methods for species
distribution modeling, then present the
theory of the method, the software for the
method, and an example “laboratory exercise”
to the class, including providing data for the
exercise. This can be the tutorial produced by
the developer of the software, or an exercise
developed by the student(s).
Work Assignments
• Instructor will conduct next week’s class:
– Theory of predictive vegetation mapping (subset of
Species Distribution Modeling)
– Selection and description of one method for
predictive vegetation mapping (Habitat Suitability)
– Tutorial for working through a Habitat Suitability
– Provision of data for working up several examples.
• Read the assigned papers linked on the syllabus.
Methods from which to Choose (See Class References)
Habitat Suitability Modeling (Binford)
Gradient Analysis (Ordination) methods (surprisingly not a major part of the literature)
Logistic Regression
Fuzzy Set Approaches (also more seen as part of habitat suitability modeling)
Biodiversity Informatics Facility – compilation of multiple software sites:
DMAP – distribution mapping software
R-forge contributions for Species Distribution Modeling
Openmodeller (multiple methods)
MAXENT (Tutorial Available)
Land-change Modeler for IDRISI and ArcGIS
Distribution of Species – Main Points
1. Individuals of each species have ecological
requirements and limits that, along with historical
factors, determine distribution.
2. Physical environmental factors create gradients of
tolerance and optimality for organisms.
1. E.g. water, light, pH, temperature, salinity, etc.
3. Biotic environment (other organisms) also create
gradients of tolerance and optimality.
1. Population growth (intraspecific interactions)
2. External (predation, competition, mutualism)
3. Internal (pathogens, parasites)
New Topic: Distribution of Species
– Secondary Points
1. Interactions of physical and biotic factors
always occur.
2. Very difficult to demonstrate mechanisms of
distribution limitation, but straightforward to
show correlations among physical and biotic
factors and distributions.
The Niche
Also implies that there can be multiple,
or combinations of limiting factors
All Models Follow the Same Principle
Mapping from species and environmental factor
distribution (Geographic Space), modeled in
Environmental Space, and re-mapped into Geographic
Figures from Elith and Leathwick 2009
Another Way to Look at it.
Candidate variables
Species occurrence data
Environmental Data
Modeling Framework
Map of predicted occurrence
Redrawn from Franklin, J. 2009. Mapping Species Distributions. Cambridge.
This Time
• A little more general theory
• Binford: Rudimentary SDM.
– Habitat Suitability/Cartographic Overlay
• National Fish and Wildlife Service/Agency
• Wildlife-habitat relationship (WHR) models
• GAP Analysis Project
– Workshop: Swallow-tailed Kite in North-central
and northwest Florida.
Predictive vegetation mapping:
Franklin 1995
• Fundamental paper reviewing practice to 1994;
30 papers!
• Principles of Predictive Species Distribution
Modeling (PSDM)
– Predictive Vegetation Mapping
– Habitat Modeling
• Dependence of predictive vegetation mapping on
ecological niche theory and gradient analysis
Cited by 363 more recent papers as of 30 August 2010
Definition: Predictive vegetation
mapping by Climate Envelopes
• Predicting the vegetation composition across
a landscape from mapped environmental
Bald Cypress
Longleaf Pine
Live Oak
Longleaf Pine – Niche Graph and Importance
Value Map
Gradient Analysis and Continuum
Concept (Whittaker 1951)
• Whittaker 1973
continuum concept more explicitly puts forth
hypotheses about species response functions
(curves) to environmental gradients, e.g., that
they are Gaussian.
Predictive vegetation mapping:
• Always starts with the development of some
type of model, followed by the application of
that model to a geographic database to
produce the predictive map, a realization of
the model.
Franklin 1995
Foundations and Premises
• Predictive vegetation mapping is founded in
ecological niche theory and vegetation
gradient analysis.
• Premise: vegetation distribution can be
predicted from the spatial distribution of
environmental variables that correlate with or
control plant distributions.
Franklin 1995
• Further, maps of the environmental variables
or their surrogates must be available, or easier
to map than the vegetation itself, in order for
predictive vegetation mapping to be a
practical or informative exercise.
Franklin 1995
Model Foundations
• In order to extrapolate over space (predictive
vegetation mapping) or time (vegetation
change modeling), direct gradients or their
surrogates must be mapped (temperature,
potential solar radiation, precipitation, soilmoisture availability, geology or soil
Spatial Focus
• Focus on the prediction of plant species
distributions or vegetation patterns at the
'regional' scale, e.g., where the mapped
extent of the predictions are generally at or
within the biogeographic range of the
dominant plant species.
Franklin 1995
Models from Franklin 1995
Statistical Methods
– Regression
Logistic (Logit)
– Baysian
Maximum Likelihood Classification
Rule-based Methods
Multivariate Methods
– Discriminant analysis
– Canonical correlation
Classification Trees
Neural Networks
Decision-tree Classification
Genetic Algorithms
Potential Vegetation vs. Actual
• Figure 1 Conceptual model showing
the relationship between direct
gradients (nutrients, moisture,
temperature), their environmental
determinants (climate, geology,
topography) and potential natural
vegetation, and the processes that
mediate between the potential and
actual vegetation cover (the latter is
sensed by a remote sensing device).
• Franklin 1995 page 479
Austin, M.P. 2002. Spatial prediction of species
distribution: an interface between ecological
theory and statistical modelling. Ecological
Modelling 157:101-118.
• Cited by 495
Rudimentary SDM: Habitat
Suitability/Cartographic Overlay
• Goes back to McHarg 1969: Design with Nature
Suitability for
Ski Areas
Not drinking water
Slope between 5% and 75%; Aspect N, W, E
Soils not erodible, not high runoff
Vegetation: grass, degraded forest
Biodiversity low
Areas suitable for ski areas
Site Suitability
Analysis, or
This project aimed at
identifying the best area
suitable for development
of a ski resort in Mitchell
and Yancey Counties in NC.
The following factors were
included in the suitability
analysis: Land cover,
access, snow precipitation,
land ownership, elevation,
aspect and slope.
Areas Suitable for Ski Resort Development
McHargian analysis: Woodlands
Works for SDM/Habitat Modeling, too
• What are habitat requirements for
– Single species
– Multiple species
– Vegetation communities, assemblages
– Biodiversity
– Dot maps vs. area maps. Fine vs. coarse scale
– Depends on question!
Animal Species Distribution Modeling – Habitat
Suitability/Wildlife-Habitat Relationship (WHR) Models
• GAP Analysis Project: “Keeping Common Species Common”
• Where are the gaps in biodiversity protection?
Research/Management Question
• How well are we protecting common plants
and animals?
– Corollary: if we are not protecting them well,
what can we do about it?
• Land-use planning
• Conservation purchases and easements
• Originally not “common species” but
GAP Procedure
• 1. Map LAND COVER of the dominant
ecological systems
• 2. Map and model SPECIES ranges and
• 3. Map land STEWARDSHIP
• 4. Conduct the ANALYSIS
• All from
GAP Procedures
Note WHRM is an
database: what
animals would be
expected to be
found in vegetation
alliance (habitat).
From Complete GAP Handbook available from
Florida GAP
Data available at FGDL data repository
Definition: Habitat Modeling
• Habitat
– where and animal lives
– the living and non-living characteristics of a landscape
that an animal uses
– what animals need to survive and reproduce
• Different kinds of habitat: Food, water, hiding
cover (prey) or ambush cover (predators),
thermal cover (against heat or cold or both), and
nest sites (or other special needs for
reproduction), the minimum amounts and spatial
arrangement of the first 5 components
Habitat Modeling
• Expert opinion
• Literature
• Compilation into database
– What should occur where
– Absences difficult to model
• Database query
– What collection of vertebrates should be in what
vegetation alliances?
– Calculate biodiversity hotspots
• Stewardship determination
– What is not already protected?
– Considers management objectives of public agencies
Gaps Identified!
• Now what?,_Data,_&_Reports/Find_Upd
Animal Species Distribution Models –
Habitat Suitability Indices
• US Fish and Wildlife Service
Habitat Suitability Modeling; WildlifeHabitat Relationship Models: Theory
• Habitat suitability index (HSI) model is intended for use
with the habitat evaluation procedures (HEP) developed by
the U.S. Fish and Wildlife Service (1980) for impact
assessment and habitat management. The model was
developed from a review and synthesis of existing
information, and includes unpublished information that
reflects the opinions of persons familiar with blackshouldered kite ecology. It is scaled to produce an index of
habitat suitability between 0 (unsuitable habitat) and 1.0
(optimally suitable habitat).
• This model is a hypothesis of species-habitat relations, not
a statement of proven cause and effect. The model has not
been field tested.
HIS: Theory – All from literature or
• Distribution and commercial importance
• Life history overview
• Habitat requirements
– Food for both adults and immatures (and
associated habitat)
– Cover for both adults and immatures
– Breeding habitat (e.g. nesting) if different
• Geographic area and season
• Minimum habitat area
Example Habitat Suitability Index: American
Alligator (Alligator mississippiensis)
• The American alligator is characteristically a resident of river swamps,
lakes, bayous, and marshes of the Gulf and Lower Atlantic Coastal Plains
from Texas to North Carolina.
• HSI publications have standard format:
Distribution and Commercial Importance
Life History Overview
Model Applicability
Model Description
Suitability Index (SI) Graphs for Model Variables
Component Index (CI) Equations and HSI Determination
Field Use of Model
Interpreting Model Outputs
American Alligator HSI Model
American Alligator HSI Model: “Data”
American Alligator HSI Model: “Data”
American Alligator HSI Model: “Data”
American Alligator HSI Model
If all of these components can be
mapped in a GIS, then the map of
habitat suitable for alligators can be
Alligator HSI
American Alligator HSI Model
If all of these components can be
mapped in a GIS, then the map of
habitat suitable for alligators can be
Several Classification Approaches
Probabilistic and Fuzzy
Classification: An area
has a XX% probability
of belonging to the xxxxx
set of land cover, or a
membership value of XX
Standard Classification:
An area either is or is
not a member of the set
of xxxxx land cover.
Source: Hill, K.E. 1997. The Representation of Categorical Ambiguity: A Comparison of Fuzzy,
Probabilistic, Boolean, and Index Approaches in Suitability Analysis. Dissertation, Harvard University
• Model the distribution of an interesting
animal using a habitat-suitability/WHR
• Swallow-Tailed Kite Elanoides forficatus
– Seen occasionally in Florida
• Data and ArcGIS project available at
• Download the folder to your computer. It is big!
Swallow-Tailed Kite Elanoides
Diminishing population
Habitat loss, fragmentation
Formerly in 21 States
Possibly fewer than 5000 remain, with 60-65%
of the population breeding in Florida during
the summer months.
• Habitat Modeling is said to be a major longterm goal of STK conservancy
Swallow-Tailed Kite Conservation Status
• Listed as “imperiled” by The Nature Conservancy. The
predominant identified threat is the loss of suitable
nesting habitat.
• National Audubon Society’s Watchlist as a “species of
critical concern”
• Designated by the U. S. Geological Survey’s Biological
Resources Division as a “Species at Risk”
• IUCN – “Species of Least Concern”
• FL Freshwater Fish and Wildlife Conservation
Commission: “one of Florida's most vulnerable and
poorly understood species”
• Habitat Modeling is said to be a major long-term goal
of STK conservancy
Swallow-Tailed Kite Conservation Status
• The current challenges to kite conservation
include wetland loss and drainage, extensive
clear-cutting, short rotation timber harvesting,
and significant land use changes along
migration routes and wintering habitats in
South America (Gruber 2009).
Swallow-Tailed Kite Elanoides
From Gruber, 2009
The disappearance of this species from three-fourths of its breeding range between 1880 and 1910 (Figure 1)
was one of the most dramatic range contractions of any bird species before the highly publicized post-WWII
Peregrine falcon crash (Cely 2005).
Swallow-Tailed Kite Elanoides
forficatus: Natural
• Predator
– Insects (airborne)
– Herps
– Small Mammals
– Small birds (nests!)
• Neo-tropical Migrant
• Breeds in SE US, winters
in South America
Gruber 2009
Habitat Requirements (Gruber 2009):
Different Scale than Area Map
• Riparian and bottomland forest
• Mixed pine
• Strong preference for nesting in dominant or co-dominant
loblolly pine (Pinus taeda) stands growing near or within
wetlands (Cely 2005).
• The physical structure of the forest stand is perhaps more
important than the specific vegetation communities.
• Tall, easily accessible nest trees near open areas that
provide sufficient prey (Meyer 1995).
• Nests are typically built at the top of the tallest trees in the
stand; the preferred surrounding stand is usually low
density and has an uneven height/age structure (Meyer
Habitat Requirements (Gruber 2009)
• Kites have large home ranges encompassing thousands
of acres, and will often commute long distances, up to
24 km, from the nest site to forage.
• The main food source of kites is large insects caught on
the wing; a variety of other prey such as snakes,
anoles, frogs, nestling birds, and wasps nests, are
gleaned from vegetation.
• Large communal roosts near nesting areas are
• Social nesting behavior: clustered distribution.
– May restrict dispersal into unused habitat.
Ridgely, R. S., T. F. Allnutt, T. Brooks, D. K. McNicol, D. W. Mehlman, B. E. Young, and J. R.
Zook. 2007. Digital Distribution Maps of the Birds of the Western Hemisphere, version 3.0.
NatureServe, Arlington, Virginia, USA.
Data: Local Data
• and others
– Habitat (vegetation, land-cover)
– Topography (elevation, slope, aspect)
– Soils (SSURGO)
– Hydrography (all water bodies including wetlands)
• Species data: difficult to come by.
– Found sightings in SPECIES_OBS_APR10, data
available from FGDL
Habitat Model
• The first component of the habitat model was
to identify land cover types most likely to be
used as nesting sites (Beyeler 2008). Nest
points were buffered by 1000 m, and all land
use/land cover types found within those 3.14
km2 buffers were identified as potentially
suitable for nesting (Table 1)
(Gruber 2009)
Water Management District
Classification: FLUCCS Code – Suitable
Cely, J.E. 2005. Swallow-tailed Kite (Elanoides forficatus) Fact Sheet. South Carolina Department
of Natural Resources.
Gruber, J. 2009. Targeting Potential Conservation Sites for Swallow-tailed Kites (Elanoides
forficatus) in Levy County, Florida. Unpublished M.E.M. thesls. Duke University
Meyer, K. D, and M. W Collopy. 1990. Status, distribution, and habitat requirements of the
American Swallow-tailed Kite (Elanoides forficatus) in Florida. Final report. Florida Nongame
Wildlife Program, Florida Game and Fresh Water Fish Commission, Tallahassee.
Meyer, K.D. 1995. Swallow-tailed Kite (Elanoides forficatus). In A. Poole and F. Gill, editors. The
Birds of North America, Number 138. Academy of Natural Science, Philadelphia, PA, and
American Ornithologists Union, Washington, D.C., USA.
Ridgely, R. S., T. F. Allnutt, T. Brooks, D. K. McNicol, D. W. Mehlman, B. E. Young, and J. R. Zook.
2007. Digital Distribution Maps of the Birds of the Western Hemisphere, version 3.0.
NatureServe, Arlington, Virginia, USA.
Sykes Jr, P. W, C. B Kepler, K. L Litzenberger, H. R Sansing, E. T.R Lewis, and J. S Hatfield. 1999.
Density and Habitat of Breeding Swallow-Tailed Kites in the Lower Suwannee Ecosystem,
Florida (Densidad y Habitat Reproductivo de Elanoides forficatus en la Parte Inferior del
Ecosistema Suwanee, Florida). Journal of Field Ornithology 70, no. 3: 321–336.
Wright, M. H, R. O Green, and N. D Reed. 1970. A collection of observations and field notes on the
nesting activities of the swallow-tailed kite (Elanoides forficatus) in the Everglades National

similar documents