Transcript Species Distribution Models

Strategic Habitat Conservation:
Modeling to support cooperative,
adaptive, science-based
management
USGS-USFWS Science Support Partnership
Ashton Drew
Outline
 SSP project context & objectives
 Building a tool to meet SHC science and
management objectives
 Species-habitat modeling approach
 Future directions
SSP & SHC Challenge
► Move
from static to dynamic thinking
regarding how you collect, summarize,
utilize, and share data…




Scaling: stepping-down & stepping-up
Communicating: science & management
Modeling: general (what) & specific (where)
Management: acting & monitoring
SHC Highlights
Selecting species suitable for modeling
► Maximizing benefits from existing data &
expertise
► Knowledge summary & communication tools
►
►
Conservation Design
►
Biological Planning
►
Hypotheses &
sampling design
based on ecological
assumptions and
►
predicted
management
outcomes
►
Regular maintenance
of GIS and biological
data layers
Delivery of
Temporal cautionary
Conservation Actions
note
► Multiple scales, on and off refuge lands
► Must be documented in a GIS
Monitoring & Research
►
Decision support tools
to evaluate alternative
actions
Integration of value
systems into ecological
model
Decisions based on
available science with
documented
assumptions and
alternatives considered
Pilot Project Objective
► Aid
with step-down of national
population & habitat objectives
Partners in Flight 2004
National Goals
Bachman’s sparrow (250,000) – Increase 100%
Brown-headed nuthatch (1.5 mil) – Increase 50%
Ecosystem?
National Wildlife Refuges?
Other protected lands?
Errol Taskin
www.birdsource.org
Management Context & Priorities
► State
and refuge level planning
documents
 Reference national and international
plans
 Set management priorities in
ecosystem context
 Partnership for coordinated
management in time and space
 Shift from few to many species and
habitats
► Quantitative
success
goals & measures of
RTNCF Pilot Model Guidelines
►
►
►
►
►
Two spatial scales
Terrestrial & aquatic
species
Data-rich & data-poor
(expert opinion) scenarios
Start with GAP products
Design for adaptive
management use
Bayesian Approach?
Starting on the same page...
►
►
►
►
►
Set population objectives for species
Set abundance goals for RTNCF natural communities
Convert population/abundance objectives into habitat
objectives
Map potential conservation areas where deficits exist
Step down population/abundance objectives to individual
refuges and partner lands
What do managers want?
&
What can a model provide?
&
What are the objectives of SHC?
Starting on the same page...
►
►
►
►
►
Set population objectives for species
Set abundance goals for RTNCF natural communities
Convert population/abundance objectives into habitat
objectives
Map potential conservation areas where deficits exist
Step down population/abundance objectives to individual
refuges and partner lands
Models don’t set targets...
People do!
Starting on the same page...
►
►
►
►
►
Set population objectives for species
Set abundance goals for RTNCF natural communities
Convert population/abundance objectives into habitat
objectives
Map potential conservation areas where deficits exist
Step down population/abundance objectives to individual
refuges and partner lands
Managers starts with national goals...
Modeling starts with local knowledge
Starting on the same page...
►
►
►
►
►
Set population objectives for species
Set abundance goals for RTNCF natural communities
Convert population/abundance objectives into habitat
objectives
Map potential conservation areas where deficits exist
Step down population/abundance objectives to individual
refuges and partner lands
Is habitat acquisition the only management
action under consideration?
Starting on the same page...
►
►
►
►
►
Set population objectives for species
Set abundance goals for RTNCF natural communities
Convert population/abundance objectives into habitat
objectives
Map potential conservation areas where deficits exist
Step down population/abundance objectives to individual
refuges and partner lands
Single descriptive outcome
= knowledge communication tool
Multiple predictive outcomes
= predictive decision support tool
Starting on the same page...
►
►
►
►
►
Set population objectives for species
Set abundance goals for RTNCF natural communities
Convert population/abundance objectives into habitat
objectives
Map potential conservation areas where deficits exist
Step down population/abundance objectives to individual
refuges and partner lands
STATIC vs. DYNAMIC OBJECTIVES
►
►
►
►
Quantify refuge contributions to populations and habitats
Identify where and how refuge-scale management actions
may contribute to regional objectives
Identify where and what additional research would be most
beneficial
Coordinate activities with partner agencies’ managers to
step-down objectives and track regional progress
Ecological “Step-down”
Policy
Guidelines
SPACE
Strategic Land
Use Plans
Refuge
Management
Plans
TIME
Ecological “Step-down”
Policy
Guidelines
Biogeographic
Range
SPACE
Strategic Land
Use Plans Habitat
Refuge
Management
Plans
Distribution in
Regional
Landscape
Patchy
Resources
within Habitat
TIME
Knowledge & Assumptions
Vary with Scale
Policy
Guidelines
Biogeographic
Range
SPACE
Strategic Land
Use Plans Habitat
Refuge
Management
Plans
Distribution in
Regional
Landscape
Patchy
Resources
within Habitat
TIME
Good GIS data
sources, limited
knowledge
Knowledge & Assumptions
Vary with Scale
SPACE
Policy
Guidelines
Reasonable
knowledge,
limited GIS
Refuge
Management
Plans
Biogeographic
Range
Strategic Land
Use Plans Habitat
Distribution in
Regional
Landscape
Patchy
Resources
within Habitat
TIME
Effective Knowledge Transfer
(Perera et al. 2007)
Policy
Guidelines
Biogeographic
Range
SPACE
Strategic Land
Use Plans Habitat
Refuge
Management
Plans
Distribution in
Regional
Landscape
Patchy
Resources
within Habitat
TIME
Species-Habitat Model
Model habitat location and
quality based on expert opinion
and literature review
Field validation
& model updating
Validated and updated
habitat model
Amount of habitat, Number of individuals
(total, % protected, spatially-explicit)
Significant sources of uncertainty
Species-Habitat Model
Model habitat location and
quality based on expert opinion
and literature review
Field validation
& model updating
Validated and updated
habitat model
Decision-Support Extension
Management
Scenarios
Science
Scenarios
Action Set
A vs. B
Hypothesis Set
A vs. B
Model habitat & population under
alternate scenarios
Evaluate costs & risks to
compare value
Perform selected management
action or research
Amount of habitat, Number of individuals
(total, % protected, spatially-explicit)
Significant sources of uncertainty
Species-Habitat Model
King Rail
Rallus elegans
Coarse Scale Habitat Models
► SE
GAP provides Potential Occurrence in SE region
King Rail live in
Fresh or Brackish
Marsh Habitat
(red)
Refuge-level Habitat Variability
King Rail
Rallus elegans
Bayesian Modeling Approach
Prob (
)
Prior Probability
(Model)
Likelihood
(Data)
400m grid cells containing GAP
potential King Rail habitat
Posterior Probability
(Model given the Data)
Bayesian Belief Network
Prob (
►
►
)
P (detect KIRA) varies within GAP predicted habitat
Variables from literature and experts
Bayesian Belief Network
Prob (
Occurrence
►
►
)
Foraging
Courting
Brooding
Wintering
Occurrence patterns depend on activity and time of year
**Availability for detection varies by activity and time of year
Bayesian Belief Network
Prob (
Occurrence
Habitat
►
►
)
Foraging
Landcover
Courting
Brooding
Distance to
Open Water
Wintering
Water Depth
Hierarchical habitat selection: macro and microhabitat
Limited GIS data at relevant temporal & spatial scale
Bayesian Belief Network
Prob (
Occurrence
Habitat
►
)
Foraging
Landcover
Courting
Brooding
Distance to
Open Water
Wintering
Water Depth
Relationships from literature and expert opinion
Bayesian Belief Network
Prob (
Occurrence
Habitat
Management
Choices
►
►
)
Foraging
Courting
Landcover
Burning
Brooding
Distance to
Open Water
Flooding
Wintering
Water Depth
Acquisition
Restoration
Management choices influence occurrence patterns via habitat
Again, choices occur at multiple scales
Bayesian Belief Network
Prob (
Occurence
Habitat
Management
Choices
►
►
)
Decision
Foraging
Courting
Landcover
Burning
Brooding
Distance to
Open Water
Flooding
Wintering
Water Depth
Acquisition
Restoration
Manager defines potential habitat management actions
Manager decides how to act in given situation based on
probability and uncertainty associated with probability
Model Validation & Monitoring
Prob (
)
►
…depends on:
 patch size, cell context, distance
from open water, salinity, water
depth
Stratify survey on GIS
relevant assumptions
► Checking for ommission &
commission
► Collect microhabitat to
distinguish false assumptions
from inadequate data
►
400m grid cells containing GAP
potential King Rail habitat
Science – Management Feedback
All SEGAP Marsh Patches
►
SEGAP Marsh Patches >1 acre
Experts all suspect a minimum patch size, but disagree
about how small is “too small”
Science – Management Feedback
All SEGAP Marsh Patches
►
►
SEGAP Marsh Patches >1 acre
Source of uncertainty in population and habitat estimates
Uncertainty passes to management decisions
Science – Management Feedback
All SEGAP Marsh Patches
►
►
SEGAP Marsh Patches >1 acre
Take management action based on knowledge
Select monitoring sites to test patch size hypothesis that
underlies action
Pilot Project Models
vs.
“The Real Thing”
Future Directions?
► Five





things I can’t deliver (by June 2009)…
pretty GUI interface
interactive decision support
multi-year predictions
population viability assessment
GIS to track management actions
► …but
all are feasible additions to the
framework I am developing
Pilot Model Species
►
King Rail
 USFWS Focal Species
 Fresh & brackish wetlands
 Back Bay, Cedar Island, Currituck,
MacKay Island, Pea Island, &
Swanquarter
►
Swainson’s Warbler
 PIF Priority Species
 Bottomland & upland hardwood forest
 Alligator River, Great Dismal Swamp,
Pocosin Lakes, Roanoke River
►
Blueback Herring
 NOAA Species of Concern
 Anadromous fish
 Roanoke River, Alligator River
Modeling Method to Support SHC
► Pilot







project to establish protocol for:
Gathering, summarizing existing data
Gathering, summarizing expert opinion
Communally constructing a belief network
Asking science and management “what-ifs”
Designing a monitoring protocol to reduce uncertainty
Updating model with new information
Recommending adjustments to management and/or
monitoring
Bayesian Belief Network
Prob (
Occurence
Habitat
Management
Choices
►
►
)
Decision
Foraging
Courting
Landcover
Burning
Brooding
Distance to
Open Water
Flooding
Wintering
Water Depth
Acquisition
Restoration
Manager defines potential habitat management actions
Manager decides how to act in given situation based on
probability and uncertainty associated with probability
Bayesian Belief Network
Prob (
Occurrence
Habitat
Management
Choices
►
►
)
Foraging
Pool/Riffle
Landcover
Decision
Spawning
Migrating
Substrate
Water
Quality
Shading
Riparian
Mgmt.
Dam
Removal
Fish
Ladder
Ecological relationships from literature and experts
Manager decides how to act in given situation based on
probability and uncertainty associated with probability
Bayesian Belief Network
Prob (
Occurrence
Habitat
Management
Choices
►
►
)
Decision
Eggs
Tadpoles
Landcover
Restoration
Shading
Hybernating
Breeding
Water
Quality
Aqcuisition
# Dry
Days
Artificial
Ponds
Ecological relationships from literature and experts
Manager decides how to act in given situation based on
probability and uncertainty associated with probability
Many Thanks To…
► GIS
Data: SE-GAP & BaSIC
► Lit Review: E. Laurent, Q. Mortell
► Expert Opinions: Anonymous (USFWS, TNC,
Natural Heritage Program, Wildlife
Resources Commission, NC Museums)
► KIRA-CAP: National cooperation on
research, modeling, and funding
► Model and Validation Funding: USGS &
USFWS
RTNCF SSP Questions:
Ashton Drew: [email protected] or 919-513-0506
Project Website: www.basic.ncsu.edu/proj/SSP.html
►
►
►
►
Quantify refuge contributions to populations and habitats
Identify where and how refuge-scale management actions
may contribute to regional objectives
Identify where and what additional research would be most
beneficial
Coordinate activities with partner agencies’ managers to
step-down objectives and track regional progress