Coordination Plan for

1

• Co-ordination Plan for
• Predictive Modeling of Marine Mammal Habitats
• Duke University
• CS-1390
• Southwest Fisheries Science Center
• CS-1391
• Dr. Patrick N. Halpin
• Nicholas School of the Environment Earth
  Sciences
• Duke University Marine Laboratory
• Brief to The Scientific Advisory Board
• October 14, 2004

2
Performers

• CS-1391
• Dr. Jay Barlow, NOAA La Jolla
• Population assessment
• Dr. Lisa Ballance, NOAA La Jolla
• Marine mammal ecology
• Dr. Karin Forney, NOAA Santa Cruz
• Population assessment
• Dr. Steve Reilly, NOAA La Jolla
• Marine mammal ecology
• Dr. Paul Fiedler, NOAA La Jolla
• Biological oceanography
• Dr. Tim Gerrodette, NOAA La Jolla
• Population assessment
• Dr. Jessica Redfern, NOAA La Jolla
• Ecological modeling
• Dr. Megan Ferguson Scripps Institution
• Ecological modeling
• Ms. Elizabeth Becker UC Santa Barbara

CS-1390 Dr. Andy Read, Duke University Marine
mammal ecology Dr. Pat Halpin, Duke University
Spatial analysis and GIS Dr. Song Xian, Duke
University Bayesian statistics Mr. Ben Best,
Duke University Geospatial analysis,
modeling Mr. Rob Schick. Duke University Geospati
al analysis, modeling Dr. Larry Crowder, Duke
University Population modeling Dr. Cisco
Werner, University of North Carolina Physical
oceanography Dr. Debi Palka, NOAA - Woods Hole
Population assessment
3
Problem Statement

• The Navy must conduct training exercises to
  maintain readiness
• Predicting the potential for encountering
  populations of marine mammals during training
  exercises may help to minimize adverse
  interactions
• There are no current analytical tools with which
  to predict the distribution or density of marine
  mammals at spatial and temporal scales of
  relevance to these exercises

4
Contrasting Approaches
CS-1390
CS-1391
5
CS-1390 Approach
(2) Remote sensing data fusion
Data
(1) Archival data mining
Information
(3) Automated modeling system
(4) Web services SDSS
6
CS-1391 Approach
(1) Marine mammal observation In situ
environmental data collection
(2) Modeling

• A. Model Design
• Separate encounter rate and group size models
• Forward/backward stepwise selection
• AIC
• B. Evaluate predictive performance
• Cross-validation
• Average squared prediction error

7
Co-ordination Plan

• Towards a Consilience of Approaches
• CS-1390 CS-1391 Habitat and density models

Density Model
Habitat Model
8
Accomplishments

• CS-1391
• Explore modeling techniques
• GLM (maximum likelihood and Bayesian), GAM, CART
• Evaluate the effects of scale on cetacean-habitat
  models
• Evaluate interpolation techniques
• Develop methods for estimating variance
• Compare models built using in situ and remotely
  sensed data
• Explore the effect of using prey data in
  cetecean-habitat models

• CS-1390
• Explore modeling techniques
• GLM, GAM, CART
• Spatio-temporal / Bayesian
• Develop an automated modeling analysis system
• Assess model robustness for cetacean Atlantic
  Gulf species
• Develop prototype tools
• Develop web-services system
• Begin development of prototype spatial decision
  support system

9
Co-ordination Plan

• Outline
• 1. Initial co-ordination planning workshop
  (June 2004)
• 2. Annual co-ordination meetings (2005, 2006
  2007)
• 3. Preparation of a joint review manuscript (In
  Press)
• 4. Ensure consistency of modeling approaches (In
  Progress)
• 5. Joint toolbox development (Planned)
• 6. Joint dissemination of SDSS to Navy end-users
  (Planned)

10
Co-ordination Plan
Initial co-ordination planning workshop (June
2004)

• SERDP Marine Mammal Modeling Experts Workshop
• Duke University, June 7-9 2004

Nancy Friday National Marine Mammal Lab Lance
Garrison Southeast Fisheries Science Center Glen
Gawarkiewicz Woods Hole Oceanographic
Institution Bob Gisiner Office of Naval
Research Caroline Good Duke University Pat
Halpin Duke University Toshihide Hamazaki Alaska
Department of Fish Game David Hyrenbach Duke
University Kristen Kaschner University of British
Columbia Erin LaBrecque Duke University
Jay Barlow Southwest Fisheries Science Center
Mark Baumgartner Woods Hole Oceanographic
Institution Joel Bell Naval Facilities
Engineering Command Ben Best Duke
University Chris Calloway University of North
Carolina, Chapel Hill Michael Coyne Duke
University Per Fauchald Norwegian Institute for
Nature Research Megan Ferguson Scripps
Institution of Oceanography Karin
Forney Southwest Fisheries Science Center Sloan
Freeman Duke University
Debi Palka Northeast Fisheries Science
Center Andy Pershing Cornell University Song
Qian Duke University Andy Read Duke
University Jessica Redfern Southwest Fisheries
Science Center Jason Roberts Duke
University Leigh Torres Duke University Kathy
Raposa University of Rhode Island Cisco
Werner University of North Carolina
11
Co-ordination Plan

• Outline
• 1. Initial co-ordination planning workshop
  (June 2004)
• 2. Annual co-ordination meetings (2005, 2006
  2007)
• 3. Preparation of a joint review manuscript (In
  Press)
• 4. Ensure consistency of modeling approaches (In
  Progress)
• 5. Joint toolbox development (Planned)
• 6. Joint dissemination of SDSS to Navy end-users
  (Planned)

12
Co-ordination Plan

• Annual co-ordination meeting (SWFSC, June 2005)
• Attendees
• 1390 - Dr. Andrew Read, Dr. Patrick Halpin, Dr.
  Song Qian, Ben Best, and Robert Schick
• 1391 - Dr. Jay Barlow, Dr. Karin Forney, Dr. Lisa
  Ballance, Dr. Paul Fiedler, Dr. Jessica Redfern,
  Dr. Megan Ferguson, Elizabeth Becker, Nacho
  Vilchis, and Liz Zele.
• Discussion topics
• exploring alternative statistical techniques for
  building cetacean-habitat models
• improve techniques for estimating variance in
  predicted cetacean densities
• comparing models built from varying spatial
  extents
• improving methods for model selection
• modeling estimated density rather than encounter
  rate and group size separately
• combining depth and slope as predictor variables
  in habitat models,
• using distance from geographic features as
  predictor variables in habitat models.

13
Co-ordination Plan

• Annual co-ordination meeting (SWFSC, June 2005)
• (Continued)
• Priority modeling tasks identified by the SERDP
  teams
• estimating density for any arbitrarily defined
  polygon
• developing rigorous smoothing or interpolation
  methods to express model results as gridded and
  contoured plots
• estimating uncertainty for predicted cetacean
  densities and graphically displaying this
  uncertainty
• graphically displaying model goodness-of-fit.
• developing software to support the Navys use of
  the SERDP project results

14
Co-ordination Plan

• Annual co-ordination meeting (SWFSC, June 2005)
• (Continued)
• New directions discussed
• Data automation systems
• Analytical tool development (desktop GIS systems)
• Spatial Decision Support Systems (SDSS)

15
Co-ordination Plan

• Outline
• 1. Initial co-ordination planning workshop
  (June 2004)
• 2. Annual co-ordination meetings (2005, 2006
  2007)
• 3. Preparation of a joint review manuscript (In
  Press)
• 4. Ensure consistency of modeling approaches (In
  Progress)
• 5. Joint toolbox development (Planned)
• 6. Joint dissemination of SDSS to Navy end-users
  (Planned)

16
Co-ordination Plan

• Joint modeling review manuscript (In Press)

Redfern et al. 2006. Techniques for Cetacean
Habitat Modeling A Review. Marine Ecological
Progress Series
19 co-authors
Outline 1. INTRODUCTION 2. MODELING PURPOSES 3.
SCALE 4. DATA COLLECTION 5. DATA CONSIDERATIONS
6. DESCRIPTIVE TECHNIQUES 7. MODELING TECHNIQUES
8. MODEL FITTING 9. MODEL EVALUATION 10.
CONCLUSION
17
Co-ordination Plan

• Outline
• 1. Initial co-ordination planning workshop
  (June 2004)
• 2. Annual co-ordination meetings (2005, 2006
  2007)
• 3. Preparation of a joint review manuscript (In
  Press)
• 4. Ensure consistency of modeling approaches (In
  Progress)
• 5. Joint toolbox development (Planned)
• 6. Joint dissemination of SDSS to Navy end-users
  (Planned)

18
Co-ordination Plan

• Ensure consistency of modeling approaches (In
  Progress)

• Towards a Consilience of Approaches
• Use of derived oceanographic products from ocean
  circulation models

19
Co-ordination Plan

• Outline
• 1. Initial co-ordination planning workshop
  (June 2004)
• 2. Annual co-ordination meetings (2005, 2006
  2007)
• 3. Preparation of a joint review manuscript (In
  Press)
• 4. Ensure consistency of modeling approaches (In
  Progress)
• 5. Joint toolbox development (Planned)
• 6. Joint dissemination of SDSS to Navy end-users
  (Planned)

20
CS-1390 Approach
(2) Remote sensing data fusion
Data
(1) Archival data mining
Information
(3) Automated modeling system
(4) Web services SDSS
21
CS-1390
Habitat Modeling System
22
CS-1390
Habitat Modeling System
17 guilds x 2 regions x 4 seasons x 3 methods
(GLM, GAM, CART)
23
Co-ordination Plan

• Outline
• 1. Initial co-ordination planning workshop
  (June 2004)
• 2. Annual co-ordination meetings (2005, 2006
  2007)
• 3. Preparation of a joint review manuscript (In
  Press)
• 4. Ensure consistency of modeling approaches (In
  Progress)
• 5. Joint toolbox development (Planned)
• 6. Joint dissemination of SDSS to Navy end-users
  (Planned)

24
GIS Web Services for Naval Planning
CS-1390
Habitat Modeling System
Joint dissemination of SDSS to Navy end-users
(Planned)

• Remote - Duke
• Observations, Effort
• By Dataset
• By Species
• Range Maps
• Predictive Models
• Expert-Reviewed
• Up-to-date
• Local Navy
• Naval OPAREAs
• Remote - JPL
• Satellite Data
• SST, SSH, Wind
• Near Real-time

25
GIS Web Services for Naval Planning
CS-1390
Habitat Modeling System
EIMS
Web Link
Navy users PKI-enabled (SSL) computers
INTERNET
IIS 6Web Server
Other Web Content
Internet Map Server (ArcIMS)
Navy-owned server, database, and associated
software
ArcSDE Gateway
Oracle Database
Database model
GIS Views
Relational Tables
New info update
Spatial Tables
Modified by P. Halpin
26
CS-1390
Habitat Modeling System
27
Co-ordination Plan

• Outline
• 1. Initial co-ordination planning workshop
  (June 2004)
• 2. Annual co-ordination meetings (2005, 2006
  2007)
• 3. Preparation of a joint review manuscript (In
  Press)
• 4. Ensure consistency of modeling approaches (In
  Progress)
• 5. Joint toolbox development (Planned)
• 6. Joint dissemination of SDSS to Navy end-users
  (Planned)

Review
28
Co-ordination Plan

• Future Annual Co-ordination Meetings
• Summer 2006 2007

1. Compare predictive capability of both
modeling approaches 2. Agree on full
specifications of SDSS for each project 3.
Review progress on construction of SDSS for each
project 4. Review beta-testing of SDSS by select
Navy end-users 5. Plan combined product
dissemination to Navy end-users
29
BACKUP SLIDES
30
CS-1390
Habitat Modeling System
Baleen Whales, Beaked Whales and Tursiops
31
Baleen whales
Tursiops
Beaked whales
CS-1390
Habitat Modeling System
Example model prediction period July 2004
32
Baleen whales
Tursiops
Beaked whales
CS-1390
Habitat Modeling System
Example model prediction period July 2004
33

• Model results
• Atlantic region only
• GLM models

Summer GLM models True positives Min
77 Max 99 Mean 91 R2 Min 0.22 Max
0.66 Mean 0.41