wildtrack ppt for jmp

1
University of Florida USA/Guatamala
Sky Alibhai Zoe Jewell
wildtrack_at_clix.pt
www.wildtrack.org
Centro de Estudos e Investigação Científica
Tapir Conservation Project, Argentina
2
STRIPED HYAENA, TURKEY
POLAR BEAR, CANADA
Some of the WildTrack projects monitoring
endangered or elusive species using the Footprint
Identification Technique (FIT)
BENGAL TIGER, INDIA
BROWN BEAR, GREECE
BLACK RHINO, SOUTH AFRICA
CHEETAH, NAMIBIA
LOWLAND TAPIR, BRAZIL
3
FIT
FIT is designed to identify animals at the
species, individual, gender and age-class levels
from their footprints. FIT utilises
software from to create a geometric
profile of the footprint and for data analyses.
The advantages of FIT

• Non-invasive

• Cost-effective

• Highly accurate

• Integrates local communities and expertise

Angie Nash/
4
THE IMPETUS FOR DEVELOPING NON-INVASIVE METHODS
Our research with black rhino in Zimbabwe
indicated that routine immobilization required to
maintain radio-collars for telemetric monitoring
of individuals had a negative effect on female
fertility
COLLAR FAILURE RATE
FEMALE FERTILITY
M
F
MF
Alibhai Jewell (2001). Oryx 35(4), 284-288
Alibhai, Jewell Towindo (2001) J. Zool. Lond.
253, 333-345.
5
FIT How does it work?
6
SUMMARY OF FOOTPRINT IDENTIFICATION TECHNIQUE
DIGITAL IMAGE OF FOOTPRINT
LEFT HIND FOOTPRINT
GEOMETRIC PROFILE IN
GENERATING gt100 ANGLES, DISTANCES POLYGONS
DATA ANALYSIS IN
7
SUMMARY OF FOOTPRINT IDENTIFICATION TECHNIQUE
DIGITAL IMAGE OF FOOTPRINT
LEFT HIND FOOTPRINT
GEOMETRIC PROFILE IN
GENERATING gt100 ANGLES, DISTANCES POLYGONS
DATA ANALYSIS IN
ALGORITHM DEVELOPMENT ,TESTING AND PRACTICAL
APPLICATION FOR
INDIVIDUAL ID
GENDER ID
AGE CLASS
LATERALITY
SPECIES ID
M
?
?
F
8
Captive tiger footprint
Photography protocol
Replicates required
9
Initial photo manipulation
Raw image
Greyscale
Photo-optimise
First 2 Landmark points
Rotate image
10
Drag and drop from image
processing software into JMP
11
Input animal data, and identify remaining
landmark points
12
LANDMARK POINTS
DERIVED POINTS
13
Animal data, with measurements of footprint
lengths, angles and areas are exported to a JMP
data table to provide geometric profile of
footprint
14
DISCRIMINATION TECHNIQUES
TRACK 2
RCV

CANONICAL 2

TRACK 1
CANONICAL 1
15
IDENTIFICATION OF INDIVIDUALS USING FIT
SCENARIO A
SCENARIO B
KNOWN POPULATION SIZE (MONITOR)
UNKNOWN POPULATION SIZE (CENSUS)
HOW MANY ARE WE?
WHO AM I?
16
CANONICAL CENTROID PLOT TECHNIQUE TRACK VS TRACK
(CENSUS SCENARIO)
CLASSIFIER
CLASSIFIER
TRACK 2

TRACK 2
RCV

RCV



TRACK 1

TRACK 1
RCV REFERENCE CENTROID VALUE
Jewell, Alibhai Law (2001) J. Zool. Lond. 253,
333-345.
17
A. THE REFERENCE CENTROID VALUE (RCV) We found
that the easiest and most effective way of coming
up with the RCV was to simply repeat the data in
our database of known individuals and relabelling
it RCV. In this example, the database consisting
of 1269 footprints from 40 known individuals was
duplicated to constitute the RCV.
RCV 1269 ROWS
Jewell, Alibhai Law (2001) J. Zool. Lond. 253,
333-345.
18
B. WHICH VARIABLES AND HOW MANY?

• VARIABLES SELECTED STEPWISE BASED ON F RATIOS IN
  JMP

INCREASING NUMBER OF VARIABLES ELLIPSES FURTHER
APART
RCV
TRACK 1
TRACK 2
TRACK 1
TRACK 2
TRACK 2
TRACK 1




CANONICAL 2


DECREASING NUMBER OF VARIABLES ELLIPSES CLOSER
TOGETHER
CANONICAL 1
19
C. TRACK SIZE HOW MANY FOOTPRINTS PER TRACK?
DECREASING TRACK SIZE INCREASES ELLIPSE DIAMETER
RCV
TRACK 1
TRACK 2


CANONICAL 2
INCREASING TRACK SIZE DECREASES ELLIPSE
DIAMETER
CANONICAL 1
20
C. HOW MANY FOOTPRINTS PER TRACK?
n152 tracks with a total of 1269 footprints from
40 known white rhino
Alibhai, Jewell Law (2008). Endang Spec Res
21
ALGORITHM DEVELOPMENT AND TESTING FOR INDIVIDUAL
IDENTIFICATION OF WHITE RHINO USING THE CANONICAL
CENTROID PLOT TECHNIQUE
WHITE RHINO
LEAVE ONE PAIR OUT
TRACK VS TRACK ACCURACY 99
12 VARIABLES SELECTED STEPWISE 11,329 TRACK
VERSUS TRACK COMPARISONS IN PRESENCE OF RCV 104
MISCLASSIFICATIONS
40 INDIVIDUALS 1269 FOOTPRINTS 152 TRACKS
GROUP ASSOCIATIONS PREDICTED 42 RHINOS (95 ACC)
Alibhai, Jewell Law (2008). Endang Spec Res
22
IDENTIFICATION OF INDIVIDUALS USING FIT
SCENARIO A
SCENARIO B
UNKNOWN POPULATION SIZE
KNOWN POPULATION SIZE (MONITOR)
HOW MANY ARE WE?
WHO AM I?
23
ACTUAL CLASSIFICATION OF TEST TRACK
TEST TRACK
LINEAR DISCRIMINANT ANALYSIS USING VARIABLES
SELECTED STEPWISE IN JMP
?
FOR THIS TEST TRACK, ONLY TWO OUT OF EIGHT
FOOTPRINTS CLASSIFIED CORRECTLY
?
24
CANONICAL ELLIPSE REDUCTION TECHNIQUE (CERT)
TRACK VS SET (MONITOR SCENARIO)
DATA FROM ALL FOOTPRINTS INCLUDED IN A TWO-WAY
CANONICAL PLOT USING THE SAME VARIABLES AS FOR
LDA. ELLIPSES NOT OVERLAPPING WITH THE TEST
ELLIPSE THEN EXCLUDED FROM THE ANALYSIS UNTIL
TEST TRACK OVERLAPS WITH ONLY ONE OTHER ELLIPSE.
TEST TRACK
TEST TRACK
TEST TRACK
Alibhai, Jewell Law (2008). Endang Spec Res
25
               
GENDER IDENTIFICATION IN THE BENGAL TIGER
F
M
WILD TIGER TEST TRACK
26
SPECIES IDENTIFICATION
SEPARATING INDIVIDUALS OF SYMPATRIC SPECIES
WHITE RHINO
BLACK RHINO
Alibhai, Jewell Law (2008). Endang Spec Res
27
DATABASE, ALGORITHM DEVELOPMENT AND TESTING USING
FIT
SPECIES
TEST MODEL
ACCURACY
WHITE RHINO
LEAVE-ONE-OUT
BLACK RHINO
50 MODEL50 TEST
BENGAL TIGER
gt 90 IN ALL CASES
LOWLAND TAPIR
EQUAL IMAGES PER SET FOR MODEL
BAIRDS TAPIR
TEST IMAGES FROM OTHER SOURCES
AFRICAN LION
POLAR BEAR
28
Refining and further developing FIT
Automated Image Extraction Computer
Vision Polynomial texture maps
Automation of JMP routines and development of
user-interface Sabbatical at SAS and NC State
2009
Planning with SAS and JMP to manage large
databases and open-access management
29
FIT Base An endangered species footprint
database
Building a database of
footprints from known individuals
of endangered species to help their wild
counterparts
FIT geometric profiling and algorithm development
FIT BASE
Angie Nash/
30
A NON-INVASIVE INUIT-LED PROTOCOL FOR POLAR BEAR
MONITORING NUNAVUT TERRITORY, CANADA
AN INTEGRATED APPROACH
Conserving wildlife through the application of
traditional Inuit and scientific
knowledge nwmb.com
DNA FROM HAIR AND FOOTPRINTS
FIT
TRADITIONAL INUIT FOOTPRINT TRACKING
Angie Nash/
31
Helping to protect and monitor endangered species
FIT, a sustainable way forward
Cost-effectiveness Monitoring Polar bears
using FIT, genetics and Inuit TEK is estimated to
be only 10 of the cost of current techniques (de
Groot, pers.comm.)
Community-strength FIT incorporates the skills
of indigenous communities. Local people become
active stakeholders with proven long-term
conservation benefits.
Environmental responsibility FIT does not
require aircraft, helicopters, collars,
transmitters or drugs. It can utilise existing
labour and vehicles where required. It is
non-invasive to the animals being studied.
Corporate and national responsibilities are
increasingly key in protecting environment and
biodiversity. Sustainable monitoring is the key
to evaluating progress in meeting targets in
ecological risk management.
32
The costs of biodiversity losses are not only
greater than those of the current financial
problems, but in many cases, the losses are
irreparable. IUCN World Conservation Congress
statement Barcelona, October 2008