Many arboreal (tree-dwelling) animals repeatedly use pathways called “canopy highways” to move through their three-dimensional habitat. Using light detection and ranging (LiDAR) forest structure maps and arboreal primate movement data, I am creating a computer model to identify potential highways, which I will verify in the field using camera traps.

1. Characterization of arboreal mammal
movement pathways in neotropical
forest canopies
Kevin McLean
Sigma Xi Student Research Showcase
March 2013

2. Introduction
My name is Kevin McLean, I am a third year doctoral student at
the Yale School of Forestry and Environmental Studies. The
project that I have outlined here is a portion of my dissertation
research. I will present the background and hypotheses,
methods, justification, and preliminary results by way of the
four most common questions that are asked of me by my
friends, colleagues, parents, and peers:
“So, what do you do?”
“How do you do that?”
“Why bother?”
“What have you found?”

3. Background
“So, what do you do?”
One of the first questions that is often asked of me in
any given conversation is about the general goal of my
life or research, which at times are one and the same. I
strive with varying success to offer a clear, succinct, and
accurate portrayal of my research interests no matter
who my audience may be.

4. Background: “So, what do you do?”
The conversation pictured
on the right with my high
school friend is one of the
more successful attempts in
terms of providing an idea
of my research focus. Before
delving into the details of
canopy highways and how I
study them, let me provide a
short backstory of how I
arrived at this seemingly
obscure topic…

5. In 2010 I worked with the
Terrestrial Ecosystem
Assessment and Monitoring
(TEAM) Network on part of a
project that aimed to monitor
mammal populations
throughout the tropics using
motion-sensitive cameras, or
“camera traps” (top left).
As many species in the forest
are nocturnal or otherwise
difficult to observe directly
(“cryptic”), camera traps are
widely-used and highly effective
means of studying animals as
they move naturally through
their environment.

6. Shortly after placing one of our
cameras, I spotted this
Northern Tamandua (a type of
anteater) traversing from tree
to tree – right over the camera
we had just placed.
This got me thinking about all
the arboreal (tree-dwelling)
mammals – monkeys,
opossums, rats, squirrels – that
we were missing by just
focusing on the ground.

7. After a little reading, I found out that neotropical
forests have the highest diversity of non-volant
(non-flying) arboreal mammals of any ecosystem
in the world, with up to 60% of mammal species
considered at least partially arboreal.

8. Given the tendency of ground-
level animals to travel along
trails and the more limited
substrate available for
movement in the canopy, it
follows logically that animals
would repeatedly use specific
pathways to move through the
treetops. Researchers have
referred to these as arboreal
routes or pathways, arboreal
highways, or in some cases,
canopy highways.

9. Only a few studies have specifically
examined this “route-based movement”
and characterization of the pathways that
animals use is extremely challenging.
Two studies that have done so have found
that food resource availability, forest
structure, and landscape topography are
significant predictors of the routes that
arboreal primates will use repeatedly.
Pathways used by spider monkeys and
woolly monkeys in Amazonian Ecuador
(Di Fiore & Suarez 2007) Forest structure has historically been
particularly difficult to quantify and thus
requires further study in order to
determine its importance. Recent advances
in remote sensing (e.g. high-resolution
LiDAR) provide far more accurate, three-
dimensional means of measuring forest
structural properties than the two-
dimensional measures (e.g. leaf area index)
Pathways used by howler monkeys on that have been traditionally used.
Barro Colorado Island, Panama
(Hopkins 2011)

10. Hypotheses
The knowledge of repeated pathway use in combination
with the technological innovations that provide an
improved understanding of forest structure along these
pathways, I planned out a research project that would
allow me to investigate the following hypotheses:
1. Primate pathways are non-random with respect to
forest structural characteristics.
1. Forest structural features can be used to predictively
model primate movement behavior.

11. Study Site and Focal Species
I chose to conduct my research on Barro Colorado
Island (BCI), a reserve operated by the Smithsonian
Tropical Research Institute (STRI) in the Panama Canal.
BCI is uniquely suited for this project because of the
wealth of data available on forest composition and
animal movement.
The Carnegie Airborne Observatory has mapped 98% of
BCI using airborne Light Detection and Ranging
(LiDAR). These data provide information on the crown
height and density of canopy and understory structure
for the entire island, essentially creating a map of
arboreal connectivity.
I am currently focused on two arboreal species for
which collaborators have acquired movement data on
BCI: white-faced capuchin monkeys (bottom left) and
mantled howler monkeys (bottom right). Movement
data have been collected for additional arboreal species
on BCI as well, including black-handed spider monkeys,
kinkajous, and northern tamanduas. Many of these
species have known differences in their movement and
ranging behavior; including a diverse set of will facilitate
robust models of forest connectivity.

12. Methods: “How do you do that?”
I have organized my research
plan into three phases that will
take place over the course of
the next two years:
Phase 1: Development of
arboreal camera trapping tools
and methods.
Phase 2: Spatial modeling of
arboreal primate movement
behavior.
Phase 3: Field-based model
evaluation with arboreal
camera traps.

13. Phase 1
Arboreal Camera Trapping Methods
In anticipation of the final field-based
model verification, I began testing
arboreal camera trapping methods in
December 2012 following extensive
training in canopy access techniques.
Using custom-built camera
mounts, I deployed six cameras
over the course of three months
for a total of over 200 trap-nights.
The primary challenge beyond simply accessing the
canopy was placing the cameras in a manner that would
detect animal activity with minimal false deployments.

14. Phase 2
Movement Modeling
Movement Data Forest Structure Data
Movement trajectory data for the focal species Light Detection and Ranging (LiDAR) data
have been collected by collaborating researchers collected by the Carnegie Airborne Observatory
using telemetry (pictured below) and direct use multi-return laser pulses across the entire
observation for capuchin and howler monkeys, island. This intensive approach produces a
respectively. To make trajectory data comparable detailed three-dimensional representation of
across species, I divided all trajectories into daily forest structure, rendered as 1.2x1.2x1.0-m3 (60
segments to evaluate if distinct forest structure cm) voxels reaching from the forest floor to tree
characteristics are associated with pathways used crowns nearly 60 m above ground level (Asner et
repeatedly. al. 2011).
LiDAR
(Light Detection and Ranging)

15. Phase 2
Movement Modeling
Generation of a movement model is a multi-
step process. First, in order to construct a set
of predictor variables from the LiDAR data, I
am currently generating null models for
canopy height and each 1-m slice of
vegetation density. The null models provide
a baseline against which to compare the
correlation between real movement
trajectories and environmental variables (e.g.
MCH, vegetation density by height, ground-
level elevation) The output is a series of Preliminary null model analysis of
distributions that indicate whether the capuchin monkey trajectory using
environmental variables along the random shift rotation (RSR). Shape and
trajectories analyzed are significantly distinct length of path preserved and placed in
from the random distribution of these array of positions and orientations
variables across the landscape. across the landscape.

16. Phase 2
Movement Modeling
The next step in the movement model will be to apply step
selection functions (SSF). SSF use logistic regression to
quantify relationships between animal movement trajectories
and environmental variables. Features present along each
successive 10-minute “step” in the trajectories for each animal
are analyzed. In effect, the model is comparing the movement
taken to the movement possible for each successive time step,
providing an estimate of how suitable a pathway is for
movement.
Used features are compared to
available features by generating steps
with the same starting point, which
preserves the length and tortuosity of
the entire trajectory.
Zeller et al. 2012

17. Phase 2
Movement Modeling
The outputs of the null model and the step selection
functions are probability distributions for selection of
environmental variables. For model selection, I will use
Akaike information criterion (AIC) to select optimal
combinations of these variables to produce habitat
suitability models that indicates the likelihood that animals
will cross through the available habitat.
k: Number of parameters
L: Maximized value for likelihood function
Akaike 1974

18. Phase 2
Movement Modeling
The final step in the movement modeling
phase will be generation of a resistance
surface. Resistance surfaces are a common Simplified Resistance Surface
means of visually representing predicted
movement from habitat suitability models.
Resistance surfaces operate under the
principle of “friction” associated with a
landscape - areas that are preferred for
movement are those with the least friction,
or resistance. As habitat suitability models
represent features identified as preferred for
movement, I will use an inversion function (f
= h-1) to transform measures of suitability Low Resistance
into measures of resistance. The output will (High probability of use)
be a resistance surface map showing
corridors across the landscape (i.e. High Resistance
movement pathways) with the least (Low probability of use)
resistance to animal movement.

19. Phase 3
Field-Based Model Evaluation
Installation of camera traps at fruiting
X1B trees has been shown to provide an
X4B accurate means of estimating foraging
X1A X2A frequency. In order to test the
X5A movement models in the field, I will
X4A X5B install cameras at paired trees at similar
X2B fruiting stages within and outside the
X3A
X3B least-cost corridor (low resistance area).
Low Resistance I will deploy each camera in the field for
High Resistance 30 days and analyze all photos to
X Fruiting Tree determine frequency and species
Increased detection from cameras at composition of visitation (foraging)
sites identified as least-cost corridors events. This will serve as a proxy of
(low resistance) in the resistance habitat use in order to evaluate the
surface will serve as validation of the performance of the model.
model’s performance.

20. Justification: “Why bother?”
For those brave enough to stick
around and hear me out on my
often lengthy tale of what I do
and how I plan to do it, the next
inevitable question generally
focuses on why I’ve decided to do
it. The justification for basic
research is as important in
scientific circles as it is in social
ones, so in planning out my
research, I have been mindful of
how the outcome of my project
will inform subsequent ecological
research as well as conservation
and land use planning.

21. Ecological Research Conservation Applications
 Animals distribute seeds and waste in patterns  Developing a means of identifying important
that reflect movement behavior. pathways for arboreal animals is critical for their
 Some areas that receive high densities of protection.
animal-distributed seeds also exhibit higher
seedling success and sapling growth.  Habitat assessments for conservation planning
 Characterization of animal movement can rarely include an assessment of aboveground
provide predictive insight into forest habitat, instead relying on ground-based
regeneration patterns. measures of habitat quality that may mask
 Congruence between animal movement and complex forest structure.
their ecological effects would serve to  Suitable habitat as measured from the ground
conceptually “close” the feedback loop may not necessarily indicate suitable habitat in
between structure, animals, and forest the canopy.
regeneration.
 Behaviorally-informed habitat suitability models
using high-resolution LiDAR technology solves this
problem.

22. Results: “What have you found?”
As I am currently in the midst of
my research, most of the
progress I have made thus far is in
the first phase of my project.
Through trial and error I was able
to get the camera traps to run
successfully in the trees, and
though I will have to wait until I
have finished with the movement
modeling to answer my original
questions, the photos that the
cameras have collected thus have
sparked my interest in some
additional questions that I may be
able to address.

23. Preliminary Findings
Phase 1: Repeated Pathway Use?
Climbing Rat The selection of photos shown
here revealed a climbing rat
5-Jan (Tylomys spp.) moving along a
branch on several consecutive
nights (more than shown).
Further investigation of this
6-Jan behavior would require some
form of unique identification
or marking to determine
whether this is the same
individual or several different
7-Jan individuals using the same
pathway. In addition, a longer
monitoring period may reveal
whether use of this particular
8-Jan pathway changes with season,
fruiting stage of the
surrounding trees, weather,
etc.

24. Preliminary Findings
Phase 1: Multi-Species Canopy Highway?
Much like trails on the ground, the structural or environmental properties that make a
certain pathway ideal for movement for one species may translate for other species as
well. Further monitoring of this and surrounding branches would reveal the species
diversity and frequency of use changes over with time or environmental change.
Iguana
Capuchin
Monkey
Rodent

25. Summary
From a scientific perspective, my research
employs a novel integration of behavioral
ecology with previously unprecedented
information about the habitat in which animals
behave, which may provide more accurate
models upon which subsequent research and
management decisions can be made. From a
personal perspective, I have found a topic,
research questions, and methods that are are
both rigorous and enjoyable. While I still have
a significant portion of my project ahead of me
and will undoubtedly hit a few bumps along
the way, this project has and will continue to
provide me with the experience necessary to
grow and develop as a scientist. I will continue
to rely and build on the skills and knowledge
that I acquire from this project throughout my
career.

26. Acknowledgements
People Organizations and Funding
• Dr. Oswald Schmitz • National Geographic Society
• Dr. Mark Bradford • Yale Institute for Biospheric Studies
• Dr. Patrick Jansen • Yale Tropical Resources Institute
• Dr. Greg Asner • Smithsonian Tropical Research
• Dr. Meg Crofoot Institute
• Dr. Mariah Hopkins • Institute for Tropical Ecology and
• Dr. Anne Trainor Conservation
• Joe Maher
• Dr. Tremaine Gregory