5. Machine Learning Process with Google Streets
Geo-localization of tree
canopy (Step 1)
● Aerial imagery is
used to identify
where trees are.
● Canopy pixels are
extracted and
vectorized to define
the boundary called
the tree canopy
zone.
Estimating tree
count (Step 2)
● Within the tree
canopy zone, street
view imagery is used
to find the trees
under street view
Estimating distance
from observer (Step 3)
● A heat map is generated
that defines the distance
of each pixel from the
observer.
● Using this, the average
distance of tree pixels is
calculated within the
bounding box extracted
in step 2
Identifying location of
individual trees (Step 4)
● Observer location and
field of view is projected
in aerial view (the right
angle in blue above)
● Using the distance
calculated in step 3,
individual trees are
placed on aerial image
map (yellow points).
Photo credit - SiteRecon
12. Implementing Tree Monitoring Program
Year 1
Initiate tree monitoring
program
Perform advanced
assessments
Install TreeKeeper 9
Year 2
Implement information
via TreeKeeper 9
Year 3
Perform tree
monitoring data
collection
Perform advanced
assessments of
flagged trees
Perform change
analysis
Update TreeKeeper 9
Year 5
Perform tree
monitoring data
collection
Perform advanced
assessments of
flagged trees
Perform change
analysis
Update TreeKeeper 9
Year 4
Implement information
via TreeKeeper 9
Photo credit - greehill
13. Initial assessment
greehill drives streets &
parks per contract specs
Data Delivery
Data is delivered into
TreeKeeper 9
Data extraction
Data is processed via
machine learning to provide
information per data specs.
Advanced Assessments
Davey provides Level 2 or 3
assessments to flagged trees.
Flagging Trees
Based on results of data,
client goals, & budget a
certain # of trees are
identified for advanced
assessments
Tree Monitoring
Program
Operation workflow
Photo credit - greehill
14. Using Machine Learning for
Tree Inventories
Josh Behounek
573-673-7530
Josh.Behounek@davey.com
15.
16. A Scalable Remote Sensing Model for Urban
Forests
From airborne missions to satellites
Jonathan Pando Ocón, UCLA
E. Natasha Stavros, CU Boulder
Thomas W. Gillespie, UCLA
Justin Robertson, LA County
Steven J. Steinberg, LA County
Image: Louis Reed, Unsplash
18. Project Stakeholders
Image: Ev Milee, Unsplash
To ensure our project goals are met, we sought input from stakeholders to identify the
needs and priorities of those that will be using our product in day-to-day operations:
Stakeholder Advisory Group
End users and department managers
Local advocacy and conservation groups
19. Stakeholder Needs
Image: Henry Perks, Unsplash
Individual tree species identification
Canopy cover metrics
Health assessment
20. Remote Sensing of Urban Environments
Depending on the mode of image
capture, the spatial and spectral
resolutions of your chosen dataset can
capture a high amount of noise:
High species richness
High density of (in)organic
materials
22. Existing Approaches to Unmixing Pixels
Spectral Mixture Analysis
Multiple Endmember Spectral Mixture Analysis (MESMA) is used to account for
subpixel spectral mixing
Computationally expensive
Complex workflow
23. Success Stories
Able to untangle spectral confusion relatively well in less complex scenes.
Image: Johannes Mandel, Unsplash
24. Missing the Mark
Prone to misclassifications in complex urban environments.
Image: Denys Novazhai, Unsplash
25. Science-led Resources
Science-led projects can deploy the state-of-the-art in remote sensing image analysis using
complicated workflows, computationally expensive models, and often a number of technically savvy
individuals to inform the model.
Uneven distribution of resources
Public and governmental stakeholders may or may not have this.
26. Stakeholder-led Resources
Adhering to our stakeholder needs informs the research design immensely.
Considerations for processing capacity, and technical know-how are necessary to ensure end-users
can continue to use and update the product.
1. Who will run the model next year? The year after?
2. Are additional image acquisitions needed?
3. What is the best product format for day-to-day operations?
27. Probabilistic Urban Forest Inventories
High Confidence
Ficus rubiginosa
Low
Confidence
Uncertai
n
No Presence
29. Thank you!
Jonathan P. Ocón
Ph.D. Candidate, UCLA
Contact:
jonocon[at]ucla.edu
Image: Louis Reed, Unsplash
30.
31. Partners in Community Forestry Conference
Nov 16th, 2022
Urban Tree Water Use and Implications for
Stormwater Management
Sarah Ponte, Nancy Sonti, A. Christopher Oishi,
Dexter Locke, Tuana Phillips, and Mitchell Pavao-
32. Urban development alters the natural hydrologic
cycle
Askarizadeh et al., 2015,
Environ. Sci. Technol.
34. Image Credit: Jeffrey Milstein
Trees in urban areas vary by
Ecohydrological Landscape
Characteristics (ELCs) (Blood and Day)
35. Transpiration Rates and Whole-Tree Water Use
by Deciduous Species
Objective:
To quantify the use of trees to
meet stormwater management
requirements.
Single trees over turfgrass
Cluster of trees over turfgrass
Closed canopy
Baltimore City
Montgomery County
37. Sap flux is a proxy for transpiration rates
Granier-type thermal dissipation sap flux sensors (built by the UMD
Project Development Center)
38. Different management contexts have significantly
different daily sap flux (Js) distribution
Repeated
measures ANOVA
p < 0.0001
Ponte et al., 2021
Scientific Reports
Single red
maple trees
had nearly
three times the
daily sum of
sap flux density
(Js) of closed
canopy trees
Sap
Flux
39. Water use among these common, diffuse-
porous, deciduous, eastern species is similar
Sap
Flux
40. Species differences in sap flux density were
observed at the 24h time-scale
July 19, 2019 – high soil moisture
Aug 30, 2019 – drought conditions
Tulip poplar was the
most sensitive to
drought
41. Whole-tree water use estimates based on a
predictive model
0
50
100
150
200
250
300
0 10 20 30 40 50 60
L
day
-1
tree
-1
DBH (cm)
A. rubrum - Single
L. styraciflua - Cluster
A. rubrum - Cluster
A. rubrum - Closed Canopy
L. tulipifera - Closed Canopy
L. styraciflua - Closed Canopy
42. 0
100
200
300
400
500
mm
H
2
O Cumulative Plot-Level Transpiration -
900 m2 plots
Single Cluster Closed Canopy
2019
Single: 9% of precipitation
Cluster: 71% of precipitation
Closed Canopy: 52% of precipitation
43. • Tree water use depends on
management context
• Water use among these diffuse-
porous, deciduous, eastern
species is similar; however,…
• …they do exhibit differences in
response to drought
Main takeaways
44. This study provides a
foundational framework for
estimating how a proposed
urban tree/forest project
would affect the hydrologic
balance with useful
implications for stormwater
managers
e. sponte@umd.edu
@sarahpcabral
45.
46. i-Tree is a
Cooperative Initiative
among these partners
Tree Equity:
How i-Tree is Helping
“…enough trees in specific
neighborhoods or
municipalities for everyone
to experience the health,
economic and climate
benefits that trees provide.”
Landscape.itreetools.org
47. i-Tree is a
Cooperative Initiative
among these partners
Tree Equity:
How i-Tree is Helping
“…enough trees in specific
neighborhoods or
municipalities for everyone
to experience the health,
economic and climate
benefits that trees provide.”
Landscape.itreetools.org
“…a map of tree cover in
America’s cities is too often
a map of income and race.”
48. i-Tree is a
Cooperative Initiative
among these partners
Tree equity is hard work
• Skeptical
communities
• Overcoming
negative
opinions
• Dealing with tree
mortality
49. i-Tree is a
Cooperative Initiative
among these partners
How can i-Tree help?
Identify – Find where trees can do
the most good.
Engage – Communicate the
benefits of trees.
Account – Ensure delivery of tree
benefits.
50. i-Tree is a
Cooperative Initiative
among these partners
Free software
Estimate the
benefits of trees
Based on US
Forest Service
science
Technical support
i-Tree is a
Cooperative
Initiative
www.itreetools.org
i-Tree: Putting Urban Forest Science into the
Hands of Users
www.itreetools.org
51. i-Tree is a
Cooperative Initiative
among these partners
The i-Tree Framework
Structur
e
Value
Function
52. i-Tree is a
Cooperative Initiative
among these partners
Identify: Where to plant
Using i-Tree
Landscape
To prioritize
where tree
planting is
equitable
Landscape.itreetools.org
53. i-Tree is a
Cooperative Initiative
among these partners
Identify: Where to plant
Using i-Tree
Landscape
To prioritize
where tree
planting is
equitable
Place Priority Index
Darby 100
Camden 91
Millbourne 87
Chester 86
Woodlynne 77
Colwyn 77
Warminster 76
Coatesville 75
East Lansdowne 72
Norristown 71
Upland 67
Yeadon 65
Collingdale 65
Sharon Hill 64
South Coatesville 60
Avondale 58
Lansdowne 58
Clifton Heights 57
Pottstown 56
Bridgeport 56
Oxford 55
Landscape.itreetools.org
54. i-Tree is a
Cooperative Initiative
among these partners
Identify: Where to plant
i-Tree Design – Energy
Savings
i-Tree Landscape – Heat Island
Landscape.itreetools.org
Design.itreetools.org
55. i-Tree is a
Cooperative Initiative
among these partners
Engage: i-Tree for Education
MyTree.itreetools.org
56. i-Tree is a
Cooperative Initiative
among these partners
Account: MyTree Accountability Dashboard
MyTree.itreetools.org
57. i-Tree is a
Cooperative Initiative
among these partners
Account: Monitor Impact Over Time
2007 2017
Canopy.itreetools.org
58. i-Tree is a
Cooperative Initiative
among these partners
i-Tree: Make tree equity count
Identify where to plant trees to maximize
benefits.
Engage communities to support
stewardship.
Account for how tree equity efforts benefit
underserved communities.
59. i-Tree is a
Cooperative Initiative
among these partners
i-Tree: Make tree equity count
Jason Henning PhD
The Davey Institute and
USDA Forest Service, Philadelphia Field Station
jason.henning@davey.com
www.itreetools.org
MyTree Design
Landscape OurTrees
Canopy