Using Airborne LiDAR

Extraction of Mangrove Forest Parameters
Using Airborne LiDAR
By Ms. Wasinee Cheunban
Asian Institute of Technology
RS&GIS FOS, School of Engineering and Technology

Contents
1. Background
2. Objective
3. Methodology
4. Results
5. Conclusions

Technical challenges for REDD+
REDD+ (The Reduced Emissions From Deforestation and Degradation)
Forest carbon baseline and monitoring (Tier III)
Monitoring, Reporting and verification (MRV)
The BAU minus intervention of a REDD program =Credits

Objectives
1. To extract mangrove forest parameters at individual tree level
from LiDAR.
2. To implement functions for effective body mass and projected
area estimation according to water level.
The procedure
(1) Construction of Canopy Height Model (CHM).
(2) Extraction of mangrove forest biophysical parameters based on individual
tree level by using LiDAR; 1)Trees location 2)Tree height 3)Crown diameter
(3) Development of the DBH algometric model of Avicennia marina tree.
(4) Development of trunk diameter function for computing tree volume and
projected area at each height level
(5) Calculation of the effective body mass (Vo) and projected area (Ao) under
water depth.

Introduction

Study area
The study site is located at Bang Poo sub-district, the coastline of Samut Prakan
province, Thailand
E
N
S
W
120m
- Plot size = 100 x 120 m.
- In a flat at a tidal and mud zone.
- There are many mangrove forest with Avicennia marina
(Sa Mae Taray) species group
- Around of mangrove plot are many shrimp farm and factory
100m

Avicennia marina
Characteristics of Avicennia marina (Sa Mae Taray)
Science name: Avicennia marina
Family: AVICENNIACEAE
Local name: Sa Mae Taray
(Thailand)
Function Characteristics at study area
DBH 13- 38 cm.
Height 4- 13 m.
Crown width 4 to 8 m.
Crown shape
Leave
Age
usually egg-shaped, elliptic
light green about 10 cm long
> 12 year

LiDAR
LIDAR (Light Detection And Ranging) is
a technology for determining the shape
of the ground surface.
GPS
Z
Y
X
Lidar systems are active remote sensing
devices that measure the time of travel
needed for pulse of laser energy sent
from the airborne system to reach the
ground or object on the earth surface
and reflect back to the sensor. The time
measurement is converted into a
distance.
Laser
scanner
TL
INS
Z
Y
X
first return
- The first return shows the highest
features such as the tree canopy,
buildings etc.
- The last return is ground level.
last return
start pulse
last pulse
GPS
Z
Y
X

LiDAR
- Forest density and structure (and thus carbon stocks)
- High accuracy over large geographic area
- Limited ground plot measurement (Tropical rain forest)

LiDAR
LiDAR data in study plot
Top view of LiDAR image
crown area = 77.42 m 2
Laser point = 230 point
Average = 2.97 point/m 2
LiDAR image on the front of the view.

Methodology
1
Filter
Window
size
T = Tree position
H = Tree height
CW = Crown width
N = Number of tree
CHM = Canopy Height Model

Methodology
2
To implement functions for effective body mass and projected area
estimation at each level height.
Against Tsunami Energy
Tree Volume
V = V(h)
Projected area
A = A(h)
Tsunami Model
Water level
Vo and Ao
V 0 :
Effective body mass of trees under water
A 0
: Effective projected area of trees under water

Forest inventory
Data collection
Cw= (W 1
+ W 2
)/2
Rope (Y)
5
4
Tree height
3
Reference
point
2
1
X
0 1 2 3 4 5 6
Rope (x)
DBH = 1.3 m.
Tree position
measurement
DBH and Height
measurement
Crown width
measurement

Canopy
Height
Model: CHM
The CHM is a lidar-derived three-dimensional surface that contains
information on vegetation height above the ground surface.
CHM will be used to be data source for the tree location, height and
crown width extraction.
CHM = DSM - DEM
DSM
Lidar DEM
CHM

CHM Extraction
LiDAR provides 3-dimensional canopy surface with individual tree crown
The last return laser points Interpolate to regular grid by 50 cm DEM
The first return laser points Interpolate to regular grid by 50 cm DSM
3-dimensional canopy height model on the study plot.

TreeVaW
Tree Variable Window
(©Sorin Popescu)
Tree height estimates were base on single tree
identification using adaptive technique for local
maximum filtering with vary circular window size.
The derivation of the appropriate window size to
search for tree tops is based on a relationship
between the height of the trees and their crown width
The local maximum filter works best for trees with a
single, well defined apex..
+
+
+
+ +
+
+ +
+
+
+ +
+
+
+
+
+
+
Smaller
window size
Larger
window size
+
+
+
+
+
+ +
+ +
+
+
+
+
+ + + +
Circular window

Tree parameter extraction
TreeVaW is a canopy height model (CHM) algorithm implemented in IDL for locating and
measuring individual trees.
Input is a LiDAR-derived canopy height model (CHM) in header file format.
Parameter requirement
1. Min – Max crown width (for calculating the Min - Max window size)
2. Min tree height.
3. Median filtering size 3x3 pixel (pixel size = 0.5 m.)
4. Crown width approximation equation[1] Cw = 1.4334H – 1.7675
Outputs consist of individual tree position, tree height, and crown radius.
Input
Output
CHM
TreeVaW
tree positions
(Trees top)

Circular filtering window size
Tree location and tree height
The relationship between tree height and crown width
Field Observation
Crown width
Regression Model
Tree height
Used to determine the appropriate
circular filtering window size for
searching the tree top and crown
width from LiDAR.
Cw = f(H) ------Eq [1]
where: Cw = Crown width
H = Tree height

Crown Width Equation
Cw equation is relationship between tree height and crown width
Tree height has high correlation with crown
diameter explained by linear regression with
R 2 of 0.791
16.00
y
14.00
R 2 = 0.791
Cw = 1.4334H – 1.7675
---- [1]
12.00
10.00
Cw (m)
8.00
6.00
4.00
2.00
0.00
0.00 2.00 4.00 6.00 8.00 10.00 12.00 14.00 16.00
Ht (m)

Tree parameter extraction
Tree location and tree height
(implemented in IDL code)
1) Read one pixel (height value), 2) calculate
window size by Eq[1] and 3) consider, Is it a
maximum? but 5 is not a maximum, so turn
to read next pixel.
6 3 5 7 8 8 5
Cw = 1.4334H – 1.7675
6 6 6 5 6 8 7
6 7 9 9 8 9 7
Windows size = 5
7 9 5 10 9 8 6
8 8 9 9 9 8 5
6 7 8 7 8 6 7
5 is not maximum
6 7 9 6 7 9 7
6 3 5 7 8 8 5
6 6 6 5 6 8 7
6 7 9 9 8 9 7
Windows size = 7
7 9 5 10 9 8 6
8 8 9 9 9 8 5
6 7 8 7 8 6 7
10 is maximum
6 7 9 6 7 9 7
Remark : Pixel size =50 x 50 cm

Tree parameter extraction
Crown Width
(implement in TreeVaW IDL code)
Algorithm
Locate all trees on the CHM
Get one tree location
CHM
+
x
+
Extract two perpendicular profiles of
the CHM centered on the tree top
Fit a polynomial on each profile
x
+
+ x
Find critical points of the fitted function
around the tree top
x
+
+
Calculate crown width on each profile as the
distance between local minimum critical
points on each side of the tree top
+
Average the crown widths along the two
profiles
Remark : Pixel size =50 x 50 cm
+ = tree location (tree top)
Processed all tree tops ?
Yes
Output crown width
No

Crown Width
Find critical points of the fitted function around the tree top
20
15
10
ox
x
ox
Horizontal
x
ox
20
15
10
o
x x x
x
ox
x
Vertical
ox
x
5
0 2.5 5 7
Fit profile
Original profile
Distance (m)
X = critical point
O = local minimum critical point
O = local maximum critical point
5
0 2.5 5 7
Distance (m)
Window size = 15 x 15 cell size
Pixel size = 0.5 m.
Crown width = 5.5 m.

Results of parameter extraction
No. TREE_ID X Y Cw H
1 1/1 681741.87 1494014.75 12.62 9.15
2 1/2 681748.87 1494023.25 10.50 9.36
3 1/3 681749.87 1494006.25 9.26 8.35
4 1/4 681749.87 1494037.25 12.00 8.31
5 10 681754.37 1494031.75 9.00 7.11
6 11/1 681759.87 1494061.25 7.26 6.86
7 11/3 681759.87 1494065.25 6.50 7.29
8 12 681761.37 1493982.75 11.00 9.23
9 13 681761.87 1494071.25 6.26 6.53

Accuracy assessment
Tree detection
Type Detected Missing Accuracy
Tree 30 tree 2 93.55%
Branch 51 branch 4 92.15%
RMSE
n
i=
1
=
∑(
X
esti
−X
N
ref
2
)
= true position from field
= predicted tree position from
LiDAR
RMS Error (m)
X 0.155
Y 0.290
H 0.300
Cw 0.281

Accuracy assessment
Tree height
LiDAR –estimated height vs. observed height
LiDAR- estimated Height estimate Height (m)
12.00
y = 0.9598x + 0.2384
RR 2 = 2 0.9679
0.96
10.00
8.00
6.00
4.00
2.00
0.00
0.00 2.00 4.00 6.00 8.00 10.00 12.00
Observed Height Height refference (m)

Accuracy assessment
Crown width
LiDAR –estimated Crown width vs. observed crown width
LiDAR- estimated Crowm width (m)
R 2 = 0.7062
16.00
14.00
12.00
10.00
8.00
6.00
4.00
2.00
0.00
0.00 2.00 4.00 6.00 8.00 10.00 12.00 14.00 16.00 18.00
Observed Crown width (m)

DBH Algometric Model
DBH=1.801Cw+1.677H-4.583
Accuracy Assessment

DBH Algometric Model
LiDAR – estimated DBH vs. Observed DBH
40.00
R 2 = 0.8194
35.00
30.00
Observed DBH (cm)
25.00
20.00
15.00
10.00
5.00
0.00
0.00 5.00 10.00 15.00 20.00 25.00 30.00 35.00 40.00
LiDAR- estimated DBH (cm)

Tree parameters extracted from LiDAR
Examples of output
ID TREEID X Y CROWN_SIZE HEIGHT DBH_LiDAR(cm)
1 1.1 681767.58 1493966.31 7.20 8.01 21.82
2 1.2 681769.45 1493964.20 9.26 8.16 25.78
3 1.3 681772.37 1493966.25 13.06 8.25 32.77
4 1.4 681770.60 1493968.78 6.38 7.44 19.38
5 2 681761.37 1493982.75 11.00 9.23 30.71
6 3 681822.37 1493991.75 7.50 7.16 20.93
7 4 681812.87 1493990.25 7.76 6.79 20.78
8 6 681785.37 1494004.25 6.76 5.39 16.63
9 7 681795.87 1493996.25 10.00 8.50 27.68
10 8.1 681759.87 1494061.25 7.26 6.86 20.00
11 8.2 681767.37 1494064.25 8.00 6.82 21.26
12 8.3 681759.87 1494065.25 6.50 7.29 19.35
13 8.4 681761.87 1494071.25 6.26 6.53 17.64
14 8.5 681765.87 1494069.75 5.76 6.18 16.15
15 9.1 681778.87 1494030.25 7.34 6.79 20.02

Trunk diameter function
Komiyama’ volume model
Vs =
×
0.724
2
0.0687
× 10 × ( DBH H
)
0.931
------------ [3]
where: DBH(m): Diameter at breast height
H(m): Tree height
Vs(m 3 ): Trunk volume
BH: Breast height = 1.3m
Vs = πC(
DBH
α = 0.931
× H )
C = 0.115826215
2
α

Assumption of the function R(x)
Finding Tree Diameter Function
Obtaining the function r(h)
R( x)
= r(
H − h)
x =
distance from the tree top to ground
x=0
R ( x)
=
x
b
ax
R(x)
R(x)=ax b
h 1
h 2
BH=1.3m
x
h
dh
r
ax b
R(x) = 0
r
dh
x=H
x
x=H
R(x)

Trunk diameter function
Equation to derive tree trunk shape radius r(h), where h: height from ground.
H
V = ∫πr
BH
( h)
dh + πDBH
2. r(
BH ) = DBH
DBH
r( BH ) = R( H − BH ) =
2
2
H H −BH
2 2 2α
α DBH
∫ π r ( h) dh = R ( x)
dx C DBH H BH
BH ∫ π = π ⋅ −π
0
4
2
H −VH
2 2b
2α
α DBH
π ∫ a x dx = πC ⋅ DBH H −π
BH
0
4
2
H −BH
2 2
2b+
1⎤
2b+
1
2α
α
⎡ a a DBH
π ⎢ x ⎥ = π ( H − BH ) = πC ⋅ DBH H −π
BH
⎣2b
+ 1 ⎦ 2b
+ 1 4
0
BH = πC(
DBH
2 2
a
2b+
1 2α
α DBH
( H − BH ) = C ⋅ DBH H − BH
2b
+ 1 4
2
2α
α DBH
2
C ⋅ DBH H − BH
a
2b
( H − BH ) =
4
2b + 1
H − BH
⎧
DBH
⎪ 2
⋅ −
a
2b
⎪ ( H − BH ) =
4
⎨2b + 1
H − BH
⎪
b DBH
⎪a( H − BH ) =
⎩
2
H )
2 2
2 α
2
2α
α
C DBH H BH
a
DBH
= ( H − BH )
2
−b
------------ [4]

Trunk diameter function
step to derive ‘b’
DBH
−b
a = ( H − BH )
2
2 2
DBH
−2b
2α
α DBH
( H − BH )
C ⋅ DBH H − BH
4 2b
( H − BH ) =
4
2b + 1
H − BH
2 2
DBH
2α
α DBH
C ⋅ DBH H − BH
4 =
4
2b + 1
H − BH
2
H − BH DBH
2b
+ 1 =
2
2α
α DBH 4
C ⋅ DBH H − BH
4
2
1 ⎛ ⎛ ( H −−BH
) DBH ⎞ ⎞
b =
−
⎜
−1
⎟
2α
α
2
2 ⎝ 4C
⋅ DBH H − DBH BH ⎠
b
1 ⎛ H − BH ⎞
= ⎜
−1
2( α −1)
α
⎟
2 ⎝ 4C
⋅ DBH H − BH ⎠
------------ [5]

Trunk Volume and Projected Area
1) Volume of a mangrove under water level (depth = d)
if water depth (d) ≤ 1.3 meter
DBH
= π
V h < d
4
2
d
if water depth (d) ≥ 1.3 meter
------------ [6]
H d
2
2
DBH a
2b−1
Vh> d
= π 1.3 + π (( H −1.3)
− ( H − d)
4 2b
+ 1
2b+
1
)
------------ [7]
2) Projected area of a mangrove under water (depth = d)
if water depth (d) ≤ 1.3 meter
A h < d
= DBH⋅
d
if water depth (d) ≥ 1.3 meter
------------ [8]
a
b
Ah> d
= DBH ⋅ d + (( H −1.3)
− ( H − d)
b + 1
2 +1 b+
1
)
------------ [9]

Trunk shape of mangrove tree
H : 8.25 m. DBH:0.33 m.
a= 0.03090, b= 0.86046
0.25
0.20
0.15
0.10
Tree Radius
Cylinder and y=ax b
Radius Radius(m)
0.05
0.00
0.0
-0.05
1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0
-0.10
-0.15
-0.20
-0.25
h or H-x: Tree height(m)

Trunk shape of mangrove tree
Case 1: y=ax^b
Case 2: Cylinder and y=ax^b
H : 8.25 m. DBH:0.33 m.
Case 1 a= 0.02283, b= 1.01715
Case 2 a= 0.03090, b= 0.86046
Radius(m) (m)
0.25
0.20
0.15
0.10
0.05
0.00
0.0
-0.05
1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0
-0.10
-0.15
-0.20
-0.25
h or H-x: Tree height(m)

Trunk shape of mangrove tree
H DBH Vs a b
7.16 0.21 0.123 0.0264 0.7775
Height level 0.5 m 1.0 m 1.5 m 2.0 m 2.5 m 3.0 m 3.5 m 4.0 m 4.5 m 5.0 m H
Volume (m 3 ) 0.017 0.034 0.051 0.067 0.080 0.091 0.100 0.107 0.113 0.117 0.124
Projected area
(m 2 ) 0.105 0.209 0.301 0.369 0.433 0.491 0.544 0.592 0.634 0.671 0.752
0.6
Trunk diameter: ax^b (m)
0.4
0.2
0
-0.2
-0.4
-0.6
0 1 2 3 4 5 6 7 8
Tree height: x (m)

Trunk shape of mangrove tree
Case 1: y=ax^b
Case 2: Cylinder and y=ax^b
Case 1 a= 0.17430, b= 1.04889
Case 2 a= 0.03090, b= 0.86046
0.25
0.20
0.15
0.10
Radius Radius(m)
0.05
0.00
0.0
-0.05
1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0
-0.10
-0.15
-0.20
-0.25
h or H-x: Tree height(m)

Trunk Volume and Projected Area
Plot trunk volume and projected area from 0.5 m to 5 m.
Height level (m) Volume(m 3 ) Projected Area (m 2 )
0.5 1.04 5.40
1.0 2.07 10.79
1.5 3.11 15.51
2.0 4.03 19.03
2.5 4.84 22.30
3.0 5.54 25.32
3.5 6.13 28.08
4.0 6.63 30.57
4.5 7.03 32.80
5.0 7.36 34.73
Total 8.18 41.35

Trunk Volume and Projected Area
The total trunk volume and projected area each water depth
1) Trunk volume at each water level 2) Projected area at each water level
8.0
40.0
7.0
35.0
6.0
30.0
Trunk volume (m 3 )
volume(m^3
5.0
4.0
3.0
2.0
Projected area (m 2 )
Projected Area (m^
25.0
20.0
15.0
10.0
1.0
5.0
0.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0
water height level depth (m) (m)
0.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0
water height level depth (m) (m)

Thank you