tropical forest monitoring using synthetic … › jtetukoss › documents › jtetukoss...tropical...
TRANSCRIPT
Tropical Forest Monitoring using Synthetic Aperture Radar
- Theories and Applications -
January 2002
Josaphat Tetuko SRI SUMANTYO
Graduate School of Science and Technology
Chiba University
ii
If you download/copy/refer this manuscript,
Please, donate for Indonesian Children Scholarship: *)
Pandhito Panji Foundation
c/o Innes Indreswari (Executive Secretary)
TAPLUS BNI 1946, ITB Branch
Jalan Ganesha 10 Bandung 40132 Indonesia
No. Account 236.003094537.901
*) This announcement is not including in real dissertation.
iii
iv
v
Declaration
This document certifies that the research and its results in this dissertation
have never been submitted elsewhere for an award of any degree or
diploma.
Chiba, January 2002
Candidate
(Josaphat Tetuko SRI SUMANTYO)
Supervisors
(Prof. Dr. Nobuo TAKEUCHI) (Ass. Prof. Dr. Ryutaro TATEISHI)
vi
vii
To my lovely wife Innes Indreswari, son Johannes Pandhito Panji Herdento,
parents Michael Suman Juswaljati and Florentina Srindadi
for their love, support and encouragement.
Chiba, January 2002
viii
ix
Abstract
Recently, remote sensing technology, especially Synthetic Aperture Radar (SAR)
sensor, has been an efficient and helpful tool to monitor tropical forest area where always in
cloudy condition. However, SAR data are not easily interpreted due to the complex relations
of the radar scattering mechanism between microwaves and tropical forest parameter (i.e.
diameter of tree trunk, thickness of burnt coal seam, surface roughness and soil moisture).
Hence, in this research, numerical methods were developed to analyse the interaction of L
Band microwaves with a tree trunk of tropical forest and burnt coal seam of forest fire scars.
In analysis of scattered wave from a tropical tree trunk, the proposed method
approximates a trunk as an infinite length of two and three layers of cylindrical dielectric
media. These layers are skin and heartwood; and skin, xylem and heartwood. The
horizontally and vertically polarised scattered fields are derived in order to calculate the
relationship between trunk diameter and backscattering coefficient. The analysis result is
confirmed by simulating the scattered wave from a tree trunk using Finite Difference Time
Domain (FDTD) method. The model uses the equations of scattered electromagnetic fields
that are derived from Maxwell’s equations. Both analysis and simulation results are similar.
Then the relationship is used to estimate tree trunk diameters of pine forest around Saguling
lake and tropical forest at Mount Gede Pangrango National Park, west Java, Indonesia from
Japanese Earth Resources Satellite (JERS-1) SAR data.
In analysis of scattered wave from burnt coal seam, two types of methods (simple and
complicated) are conducted to analyse scattered waves from burnt coal seam in order to
estimate thickness of forest fire scars. The model is composed of three media namely; free
space (air), burnt coal seam and peat (a perfectly conductor). For computation purposes, the
equivalent circuit of this model is conducted using classical transmission line circuit method
for a simple analysis, and the advanced stationary-phase approximation is used to analyse
scattered wave from complicated rough burnt coal seam. The relationship between
backscattering coefficient and thickness of burnt coal seam is obtained. The analysis result is
confirmed by simulation using FDTD method. The simulation is done using a two-
dimensional finite-difference model for scattered waves from the burnt coal seam. Both
analysis and the simulation results are similar. Subsequently, the developed model is applied
to estimate the thickness of burnt coal seam in central Borneo fire events in 1997 using
JERS-1 SAR data. The estimated result agrees with ground measurement that was collected
in period of 1995 to 1997.
x
xi
xii
xiii
Acknowledgements
The author wishes to express his gratitude to the members of his committee and to the
Department of Geoinformation Analysis and Department of Sensor and Atmospheric
Radiation, Center for Environmental Remote Sensing (CEReS), Chiba University for
supporting this research. Special thanks are due to Associate Professor Ryutaro TATEISHI
and Professor Nobuo TAKEUCHI for their continuing guidance and invaluable support
throughout the course of this research.
He would also like to thank the following friends and colleagues for their friendship and
enlightening conversations during his graduate studies: Dr. Peter Gunin, Dr. Nicolai Kharin,
Dr. WEN Cheng-Gang, Dr. ZHU Lin, Dr. PARK Jong-Hyun, Dr. Masayuki MATSUOKA, Dr.
PARK Jong-geol, Dr. Ketut Wikantika, Dr. Hussein HARAHSHEH, Dr. Kamal Sarabandi,
WU Jian-yu, Shin ICHIHASHI, ZHANG Xian-Ji, Satoshi AKAGAWA, KIM Dong-Hee,
Renchin Tsolmon, Hiroshi SATOH, Hussam AL-BILBISI, ABAN Jose Edgardo, Alessandra
DE CONTI, Ngigi Thomas CATHUNGU, Mitsuhiko EBATA, Yasunobu SHIMAZAKI,
Hokuto KANO, Jun OSOZAWA, Mohamed ABOEL GHAR, Adi Junjunan Mustafa, Yosuke
ORISHIMO, Rokhmatuloh, Faten Ali Kayed KHRAISHA, Hideyuki HANAWA, Kentaro
HARAGUCHI, and Wihartini NAZORI.
He also grateful to Dr. Kasdi Subagyo and Nuraini (Centre for Soil and Agroclimate
Research) for ground data of central Borneo; Dr. Abdul Hadi (Universitas Lambung
Mangkurat) and Prof. Kenichiro YASHIRO for valuable discussion; Maksum (Cicurug
National Park), Lucia Tri Erowadanti SRI SUMANTYO (Universitas Sebelas Maret) for
tropical tree samples; Franciscus Dwi Koco SRI SUMANTYO (Sarana Putra Makmur) and
Pandhito Panji Foundation (Research Center and Indonesian Databank) for ground data,
topographic and digital maps of the study area in this study; Prof. Koichi ITO, Dr. Ichiro IDA
xiv
and Dr. Kazuyuki SAITO (Chiba University) for their assistance to measure the dielectric
constant of tropical forest samples; MITI/NASDA for the courtesy of JERS-1 VNIR and SAR
data; CNES for SPOT-HRV data; BAKOSURTANAL for topographic and Digital Elevation
Model (DEM) data; Pandhito Panji Foundation (PPF), Okamoto Scholarship Foundation
(OSF), Satoh International Scholarship Foundation (SISF) and Atsumi International
Scholarship Foundation (AISF) for supporting this research.
xv
Contents
Title (in English) i
Title (in Japanese) iii
Declaration v
Dedication vii
Abstract (in English) ix
Abstract (in Japanese) xi
Acknowledgements xiii
Contents xv
List of Appendices xix
List of Figures xxi
List of Tables xxviii
Chapter I. Introduction
1.1 Motivations and objectives 1
1.2 Deforestation and forest fire 2
1.3 Japanese Earth Resources Satellite (JERS-1) Synthetic Aperture Radar 6
1.4 Physical characteristic measurements 8
References 13
Chapter II. Analysis of Scattered Waves from Two Layers of Tree Trunk
2.1 Introduction 17
2.2 Analysis 18
2.3 Simulation 22
2.4 Results and discussion 29
2.5 Application 33
2.5.1 Study area 33
2.5.2 Data processing 33
2.6 Conclusions 37
References 38
xvi
Chapter III. Analysis of Scattered Waves from Three Layers of Tree Trunk
3.1 Introduction 41
3.2 Analysis 42
3.3 Simulation 47
3.4 Results 49
3.5 Application 55
3.5.1 Study area 55
3.5.2 Data processing 60
3.6 Conclusions 63
References 65
Chapter IV. Analysis of Scattered Waves from Burnt Coal Seam
4.1 Introduction 67
4.2 Analysis 78
4.3 Simulation 82
4.4 Results 84
4.5 Application 89
4.5.1 Study area 89
4.5.2 Data processing 94
4.6 Conclusions 96
References 98
Chapter V. Analysis of Scattered Waves from Rough Burnt Coal Seam
5.1 Introduction 101
5.2 Analysis 102
5.2.1 Scattered fields on burnt coal seam surface (1) or interface 1 106
5.2.1.1 Scattering field on medium 1 106
5.2.1.2 Scattering field on medium 2 108
5.2.2 Scattering field on peat surface or interface 2 110
5.2.3 Scattering field on burnt coal seam surface (2) or interface 1 113
5.2.3.1 Scattering field on medium 2 113
5.2.3.2 Scattering field on medium 1 115
5.2.4 Scattering coefficient 117
xvii
5.3 Results and Discussion 124
5.4 Application 126
5.4.1 Study area 126
5.4.2 Data processing 126
5.5 Conclusions 128
References 129
Chapter VI. Summary and Recommendations
6.1 Summary 131
6.2 Future work and Recommendations 132
Appendices 135
Publication list 183
Biography 187
xviii
xix
List of Appendices
Appendix A Analysis of Scattered Waves on Two Layers of Tree Trunk (TM
mode)
135
Appendix B Analysis of Scattered Waves on Three Layers of Tree Trunk (TM
mode)
139
Appendix C Finite Difference Time Domain Method (TE mode) 143
Appendix D Ground Data Around Opening Peatland Area (One Million
Hectares Peatland Project, central Borneo, Indonesia)
149
Appendix E Wave Analysis in Cylindrical Coordinate System (TM mode) 157
Appendix F Ground Data : Mount Gede Pangrango National Park 159
Appendix G Derivation of the Scattered Fields in the Medium 1 at Air and
Burnt Coal Seam Interface
165
Appendix H Derivation of the Scattered Fields in the Medium 2 at Burnt Coal
Seam and Peat Interface
169
Appendix I Derivation of the Scattered Fields in the Medium 2 on Air and
Burnt Coal Seam Interface
173
Appendix J Derivation of the Scattered Fields in the Medium 1 on Air and
Burnt Coal Seam Interface
177
Appendix K Derivation of Horizontally and Vertically Polarized Surface-
Current Density
181
xx
xxi
List of Figures
Figure 1.1 Distribution of tropical forest at Indonesia archipelago 2
Figure 1.2 The spread of forest plantations across Indonesia and the species
planted in each province
3
Figure 1.3 Borneo’s forest cover and 1997-1998 fire hot spots 5
Figure 1.4 Instruments onboard on JERS-1 satellite (source: NASDA) 7
Figure 1.5 Synthetic Aperture Radar on JERS-1 satellite (source: NASDA) 7
Figure 1.6 Photograph of dielectric constant measurements
(a) Tree trunk 9
(b) Burnt coal seam 9
Figure 1.7 Dielectric constant of burnt coal seam 10
Figure 1.8 (a) Dielectric constants measurement of tropical forest tree trunk
(skin): r-real part and i-imaginary part
11
(b) Dielectric constants measurement of tropical forest tree trunk
(xylem): r-real part and i-imaginary part
12
Figure 2.1 Photograph and Geometry of the analysis.
(a) Photograph of a cross section of pine (Pinus merkusii) trunk. 19
(b) Geometry of scattered waves from a pine trunk. 19
Figure 2.2 Simulation model. 25
Figure 2.3 Pulse of incident wave.
(a) Gaussian pulse. 27
(b) Fast fourier transformed Gaussian pulse. 27
Figure 2.4 Geometry of incident wave 28
xxii
Figure 2.5 Scattered waves in simulation space from tt ∆= 50 s to t∆300 s. A
and B are scattered waves from skin and heartwood, C and D are
scattered wave from trapped waves in skin layer, E is forwarded wave
that occurred by clipping pulse that flows on the trunk surface and
scattered to back of trunk, F and G are heartwood and skin,
respectively. P is the observed point.
30
Figure 2.6 Scattered waves at observed point P: A and B, C and D are scattered
pulse from skin and heartwood, and trapped wave in the skin layer,
respectively.
31
Figure 2.7 Relationship between tree trunk diameter and backscattering
coefficient
32
Figure 2.8 Map of the study area 34
Figure 2.9 Photograph of pine forest in the study area and the supervised
classification results of JERS-1 SAR data (path 106, row 312, 13
May 1997).
35
Figure 3.1 Tree trunk media 42
Figure 3.2 Geometry of analysis 43
Figure 3.3 Geometry of simulation space. Remarks: P is observed point. A, B
and C are heartwood, xylem, and skin, respectively. Simulation space
is divided into INX x INY grids of meshes.
48
Figure 3.4 Distribution of scattered electric field intensity SyE with tt ∆= 50 to
t∆300 s.
51
xxiii
Figure 3.5 Distribution of scattered electric field SyE in tt ∆= 300 s, where A, B
and C are scattered wave from skin, xylem and heartwood,
respectively. P is observed point. E, F and G are skin, xylem and
heartwood, respectively. D is forwarded wave that is occurred by
clipping wave that flows on the trunk surface and scattered to the
backward of tree trunk.
52
Figure 3.6 Scattered electric field intensities at observed point P. A, B and C are
scattered pulse from skin, xylem and heartwood, respectively.
53
Figure 3.7 Analysis and simulation results for four species of Indonesian tropical
forest, where the diameter of tree trunk is equal with 2c, where c is
radius of tree trunk.
54
Figure 3.8 JERS-1 VNIR data of the study area (Path 107 Row 312, 19970930):
Gede Pangrango National Park, west Java, Indonesia. Remark: A and
B show northern and southern part of the National Park, respectively.
56
Figure 3.9 JERS-1 SAR data of the study area (Path 107 Row 312, 19970810):
Gede Pangrango National Park, west Java, Indonesia. Remark: A and
B show northern and southern part of the National Park, respectively.
57
Figure 3.10 Altitude distribution of the study area : Mount Gede Pangrango
National Park, west Java, Indonesia.
58
Figure 3.11 Location of the study area: Gede Pangrango National Park (part of
area A in figure 3.8 and 3.9)
59
Figure 3.12 Classification result assigned the distribution of classes in the study
area. Test area shows classes distribution in ecosystem zones and its
terrain conditions. Ecosystem zones are settlement and paddy (sp),
sub-montane (sm) and montane (mt).
61
xxiv
Figure 4.1 Photographs of field survey expeditions in period 1995 to 1997. A
and B show main vegetations that found around study sites;
Tengkawang (Dipterocarpaceae spp.) and purun grass, respectively.
C shows burnt forest that remained burnt tree trunk and burnt coal
seam. D shows staffs measured thickness of coal seam.
68
Figure 4.2 (a) Digital map of the study area: One Million Hectares Peatland
Project (PLG), central Borneo, Indonesia (DEPHUTBUN 1999).
69
(b) Study area: master plan of ‘One Million Hectares Peatland Project
(PLG)’ at central Borneo, Indonesia. This figure shows thickness
of coal seam that collected in field survey expeditions in period
1995 to 1997. Dotted lines show the area covered by JERS-1 SAR
and SPOT HRV data.
70
(c) Distribution of the thickness of coal seam: One Million Hectares
Peatland Project, District B (PLG-B) at central Borneo, Indonesia
71
(d) Distribution of the thickness of coal seam: One Million Hectares
Peatland Project, District D (PLG-D) at central Borneo, Indonesia
72
Figure 4.3 SPOT HRV data of fire events in the study area
(a) 6 June 1997 (prior to the fire). 74
(b) 29 July 1997 (during fire). 75
(c) 7 August 1997 (during fire). 76
(d) 8 September 1997 (after fire). 77
Figure 4.4 Geometry of analysis
(a) Analysis model. Remarks: ① : burnt coal seam-obstacles
scattering; ② : obstacles – burnt coal seam scattering
79
(b) equivalent circuit 79
xxv
Figure 4.5 Geometry of wave propagation in two media. 80
Figure 4.6 Measurement of burnt coal seam sample using dielectric constant kit
HP85070B (see sub-figure)
80
Figure 4.7 Simulation model. 83
Figure 4.8 Scattered waves in simulation space. 85
Figure 4.9 Intensity of scattered wave in observed point Q.
(a) Electric field SyE 87
(b) Magnetic field SzH 87
Figure 4.10 Relationship between burnt coal seam thickness and backscattering
coefficient in two dimensional analysis and simulation.
88
Figure 4.11 (a) Raw data of JERS-1 SAR: 15 May 1996 90
(b) Raw data of JERS-1 SAR: 3 February 1997 91
(c) Raw data of JERS-1 SAR: 29 July 1997: dotted line is the study
area in this study
92
Figure 4.12 Composite of JERS-1 SAR data: red – 29 July 1997, green – 3
February 1997, blue – 15 May 1996
93
Figure 4.13 A SPOT-HRV data and supervised classification results of a JERS-1
SAR data (path 95, row 305, 27 July 1997).
95
Figure 5.1 Geometry of the scattered waves analysis. 103
Figure 5.2 Relationship between the backscattering coefficient and the thickness
of burnt coal seam.
125
Figure A.1 Geometry of analysis of scattered TM mode wave (two layers) 135
Figure B.1 Geometry of analysis of scattered TM mode wave (three layers) 139
Figure C.1 Portion of the finite-difference grid. 145
Figure D.1 Opening of peatland area at central Borneo, August 1996 (a). 149
xxvi
Figure D.2 Opening of peatland area at central Borneo, August 1996 (b). 149
Figure D.3 Vegetations around peat swamp at Mentangai river near Bunter lake
‘One million hectares peatland project (PLG)’, August 1996.
150
Figure D.4 Vegetation at peatland (± 8m) around Kurun river (Black water
river), PLG area, August 1996.
150
Figure D.5 Converting peatland area to be agricultural area at Dadahup, central
Borneo, August 1996.
151
Figure D.6 Canals at Tabukan, central Borneo, August 1996. 151
Figure D.7 Staffs are boring peatland to explore the depth and type of peatland at
central Borneo, August 1996.
152
Figure D.8 Converted peatland area at Siantan, west Borneo, 1995. 152
Figure D.9 Peatland identification with ± 8m pipe at backswamp around Kurun
river (black water river), central Borneo, August 1996.
153
Figure D.10 Burnt tengkawang and pule grass around Bunter lake, reaches of
Mentangai river, central Borneo, August 1996.
153
Figure D.11 Purun grass as main vegetation at backswamp (Kurun river) to
indicate that this area is ferrit land, August 1996.
154
Figure D.12 Burnt peatland at Berengbengkel, Kecamatan Pakandut, Kodya
Palangkaraya, central Borneo, 1995.
154
Figure D.13 Paddy field at Sakalagun, central Borneo, 1995. 155
Figure F.1 Map of field survey that was held in July 2000. 159
Figure F.2 Mount Gede (right) and Pangrango (left), west Java, Indonesia (figure
F.1 ①)
160
xxvii
Figure F.3 Hill of Mount Gede (900m asl), Rasamala forest, rock, and tea
plantation around Cipelang river that flows from Mount Gede to
Selabintana, Sukabumi (figure F.1 ②).
160
Figure F.4 Tea plantation around Mount Gede at Pasiripis district, Pondok
Halimun, Cipelang Selabintana Sukabumi (figure F.1 ③).
161
Figure F.5 Hill of Mount Gede with variation of grass, fern, and rasamala.
Pondok Halimun district in Gede Pangrango National Park. Land
surface around this area is wavy (figure F.1 ④)
161
Figure F.6 Tea and tobacco plantation at Wanasari district, Sukabumi, with
background Mount Gede Pangrango (figure F.1 ⑤).
162
Figure F.7 Rasamala forest and tea plantation near Cipelang river (850m
asl)( figure F.1 ⑥)
162
Figure F.8 Variation of Montane forest, fern, moss and bush (950m asl)( figure
F.1 ⑦)
163
Figure F.9 Small tree trunk of rasamala that nominates hill of Mount Gede
Pangrango (1030m asl) in Cicurug National Park (figure F.1 ⑧).
163
Figure F.10 Rasamala (Altingia exelsa) in Mount Gede Pangrango with diameter
50 – 200cm and height about 7 – 20m. It grows at attitude 700 –
1750m asl (figure F.1 ⑨).
164
xxviii
List of Tables
Table 1.1 Specification of JERS-1 Synthetic Aperture Radar (SAR) 8
Table 1.2 Specification of the dielectric probe kit HP85070B 9
Table 1.3 List of tropical forest species at Indonesia 10
Table 2.1 Classification and estimation results 36
Table 3.1 Dielectric constants of Indonesian tropical forest trees at the
frequency of JERS-1 SAR ( 275.1=f GHz)
47
Table 3.2 Relationship between backscattering coefficients and tree trunk
diameters of rasamala forests in the study area
62
Table 4.1 Thickness of burnt coal seam in the study area 96
Table 5.1 Thickness of burnt coal seam in the study area 127
1
Chapter I
Introduction
1.1. Motivations and objectives
Tropical forest is covering less than 7% of the terrestrial surface and has the important
role in carbon cycle (Tucker et al. 2000) and mega-biodiversity are living inside, especially in
Indonesian tropical forest (Tateishi et al. 2000). Figure 1.1 shows the tropical forest
distribution at Indonesia (WRI 1999 and DEPHUTBUN 1997). The type of tropical forest that
cover Indonesian area are lowland forest, montane forest, mangrove forest, swamp forest, and
lowland monsoon forest. Kyoto protocol (UN 1997) was signed in 1997, but deforestation is
being still done in many places in the world, especially at tropical forest area. Twelve million
hectares of tropical forest was cleared annually (FAO 1997). Another problem is leak of tools
and methods to manage and monitor the tropical forest in a large area. Indonesia is large
archipelago that consists of more than 1,700 large islands with land territory covers
approximately 1.9 millions square kilometres. It is very difficult to estimate age or volume of
tree trunk using conventional techniques that spent much time and cost to collect its ground
data.
Recently, remote sensing technology has been an efficient and helpful tool to monitor
tropical forest and plantation in a large area. The main problem in monitoring tropical areas,
as Indonesia, is cloudy condition. The best instrument to monitor these areas is Synthetic
Aperture Radar (SAR), as it works effectively in spite of cloudy conditions. However, SAR
data are not easily interpreted due to the complex relations of the radar backscattering
mechanism between microwaves and tropical forest parameter (i.e. diameter of tree trunk,
thickness of burnt coal seam, roughness of soil surface, soil moisture) (Kamal 1989). In this
2
Figure 1.1. Distribution of tropical forest at Indonesia archipelago
study, the author attempts to find the relationship between the radar backscattering
coefficients and its parameters that are found in tropical forest area. Figure 1.2 shows the
spread of forest plantation types across Indonesia and the species planted in each province
with major species are Acacia mangium, Agathis spp., Pinus merkusii, Gmelina arborea,
Shorea spp, Swietenia macrophylla etc (Nair 2000). In this research, the author also applies
the proposed method to monitor these species using remote sensing techniques or SAR sensor.
1.2. Deforestation and forest fire
Deforestation in tropical regions went at a quite stable pace during 1980 – 1995, of
0.7% per year, which is about 12 million hectares annually (including reforestation) of
tropical forest is lost each day. This boils down to 33,000 hectare per day. Nearly half of all
species of plants and animals on earth are living in tropical forests. According to some experts
something like 100 species become extinct each day, and most of them as a direct result of
deforestation. Indonesia has second deforestation speed in the world, so it has high
South China Sea
Indian Ocean Sahul Shelf
Pacific Ocean
Sunda Shelf
Sulawesi Sea
Java Sea
Malaysia
Malaysia
Philippine
Legend
Lowland forest
Montane forest >1000m) Mangrove forest
Swamp forest
Non-productive dryland Non-productive wetland
Agricultural area Plantation Other landcovers
Lowland monsoon forest
N
3
Figu
re 1
.2. T
he s
prea
d of
fore
st p
lant
atio
ns a
cros
s In
done
sia
and
the
spec
ies
plan
ted
in e
ach
prov
ince
Eas
t K
alim
anta
n
Aca
cia
man
gium
E
ucal
yptu
s sp
p.
Eus
ider
oxyl
on z
wag
eri
Gm
elia
arb
orea
Pa
rase
rian
thes
fal
cata
ria
Pero
nem
a ca
nesc
ens
Shor
ea s
pp
Swie
teni
a m
acro
phyl
la
Tec
tona
gra
ndis
Cen
tral
Su
law
esi
Och
rom
a sp
p Pa
rase
rian
thes
Cen
tral
Su
law
esi
Och
rom
a sp
p Pa
rase
rian
thes
Mol
luca
s A
caci
a m
angi
um
Gm
elin
a ar
bore
a Pa
rase
rian
thes
fal
cata
ria
Sou
th K
alim
anta
n
Aca
cia
man
gium
E
ucal
yptu
s sp
p E
usid
erox
ylon
zw
ager
i G
mel
ina
arbo
rea
Para
seri
anth
es f
alca
tari
a Pe
rone
ma
cane
scen
s Sh
orea
spp
. T
ecto
na g
rand
is
Mol
luca
s A
caci
a m
angi
um
Gm
elin
a ar
bore
a Pa
rase
rian
thes
fal
cata
ria
Eas
t N
usa
Ten
ggar
aPa
rase
rian
thes
fal
cata
ria
Tec
tona
gra
ndis
Cen
tral
Kal
iman
tan
Aca
cia
man
gium
A
ntho
ceph
alus
spp
D
yera
spp
G
mel
ina
arbo
rea
Koo
mpa
ssia
spp
O
chro
ma
spp
Para
seri
anth
es f
alca
tari
a P
eron
ema
cane
scen
s Sh
orea
spp
.
Wes
t K
alim
anta
n
Aca
cia
man
gium
A
lsto
nia
spp
Euc
alyp
tus
spp
Gm
elin
a ar
bore
a Pa
rase
rian
thes
fal
cata
ria
Pero
nem
a ca
nesc
ens
Shor
ea s
pp
Ria
u
Aca
cia
man
gium
O
ther
Aca
cia
spp
Ant
hoce
phal
us s
pp
Aza
dira
chta
exc
elsa
G
mel
ina
arbo
rea
Mae
sops
is e
min
ii O
ctom
eles
sum
atra
na
Para
seri
anth
es f
alca
tari
a Pe
rone
ma
cane
scen
s
D.I
. Ace
h
Aca
cia
man
gium
E
ucal
yptu
s sp
p Pa
rase
rian
thes
fal
cata
ria
Pinu
s m
erku
sii
Nor
th S
um
ater
a A
caci
a m
angi
um
Oth
er A
caci
a sp
p A
ntho
ceph
alus
spp
A
zadi
rach
ta e
xcel
sa
Dua
bang
a co
lucc
ana
Euc
alyp
tus
spp
Gm
elin
a ar
bore
a M
aeso
psis
em
inii
Para
seri
anth
es f
alca
tari
a Pe
rone
ma
cane
scen
s Pi
nus
mer
kusi
i Sw
iete
nia
mac
roph
ylla
Wes
t S
um
ater
a G
mel
ina
arbo
rea
Para
seri
anth
es f
alca
tari
a
Jam
bi
Aca
cia
man
gium
O
ther
Aca
cia
spp
Dye
ra s
pp
Euc
alyp
tus
spp
Gm
elin
a ar
bore
a G
onys
tylu
s ba
ncan
us
Para
seri
anth
es f
alca
tari
a Pe
rone
ma
cane
scen
s Sh
orea
spp
Sou
th S
um
ater
a A
caci
a m
angi
um
Als
toni
a sp
p G
mel
ina
arbo
rea
Para
seri
anth
es f
alca
tari
a Sh
orea
spp
Lam
pu
ng
Aca
cia
man
gium
G
mel
ina
arbo
rea
Para
seri
anth
es f
alca
tari
a Sh
orea
spp
.
Wes
t ja
va
Aca
cia
man
gium
A
gath
is s
pp
Man
grov
e sp
ecie
s Pi
nus
mer
kusi
i Sh
orea
spp
. Sw
iete
nia
mac
roph
ylla
T
ecto
na g
rand
is
Sou
thea
st S
ula
wes
iA
caci
a m
angi
um
Gm
elin
a ar
bore
a Sw
iete
nia
mac
roph
ylla
T
ecto
na g
rand
is
Sou
th S
ula
wes
i G
mel
ina
arbo
rea
P
aras
eria
nthe
s fa
lcat
aria
Eas
t Ja
va
Aga
this
spp
M
elal
euca
caj
uput
i Pa
rase
rian
thes
fal
cata
ria
Pinu
s m
erku
sii
Schl
eich
era
oleo
sa
Swie
teni
a m
acro
phyl
la
Tec
tona
gra
ndis
Cen
tral
Jav
a
Aga
this
spp
M
elal
euca
caj
uput
i Pa
rase
rian
thes
fal
cata
ria
Pinu
s m
erku
sii
Schl
eich
era
oleo
sa
Swie
teni
a m
acro
phyl
la
Tec
tona
gra
ndis
Mol
ucca
s
Cen
tral
Su
law
esi
Sout
heas
t Su
law
esi
Sout
h
Sula
wes
i
Eas
t
Nus
a T
engg
ara
Eas
t
Kal
iman
tan
Cen
tral
K
alim
anta
n
Sout
h
Kal
iman
tan
Wes
t
Kal
iman
tan
Eas
t Ja
va
Cen
tral
Ja
va
Wes
t Ja
va
Lam
pung
Sout
h Su
mat
era
Jam
bi
Wes
t
Sum
ater
a
Ria
u
Nor
th
Sum
ater
a
D.I
. Ace
h
4
contribution in environmental destruction and indirectly influence to the carbon dioxide
production over the earth (WRI 1999). In this reason, the author attempts to analyse the
scattered waves from burnt coal seam to estimate the thickness of burnt coal seam. In near
future, this result can be applied to estimate the carbon dioxide volume that is produced by
deforestation or forest fire area, especially at Indonesian region.
The large-scale forest fire was occurred in Borneo island, Indonesia in period of 1997 to
1998 (Nakayama et al. 1999, Siegert et al. 2000, Liew et al. 1999). This is coinciding with the
El-nino Southern Oscillation (ENSO) of 1997-1998 (Muraleedharan et al. 2000a). It has
become air pollution episode due to the out-of-control biomass burning for agricultural
purposes started in June 1997 that to be a severe environmental problem for itself and the
neighbouring countries. Its impact on the health and ecology in the affected areas is expected
to be substantial, costly and possibly long lasting (Fang et al. 1999). Figure 1.3 shows the
distribution of forest fire at Borneo island, Indonesia in 1997 – 1998, where about one million
hectares of forest area was devastated (Charles et al. 2000).
Many researchers attempted to know hotspot of forest fire and quantity of air pollution
using many sensors with study sites were countries around Indonesia (Fang et al. 1998,
Miroslav et al. 1999, Wolfgang et al. 1999). Biomass fires have been responsible for several
regional haze episodes in southeast Asia, and most of the fires were in areas where peat is the
dominant biomass fuel (Muraleedharan et al. 2000b). Muraleedharan reported that the
chemical components present in the haze in southeast Asia with special emphasis on those
substances with potential health impacts (Muraleedharan et al. 2000a). Based on these reasons,
this research attempts to analyse the relationship between the thickness of burnt coal seam and
radar backscattering coefficient. Then the result will be applied to estimate the thickness of
burnt coal seam using Synthetic Aperture Radar (SAR) data. The study area was chosen
5
Figure 1.3. Borneo’s forest cover and 1997-1998 fire hot spots
1997
1998
Frontier forest under low threat
Threatened frontier forest
Non-frontier forest
Others
6
around ‘One Million Hectares Peatland Project (PLG)’ at central and south Borneo area. To
realise this purpose, Japanese Earth Resources Satellite (JERS-1) SAR data (L band) were
employed.
1.3. Japanese Earth Resources Satellite (JERS-1) Synthetic Aperture Radar
JERS-1 is a Japanese National Space Development Agency (NASDA) satellite whose
mission objectives of JERS-1 are twofold; Firstly, the assessment of the newly developed
onboard sensors and the spacecraft itself, and secondly, the establishment of an integrated
system for observing the Earth that is focused on observations of earth resources, geology,
agriculture, forestry, land use, sea ice monitoring and coastal monitoring.
The spacecraft contains two instruments: a Synthetic Aperture Radar (SAR) and an
Optical Sensor (OPS), see figure 1.4. The SAR is an active microwave sensor that transmits
microwave and detects the wave that is reflected back by objects (see figure 1.5 and table 1.1).
It enables fine-resolution, high contrast observation and accurate determinations of
topographical features. Since it is totally reliant on microwave data (1.275 GHz), it operates
independent of weather conditions and cloud cover. The OPS separates the light reflected
from the ground into seven spectral bands from visible to short wave infrared. It is made up of
two sensors: the Visible and Near Infrared Radiometer (VNIR) and the Short Wave Infrared
Radiometer (SWIR).
JERS-1 was launched on 11 February 1992 from Tanegashima Space Centre in
Kagoshima, Japan on a 2 stage H-1 launch vehicle. The satellite had approximate dimensions
1.0m x 1.8m x 3.1m with payload weights approximately 1.4 tons. The spacecraft has a solar
array that is approximately 8.0m x 3.4m. JERS-1 had an onboard Mission Data Recorder
(MDR) that allows it to collect data even when a ground station is not in view. JERS-1 had
already exceeded its 2 years design life until 11 October 1998.
7
Figure 1.4. Instruments onboard on JERS-1 satellite (source: NASDA)
Figure 1.5. Synthetic Aperture Radar on JERS-1 satellite (source: NASDA).
R
R′
8
Table 1.1. Specification of JERS-1 Synthetic Aperture Radar (SAR)
Specification
Flight attitude 568km Flight inclination angle 97.7o
Frequency 1.275GHz (L band) Wavelength 23.5cm Polarisation HH Off nadir angle 35o
Incidence angle 38.7o
Swath width 75km Azimuth resolution 18m (3 looks) Range resolution 18m Peak Power 325 W (specification 1.3kW) Band width 15MHz Antenna size 2.2m x 12m
1.4. Physical characteristic measurements
Table 1.2 shows specification of the dielectric probe kit HP85070B (HP 2000). The
dielectric constant of each sample is important parameter in modelling and analysis of the
relationship between backscattering coefficient and tropical forest characteristics (i.e. tree
trunk diameter, thickness of burnt coal seam, roughness of soil surface, and soil moisture).
Figure 1.6 shows photographs of dielectric constant measurements that were done by the
author to obtain the dielectric constant of samples, where figure 1.6 (a) and (b) shows the
measurement of dielectric constant of tropical forest tree trunk media and burnt coal seam,
respectively. In field survey that was done in 1999, the tree trunk samples of 17 species of
tropical forest plants were collected in Indonesia. Table 1.3 shows the list of these samples.
The measurement was done in frequency range from 0.3 to 3GHz. Figure 1.7 and 1.8 shows
the measurement results from samples of burnt coal seam and tropical forest tree trunks,
respectively. Tree trunk media were skin, xylem and heartwood. The dielectric constant of
9
Table 1.2. Specification of the dielectric probe kit HP85070B
Specification
Band width 200 MHz ~ 20 GHz
Operating temperature - 40 ~ +200oC
Probe 3.5mm connector type
Flexible cable 1m length
Terminal 50Ω Open / Short / Load: 3.5mm connector
Probe stand 24inch, diameter 1/2inch
(a) Tree trunk (b) Burnt coal seam
Figure 1.6. Photograph of dielectric constant measurements
each medium was measured, but the results of heartwood measurement were similar with
water. Consequently, figure 1.8 shows results from the measurement of skin and xylem only.
10
Table 1.3. List of tropical forest species at Indonesia
No Trade name Botanical name Family 1 Acacia Acacia mangium - 2 Coconut Cocos mucifera - 3 Mahagony Swietenia mahagony - 4 Rasamala Altingia exelsa noronhae Hamamelidaceae 5 Pine Pinus Merkusii Pinaceae 6 Rattan Calamus - 7 Tekik Parasianthes lebbeck Benth. Mimosaceae 8 Tamarind Tamarindus indica - 9 Teak Tectona grandis - 10 Orange Citrus aurantium sinensis - 11 Petai Parkia Speciosa Hassk Mimosaceae 12 Mango Magnifera indica Anacardiaceae 13 Mete Anacardium occidentale - 14 Mulwo - - 15 Munggur Enterolobium saman prain - 16 Randu (capok tree) Cinnamomum spp. Lauraceae 17 Sengon Albizzia chinensis Merr. Mimosaceae
Figure 1.7. Dielectric constant of burnt coal seam
-1
0
1
2
3
4
0.24 0.84 1.44 2.04 2.64
Frequencies [GHz]
Die
lect
ric c
onst
ants
real
imaginary
11
0
1
2
3
4
5
6
7
0.240.54
0.84
1.14
1.44
1.74
2.04
2.34
2.642.94
Frequency [GHz]
Die
lect
ric c
onst
ant
acacia (r) acacia (i)
coconut (r) coconut (i)
mahogany (r) mahogany (i)
rasamala (r) rasamala (i)
pine (r) pine (i)
rattan (r) rattan (i)
tekik (r) tekik (i)
tamarind (r) tamarind (i)
teak (r) teak (i)
orange (r) orange (i)
petai (r) petai (i)
mango (r) mango (i)
mete (r) mete (i)
mulwo (r) mulwo (i)
munggur (r) munggur (i)
randu (r) randu (i)
sengon (r) sengon (i)
Figure 1.8 (a) Dielectric constants measurement of tropical forest tree trunk (skin): r – real
part and i – imaginary part.
i
r
12
0
2
4
6
8
10
12
14
16
18
0.24 0.84 1.44 2.04 2.64
Frequency [GHz]
Die
lect
ric c
onst
ant
teak (r)
teak (i)mahagony (r)
mahagony (i)pine (r)pine (i)
rasamala (r)rasamala (i)
Figure 1.8 (b) Dielectric constants measurement of tropical forest tree trunk (xylem): r – real
part and i – imaginary part
i
r
13
References
1. CHARLES V. B., JAMES S., 2000, Trial by Fire, Forest Fire and Forestry Policy in
Indonesia’s Era of Crisis and Reform. World Resource Institute (WRI), Forest Frontiers
Initiatives in Collaboration with WWF-Indonesian & Telapak Indonesia Foundation
(Washington: WRI).
2. DEPHUTBUN, 1997, Land Use by Concensus Map, Directorate General Forest Inventory
and Land Use Planning, Indonesian Ministry of Forestry and Estate (Jakarta: dephutbun).
3. FANG, M., HUANG, W., 1998, Tracking the Indonesian forest fire using
NOAA/AVHRR images, International Journal of Remote Sensing, Vol.19, No.3, pp.387-
390 (London: Francis and Taylor).
4. FANG, M., ZHENG, M., WANG, F., To, K.L., JAAFAR, A.B., TONG, S.L., 1999, The
solvent-extractable organic compounds in the Indonesia biomass burning aerosols –
characterisation studies. Atmospheric Environment, Vol. 33, pp. 783-795 (Pergamon).
5. FAO, 1997, State of the World's Forests 1997, Food and Agriculture Organization of the
United Nations, p.16 (Rome: FAO).
6. HP, 2000, Measurement Tutorial HP85070M. Dielectric Constant Measurement System,
Japan Hawlett Packard (Tokyo: HP)
7. KAMAL SARABANDI, 1989, Electromagnetic Scattering from Vegetation Canopies.
Radiation Laboratory, Department of Electrical Engineering and Computer Science, The
University of Michigan (Dissertation) (Michigan: University of Michigan Press).
8. LIEW, S.C., KWOH, L.K., PADMANABHAN, K., LIM, O.K., LIM, H., 1999,
Delineating land/forest fire burnt scars with ERS interferometric synthetic aperture radar,
Geophysical Research Letters, Vol.26, No. 16, pp.2409-2412 (American Geophysical
Union)
14
9. MIROSLAV R., HASNAH H., 1999, Air quality in Brunei Darussalam during the 1998
haze episode. Atmospheric Environment, Vol.33, pp. 3651-3658 (Pergamon).
10. MURALEEDHARAN, T. R., MIROSLAV R., ALLAN W. ANTHONY C., 2000a,
Chemical characterisation of the haze in Brunei Darussalam during the 1998 episode.
Atmospheric Environment, Vol. 34, No. 17, pp. 2725-2731.
11. MURALEEDHARAN, T.R., MIROSLAV R., ALLAN W., ANTHONY C., 2000b,
Emissions from the combustion of peat: an experimental study. Atmospheric Environment,
Vol. 34, pp. 3033-3035 (Pergamon).
12. NAIR, K.S.S. (Editor), 2000, Insect pests and diseases in Indonesian forests – An
assessment of the major threats, research efforts and literature, Center for International
Forestry Research (Jakarta: Grafika Desa Putera).
13. NAKAYAMA, M., MAKI, M., ELVIDGE, C.D., LIEW, S.C., 1999, Contextua l algoritm
adapted for NOAA-AVHRR fire detection in Indonesia, International Journal of Remote
Sensing, Vol. 20, No. 17, pp. 3415-3421 (London: Taylor and Francis).
14. SIEGERT, F., HOFFMANN, A.A., 2000, The 1998 forest fires in east Kalimantan
(Indonesia): A quantitative evaluation using high resolution, multitemporal ERS-2 SAR
images and NOAA-AVHRR hotspot data, Remote Sensing of Environment, Vol. 72, No.1,
pp.64-77 (New York: Elsevier).
15. TATEISHI, R. and HASTINGS, D. (Editors), 2000, Global Environmental Databases –
Present Situation; Future Directions, Chapter 6 Biodiversity Data and Information,
International Society for Photogrammetry and Remote Sensing (ISPRS), pp. 126-156, 1st
edition (Hongkong: ISPRS).
16. TUCKER, C.J. and TOWNSHEND, J.R.G., 2000, Strategy for monitoring tropical
deforestation using satellite data, International Journal of Remote Sensing, Vol. 21, No.6
& 7, pp.1461-1471.
15
17. UN, 1997, Kyoto Protocol to the United Nations Framework Convention on Climate
Change, United Nations.
18. WOLFGANG V.H.H., TORSTEN S., SIGURD S., CHAN A.K., LIM J.T., 1999,
Climate-relevant aerosol parameters of South-East-Asian forest fire haze. Atmospheric
Environment, Vol. 33, pp.3183-3190 (Pergamon).
19. WRI, 1999, World Resources 1998-1999: Chapter 11 Forests and Land Cover, World
Resources Institute (WRI) – Food and Agriculture Organisation of the United Nations
(FAO) – International Tropical Timber Organisation (ITTO) (Washington: WRI).
16
17
Chapter II
Analysis of Scattered Waves from Two Layers of Tree Trunk
2.1. Introduction
Pine (Pinus Merkusii) is the important plant in Indonesia, especially Java and Sumatra
islands as a source of turpentine or volatile oil (Coppen et al. 1993), see figure 1.2.
Turpentine is distilled from the pine resin. Traditionally, turpentine has been employed as a
solvent or cleaning agent for paints and varnishes. Most turpentine nowadays is used as a
source of chemical isolations that are then converted into a wide range of products. Many of
these, including the biggest single turpentine derivative and synthetic pine oil, are employed
for fragrance and flavour use, although there are also many important non-aromatic
applications such as polyterpene resins. Pine oil is used in disinfectants, cleaning agents and
other products having a pine odour (FAO 1995). Turpentine is obtained via tapping of living
pine trees (whether natural stands or plantations). Pine is multi purposes plant, if its tapped
then the felled trees provide income from sale of the logs for timber or pulp.
Indonesia is large archipelago that consists of more than 1,700 large islands with land
territory covers approximately 1.9 million square kilometres. It is very difficult to estimate
age or volume of pine using conventional techniques that spent much time and cost to collect
its ground data. Recently, remote sensing technology has been an efficient and helpful tool to
monitor forest and plantation in a large area. The main problem in monitoring tropical areas,
as Indonesia, is cloudy conditions. The best instrument to monitor these areas is synthetic
aperture radar (SAR), as it works effectively in spite of cloudy conditions. However, SAR
data are not easily interpreted due to the complex relations of the radar backscattering
mechanisms between microwaves and pine trunk. In this study, the author attempts to find the
18
relationships between the radar backscattering coefficients and the diameter of pine trunk that
are found in Indonesian forests and plantations.
In this study, a simple analysis of scattered wave from a pine trunk was done in order
to estimate the relationship between diameter of tree trunk and its backscattering coefficients
oσ . In section 2.2, the modelling and formulation of scattering problems on pine are
discussed. In section 2.3, the simulation of transverse electric (TE) wave propagation is done
using Finite Difference Time Domain (FDTD) method. In section 2.4, the analytical results
are verified by comparing them with the simulated results. The application of this research
will be introduced in section 2.5. Finally, conclusions are given in section 2.6.
2.2. Analysis
Figure 2.1(a) shows photograph of pine trunk that found at west Java, Indonesia.
Actually, its tree trunk is composed of two media; skin and heartwood. In this study, the
analysis of scattered wave from its tree trunk is discussed in order to investigate the
correlation of backscattering coefficient oσ and the diameter of tree trunk. This scattering
problem in its tree trunk is analysed using mode expansion method (Tetuko et al. 2001). The
two-dimensional model of tree trunk is shown in figure 2.1(b). Two layers of media compose
this model of a tree trunk with infinite length in z-axis. The radii of heartwood and skin layer
are a and b, respectively. Several trunks of pine trees that were found around the study area,
were measured and the results showed that approximately each component had relations
a=0.5b. The properties of skin are determined by complex dielectric constant rε and complex
permeability rµ . The dielectric constant rε of several skin medium of pine samples were
measured experimentally by the author using dielectric probe kit HP85070B and the result
was 3.1-j0.4 (see figure 1.8 (a)). The water content of heartwood is high, consequently,
heartwood was assumed to be an infinite length of perfect conductor or electromagnetic fields
19
Figure 2.1. Photograph and geometry of the analysis.
heartwood
skin
(a) Photograph of a cross section of pine (Pinus merkusii) trunk.
heartwood
y
x z
skin
a
b
r
(region I) (region II)
scattered waves
incident waves
transmitted waves
observed point (P)
SEφ
mEφ
IEφφ
(b) Geometry of scattered waves from a pine trunk.
20
in heartwood is zero. Here, incident wave is assumed as a plane wave that has transverse
electric (TE) mode and incident angle φ with respect to direction of observed point P from
origin of coordinate. This wave propagation is –x direction. Based on this figure and
Appendix E, the z component of the magnetic fields in free space and skin are determined as
Incident wave ∑∞
=
=0
cos)(m
momm
Io
Iz mjrkJUHH φ ( br > ) (2.1)
Scattered wave ∑∞
=
=0
)2( cos)(m
ommIo
Sz mrkHbHH φ ( br > ) (2.2)
Transmitted wave φmkrNakrJaHHm
mmmmIo
mz cos)()(
0∑
∞
=
′+= ( bra ≤≤ ) (2.3)
Where the wave number of skin is rrokk εµ= and ok is wave number in free space. ma to
mb are amplitude coefficients. IoH is initial amplitude of incident magnetic field. mJ , mN ,
and )2(mH are m-th order of Bessel function, Neumann function, and 2nd kind of Hankel
function. Where
==
=),3,2,1(2
)0(1
Lm
mUm (2.4)
By substituting (2.1) to (2.3) into Maxwell’s equations below
t∂∂=×∇ E
H ε (2.5)
the electric field of each medium was derived as
∑∞
=
′−=0
cos)(m
momm
o
IooI mjrkJU
j
HkE φ
ωεφ ( br > ) (2.6)
∑∞
=
′−=0
)2( cos)(m
ommo
IooS mrkHb
j
HkE φ
ωεφ ( br > ) (2.7)
∑∞
=
′′+′−=0
cos)()(m
mmmm
Iom mkrNakrJa
jkH
E φωεφ ( bra ≤≤ ) (2.8)
21
Further, by substituting (2.1) to (2.3) and (2.6) to (2.8) into the boundary condition of each
interface between media given below:
r=a 0=mEφ (2.9)
r=b SIm EEE φφφ += (2.10)
Sz
Iz
mz HHH += (2.11)
the amplitude coefficient mb of scattered wave from tree trunk SEφ was obtained as;
( )
′−
′−−=
)()(
)()(
)2()2( bkHbkHZ
bkJbkJZjUb
ommomrm
ommomrmm
mm
βα
βα (2.12)
where
rrrZ εµ=
)()()()( kbJkaNkaJkbN mmmmm ′′−′′=α
)()()()( kbJkaNkaJkbN mmmmm ′−′=β
Finally, by substituting the amplitude coefficient mb of (2.12) into (2.7), the scattered electric
field is obtained.
In the same manner, fields of transverse magnetic (TM) mode that scattered from two
layers of tree trunk could be derived (see Appendix A) and are obtained as below. Where the
electric fields are
Incident wave ( )∑∞
=
=0
cosm
momm
Io
Iz mjrkJUEE φ ( br > ) (2.13)
Scattered wave ( ) ( )∑∞
=
=0
2 cosm
ommIo
Sz mrkHbEE φ ( br > ) (2.14)
Transmitted wave ( ) ( ) ∑∞
=
′+=0
cosm
mmmmIo
mz mkrNakrJaEE φ ( bra ≤≤ ) (2.15)
and the magnetic fields are
22
( )∑∞
=
′=0
cosm
momm
o
IooI mjrkJU
j
EkH φ
ωµφ ( br > ) (2.16)
( )∑∞
=
′=0
)2( cosm
ommo
IooS mrkHb
j
EkH φ
ωµφ ( br > ) (2.17)
( ) ( ) ∑∞
=
′′+′=0
cosm
mmmm
Iom mkrNakrJa
jkE
H φωµφ ( bra ≤≤ ) (2.18)
Finally, amplitude coefficient mb for transverse magnetic mode is derived as
( ) ( )
( ) ( )bkHbkHZ
bkJbkJZjUb
ommomrm
ommomrmm
mm
)2()2( βα
βα
−′
−′−= (2.19)
where
( ) ( ) ( ) ( )kaNkbJkbNkaJ mmmmm −=α
( ) ( ) ( ) ( )kaNkbJkbNkaJ mmmmm ′−′=β
To confirm the analysis result, the simulation of scattered wave from a pine tree trunk
using Finite Difference Time Domain (FDTD) method is discussed in the next section.
2.3. Simulation
The JERS-1 SAR operated in horizontal (H) polarization on both transmits and
receives. Hence horizontal polarization or transverse electric (TE) mode is considered in this
study. Hence, electromagnetic field components are considered as )0,,( yx EE and ),0,0( zH .
Figure C.1 (in Appendix C) shows the position of field components in the finite-difference
grid (unit of sample spacing). By referring this figure, the detail scattered electromagnetic
fields SE and SH that are derived from Maxwell’s equations are discussed in Appendix C,
where the electromagnetic fields are shown in (2.20) to (2.22). In the derivation of these
equation, t∆ is time step, and x∆ , y∆ are sample spacing in x and y-axis, respectively. Here,
the notation of Yee (Yee 1966) is used to replace tnH ∆+ )( 21
as )( 21+nH . Similarly, scattered
23
fields ),( yxS,tE and ),(, yxtSH are expressed as ),( jiS,nE and ),(, jinSH , where tnt ∆= ,
xix ∆= and yjy ∆= . Sampling lead to the characteristic staggered finite-difference grid. In
this grid, the electromagnetic field components are offset by 2t∆ in time and 2x∆ and
2y∆ in space. In the former, the field components are updated sequentially in time.
( ) ( ) ( )( ) ( ) ( )
( )( ) ( )
( ) ( )y
jiHjiHjitji
jit
jiEjitjijitji
jiE
nSz
nSz
nSx
nSx
∆−+−++
+∆+++∆
+
++∆+++∆+−
=+
−−
−
21
21,
21
21,
21
21
21
211,
21
21
21
21
21,
,,,2,1
,
,,2,1,2,1
,
21
21
εσε
εσεσ
( ) ( ) ( )( ) ( )( ) ( ) ( )
( ) ( ) ( )( ) ( )( ) ( ) ( )jiE
jitjijijijitji
jiEjitji
jijijitji
nIx
o
nIx
o
,,2,1
,,,2,
,,2,1
,,,2,
21,
21
21
21
21
21
21
211,
21
21
21
21
21
21
++∆++
+−+++∆+−
++∆++
+−+−+∆+− −
εσεεεεσ
εσεεεεσ
(2.20)
( ) ( ) ( )( ) ( ) ( )
( )( ) ( )
( ) ( )
( ) ( ) ( )( ) ( )( ) ( ) ( )2
11,
21
21
21
21
21
21
21
21,
21
21,
21
21
21
211,
21
21
21
21
21,
,,2,1
,,,2,
,,
,2,1
,
,,2,1
,2,1,
21
21
++∆++
+−+−+∆+−
∆+−−++
+∆+++∆
−
++∆+++∆+−
=+
−
−−
−
jiEjitji
jijijitji
x
jiHjiH
jitji
jit
jiEjitji
jitjijiE
nIy
o
nSz
nSz
nSy
nSy
εσεεεεσ
εσε
εσεσ
( ) ( ) ( )( ) ( )( ) ( ) ( )2
1,
21
21
21
21
21
21
,,2,1
,,,2,+
+∆+++−+++∆+
− jiEjitji
jijijitji nIy
o
εσεεεεσ
(2.21)
( )
( ) ( ) ( ) ( )( )
( ) ( ) ( )( )( )
( ) ( ) ( )( )21
21,
21
21,
21
21
21
21
21,
21,
21
21
21,
21,
21
212
121,
21
21,
,,,
,
,1,,
,,1,
,
,
21
21
21
21
++−++++
−++−
+−++∆++
∆+
+−++∆++
∆−++
=++
−+
−
+
jiHjiHji
ji
jiEjiEyji
t
jiEjiExji
tjiH
jiH
nI
z
nI
zo
nSx
nSx
nSy
nSy
nSz
nS
z
µµµ
µ
µ
(2.22)
24
where ),( jiµ , ),( jiε , ),( jiσ are the characteristics of wave propagation media in
simulation space. SE and SH are scattered electromagnetic fields that derived from
Maxwell’s equations. oε and oµ are dielectric constant and permeability of free space,
respectively. IE and IH are space functions of incident electromagnetic fields. In this
research, horizontal polarization or transverse electric wave was considered. Consequently, in
this simulation, electromagnetic field components were considered as )0,,( yx EE and
),0,0( zH .
When implementing the finite-difference scheme, boundary conditions must be treated
in a special manner. Two different kinds of boundaries: the internal boundaries (i.e.,
boundaries within the medium marked by a change in material properties) and the external
boundaries (i.e., the grid edges). The conditions at internal boundaries (i.e., at the interfaces
between different media) are usually satisfied implicitly. However, to ensure numerical
stability, the material properties must be averaged for components on boundary. For
transitions between similar materials, the averaging may be omitted. However, it is necessary
at an interface between media with greatly different material properties (for example, at an air
– skin interface) in order to maintain the stability. The finite-difference model is implemented
in two-dimensions (2-D) as shown in figure 2.2. In this figure, simulation space is sampled
into INX x INY grids. FDTD method can only simulate a finite space, but real scattering
problems are often in the infinite formations. In this case, artificial external boundaries must
be applied in the FDTD method. To prevent these artificial boundaries from reflecting
electromagnetic waves, absorbing boundary conditions are used. In 1981, Gerrit Mur
introduced simple absorbing boundary conditions to truncate FDTD meshes (Gerrit 1981).
The second kind of Mur method was applied in this analysis, because it involves small
calculation-memory size and its accuracy is assured (Uno 1998). In this study, for example,
electric field in i=1 is determined as;
25
Figure 2.2. Simulation model.
tree trunk
B
P
1 2 3 4 … … … INX-1 INX
1
2
3
4
:::
::
INY-1
INY
abso
rbin
g bo
unda
ry c
ondi
tion
simulation space
A
incident wave
26
( )
( )
)1,2(1,
),2(1,
2)1,2(1,
)1,1(1,
),1(1,
2
)1,1(1,
22
2),2(
1,),1(
1,2
),1(2,
),2(,
),2(2,
),1(,
−−
+−
−+−
+−−
+−
−
+−
∆+∆∆
∆∆+
−+
−
∆+∆
∆+
−+
∆+∆
∆−∆+
−−=
jnS
EjnS
EjnS
EjnS
EjnS
E
jnS
Extvy
tvxj
nSEj
nSE
xtv
x
jnS
EjnS
Extv
xtvj
nSEj
nSE
yyyyy
yyy
yyyy
(2.23)
In the same manner, the other components of electric field in i=INX, j=1, and j=INY can be
derived, where ν is wave speed.
Figure 2.2 depicts the simulation model where A and B are heartwood and skin layer,
respectively. This model was done in two-dimension, where each medium has infinite length
in z-axis. This simulation space is divided into INX x INY grids (unit of cell size). Incident
wave is a plane wave of intensity as that shown by Gaussian pulse, where the power spectrum
of it is smooth and it is easy to sample. This pulse propagates from left to right of the
simulation space in light speed. The Gaussian pulse is defined by function
≤≤=
−−
otherwise
tetpt
,0
)20(,)( 0)( 2
0 ττα
(2.24)
Where t is running time, 0τ is pulse width, and ( )204 τα = . Figure 2.3 (a) and (b) show
Gaussian pulse with 90 109 −×=τ s and its spectrum, respectively.
The geometry of incident wave is shown in figure 2.4, where plane wave propagates at
angle φ with respect to x-axis. The propagation direction of the incident wave is determined
as
φφ sinˆcosˆ0 yxr += (2.25)
where x and y are unit vectors. Hence the incident electromagnetic fields are derived as
( )000
0 ˆ),( tcrtpZE
tH Iz +⋅+= rr (2.26)
)ˆ()ˆcosˆsin(),( 000 trtpyxEtI +⋅++−= crrE φφ (2.27)
27
Figure 2.3. Pulse of incident wave.
0.0
0.2
0.4
0.6
0.8
1.0
0.0 0.5 1.0 1.5 2.0
Times
Gau
ssia
n pu
lse
0τt
(a) Gaussian pulse.
-180
-160
-140
-120
-100
-80
-60
-40
-20
0
0 2 4 6 8 10
FFT
inte
nsiti
es
[dB
]
0τf(b) Fast fourier transformed Gaussian pulse.
28
Figure 2.4. Geometry of incident wave
where c is speed of light. 0E and 0Z are initial intensity of incident electric field and wave
impedance in free space, respectively. )(tp is pulse function excited by Gaussian pulse
(2.24). By substituting the components of electromagnetic fields in (2.26) and (2.27), and
considering the Yee’s notation, each component of electromagnetic fields of incident wave is
acquired as follows;
( ) ( ) ( ) ( )cdcyjxitnpZE
jiH nIz −∆++∆++∆+⋅=+++ φφ sincos),( 2
121
21
0
021
21, 2
1
(2.28)
( ) ( ) ( )cdcyjxitnpEjiE nIx −∆++∆++∆⋅−⋅=++ φφφ sin1cos)sin()1,( 2
102
1,
(2.29)
( ) ( ) ( )cdcyjxitnpEjiE nIy −∆++∆++∆⋅⋅=++ φφφ sincos1cos),1( 2
102
1, (2.30)
where cdt −=0 means that the pulse head is at a distance d from the origin of coordinate at
initial time ( 0=t s). For each time step, the value of excitation is added to the value
calculated from the finite-difference scheme.
To arrange the time step and cell size, first, frequency maxf must be decided from
figure 2.3(b). Here maxf is frequency in 120dB from maximum intensity of frequencies
spectrum. It means that the accuracy of calculation is assured to six digits. By referring to this
figure, 6.2max =f GHz is obtained. By using empirical equation max10 fvyx =∆=∆ (Uno
1998), 21025.1 −×=∆=∆ yx m is obtained, where v is wave speed, in this study it is assumed
IEφ
IzH
y
z x
0r0φ
d
29
to be the same with wave speed in free space ( smc 8103×= ). Finally, according to Courant
condition (Uno 1998)
( ) ( )22 111 yxtc ∆+∆≤∆ (2.31)
the time step t∆ is obtained as st 11105.2 −×=∆ . Finally, the two dimensional backscattering
coefficient oσ is defined as
GHzf
Iy
Sy
E
E
lR
275.1
2
2
0 2
=
= πσ (2.32)
where IyE is observed or incident electric field intensity in frequency 275.1=f GHz on the
trunk surface. SyE is the intensity of electric field in observed point. l is effective surface of
tree trunk for scattered wave, where it is assumed as bπ . To simplify the analysis, the trunk
number in targeted area is not considered and the obtained backscattering coefficient is
considered as the average backscattering coefficient of total tree trunk in a unit area.
2.4. Results and discussion
In simulation space, see figure 2.2, infinite length of tree trunk is considered. This tree
trunk is composed of two media; skin and heartwood. The radius of tree trunk b varies from 0
to 40 grids (or 0 to 0.5 m). The simulation space edges are surrounded by artificial absorbing
boundary condition (Mur method). Incident wave is a plane wave of intensity as that shown
by Gaussian pulse, which propagates from left to right of the simulation space in speed of
light. Parameters of simulation are the simulation space grids 300== INYINX , space-
increments 21025.1 −×=∆=∆ yx m, time- increment 11105.2 −×=∆t s, maximum intensity of
initial electric field 100 V/m, and running time tt ∆= 600 s. Figure 2.5 shows scattered wave
from pine tree trunk with tt ∆= 50 to t∆300 s. Additionally, figure 2.5 at tt ∆= 300 s shows
30
Figure 2.5. Scattered waves in simulation space from tt ∆= 50 s to t∆300 s. A and B are
scattered waves from skin and heartwood, C and D are scattered wave from
trapped waves in skin layer, E is forwarded wave that occurred by clipping pulse
that flows on the trunk surface and scattered to back of trunk, F and G are
heartwood and skin, respectively. P is the observed point.
A B
C D
E
F G P
31
Figure 2.6. Scattered waves at observed point P: A and B, C and D are scattered pulse from
skin and heartwood, and trapped wave in the skin layer, respectively.
details of scattered wave, where A and B are scattered waves from skin and heartwood,
respectively. C and D are scattered wave from trapped waves in skin layer. E is forwarded
wave. Then F and G are heartwood and skin, respectively. The observed point P is at 1.5 m
from centre of tree trunk. This point is used to observe the intensities of scattered
electromagnetic fields. In this study, the author observed only the horizontal polarisation
component (transverse electric wave) or electric field SyE in backscattering direction
( oo 0=φ ). The backscattered electric field is computed and is shown in figure 2.6. In this
figure, A and B, C and D show scattered waves from skin and heartwood, and trapped wave
in the skin layer, respectively. Further, fast Fourier transform is employed to obtain the
electric field intensity of preferred frequency, in this case, frequency of Japanese Earth
Resources Satellite (JERS-1) SAR, 275.1=f GHz, is used. Finally, the backscattering
coefficient is calculated using (2.32) and the results are shown in figure 2.7 (£ - simulation).
-20
-10
0
10
20
0.0 2.5 5.0 7.5 10.0 12.5
Times (ns)
Inte
nsity
(V/m
) A B
C D
32
Figure 2.7. Relationship between tree trunk diameter and backscattering coefficient
-16
-14
-12
-10
-8
-6
-4
-2
0
2
0.254 0.262 0.270 0.278 0.286 0.294
Diameter of tree trunk [m]
Bac
ksca
tterin
g co
effic
ient
[dB
]
analysis
simulation
a
b
pine, a=0.5b
33
In the analysis, direction of scattered wave is the same as observed point P or oo 0=φ .
Consequently, SEφ in analysis is equivalent to SyE that calculated in the simulation. By
substituting the parameters of JERS-1 SAR in the equations of analysis, the analysis results
are obtained and are as shown in figure 2.7 (¢ - analysis). The results compare well with
simulation ones for a tree trunk (£ - simulation). However, a small error was found. It is
considered that the error is generated by the Finite Difference Time Domain (FDTD)
calculation error caused by calculation using the sampled space. In this analysis, the soil is
assumed as perfectly conductor, hence figure 2.7 shows only pure backscattering coefficient
of a tree trunk. In the next section, this result will be applied to estimate pine trunk diameter
in the study area using JERS-1 SAR data.
2.5. Application
2.5.1. Study area
The study area is pine forest around Saguling lake, west Java, Indonesia (figure 2.8).
The region has altitude ranging from 7m to 127m. Biomes of this area are pine forest, mixed
vegetation area, settlement and paddy fields called ladang (dry paddy fields). The soil
condition around the study area is wet. The ground data of the study area were collected in
1999 (Ketut 2001). The annual average rainfall of this area was 233 days, while the annual
average temperature is 23oC.
2.5.2. Data processing
The JERS-1 SAR data was examined in order to estimate the diameters of pine tree
trunk in the study area. The data (path 106, row 312) was acquired on 13 May 1997 during the
34
Figure 2.8. Map of the study area
Indian Ocean
Jakarta
Bogor
I N D O N E S I A
Bandung
Bandung
Saguling Lake
highway
N
0 5 km
107o 48` E 107o 18` E
6o 42` S
7o 00` S
to Jakarta
study area (pine forest)
Sumatera island
Java island
35
Figure 2.9. Photograph of pine forest in the study area and the supervised classification results
of JERS-1 SAR data (path 106, row 312, 13 May 1997).
Legend
kilometres
1. river/paddy field
2. mixed vegetation area
3. forest 1
4. forest 2
5. forest 3
6. dry land (ladang) 7. settlement
6o55`S
6o55`S
107o27`E 107o28`E 107o29`E
10
10
10
10
10
10
15
15
15 20
15
20
15 15
20
20 20
20 25
25
25
30
30
30
35 35
35
35
40
40
40
45
45
0 1 2 3
Altitude units in metres
Pine forest in the study area
36
Table 2.1. Classification and estimation results
Class names Backscattering coefficient oσ (dB)
Diameter of tree trunk (m)
Standard deviation (m)
forest 1 -10 0.260 0.0025
forest 2 -8 0.265 0.0025
forest 3 -5 0.270 0.0025
dry season in the study area. This image was processed at level 2.1 or standard geocoded data
and was resampled to the Universal Transverse Mercator (UTM) projection by the Earth
Observation Research Centre (EORC) of the National Space Development Agency (NASDA)
of Japan. Firstly, a 3x3 median filter was employed and second ly, process used a 5x5 average
filter to reduce inherent speckle noise (Sunar et al. 1998). At the same time, the data was also
referenced to the UTM co-ordinate system, through a polynomial rectification using 30
ground control points collected from topographic maps scale of 1:25.000
(BAKOSURTANAL 1990). This procedure yielded a geometric accuracy of 0.1 pixels. Then
the spatial resolution of SAR data was resampled to 12.5m.
A supervised classification was performed to classify the satellite data into seven
classes. The topographic maps and ground data were used to select training sites (Ketut 2001),
i.e. river and paddy field, bush, forest 1, forest 2, forest 3, mixed forest, and settlement (figure
2.9). By supervised classification, the average pixel intensity I of each class was obtained.
These values were substituted in the equation 2.68log20 −= Ioσ dB (Shimada 1998) to
obtain backscattering coefficients. By plotting the results on figure 2.7, the average diameter
of each class was obtained (see table 2.1). The estimation result shows the diameter of pine in
37
the study area was between 0.26m and 0.27m and has standard deviation 0.0025m. A sampling of
10 locations yielded similar results to the ground data that collected in 1999 (Ketut 2001).
2.6. Conclusions
A simple numerical analysis was conducted to analyse the rela tionship between the
backscattering coefficients oσ and diameter of pine tree trunk. The analysis results were
confirmed by simulation using Finite Difference Time Domain (FDTD) method. These results
are in good agreement. A variation of this analysis, it could be applied to estimate diameter of
tropical tree trunk from Synthetic Aperture Radar (SAR) data, which this information is very
important to estimate the forest volumes or biomass effectively and accurately. These results
succeeded in estimating tree trunk diameter of pine (Pinus merkusii) that is widely distributed
in the west Java forest, Indonesia from JERS-1 SAR data.
While this study focused on single site in Indonesia, it is reasonable to expect that this
method or variations should be successful in estimating tree trunk diameters in similar forest
regions of the world using SAR data.
38
References
1. BAKOSURTANAL, 1990, Topographic maps; 1209-223, 1209-241, 1209-242.
Indonesian National Coordination Agency for Surveys and Mapping, 1st edition
(Cibinong: Bakosurtanal).
2. COPPEN, J.J.W., GAY, C., JAMES, D.J., ROBINSON, J.M. and SUPRIANA, N., 1993,
Variability in xylem resin composition amongst natural populations of Indonesian Pinus
merkusii. Phytochemistry, 33, 129-136.
3. DAVID, P., STELLA, E.B., STHEPHEN, J.M., 1997, Terrain influences on SAR
backscatter around Mt. Taranaki, New Zealand. IEEE Transactions on Geoscience and
Remote Sensing, 35, 924-932.
4. FAO, 1995, Flavours and fragrances of plant origin. Food and Agriculture Organisation
(FAO), United Nations, Chapter 8, 1st edition (Rome: FAO-UN).
5. GERRIT MUR, 1981, Absorbing boundary conditions for the finite-difference
approximation of the time-domain electromagnetic-field equation. IEEE Transactions on
Electromagnetic Compatibility, 23, 377-382.
6. KETUT WIKANTIKA, 2001, Spectral and textural aspects of multisensor and
multitemporal satellite data for land use / land cover mapping in a tropical area. Ph.D
Dissertation, Chiba University, January 2001 (Chiba: Chiba University).
7. SHIMADA, M., 1998, User’s guide to NASDA’s SAR products. Earth Observation
Research Centre, National Space Development Agency (NASDA), 2nd edition (Tokyo:
NASDA).
8. SUNAR, F., TABERNER, M., MAKTAV, D., KAYA, S., MUSAOGLU, M., and YAGIZ,
E., 1998, The use of multi temporal radar data in agriculture monitoring: a case study in
Kyocegiz-Dalaman ecosystem, Turkey. International Archives of Photogrammetry and
Remote Sensing, 22, 559-565.
39
9. TETUKO S.S., J., R. TATEISHI, K. WIKANTIKA, 2001, A method to estimate tree
trunk diameter and its application to discriminate Java-Indonesia tropical forests.
International Journal of Remote Sensing, 22, 177-183.
10. UNO TORU, 1998, Finite difference time domain method for electromagnetic field and
antenna analyses. 1st edition (Tokyo: Corona).
11. YEE, K. S., 1966, Numerical solution of initial boundary value problems involving
Maxwell’s equations in isotropic media. IEEE Transactions on Antennas Propagation, 14,
302-307.
40
41
Chapter III
Analysis of Scattered Waves from Three Layers of Tree Trunk
3.1 Introduction
In Chapter II, the author has discussed about the analysis of the scattered wave from
two layers of a tree trunk. The analysis was only focusing in two layers, but actually the tree
trunk has three main layers (Kamal 1989). Hence in this study, the analysis of scattered wave
from three layers of a tree trunk is discussed to explore the complex relationships of the radar
backscattering mechanisms between microwaves and tropical vegetation types. Many
researchers have developed electromagnetic modelling of vegetations (Kamal 1989, Li et al.
1999), but here the author attempts to develop a method to find the relationships between the
radar backscattering coefficients and the characteristics of tropical forest vegetation,
particularly species that are found in Indonesian tropical forests.
In this study, analysis of scattered wave from a tree trunk has been done in order to
estimate the relationship between the diameter of a tree trunk and its backscattering
coefficients oσ . In section 3.2, the modelling and formulation of scattering problems in three
layers of a tropical tree trunk are discussed. Then the simulation of transverse electric (TE)
wave propagation in a tropical tree trunk is done using the Finite Difference Time Domain
(FDTD) method (Uno 1998, Yee 1966), where the Mur method (Mur 1981) is applied as the
42
Figure 3.1. Tree trunk media
absorbing boundary condition to absorb the outgoing electromagnetic waves in a simulation
space edges are discussed in section 3.3. In section 3.4, the analytical results are verified by
comparing them with the simulated results. The application of the proposed method is
discussed in section 3.5. Finally, conclusions are given in section 3.6.
3.2 Analysis
Actually, a tree trunk is composed of three media; skin, xylem, and heartwood,
referring to figure 3.1 (Kamal 1989). Additionally, a skin medium is structured by cork,
phleom and cambium layers. In this study, the scattering problems in a tropical tree trunk is
discussed in order to investigate the correlation of backscattering coefficient oσ and the
diameter of a tree trunk. This scattering problem in a tropical tree trunk is analysed by
cork
phleom
cambium
xylem
heartwood
skin
43
Figure 3.2. Geometry of analysis
deriving the scattered fields at each medium. The two-dimensional model of a tree trunk is
shown in figure 3.2. Three layers of media compose this model of a tree trunk with infinite
length in z-axis. The radii of heartwood, xylem and skin layer are a, b, and c, respectively.
Where these radii have relations as a=0.5c and b=0.9c.
In this study, the scattered wave from a tropical tree trunk is analysed. For this
purpose, several trunks of tropical trees were collected and measured. The properties of xylem
and skin are determined by 1rε , 1rµ and 2rε , 2rµ respectively. Where, riε and riµ (i=1,
2) are complex dielectric constant and complex permeability, respectively. The water content
of heartwood is high, consequently, heartwood may be assumed to be an infinite length of
b
c
a
x
y
P
φ
r
I (xylem)
II
IEφ
IzH
III (free surface, air) (skin)
(heartwood)
z
SEφ
1mEφ
2mEφ
Incident wave
Scattered wave
Transmitted
wave at I
Transmitted
wave at II
Observed point
44
perfect conductor or electromagnetic fields in heartwood is zero. Here, incident wave is
assumed as a plane wave that has transverse electric (TE) mode and incident angle φ with
respect to direction of observed point P from the origin of coordinate (0, 0). This wave
propagates in –x direction. Based on this figure, the φ component of the electromagnetic
fields in a free space, xylem and skin are determined as
xjkIo
Iz
oeHH = )( cr > (3.1)
∑∞
=
=0
)2( cos)(m
ommIo
Sz mrkHbHH φ )( cr > (3.2)
∑∞
=
′+=0
11111 cos)()(
mmmmm
Io
mz mrkNarkJaHH φ )( bra ≤< (3.3)
∑∞
=
′+=0
22222 cos)()(
mmmmm
Io
mz mrkNarkJaHH φ )( crb ≤< (3.4)
where the wave numbers of each medium are 111 rrokk εµ= and 222 rrokk εµ= , and ok
is the wave number in free space. ma1 to mb are amplitude coefficients. IoH is initial
amplitude of incident magnetic field. mJ , mN , and )2(mH are m-th Bessel function,
Neumann function, and 2nd kind of Hankel function. By referring to Appendix E, (3.1) is
transformed to be
∑∞
=
=0
cos)(m
momm
Io
Iz mjrkJUHH φ )( cr > (3.5)
where
==
=L,2,1,2
0,1
m
mUm (3.6)
By substituting (3.1) to (3.5) into source free Maxwell’s equations below
t∂∂=×∇ E
H ε (3.7)
45
the magnetic field of each medium is derived as
∑∞
=
′−=0
cos)(m
momm
o
IooI mjrkJU
j
HkE φ
ωεφ )( cr > (3.8)
∑∞
=
′−=0
)2( cos)(m
ommo
IooS mrkHb
j
HkE φ
ωεφ )( cr > (3.9)
∑∞
=
′′+′−=0
11111
11 cos)()(m
mmmm
Iom mrkNarkJa
j
HkE φ
ωεφ )( bra ≤< (3.10)
∑∞
=
′′+′−=0
22222
22 cos)()(m
mmmm
Iom mrkNarkJa
jHk
E φωεφ )( crb ≤< (3.11)
Further, by substituting (3.1) to (3.5) and (3.8) to (3.11) into the boundary condition of each
interface between media given below:
ar = 01 =mEφ (3.12)
br = 21 mm EE φφ = and 21 mz
mz HH = (3.13)
cr = ISm EEE φφφ +=2 and Iz
Sz
mz HHH +=2 (3.14)
the amplitude coefficient mb of scattered wave from tree trunk SEφ is obtained as;
[ ])()(
)()(
)2(7
)2(6
76
ckHckH
ckJckJjUb
ommomm
ommommm
mm ′
−
′−−=
αα
αα (3.15)
where
)()(
1
11 akJ
akN
m
mm ′
′=α (3.16)
( ) ( )bkJbkN mmmm 1112 ′−′= αα (3.17)
)()( 1113 bkJbkN mmmm αα −= (3.18)
( ) ( )bkJk
bkJk
mmmmm 232
222
1
14 ′−= α
εα
εα (3.19)
( ) ( )bkNk
bkNk
mmmmm 232
222
1
15 ′−= α
εα
εα (3.20)
46
( ) ( )
′−′= ckJckNkk
mm
mm
o
om 2
4
52
2
26 α
αεε
α (3.21)
)()( 24
527 ckJckN m
m
mmm α
αα −= (3.22)
Finally, by substituting the amplitude coefficient mb of (3.15) into (3.9), the scattered
electric field is obtained.
In the same manner, transverse magnetic (TM) mode of scattered waves from three
layers of tree trunk was derived as Appendix B. The electric fields in each medium will be
Incident wave ( )∑∞
=
=0
cosm
momm
Io
Iz mjrkJUEE φ ( cr > ) (3.23)
Scattered wave ( )∑∞
=
=0
)2( cosm
ommIo
Sz mrkHbEE φ ( cr > ) (3.24)
Medium I ( ) ( ) ∑∞
=
′+=0
11111 cos
mmmmm
Io
mz mrkNarkJaEE φ ( bra ≤< ) (3.25)
Medium II ( ) ( ) ∑∞
=
′+=0
22222 cos
mmmmm
Io
mz mrkNarkJaEE φ ( crb ≤< ) (3.26)
and the magnetic fields will be
Incident wave ( )∑∞
=
′=0
cosm
momm
o
IooI mjrkJU
j
EkH φ
ωµφ ( cr > ) (3.27)
Scattered wave ( )∑∞
=
′=0
)2( cosm
ommo
IooS mrkHb
j
EkH φ
ωµφ ( cr > ) (3.28)
Medium I ( ) ( ) φωµφ mrkNarkJaj
EkH
mmmmm
Iom cos
01111
1
11 ∑∞
=
′′+′= ( bra ≤< ) (3.29)
Medium II ( ) ( ) φωµφ mrkNarkJaj
EkH
mmmmm
Iom cos
02222
2
22 ∑∞
=
′′+′= ( crb ≤< ) (3.30)
By substituting these fields into the boundary conditions, the amplitude coefficient mb is
obtained as
( ) ( )
( ) ( )ckHckH
ckJckJjUb
ommomm
ommommm
mm ′−
−′=
)2(6
)2(7
76
αα
αα (3.31)
47
Table 3.1. Dielectric constants of Indonesian tropical forest trees at the frequency of
JERS-1 SAR ( 275.1=f GHz)
skin ( rrr jεεε ′′−′=2 ) xylem ( rrr jεεε ′′−′=1 ) species names
rε′ rε ′′ rε′ rε ′′
teak 3.1 0.4 11.5 2.6
mahogany 2.7 0.3 10.2 2.1
pine 3.4 0.4 13.6 3.0
rasamala 2.5 0.3 9.4 2.1
where the constants of (3.31) can be referred in Appendix B.
The dielectric constant rε of several sample of tree trunks were measured
experimentally using dielectric probe kit HP85070B (figure 1.6.(a)), and the results are shown
in figure 1.8 and table 3.1 which shows dielectric constants in frequency 275.1=f GHz
(JERS-1 SAR).
3.3 Simulation
In this study, the author considers the scattered waves from tropical tree trunk to
explore the relationship between backscattering coefficient oσ and the diameter of three
layers of tree trunk. The finite-difference model (Yee 1966) is implemented in two dimensions
(2-D) as shown in figure 3.3. In this figure, simulation space is sampled into INX x INY grids.
48
Figure 3.3. Geometry of simulation space. Remarks: P is observed point. A, B and C are
heartwood, xylem, and skin, respectively. Simulation space is divided into INX x
INY grids of meshes.
Trunk
C P
1 2 3 4 … … … … INX-1 INX
1 2 3 4
:::::
INY-1 INY
Incident wave
Abs
orbi
ng b
ound
ary
cond
ition
Simulation space
B A
49
Referring to the simulation in Chapter II with the same parameters, i.e. incident pulse using
Gaussian pulse, boundary condition using Mur method (Mur 1981). Finally, the two
dimensional backscattering coefficient oσ is defined by (3.32), where IyE is observed
electric field intensity on the trunk surface in frequency 275.1=f GHz.
GHzf
Iy
Sy
E
E
lR
275.1
2
2
0 2
=
= πσ (3.32)
Where R is distance from centre of tree trunk to the observation point. l is effective scattered
surface of tree trunk, in this case, is assumed as bπ . In the calculation of its backscattering
coefficient in the analysis, SyE and I
yE are equal to SEφ and IEφ respectively at incident
angle 0o.
3.4. Results
In simulation space, see figure 3.3, scattered wave from an infinite length of tree trunk
is considered. This tree trunk is composed of three media; skin, xylem, and heartwood. The
radius of tree trunk varies from 0 to 40 grids (or 0 to 0.5m). The simulation space edges
(external boundaries) are surrounded by artificial absorbing boundary condition (Mur
method), refer the simulation in Chapter II. Incident wave is a plane wave with intensity as
shown by Gaussian pulse, which propagates from left to right of the simulation space in speed
of light. Parameter of simulation is simulation space grids 300== INYINX ,
50
space-increments 21025.1 −×=∆=∆ yx m, time- increment =∆t 11105.2 −× s, maximum
intensity of initial electric field 100 V/m, and running time tt ∆= 600 s. Figure 3.4 shows
scattered wave from tree trunk with tt ∆= 50 to t∆300 s. Additionally, figure 3.5 shows
details of scattered wave when tt ∆= 300 s, where A, B, C, and D are scattered waves from
skin, xylem, heartwood, and forwarded wave that is occurred by clipping wave on the tree
trunk surface, respectively. Then E, F, and G are skin, xylem, and heartwood, respectively.
The observed point P is at 1.5 m from centre of tree trunk. This point is used to observe
intensities of scattered electromagnetic fields.
In this study, the author observed only the horizontal component (TE wave) or
electric field SyE in backscattering direction ( o0=φ ). The backscattered electric field is
computed and is shown in figure 3.6. In this figure, A, B, and C show scattered waves from
skin, xylem, and heartwood, respectively. Further, fast Fourier Transform is employed to
obtain power spectrum of it and electric field intensity of preferred frequency, in this case,
frequency of Japanese Earth Resources Satellite (JERS-1) SAR, 275.1=f GHz was used.
Finally, the backscattering coefficient is calculated using (3.32) and the result is shown in
figure 3.7 (r - simulation).
In the analysis, direction of incident wave is the same as of observed point P or
o0=φ . Consequently, SEφ is equivalent to SzE that obtained in simulation where the
distance between the centre of tree trunk to the observed point is 1.5 m. By considering the
51
Figure 3.4. Distribution of scattered electric field intensity SyE with tt ∆= 50 to t∆300 s.
52
Figure 3.5. Distribution of scattered electric field SyE in tt ∆= 300 s, where A, B and C are
scattered wave from skin, xylem and heartwood, respectively. P is observed point.
E, F and G are skin, xylem and heartwood, respectively. D is forwarded wave that
is occurred by clipping wave that flows on the trunk surface and scattered to the
backward of tree trunk.
X [ m ]
Y [
m]
0 0 . 5 1 1 . 5 2 2 . 5 3 3 . 5
0
0 . 5
1
1 . 5
2
2 . 5
3
3 . 5
-100
-90
-80
-70
-60
-50
-40
-30
-20
-10
0n=300 Rasamala (Altingia exelsa) Intensities S
yE
(dB)
P
A B
C D
E
F G
53
Figure 3.6. Scattered electric field intensities at observed point P. A, B and C are scattered
pulse from skin, xylem and heartwood, respectively.
parameter of JERS-1 SAR, the result of analysis is obtained and is shown in figure 3.7 (p -
rasamala). This result is obtained from the analysis of scattered wave from a tree trunk of
rasamala (Altingia exelsa) in analysis area. The analysis result compares well with simulation
ones. However, a small error was found. It can be considered that the error is generated by the
FDTD calculation error caused by calculation using the sampled grids. Additionally, in the
same figure, the results of analysis for teak (Tectona grandis), pine (Pinus merkusii), and
mahagony (Swietenia macrophylla) are depicted too.
From these results we know that the increment of tree trunk diameter was directly
proportional to increment of the backscattering coefficient. It means that backscattering is
influenced by the width of tree trunk surface and volume of it. When the surface is wide, the
-15
-10
-5
0
5
10
15
0.0 2.5 5.0 7.5 10.0 12.5
Times (ns)
Bac
ksca
ttere
d A
mpl
itude
(V/m
)
A B
C
54
Figure 3.7. Analysis and simulation results for four species of Indonesian tropical forest,
where the diameter of tree trunk is equal with 2c, where c is radius of tree trunk.
-20
-18
-16
-14
-12
-10
-8
-6
-4
-2
0
0.252 0.262 0.272 0.282 0.292 0.302 0.312 0.322
Diameter of tree trunk [m]
Bac
ksca
tterin
g co
effic
ient
[dB
]
mahagonypinerasamalateaksimulation
55
reflected or scattered field intensity or energy will be higher, and in the contrary of it.
The application of this study will be applied to monitor rasamala (Altingia exelsa)
species that is mainly distributed around Gede Pangrango National Park, west Java, Indonesia
using JERS-1 Synthetic Aperture Radar (SAR, L band) data. The application will be
discussed in the next section.
3.5. Application
3.5.1 Study area
The study area was Cisarua tropical forest, part of Gede Pangrango National Park,
west Java, Indonesia (part of A area in figure 3.8 and 3.9). Figure 3.8 and 3.9 show JERS-1
Visible Near Infra Red (VNIR) and Synthetic Aperture Radar (SAR) data respectively that
acquired at the same area (Path 107 Row 312) on 30 September 1997 and 10 August 1997.
This area is one of the wettest parts of Java with an average annual rainfall of around 3000mm
to 4200mm, while the relative humidity varies between 80% and 90%. Figure 3.10 shows the
altitude distribution of the study area where the study area has altitude ranging from 627m to
2030m as seen in figure 3.11 (DEM 1990). Biomes of this area are the sub-montane (1100m
to 1500m) and montane (1500m to 2030m) tropical forest. In addition to that, settlement and
dry paddy fields called ladang are distributed in the region with altitude between 627m and
1100m above sea level (asl). Tea, which has tall canopy about 1m, is usually planted at
56
Figure 3.8. JERS-1 VNIR data of the study area (Path 107 Row 312, 19970930): Gede
Pangrango National Park, west Java, Indonesia. Remark: A and B show northern
and southern part of the National Park, respectively.
JERS-1 VNIR
Path 107 – Row 312
19970930
A
B
clouds
Study area
57
Figure 3.9. JERS-1 SAR data of the study area (Path 107 Row 312, 19970810): Gede
Pangrango National Park, west Java, Indonesia. Remark: A and B show northern
and southern part of the National Park, respectively.
A
B JERS-1 SAR
Path 107 – Row 312
19970810
Study area
58
Figure 3.10. Altitude distribution of the study area : Mount Gede Pangrango National
Park, west Java, Indonesia.
Study area
S6o45’
S6o50’
E106o55’ E107o00’
59
Figure 3.11. Location of the study area: Gede Pangrango National Park (part of area A in
figure 3.8 and 3.9)
Indonesia
106o56`E
6o38` S
6o44`S
107o02`E
N
800m
1000m
1400m
1200m
1600m
1600m
1200m
1400m
1000m
1400m
1600m
1 0 2km
60
altitude between 627m and 2030m by government companies. Appendix F shows part of the
ground data of the study area. The sub-montane tropical forest has the highest diversity of
plant life and is characterized by large trees forming diameter trunks between 0.2m and 0.4m.
The dominant species in this ecosystem is rasamala (Altingia exelsa). Besides a rich ground
flora containing begonias and ferns, many species of epiphytes are found growing
non-parasitically on twigs and branches (e.g. orchids, lianas, and herbs).
Montane tropical forest has a lower diversity of plants with fewer herb species than
the sub-montane zone. Common trees included rasamala; also noticeable are puspa (Schima
walichii) and conifers (Dacrycarpus imbricatus and Podacarpus neriifolius).
3.5.2 Data processing
The JERS-1 SAR data (figure 3.9) was examined in order to estimate the diameters
of rasamala in the study area. The data (path 107, row 312) was acquired on 10 August 1997
during the dry season in the study area. This data was processed at level 2.1 or standard
geocoded data and was resampled to Universal Transverse Mercator (UTM) projection by the
Earth Observation Research Centre (EORC) of National Space Development Agency
(NASDA) of Japan. Firstly, a 3x3 median filter was employed and the second process used a
5x5 average filter to reduce inherent speckle noise (Sunar et al. 1998). At the same time, the
image was also referenced to the UTM co-ordinate system, through a polynomial rectification
61
Figure 3.12. Classification result assigned the distribution of classes in the study area. Test
area shows classes distribution in ecosystem zones and its terrain conditions.
Ecosystem zones are settlement and paddy (sp), sub-montane (sm) and montane
(mt).
6o38 S
6o44 S
106o56 E 107o00 E
km
sp sp sm sm mt
test area
Altitude units in metres
62
Table 3.2. Relationship between backscattering coefficients and tree trunk diameters of
rasamala forests in the study area
Class names Backscattering
coefficient (dB) Trunk diameter (m) Standard deviation (m)
forest 1 -12 0.290 0.0025
forest 2 -9 0.295 0.0025
forest 3 -7 0.300 0.0025
using 30 ground control points collected from topographic maps scale of 1:25 000
(BAKOSURTANAL 1990). This procedure yielded a geometric accuracy of 0.1 pixels. Then
the spatial resolution of SAR image was resampled to 12.5m.
A supervised classification was performed to classify the data. The study area was
classified into six classes based on topographic maps (BAKOSURTANAL 1990). They were
namely forest 1, forest 2, forest 3, bush (tea), ladang (dry paddy fields), and settlement.
Figure 3.12 shows classification results and the terrain characteristics of each ecosystem zone
in the study area. The classification results contained only 1% error in comparison with the 30
training sites that were sampled from topographic maps. The terrain characteristics of the
study area were generated using digital elevation model (DEM 1990). Figure 3.12 shows that
forest classes distributed in the study area are at the altitude above 1100m or at the
sub-montane and montane zones.
63
The statistical value of each forest class was derived and then the backscattering
coefficient of each class was calculated using NASDA calibrated equation (Shimada 1998).
These results are shown in table 3.2. Where the species included in forest classes in each pixel
were assumed to be rasamala (Altingia exelsa), because it was the dominant species in the
study area (refer Appendix F). By comparing the backscattering coefficient of each forest
class to the curve of rasamala in figure 3.7, as seen in table 3.2, the trunk diameter of each
forest class was obtained. This result shows that trees in sub-montane and montane tropical
forest zones have a trunk diameter between 0.29m and 0.30m with standard deviation of each
class is 0.0025m. These results matched well with the ground data.
3.6 Conclusions
Numerical analysis was conducted to analyse the relationship between the
backscattering coefficients oσ and diameter of tropical tree trunk. The analysis results were
confirmed by simulation using Finite Difference Time Domain (FDTD) method. These results
are in good agreement. A variation of this analysis, it could be applied to estimate diameter of
tropical tree trunk from Synthetic Aperture Radar (SAR) data, which this information is very
important to estimate the forest volumes effectively and accurately. These results succeeded in
estimating tree trunk diameter of rasamala (Altingia exelsa) that is widely distributed in the
Cisarua tropical forest, part of Gede Pangrango National Park, west Java, Indonesia (Tetuko et
64
al. 2001) from JERS-1 SAR data. While this study focused on single site in Indonesia, it is
reasonable to expect that this method or variations should be successful in estimating tree
trunk diameters in similar tropical forest regions of the world using SAR data.
65
References
1. BAKOSURTANAL, 1990, Topographic maps; Cisarua 1209-142, Gunung Hambalang
1209-144. Indonesian National Coordination Agency for Surveys and Mapping, 1st edn
(Cibinong: BAKOSURTANAL).
2. DAVID, P., STELLA, E.B., STHEPHEN, J.M., 1997, Terrain influences on SAR
backscatter around Mt. Taranaki, New Zealand. IEEE Transactions on Geoscience and
Remote Sensing, 35, 924-932.
3. DEM, 1990, Digital Elevation Model: Cisarua 1209-142, Gunung Hambalang 1209-144.
Indonesian National Coordination Agency for Surveys and Mapping, 1st edn (Cibinong:
BAKOSURTANAL).
4. KAMAL SARABANDI, 1989, Electromagnetic Scattering from Vegetation Canopies,
Ph.D Dissertation in the University of Michigan, Michigan: Michigan University.
5. LI, W., KOH, J., YEO, T., KOOI, P., 1999, “Analysis of Radiowave Propagation in a
Four- layered Anisotropic Forest Environment,” IEEE Transactions on Geoscience and
Remote Sensing, vol.37, pp.1967-1979.
6. MUR, G., 1981, “Absorbing boundary Conditions for the finite-difference approximation
of the time-domain electromagne tic- field equation,” IEEE Transactions on
Electromagnetic Compatibility, vol.23, no.4, pp.377-382.
7. SHIMADA, M., 1998, User’s guide to NASDA’s SAR products. Earth Observation
66
Research Centre (EORC), National Space Development Agency (NASDA), 2nd edn
(Tokyo: EORC-NASDA)
8. SUNAR, F., TABERNER, M., MAKTAV, D., KAYA, S., MUSAOGLU, M., and YAGIZ,
E., 1998, The use of multi temporal radar data in agriculture monitoring: a case study in
Kyocegiz-Dalaman ecosystem, Turkey. International Archives of Photogrammetry and
Remote Sensing, 22, 559-565.
9. TETUKO S.S., J., TATEISHI, R., WIKANTIKA, K., 2001, “A method to estimate tree
trunk diameter and its application to discriminate Java-Indonesia tropical forest,”
International Journal of Remote Sensing, vol. 22, no.1, pp.177-183.
10. UNO TORU, 1998, Finite Difference Time Domain Method for Electromagnetic Field
and Antenna Analyses. 1st edition, Tokyo: Corona.
11. YEE, K. S., 1966, “Numerical solution of initial boundary value problems involving
Maxwell’s equations in isotropic media,” IEEE Transactions on Antennas Propagation,
vol.14, no.4, pp.302-307.