EARLY DETECTION OF POTENTIAL FOREST FIRES USING SATELLITE
REMOTE SENSING TECHNIQUES
AIDA HAYATI BINTI MOHD HASSAN
A thesis submitted in fulfillment of the requirements for the award of the degree of
Master of Science (Remote Sensing)
Faculty of Geoinformation Science and Engineering
JULY, 2008
iii
I owe huge gratitude to my family for their love and support, without which this thesis
would most definitely never have been born. The ever patient and encouraging clan, the
big engine that is such a motivation to this little train; Ijan, Ciputt, Bantai, Ina, Mamah,
Aleen, Inon, Rina , Nora, Dayah and Vera.
iv
ACKNOWLEDGEMENTS
Look for good in life and you will find it. I want to express my gratitude to God
who gave me the strength and guide me all the way to look for the good things in my
life.
I wish to thank my honoured supervisor, Profesor Dr Mazlan bin Hashim for his
guidance, patience, and ideas. All his advice and thoughts will always be helpful to me
in the future.
I want to give credit to Professor Felix Kogan for spending his time and thoughts
for this research. I received a lot of material from his journals and research.
My appreciations also go to Encik Hassan Abdul Majid, Encik Wan Hazli Bin
Wan Kadir and Encik Abdul Wahid Bin Rasib of the Department of Remote Sensing.
Last but not least, my thanks to colleagues in the Radar Laboratory for their
strong support and enthusiastism. They are among the people who always look at the
sunny side of everything and think only the best.
v
ABSTRACT
In 1997/1998, Malaysia experienced one of the most severe forest fire episodes
in history as a consequence of a prolonged dry season following the El- Niňo
phenomenon. Since then, uncontrolled fires, atmospheric pollutions and haze remained
as a common problem throughout the dry period in this region. The estimated cost of the
damage caused by forest fires in Malaysia is about RM816.47 million a year. The loss
by forest fire episodes has brought to light the importance of developing better tools for
effective forest fire management systems. In this research, three sets of computer
programmes were designed for: detecting hot spots; computing the fire risk index and
generating spatial analysis for detected fires. Remote sensing and GIS techniques have
both been integrated in this work. Eventually, a simple yet robust early warning system
for forest fire detection in Malaysia has been devised. Thermal bands of MODIS
(Moderate Resolution Imaging Spectroradiometer) were used to extract hot spot
information and to generate a fire risk map. Proximity analysis was carried out using an
extension in ArcView GIS software. The results from this research were compared with
forest fire occurrence information from the Fire and Rescue Department of Malaysia
(FRDM) and information of rainfall and temperature from the Malaysian Meteorological
Services (MMS). High correlation (R2 = 0.8) was found between temperature derived
from MODIS and the temperature obtained from the MMS. Forest fire map generated
from the study also gave a high accuracy (71%). Normalized Difference Vegetation
Index (NDVI) values derived from MODIS were found to be highly correlated (R2 = 0.7
and R2 = 0.85) with rainfall and temperature data obtained from the MMS, respectively.
Hence, the output of the research shows that this system can be used as an early warning
system mechanism to mitigate forest fire incidence and can be upgraded into a more
complex system to enhance its functioning.
vi
ABSTRAK
Pada tahun 1997/98, Asia Tenggara telah melalui episod kebakaran hutan yang
paling ekstrem di dalam sejarah akibat musim kemarau yang panjang kesan fenomena
El-Niňo. Berikutan daripada peristiwa itu, pencemaran atmosfera dan jerebu telah
menjadi masalah yang lazim pada musim kemarau di sini. Anggaran kerugian daripada
kebakaran hutan di Malaysia ialah sebanyak RM816.47 juta setahun. Kerugian yang
dialami akibat daripada episod kebakaran hutan tersebut telah menyedarkan banyak
pihak tentang kepentingan untuk membina sistem pengurusan kebakaran hutan yang
efektif. Di dalam kajian ini, 3 set program komputer telah direka untuk: mengesan titik
panas; mengira indeks risiko kebakaran, dan menjana analisa spatial. Teknik remote
sensing dan GIS telah digabungkan di dalam kajian ini. Dengan itu, sebuah sistem
amaran awal yang ringkas tetapi efektif untuk mengesan kebakaran hutan telah dicipta.
Jalur termal dari MODIS (Moderate Resolution Imaging Spectroradiometer) telah
digunakan untuk mengekstrak informasi titik panas dan menjana peta risiko kebakaran
hutan. Analisa spatial dilakukan dengan menggunakan fungsi dari perisian ArcView.
Hasil dari kajian ini dibandingkan dengan data kebakaran hutan dari Jabatan Bomba dan
Penyelamat, Malaysia, dan maklumat taburan hujan serta suhu dari Jabatan Kajicuaca,
Malaysia. Nilai korelasi yang tinggi (R2 = 0.8) telah diperolehi diantara suhu yang
diekstrak dari MODIS dengan suhu dari Jabatan Kajicuaca. Peta kebakaran hutan yang
diperolehi juga mempunyai ketepatan yang tinggi (71%). Nilai Normalized Difference
Vegetation Index (NDVI) yang diperoleh dari MODIS juga mencatatkan korelasi yang
tinggi (R2 = 0.7 and R2 = 0.85) dengan data jumlah hujan dan suhu dari Jabatan
Kajicuaca. Dengan itu, hasil dari kajian ini menunjukkan bahawa ianya boleh digunakan
sebagai satu mekanisma sistem amaran awal untuk mengurangkan kejadian kebakaran
hutan dan boleh diperkembangkan lagi menjadi sebuah sistem yang lebih kompleks
untuk meningkatkan lagi fungsinya.
vii
TABLE OF CONTENTS
CHAPTER
TITLE
PAGE
DECLARATION
I
II
DEDICATION
iii
ACKNOWLEDGEMENTS
iv
ABSTRACT
v
ABSTRAK
vi
TABLE OF CONTENTS
vii
LIST OF TABLES
xii
LIST OF FIGURES
xiii
LIST OF SYMBOLS
xvi
INTRODUCTION
1.1
General Introduction
1
1.2
Problem Statement
4
1.3
Objectives
6
1.4
Scope
6
1.5
Significance of the Study
7
1.6
Area of Study
8
1.7
Thesis Structure
9
FOREST FIRE MODEL
2.1 Introduction
11
viii
2.2 Terminology of Forest Fires
12
2.3 Influencing Factors
13
2.4 Causes of Forest Fires
14
2.5 Types of Forest Fires
15
2.5.1
Ground Fire
15
2.5.2
Surface Fire
16
2.5.3
Crown Fire
16
2.6 Fire Potential Measurement Components
19
2.6.1
Changes in Fuel Load
20
2.6.2
Fuel availability
20
2.6.3
Weather Variables
21
2.7 Evaluation of Fire Risk
21
2.7.1
Structural Fire Indices
22
2.7.2
Dynamic Risk Indices
23
2.7.3
Advanced Forest Fire Indices
24
2.8 Phases of forest fire development
24
2.8.1
Pre-ignition
25
2.8.2
Flaming phase
26
2.8.3
Glowing phase
26
2.8.4
Smouldering phase
26
2.9 Review of Forest Fire Models
27
2.10
31
Review of Existing Computer Code and Running
Application for Forest Fire
2.10.1 BEHAVE
32
2.10.2 FireLib
33
2.10.3 FARSITE
33
2.10.4 FORFAIT
34
2.10.5 FOMFIS
34
2.11 Review of Fire Detection Techniques Using Remote
35
Sensing
2.11.1 GOES Fire Detection
36
ix
2.11.2 Defense Mapping Satellite Program (DMSP) Fire
37
Detection
2.11.3 Landsat Fire Detection
37
2.11.4 NOAA AVHRR Fire Detection
38
2.12
III
Summary
39
METHODOLOGY
3.1 Introduction
41
3.2 Data Acquisition
43
3.2.1 Satellite Data
43
3.2.1.1 MODIS
43
3.2.2 Ancillary Data
46
3.2.2.1 Topographic Map
46
3.2.2.2 Vector Data
47
3.2.2.3 Forest Fire Occurrences Data
48
3.2.2.4 Fire and Rescue Department Distribution
48
3.3 Data Pre-processing
3.3.1 Geometric Correction
3.4 Data Processing
48
48
50
3.4.1 Normalized Difference Vegetation Index (NDVI)
50
3.4.2 Vegetation Condition Index (VCI)
51
3.4.3 Brightness Temperature (BT)
52
3.4.4 Temperature Condition Index (TCI)
53
3.4.5 Vegetation Health Index (VH)
53
3.4.6 Fire Risk Map
54
3.4.7 Hot Spot of MODIS
54
3.4.8 Proximity Analysis
58
3.4.9 Development of Forest Fire Interface
59
3.5 Accuracy Assessment
59
3.6 Summary
59
x
IV
V
FOREST FIRE INTERFACE
4.1. Introduction
61
4.2. System Development Life Cycle
62
4.3. Overview of Forest Fire Interface
64
4.4. Forest Fire Interface
68
4.5. Summary
76
RESULTS AND ANALYSIS
5.1. Introduction
77
5.2. Hot Spots Detected From MODIS
78
5.3. Comparison of Temperatures Extracted From MODIS
86
With Temperature Data From MMS
5.4. Generating A Fire Risk Map
5.4.1. Daily Normalized Difference Vegetation Index
88
89
(NDVI)
5.4.2. Smoothed Weekly Normalized Difference
89
Vegetation Index (NDVI)
5.4.3. Minimum and Maximum Normalized
90
Difference Vegetation Index
5.4.4. Brightness Temperature (BT)
95
5.4.5. Smoothed Weekly Brightness Temperature
95
(BT)
5.4.6. Minimum and Maximum Brightness
95
Temperature
5.4.7. Vegetation Condition Index (VCI)
100
5.4.8. Temperature Condition Index (TCI)
100
5.4.9. Vegetation Health Index (VH)
100
5.5. Fire Risk Map
101
5.6. Vegetation Index Analysis
105
5.7. Fire Risk Map Analysis
109
5.8. Summary
109
xi
VI
CONCLUSIONS AND RECOMMENDATION
6.1. Introduction
111
6.2. Conclusions
111
6.3. Recommendation
112
REFERENCES
114
Appendices A-C
125-164
xii
LIST OF TABLES
TABLE NO.
TITLE
PAGE
1.1
Area of forest type burned in 1998 in Malaysia
3
1.2
Costs of the damage caused by smoke haze in Malaysia.
4
2.1
Existing computer code and running application for forest
fire.
35
2.2
GOES-7 VAS satellite spectral channels
36
3.1
MODIS spectral bands characteristics
44
3.2
Number of datasets processed in this study for each year
from January 2000 to April 2005
46
3.3
Topographic maps used to execute geometric correction
47
3.4
List of maps used to extract vector layers
47
3.5
List of ground control point for MODIS data on 27
January 2005
49
3.6
Attributes generated from the MOD14 data
56
4.1
Explanation of system development life cycle
63
5.1
List of hot spots detected from MODIS from 1 January to
30 April 2005
80
5.2
Peat utilization in Malaysia.
84
xiii
LIST OF FIGURES
FIGURE
NO.
TITLE
PAGE
1.1
Peninsular Malaysia
9
2.1
(a) Fire fundamentals triangle; and (b) fire environment triangle
13
illustrating the factors of fire combustion and factors controlling
fire propagation, respectively.
2.2
Types of fires; (a) Ground fires, (b) Surface fires, (c) Crown fires.
18
2.3
Proposed approaches for the evaluation of forest fire risk indices
22
2.4
Phases of forest fire development
25
2.5
Structure of Canadian Forest Fire Danger Rating System
(CFFDRS).
29
2.6
The structure of Fire Weather Index (FWI)
30
2.7
The structure of Fire Behaviour Index (FBP), including the
necessary inputs and the outputs produced by the system
31
3.1
Flow chart of operational methodology
42
3.2
MODIS data on 28 January 2005 with composite of band 1, 2 and
4
43
3.3
Twelve GCP’s were used to geometrically correct the MODIS
49
image of 27 January 2005
3.4
Geometrically corrected image of MODIS on 27 January 2005
50
4.0
Diagram of system development life cycle
62
4.1
Explanation of system development life cycle
63
4.2
Forest fire interface diagram component
64
4.3
Diagram of vegetation index components namely Vegetation
Condition Index (VCI), Temperature Condition Index (TCI),
65
xiv
Vegetation Health Index (VH) and Fire Risk Index
4.4
Flow of NDVI and brightness temperature data as input to carry
out Fire Risk Index
66
4.5
Input data of MOD021KM and MOD03 used to extract the hot
spot
66
4.6
The main window for developed forest fire interface
68
4.7
Vegetation Index window
69
4.8
Information window of the displayed image
69
4.9
NDVI window prompt
70
4.10
VCI window prompt
71
4.11
TCI window prompt
71
4.12
VH window prompt
72
4.13
Fire Risk Map window prompt
73
4.14
Prompt window to set up the number of input weeks to Fire Risk
Map
73
4.15
Hotspot Index window prompt
74
4.16
MS DOS command used to derive the MODIS fire mask using
the MODIS Level 1B Radiances and Geolocation products
74
4.17
MS DOS window prompt to convert hotspot data into shapefile
75
4.18
ArcView function, ‘Closest Feature submenu used to extract
proximity analysis
76
5.1
Distribution of hot spots by land use type
83
5.2
Extreme drought occurrence in (a) Kampong Teluk Jambu
Bintong, Kangar, Perlis; and (b) Firemen battling bush fire
outside the Penang International Airport cargo complex in Bayan
Lepas
85
5.3
Correlation between temperature from Malaysian Meteorological
Services (MMS) and observed temperature from MODIS
87
5.4
NDVI derived from single MODIS dataset on 6 April 2000
91
5.5
Smoothed weekly MODIS datasets from 5th week of 2005
92
5.6
Multi-year maximum MODIS datasets
93
5.7
Multi-year minimum MODIS datasets
94
5.8
Brightness temperature derived from single MODIS dataset on 5
April 2000
96
5.9
Smoothed weekly brightness temperature of MODIS datasets
derived from 14th week of 2005
97
xv
5.10
Multi-year maximum of brightness temperature derived from
MODIS
98
5.11
Multi-year minimum of brightness temperature derived from
MODIS
99
5.12
Vegetation Condition Index derived from 15th week of 2005
102
5.13
5.14
th
Temperature Condition Index derived from 15 week of 2005
th
Vegetation Health Index derived from 15 week of 2005
103
104
5.15
th
Fire Risk Map of 11 week of 2005
106
5.16
The relationship between VCI and rainfall
107
5.17
Relationship between TCI and temperature
108
5.18
The relationship between VH and temperature
108
xvi
LIST OF SYMBOLS
EEPSEA
-
Economy and Environment Program for Southeast Asia
NDVI
-
Normalized Difference Vegetation Index (NDVI)
BT
-
Brightness Temperature
VCI
-
Vegetation Condition Index
TCI
-
Temperature Condition Index
VH
-
Vegetation Health Index
FRDM
-
Fire Rescue Department of Malaysia
MMS
-
Malaysian Meteorological Services
KBDI
-
Keetch and Byram Drought Index
FWI
-
Canadian Forest Fire Weather Index
FBP
-
Canadian Forest Fire Behavior Prediction System
CFFDRS
-
Canadian Forest Fire Danger Rating System
FOP
-
Fire Occurrence Prediction System
FFMC
-
Fine Fuel Moisture Code (FFMC)
DMC
-
Duff Moisture Code
DC
-
Drought Code
ISI
-
Initial Spread Index
BUI
-
Buildup Index
FFDM
-
Forest Fire Danger Meter
API
-
Application Programming Interface
GIS
-
Geographical Information System
VAS
-
Visible Infrared Spin Scan Radiometer and Atmospheric Sounder
xvii
SWIR
-
Shortwave infrared
LWIR
-
Longwave infrared
μm
-
Micrometer (1 meter = 1 000 000 μm)
km
-
kilometer
mm
-
Milimeter
DMSP
-
Defense Mapping Satellite Program
TM
-
Thematic Mapper
K
-
Kelvin
NOAA
-
National Oceanic and Atmospheric Administration
AVHRR
-
Advanced Very High Resolution Radiometer
MSS
-
Multispectral Satellite
MODIS
-
Moderate Resolution Imaging Spectroradiometer
SNR
-
Signal noise to ratio
NEΔT
-
Noise-equivalent temperature difference
CZCS
-
Nimbus Coastal Zone Color Scanner
HRIS
-
High Resolution Infrared Sounder
GES DAAC
-
Goddard Earth Sciences Distributed Active Archive Centre
GCPs
-
Ground control points
RMS
-
Root Mean Square
VIS
-
Visible
NIR
-
Near infrared
NDVImin
-
Multiyear absolute minimum of NDVI
NDVImax
-
Multiyear absolute maximum of NDVI
h
-
Planck’s constant (Joule per hertz)
c
-
Speed of light in vacuum (m/s)
k
-
Boltzmann gas constant (Joule/Kelvin)
λ
-
Band or detector centre wavelength (m)
T
-
Temperature (Kelvin)
BTmin
-
Absolute minimum of smoothed weekly brightness temperature
BTmax
-
Absolute maximum of smoothed weekly brightness temperature
T4
-
Brightness temperature of 4 micrometer channel
xviii
T11
-
Brightness temperature of 11 micrometer channel
Ť
-
Respective mean of the channel for valid neighbouring pixel
Naw
-
Number of water pixel adjacent to the fire pixel
Nac
-
Number of cloud pixel adjacent to the fire pixel
δ
-
Mean absolute deviation of the respective channel for valid
neighboring pixel
C
-
Confidence level
TIFF
-
Tagged Image File Format
UTM
-
Universal Transverse Mercator
MOD01
-
Raw MODIS data
MOD02
-
Level 1B MODIS data
MOD03
-
Geolocation MODIS data
MOD035
-
MODIS Cloud Mask product
MOD14
-
MODIS Fire product
xix
LIST OF APPENDICES
APPENDIX
TITLE
PAGE
A
Forest Fire Record From FRDM
125
B
Fire risk level for each forest fire occurrence in time range
127
of 1 January 2005 to 30 April 2005.
C
Visual C++ source code
129
CHAPTER I
INTRODUCTION
1.1
General Introduction
By the end of 1990, an estimated area of 5.55 million hectares of forest
covered 42.2% of Peninsular Malaysia’s total land area (Khali, 2001). Of the total,
5.51 million hectares are classified as evergreen rainforest consisting of 4.94 million
hectares of Dipterocarp Forests, 0.46 million hectares of peat swamp forest, 0.11
million hectares of mangrove forest while the remaining 0.04 million hectares
consists of estate and agricultural areas developed since as far back as 1957 (Mohd
Shahwahid, 2004). Generally, the low temperature and moist condition in the natural
forests that give rise to a high rate of litter decomposition contribute to the low
occurrence of large scale forest fire in Malaysia.
In the last three decades however, large tracts of forestlands have been
planted with monoculture crops. Some 1.65 million hectares of rubber and 2.62
million hectares of oil palm have been established, posing a higher fire risk than the
natural forests (Hussin, 2000). Also logging activities in the natural forests produce a
lot of waste, thereby increasing flammable material, and opening canopies
2
which reduce the water retention capacity of the forests, become more susceptible to
fire.
In the past 10 years, there has been an increasing incidence of major fires
especially peat forest fires in the Southeast Asia region. In Malaysia, the worst forest
fire was reported in Sabah from 1983 to 1985 (Mat Isa, 2001). In these incident, over
one million hectares of mostly logged-over forests were burnt. In East Kalimantan,
Indonesia, a fire occurrence that started in September 1982 lasted for 10 months and
affected more than 35,000 hectares of peatlands (Suhaili and Mohd Yunus, 1999). It
has been said that Indonesia needs about 500 years to return the forest back to its
primary condition. Anyway, to capture back the ecosystem equalization and variation
in the biology seems to be impossible. This problem is further compounded by the
fact that some of the affected areas have been burned twice or more. If left unabated,
peat areas that will be at risk to burn will be on the increase.
In terms of forest type, the peat forests suffered the most with 63,331 hectares
(98%) burned in 1998 (Table 1.1). Land clearing for agriculture by the farmers in the
dry season was identified as the most likely cause of the forest fires.
Forest fires could cause a serious environmental problem. During the 1997
big fire event, about 70 million people were forced to breathe polluted air because of
the haze from the forest fire. At least six people died from direct health effects and
many more were hospitalised (Wan Ahmad, 2001). Poor visibility that was brought
about by haze increased the risk of accidents on land. In some regions, schools and
shops were forced to close down and business matters were hindered. The tourism
industry experienced the worst impact from this event due to the decrease in the
number of visitors and a large number of flights cancelled as a precaution.
3
Table 1.1: Area of forest type burned in 1998 in Malaysia.
Area
Forest Type
Peat Forests
Probable cause
(Hectares)
63,331 Land clearing by farmers and indigenous
people, hunting and other unknown causes.
Secondary Forest
432 Land clearing by farmers
Degraded Heath Forest
310 Land clearing by farmers
Heath Forest
250 Unidentified
Logged-over Forest
120 Unidentified
Forest Plantation
26 Snapped
electrical
transmission
lines,
cigarettes.
Montane Forest
15 Campers
Coastal Swamp Forest
15 Clearing by fishing villagers
Total
64,499
(Source: Ahmad Zainal, 2000)
Consequently, the risk from forest fires to private property and human life has
increased making fire fighting more complicated, expensive, and dangerous.
Indonesia lost RM16.72 million in the battle to fight the fires (Suhaili and Mohd
Yunus, 1999). According to the Economy and Environment Program for Southeast
Asia (EEPSEA) study, the estimated incremental cost of the haze damage to
Malaysia during the months of August to October in 1997 was RM816 million
(Table 1.2).
The transboundary nature that exists in the forest fire problems has suggested
a network approach for sharing of information and experience. In October 1997,
1262 firefighters from Malaysia were deployed to Sumatra and Kalimantan to
combat the forest fires (Wan Ahmad 2001). Besides supplying help, Malaysia also
gained some experience and knowledge which is useful for addressing similar issues.
4
Table 1.2: Costs of the damage caused by smoke haze in Malaysia.
Type of Damages
RM Million
Percentage
Adjusted cost of illness
36.16
4.43
Productivity loss during the emergency
393.51
48.19
Tourist arrival decline
318.55
39.02
Flight cancellations
0.45
0.06
Fish landing decline
40.72
4.99
Cost of fire fighting
25.00
3.06
Cloud seeding
2.08
0.25
816.47
100.00
Total
(Source: Mohd Shahwahid and Jamal,1999)
1.2
Problem Statement
Forest fires arise from a sudden encounter between oxygen and fuel material
at high temperatures. They would usually damage the forest ecosystems resulting in a
decline in biological diversity, environmental degradation, soil erosion, atmospheric
and water pollution. The main pollution is the occurrence of haze. Only in the past
decade have researchers realized the important contribution of biomass burning to
the global budgets of many radiatively and chemically active gases such as carbon
dioxide, carbon monoxide, methane, nitric oxide, tropospheric ozone and elemental
carbon particulates (Kaufman, 1998; Hashim et. al., 2004). These factors result in
negative implications on the socio-economic, health and well being of the human.
Besides that, the effects of forest fires are not local, they are cumulative and
contribute to regional and global problems such as deforestation, global warming, or
desertification.
5
Peat forest fires are a common problem in South East Asia including
Malaysia. Controlling peat fires is extremely difficult. Peat will burn from the bottom
of the soil layer. The fires burn in a slow and patchy manner, and are widespread.
The fires spread slowly through the thick peat layers, making it extremely difficult to
detect and extinguish them. In the case of peat, although the surface fires are
extinguished, the underground peat will continue to burn unless a large amount of
water is used to completely drench the peat layers.
Usually, peat fires are hard to notice. Peat fires are deep underground and can
burn uncontrolled and unseen for several months. This fire is potentially a great
threat to human health than the Kuwaiti oil fires and hard to extinguish. One cannot
notice underground fires (such as peat fires) until occurrence of smouldering smoke,
which are then naturally followed by fires. During this time, fires have already
burned all the organic material under the ground. Fortunately, much has changed in
the field of fire fighting. Today, an early warning indicator is an important
component in the total fire management system. Due to the difficulty of detecting
underground peat fires and the absence of detailed fire records in many areas, the
possibility of using remote sensing satellite observations will be examined.
Human eyes cannot detect subtle differences in thermal infrared energy
emanating from the underground peat fires because they are only primarily sensitive
to short wavelength visible light from 0.4 to 0.7μm. Human eyes also are not
sensitive to the reflective infrared in the range of 0.7 to 3.0μm, or thermal infrared
energy in the range of 3.0 to 14μm. Nowadays, thermal detectors which are sensitive
to thermal infrared radiation have been designed and manufactured widely. With this
advance thermal sensor, thermal remote sensing related studies such as thermal
characteristic of the landscape, forest fires and urban heat are possible.
6
1.3
Objectives
The objectives of the study are divided into three:
i)
To determine hot spots from MODIS data;
ii)
To derive a fire risk map based on biophysical parameters from MODIS data,
and
iii)
To develop forest fires interface using remote sensing data derived from
objectives (i) and (ii).
1.4
Scope
MODIS datasets were used in this study for several reasons; namely (i) the
thermal (10.78-11.28μm) bands of MODIS were specifically developed to detect hot
spots along with well developed algorithm for forest fire purpose (Kaufman, 1998).
In addition, this band is also used in extracting brightness temperature (BT) as a
parameter in generating the vegetation index; and (ii) the visible (620-670nm) and
near infrared (841-876nm) bands of MODIS are suitable to extract Normalized
Difference Vegetation Index (NDVI), which is a parameter to generate the fire risk
map.
Two biophysical parameters are used to generate the vegetation indices,
namely Normalized Difference Vegetation Index (NDVI) and brightness temperature
(BT). Three indices were developed from the biophysical parameters; namely (i)
Vegetation Condition Index (VCI), for estimation of cumulative moisture impacts on
vegetation; (ii) Temperature Condition Index (TCI), for estimation of thermal
7
impacts on vegetation; and (iii) Vegetation Health Index (VH), for estimation of
moisture and thermal impacts on vegetation. Cumulative of four weeks of VH were
used to generate the fire risk map.
Cumulative daily MODIS datasets from 2000 to 2005 were processed in
order to generate the fire risk maps. This is however excluding the datasets that
covered more than 10% cloud. In order to minimize the cloud effects, the NDVI and
BT datasets were composited over a 7-day period. This will reduce the cloud effects.
1.5
Significance of the Study
During the 1997-98 Indonesian forest fires, many national governments and
international agencies tried their best to help Indonesia and other affected countries
to suppress the fires in order to minimize the impacts. However, there was a lack of
information about active forest fires and base line data about water resources,
transport networks, land use patterns, topography types and location of settlements.
Although emphasis was given to immediate fire suppression, it is realized that there
should be long-term strategy measures to mitigate the occurrence of disasters due to
forest fires in future. The establishment of an effective early warning mechanism was
the focus of discussion by most of the workers involved with forest fire mitigation in
the ASEAN region.
This study can provide an alternative to conventional techniques in detecting
and monitoring forest fire in an effective way. It is necessary to develop an effective
system for forest management in order to control forest fires occurrence. The final
result of this study will provide a valuable tool for fire agencies as a fire-prediction
strategy. It will provide benefit as one of the technologies used to assist monitoring
8
and evaluation of forest fire risk. Nowadays, with the existence of advanced
technology, we can reduce the damage of forest fires.
The development of a detecting and mitigating system in this study can be an
advantage to the Department of Environment and the Department of Forestry, and
also to other related agencies in order to improve forest management and
sustainability. It can also be utilized for various purposes such as teaching and
increasing awareness among decision makers in the region about the availability of
remote sensing technology to mitigate disasters. Satellite remote sensing monitoring
can also provide advance weather information or data that can assist in studying the
evolution of fires as they develop.
1.6
Area of Study
The area of study is Peninsular Malaysia, located from 1o 20’ N to 6o 40’ N
latitude, and 99o 35’ E to 104o 20’E longitude (Figure 1.1). The total area of
Peninsular Malaysia is 131,794 km2. Topographically, Peninsular Malaysia is
characterized by extensive coastal plains in the east and west, hilly and mountainous
region with steep slopes in the middle and undulating terrain in other parts of the
peninsular.
The climate of Malaysia is typical of the humid tropics and is characterized
by year-round high temperature and seasonal heavy rain. Temperature ranges from
26oC to 32oC and rainfall ranges from 2,000 mm to 4,000 mm per annum.
9
Figure 1.1: Peninsular Malaysia
1.7
Thesis Structure
This thesis will consist of 6 chapters namely; (1) Introduction; (2) Forest Fire
Model; (3) Methodology; (4) Forest Fire Interface; (5) Result and Analysis; and (6)
Conclusions and Recommendations.
The first chapter of this study discusses introduction to forest fires
occurrences. The objectives, scope, problem statement and significance of the study
are also included in this chapter.
Chapter 2 explains about forest fires briefly from the terminology of the fire
itself, to the method used to detect and monitor it. The methods discussed include
10
conventional techniques, as well as the combination of conventional and remote
sensing technology.
The method used in this study is discussed in Chapter 3. This chapter
explains in detail the materials used in this study. The methods used to extract
hotspots, Normalized Difference Vegetation Index (NDVI), brightness temperature
(BT), Vegetation Condition Index (VCI), Temperature Condition Index (TCI) and
Vegetation Health (VH) are explained together with the analysis.
Chapter 4 explains the forest fire interface from the early phase of its
development until the end users requirement.
Chapter 5 consists the results obtained from this study. The analysis is carried
out at the end of the chapter. For hot spot analysis, the results from this study are
compared to ground truth forest fire occurrences data obtained from the Fire Rescue
Department of Malaysia (FRDM). In order to determine the accuracy of the
temperature derived from MODIS data, the derived temperature were compared with
field temperature obtained from Malaysian Meteorological Services (MMS).
The last chapter, namely Chapter 6, discusses the conclusions derived from
this study and also the recommendations to improve this study in future.
CHAPTER II
FOREST FIRE MODEL
2.1
Introduction
Forest fire is a prominent global phenomenon. It has occurred long before the
advent of humans in order to shape landscape structure, pattern and ultimately the
species composition of ecosystems. The ecological role of fire is to influence several
factors such as plant community development, soil nutrient availability and
biological diversity (Levine et al., 1999). Forest fire is a vital and natural process that
initiates natural cycles of vegetation succession and maintains ecosystem viability.
In recent years, the combination of climate and human activity account for
the majority of forest fire occurrence. The extensive use of natural and forest regions
as recreational areas, for agricultural purpose and other development factors have
increased the number of human caused fires. During 1997 and 1998, combination of
drought conditions brought by El Niňo and uncontrolled burning practices caused
unprecedented levels of forest fires across the globe (Anderson et. al., 1999).
12
In times of severe droughts, forest fire occurrence may bring about a massive
destruction of forest richness and severe transboundry pollution in the form of smoke
and haze. Biomass burning resulting from forest fires can also be identified as a
significant source of aerosols, carbon fluxes and trace gases, which pollute the
atmosphere and contribute to radiative forcing responsible for global climate changes
(Kaufman, 1990).
2.2
Terminology of Forest Fires
By definition a forest fire is any wild land fire not prescribed for the area by
an authorised plan. A more descriptive definition is uncontained and freely spreading
combustion, which consumes the natural fuels of a forest that is duff, litter, grass,
dead branch wood, snags, logs, stumps, weeds, brush, foliage, and to a limited
degree, green trees (Brown and Davis, 1973). When referring to the South East Asia
situation, fires on land within natural forest area often burn a little except grassland
and scrub. In areas outside forest land, it is again agricultural waste, land clearance
and peat soil that are most vulnerable to fire. Fires in these non-forested areas are
sometimes referred to as forest fires (Anderson et. al., 1999).
According to FAO’s terminology (FAO, 1986) forest fire risk is the
probability of fire initiation due to the presence and activity of a causative agent.
Bachman and Allgower (1998) define fire risk as the probability of a fire to occur at
a specified location and under given circumstances and its expected outcome as
defined by the impacts on the affected objects.
13
2.3
Influencing Factors
Basically, fire is a chemical reaction called rapid oxidation. Fire is produced
only when heat, fuel and oxygen (Johnson, 1999) are present in the right amounts.
The basic principle of fire control is to remove one or more of these elements in the
quickest and most effective manner.
From Figure 2.1(a), fuel generally is available in ample quantities in the
forest. Fuel must contain carbon. It comes from living or dead plant materials or
organic matter. Trees and branches lying on the ground are a major source of fuel in
a forest. Trees and branches left on the ground after a logging operation can become
fuel too. Peat is also one of the sources of fuel in a forest. Peat is an accumulation of
partially decomposed and disintegrated plant remains, which has been fossilized
under conditions of incomplete aeration and high water content (Bell, 1983). Dried
peat is easier to ignite.
Fire
Fundamental
Triangle
(a)
Fire
Environment
Triangle
(b)
(Modified from Andrew and Williams, 1998)
Figure 2.1: (a) Fire fundamentals triangle; and (b) fire environment triangle
illustrating the factors of fire combustion and factors controlling fire propagation,
respectively.
14
Oxygen is present in the air. As oxygen is used up by fire, it is replenished
quickly by the wind, while heat is required to start and maintain fire. Heat can be
supplied by nature through solar radiation. Human beings can also supply a heat
source through misuse of matches, campfires, trash fires, and cigarettes. Logging
equipment, trains, and automobile exhaust systems can also supply a heat source for
a fire. Once fire has started, it provides its own heat source as it spreads.
Once sustained ignition occurs, elements of the fire environment triangle
from Figure 2.1(b) become the controlling factors in fire propagation. The fire
environment triangle represents the influence of fuel, topography and weather in a
fire occurrence. These are the influencing factors that must be examined, as they
change in space and time, in order to assess the risk of uncontrolled fire.
Fuel moisture content is a function of short term weather variables such as
precipitation and relative humidity, while topography can concentrate winds. Wind
will contribute oxygen to the fire, increasing fire intensity and rate of spread.
Andrews and Williams (1998) found that positive slopes place unburned fuels
physically closer to the heat and flames, also increasing the rate of spread and fire
intensity.
2.4
Causes of Forest Fires
Fire has co-existed with human activities for many years. Burning is a
universal human cultural trait. Therefore, motivation for starting fires are ancient
patterns that have remained the same throughout the ages. The extensive use of
natural and forest regions as a recreational area has increased the number of human
caused fires. On the other hand, the decrease of rural population and abandonment of
agricultural regions has led to the build up of fire fuels that increase the potential of
15
forest fire incident. Human are considered responsible for about 90% of global
biomass burning (Levine et al., 1999) and that is where most of the forest fires
research emphasis is laid on.
Yet depending on the regional patterns, lightning can be quite important
cause for wildfires. Pyne et al. (1996) found that in the Pacific regions of the United
States of America (USA), 31% of wildfires are caused by lightning. This number
roughly corresponds to the results of a study in Switzerland where 26% of the
recorded fires are caused by lightning (Langhart et al., 1998). However, in general,
natural causes do not seem to be of great interest for the wildfire research community
(Bachmann and Allgower, 1998).
2.5
Types of Forest Fires
There are basically three common fire types; namely (a) ground fire, (b)
surface fire, and (c) crown fire. Illustration of these fire types is shown in Figure 2.2.
Classification of fires is based on the degree to which fuels from organic soil upward
to top of the tree were involved in the combustion.
2.5.1
Ground Fire
As illustrated in Figure 2.2(a), ground fire burned in ground fuels such as
organic soils, roots and buried logs. Ground fuels are characterized by higher bulk
density than surface and canopy fuels (Scott and Reinhardt, 2001). Ground fires burn
with very low spreading rates but can be sustained at relatively high moisture
16
contents (Frandsen, 1987, 1991). Fuel consumption through ground fire can be great,
causing significant injury to trees and shrubs. In ground fires, the heat is so intense
that it might cause to the loss of underground structure and death of microorganisms.
Organic matter and soil nutrients such as nitrogen become volatilized and escape to
the atmosphere. Less intense fires have the advantage of breaking standing litter into
smaller pieces, thereby making it more available for decomposition. Pathogenic
fungi may also be killed, allowing more seedlings to emerge (Bozniak et al., 2005).
Ground fire is the most intense fire compared to the other types of fire.
2.5.2
Surface Fire
As illustrated in Figure 2.2(b), surface fire burns in the surface fuel layer,
which lies immediately above the ground fuels but below the canopy, or aerial fuels.
Surface fuels consists of leaves, grass, dead and fallen branches, wood and logs,
shrubs, low brush, and short trees (Brown and Davis, 1973). Commonly, it occurs in
grasslands, where there are no trees and the litter and vegetation is spread fairly
evenly over the surface. Surface fire is the least intense among the three types of fire.
The soil surface may become quite hot, but nutrients, roots and microorganisms
below the surface are left intact.
2.5.3
Crown Fire
Crown fires occur when the tops of trees are heated sufficiently to ignite so
that the fire travels explosively from tree to tree as shown in Figure 2.2. Canopy fuels
normally consumed in crown fires consist of the live and dead foliage, also fine live
and dead branchwood found in a forest canopy (Call and Albini, 1997). Essentially
17
all the vegetation is burned, some areas of ground may even be heated sufficiently to
burn out humus and kill roots. Forests of lodge pole pine, Bishop pine, sand pine and
jack pine that are usually found in Europe typically burn with crown fires
(Casagrandi and Rinaldi, 1999). Canopy fires are more intense than surface fires due
to the fact that there is more fuel involved, therefore the heat generated is greater
than that found in surface fires. But on the other hand, surface fire is less intense than
ground fire.
According to Casagrandi and Rinaldi (1999), there are four major groups of
forest that produce different type of fuel namely rain forests, boreal forests, savannas
and Mediterranean forests. Each group of forest has different unique characteristics.
Rain forests are dense and humid, and layers are not easily recognizable. While
boreal forests are characterized by high density of large conifers and scarcity of bush,
the prevalent understory species being bryophytes and lichens. By contrast, trees are
quite rare while herbs are very dense in savannas. Finally, in Mediterranean forests,
both understory and overstory species are important. Fires in rain forests usually
occur by accident, while they are recurrent in other forests. Ground fire can occur in
rain forests on peat soil. But fire regimes in boreal forest, savannas and
Mediterranean forests are remarkably different. Payette (1989) found that fires in
boreal forests prevalently involve crowns, and the fire return times are typically 50200 years. Fires in savannas are surface fires and return times are typically 1-2 years
in moist savannas and 5-10 years in arid savannas (Rutherford, 1981). While in
Mediterranean forests, fire can be either crown or surface fires and occur in an
apparently random sequence.
18
(a)
(b)
(c)
(Modified from Scott and Reinhardt, 2001)
Figure 2.2: Types of fires; (a) Ground fires, (b) Surface fires, (c) Crown fires.
19
2.6
Fire Potential Measurement Components
Different approaches are used to compute potential fire occurrence. Fire risk
indices are used in Europe, while fire danger rating systems are used in the United
States. In Malaysia, an extreme dry season is always used as an early warning of fire
occurrence. Practically, scientists used numerical systems to provide early warning
of conditions conducive to the onset and development of fire occurrence. The
concept of fire potential involves both tangible and intangible factors, physical
processes and hazard events. By definition, fire potential is a general term used to
express an assessment of both constant and variable fire danger factors affecting the
inception, spread, intensity and difficulty of control of fires and the impact they
cause (Bachmann and Allgower, 1998).
Fire potential produce qualitative and numerical indices of fire danger that
function as an early warning to fire threat (Andrews and Williams, 1998). Different
systems with varying complexity have been developed throughout the world with
respect to the severity of the fire climate and environment. The simplest systems use
only temperature and relative humidity to provide an index of the potential for fire
starts such as the Angstrom Index (Skvarenina, 2003). The Angstrom Index was
devised in Sweden and has been used all over Scandinavia.
While a single fire potential index may be useful to provide early warning of
a fire situation over broad areas, it is unable to give a complete picture of the daily
fire danger with a single index. Therefore, it is necessary to break a fire potential
system into its major components to appreciate where early warning systems for
single factors fall into the overall picture of potential. These fall into three broad
categories as explained below.
20
2.6.1
Changes in Fuel Load
The first element of early warning for a potential fire risk is a major shift in
the forest fuel condition. Farmer’s activity draining the water of peat swamp for
agricultural purposes will decrease the thickness of the soil. When the dry season has
arrived, peat soil become drier and has a high possibility to ignite. Fire can change
the natural soil moisture content. Alfian (1999) found that moisture content for peat
is 47.65%, but the moisture content of the soil will decrease to 28.15% once it is
burned. As for vegetation, a prolonged dry season causes extreme vegetation stress.
The physiological condition of live vegetation has a strong bearing on fire
occurrence (Kogan et al., 2003).
2.6.2
Fuel availability
The seasonal change in fuel availability as fuels dry out during the drought
period sets the stage for severe fire occurrence. Under drought conditions fuel is
available for combustion. Organic soils may dry out and become combustible.
Drought stress on living vegetation not only reduces the moisture content of the
green foliage but also dried plant matter such as leaves and bark can be shed adding
to the total load of the surface fuel. However, most devastating fires occur when
severe fire weather variables are combined with extreme drought.
21
2.6.3
Weather Variables
Less rain, low humidity and high temperature are very significant for fire
occurrence. Regular charting of bookkeeping systems such as the Keetch-Byram
Drought Index (Keetch and Byram, 1968) and the Mount Soil Dryness Index are
particularly useful in monitoring the total rainfall during the normal or drought
season. In some regions, there were indices which indicate the changes in the global
circulation patterns which may provide warning as early as 6 to 9 months in advance
of extremely dry conditions. One of these is the Southern Oscillation Index (Torrence
and Webster, 2000) which records the difference in atmospheric pressure between
two regions in Australia which can be related to the El Niño events.
2.7
Evaluation of Fire Risk
Forest fire risk can be evaluated from several perspectives. Fire risk indices
are classified into long-term indices, short-term indices and dynamic indices (SanMiguel-Ayanz, 1999). The indices which are derived from factors that do not change
in a short elapse of time are referred to as long-term or structural indices. These
stable factors include elevation and fuel type. On the contrary, dynamic indices are
derived from factors that vary in short periods of time, such as the vegetation status
or the meteorological conditions. The third type of indices combine structural and
dynamic variables. Figure 2.3 shows a general scheme of the different approaches to
compute three types of indices that have been discussed above.
22
2.7.1
Structural Fire Indices
Structural fire indices are based on parameters that do not change in a short
period of time. These include variables that are fairly static such as the topography
and other variables whose rate of change is so slow that they can be considered stable
for a given period that is usually not less than a year (San-Miguel-Ayanz, 1999;
Spano, 2003). In order to provide stable indices over time, the values used for some
variables are the average values over a given period of time. This is the case of the
statistical approach of this type of index. Structural forest fire indices are indicators
of stable conditions that favor fire occurrence. In practice they are used to determine
areas with high risk of fire due to their intrinsic conditions.
Evapotranspiration
Relative Humidity
FIRE RISK
ASSESSMENT
Wind
VEGETATION
STRESS
DYNAMIC
DRASTIC
METEOROLOGICAL
CONDITIONS
Temperature
Fuel Types
ADVANCED
OR
INTEGRATED
Fire History
Population
STRUCTURAL
Topography
Soils
Proximity to Roads
(Source: San-Miguel-Ayanz, 1999)
Figure 2.3: Proposed approaches for the evaluation of forest fire risk indices.
23
The derivation of long term fire indices consists in choosing the variables that
are suitable to the environment and producing a weighted model in a GIS
environment (Chuvieco and Congalton, 1989). These variables will then be classified
into groups, and each variable will be assigned a specific weight according to its
potential contribution to the risk.
2.7.2
Dynamic Risk Indices
Dynamic indices are focused on determining the probability of forest fire
ignition and on the capability of fire spread. They are usually based on determining
vegetation status. These indices were accomplished directly through the use of
meteorological variables, or through the analysis of vegetation indices computed
from remotely sensed data.
Dynamic fire risk indices are developed to determine the probability of forest
fire ignition and also the capability of fire spread. However, the prediction capability
of a forest risk index depends closely on the quality of the data used, and the data
range involved in the developed model.
Since weather is the most significant component for forest fire ignition and
propagation, many forest fire risk indices have been developed in respect to
meteorological information. Therefore, weather information is the most commonly
used component in dynamic fire risk indices. Another popular type of dynamic risk
index is based on the fuel condition. Vegetation structure and moisture condition has
a strong influence on the ignition and the propagation of forest fires. Many studies
evaluated vegetation stress by quantifying the amount of water in the plants, which is
closely related to vegetation stress (Chuvieco et al., 2000; Kogan et al., 2003; and
Burgan et al., 1998). Remote sensing techniques can be used to determine vegetation
24
stress. The Normalized Difference Vegetation Index (NDVI) is probably the most
often used vegetation index for this purpose.
2.7.3
Advanced Forest Fire Indices
Advanced forest fire risk indices agglomerate several factors that are
independently taken into account by the long term and dynamic indices. Recent
advanced forest fire indices have been built as an early warning system to predict the
potential of forest fire, detect forest fire, predict fire propagation and provide burn
scar analysis of forest fire (Cantelaube et al., 2002). Scientists combined structural
indices from a long period of time data and dynamic indices to build an advanced
forest fire index.
2.8
Phases of forest fire development
It is more logical to consider fire in four phases namely, pre-ignition;
flaming; smoldering and glowing as shown in Figure 2.4.
FLAMING
CO
CO2
HOT AIR GASES
RADIATION
ORGANIC GASES
H2O EVAPORATING
25
……….
SMOLDERING
GLOWING
ASH
(Modified from Johnson, 1998)
Figure 2.4: Phases of forest fire development.
2.8.1
Pre-ignition
The unburned material, under certain conditions of the surrounding
atmosphere, soil, and species, has a certain amount of moisture. Part of the incoming
heat will be invested in evaporating such water previous to the pre-heating and
combustion of the material. The combustion process is the result of a series of events
that begins with the rise in temperature of the woody material, to a temperature that
decomposes these materials into gases, tars and char. This process is known as
pyrolysis and ends up in the ignition of the resulting gases into flames and glow
(Johnson, 1998).
26
2.8.2
Flaming phase
Combustion gases and vapours from the pyrolysis process rise above the fuels
and mix with oxygen. Flaming occurs when they are heated to the ignition point. The
heat from the flaming reaction accelerates the rate of pyrolysis. This will cause the
release of greater quantities of combustible gases, which also oxidize and increase
the amounts of flaming.
2.8.3
Glowing phase
Once a fire reaches the glowing phase, most of the volatile gases have been
driven off. Oxygen contact directly with the surface of the fuel causes the fuel to
oxidize. The fuel then burns with a glowing characteristic. This process continues
until the temperature drops so low that combustion can no longer occur, or until all
the combustible material is gone.
2.8.4
Smouldering phase
Smoldering is a smoky process occurring after the active flaming front has
passed. Combustible gases are still being released by the process of pyrolysis, but the
rate of release and the temperature become more stable. The temperatures are not
high enough to maintain flaming combustion.
27
2.9
Review of Forest Fire Models
The need for a method to mitigate forest fire was recognized at least as far
back as 1940 (Burgan et al., 1998). Some fire models are stand alone while others
are modules within larger land cover dynamic models. Fire models operate at many
scales, use different predictive equations, and produce numbers or maps representing
fire frequency, severity, spread rate, burn pattern or risk. Many fire models were
originally developed for use in conifer-dominated forests of the Western United
States, pine forests in Canada and bush forests in Australia (Casagrandi and Rinaldi,
1998). However, ecosystem differences may take these model structures
inappropriate for other regions. Although none of the major indices is inherently
superior to the rest in all circumstances, some indices are better suited than others in
certain uses. Most of the original indices are measured using conventional
techniques.
From Baines (1990), the majority of current fire models that are in use today
are based on fire spread relationships developed using the drought index (Rothermel,
1972; Keetch and Byram, 1968; Van Wagner, 1969; McArthur, 1966). Within these
models, the most complex systems are developed by Rothermel and Van Wagner
which combine measures of fuel, topography, weather and risk of ignition to provide
indices of fire occurrence or fire behaviour. Keetch and Byram model forest fires
based on information from a cumulative period of time data. While McArthur model
is easier to use where this model measures drought and weather as applied to a
standard fuel type to predict the speed of a fire or its difficulty of suppression.
Rothermel Model’s equations require a description of fuel which includes
depth, loading, percentage of dead fuel, moisture of extinction, heat content, surface
area to volume ratio, mineral content, silica content, and particle density. A complete
set of fuel information is needed as an input to use the Rothermel Fire Spread Model
(Watson et al., 2000). It also requires environmental variables such as wind speed at
28
half-flame height, slope and fuel moisture content. Rothermel’s equations are only
valid for surface fires (Scott and Burgan, 2005).
In 1968, Keetch and Byram developed an index to describe the dryness of
fuels namely the Keetch and Byram Drought Index (KBDI). It is a cumulative or
book keeping index. To start the index, it is necessary to begin accounting after a
very heavy rainfall event. It is assumed the soils and fuel layers were saturated after a
period of abundant rainfall, at 5mm within a week and the index will start at or near
zero. From that time onwards, the Drought Index is calculated for each subsequent
day which becomes a simple bookkeeping. The drying factor is large for very low
KBDI and close to zero when the rainfall is 200mm. This index has proved to be a
useful early warning tool and is now incorporated into the US National Fire Danger
Rating System (Pyne et al., 1996) and the Australian Forest Fire Danger Rating
System (McArthur, 1967).
Countries in South East Asia are using the Canadian Forest Fire Danger
Rating System (Andrews and Williams, 1998) in predicting forest fire occurrence.
Forest fire danger rating research in Canada was initiated in 1968 (Beall, 1990). The
original approach has been evolutionary, building on previous systems and using
field experiments and empirical analysis extensively. The Canadian Forest Fire
Weather Index System consists of two subsystems, namely Canadian Forest Fire
Weather Index (FWI) System and the Canadian Forest Fire Behavior Prediction
(FBP) System. The components are expressed in the diagram in Figure 2.5.
29
INPUTS
Human cause,
lightning
Weather
Topography
Fuels
Fire Weather
Index (FWI)
System
Fire Occurrence
Prediction
(FOP) System
Accessory Fuel
Moisture System
Fire Behavior
Prediction (FBP)
System
Canadian Forest Fire Danger Rating System (CFFDRS)
(Source: Taylor and Alexander, 2006)
Figure 2.5: Structure of Canadian Forest Fire Danger Rating System
(CFFDRS).
Inputs to the FWI System include elevation and current daily weather data
from a variety of sources. The FWI System consists of six components as illustrated
in Figure 2.6. This components are functions to account for the effects of fuel
moisture and wind on fire behaviour.
As illustrated in Figure 2.6, the first three components of FWI namely Fine
Fuel Moisture Code (FFMC), Duff Moisture Code (DMC) and Drought Code (DC)
are numerical ratings of the moisture content of litter and other fine fuels, average
moisture content of organic layers, also average moisture content of deep, compact
or organic layers respectively.
The remaining three components namely Initial Spread Index (ISI), Buildup
Index (BUI) and Fire Weather Index (FWI) represent the rate of fire spread, the fuel
available for combustion and the frontal fire intensity respectively. Their values rise
as the fire danger increases.
30
Fire weather
observation
Fuel moisture
code
Fire behaviour
indices
Temperature, relative
humidity, wind, rain
Temperature, relative
humidity, rain
Wind
FINE FUEL
MOISTURE
CODE (FFMC)
DUFF
MOISTURE
CODE (DMC)
INITIAL
SPREAD
INDEX (ISI)
Temperature,
rain
DROUGHT
CODE
(DC)
BUILDUP
INDEX
(BUI)
FIRE
WEATHER
INDEX(FWI)
(Source: Andrews and Williams, 1998)
Figure 2.6: The structure of Fire Weather Index (FWI).
As illustrated in Figure 2.7, the Fire Behaviour Prediction System (FBP)
provides quantitative estimates of potential head fire spread rate, fue l consumption,
and fire intensity, as well as fire descriptions. With the aid of an elliptical fire growth
model, it gives estimates of fire area, perimeter, perimeter growth rate, also flank and
back fire behaviour. Back fire is a fire spreading, or set to spread, into or against the
wind.
The McArthur Forest Fire Danger Meter (FFDM) first appeared in
operational use in 1967. It is an index that relates to the chances of a fire starting, its
rate of spread, its intensity, and its difficulty of suppression, according to various
combinations of air temperature, relative humidity, wind speed and both the long and
short-term drought effects. This model is a key tool for assessing broad scale fire
danger throughout eastern Australia to fight against bush fires (Griffith, 1999).
31
FBP System
fuel type
FFMC, ISI,
BUI, wind speed
and direction
Percent slope,
upslope
direction
Elevation,
coordinate,
date
Elapsed time,
point/line
ignition
Fuels
Weather
Topography
Foliar Moisture
Content
Type &
Duration of
Prediction
Fire Behaviour Prediction
(FBP)
Primary Output
•
•
•
•
Rate of Spread
Total Fuel Consumption
Head Fire Intensity
Fire Description
Secondary Output
• Head, flank and back fire
spread distance
• Flank and Back Fire Rates of
Spread
• Flank and Back Fire Intensities
• Elliptical Fire Area and
Perimeter
• Rate of Perimeter Growth
• Length to Breadth Ratio
Figure 2.7: The structure of Fire Behaviour Index (FBP), including the
necessary inputs and the outputs produced by the system.
2.10
Review of Existing Computer Code and Running Application for Forest
Fire
Recently developed forest fire models are based on both ground and remote
satellite observation. Scientists also simulate fire behaviour with the support of
robust ancillary data through computer systems. Many of the recently developed
models are theoretical approaches which have an associated formulation combined in
a computer code. Currently there are a number of theoretical models which are
suitable to be converted in simulation applications or to form part of an integrated
information system. Table 2.1 gives some examples of the most popular forest fire
32
models that have been developed and widely used along with their characteristics.
Explanation of each models in Table 2.1 are briefly given below.
2.10.1 BEHAVE
A widely used semi-empirical fire spread model is based on the Rothermel
Model (Rothermel, 1972). It has been adapted in many applications to solve one or
more fire management problem (Burgan, et al, 1994; Andrews, 1996).
BEHAVE consists of algorithm which describing fuels, fuel moisture, wind,
topography, fire size and shape, flame length, spotting fire, ignition probability and
tree scorch in user-friendly programs. Simplified methods of fuel appraisal are
described that do not require extensive inventory in which the methods are applicable
to surface fires.
BEHAVE is run by user-supplied input. Requested values depend on the
modeling choices made by the user such as fuel model, fuel moisture, wind speed
and direction, terrain slope are used to calculate rate of spread, flame length and
intensity. The original BEHAVE is written in FORTRAN. The recent developed is
being written so that it will run under various operating systems including Windows,
NT and Unix.
33
2.10.2 FireLib
FileLib is a C language function library for predicting the spread rate,
intensity, flame length and scorch height of free-burning surface fires. This
application is widely used in United State. It is derived directly from the BEHAVE
fire behaviour algorithms for predicting fire spread in 2 dimensions (Bevins, 1996).
FireLib was developed to give fire growth modelers a simple, common, and
optimized application programming interface (API) to use in their simulations. This
application can be run on a variety of personal computers and workstations.
2.10.3 FARSITE
FARSITE is a fire growth simulation model. It uses spatial information on
topography and fuels along with weather and wind files (Finney, 1998; Cloeman and
Sullivan, 1996). FARSITE incorporates the existing models for surface fire, crown
fire and fire acceleration into a 2-dimensional fire growth model. FARSITE runs
under Microsoft Windows operating systems and features a graphical interface. The
user can also link the result to a geographical information system (GIS) for further
analysis.
FARSITE has been distributed mostly in the continental Unites States of
America, and also to users in Europe and South America. Other users are associated
with universities, state and private land management agencies and private consulting
companies.
34
2.10.4 FORFAIT
FORFAIT sets out to develop and demonstrate a decision support system to
assist planners, regulators and industry in optimizing the management of forest fire
risks. It will aid in implementing measures that eliminate or mitigate harm to
humans, the environment and business, using a cost benefit approach where
appropriate and depending on local regulatory requirements.
FORFAIT uniquely combines generic and time varying site-specific
information through electronic links from field and satellite data sources, state-ofthe-art predictive models and expert knowledge. Recognizing that in general there
are many different decisions that could be made in any given situation, the system
uses fuzzy logic to suggest the most appropriate course of action and a probabilistic
framework to take account of uncertainty in the parameters. The fire model used for
FORFAIT is the Fire Behaviour Model based on the algorithm proposed by R.
Rothermel.
2.10.5 FOMFIS
The FOMFIS Fire Behaviour Model has been designed for the computation
of the fire spread under non-uniform conditions using geographical information in
the form of raster maps, namely digital terrain model, forest fuel and wind field
(Kallidromitou et al., 2000).
The FOMFIS system is simulating forest fire behaviour in its expansion for
every generated fire starting point till it is naturally extinguished or suppressed by
fire fighting operations.
35
BEHAVE
FireLib
FARSITE
FORFAIT
FOMFIS
Table 2.1: Existing computer code and running application for forest fire.
United
United
United
Europe
Europe
States
States
States
Algorithm
Rothermel
Rothermel
Ground Fire
No
No
No
No
No
Surface Fire
Yes
Yes
Yes
Yes
Yes
Crown Fire
No
No
Yes
Yes
No
Spot Fire
Yes
No
Yes
No
No
Fire
Acceleration
& Interaction
Wind
No
No
Yes
No
No
Yes
Yes
Yes
Yes
Yes
Weather
Yes
Yes
Yes
Yes
Yes
Fuel Moisture
Yes
Yes
Yes
Yes
Yes
Predict
Human
Casualties
Support GIS
integration
No
No
No
Yes
No
No
No
Yes
Yes
Yes
Country
Rothermel
(Modified from: Sánchez, 2001)
2.11
Review of Fire Detection Techniques Using Remote Sensing
Satellite remote sensing has played a growing role in fire detection and
monitoring over the past two decades. Infrared sensors on board satellites have been
widely used to collect statistics on fires. Several studies using this recent technology
are explained below.
36
2.11.1 GOES Fire Detection
Recent work has shown the potential of meteorological geostationary
satellites to monitor biomass burning activities. The Visible Infrared Spin Scan
Radiometer and Atmospheric Sounder (VAS) available on the GOES satellites can
be used to locate active fires by utilizing the visible, shortwave infrared (SWIR) in
the 4μm channel and longwave infrared (LWIR) in the 11μm window data. GOES
VAS instrument have provided multispectral monitoring of the earth in the visible
and 12 infrared spectral bands between 4μm and 15μm as shown in Table 2.2.
Table 2.2: GOES-7 VAS satellite spectral channels.
Channel
1
2
3
4
5
6
7
8
9
10
11
12
Wavelength (μm)
14.6-14.8
14.3-14.7
14.1-14.4
13.8-14.2
13.2-13.5
4.5-4.6
12.5-12.8
10.4-12.1
7.2-7.4
6.4-7.1
4.4-4.5
3.8-4.1
Spatial Resolution (km)
13.8
13.8
6.9 or 13.8
6.9 or 13.8
6.9 or 13.8
13.8
6.9 or 13.8
6.9 or 13.8
6.9 or 13.8
6.9 or 13.8
13.8
13.8
(Source: Prins and Menzel, 1993)
The technique of Matson and Dozier (1981) was adapted to GOES VAS data
to monitor biomass burning in selected regions of South America by Prins (1993).
The fires were identified in GOES VAS imagery by manually locating hot spots in
the 4μm channel and verifying the presence of fire by the presence of smoke plumes
in the visible imagery. Furthermore, to be designated as a fire pixel, the pixel must
exhibit elevated brightness temperature values in both the 4μm and 11μm channels.
Although the spatial resolution of GOES VAS data which is 7km to 14km is a
limiting factor in detecting fires, comparison with NOAA AVHRR indicated that the
37
decreased spatial resolution does not severely impede the ability of the VAS
instrument to detect subpixel fires.
2.11.2 Defense Mapping Satellite Program (DMSP) Fire Detection
The orbital characteristics of the DMSP satellites are quite similar to those of
the NOAA satellites. It is in a polar sunsynchronous orbit, with an equatorial
crossing time of 10:50 and 22:50 local solar time. The unique characteristics of the
DMSP platform is the low light visible sensor which is operated routinely at night
(Sullivan, 1989). Even though the instrument was designed for meteorological
purposes, the visible-light instrument is sensitive to other terrestrial light sources,
such as city lights, gas fires, auroras, lightning and fires (Sullivan, 1989). This
imagery has been demonstrated to be a useful tool for monitoring fire activity at
night, particularly for remote regions where there are few city lights. Historically, the
DMSP archive was primarily in photographic form. The instrument has an adjustable
gain and the archived data does not include instrument calibration information
thereby making any quantitative comparison between light sources impossible.
2.11.3 Landsat Fire Detection
The high spatial resolution sensor of Landsat Thematic Mapper (TM)
includes a middle infrared channel (2.08-2.35μm) with a 30m spatial resolution,
which permits active fires to be detected. A 700K fire that occupies 20% of the 30m
pixel will saturate the middle infrared TM channel. A study using the TM 1.6μm
channel has also provided some interesting preliminary results, estimating fire
38
intensity through interpretation of the spectral response from the resulting ash layer
(Riggan et al., 1993).
2.11.4 NOAA AVHRR Fire Detection
The current sequence of National Oceanic and Atmospheric Administration
(NOAA) satellites has been in continuous operation since October 1978. The
AVHRR is a scanning radiometer measuring reflected and emitted radiation in four
channels on board the satellite NOAA 6, 8, 10 and 12, and in five channels on board
NOAA 7, 9, 11, 13 and 14. The term ‘very high resolution’ refers to a high
radiometric resolution. AVHRR data are recorded to 10 bit precision. For the thermal
infrared channel, on-board calibration exists through the regular measurement of
deep space and a blackbody of known temperature on board.
Lee and Tag (1990) presented an approach to non-interactive fire detection.
Essentially they subjectively chose a threshold fire temperature and used the Dozier
model to develop a look-up table specifying which combinations of satellite
measurements constituted a positive fire detection. Atmospheric corrections were
included in the estimation of background temperatures using the method of McClain
et al. (1985). They applied their technique to nighttime imagery over the San
Francisco area and the Persian Gulf.
Kaufman et al. (1990) applied fire detection with NOAA 9 data to the
estimation of emissions from biomass burning. Fires were detected in pixels that met
three detection criteria. The first criterion was the channel 3 brightness temperature
to be elevated above a set threshold indicating a fire to be present. The second
criterion specified that the channel 3-channel 4 temperature difference must be at
39
least 10K and the third criterion used the channel 4 temperature to eliminate false
detections from cool clouds that are highly reflective in the 3.8μm band.
Justice et al. (1996) and Scholes et al. (1996) combined AVHRR fire
information in a dynamic model to generate improved trace gas and particulate
emission estimates for Southern Africa. The approach combined satellite data on fire
distribution and timing with fuel load calculated by a simplified ecosystem
production model and ground based measurements of emission ratios. Daily fires
detected by the AVHRR for the entire burning season were calibrated to provide
burned area estimates using Landsat MSS data.
2.12
Summary
This chapter discussed briefly the definition of forest fires to the latest
research utilizing remote sensing techniques. By definition, forest fires are any wild
land fire not prescribed for the area by an authorized plan. There are three factors for
fires to ignite which is called fire fundamental triangle; namely fuel, heat and
oxygen. There are also three controlling factors in fire propagation namely fire
environment triangle; namely fuel, weather and topography.
Forest fires are classified as ground, surface and crown fires. Ground fires are
the most intense fires, followed by crown fires and the less intense is surface fires.
Common occurrences of forest fires in South East Asia region are ground fires,
which is the most difficult type of fires to extinguish.
Fire risk indices are categorized into three different classes namely structural
fire indices, dynamic fire indices and advanced fire indices. Structural fire indices
40
depend on parameters that do not change in a short period of time, while dynamic
fire indices depend on dynamic parameters such as weather and vegetation indices.
Advanced fire indices are developed from the combination of structural and dynamic
indices.
Several fire models are also discussed in this chapter. Most of the models
used today are based on fire spread relationship developed using drought index by
Rothermal (1972), Van Wagner (1969) and McArthur (1966) which is widely used in
the United States, Canada and Australia, respectively. Drought index by Keetch and
Byram (1968) is also used widely in developing forest fire model. Computer code
and running application were designed using the developed models such as
BEHAVE, FireLib, FARSITE, FORFAIT and FOMFIS.
Current fire detection procedures using remote sensing techniques are also
discussed. A number of satellites have been used to detect fires since 25 years,
namely GOES, DMSP, Landsat and NOAA AVHRR.
CHAPTER III
METHODOLOGY
3.1
Introduction
In the present work a forest fire system is developed from several satellite
data processing. The methodology in this study is categorized into four different
stages namely, data acquisition, data pre-processing, data processing and data
analysis. A variety of data is used in this study, namely satellite data and ancillary
data. Geometric and radiometric correction was applied to the satellite data in order
to carry out the pre-processing task. The processing stage involved the derivation of
hotspots, Normalized Difference Vegetation Index (NDVI), brightness temperature
(BT) Vegetation Condition Index (VCI) and Temperature Condition Index (TCI).
Vegetation Health Index (VH) was then carried out followed by fire risk mapping. In
order to carry out the analysis, temperatures extracted from satellite and weather data
were compared to evaluate the accuracy of the satellite-derived temperatures.
The fire risk map is classified into three categories namely stressed, fair and
favorable based on present vegetation and temperature conditions. Further details of
the approach utilized are discussed in this chapter. The flow of operational
methodology is shown in Figure 3.1.
42
Geometric
Correction
Data Preprocessing
Topographic
Map
Vector
Layers
Data
Acquisiiton
MODIS
data
Calibration
Brightness
Temperature
(BT)
Vegetation
Condition
Index (VCI)
Temperature
Condition
Index (TCI)
Vegetation
Health Index
(VH)
Data Processing
Normalized
Difference
Vegetation
Index (NDVI)
Hot
Spots
Proximity
Analysis
Fire Risk
Map
Figure 3.1: Flow chart of operational methodology.
Results
Integrated Forest Fire
System
Accuracy
Assessment
Accuracy
Assessment
43
3.2
Data Acquisition
Satellite data together with a variety of ancillary data were used in this study.
Characteristics of each dataset and approach of processing them are described below.
3.2.1
Satellite Data
3.2.1.1 MODIS
MODIS stands for Moderate Resolution Imaging Spectroradiometer on board
the Terra satellite. This satellite was launched on December 19, 1999 in a near sunsynchronous polar orbit descending southward with 10.30 am local equator crossing
time. It is a 36-band spectroradiometer that covers a broad spectral range for a
variety of applications as shown in Table 3.1 (Barnes et. al., 1998).
Figure 3.2: MODIS data on 28 January 2005 with composite of bands 1, 2 and 4.
44
Table 3.1: MODIS spectral bands characteristics.
Band
(1km)
1**
2**
3*
4*
5*
6*
7*
8
9
10
11
12
13
14
15
16
17
18
19
Bandwidth
(nm)
620-670
841-876
459-479
545-565
1230-1250
1628-1652
2105-2155
405-420
438-448
483-493
526-536
546-556
662-672
673-683
743-753
862-877
890-920
931-941
915-965
Spectral
Radiance1
21.8
24.7
35.3
29.0
5.4
7.3
1.0
44.9
41.9
32.1
27.9
21.0
9.5
8.7
10.2
6.2
10.0
3.6
15.0
Required
SNR2
128
201
243
228
74
275
110
880
838
802
754
750
910
1087
586
516
167
57
250
Primary
Use
Land/Cloud/Aerosols
Boundaries
Band
Bandwidth
(μm)
3.660-3.840
3.929-3.989
3.929-3.989
4.020-4.080
4.433-4.498
4.482-4.549
1.360-1.390
6.535-6.895
7.175-7.475
8.400-8.700
9.580-9.880
10.780-11.280
11.770-12.270
13.185-13.485
13.485-13.785
13.785-14.085
14.085-14.385
Spectral
Radiance1
0.45 (300K)
2.38 (335K)
0.67 (300K)
0.79 (300K)
0.17 (250K)
0.59 (275K)
6.00
1.16 (240K)
2.18 (250K)
9.58 (300K)
3.69 (250K)
9.55 (300K)
8.94 (300K)
4.52 (260K)
3.76 (250K)
3.11 (240K)
2.08 (220K)
Required
NEΔT3 (K)
0.05
2.00
0.07
0.07
0.25
0.25
150 (SNR)
0.25
0.25
0.05
0.25
0.05
0.05
0.25
0.25
0.25
0.35
Primary
Use
Surface/
Cloud
Temperature
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
* 500m spatial resolution
** 250m spatial resolution
The rest is 1km resolution
Land/Cloud/
Aerosols Properties
Ocean Color/
Phytoplankton/
Biogeochemistry
Atmospheric
Water Vapor
Atmospheric
Temperature
Cirrus Clouds
Water Vapor
Ozone
Surface/Cloud
Temperature
Cloud
Top Altitude
1
Spectral Radiance values are in W/m^2-um-sr
SNR = Signal-to-noise-ratio
3
NEΔT = Noise-equivalent temperature difference
2
(Modified from: Barnes et al., 1998)
45
There are three variety of resolution within MODIS datasets which is 250 m
resolution consists of band 1 and 2, 500 m resolution consists of band 3, 4, 5, 6, and
7 and 1 km resolution for the remaining 29 bands. Within the 36 bands on MODIS,
there are three major band segments (Barnes et al., 1998). There are seven bands,
namely bands numbered 1-7 in Table 3.1, that will observe land cover features plus
cloud and aerosol properties (King et. al., 1992). The spectral placement of these
bands are derived so as to be very similar to the bands on the Landsat TM though the
spatial resolutions are 250 m and 500 m. There are nine ocean color bands, namely
bands numbered 8-16 in Table 3.1, that were chosen as a result of studies of the
Nimbus Coastal Zone Color Scanner (CZCS) and the SeaWiFS instrument (Gordon,
1990). Another segments of bands, namely bands 20-25 and 27-36 in Table 3.1, were
drawn from the bands of the High Resolution Infrared Sounder (HRIS), with
emphasis on those HRIS bands that sense properties of the troposphere and the
surface (Giglio et al., 2003).
Over 40 data products are produced from raw MODIS data (Suraiya Ahmad
et al., 2002) to enable advanced studies of land, ocean and atmospheric processes for
distribution to the public. MODIS Level 1B 1KM Radiances Product namely
MOD021KM, together with the MODIS Geolocation Product namely MOD03, were
used in this study. The data was acquired from Goddard Earth Sciences Distributed
Active
Archive
Centre
(GES
DAAC)
website
http://daac.gsfc.nasa.gov/data/dataset/MODIS/01_Level_1/index.html.
at
The
cumulative dataset of MODIS from January 2000 to April 2005 were used in this
study. There are 1211 datasets processed in this study and the fraction of processed
datasets are shown in Table 3.2 below. From the table, there are 187 datasets from the
year 2000, 263 datasets from 2001, 231 datasets from 2002, 221 from 2003, 234
from 2004 and 75 datasets from 2005. Datasets of 2005 is only for January till April.
46
Table 3.2: Number of datasets processed in this study for each year from January
2000 to April 2005.
3.2.2
Year
Number of processed datasets
2000
187
2001
263
2002
231
2003
221
2004
234
2005
75
Total
1211
Ancillary Data
Several sets of ancillary data were used in this study, namely topographic
map, vector data, weather data, fire occurrence record and location of Fire and
Rescue Department Station distribution over Peninsular Malaysia. Sources and
function of each data are explained in detail below.
3.2.2.1 Topographic Map
A topographic map is used to execute geometric correction of satellite
imagery. The details of the map used are shown in Table 3.3.
47
Table 3.3: Topographic map used to execute geometric correction
Publisher
Type
Topographic map
DNMM 2001
Scale
1:750,000
Date of
Publication
2001
Edition 1-PNMM Semenanjung
Publisher: Department of Survey
and Mapping, Malaysia.
3.2.2.2 Vector Data
Vector data acquired in this study were soil type and road network. These data
were digitized from several maps shown in Table 3.4.
Table 3.4: List of maps used to extract vector layers.
Title
Publisher
Land Use Map of Soil Management Division,
Semenanjung Malaysia Dept of Agriculture,
Peninsular Malaysia
Department of Survey and
• Road Network Map
Mapping, Malaysia
Siri L4010
Lembar 1
Edisi 7 - PPNM
Scale
1:75,000
1:500,000
Vector
layer
Land use
Road
network
48
3.2.2.3 Forest Fire Occurrences Record
Forest fire occurrences record were acquired from Fire and Rescue
Department, Malaysia. Records of fire occurrence in Peninsular Malaysia from
January 2005 to April 2005 were used in this study shown in APPENDIX A.
3.2.2.4 Fire and Rescue Department Distribution
Vector data of Fire and Rescue Department distribution over Peninsular
Malaysia was also used in this study to carry out the proximity analysis. This data
was acquired from Fire and Rescue Department of Malaysia.
3.3
Data Pre-processing
3.3.1
Geometric Correction
The MODIS data were geometrically corrected by using the latitude and
longitude of ground control points (GCPs) using the WGS 84 datum. Twelve (12)
GCPs were used to geometrically corrected the MODIS image of 27 January 2005 as
shown in Figure 3.3. The RMS (Root Mean Square) error of the geometrically
corrected image is as shown in Table 3.5. From the table, X Input and Y Input is the
input coordinate from original MODIS data, while X Ref and Y Ref is the input
coordinate acquired from vector data of the study area. The average RMS error for
this image is 0.07 pixel. The geometrically corrected image is shown in Figure 3.4.
49
Table 3.5: List of ground control point for MODIS data on 27 January 2005.
Point
X Input
Y Input
X Ref
Y Ref
RMS Error
GCP1
99.65
6.28
99.71
6.23
0.01
GCP2
100.15
5.47
100.18
5.46
0.14
GCP3
100.92
4.04
100.85
3.99
0.00
GCP4
101.54
2.78
101.42
2.78
0.22
GCP5
102.86
1.84
102.69
1.85
0.12
GCP6
103.59
1.27
103.49
1.24
0.14
GCP7
104.31
1.44
104.29
1.39
0.12
GCP8
103.87
2.63
100.76
2.61
0.09
GCP9
103.61
3.50
103.45
3.51
0.07
GCP10
101.26
6.87
101.28
6.85
0.01
GCP11
103.64
4.30
103.49
4.24
0.03
GCP12
104.28
2.71
104.22
2.66
0.07
Average Control Point Error
0.07
Figure 3.3: Twelve GCPs were used to geometrically correct the MODIS image of
27 January 2005.
50
Figure 3.4: Geometrically corrected image of MODIS on 27 January 2005.
3.4
Data Processing
3.4.1
Normalized Difference Vegetation Index (NDVI)
Forest fire normally occurs in a dry season. In a normal season, green and
vigorous vegetation reflects little radiation in the visible (VIS) part of the solar
spectrum, due to high chlorophyll absorption, and more radiation in the near-infrared
(NIR), due to scattering of the light by leaf internal tissues and water content (Kogan,
2002). Following these properties, the difference between NIR and VIS becomes
greater. The dry season consequently depresses vegetation greenness and vigour, due
to reduction of chlorophyll and water content. This situation leads to an increase in
51
the VIS, decrease in the NIR and reduction in NIR-VIS. This principle is applied in
the construction of normalized difference vegetation index or NDVI;
NDVI =
( NIR – VIS)
(NIR + VIS)
… (3.01)
where;
NDVI = Normalized Difference Vegetation Index
NIR = Near-infrared band
VIS = Visible band
In this study, NDVI are extracted from MODIS data through;
NDVI =
( Channel 2 – Channel 1)
( Channel 2 + Channel 1)
… (3.02)
where;
Channel 1 = Channel 1 of MODIS data
Channel 2 = Channel 2 of MODIS data
NDVI have been used for monitoring land surface and the environment since
the mid-1980s (Kogan, 2002). One of the important advantages of this method is the
combination of NDVI, normally used alone, and the thermal channel. Assessment of
temperature conditions permits one to identify subtle changes in vegetation health
because the effect of drought becomes obvious if a shortage of rainfall is
accompanied by excessive temperatures.
3.4.2
Vegetation Condition Index (VCI)
The Vegetation Condition Index (VCI) was derived from the following
algorithm:
VCI =
(NDVI – NDVI min)
*100
(NDVImax - NDVImin )
… (3.03)
52
where;
NDVI = smoothed weekly NDVI
NDVImin = multiyear absolute minimum of NDVI
NDVImax = multiyear absolute maximum of NDVI
Extreme conditions were derived by measuring the maximum and minimum
NDVI values from weekly cumulative readings. Cumulative reading is used to
eliminate high-frequency temporal variations from the NDVI time series (van Dijk et
al., 1986; Kogan, 1990). Composite of 7-day satellite data is used to minimize cloud
effects.
3.4.3
Brightness Temperature (BT)
Brightness temperature is a measure of the intensity of radiation thermally
emitted by an object, given in units of temperature because there is a correlation
between intensity of the radiation emitted and physical temperature of the radiating
body which is given by the Planck’s law.
Brightness temperature of MODIS is calculated from Planck’s Law, expressed as;
BT =
2hc2 λ-5
ehc/kλT – 1
where;
BT = Brightness temperature (Js-1m-2sr-1Hz-1)
h = Planck’s constant (Joule per hertz)
c = Speed of light in vacuum (m/s)
k= Boltzmann gas constant (Joule/Kelvin)
λ = Band or detector center wavelength (m)
T = Temperature (Kelvin)
…(3.04)
53
Inverting algorithm (3.04) to solve for temperature gives;
T=
3.4.4
hc
kλ
1
ln ( 2hc 2λ-5 BT-1 + 1)
…(3.05)
Temperature Condition Index (TCI)
The Temperature Condition Index can be measured from a simple algorithm
below:
VCI =
(BTmax – BT)
*100
(BTmax – BTmin)
… (3.06)
where;
BT = smoothed weekly brightness temperature
BTmin = absolute minimum of smoothed weekly brightness temperature
BTmax = absolute maximum of smoothed weekly brightness temperature
This algorithm was applied to the thermal channel of MODIS data.
3.4.5
Vegetation Health Index (VH)
Fluctuations were estimated relative to the maximum and minimum intervals
of both NDVI and BT variations and named the Vegetation Health Index (VH). VCI
and TCI were combined in VH index in order to express their additive approximation
of vegetation stress:
(VCI + TCI)
VH =
2
where;
*100
… (3.07)
54
VH = Temperature health index.
3.4.6
Fire Risk Map
Fire potential estimation is based on the intensity and duration of combined
moisture and thermal stress. If severe stress, for which the index value is less than
15, continues for one week, it indicates that the fire potential is minimal, while the
fire potential reached maximum if this condition continues for four weeks and longer.
Fire potential is higher if vegetation stress is severe and persistent. High level of fire
potential due to both moisture and temperature conditions can be used as a fire
danger warning.
3.4.7
Hot Spots of MODIS
Fire detection on MODIS is performed using the 4 and 11 micrometer
channel of brightness temperatures (T4 and T11). The MODIS instrument has two
4μm channels, which is bands 21 and 22, both of which are used by the detection
algorithm. Channel 21 saturates at about 500 K, and channel 22 saturates at about
335 K. Since channel 22 is less noisy and has a smaller quantization error, T4 is
derived from channel 22 when possible. When channel 22 saturates, or has missing
data, channel 21 is used to derive T4 instead. T11 is always computed from channel 31
which saturates at approximately 400 K (Kaufman, 2000).
55
The fire detection strategy is based on absolute detection of the fire, if the fire
is strong enough, and on detection relative to the background to account for
variability of the surface temperature and reflection by sunlight (Kaufman, 1998).
To avoid false detection, all pixels that fulfill the following conditions in algorithm
3.08 and 3.09 are immediately excluded as fires.
T4 < 315 K (305 K at night)
… (3.08)
ΔT41 – T11 < 5 K (3 K at night)
… (3.09)
where;
T4 = 4µm channel of brightness temperatures
T11 = 11µm channel of brightness temperatures
K = Kelvin
From the remaining pixel, pixel is classified as fire if at least one of the
following five logical conditions is met:
T4 > T4b + 4δT4b
… (3.10)
T4 > 320K (315K at night)
… (3.11)
ΔT41 > ΔT41b + 4δΔT41b
… (3.12)
T41 > 20K (10K at night)
… (3.13)
T4 > 360K (330K)
… (3.14)
where;
T4b = T4 background pixel
δT4b = Standard deviation of T4 background pixel
δΔT41b = Standard deviation difference between 4µm and 11µm of background pixel
The hot spot is generated using a small DOS program provided by ScanEx
Research and Development Centre, a Russian Remote Sensing ground station in
Moscow at their website (http://eostation.scanex.ru/software.html). The program is
run from the command prompt. It accepts a set of options and input and output file
names in definitive order. To run fire algorithm one must provide MOD021KM
56
(Level 1B) and corresponding MOD03 (Geolocation) files as input. The output of
this algorithm is known as MODIS Thermal Anomalies and Fire Daily Data Product
(MOD14). MOD14 is the most basic fire product in which active fires and other
thermal anomalies, such as volcanoes are identified. This product generates all of the
higher-level fire products, and contains several components, such as line, sample,
latitude, longitude, confidence level, adjacent cloud, adjacent water, reflectance of
Channel 2 and brightness temperature of Channel 21 or 22. Detail explanation for
each component generated in MOD14 data shown in Table 3.6.
Table 3.6: Attributes generated from the MOD14 data.
Attribute
Description
Line
Granule line of fire pixel
Sample
Granule sample of fire pixel
Latitude
Latitude of fire pixel (Degrees)
Longitude
Longitude of fire pixel (Degrees)
Confidence
Detection confidence level (%), in range of 0-100
AdjCloud
Number of adjacent cloud pixels in 3x3 window
AdjWater
Number of adjacent water pixels in 3x3 window
R2
Channel 2 reflectance of fire pixel
T21
Channel 21 or 22 brightness temperature of fire pixel (Kelvin)
T31
Channel 31 brightness temperature of fire pixel (Kelvin)
MeanT21
Channel 21 brightness temperature of background (Kelvin)
MeanT31
Channel 31 brightness temperature of background (Kelvin)
MedianDT
Median background brightness temperature difference (Kelvin)
StdDevT21
Channel 21 or 22 brightness temperature standard deviation of
background
StdDevT31
Channel 31 brightness temperature standard deviation of
background
StdDevDT
Standard deviation of background brightness temperature
difference (Kelvin)
Power
Total emitted power (Megawatts)
57
A measure of confidence for each detected fire pixel produced is based on the
approach of Giglio et al. (2003). The confidence assigned to each fire pixel is
composed of a combination of five sub-confidences, labeled C1 to C5, each having a
range of 0 (lowest confidence) to 1 (highest confidence). The levels of confidence
defined as;
C1 = S(T4)
…(3.15)
C2 = S(zΔT;)
…(3.16)
C3 = S(z4)
…(3.17)
C4 = 1 - S(Nac)
…(3.18)
C5 = 1 - S(Naw)
…(3.19)
The standardized variables of z4 and zΔT, are defined as follows;
z4 =
ΔT - Δ Ť
δΔT
…(3.20)
zΔT =
T4 – Ť4
δ4
…(3.21)
where;
T = brightness temperature
Ť = Respective mean of the channel for valid neighboring pixel
Naw = Number of water pixel adjacent to the fire pixel
Nac = Number of cloud pixel adjacent to the fire pixel
δ = Mean absolute deviation of the respective channel for valid neighboring pixel
For C1, 310K represents the minimum temperature required for a pixel to be
considered a fire pixel, while 340K represents a typical value for a reasonably
obvious fire. For C2, z4 = 2.5 is the minimum value required of fire pixels by
58
detection algorithm, whereas z4 = 6 represents a typical value for an unambiguous
fire. A similar rationale applies to the definition of C3. C4 reduces the detection
confidence as the number of adjacent cloud pixel increases, accounting for the fact
that fire pixels detected along cloud edges are more likely to suffer from cloud
contamination, potentially triggering a false alarm via reflected sunlight.
Finally, C5 reduces the confidence as the number of adjacent water pixels
increases, reflecting the greater likelihood that the detected fire pixel is instead a
coastal false alarm. Following Giglio et al. (2003), the detection confidence C is then
defined as the geometric mean of the subconfidences as follows
C = 5√ C1 C2 C3 C4 C5
3.4.8
…(3.22)
Proximity Analysis
To provide more useful information about the locations of the fires, a
proximity analysis is done by measuring the distance of the fire pixel from the
nearest town or city, road and river. Distances from the closest town will assist in
quick dispatch to control or extinguish the fires detected in real time. Besides that,
distance from the closest road and fire fighter station is calculated to show the
accessibility to the fires and determine the remoteness of its location. Distance to the
closest rivers provides valuable information as they provide alternative water supply
to extinguish fires particularly in remote areas. Besides that, they also serve as break
or pathways for the suppression of fires (Chuvieco and Martin, 1989). The proximity
of these features is determined with an ArcView extension called Alaska Pak. The
extension provides a function called closest feature that determines the closest
features from a selected theme to the fires and the distance and then adds them as
additional attributes.
59
3.4.9
Development of Forest Fire Interface
Forest fire interface is a computer program which consists of hot spot, fire
risk map and proximity analysis application that has been discussed earlier in this
study. The interface is developed using Microsoft Visual C++. The forest fire
interface is discussed briefly in Chapter 4.
3.5 Accuracy Assessments
Accuracy assessments in this study were divided into several categories as
follows:
1. Comparison was made of hot spot extracted from MODIS with recorded forest
fire occurrence data from Fire and Rescue Department of Malaysia (FRDM).
From the comparison, all the hot spot detected by MODIS were also recorded by
FRDM.
2. Comparison of temperature extracted from MODIS with temperature data from
Malaysian
Meteorological
Service
(MMS).
Correlation
between
both
temperatures from MODIS and MMS were also examined.
3. Validation of fire risk map with fire occurrence data from FRDM.
3.6
Summary
This chapter discussed the methodology used in this study from early stage of
data acquisition, data pre-processing, data processing, data analyzing and obtaining
results.
60
MODIS data along with several ancillary data were used in this study. Two
types of MODIS data were used in this study namely MODIS Level 1B 1KM
Radiances Product (MOD021KM) and MODIS Geolocation Product (MOD03).
In order to extract fire risk map, several processing techniques were applied
to the datasets namely Normalized Difference Vegetation Index, brightness
temperature, Vegetation Condition Index, Temperature Condition Index (TCI) and
Vegetation Health Index. The hot spots are extracted with several attributes as shown
in Table 3.6, that it can be analyzed with GIS software like ArcView.
CHAPTER IV
FOREST FIRE INTERFACE
4.1
Introduction
In order to develop the forest fire interface, several approaches were applied
to assessing system products throughout the system development life cycle. This is to
ensure that quality is built into the software and the software satisfies user
requirements.
Forest fire interface is developed mainly using Microsoft Visual C++. Take
an advantage of new technology namely Object Linking Embedding, the forest fire
interface is also link to DOS prompt and ArcView software.
62
4.2
System Development Life Cycle
The system development life cycle is the process of developing information
system through investigation, analysis, design, implementation and maintenance.
Figure 4.1 shows the steps involved in the system development life cycle.
Reference and data
collection
Early
investigation
Start
Requirement analysis
Development
phase
System framework
Source code
development
System testing
Error
No
End
Figure 4.1: Diagram of system development life cycle.
Evaluation
phase
Yes
63
Each step in the system development life cycle is explained in the Table 4.1 below:
Table 4.1: Explanation of system development life cycle.
Step
Explanation
Reference and data
collection
The first step is to identify a need for the new system.
This will include determining the main objective of the
developed
system,
problem
statement
and
data
acquisition.
Requirement analysis
Requirements analysis is the process of analyzing the
information needs of the end users, any system presently
being used and developing the functional requirements of
a system that can meet the needs of the users. Also, the
requirements should be recorded user interface storyboard
or executable prototype. The requirements documentation
should be referred throughout the rest of the system
development process to ensure the developing project
aligns with user needs and requirements.
System framework
After the requirement has been determined, the necessary
specifications for the hardware, software, people and data
resources that will satisfy the functional requirements of
the proposed system can be determined. The design will
serve as a blueprint for the system and help detect
problems before these errors or problems are built into the
final system.
Source code
development
Coding and debugging is the act of creating the final
System testing
The system must be tested to evaluate its actual
system.
functionality in relation to expected or intended
functionality. Demonstration using real dataset is
required. End users will be key in determining whether
the developed system meets the intended requirements.
64
4.3
Overview of Forest Fire Interface
Forest fire interface consists of a group of computer programs to process
satellite data and to carry out proximity analysis (Figure 4.2). There are three main
components created in order to develop this interface namely vegetation index, hot
spot and proximity analysis. Vegetation index and hot spot are computer programs
written in C++ language before it convert into MS DOS, while proximity analysis is
an extension from ArcView software, which was written in Avenue language. This
interface has the ability to connect to another program through Object Linking
Embedding (OLE). OLE refers to the process of inserting linked and embedded
objects in one application that is created by another application (Perry and Hettihewa,
1998). Therefore in this study, forest fire interface will connect to other programs
namely MS DOS and ArcView.
Vegetation Index
Forest Fire
Interface
Hot Spot
Proximity Analysis
Figure 4.2: Forest fire interface diagram component.
Vegetation index was developed from four computer program written in C++
language to process MODIS data. The computer program was written to extract
Vegetation Condition Index (VCI), Temperature Condition Index (TCI), Vegetation
Health Index (VH) and Fire Risk Index (Figure 4.3).
65
Vegetation
Condition Index
(VCI)
Temperature
Condition Index
(TCI)
Vegetation Index
Vegetation Health
Index (VH)
Fire Risk Index
Figure 4.3: Diagram of vegetation index components namely Vegetation Condition
Index (VCI), Temperature Condition Index (TCI), Vegetation Health Index (VH) and
Fire Risk Index.
All the components of vegetation index in Figure 4.3, which is written in C++,
were then convert into MS DOS format. The input data for vegetation index is
Normalized Difference Vegetation Index (NDVI) and brightness temperature (BT) of
MODIS data. Both data must be geometrically corrected using Universal Transverse
Mercator (UTM) Projection with dimension of 382 pixel height by 425 pixel width.
The input data must be in Tagged Image File Format (TIFF) format. As shown in
Figure 4.4, input of NDVI data was used to carry out VCI, while input of BT data
was used to carry out TCI. Both VCI and TCI data were then used to extract VH. VH
results for four weeks were then used to carry out Fire Risk Index.
66
Brightness
Temperature data
NDVI data
Vegetation Condition
Index (VCI)
Temperature
Condition Index (TCI)
Vegetation Health Index
(VH)
Fire Risk Map
Figure 4.4: Flow of NDVI and brightness temperature data as input to carry out Fire
Risk Index.
Another component in forest fire interface is hot spot. Hot spot was
developed from two computer programs written in C++ language before it is
converted into MS DOS format (Figure 4.5). The computer programs namely
MOD14
and
MOD142shp
was
downloaded
from http://eostation.scanex.ru/software.html. MOD14 is a computer program to
execute MODIS fire detection algorithm. There are 44 product produced by MODIS
including raw data (MOD01), level 1B data (MOD02), geolocation data (MOD03)
and cloud mask product (MOD35). MODIS fire product was named MOD14 by
NASA. The input of MOD14 program is level 1B of MODIS data with 1km
resolution namely MOD021KM and geolocation data (MOD03). This will produce
hot spot data that were named MOD14 by NASA. The DOS command used to derive
the MODIS fire mask is as below:
mod14.exe <MOD021KM granule> <MOD03 granule> <L2 fire output file>
67
MOD14 or hot spot data was then used as an input for MOD142shp program.
MOD142shp program will convert raster data of hotspot into vector format or more
specifically, into shapefile format. The command used to extract the fire pixels and
attributes into ESRI shapefiles is as below:
mod142shp.exe -i input_mod14_file.hdf [-o output_shp_folder]
L1B MODIS data
(MOD021KM)
Geolocation MODIS
data (MOD03)
MODIS Fire Extraction
(MOD14)
Conversion of MOD14
to shapefile
Hot spot
Figure 4.5: Input data of MOD021KM and MOD03 used to extract the hot spot.
The last component in forest fire interface is proximity analysis. Proximity
analysis is an extension in ArcView 3.2 which is written in Avenue language. This
extension calculates the distance of an active theme to the closest feature in a theme.
This extension was downloaded from http://science.nature.nps.gov/nrgis/tools/ .
68
4.4
Forest Fire Interface
Figure 4.6 shows the main window of developed forest fire interface. Forest
fire interface consists of three main components namely Vegetation Index, Hot Spot
and Proximity Analysis. The source code to create the interface can be found in
APPENDIX C.
Figure 4.6: The main window for developed forest fire interface.
Vegetation Index component (Figure 4.7) consists of four vegetation indices
function namely Vegetation Condition Index, Temperature Condition Index,
Vegetation Health Index and Fire Risk Map. In this window, users are able to view
the digital image, which is in TIFF format. Users are also able to get information
about the displayed image as shown in Figure 4.8.
69
Figure 4.7: Vegetation Index window.
Figure 4.8: Information window of the displayed image.
Each vegetation indices function are linked to a window prompt that requires
the user to browse the input filename and the key in the output filename. NDVI
function window prompt as shown in Figure 4.9 require the user to browse Channel 1
and Channel 2 of MODIS data, which refer to visible and near infrared channel,
respectively. The user also has to key in the output name before running the process.
70
Figure 4.9: NDVI window prompt.
VCI function window prompt as shown in Figure 4.10 require the user to
browse NDVI, multiyear absolute minimum of NDVI and multiyear absolute
maximum of NDVI images of MODIS data. The user also has to key in the output
name before running the process.
TCI function window prompt as shown in Figure 4.11 require the user to
browse smoothed weekly brightness temperature, absolute minimum of smoothed
weekly brightness temperature and absolute maximum of smoothed weekly
brightness temperature images of MODIS data. The user also has to key in the output
name before running the process.
71
Figure 4.10: VCI window prompt.
Figure 4.11: TCI window prompt.
72
VH function window prompt as shown in Figure 4.12 require the user to
browse VCI and TCI images of MODIS data. The user also has to key in the output
name before running the process.
Figure 4.12: VH window prompt.
Fire Risk Map function window prompt as shown in Figure 4.13 require the
user to browse VCI and TCI images of MODIS data. The user also has to key in the
output name before running the process. Fire Risk Map are generated from
accumulation of VCI and TCI for several weeks. The user can choose to accumulate
the input dataset from several weeks as shown in Figure 4.14.
73
Figure 4.13: Fire Risk Map window prompt.
Figure 4.14: Prompt window to set up the number of input weeks to Fire Risk Map.
Extracting Hotspot Index window (Figure 4.15) consists of two object linking
embedding function to MS DOS namely Hot Spot and Convert. It is also equipped
with Close function to log out from this window.
74
Figure 4.15: Hotspot Index window prompt.
Each object linking embedding function in Hotspot Index window are linked
to MS DOS prompt. To run fire algorithm the user must provide MODIS Level 1B
(MOD021KM) and corresponding geolocation (MOD03) data as input. Both of this
data must be saved in the same location as the mod14.exe application. No additional
input data are required. The user has to key in the DOS command in order to derive
the hot spot information (Figure 4.16). The command is as below:
mod14.exe <MOD021KM.hdf> <MOD03.hdf> <Output.hdf>
MOD021KM.hdf as shown above is the MODIS L1B dataset filename. MOD03.hdf
is the geolocation filename. Output.hdf is the output file name given by user.
Figure 4.16: MS DOS command used to derive the MODIS fire mask using the
MODIS Level 1B Radiances and Geolocation products.
75
Convert to Shapefile function is also linked to MS DOS prompt (Figure
4.17). This function converts hot spot information created by Hotspot function to
raster file or shapefile. The command used to extract the fire pixels and attributes
into shapefiles is as below:
mod142shp.exe -I Mod14.hdf Output_raster
Mod14.hdf is raster data from MODIS fire product that has been processed by
Hotspot function. Output_raster is a folder name where the converted raster data will
be saved into.
Figure 4.17: MS DOS window prompt to convert hotspot data into shapefile.
Forest fire interface is also linked to ArcView software. If ArcView is not
running, the application will start it. Vector data that has been generated can then be
viewed in ArcView and further analysis can be done.
Proximity analysis was carried out from ArcView interface. The hot spot data
must be in metre or kilometre map unit. From menu ‘AlaskaPak’ and ‘Closest
Features’ submenu as shown in Figure 4.18, distance of the fire pixel from the
76
nearest town or city, road and river can be measured. The nearest feature information
will be added in the database of the respective hot spots.
Figure 4.18: ArcView function, ‘Closest Feature’ submenu used to extract
proximity analysis.
4.5
Summary
Simple forest fire interface is developed in this chapter. The interface is
developed in order to process MODIS data to determine hot spots, to carry out fire
risk map and generate proximity analysis through ArcView. The interface provides
function to view and process raster data in TIFF format. In order to process the data
using this interface, raster data must be in the same dimension. However, the
interface do not provide any function to view the vector data. The user has to use
another software such as ArcView, Erdas Imagine or Envi to view the vector data.
55
CHAPTER V
RESULTS AND ANALYSIS
5.1
Introduction
This chapter discusses and analyzes the results obtained in this study. The
discussion is divided into three major categories as listed below:
i)
Discussion on derived hot spots from MODIS data.
ii)
Discussion on the comparison of temperatures extracted from MODIS with
temperature data from Malaysia Meteorological Services (MMS).
iii)
Discussion on the derived fire risk map
The analysis of hot spots from MODIS data was carried out from 1 January to
30 April 2005. The analysis can only be done during this time frame because the Fire
and Rescue Department of Malaysia (FRDM) only provide recorded forest fire
occurrences data during this period only. Analysis on the comparison of temperature
extracted from MODIS with temperature from MMS was carried out from May 2004
to April 2005. The analysis was carried out during this time frame because of
unavailability of MMS data after April 2005. The detailed results and discussion are
explained further below and compiled in Table 5.1.
78
5.2
Hot Spots Detected From MODIS
In this study, 75 datasets of MODIS data for Peninsular Malaysia area were
processed to extract hot spots beginning from 1 January 2005 to 30 April 2005. Some
of the datasets were obtained from mosaic images of two scenes for the respective
study area. Single datasets processed for each day exclude cloudy images, images
with bow-tie effect and data with technical problem from NASA. From the processed
data, there are 65 hot spots detected. The analysis of the hot spots only can be made
until 30 April 2005 due to unavailability of field data from FDRM.
From the 65 hot spots detected, 16 hotspots were found in Perak, followed by
12 hot spots detected both in Kedah and Johor, 10 hot spots detected in Pahang, 9 hot
spots detected in Selangor and 2 hot spots each detected in Pulau Pinang, Kelantan
and Terengganu.
Forest fires occur in the dry season. The southwest monsoon corresponds to
relatively drier weather in Malaysia. The monsoon is caused by land-sea temperature
differences due to heating by the sun’s radiation. During the southwest monsoon,
most states in Peninsular Malaysia experience monthly rainfall as low as 100mm.
Dry conditions in Peninsular Malaysia are also contributed by the rain shadow effect
due to the Sumatra mountain range. The dry season starts from the north of
Peninsular Malaysia and slowly heads to south of Peninsular Malaysia. Therefore,
early hot spots were detected in the north of Peninsular Malaysia and slowly the hot
spot spread to the south of Peninsular Malaysia. During this season, farmers and the
public were warned not to perform any open burning.
All the hot spots detected by MODIS were recorded by FRDM. Nevertheless
no daily MODIS data were available during this study. Clear MODIS data were
chosen instead of cloudy data. Data with the bow-tie effect (Jason, 2003) were also
excluded from the study.
79
Peat forest fire is a common problem to countries in this region. From the
Table 5.1, 16 hot spots were detected on peat. The hot spots detected in Kota Tinggi,
Johor, on 14 February 2005 occurred on peat area that was redeemed into cropland
area. Most of the peat forest fires occurred in Raja Muda Forest Reserve in Batang
Berjuntai, Selangor where forest fires occur every year (Ng Wei Loon, 2005).
Table 5.1 shows all the hot spots detected from 1 January to 30 April 2005.
The table includes several information, namely date of hot spots detected, latitude in
degree, longitude in degree, state, brightness temperature of Channel 21/22 (T4),
brightness temperature of Channel 31 (T11), confidence level, land use type and soil
type. From the table, 73% of the hot spots have a very high confidence level (80)
while 8% and 6% of it were detected with 60 and 20 confidence level, respectively.
80
Table 5.1: List of hot spots detected from MODIS from 1 January to 30 April 2005.
Fire
Date
Latitude Longitude
State
T4
1
18/01/2005
6.508
100.392
Kubang Pasu, Kedah
2
3
4
5
6
7
8
9
10
11
18/01/2005
03/02/2005
03/02/2005
03/02/2005
03/02/2005
10/02/2005
10/02/2005
10/02/2005
10/02/2005
10/02/2005
3.41
6.33
6.01
5.98
5.73
6.22
6.01
4.35
4.31
4.31
101.38
100.25
100.38
100.41
100.64
100.42
100.39
100.67
101.01
101.02
12 10/02/2005
4.18
13 10/02/2005
T11
Land Use
Soil Type
319.52
Confidence
(0-100)
299.88
80
Cropland
Kuala Selangor, Selangor
Kubang Pasu, Kedah
Kota Setar, Kedah
Pendang, Kedah
Kuala Muda, Kedah
Kubang Pasu, Kedah
Kota Setar, Kedah
Manjung, Perak
Perak Tengah, Perak
Kinta, Perak
317.11
326.91
325.93
324.65
327.81
324.92
326.55
315.81
331.17
319.27
299.62
303.30
300.93
301.58
300.96
296.73
300.50
297.00
297.45
298.27
80
80
20
80
80
80
80
80
80
80
Wetlands
Paddy
Paddy
Paddy
Cropland
Paddy
Cropland
Wetlands
Wetlands
Forest
100.97
Perak Tengah, Perak
323.27
297.30
80
Wetlands
4.17
101.01
Perak Tengah, Perak
324.32
296.19
80
Wetlands
14 10/02/2005
15 10/02/2005
16 10/02/2005
3.48
3.46
3.94
101.40
101.44
103.13
Batang Berjuntai, Selangor 321.24
Kuala Selangor, Selangor 318.7
Kuantan, Pahang
329.88
298.02
298.31
296.88
80
80
80
Wetlands
Wetlands
Forest
17 10/02/2005
2.27
102.83
Muar, Johor
320.21
296.74
80
Cropland
18 10/02/2005
19 10/02/2005
20 10/02/2005
2.03
2.33
2.07
102.73
103.33
103.37
Muar, Johor
Kluang, Johor
Kluang, Johor
335.19
315.10
316.57
300.08
295.88
298.06
80
80
80
Wetlands
Forest
Cropland
21 12/02/2005
6.18
100.31
Kota Setar, Kedah
329.36
303.50
80
Paddy
Serdang-BungorMunchong
Mined Land
Teluk-Guar
Teluk-Guar
Hutan-Semberin
Hutan-Semberin
Hutan-Semberin
Teluk Guar
Sungai,Peat Swamp
Holyrood-Lunas
Rengam-Bukit
Temiang
Telemong-AkobLocal Alluvium
Telemong-AkobLocal Alluvium
Peat
Mined Land
Telemong-AkobLocal Alluvium
Briah-Organic Clay
and Muck
Peat
Rengam-Jerangau
Batang MerbauMunchong
Chengai
81
…/Table 5.1 (continued)
22 12/02/2005
4.18
100.97
Perak Tengah, Perak
322.01
300.08
60
Forest
23 14/02/2005
24 02/02/2005
1.38
4.18
104.16
100.97
Kota Tinggi, Johor
Perak Tengah, Perak
326.34
316.30
301.82
284.80
80
80
Cropland
Forest
25
26
27
28
29
30
31
24/02/2005
24/02/2005
26/02/2005
26/02/2005
27/02/2005
28/02/2005
01/03/2005
3.48
2.55
3.48
2.55
3.48
3.48
4.18
101.4
103.59
101.4
103.59
101.4
101.4
100.97
Batang Berjuntai, Selangor
Rompin, Pahang
Batang Berjuntai, Selangor
Rompin, Pahang
Batang Berjuntai, Selangor
Batang Berjuntai, Selangor
Perak Tengah, Perak
356.24
328.28
315.10
323.75
320.20
322.08
322.01
295.60
296.22
295.88
294.42
296.74
293.80
300.08
80
80
80
80
80
20
60
Wetlands
Wetlands
Wetlands
Wetlands
Wetlands
Wetlands
Forest
32 03/03/2005
33 03/03/2005
5.98
4.18
100.40
100.97
Kota Setar, Kedah
Perak Tengah, Perak
316.30
356.23
284.80
295.62
80
80
Paddy
Forest
34 05/03/2005
35 05/03/2005
5.98
4.18
100.49
100.97
Pendang, Kedah
Perak Tengah, Perak
324.43
316.32
302.72
284.63
80
80
Paddy
Forest
36
37
38
39
05/03/2005
05/03/2005
05/03/2005
05/03/2005
3.62
3.61
3.42
2.55
103.26
103.29
103.06
103.59
Kuantan, Pahang
Pekan, Pahang
Pekan, Pahang
Rompin, Pahang
328.30
323.88
316.09
356.24
296.04
294.37
296.21
295.60
80
80
80
80
Wetlands
Wetlands
Wetlands
Wetlands
40 05/03/2005
2.44
102.82
Segamat, Johor
316.23
297.44
80
Forest
41 07/03/2005
5.46
102.96
322.01
300.08
60
Wetlands
42 07/03/2005
43 09/03/2005
44 21/03/2005
4.55
1.81
5.67
103.45
104.12
100.69
Kuala Terengganu,
Terengganu
Kemaman, Terengganu
Kota Tinggi, Johor
Baling, Kedah
322.24
329.74
329.46
293.91
301.69
300.48
20
80
80
Wetlands
Wetlands
Cropland
Telemong-AkobLocal Alluvium
Peat
Telemong-AkobLocal Alluvium
Peat
Peat
Peat
Peat
Peat
Peat
Telemong-AkobLocal Alluvium
Hutan Sembrin
Telemong-AkobLocal Alluvium
Hutan Semberin
Telemong-AkobLocal Alluvium
Peat
Peat
Holyrood-Lunas
Briah-Organic Clay
and Muck
Batu-AnamMalacca-Tavy
Peat
Rudua-Rusila
Peat
Malacca-TavyGajah Mati
82
…/Table 5.1 (continued)
45 21/03/2005
4.50
46 21/03/2005
4.41
100.93
100.89
Perak Tengah, Perak
Perak Tengah, Perak
327.26
330.44
297.11
298.12
60
60
47 01/04/2005
3.55
103.27
Pekan, Pahang
316.30
297.80
80
48 08/04/2005
49 08/04/2005
3.53
2.19
103.39
103.52
Pekan, Pahang
Kluang, Johor
322.23
326.24
297.77
293.18
80
80
50 12/04/2005
51 13/04/2005
52 13/04/2005
1.66
5.81
5.31
103.38
100.68
101.06
Pontian, Johor
Sik, Kedah
Hulu Perak, Perak
341.61
319.27
318.50
300.75
300.35
300.65
80
80
80
53 13/04/2005
54 15/04/2005
4.74
5.28
102.01
100.52
Gua Musang, Kelantan
315.39
Seberang Prai, Pulau Pinang 319.46
296.85
299.59
80
80
55 15/04/2005
56 15/04/2005
4.45
4.84
101.16
101.97
Kinta, Perak
Gua Musang, Kelantan
319.52
318.70
301.16
299.67
80
80
57 15/04/2005
58 15/04/2005
59 17/04/2005
3.93
2.81
3.84
103.14
101.59
103.22
Kuantan, Pahang
Kuala Langat, Selangor
Kuantan, Pahang
323.49
320.39
319.84
297.62
299.89
297.11
80
80
80
60 17/04/2005
61 20/04/2005
1.82
4.53
103.66
100.77
Kota Tinggi, Johor
Manjung, Perak
345.42
319.77
291.09
296.69
60
80
62 05/03/2005
63 22/04/2005
5.32
2.81
100.29
101.60
Penang
Kuala Langat, Selangor
324.70
316.30
298.21
297.84
20
80
64 22/04/2005
2.66
102.79
Segamat, Johor
335.24
297.36
80
Cropland
65 22/04/2005
2.66
102.80
Segamat, Johor
316.09
296.68
80
Cropland
Cropland
Paddy
Serdang-Kedah
Telemong-AkobLocal Alluvium
Cropland Briah-Organic Clay
and Muck
Cropland
Rudua-Rusila
Cropland
Telemong-AkobLocal Alluvium
Wetlands
Peat
Cropland Rengam-Jerangau
Cropland
Rengam-Bukit
Temiang
Cropland
Steepland
Cropland
Telemong-AkobLocal Alluvium
Wetlands
Mined Land
Cropland Durian-MunchongBungor
Forest
Holyrood-Lunas
Cropland Selangor-Kangkong
Wetlands
Telemong-AkobLocal Alluvium
Cropland Rengam-Jerangau
Cropland
Rengam-Bukit
Temiang
Urban area
Urban Land
Cropland Selangor-Kangkong
Organic Clay and
Muck
Organic Clay and
Muck
83
Figure 5.1, represents the distribution of hot spots by land use type. It can be
seen that most of the hot spots (37%) were found in wetland area, with 14 of it
detected in peat swamp. The hot spots were detected 5 times in peat swamp of Raja
Muda Musa Forest Reserve in Kuala Selangor, Selangor and twice in Rompin,
Pahang. Other peat swamp areas with hot spots are in Manjung in Perak, Muar in
Johor, Kuantan and Pekan in Pahang, Kuala Terengganu in Terengganu, and Kota
Tinggi and Pontian in Johor.
Figure 5.1: Distribution of hot spots by land use type.
The peat fires that have occurred throughout Peninsular Malaysia are the
consequence of peat swamps being drained for agriculture and forestry purposes.
From Table 5.2, there are approximately 32% of the area developed for agriculture in
Peninsular Malaysia or 100,352 hectares of peat under cultivation (Jamaluddin Jaya,
2002). The most important crop at present is oil palm. Draining out the water,
especially during the dry season causes the lowering of the water table. Hence, the
swamps dry out and act like a dry sponge and become more vulnerable to fire. Peat
can be identified by its dark and brown colour. It has been accumulating since the
last ice age from partly-decomposed plants such as descended leaves and trunks,
accumulated dead organic material, small organisms and decaying material. It is a
long natural process to compose this soil. Peat with 2 to 3 metres thickness will take
about 2,000 to 4,000 years to develop (Buol et al., 1973).
84
Table 5.2: Peat utilization in Malaysia.
Region
State/Division
Peninsular
Malaysia
Johor
Pahang
Selangor
Perak
Terengganu
Kelantan
Negeri Sembilan
298,500
282,500
194,300
107,500
88,000
7,400
6,300
Area
Developed for
Agriculture
(ha)
145,900
17,100
59,900
69,700
13,900
2,100
5,000
Peat area
developed for
Agriculture
(%)
49
6
31
65
16
28
79
984,500
313,600
32
502,466
340,374
314,585
205,479
172,353
168,733
34,730
26,827
269,571
50,836
66,114
50,836
61,112
47,591
8,715
n.a.
54
15
21
25
35
28
25
-
1,765,547
86,000
554,775
n.a.
31
n.a.
Sub-total for Sabah
86,000
Grand Total
2,836,047
n.a.
868,375
n.a.
31
Sub-total for
Peninsular Malaysia
Sarawak
Sibu
Sri Aman
Miri
Samarahan
Sarikei
Bintulu
Limbang
Kuching
Sabah
Sub-total for Sarawak
Sabah
Total
area (ha)
(Source:Jamaluddin Jaya, 2000)
A total of 34% of the hot spots were detected in cropland areas. Land clearing
by farmers for agricultural purpose is the most popular reason to start a fire. Fifteen
percent (15%) of the hot spots were found in forested areas and the most frequent hot
spots in forested areas occurred in Gunung Kenderong, Perak. The hot spots were
detected as frequent as 7 times (2 February 2005, 10 February 2005, 12 February
2005, 14 February 2005, 1 March 2005, 3 March 2005 and 5 March 2005). From
FRDM records, the fires in Gunung Kenderong occurred due to hikers who
carelessly left behind objects that may start fire easily. MODIS detected hot spots as
early as 2 February 2005, but FRDM only started to extinguish the fire on 4 February
2005 after receiving reports from the public. The forest fire continued until 17
85
February when finally rainfall put out the fire. Unfortunately the fire occurred again
on 1 March 2005 and has been brought under control on 5 March 2005.
The remaining 12% and 2% of fires were detected in paddy areas and in urban
areas, respectively. The paddy areas in north of Peninsular Malaysia (Figure 5.2(a))
was dried out due to prolonged dry season as reported in Utusan Malaysia (Nidzwan
Zainal Abidin, 2005). Due to extreme drought, farmers were urged to postpone their
paddy cultivating season.
(a)
(b)
Figure 5.2: Extreme drought occurrence in (a) Kampong Teluk Jambu Bintong,
Kangar, Perlis; and (b) Firemen battling bush fire outside the Penang International
Airport cargo complex in Bayan Lepas.
86
One of the hot spots that occurred was detected in an urban area. As reported
in The Star, a bush fire occurred near to Penang International Airport cargo complex
in Bayan Lepas as shown in Figure 5.2(b) on 5 March 2005. FRDM also recorded
that the Department had put out several bush fires in the same area in the previous
month.
5.3
Comparison of Temperatures Extracted From MODIS With Temperature
Data From MMS
In this study, brightness temperature extracted from the satellite data were
converted to temperature and compared with temperature data obtained from MMS.
The MMS data used in this study were collected from the distribution of weather
stations over Peninsular Malaysia. This analysis was carried out in order to examine
the accuracy of observed temperature from satellite data.
Figure 5.3 shows the correlation between field temperature obtained from
MMS and observed temperature from satellite data. Continuous MODIS and weather
data from January 2004 to December 2004 were used, involving 334 MODIS
datasets processed to generate the regression value. Data for each month were
represented by different colour. From the graph, it shows that temperature in
December were relatively low. Low temperature and heavy rain during this month
were brought by the Northeast Monsoon. High temperatures were recorded in the
drier season around March and August. These findings were strengthened by the fact
that most of the forest fires in Malaysia occur in drier season around March. There
also several fire occurrences reported in August due to the transition period in
between the monsoons.
87
Temperature derived from MODIS and temperature obtain from MMS were
highly correlated with 0.82 regression value. It shows that MODIS data can be used
to extract temperature information. The highest temperature derived from satellite
data was 37oC, while the lowest temperature derived was 26oC.
MODIS Temperature
Temperature(Degree
(DegreeCelcius)
Celcius
MODIS
39
y = 1.15x – 4.55
R2 = 0.82
37
Jan
Feb
35
March
April
33
May
June
31
July
August
29
September
October
27
November
December
25
25
27
29
31
33
35
37
39
Field Temperature (Degree Celcius)
Figure 5.3: Correlation between temperature from Malaysian Meteorological
Services (MMS) and observed temperature from MODIS.
88
5.4
Generating A Fire Risk Map
Vegetation index has been considered by numerous scientists as one of the
important parameters for the mapping of agricultural fields, rainfall monitoring,
estimating weather impacts, calculating biomass, crop yield and pasture production,
drought conditions and determining the vigour of the vegetation (Singh et al., 2003).
Kogan (2001) has developed a new method for early drought detection based on
estimation of vegetation stress from NOAA AVHRR indices. Unlike the two spectral
channel approach routinely applied to vegetation monitoring, this numerical method
is based on the combination of three spectral channels namely visible (0.50-0.68μm),
near infrared (0.73-1.10μm) and thermal infrared (10.30-11.30μm). The visible and
near infrared values were converted to the NDVI and brightness temperature. The
NDVI and brightness temperature were converted to the Vegetation Condition Index
(VCI), Thermal Condition Index (TCI) and Vegetation Health Index (VH).
In this study, MODIS data is used instead of NOAA AVHRR. Three spectral
channel of MODIS data were used in this numerical method namely visible (0.620.67μm), near infrared (0.84-0.87μm) and thermal infrared (10.78-11.28μm). All the
spectral channels of the MODIS data chosen have similar characteristics in terms of
the wavelength with NOAA AVHRR data used by Kogan (2001). Three indices
characterizing moisture, namely Vegetation Condition Index (VCI), thermal namely
Thermal Condition Index (TCI) and vegetation health condition (VH) were
constructed from the processed NDVI and BT following the principle of comparing a
particular year NDVI and BT with the range of their variation during the extreme
conditions. The extreme conditions were derived by calculating the maximum and
minimum (MAX-MIN) NDVI and BT values for cumulative of 5 years (2000 to
2004) of satellite data for each land pixel of the study area. Each component to
extract VH namely NDVI, BT, VCI and TCI are described below.
89
5.4.1
Daily Normalized Difference Vegetation Index (NDVI)
Radiances measured by band 1 and band 2 on board MODIS satellites were
used in the study. Green and healthy vegetation reflects less solar radiation in the
visible channel namely band 1 of MODIS data compared to near infrared channel
data. Under stress condition, visible channel values may increase and near infrared
channel values may decrease (Singh et. al., 2003). Healthy and dense vegetation
show large NDVI values. In contrast, clouds and water have larger visible reflectance
than near infrared channel. Thus, those features yield negative index values. Rock
and bare soil areas have similar reflectance both in visible and near infrared channel
and result in vegetation indices near zero.
Figure 5.4 shows NDVI values generated from MODIS data on 6 April 2000.
Minimum NDVI value derived from the data was (-0.75), while the maximum value
derived from it was 0.65. The mean value derived from the data was 0.32. The high
NDVI values indicate that the vegetation was in healthy condition.
5.4.2
Smoothed Weekly Normalized Difference Vegetation Index (NDVI)
In order to minimize cloud effects, the NDVI were smoothed or averaged
over a 7 day period. Figure 5.5 shows smoothed NDVI from fifth week of 2005.
Minimum NDVI values derived from the images was (-0.64) and the maximum
values derived from the images was 0.61, while mean values derived from the
images were 0.48. The mean value depends heavily on how much of the area was
cloud covered. Since the data used in this study were smoothed over a week and
exclude data with more than 10% cloud, the cloud effect in each composite data is
minimal.
90
5.4.3
Minimum and Maximum Normalized Difference Vegetation Index
Minimum and maximum NDVI values are the lowest and the highest weekly
values observed during the 2000 to 2004 period for each pixel, respectively. This
procedure was based on three environmental laws, namely law of minimum, law of
tolerance and the principle of carrying capacity (Hardin, 1986). These laws provide
the basis for determining the lowest and the highest potentials of an ecosystem’s
resources in response to the environment.
Basically, extreme NDVI values during the years 2000 through 2004 were
calculated for each week and pixel. The resulting multi-year maximum and minimum
NDVI (figure 5.6 and 5.7) were used as the criteria for estimating the upper
(favourable weather) and the lower (unfavourable weather) limits of the ecosystem
resources in response to extreme weather condition (Kogan et al., 2003).
91
Perlis
Kedah
P. Pinang
Kelantan
Terengganu
Perak
Pahang
Selangor
N. Sembilan
Melaka
Johor
0
100
Figure 5.4: NDVI derived from single MODIS dataset on 6 April 2000.
92
Perlis
Kedah
Kelantan
P. Pinang
Terengganu
Perak
Pahang
Selangor
N. Sembilan
Melaka
Johor
0
Figure 5.5: Smoothed weekly MODIS datasets from 5th week of 2005.
100
93
Perlis
Kedah
P. Pinang
Kelantan
Terengganu
Perak
Pahang
Selangor
N. Sembilan
Melaka
Johor
0
Figure 5.6: Multi-year maximum MODIS datasets
100
94
Perlis
Kedah
Kelantan
P. Pinang
Terengganu
Perak
Pahang
Selangor
N. Sembilan
Melaka
Johor
0
Figure 5.7: Multi-year minimum MODIS datasets.
100
95
5.4.4
Brightness Temperature (BT)
Brightness temperatures were derived from channel 31 of MODIS data
(10.78-11.28μm). Single brightness temperature values derived in this study are
shown in Figure 5.8. The measurement was generated from MODIS data on 5 April
2000. Minimum brightness temperature values derived from the data was 269.53,
while the maximum value derived from it was 303.40.
5.4.5
Smoothed Weekly Brightness Temperature (BT)
Brightness temperature data were composited over a 7 day period in
order to minimize the cloud effect. Figure 5.9 shows a composite image from 14th
week of 2001. Minimum brightness temperature value derived from the images was
263.49 and the maximum values derived from the images was 305.94.
5.4.6
Minimum and Maximum Brightness Temperature
Smooth weekly brightness temperatures were processed to extract the lowest
and highest weekly values for each land pixel in each respective year. Multiyear
minimum and maximum brightness temperature values observed during the 2000 to
2004 period for each pixel were generated and are shown in Figure 5.10 and 5.11.
96
Perlis
Kedah
P. Pinang
Kelantan
Terengganu
Perak
Pahang
Selangor
N. Sembilan
Melaka
Johor
0
100
Figure 5.8: Brightness temperature derived from single MODIS dataset on 5 April
2000.
97
Perlis
Kedah
P. Pinang
Kelantan
Perak
Terengganu
Pahang
Selangor
N. Sembilan
Melaka
Johor
0
100
Figure 5.9: Smoothed weekly brightness temperature of MODIS datasets derived
from 14th week of 2005.
98
Perlis
Kedah
P. Pinang
Kelantan
Terengganu
Perak
Pahang
Selangor
N. Sembilan
Melaka
Johor
0
100
Figure 5.10: Multi-year maximum of brightness temperature derived from MODIS.
99
Perlis
Kedah
P. Pinang
Kelantan
Terengganu
Perak
Pahang
Selangor
N. Sembilan
Melaka
Johor
0
100
Figure 5.11: Multi-year minimum of brightness temperature derived from MODIS.
100
5.4.7
Vegetation Condition Index (VCI)
The Vegetation Condition Index (VCI) concept was designed to extract the
weather component from NDVI values (Kogan et al., 2003). VCI changes from 1 to
100, correspond to changes in vegetation condition from extremely unfavourable to
optimal. Figure 5.12 shows VCI derived from 15th week of 2005. The VCI values
show generally favourable condition. But extreme conditions were detected in many
places in Johor, Melaka and Selangor.
5.4.8
Temperature Condition Index (TCI)
The Temperature Condition Index (TCI) is based on the thermal band (10.7811.28μm) which is converted into brightness temperature. TCI is used to determine
temperature related vegetation stress. Figure 5.13 shows TCI image derived from 15th
week of 2005. The image shows that most of the area in Peninsular Malaysia are in
favourable condition except in some places in Johor and Melaka.
5.4.9
Vegetation Health Index (VH)
Vegetation Health Index (VH) in Figure 5.14 estimates vegetation health from
VCI and TCI condition (Kogan, 2000). If conditions are favourable which means VH
is larger than 75, vegetation is healthy. If vegetation is stressed, VH value is below
25.
101
Figure 5.14 shows VH from 15th week of 2005. The result shows that
vegetation conditions during the above period is fair. But some areas in south of
Peninsular Malaysia such as Johor and Melaka shows vegetation stress.
5.5
Fire Risk Map
The technique for monitoring potential fire danger or fire risk is based on the
analysis of duration and intensity of vegetation stress. The highest fire risk is
estimated when intensive vegetation stress (VH<25) continues for at least 4 weeks.
This technique provided timely warning of possible fire activity. Figure 5.15 shows a
fire risk map of 10th week of 2005. In general, the vegetation condition for Peninsular
Malaysia is favourable and fair. But there are some places such as Kuala Lumpur,
Negeri Sembilan and Johor which have high risk to forest fire. High risk areas are
indicated in orange colour.
102
Perlis
Kedah
P. Pinang
Kelantan
Terengganu
Perak
Pahang
Selangor
N. Sembilan
Melaka
Johor
0
Figure 5.12: Vegetation Condition Index derived from 15th week of 2005
100
103
Perlis
Kedah
P. Pinang
Kelantan
Perak
Terengganu
Pahang
Selangor
N. Sembilan
Melaka
Johor
0
Figure 5.13: Temperature Condition Index derived from 15th week of 2005
100
104
Perlis
Kedah
P. Pinang
Kelantan
Terengganu
Perak
Pahang
Selangor
N. Sembilan
Melaka
Johor
100
Figure 5.14: Vegetation Health Index derived from 15th week of 2005
0
105
5.6
Vegetation Index Analysis
The relationship between VCI and rainfall, TCI and temperature, also VH
and temperature were examined in this study. Distribution of temperature and rainfall
were provided by MMS until 17th week of 2005 only. Therefore this analysis is
carried out from 18th week of 2004 to 17th week of 2005.
As shown in Figure 5.16, there is a good match between VCI and rainfall
data. The relationship shows a close correspondence between VCI and rainfall and
the dynamics of VCI and rainfall matched quite well. The VCI values increase in
weeks of 35 to 50, along with frequent rainfall. The VCI were low during early
weeks of the years.
106
Perlis
Kedah
Terengganu
P. Pinang
Kelantan
Perak
Pahang
Selangor
N. Sembilan
Melaka
Johor
0
Figure 5.15: Fire Risk Map of 11th week of 2005.
100
107
Figure 5.17 shows the relationship between TCI and temperature from
MMS. TCI values increased in weeks 40 to 45 of the year. However, the temperature
obtained from MMS decreased a little in weeks 40 to 45. Temperature values from
MMS do not differ much for the whole year, while TCI values dynamically changed
throughout the year.
Figure 5.18 shows the relationship between VH and temperature. VH
values increased in weeks 40 to 50, conversely while during this period, temperature
values decreased.
Relationship Between Vegetation Condition Index (VCI) and Rainfall
70
VCI
Rainfall
60
50
40
30
20
10
Weeks
Figure 5.16: The relationship between VCI and rainfall.
17
14
11
8
5
2
51
48
45
42
39
36
33
30
27
24
21
0
18
Vegetation Condition Index (%)
80
108
Relationship Between Temperature Condition Index (TCI) and Temperature
75
Temperature Condition Index (%)
70
TCI
65
Temperature
60
55
50
45
40
35
30
17
14
11
8
5
2
51
48
45
42
39
36
33
30
27
24
21
18
25
Weeks
Figure 5.17: Relationship between TCI and temperature.
Relationship Between Vegetation Health (VH) and Temperature
75
VH
Temperature
65
60
55
50
45
40
35
30
Weeks
Figure 5.18: The relationship between VH and temperature.
17
14
11
8
5
2
51
48
45
42
39
36
33
30
27
24
21
25
18
Vegetation Health Index (%)
70
109
5.7
Fire Risk Map Analysis
From the Appendix A, fire risk map derived in this study was validated with
65 forest fire occurrence data from FRDM during 1 January 2005 to 30 April 2005.
Forty six (46) of the fire occurrences were categorized as high fire risk level, while
14 were classified as moderate fire risk level and the remaining 6 fire occurrences
were found in low fire risk level. From 46 of the forest fire occurrence categorized in
high risk level, therefore the accuracy of fire risk map generated in this study is 71%
(APENDIX B).
5.8
Summary
This chapter discussed the results achieved from the study. The hot spots
detected from MODIS data were validated with recorded data of fire occurrence from
FRDM. There were 65 hot spots detected during 1 January 2005 to 30 April 2005.
Most of the hot spots were detected on peat land and has burned twice or more.
Vegetation index derived in this study, namely VCI, TCI and VH is a
numerical index which changes from 0 to 100 characterizing change in vegetation
conditions from extremely poor to excellent. This new vegetation index approach
combines the visible, near infrared and thermal radiances in a numerical index
characterizing vegetation health. This approach is extremely useful in detecting and
monitoring drought phenomenon which sets the conditions for fire occurrence.
Fire risk maps were developed from the cumulative value of VH for four
weeks. The fire risk estimate is based on intensity and duration of combined moisture
110
and thermal stress. If severe stress continues for one week, fire risk is minimal, while
it reaches maximum if this condition continues for four weeks and longer. Fire risk is
higher if vegetation stress is severe and persistent. High level of fire risk due to both
moisture and temperature conditions can be used as a fire danger warning.
CHAPTER VI
CONCLUSIONS AND RECOMMENDATIONS
6.1
Introduction
Moderate Resolution Imaging Spectroradiometer (MODIS) satellite data is used
in this study to extract hot spots and carry out fire risk map. Two types of MODIS data
were used to extract hot spots, namely MODIS Level 1B 1KM Radiances Product
(MOD021KM) and MODIS Geolocation Product (MOD03), while MOD021KM was
used to generate fire risk map.
6.2
Conclusions
From the results derived in this study, the following conclusions are made:
112
1. Moderate Resolution Imaging Spectroradiometer (MODIS) satellite data were
used to extract hot spot and biophysical parameter to derive fire risk map.
Results from this study show that MODIS can be a new alternative to NOAA
AVHRR (National Oceanic and Atmospheric Administration Advanced Very
High Resolution Radiometer) to detect hot spot and derive fire risk map. The
advantages of MODIS data over NOAA are:
a. Saturation value for NOAA data is lower than MODIS
b. Variety of wavelengths to choose from MODIS data
2. High correlation was found between temperature derived from MODIS and the
ground truth data obtained from MMS (R2 = 0.8).
3. Forest fire map generated from the study gives a high accuracy, which is 71%.
4. Simple application was successfully developed to generate fire risk map. This
application is able to process MODIS data in TIFF format.
5. New techniques were used for assessment of fire risk. Scientists usually utilized
NOAA AVHRR data to derived fire risk map and hot spots. Due to some
limitation, MODIS data were used in this study instead of NOAA AVHRR.
6.3
Recommendations
From the results obtained in this study, the following recommendations are made
to improve the results in future studies.
1. In this study, cumulative of 5 years datasets were used to extract fire risk map.
Larger datasets of at least 10 years would produce more accurate results.
113
2. The results obtained from the MODIS data with cloud cover may not be so
accurate. This could be improved by using smoothed weekly datasets. Preprocessing such as geometric correction, radiometric correction and atmospheric
correction should be emphasized before further processing is carried out in order
to achieve more accurate results. The function of developed forest fire interface
can be enhanced by attaching it to a greater forest fire system.
3. More ground truth data are needed to carry out the accuracy assessment. Limited
recorded data of forest fire occurrence from Fire and Rescue Department of
Malaysia (FDRM) is a big limitation in assessing the accuracy.
114
REFERENCES
Ahmad Zainal Mat Isa (2001). Forest Fire In Malaysia: Its Management and
Impact on Biodiversity. Workshop on Minimizing the Impact of Forest Fire
on Biodiversity in ASEAN. http://www.arcbc.org/arcbcweb/pdf/vol1no3/3135_special_reports.pdf.
Akroyd, T.N.W. (1957). Laboratory Testing In Soil Engineering. Soil Mechanics
Ltd.
Alfreds R. J. (1966). Thermal Soil Mechanics. Rutgers University Press.
Anderson, I.P., Imanda, I.D. and Muhnandar (1999). Vegetation Fires In
Indonesia: Operating Procedures For The NOAA-GIS Station In Palembang,
Sumatra. Series report by Forest Fire Prevention and Control Project,
European Union Ministry of Forestry And Estate Crops.
Andrews, P. L. (1986). BEHAVE: Fire Behavior Prediction and Fuel Modeling
System-BURN Subsystem. United State Department of Agriculture Service.
General Technical Report.
Andrews, P. L. and Williams, J. T. (1998). Fire Potential Evaluation In Support of
Prescribed Fire Risk Assessment. In Teresa L. Priden and Leonard A.B.
(eds). Fire In Ecosystem Management: Shifting The Paradigm From
Suppression To Prescription. Tall Timbers Fire Ecology Conference
Proceedings, No. 20, Tall Timbers Research Station, Tallahassee, FL. Page
64-68.
Antikidis E, Arino, O., Laur, H. and Arnaud A. (1998). Deforestration Evaluation
by Synergistic Use of ERS SAR Coherence and ATSR Hot Spots, the
Indonesian Fire Event of 1997. ESRIN, Frascati, Italy.
Atkinson, P. (1998). Radar Performance Modelling for the Real World. IEEE
Savoy Place London.
Bachman, A. and Allgower, B. (1998). The Need For A Consistent Wildfire Risk
Terminology. 3111 International Conference on Forest Fire Research.
115
Barnes, W.L., Pagano, T.S. and Salomonson, V.V. (1998). Prelaunch
Characteristics of the Moderate Resolution Imaging Spectroradiometer
(MODIS) on EOS-AM1. IEEE Transactions on Geoscience and Remote
Sensing, Vol. 36, No. 4, Pg 1088-1100.
Bell, F. G. (1983). Engineering Properties of Soils and Rocks. Butterworth & Co
Ltd, 2nd Edition.
Bevins, C. D. (1996). FireLib User Manual and Technical Reference.
Intermountain Research System, Forest Services, United State Department
of Agriculture.
Bozniak, E., Clark, S, Gatherum, D. and Wachocki, B. (2005). Resource Guide
For The Weber State University Botany Department Field Trip To Antelope
Island State Park. The Western Society of Weed Science in Cooperation with
the Western United States Land Grant Universities Cooperative Extension
Service.
Brown, A.A. and Davis, K.P. (1973). Forest Fire Control and Use. McGraw Hill
series in Forest Resources.
Brown, J. K., Oberheu, R. D., Johnston,C. M. (1982). Handbook for inventorying
surface fuels and biomass in the Interior West. General Technical
Report.INT-129. Ogden, UT:US. Dept of Agriculture, Forest Service,
Intermountain Forest and Range Experiment Station.
Buol, S. W. (1973). Soil Genesis and Classification. Ames, Iowa : Iowa State
Univ Press.
Buongiorno A, Arino, O., Zehner, C., Colagrande, P. and Gorly, P. (1997). ERS-2
Monitors Exceptional Fire Event in South-East Asia. International Journal
of Remote Sensing 15, 2235-2550.
Burgan, R. E., Barlette, R. A., (1994). Vegetation Condition and Fire Occurrence :
A Remote Sensing Connection.
Burgan, R. E., Klaver, R. W., Klaver, J. M. (1998). Fuel Models and Fire Potential
from Satellite and Surface Observations. International Journal of Wildland
Fire 8, 159-170.
116
Call, P.T. and Albini, F.A. (1997). Aerial And Surface Fuel Consumption in
Crown Fires. International Journal of Wildland Fire 7(3). Pg 259-264.
http://www.firelab.org/pdf/fbp/misc/asfuel.pdf last visited 31 March 2003
3.30pm.
Cantelaube, P., San-Miguel-Ayanz, J, Schmuck, G. and Biard, F. (2002). Towards
Forecasting Forest Fire Risk. Special Report from Institute for Environment
and Sustainability.
Carter, M. R., (1993). Soil Sampling and Methods of Analysis. Lewis
Publishers.93 S64
Casagrandi, R. and Rinaldi, S (1999). Minimal Model for Forest Fire Regimes.
The American Naturalist. Vol 153, No 5 p527-539. 1993 c2
Chuvieco, E. and Congalton R. G. (1989). Application of Remote Sensing and
Geographic Information System to Forest Fire Hazard Mapping. Remote
Sensing of Environment 29, 147-159.
Chuvieco, E., Vaughn, P., Riano, D., and Cocero, D., (2000). Fire Danger And
Fuel Moisture Content Estimation From Remotely Sensed Data.
Coleman, J. R. and Sullivan, A. L. (1996). A Real Time Computer Application
For The Prediction Of Fire Spread Across the Australian Landscape.
Simulation 67 (4): 230-240.
Fang M and Huang W. (1998). Tracking the Indonesian Forest Fire Using
NOAA/AVHRR Images. International Journal of Remote Sensing 19, 3,
387-390.
Finney, M.A. (1998).FARSITE: Fire Area Simulator-Model Development and
Evaluation. Special Report from Rocky Mountain Research Station, United
State Department of Agriculture.
Frandsen, W.H. (1987). The influence of moisture and mineral soil on the
combustion limits of smoldering forest duff. Canadian Journal of Forest
Research 17.
117
Frandsen, W.H. (1991). Burning rate of smoldering peat. Northwest Science.
65(4): 166-172.
Giglio, L., Descloitres, J., Justice, C. O. and Kaufman, Y. J. (2003). An Enhanced
Contextual Fire Detection Algorithm for MODIS. Remote Sensing of
Environment 87:273-282.
Goldammer, J. G., (1983). Fire in Tropical Biota: Ecosystems Processes and
Global Challenges. Springer, Berlin.
Gordon, H.R. (1990). Radiometric Considerations for Ocean Color Remote
Sensors. Appl. Opt. Vol 29, 3228-3236.
Griffith, D. (1999). Drought Factor Algorithm in McArthur’s Forest Fire Danger
Meter.
Australian
Bushfire
Conference,
Albury.
http://www.uni-
freiburg.de/fireglobe
Guenther, B., Godden, G.D., Xiong, Xiaoxiong, Knight, E.J., Qiu, Shi-Yue,
Montgomery, Harry, Hopkins, M.M., Mohammad G. Khayat and Hao,
Zhidong ( 1998). Prelaunch Algorithm and Data Format for the Level 1
Calibration Products for the EOS-AM1 Moderate Resolution Imaging
Spectroradiometer (MODIS). IEEE Transactions On Geoscience and
Remote Sensing. Vol 36, No 4, 1142-1151.
Haan, E. D., Termaat, R., Edil, T. B. (1993). Advances In Understanding And
Modelling The Mechanical Behaviour Of Peat. A.A. Balkema Publishers
Hardin, G. (1986). Cultural Carrying Capacity: A Biological Approach to Human
Problems. Biological Science, 36: 599-606.
http://www.mtc.com.my/statistic/stat04/2-3.html Malaysian Timber Council. Last
visit 1 July 2004. 16:29
Hungerford, R.D., Frandsen, W.H. and Ryan, K.C. (1999). Heat Transfer Into The
Duff And Organic Soil. Final Project Report, USDA Forest Service,
Intermountain Research Station, Montana.
118
Hussin, Y. A. (2000). State of The Art: Remote Sensing Applications For
Sustainable Management of Forests. Proceeding of the 20th ACRS
Conference.
Jason, J. B. (2003). The Application of Remote Sensing and GIS for the
Development of a Fire Mapping Database. Penulisan Sarjana Muda,
Universiti Teknologi Malaysia.
Jensen, J.R., (2000). Remote Sensing of the Environment, An Earth Resource
Perspective.
Johan Suhaili and Mohd Zulkifli Mohd Yunus (1999). “Peat Forest Fire”.
Makalah Seminar Kejuruteraan Awam.
Johnson, R.A. (1999). Guidance On Measures In Forest Fire Emergency Cases.
Paper presented in Health Guidelines for Vegetation Fire Events, Lima,
Peru.
Kallidromitou, D., Bonazountas, M., Papachristou, P. and Caballero, D. (2000).
Forest Fire Management and Fire Prevention System. DELFI Int.
Symposium Forest Fire Need and Innovations: 93-98.
Kaufman, Y. and Justice, C. (1998). Algorithm Technical Background Document:
MODIS Fire Products.
Kaufman, Y. J., Justice, C., Flynn, L., Kendall, J., Prins, E., Giglio, L., Darold
E.W., Menzel, P., and Setzer, A., (2000). Potential Global Fire Monitoring
from EOS-MODIS.
Kaufman, Y.J., Setzer, A., Justice, C., Tucker, C.J., Pereira, M.C. and Fung, I.,
(1990). Remote Sensing of Biomass Burning in the Tropics. Journal of
Geophysical Research, 95 : 9927-9939.
Keetch, J. J. and Byram, G. M., (1968). A Drought Index for Forest Fire Control.
U.S. Department of Agriculture-Forest Service Research Paper, Southeastern
Forest Experiment Station Ashville, North Carolina.
119
Khairuddin Kamaruddin (1992). Sbar dari Utusan Mairi Alam dan Rimba No.3:
Hutan Hujan Tropika Semenanjung Malaysia. Institut Penyelidikan
Perhutanan Malaysia.
Khali A. H. (2001). Remote Sensing, GIS and GPS As A Tool To Support
Precision Forestry Practices In Malaysia. 22nd Asian Conference On Remote
Sensing.
Kidwell, K.B. Editor (1991). NOAA Polar Orbital Data User’s Guide Washington,
B.C. NOAA/Nesdis.
King, M.D., Kaufman, Y.J., Menzel, W.P. and Tanre, D. (1992). Remote Sensing
of Cloud, Aerosol and Water Vapor Properties From the Moderate
Resolusion Imaging Spectroradiometer (MODIS). IEEE Trans. Geoscience
Renote Sensing. Vol 30, 2-27.
Kogan, F. N. (2000a). Global Monitoring Fire Potential From Operational
Satellites. Food and Agriculture Organization (FAO) Meeeting on Public
Policies Affecting Forest Fires.
Kogan, F. N. (2001). Operational Space Technology for Global Vegetation
Assessment. Bulletin of the American Meteorology Society, Vol 82, No. 9,
Pages 1949-1964.
Kogan, F. N., Gitelson, A., Zakarin, E., Spivak, L. and Lebed, L. (2003).
AVHRR-Based Spectral Vegetation Index for Quantitative Assessment of
Vegetation
State
and
Productivity:
Calibration
and
Validation.
Photogrammetric Engineering & Remote Sensing, Vol. 69, No. 8, Pages
899-906.
Kogan, F.N. (2000b). Contribution of Remote Sensing to Drought Early Warning.
Available at http://www.svs.gsfc.nasa.gov 2/5/2003, 3.45p.m.
Kogan, F.N. (2002). World Droughts in the New Milennium from AVHRR-based
Vegetation Health Indices. Eos, Transactions, American Geophysical Union,
Vol. 83, No 48, Pages 557, 562-563.
Kueh, H.S. and Tie, Y.L., (Depart. Agriculture, Kuching, Sarawak, 1989). “Report
on Research, Development and Extension Activities on Peat Soils in
120
Sarawak.” Bengkel Kebangsaan Penyelidikan dan Pembangunan Tanah
Gambut.
Lakshmi, V. and Sussking, J. (1997). Land Surface Hydrological Processes Using
Satellite Data. IEEE International. Vol 3,.1102-1104
Langhart, R., Bachmann, A. and Allgower, B. (1998). Temporal And Spatial
Patterns of Wildfire Occurrence (Canton of Grisons, Switzerland). In
Proceedings of the III International Conference on Forest Fire Research,
14th Conference on Fire and Forest Meteorology. Coimbra, Portugal. Vol. 2,
pp 2279-2292.
Lee, T.F. and Tag, P.M. (1990). Improved Detection of Hotspots Using the
AVHRR 3.7μm Channel, Bulletin of the American Meterology Society, 71
(12):1722-1730.
Leicester, L.H., (1995). Proceedings National Workshop On Fire and Timber In
Modern Building Design. “Fire Codes And Timber Structures In Australia.”
Forest Research Institute Malaysia (FRIM).
Leone, V. and Lovreglio, R. Human Fire Causes: A Challenge For Modeling. 4th
International Workshop On Remote Sensing And GIS Applications To Forest
Fire Management: Innovative Concepts And Methos In Fire Danger
Estimation; June 5-7, 2003. Chuvieco, E., Martin, P. and Justice, C. Ghent
University. Pg 90-99.
Levine, J.S., Bobbe, T., Ray, N., Singh, A. and R.G. Witt (1999). Wildland Fires
and The Environment: A Global Synthesis. Environment Information And
Assessment Technical Report. United Nation Environment Programme
(UNEP).
Liew, S.C., Kwoh, L.K., and Lim, H. (1998). A Study of the 1997 Forest Fires in
South East Asia Using SPOT Quicklook Mosaics. IGARSS ’98, Seattle.
Lim, J.C.. (1995). Proceedings National Workshop On Fire and Timber In Modern
Building Design. Building Regulations and Timber Construction in
Malaysia. Forest Research Institute Malaysia (FRIM).
121
Lorimer, C.G. (2003). Introduction to Forest Ecosystem Science and
Management. John Wiley & Sons, 3rd Edition. Page 389-411. (Buku PSZ)
Masuoka, E., Fleig, A., Wolfe, R.E. and Patt, P. (1998). Key Characteristics of
MODIS Data Products. IEEE Transactions On Geoscience and Remote
Sensing. Vol 36, No 4, 1313-1323.
Matson, M. and Dozier, J. (1981). Identification of Subresolution High
Temperature Sources Using A Thermal IR Sensor, Photogrammetry Eng.
And Remote Sensing, 47(9):1311-1318.
Matson, M. and Holben, B. (1987). Satellite Detection of Tropical Burning In
Brazil. International Journal Of Remote Sensing, 8(3):509-516.
Mazlan Hashim, Peter Sim Hock Seng, Wan Hazli Wan Kadir and Ngap Dollah
Salam (2001). Soil Moisture Estimation of Drained Peat Area Using
TOPSAR Data. Malaysian Journal of Remote Sensing & GIS. Vol. 2 pp 5161
McKee, T.B., Doesken, N.J., and Kleist, J., (1993). The Relationship of Drought
Frequency and Duration To Time Scales. 8th Conference on Applied
Climatology.
Mohd Shahwahid H.O. (2004). Symposium of Forest Certification In Malaysia.
Forest Certification in Developing and Transitioning Societies: Social,
Economic, and Ecological Effects.
Mulders, M.A., (1987). Remote Sensing In Soil Science. Elsevier Science
Publishers B.V. (S 59
Mulyanto Darmawan, Masamu Aniya and Satoshi Tsuyuki (2001). Forest Fire
Hazard Model Using Remote Sensing and Geographic Information System :
Toward Understanding of Land and Forest Degradation in Lowland Ares of
East Kalimantan, Indonesia. 22nd Asian Conference on Remote Sensing.
Muzy, A., Innoceti, E., Santucci, J. F. and Hill, R.C. (2003). Optimization of Cell
Spaces Simulation for the Modeling of Fire Spreading. Proceeding of the
36th Annual Simulation Symposium (ANSS’03).
122
Ng Wei Loon (2005). Firefighters Pray for the Rain to End Blaze. The Star
Online. Last visited 25 February 2005
Nidzwan Zainal Abidin (2005). Kering Kontang. Utusan Malaysia Online.
http://www.utusan.com.my . Last visited 8 March 2005.
Palmer, W.C. (1965). Meteorological Drought Research Paper No. 45, U.S.
Department of Commerce Weather Bereau, Washington D.C.
Payette, S., (1989). Fire As A Controlling Process In The North American Boreal
Forest. Pages 145-169. A System Analysis of the Global Boreal Forest.
Pearce, D., Francis E. P. and Vanclay, J. K. (2003). “Sustainable Forestry in the
Tropics: Panacea or Folly?” Forest Ecology and Management 172 (2003)
229-247.
Prins, E.M. and Menzel, W.P. (1993). Geostationary Satellite Detection Of
Biomass Burning in South America. International Journal of Remote
Sensing, 13:2783-2799.
Radjagukguk, B. (2000). “onversion of Peat Soil Chemicophysical Properties
Caused By Peat Soil Reclamation Used for Agriculture. Jurnal Ilmu Tanah
dan Lingkungan. Vol 2(1): 1-15.
Rutherford, M. C., (1981). Survival, Regeneration and Leaf Biomass Changes in
Woody Plants Following Spring Burns in Burkea Africana-Ocnha Pulchra
Savanna.
Sanchez, I. (2001). Fire Related Models Characteristics And recommendations on
Models Standardisation”. A document for Cluster Initiative for Flood and
Fire Emergencies (CLIFF).
San-Miguel-Ayanz, J. (1999). Methodologies for the Evaluation of Forest Fire
Risk: From Long-Term (Static) to Dynamic Indices. Available at
http://natural-hazards.jrc.it/fires/. Last visited on 13/4/2003, 9.40 a.m.
Scott, J.H. and Reinhardt, E.D. (2001). Assessing Crown Fire Potential By
Linking Models of Surface and Crown Fire Behavior. Research Paper
123
RMRS-RP-29. Fort Collins, CO. U.S. Department of Agriculture, Forest
Services, Rocky Mountain Research Station.
Shen, C., Liew, S.C., Kwoh, L.K., (2000). Spatial And Temporal Pattern Of
Forest/Plantation Fires In Riau, Sumatera From 1998 To 2000, 22nd Asian
Conference on Remote Sensing.
Siegert, F., Rucker, G. and Hoffmann, A., (2000). Evaluation of the 1998 Forest
Fires in East-Kalimantan (Indonesia) Using NOAA-AVHRR Hotspot Data
and Multitemporal ERS-2 SAR Images.
Singh, P.R., Roy, S. Kogan, K. (2003). Vegetation and Temperature Condition
Indices From NOAA AVHRR Data For Drought Monitoring Over India. Int
J. Remote Sensing, Vol. 24, No 22, 4393-4402.
Skvarenina, J., Mindas, J., Holecy, J. and Tucek, J. (2003). Forest Fire Weather
Indices During Two Largest Forest Fire Events In The Slovak Paradise
National Park. Paper presented in International Bioclimatological Workshop
2003,
Slovakia,
2-4
September
2003.
Available
online
at
http://www.fzki.uniag.sk/oldweb/BPD2003/. Last visited at 22 February
2004, 9.45 a.m..
Spano, D. (2003). A Risk Index for Mediterranean Vegetation Based on
Micrometeorological and Ecophysiological Measurements. 5th Symposium
on Fire and Forest Meteorology.
Stocks, B. J. and Duchesne, L. C. (1998). Forest Fires and Forest Fires and Forest
Fire Research in Canada. Canadian Forest Service.
Sullivan, W. T., (1989). A 10km Resolution Image of the Entire Night-time Earth
Based on Cloud-free satellite Photographs in the 400-1100 nm Band.
International Journal of Remote Sensing, 10(1): 1-5.
Suraiya Ahmad, Ouzounov, D., Leptoukh, G., Serafino, G. and Kempler, S.
(2002). Highlights of MODIS Products. IEEE Transactions On Geoscience
and Remote Sensing. 2299-2301.
124
Taylor, S. W. and Alexander, M. E. (2006). Science, Technology and Human
Factors In Fire Danger Rating: The Canadian Experience. Int Journal of
Wildland Fire, 15, 121-135.
Tie, Y. L. and Kueh, H. S. (1979). A Review of Lowland Organic Soils of
Sarawak: Technical Paper No. 4. Research Branch Department of
Agriculture Sarawak.
Torrence, C. and Webster, P. J. (2000). The Connection Between the Boreal
Spring Oscillation Persistence Barrier and Biennial Variability. J. Climate
12, 610-620.
Wan Mohd Shukri Wan Ahmad (2001). Forest Fire Situation in Malaysia.
International Forest Fire News (IFFN) No. 26: 66-74.
Watson, K., Bloms, R., Barbo, M., Bahr, D., Casiano, T., Barborinas, J., Biggs, T.,
Anderson, D., Shelley, K., Guilbert, P. (2000). Weather Station Standards. A
Publication of the National Wildfire Coordinating Group, Fire Weather
Working Team Weather Station Standards Task Group, United State
Department of Agriculture.
Xiong, Xiaoxiong, Chiang, K., Esposito, J., Guenther, B. and Barnes, W. (2003).
MODIS On-orbit Calibration and Characterization. Metrologica. Vol 40, 8992.
Zhou, W., Zhou, Y., Wang, S. and Zhao, Q. (2003). Early Warning For Grassland
Fire Danger In North China Using Remote Sensing. IEEE International
Volume 4, Pages :2505-2507.
125
APPENDIX A
Forest Fire Record From FRDM
Fire
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
Date
18/01/2005
18/01/2005
03/02/2005
03/02/2005
03/02/2005
03/02/2005
10/02/2005
10/02/2005
10/02/2005
10/02/2005
10/02/2005
10/02/2005
10/02/2005
10/02/2005
10/02/2005
10/02/2005
10/02/2005
10/02/2005
10/02/2005
10/02/2005
12/02/2005
12/02/2005
14/02/2005
02/02/2005
24/02/2005
24/02/2005
26/02/2005
26/02/2005
27/02/2005
28/02/2005
01/03/2005
03/03/2005
03/03/2005
05/03/2005
05/03/2005
05/03/2005
05/03/2005
05/03/2005
05/03/2005
05/03/2005
07/03/2005
Location
Kubang Pasu, Kedah
Kuala Selangor, Selangor
Kubang Pasu, Kedah
Kota Setar, Kedah
Pendang, Kedah
Kuala Muda, Kedah
Kubang Pasu, Kedah
Kota Setar, Kedah
Manjung, Perak
Perak Tengah, Perak
Kinta, Perak
Perak Tengah, Perak
Perak Tengah, Perak
Batang Berjuntai, Selangor
Kuala Selangor, Selangor
Kuantan, Pahang
Muar, Johor
Muar, Johor
Kluang, Johor
Kluang, Johor
Kota Setar, Kedah
Perak Tengah, Perak
Kota Tinggi, Johor
Perak Tengah, Perak
Batang Berjuntai, Selangor
Rompin, Pahang
Batang Berjuntai, Selangor
Rompin, Pahang
Batang Berjuntai, Selangor
Batang Berjuntai, Selangor
Perak Tengah, Perak
Kota Setar, Kedah
Perak Tengah, Perak
Pendang, Kedah
Perak Tengah, Perak
Kuantan, Pahang
Pekan, Pahang
Pekan, Pahang
Rompin, Pahang
Segamat, Johor
Kuala Terengganu, Terengganu
Source
Farmer activity
Bush fire
Farmer activity
Farmer activity
Farmer activity
Farmer activity
Farmer activity
Farmer activity
Bush fire
Bush fire
Forest fire
Forest fire
Forest fire
Peat fire
Peat fire
Forest fire
Bush fire
Bush fire
Bush fire
Farmer activity
Farmer activity
Forest fire
Bush fire
Forest fire
Peat fire
Peat fire
Peat fire
Peat fire
Peat fire
Peat fire
Forest fire
Farmer activity
Forest fire
Farmer activity
Forest fire
Peat fire
Peat fire
Peat fire
Peat fire
Bush fire
Peat fire
126
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
07/03/2005
09/03/2005
21/03/2005
21/03/2005
21/03/2005
01/04/2005
08/04/2005
08/04/2005
12/04/2005
13/04/2005
13/04/2005
13/04/2005
15/04/2005
15/04/2005
15/04/2005
15/04/2005
15/04/2005
17/04/2005
17/04/2005
20/04/2005
05/03/2005
22/04/2005
22/04/2005
22/04/2005
Kemaman, Terengganu
Kota Tinggi, Johor
Baling, Kedah
Perak Tengah, Perak
Perak Tengah, Perak
Pekan, Pahang
Pekan, Pahang
Kluang, Johor
Pontian, Johor
Sik, Kedah
Hulu Perak, Perak
Gua Musang, Kelantan
Seberang Prai, Pulau Pinang
Kinta, Perak
Gua Musang, Kelantan
Kuantan, Pahang
Kuala Langat, Selangor
Kuantan, Pahang
Kota Tinggi, Johor
Manjung, Perak
Penang
Kuala Langat, Selangor
Segamat, Johor
Segamat, Johor
Peat fire
Bush fire
Farmer activity
Forest fire
Forest fire
Bush fire
Bush fire
Bush fire
Bush fire
Bush fire
Bush fire
Bush fire
Bush fire
Forest fire
Bush fire
Forest fire
Peat fire
Bush fire
Bush fire
Forest fire
Urban fire
Peat fire
Bush fire
Bush fire
127
APPENDIX B
Fire risk level for each forest fire occurrence in time range of 1 January 2005 to 30
April 2005.
Fire
Date
Latitude Longitude
Location
1 18/01/2005
6.508
100.392 Kubang Pasu, Kedah
2 18/01/2005
3.41
101.38
Kuala Selangor,
Selangor
3 03/02/2005
6.33
100.25
Kubang Pasu, Kedah
4 03/02/2005
6.01
100.38
Kota Setar, Kedah
5 03/02/2005
5.98
100.41
Pendang, Kedah
6 03/02/2005
5.73
100.64
Kuala Muda, Kedah
7 10/02/2005
6.22
100.42
Kubang Pasu, Kedah
8 10/02/2005
6.01
100.39
Kota Setar, Kedah
9 10/02/2005
4.35
100.67
Manjung, Perak
10 10/02/2005
4.31
101.01
Perak Tengah, Perak
11 10/02/2005
4.31
101.02
Kinta, Perak
12 10/02/2005
4.18
100.97
Perak Tengah, Perak
13 10/02/2005
4.17
101.01
Perak Tengah, Perak
14 10/02/2005
3.48
101.40
Batang Berjuntai,
Selangor
15 10/02/2005
3.46
101.44
Kuala Selangor,
Selangor
16 10/02/2005
3.94
103.13
Kuantan, Pahang
17 10/02/2005
2.27
102.83
Muar, Johor
18 10/02/2005
2.03
102.73
Muar, Johor
19 10/02/2005
2.33
103.33
Kluang, Johor
20 10/02/2005
2.07
103.37
Kluang, Johor
21 12/02/2005
6.18
100.31
Kota Setar, Kedah
22 12/02/2005
4.18
100.97
Perak Tengah, Perak
23 14/02/2005
1.38
104.16
Kota Tinggi, Johor
24 02/02/2005
4.18
100.97 Perak Tengah, Perak
25 24/02/2005
3.48
101.4
Batang Berjuntai,
Selangor
26 24/02/2005
2.55
103.59
Rompin, Pahang
27 26/02/2005
3.48
101.4
Batang Berjuntai,
Selangor
28 26/02/2005
2.55
103.59
Rompin, Pahang
29 27/02/2005
3.48
101.4
Batang Berjuntai,
Selangor
30 28/02/2005
3.48
101.4
Batang Berjuntai,
Selangor
31 01/03/2005
4.18
100.97
Perak Tengah, Perak
32
03/03/2005
5.98
100.40
Kota Setar, Kedah
Fire Risk Level
High
High
High
Low
Moderate
High
High
Moderate
High
High
Low
High
High
Low
High
High
High
High
High
High
High
Moderate
High
High
High
High
High
Moderate
High
Low
High
High
128
33
34
35
36
37
38
39
40
41
03/03/2005
05/03/2005
05/03/2005
05/03/2005
05/03/2005
05/03/2005
05/03/2005
05/03/2005
07/03/2005
4.18
5.98
4.18
3.62
3.61
3.42
2.55
2.44
5.46
100.97
100.49
100.97
103.26
103.29
103.06
103.59
102.82
102.96
42
43
44
45
46
07/03/2005
09/03/2005
21/03/2005
21/03/2005
21/03/2005
4.55
1.81
5.67
4.50
4.41
103.45
104.12
100.69
100.93
100.89
47
48
49
50
51
52
53
54
01/04/2005
08/04/2005
08/04/2005
12/04/2005
13/04/2005
13/04/2005
13/04/2005
15/04/2005
3.55
3.53
2.19
1.66
5.81
5.31
4.74
5.28
103.27
103.39
103.52
103.38
100.68
101.06
102.01
100.52
55
56
15/04/2005
15/04/2005
4.45
4.84
101.16
101.97
57
58
15/04/2005
15/04/2005
3.93
2.81
103.14
101.59
59
17/04/2005
3.84
103.22
60
61
62
63
17/04/2005
20/04/2005
05/03/2005
22/04/2005
1.82
4.53
5.32
2.81
103.66
100.77
100.29
101.60
64
65
22/04/2005
22/04/2005
2.66
2.66
102.79
102.80
Perak Tengah, Perak
Pendang, Kedah
Perak Tengah, Perak
Kuantan, Pahang
Pekan, Pahang
Pekan, Pahang
Rompin, Pahang
Segamat, Johor
Kuala Terengganu,
Terengganu
Kemaman, Terengganu
Kota Tinggi, Johor
Baling, Kedah
Perak Tengah, Perak
Perak Tengah, Perak
High
Moderate
High
High
High
High
High
Moderate
Moderate
Low
High
High
Moderate
Moderate
Pekan, Pahang
Pekan, Pahang
Kluang, Johor
Pontian, Johor
Sik, Kedah
Hulu Perak, Perak
Gua Musang, Kelantan
Seberang Prai, Pulau
Pinang
Kinta, Perak
Gua Musang, Kelantan
High
High
High
High
Moderate
High
High
High
Kuantan, Pahang
Kuala Langat,
Selangor
Kuantan, Pahang
High
High
Kota Tinggi, Johor
Manjung, Perak
Penang
Kuala Langat,
Selangor
Segamat, Johor
Segamat, Johor
High
High
Moderate
Moderate
High
Low
High
High
Moderate
129
APPENDIX C
Visual BasicC++ source code for Window 1
Coding for main window
#include <QHBoxLayout>
#include <QStatusBar>
#include <QGroupBox>
#include <QPushButton>
#include <QMenuBar>
#include <QMenu>
#include <QAction>
#include <QFrame>
#include <QDialog>
#include <QProcess>
#include <QDir>
#include <QIcon>
#include <QMessageBox>
#include <QFile>
#include <QByteArray>
#include "MainWindow.h"
#include "WindowCalc.h"
MainWindow::MainWindow(QWidget *parent) : QMainWindow(parent)
{
QIcon icon("system\\icons\\earth.jpg");
this->setWindowIcon(icon);
frame = new QFrame();
frame->setFrameStyle(QFrame::Panel);
this->setCentralWidget(frame);
this->setWindowTitle("GIS-UTM Remote Sensing Software v1.0");
this->setGeometry(50,50,500,100);
130
statusBar = new QStatusBar(this);
this->setStatusBar(statusBar);
statusBar->showMessage("Application Ready");
menuBar = new QMenuBar();
menu_Text_File = new QMenu("File");
menu_Text_Edit = new QMenu("Edit");
act_Exit = new QAction("Exit",menu_Text_File);
connect(act_Exit,SIGNAL(triggered()),this,SLOT(close()));
act_About = new QAction("Help",menu_Text_Edit);
connect(act_About,SIGNAL(triggered()),this,SLOT(slot_about()));
menu_Text_File->addAction(act_Exit);
menu_Text_Edit->addAction(act_About);
menuBar->addMenu(menu_Text_File);
menuBar->addMenu(menu_Text_Edit);
this->setMenuBar(menuBar);
QVBoxLayout *layout_000 = new QVBoxLayout();
QGroupBox *groupBox = new QGroupBox("Main Function");
btn_m_vege = new QPushButton("Vegetation Index");
QObject::connect(btn_m_vege,SIGNAL(clicked()),this,SLOT(slot_vege()));
btn_m_hotsp = new QPushButton("Hotspot");
QObject::connect(btn_m_hotsp,SIGNAL(clicked()),this,SLOT(slot_hotsp()));
131
btn_m_prox = new QPushButton("Proximity Analysis");
QObject::connect(btn_m_prox,SIGNAL(clicked()),this,SLOT(slot_prox()));
QHBoxLayout *layout_001 = new QHBoxLayout();
layout_001->addWidget(btn_m_vege);
layout_001->addWidget(btn_m_hotsp);
layout_001->addWidget(btn_m_prox);
groupBox->setLayout(layout_001);
layout_000->addWidget(groupBox);
frame->setLayout(layout_000);
}
void MainWindow::slot_vege()
{
WindowCalc *winCalc = new WindowCalc(this);
winCalc->show();
}
void MainWindow::slot_hotsp()
{
QDialog *dialog = new QDialog(this);
dialog->setWindowTitle("Hotspot Index");
dialog->setGeometry(60,60,200,100);
QVBoxLayout *layout_001 = new QVBoxLayout(dialog);
QPushButton *btn_hotsp = new QPushButton("Hotspot",dialog);
QObject::connect(btn_hotsp,SIGNAL(clicked()),this,SLOT(slot_dlg_hotsp()));
QPushButton *btn_convert = new QPushButton("Convert",dialog);
QObject::connect(btn_convert,SIGNAL(clicked()),this,SLOT(slot_dlg_convert())
);
132
QPushButton *btn_close = new QPushButton("Close",dialog);
QObject::connect(btn_close,SIGNAL(clicked()),dialog,SLOT(close()));
layout_001->addWidget(btn_hotsp);
layout_001->addWidget(btn_convert);
layout_001->addWidget(btn_close);
dialog->setLayout(layout_001);
dialog->exec();
}
void MainWindow::slot_dlg_hotsp()
{
//QProcess *proc = new QProcess(this);
//QString program = "cmd.exe";
//QDir dir = "";
//QString str = dir.currentPath();
//QStringList arg;
//arg << "/k" << "dir mod*";
//proc->startDetached(program,arg);
QString line;
QFile file("system\\configs\\config.txt");
if (!file.open(QIODevice::ReadOnly | QIODevice::Text))
{
QMessageBox::warning(this,"Dialog - Error Opening config.txt",
"Where is the config.txt file?");
}
else
{
while (!file.atEnd())
{
line = file.readLine();
133
if ( line.contains("cmd=",Qt::CaseInsensitive))
{
line.remove(" ",Qt::CaseInsensitive);
line.remove("cmd=",Qt::CaseInsensitive);
break;
}
}
QProcess *proc = new QProcess(this);
QString program = line;
QDir dir = "";
QString str = dir.currentPath();
QStringList arg;
arg << "/k" << "cd system\\modules & dir mod*";
proc->startDetached(program,arg);
if (proc->isOpen() == 2)
{
QMessageBox::warning(this,"Dialog - Error Opening cmd.exe",
"Please set the correct path to cmd
by modifying config.txt file");
}
}
}
void MainWindow::slot_dlg_convert()
{
//QProcess *proc = new QProcess(this);
//QString program = "cmd.exe";
//QDir dir = "";
//QString str = dir.currentPath();
134
//QStringList arg;
//arg << "/k" << "dir mod*";
//proc->startDetached(program,arg);
QString line;
QFile file("system\\configs\\config.txt");
if (!file.open(QIODevice::ReadOnly | QIODevice::Text))
{
QMessageBox::warning(this,"Dialog - Error Opening config.txt",
"Where is the config.txt file?");
}
else
{
while (!file.atEnd())
{
line = file.readLine();
if ( line.contains("cmd=",Qt::CaseInsensitive))
{
line.remove(" ",Qt::CaseInsensitive);
line.remove("cmd=",Qt::CaseInsensitive);
break;
}
}
QProcess *proc = new QProcess(this);
QString program = line;
QDir dir = "";
QString str = dir.currentPath();
QStringList arg;
arg << "/k" << "cd system\\modules & dir mod*";
proc->startDetached(program,arg);
135
if (proc->isOpen() == 2)
{
QMessageBox::warning(this,"Dialog - Error Opening cmd.exe",
"Please set the correct path to cmd
by modifying config.txt file");
}
}
}
void MainWindow::slot_prox()
{
QString line;
QFile file("system\\configs\\config.txt");
if (!file.open(QIODevice::ReadOnly | QIODevice::Text))
{
QMessageBox::warning(this,"Dialog - Error Opening config.txt",
"Where is the config.txt file?");
}
else
{
while (!file.atEnd())
{
line = file.readLine();
if ( line.contains("arcview=",Qt::CaseInsensitive))
{
line.remove(" ",Qt::CaseInsensitive);
line.remove("arcview=",Qt::CaseInsensitive);
break;
}
}
136
QProcess *proc = new QProcess(this);
QString program = line;
proc->start(program);
if (proc->state() == 0)
{
QMessageBox::warning(this,"Dialog
-
Error
Opening
ARCVIEW.exe",
"Please set the correct path to
ARCVIEW by modifying config.txt file");
}
}
}
void MainWindow::slot_about()
{
QMessageBox::information(this,"GIS UTM Program - About",
"Version\:\n 1.0\n\nDeveloper\:\nAida Hayati M Hassan ");
}
Coding for NDVI
#include <QtGui>
#include "Dialog_Ndvi.h"
137
Dialog_Ndvi::Dialog_Ndvi()
{
}
Dialog_Ndvi::Dialog_Ndvi(int ind)
{
title = "NDVI File Input Dialog";
list << "channel 1" << "channel 2" << "output";
browse = ind;
init();
}
void Dialog_Ndvi::init()
{
this->setWindowTitle(title);
this->setGeometry(60,60,400,300);
QGridLayout *layout_000 = new QGridLayout(this);
label[0] = new QLabel(list[0]);
label[1] = new QLabel(list[1]);
label[2] = new QLabel(list[2]);
lineedit[0] = new QLineEdit;
lineedit[1] = new QLineEdit;
lineedit[2] = new QLineEdit;
btn[0] = new QPushButton("browse");
QObject::connect(btn[0],SIGNAL(clicked()),this,SLOT(slot_browse_000()));
btn[1] = new QPushButton("browse");
QObject::connect(btn[1],SIGNAL(clicked()),this,SLOT(slot_browse_001()));
138
btn[2] = new QPushButton("browse");
QObject::connect(btn[2],SIGNAL(clicked()),this,SLOT(slot_browse_002()));
btn_ok = new QPushButton("Process");
QObject::connect(btn_ok,SIGNAL(clicked()),this,SLOT(slot_ok()));
btn_cancel = new QPushButton("Cancel");
QObject::connect(btn_cancel,SIGNAL(clicked()),this,SLOT(close()));
layout_000->addWidget(label[0],0,0);
layout_000->addWidget(label[1],1,0);
layout_000->addWidget(label[2],2,0);
layout_000->addWidget(lineedit[0],0,1,1,4);
layout_000->addWidget(lineedit[1],1,1,1,4);
layout_000->addWidget(lineedit[2],2,1,1,4);
layout_000->addWidget(btn[0],0,6,1,1);
layout_000->addWidget(btn[1],1,6,1,1);
layout_000->addWidget(btn[2],2,6,1,1);
layout_000->addWidget(btn_cancel,4,4);
layout_000->addWidget(btn_ok,4,6);
this->setLayout(layout_000);
}
void Dialog_Ndvi::slot_ok()
{
QString str;
QByteArray ba;
str = lineedit[0]->text();
read(str,0);
str = lineedit[1]->text();
139
read(str,1);
process();
str = lineedit[2]->text();
write(str);
this->close();
}
void Dialog_Ndvi::slot_browse_000()
{
QString fileName = QFileDialog::getOpenFileName(this, tr("Open File"),"");
lineedit[0]->setText(fileName);
QImage temp(fileName);
pixel = matrix3d(browse,matrix2d(temp.width(),temp.height()));
outp = matrix2d(temp.width(),matrix1d(temp.height()));
}
void Dialog_Ndvi::slot_browse_001()
{
QString fileName = QFileDialog::getOpenFileName(this, tr("Open File"),"");
lineedit[1]->setText(fileName);
}
void Dialog_Ndvi::slot_browse_002()
{
QString
File"),"","*.tif");
fileName
=
QFileDialog::getSaveFileName(this,
tr("Save
140
lineedit[2]->setText(fileName);
}
void Dialog_Ndvi::read(QString filename,int inp)
{
image = new QImage(filename);
QRgb rgb;
for (int i=0; i<(image->width()); i++)
{
for (int j=0; j<(image->height()); j++)
{
rgb
= image->pixel(i,j);
pixel[inp][i][j] = qGray(rgb);
maxh
= j;
}
maxw = i;
}
maxh++;
maxw++;
}
void Dialog_Ndvi::process()
{
QString str;
for (int i=0;i<(image->width()); i++)
{
for (int j=0; j<(image->height()); j++)
{
outp[i][j]
=
((pixel[1][i][j]-pixel[0][i][j])
(pixel[1][i][j]+pixel[0][i][j]));
outp[i][j] = (int) (100 * (outp[i][j]) + 1);
/
141
}
}
}
void Dialog_Ndvi::write(QString str)
{
image = new QImage(maxw,maxh,QImage::Format_RGB32);
QRgb value;
for (int i=0; i<maxw; i++)
{
for (int j=0; j<maxh; j++)
{
value = qRgb(outp[i][j],outp[i][j],outp[i][j]);
image->setPixel(i,j,value);
}
}
if(! image->save(str,"TIFF",100) )
qWarning("Save Failed!");
else
{
strout = str;
}
}
Coding for VCI
#include <QtGui>
#include "Dialog_Vci.h"
142
Dialog_Vci::Dialog_Vci()
{
}
Dialog_Vci::Dialog_Vci(int ind)
{
title = "VCI File Input Dialog";
list << "NDVI" << "NDVI Min" << "NDVI Max" << "Output";
browse = ind;
init();
}
void Dialog_Vci::init()
{
this->setWindowTitle(title);
this->setGeometry(60,60,400,300);
QGridLayout *layout_000 = new QGridLayout(this);
label[0] = new QLabel(list[0]);
label[1] = new QLabel(list[1]);
label[2] = new QLabel(list[2]);
label[3] = new QLabel(list[3]);
lineedit[0] = new QLineEdit;
lineedit[1] = new QLineEdit;
lineedit[2] = new QLineEdit;
lineedit[3] = new QLineEdit;
btn[0] = new QPushButton("browse");
QObject::connect(btn[0],SIGNAL(clicked()),this,SLOT(slot_browse_000()));
btn[1] = new QPushButton("browse");
143
QObject::connect(btn[1],SIGNAL(clicked()),this,SLOT(slot_browse_001()));
btn[2] = new QPushButton("browse");
QObject::connect(btn[2],SIGNAL(clicked()),this,SLOT(slot_browse_002()));
btn[3] = new QPushButton("browse");
QObject::connect(btn[3],SIGNAL(clicked()),this,SLOT(slot_browse_003()));
btn_ok = new QPushButton("Process");
QObject::connect(btn_ok,SIGNAL(clicked()),this,SLOT(slot_ok()));
btn_cancel = new QPushButton("Cancel");
QObject::connect(btn_cancel,SIGNAL(clicked()),this,SLOT(close()));
layout_000->addWidget(label[0],0,0);
layout_000->addWidget(label[1],1,0);
layout_000->addWidget(label[2],2,0);
layout_000->addWidget(label[3],3,0);
layout_000->addWidget(lineedit[0],0,1,1,4);
layout_000->addWidget(lineedit[1],1,1,1,4);
layout_000->addWidget(lineedit[2],2,1,1,4);
layout_000->addWidget(lineedit[3],3,1,1,4);
layout_000->addWidget(btn[0],0,6,1,1);
layout_000->addWidget(btn[1],1,6,1,1);
layout_000->addWidget(btn[2],2,6,1,1);
layout_000->addWidget(btn[3],3,6,1,1);
layout_000->addWidget(btn_cancel,4,4);
layout_000->addWidget(btn_ok,4,6);
this->setLayout(layout_000);
}
144
void Dialog_Vci::slot_ok()
{
QString str;
QByteArray ba;
str = lineedit[0]->text();
read(str,0);
str = lineedit[1]->text();
read(str,1);
str = lineedit[2]->text();
read(str,2);
process();
str = lineedit[3]->text();
write(str);
this->close();
}
void Dialog_Vci::slot_browse_000()
{
QString fileName = QFileDialog::getOpenFileName(this, tr("Open File"),"");
lineedit[0]->setText(fileName);
QImage temp(fileName);
pixel = matrix3d(browse,matrix2d(temp.width(),temp.height()));
outp = matrix2d(temp.width(),matrix1d(temp.height()));
}
void Dialog_Vci::slot_browse_001()
145
{
QString fileName = QFileDialog::getOpenFileName(this, tr("Open File"),"");
lineedit[1]->setText(fileName);
}
void Dialog_Vci::slot_browse_002()
{
QString fileName = QFileDialog::getOpenFileName(this, tr("Open File"),"");
lineedit[2]->setText(fileName);
}
void Dialog_Vci::slot_browse_003()
{
QString
fileName
=
QFileDialog::getSaveFileName(this,
File"),"","*.tif");
lineedit[3]->setText(fileName);
}
void Dialog_Vci::read(QString filename,int inp)
{
image = new QImage(filename);
QRgb rgb;
for (int i=0; i<(image->width()); i++)
{
for (int j=0; j<(image->height()); j++)
{
rgb
= image->pixel(i,j);
pixel[inp][i][j] = qGray(rgb);
maxh
}
maxw = i;
= j;
tr("Save
146
}
maxh++;
maxw++;
}
void Dialog_Vci::process()
{
for (int i=0;i<(image->width()); i++)
{
for (int j=0; j<(image->height()); j++)
{
if ((pixel[2][i][j]-pixel[1][i][j]) ==0)
{
outp[i][j] = 0;
}
else
{
outp[i][j] = ( (pixel[0][i][j]-pixel[1][i][j]) / (pixel[2][i][j]pixel[1][i][j]) );
outp[i][j] = (int) (100 * (outp[i][j]));
}
}
}
}
void Dialog_Vci::write(QString str)
{
image = new QImage(maxw,maxh,QImage::Format_RGB32);
QRgb value;
for (int i=0; i<maxw; i++)
{
147
for (int j=0; j<maxh; j++)
{
value = qRgb(outp[i][j],outp[i][j],outp[i][j]);
image->setPixel(i,j,value);
}
}
if(! image->save(str,"TIFF",100) )
qWarning("Save Failed!");
else
{
strout = str;
}
}
Coding For TCI
#include <QtGui>
#include "Dialog_Tci.h"
Dialog_Tci::Dialog_Tci()
{
}
Dialog_Tci::Dialog_Tci(int ind)
{
title = "TCI File Input Dialog";
list << "BT" << "BT Min" << "BT Max" << "Output";
browse = ind;
init();
}
148
void Dialog_Tci::init()
{
this->setWindowTitle(title);
this->setGeometry(60,60,400,300);
QGridLayout *layout_000 = new QGridLayout(this);
label[0] = new QLabel(list[0]);
label[1] = new QLabel(list[1]);
label[2] = new QLabel(list[2]);
label[3] = new QLabel(list[3]);
lineedit[0] = new QLineEdit;
lineedit[1] = new QLineEdit;
lineedit[2] = new QLineEdit;
lineedit[3] = new QLineEdit;
btn[0] = new QPushButton("browse");
QObject::connect(btn[0],SIGNAL(clicked()),this,SLOT(slot_browse_000()));
btn[1] = new QPushButton("browse");
QObject::connect(btn[1],SIGNAL(clicked()),this,SLOT(slot_browse_001()));
btn[2] = new QPushButton("browse");
QObject::connect(btn[2],SIGNAL(clicked()),this,SLOT(slot_browse_002()));
btn[3] = new QPushButton("browse");
QObject::connect(btn[3],SIGNAL(clicked()),this,SLOT(slot_browse_003()));
btn_ok = new QPushButton("Process");
QObject::connect(btn_ok,SIGNAL(clicked()),this,SLOT(slot_ok()));
btn_cancel = new QPushButton("Cancel");
QObject::connect(btn_cancel,SIGNAL(clicked()),this,SLOT(close()));
149
layout_000->addWidget(label[0],0,0);
layout_000->addWidget(label[1],1,0);
layout_000->addWidget(label[2],2,0);
layout_000->addWidget(label[3],3,0);
layout_000->addWidget(lineedit[0],0,1,1,4);
layout_000->addWidget(lineedit[1],1,1,1,4);
layout_000->addWidget(lineedit[2],2,1,1,4);
layout_000->addWidget(lineedit[3],3,1,1,4);
layout_000->addWidget(btn[0],0,6,1,1);
layout_000->addWidget(btn[1],1,6,1,1);
layout_000->addWidget(btn[2],2,6,1,1);
layout_000->addWidget(btn[3],3,6,1,1);
layout_000->addWidget(btn_cancel,4,4);
layout_000->addWidget(btn_ok,4,6);
this->setLayout(layout_000);
}
void Dialog_Tci::slot_ok()
{
QString str;
QByteArray ba;
str = lineedit[0]->text();
read(str,0);
str = lineedit[1]->text();
read(str,1);
str = lineedit[2]->text();
read(str,2);
process();
150
str = lineedit[3]->text();
write(str);
this->close();
}
void Dialog_Tci::slot_browse_000()
{
QString fileName = QFileDialog::getOpenFileName(this, tr("Open File"),"");
lineedit[0]->setText(fileName);
QImage temp(fileName);
pixel = matrix3d(browse,matrix2d(temp.width(),temp.height()));
outp = matrix2d(temp.width(),matrix1d(temp.height()));
delete &temp;
}
void Dialog_Tci::slot_browse_001()
{
QString fileName = QFileDialog::getOpenFileName(this, tr("Open File"),"");
lineedit[1]->setText(fileName);
}
void Dialog_Tci::slot_browse_002()
{
QString fileName = QFileDialog::getOpenFileName(this, tr("Open File"),"");
lineedit[2]->setText(fileName);
}
void Dialog_Tci::slot_browse_003()
151
{
QString
fileName
=
QFileDialog::getSaveFileName(this,
File"),"","*.tif");
lineedit[3]->setText(fileName);
}
void Dialog_Tci::read(QString filename,int inp)
{
image = new QImage(filename);
QRgb rgb;
for (int i=0; i<(image->width()); i++)
{
for (int j=0; j<(image->height()); j++)
{
rgb
= image->pixel(i,j);
pixel[inp][i][j] = qGray(rgb);
maxh
= j;
}
maxw = i;
}
maxh++;
maxw++;
}
void Dialog_Tci::process()
{
for (int i=0;i<(image->width()); i++)
{
for (int j=0; j<(image->height()); j++)
{
if ((pixel[2][i][j]-pixel[1][i][j]) ==0)
tr("Save
152
{
outp[i][j] = 0;
}
else
{
outp[i][j] = ( (pixel[0][i][j]-pixel[1][i][j]) / (pixel[2][i][j]pixel[1][i][j]) );
outp[i][j] = (int) (100 * (outp[i][j]));
}
}
}
}
void Dialog_Tci::write(QString str)
{
image = new QImage(maxw,maxh,QImage::Format_RGB32);
QRgb value;
for (int i=0; i<maxw; i++)
{
for (int j=0; j<maxh; j++)
{
value = qRgb(outp[i][j],outp[i][j],outp[i][j]);
image->setPixel(i,j,value);
}
}
if(! image->save(str,"TIFF",100) )
qWarning("Save Failed!");
else
{
strout = str;
153
}
}
Coding for VH
#include <QtGui>
#include "Dialog_Vh.h"
Dialog_Vh::Dialog_Vh()
{
}
Dialog_Vh::Dialog_Vh(int ind)
{
title = "VH File Input Dialog";
list << "VCI" << "TCI" << "output";
browse = ind;
init();
}
void Dialog_Vh::init()
{
this->setWindowTitle(title);
this->setGeometry(60,60,400,300);
QGridLayout *layout_000 = new QGridLayout(this);
label[0] = new QLabel(list[0]);
label[1] = new QLabel(list[1]);
label[2] = new QLabel(list[2]);
lineedit[0] = new QLineEdit;
lineedit[1] = new QLineEdit;
lineedit[2] = new QLineEdit;
btn[0] = new QPushButton("browse");
QObject::connect(btn[0],SIGNAL(clicked()),this,SLOT(slot_browse_000()));
btn[1] = new QPushButton("browse");
QObject::connect(btn[1],SIGNAL(clicked()),this,SLOT(slot_browse_001()));
btn[2] = new QPushButton("browse");
QObject::connect(btn[2],SIGNAL(clicked()),this,SLOT(slot_browse_002()));
btn_ok = new QPushButton("Process");
QObject::connect(btn_ok,SIGNAL(clicked()),this,SLOT(slot_ok()));
btn_cancel = new QPushButton("Cancel");
154
QObject::connect(btn_cancel,SIGNAL(clicked()),this,SLOT(close()));
layout_000->addWidget(label[0],0,0);
layout_000->addWidget(label[1],1,0);
layout_000->addWidget(label[2],2,0);
layout_000->addWidget(lineedit[0],0,1,1,4);
layout_000->addWidget(lineedit[1],1,1,1,4);
layout_000->addWidget(lineedit[2],2,1,1,4);
layout_000->addWidget(btn[0],0,6,1,1);
layout_000->addWidget(btn[1],1,6,1,1);
layout_000->addWidget(btn[2],2,6,1,1);
layout_000->addWidget(btn_cancel,4,4);
layout_000->addWidget(btn_ok,4,6);
this->setLayout(layout_000);
}
void Dialog_Vh::slot_ok()
{
QString str;
QByteArray ba;
str = lineedit[0]->text();
read(str,0);
str = lineedit[1]->text();
read(str,1);
process();
str = lineedit[2]->text();
write(str);
this->close();
}
void Dialog_Vh::slot_browse_000()
{
QString fileName = QFileDialog::getOpenFileName(this, tr("Open File"),"");
lineedit[0]->setText(fileName);
QImage temp(fileName);
pixel = matrix3d(browse,matrix2d(temp.width(),temp.height()));
outp = matrix2d(temp.width(),matrix1d(temp.height()));
}
155
void Dialog_Vh::slot_browse_001()
{
QString fileName = QFileDialog::getOpenFileName(this, tr("Open File"),"");
lineedit[1]->setText(fileName);
}
void Dialog_Vh::slot_browse_002()
{
QString fileName = QFileDialog::getSaveFileName(this, tr("Save
File"),"","*.tif");
lineedit[2]->setText(fileName);
}
void Dialog_Vh::read(QString filename,int inp)
{
image = new QImage(filename);
QRgb rgb;
for (int i=0; i<(image->width()); i++)
{
for (int j=0; j<(image->height()); j++)
{
rgb
pixel[inp][i][j] = qGray(rgb);
maxh
= j;
}
maxw = i;
}
maxh++;
maxw++;
}
= image->pixel(i,j);
void Dialog_Vh::process()
{
QString str;
for (int i=0;i<(image->width()); i++)
{
for (int j=0; j<(image->height()); j++)
{
outp[i][j] = (pixel[1][i][j] + pixel[0][i][j]) / 2;
//outp[i][j] = (int) (100 * (outp[i][j]) + 1);
}
}
}
void Dialog_Vh::write(QString str)
{
156
image = new QImage(maxw,maxh,QImage::Format_RGB32);
QRgb value;
for (int i=0; i<maxw; i++)
{
for (int j=0; j<maxh; j++)
{
value = qRgb(outp[i][j],outp[i][j],outp[i][j]);
image->setPixel(i,j,value);
}
}
if(! image->save(str,"TIFF",100) )
qWarning("Save Failed!");
else
{
strout = str;
}
}
Coding for Fire Risk Map
#include <QHBoxLayout>
#include <QStatusBar>
#include <QGroupBox>
#include <QPushButton>
#include <QMenuBar>
#include <QMenu>
#include <QAction>
#include <QFrame>
#include <QDialog>
#include <QProcess>
#include <QDir>
#include <QLabel>
#include <QRect>
#include <QGridLayout>
#include <QFileDialog>
157
#include <QLineEdit>
#include <QMessageBox>
#include <QtGui>
#include <QKeySequence>
#include <QImage>
#include "WindowCalc.h"
#include "Dialog.h"
WindowCalc::WindowCalc(QWidget *parent) : QMainWindow(parent)
{
this->showMaximized();
frame = new QFrame();
frame->setFrameStyle(QFrame::Panel);
this->setWindowTitle("Calculation Window");
//this->setGeometry(50,50,800,600);
statusBar = new QStatusBar(this);
this->setStatusBar(statusBar);
statusBar->showMessage("None currently processed.");
menuBar = new QMenuBar();
menu_Text_File = new QMenu("File");
menu_Text_Edit = new QMenu("Edit");
act_Open = new QAction("Open File",menu_Text_File);
act_Open->setShortcut(QKeySequence("Ctrl+O"));
connect(act_Open,SIGNAL(triggered()),this,SLOT(slot_open()));
act_Exit = new QAction("Exit",menu_Text_File);
158
act_Exit->setShortcut(QKeySequence("Ctrl+E"));
connect(act_Exit,SIGNAL(triggered()),this,SLOT(close()));
act_SetFire = new QAction("Set Fire Input",menu_Text_Edit);
connect(act_SetFire,SIGNAL(triggered()),this,SLOT(slot_setFire()));
act_FileInfo = new QAction("File Information",menu_Text_Edit);
connect(act_FileInfo,SIGNAL(triggered()),this,SLOT(slot_fileInfo()));
act_FileInfo->setEnabled(false);
act_clearImage = new QAction("Close File",menu_Text_Edit);
act_clearImage->setShortcut(QKeySequence("Ctrl+p"));
connect(act_clearImage,SIGNAL(triggered()),this,SLOT(slot_clearImage()));
menu_Text_File->addAction(act_Open);
menu_Text_File->addAction(act_clearImage);
menu_Text_File->addAction(act_Exit);
menu_Text_Edit->addAction(act_SetFire);
menu_Text_Edit->addAction(act_FileInfo);
menuBar->addMenu(menu_Text_File);
menuBar->addMenu(menu_Text_Edit);
this->setMenuBar(menuBar);
btn_ndvi = new QPushButton("NDVI");
QObject::connect(btn_ndvi,SIGNAL(clicked()),this,SLOT(slot_ndvi()));
btn_vci = new QPushButton("VCI");
QObject::connect(btn_vci,SIGNAL(clicked()),this,SLOT(slot_vci()));
btn_tgi = new QPushButton("TCI");
159
QObject::connect(btn_tgi,SIGNAL(clicked()),this,SLOT(slot_tci()));
btn_vh = new QPushButton("VH");
QObject::connect(btn_vh,SIGNAL(clicked()),this,SLOT(slot_vh()));
btn_fire = new QPushButton("Fire Risk Map");
QObject::connect(btn_fire,SIGNAL(clicked()),this,SLOT(slot_fire()));
label = new QLabel();
label->resize(600,600);
label->setFrameStyle(QFrame::Box);
label->setScaledContents(true);
layout_000 = new QGridLayout();
layout_000->setSizeConstraint(QLayout::SetMaximumSize);
groupBox1 = new QGroupBox("Main Function");
groupBox2 = new QGroupBox("Image Display - No Image Loaded");
layout_001 = new QVBoxLayout();
layout_002 = new QVBoxLayout();
const QRect rect(0,0,100,600);
layout_001->setGeometry(rect);
layout_001->addWidget(btn_ndvi);
layout_001->addWidget(btn_vci);
layout_001->addWidget(btn_tgi);
layout_001->addWidget(btn_vh);
layout_001->addWidget(btn_fire);
160
layout_002->addWidget(label);
groupBox1->setLayout(layout_001);
groupBox2->setLayout(layout_002);
layout_000->addWidget(groupBox1,0,0);
layout_000->addWidget(groupBox2,0,1,2,4);
frame->setLayout(layout_000);
this->setCentralWidget(frame);
// by default, setFire is 3 input (3-1=2)
setFire = 2;
}
void WindowCalc::slot_ndvi()
{
dialog_ndvi = new Dialog_Ndvi(3);
dialog_ndvi->exec();
QString str = dialog_ndvi->strout;
if (! str.isEmpty())
{
str_img = str;
act_FileInfo->setEnabled(true);
image = new QImage(str);
label->setPixmap(QPixmap::fromImage(*image));
str = "Image Display - Loaded: " + str;
groupBox2->setTitle(str);
}
}
void WindowCalc::slot_vci()
{
dialog_vci = new Dialog_Vci(4);
161
dialog_vci->exec();
QString str = dialog_vci->strout;
if (! str.isEmpty())
{
str_img = str;
act_FileInfo->setEnabled(true);
image = new QImage(str);
label->setPixmap(QPixmap::fromImage(*image));
str = "Image Display - Loaded: " + str;
groupBox2->setTitle(str);
}
}
void WindowCalc::slot_tci()
{
dialog_tci = new Dialog_Tci(4);
dialog_tci->exec();
QString str = dialog_tci->strout;
if (! str.isEmpty())
{
str_img = str;
act_FileInfo->setEnabled(true);
image = new QImage(str);
label->setPixmap(QPixmap::fromImage(*image));
str = "Image Display - Loaded: " + str;
groupBox2->setTitle(str);
}
}
void WindowCalc::slot_vh()
{
162
dialog_vh = new Dialog_Vh(3);
dialog_vh->exec();
QString str = dialog_vh->strout;
if (! str.isEmpty())
{
str_img = str;
act_FileInfo->setEnabled(true);
image = new QImage(str);
label->setPixmap(QPixmap::fromImage(*image));
str = "Image Display - Loaded: " + str;
groupBox2->setTitle(str);
}
}
void WindowCalc::slot_fire()
{
dialog_fire = new Dialog_Fire(setFire);
dialog_fire->exec();
QString str = dialog_fire->strout;
if (! str.isEmpty())
{
str_img = str;
act_FileInfo->setEnabled(true);
image = new QImage(str);
label->setPixmap(QPixmap::fromImage(*image));
str = "Image Display - Loaded: " + str;
groupBox2->setTitle(str);
}
}
void WindowCalc::slot_setFire()
163
{
dlg_queryFire = new Dlg_QueryFire(setFire);
dlg_queryFire->exec();
setFire = dlg_queryFire->num;
}
void WindowCalc::slot_clearImage()
{
label->clear();
image = NULL;
str_img = "";
act_FileInfo->setEnabled(false);
groupBox2->setTitle("Image Display - No Image Loaded");
}
void WindowCalc::slot_open()
{
QString str = QFileDialog::getOpenFileName(this, tr("Open File"),"","Images
(*.tif *.tiff)");
image = new QImage(str);
str_img = str;
label->setPixmap(QPixmap::fromImage(*image));
str = "Image Display - Loaded: " + str;
groupBox2->setTitle(str);
act_FileInfo->setEnabled(true);
}
void WindowCalc::slot_fileInfo()
{
QString str;
QString temp;
164
if (! str_img.isEmpty())
{
QFileInfo file(str_img);
temp = file.fileName();
str = str + "File Name: " + temp + "\n";
temp = file.absolutePath();
str = str + "File Path: " + temp + "\n";
str = str + "Image Depth Size: " + temp.setNum(image->depth()) + "\n";
str = str + "Image Width: " + temp.setNum(image->width()) + "\n";
str = str + "Image Size: " + temp.setNum(image->height()) + "\n";
if (image->isGrayscale())
temp = "yes";
else
temp = "no";
str = str + "Is GrayScale: " + temp + "\n";
QMessageBox::information(this,"File Information",str);
}
}