SMOKE DETECTION USING REMOTE SENSING TECHNIQUE

(1)

SMOKE DETECTION USING REMOTE SENSING

TECHNIQUE

ASMALA BIN AHMAD

PUSAT PERKHIDMATAN AKADEMIK

KOLEJ UNIVERSITI TEKNIKAL KEBANGSAAN

MALAYSIA

(2)

ABSTRACT

SMOKE DETECTION USING REMOTE SENSING TECHNIQUES

(Key Words: PM 10, NOAA-14 AVHRR, AQI, Regression, RMSE)

Particulate Matter sizing less than 10 micrometers or widely known as PM 10 is one of the major constituents of the thick smoke haze phenomenon, which occurred in Malaysia during September 1997. In this study, seven scenes of NOAA-14 AVHRR satellite data were acquired in order to det,ermine and map PM 10 over Malaysia. Five locations of air pollution station 'were chosen where PMlO was measured. Initially, pre-processing tasks namely; atmospheric and geometric co ection was implemented before further digital image processing job. Next, band 1 of the AVHRR sensor with wavelength ranging fiom 0.58 to 0.68 micrometers were calibrated to compensate for post-launch sensor degradation. Cloud separation was then carried using visual and thresholding technique. Relationship between the satellite reflectance and the corresponding PM 10 AQI (Air Quality Index) at the selected stations was established using linear regression model. The model was then used to map the concentration of PMlO over Malaysia. The result indicates that remote sensing technique using bands 1 of NOAA-14 AVHRR data was capable to determine PM 10 concentration quantitatively and spatially in continuous manner. Finally, accuracy was assessed using Root-mean-square error technique.

Key Researchers: Asmala Ahmad Mohamad Nizam Ayof

Agus Setyo Budi Harnzah Sakidin Sharifah Sakinah Syed Ahmad

(3)

TABLE OF CONTENT

Abstract

...

1

Key Researchers: ... 1

Acknowledgement ... 2

Table of Content

...

3

List of Figures

...

4

...

12

4.1.2. Satellite Data

...

17

4.1.3. Ancillary Data

...

19

4.2. Method

...

19

...

4.2.1. Acquisition Dates And Area Selection 19 4.2.2. Post-Launch Calibration

...

2 0 4.2.3. Atmospheric Correction

...

20

4.2.4. Geometric Correction

...

21

4.2.5. Cloud Masking

...

21

4.2.6. Derivation of Smoke haze Model ... 23

4.2.7. Regression Analysis

...

24

4.2.8. Generation of Digital Smoke haze Image

...

24

Between July and November 1997, it is estimated that 45,000 km2 of forest and

land burnt on the islands Sumatra and Kalimantan (ASEAN, 2006). Fire-related air

pollution episodes are a recurrent phenomenon in Southeast Asia. Nine such incidents

have been reported over the last 20 years, of which the 1997 smoke haze episode

attracted the broadest attention.

Among the worst ASEAN countries hit by the catastrophe includes Singapore

and Malaysia, Indonesia does not yet have an inteyted air quality monitoring

network which could provide real-time, region-covering air quality information. Due

to the absence of such information, an assessment of the severity of the fire-related air

pollution episodes is limited. As a surrogate, horizontal visibility was frequently used

's

to report the status of ambient air pollution. However, even though sufficient

information on the status of air quality was available in Singapore and Malaysia,

much uncertainty existed on the impacts of such air pollution episodes and on how to

response adequately to them. The governments of the affected countries

recommended the public to remain indoors as much as possible, to avoid physical

exertion and to wear respiratory masks outdoors. In Kuching, Borneo-Malaysia, the

state of emergency was proclaimed for 10 days in 1997, leading to the closure of

schools, public offices and factories. Dubious statements in the media on the impacts

of the smoke haze were disseminated

-

such as the daily exposure would equal to 20

to 40 cigarettes; panic easily evolved (ASEAN, 2006).

The air pollutant that predominantly caused violations of ambient air quality

standards was particulate matter. Particulate matter may cause acute and chronic

respiratory diseases such as bronchitis, asthma and upper respiratory tract infections.

Increased ambient particle concentrations are suspected to be linked with increased

daily mortality. By scattering and absorbing light, particulates also result in reduced

(5)

1.1. Smoke haze Development In 1997

The influence of the 1997 fires in Kalimantan and Sumatra on ambient air quality

was discernible by July, peaked in SeptemberIOctober and weakened by November,

when the delayed monsoon rain extinguished the fires and scavenged the atmosphere.

During the peak episode, satellite imagery (NASAITOMS aerosol index maps)

showed a smoke haze layer, which expanded, over an area of more than 3 million

km2, covering large parts of Sumatra and Kalimantan. Its northward extension

partially reached Malaysia, Singapore, Brunei and Thailand. During this period,

particulate matter concentrations frequently exceeded national ambient air quality \

standards. Air quality measurement for areas close to fires in Kalimantan and

Sumatra indicate that ambient particle concentration was roughly 20 to 40 times the

normal (non-haze) background concentration and exceeded levels categorised as

'h&doust (or 'significant harm level'). Monthly mean horizontal visibility at most

locations in Sumatra and Kalimantan in September was below 1 km and daily

maximum visibility was fiequently below 100 metres (ASEAN, 2006).

The neighbouring region most affected by pyrogenic transmissions in 1997 was

Sarawak, Borneo-Malaysia. In the city of Kuching, ambient particle concentration

rose roughly 5 to 20 times above background levels, with in total 32 days in the

'unhealthy to hazardous' range. Visibility decreased from generally above 15 km to

below 0.5 km during this period (ASEAN, 2006).

In Singapore and Peninsular Malaysia, a 2 to 5-times rise in ambient particle

concentration was recorded. 12 and more than 40 days, respectively, were in the

'unhealthy' range in Singapore and Kuala Lumpur. Visibility below 2 km

(6)

1.2. Impacts of the smoke haze episodes

In all countries affected by the smoke haze, an increase of acute health outcomes

was observed. Health outcomes included emergency room visits due to respiratory

symptoms such as asthma, upper respiratory infection, decreased lung function and

eye and skin initation. In Malaysia, for instance, health surveillance showed a 30%

increase in hospital attendance for haze-related illnesses. Generally, individuals such

with pre-existing respiratory and cardiac diseases, but also elderly and children are

most susceptible to adverse health outcomes from smoke haze exposure. In addition

to the acute effects, chronic diseases are likely follow. The smoke haze episodes add

to the urban and industrial air pollution in Asia, which has reached

alarming levels in many metropolitan areas. The synergistic effects of smoke haze

and background air pollution are uncertain (ASEAN, 2006).

Besides health impacts, impaired visibility seriously affected the economies of

Indonesia, Malaysia and Singapore. Among the economic sectors affected most were

air, land and sea transportation, construction, tourism and agro based industry. WWF

roughly estimated the economic value of the damages caused by the 1997 fires and

smoke haze. They estimated

R M

5 billion of smoke haze-related damages for

Indonesia, Malaysia and Singapore. However, a variety of the damages such as

decreased quality of life, losses of biodiversity and atmospheric impacts are difficult

to monitories (WWF, 2006).

Fire-related smoke haze episodes also reveal a social component: a large part of

the population in Southeast Asia do not have the financial means to buy protective

measures such as respiratory masks and air conditioning nor are they able to refrain

from outdoor work when air pollution is high. The same applies to medical treatment

(7)

1.3. Response to the 1997 smoke haze episode

The 1997 smoke haze episode resulted in an intensification of regional measures

towards cooperation in fire and smoke management which were initiated in the aftermath of the 1991 and 1994 smoke haze episodes. These measures include the

establishment of ASEAN Haze Technical Task Force and the implementation of Regional and National Haze Action Plans. These plans define the ASEAN's countries contribution to fire prevention, monitoring, fighting and other mitigation measures.

Among others, it is also targeted to upgrade the national air quality and

meteorological monitoring networks in order to strengthen the region's early warning

and monitoring system in respect to smoke haze (ASEAN, 2006).

The 1997 once again made evident that in addition to a sound fire management a

fu&mental revision of the current land conversion and fire use policies is required

to prevent the reoccurrence of similar episodes. Ground based and airborne

investigations of the smoke haze 1997 indicated that fires on peat swamp vegetation

made a substantial contribution to the smoke haze development, which, however, are

estimated to have contributed only 30% to the total area burnt. Given this apparent particular relevance of peat swamp fires to the development of transboundary smoke

haze, emission reduction and control strategies will have to focus on the prevention of fires in this type of vegetation as a matter of priority (ASEAN, 2006).

Thus, immediate plan regarding development of smoke haze early warning

system using latest technology such as satellite remote sensing will have to be

initiated. The economic and health impacts damages of the 1997 smoke haze tragedy

demonstrated that proper and efficient measures for future smoke haze events

represents an influencing factor for public and economic prosperity especially in

(8)

L

Background

Smoke can be defined as foreign particulate matter in the atmosphere resulting

h m combustion processes. Smoke that has come a great distance from its source,

such as from forest fires, usually has a light greyish or bluish colour and is evenly

distributed in the upper air.

Smoke originating from open burning or forest fire contains large amount of

particulate matter (e.g., organic matter, graphitic carbon). This particulate matter is

hazardous to health, especially associated with lung aqd eye deceases. Besides that it

is capable of increasing the atmospheric greenhiuse effects and affecting the

tropospheric chemistry.

Conventionally, PM 10 can be measured from ground instruments such as air

sampler, sun photometer and optical particle counter, however these instruments is

impractical if measurement are to be made over relatively large areas or for

continuous monitoring.

The smoke haze episode, which occurred during mid-May to November 1997,

is considered the worst since 1980 (five similar haze episodes had occurred in April

1983, August 1990, June 1991, October 1991 and August 1994). On lgh September

1997 Malaysian government had declared that Kuching (capital of Sarawak) was in

the state of emergency when the PMlO API (Air Pollution Index) exceeded 650

(hazardous level). By 23d September 1997 the condition worsened as Kuching's

PMlO API reached 839, the highest ever been recorded by the country.

Meteorologist had revealed that it was due to the injection of suspended ash

particles from large scale forest fire in Sumatra and Kalimantan. In addition, the

occurrence of shallow localized haze in big cities (e.g., Kuala Lumpur, Kelang, Johor

Bahru) caused mainly by vehicle and industrial emissions which were stimulated by

the South West Monsoon Season which acted as the minor contributor that made the

(9)

early warning system so that precaution measures can be disseminated to public

effectively.

Research on estimation of smoke have been extensively carried out by many

researchers elsewhere. The most commonly used method of identifjmg smoke is to

assign different colours to different channels or channel combinations. The resulting

false colour images can provide visual separation of smoke from other objects (Li et

al., 2001). Kaufman et al. (1990) assigned AVHRR channel 1 to red, channel 2 to

green and channel 4 to blue, generating a composite image showing smoke plumes.

Another popular approach used by other researchers is pesholding. Christopher et al.

(1996) examined various AVIIRR channels and their combinations to distinguished

smoke. He then applied texture analysis to these channels and their combinations.

i

This study developed new remote sensing techniques of using "slqlight"

approach for detecting smoke. Unlike many previous studies which dealing with

mainly colours combinations, the techniques proposed here introduced the using of

statistical approach to estimate smoke quantitatively (Asmala and Mazlan, 2000,

2002). Figure 1 shows an operational system for smoke haze detection system.

Penghantaran Data Satelit Melalul

[image:9.505.46.395.409.613.2]

Rangkaian Komputer

Figure 1 : Operational of Smoke haze Detecting System

As the visibility of smoke being dominated by very tiny constituent known as

particulate matter sizing less than 10 micrometers in diameter or PMIO, thus this

(10)

reports focus on results of a study to determine PM 10 from NOAA-14

AVHRR

satellite data.

Data and materials used, technique employed and result obtained are

explained in detailed in the following sections. The performance of the techniques is

evaluated using suitable accuracy assessment method in the last part of this report.

3. Statement of Problem

(i) Existing smoke detection devices are incapable~f providing an early warning to

the public, time consuming especially for large area and expensive for

continuous monitoring.

(ii) Overlapping spectral signature of satellite measurements between smoke, clouds

and background surface.

(iii) It is critical for public to know daily typical regional drift pattern of smoke so

that immediate precautions can be carried out accordingly.

4. Approach

This study focuses in the 1997 thick smoke haze episode in Malaysia. However the

result are applicable for current and future use.

4.1. Materials

Three types of data namely; ground-truth data, satellite data and ancillary data will be

used.

4.1.1. Ground-truth data

Conventional measurements of smoke haze were complementarily used

(11)

messwements in micrograms per meter cube (pgm3) from lSt to 3oh September 1997

wge canied out by Alam Sekitar Malaysia Sdn. Bhd. (ASMA) to represent the actual

I haze intensity over the study area. For the purpose of this study, the PM 10

ooecentration in pgrnJ was later converted to AQI. AQI Calculator was used for this se 2006a). Some of the displays are shown in Figure 2 and Figure 3.

alculated using pollutant concentration data, the breakpoints in Table 1 ge, and the following equation (linear interpolation):

c

s the rounded concentration of pollutant

g p = the breakpoint that is greater than or equal to C Him

pp

= the breakpoint that is less than or equal to C L,,,

5

the AQT value corresponding to BP Hi Hi

(12)

I..

kf..

3 ' -

-

ooc-102 WI-$9'0 wo

-

soco vw- S'ZI nt. rsr VW-T'(KI

4-n (WIL- ISI 6) mr'o

-

szra V S I - S'ZI SF

-

ssz C'KI

-

ST^

(13)

(14)

(15)