UBC Faculty Research and Publications

Regional assessment of air pollution exposure during the 1997-98 Southeast Asian air pollution episodes Brauer, Michael; Ostermann, Kathryn; Steyn, Douw 2001

You don't seem to have a PDF reader installed, try download the pdf

Item Metadata

Download

Media
[if-you-see-this-DO-NOT-CLICK]
cida_apec.pdf [ 3.57MB ]
Metadata
JSON: 1.0048199.json
JSON-LD: 1.0048199+ld.json
RDF/XML (Pretty): 1.0048199.xml
RDF/JSON: 1.0048199+rdf.json
Turtle: 1.0048199+rdf-turtle.txt
N-Triples: 1.0048199+rdf-ntriples.txt
Original Record: 1.0048199 +original-record.json
Full Text
1.0048199.txt
Citation
1.0048199.ris

Full Text

Regional assessment of air pollution exposure during the 1997-98 Southeast Asian air pollution episodes Michael Brauer1* Kathryn Ostermann2 Douw Steyn2 1  School of Occupational and Environmental Hygiene 2 Atmospheric Sciences Program The University of British Columbia  *  Address for correspondence: Michael Brauer School of Occupational and Environmental Hygiene 2206 East Mall, Vancouver BC V6T 1Z3 Canada Tel: 604-822-9585 Fax: 604-822-9588 brauer@interchange.ubc.ca  Draft final report, February 9, 2001  DRAFT  DO NOT QUOTE OR CITE  DRAFT  TABLE OF CONTENTS Executive Summary......................................................................... 3 INTRODUCTION ........................................................................... 8 Statement of problem .............................................................................................................. 8 Objectives................................................................................................................................ 9 Background ............................................................................................................................. 9 Health Impacts....................................................................................................................... 15 TOMS - Introduction............................................................................................................ 17 Aerosol index ........................................................................................................................ 17 Surface monitoring................................................................................................................ 18  METHODS .................................................................................... 21 Overview ................................................................................................................................... 21 Data overview ........................................................................................................................... 22 TOMS.................................................................................................................................... 22 PM10 measurements............................................................................................................... 23 Population density................................................................................................................. 23 Processing of TOMS data ......................................................................................................... 24 Processing of PM10 data........................................................................................................... 25 Comparisons of TOMS/PM10 .................................................................................................. 26 Generation of estimated PM10 maps ........................................................................................ 26  RESULTS AND DISCUSSION.................................................... 27 LIMITATIONS.............................................................................. 43 CONCLUSIONS ........................................................................... 44 ACKNOWLEDGMENTS ............................................................. 45 BIBLIOGRPAHY ......................................................................... 46 Appendix........................................................................................ 65 Technical Advisory Committee Members: ............................................................................... 65  2  Executive Summary Introduction While the vegetation fire episodes of 1997 and 1998, including country-level air quality monitoring programs and health surveillance activities, were documented in the media, there has not been a comprehensive regional assessment of air quality or health impacts that incorporates the various country-level data sources. The project described in this report was conducted to assimilate available air quality and health impact data for the region that were collected during the episode periods. It is believed that Health and Environment ministries in the affected countries would benefit from a more detailed assessment of the concentrations and exposures than previously available. A regional determination of concentrations and population exposure will help support the fire and pollution prevention activities of authorities within the affected countries. This project is an initial test of the use of remote sensing as a tool to rapidly assess the health impact of such large-scale vegetation fire air pollution episodes. In this project we describe the methodology needed to use remote sensing data in the analysis of human exposure and health impacts. As such, the methodology developed in this project can be used in future preventive efforts. Additionally, by conducting a regional assessment of the air pollution exposures, and specifically the extent of the population currently covered by ground-based air monitoring stations, we were able to identify areas which were not adequately monitored. This information can be used by national governments as well as by regional organizations (ASEAN, APEC) as guidance for location of future monitoring stations within the region. Accordingly, this project was conducted with the following objectives: i) To use remote sensing (satellite) data, in combination with traditional ground-based air monitoring data, to develop a regional database of particulate air pollution levels associated with the Southeast Asian forest fires 1997-98. ii) To combine the regional particulate database with population density and other demographic data, to assess population exposure to air pollution associated with the Indonesian forest fires.  3  iii) To gather available data from affected countries (Malaysia, Thailand, Singapore, Indonesia, Brunei) within the region to assess the feasibility of conducting a regional assessment of health impacts iv) To conduct a regional assessment of the health impacts of the regional air pollution episodes occurring in 1997-1998 in Southeast Asia, using local data or a risk assessment approach.  This project originated from the Biregional Workshop on the Health Impacts of “Haze-Related” Air Pollution sponsored by WHO and held in Kuala Lumpur, Malaysia during June 1998. Several of the workshop conclusions support a regional analysis of the health impacts of the 1997-1998 “haze” episodes and call for •  “With regard to air quality monitoring and episode forecasting, from the health sector’s perspective, information on the nature and extent of human exposure to environmental pollutants is essential to impact assessment.”  •  “the implementation of joint studies on the health impacts of the 1997 haze, including the assessment of needs for air quality monitoring data from a public health point of view;”  •  “a regional study of short-term health impacts using standardized methodologies and routinely-collected data;”  The first two points listed above have been addressed in this project. We have provided a methodology to assess human exposure and analyzed exposures during the vegetation fire episodes. It is clear that large numbers of people were exposed to high levels of particulate air pollution although the specific health impacts of these exposures are not clear. With regard to air monitoring we show that monitoring was limited in regions of high population. While this situation may have improved somewhat since 1997-98, there appears to be little correspondence on a regional scale between population density and the density of air monitoring stations. The final project objective was not feasible due to the inability to collect data from the entire region and due to the lack of standardization of health impact data within the region. In terms of risk assessment we have produced a reasonable estimate of the numbers of people exposed above different level of particle concentrations for each day during the study period. These data may be used to estimate health impacts based on data from epidemiological studies that have been 4  conducted elsewhere in the world. However, due to the lack of epidemiological data associating health impacts from vegetation fires within this region we have not produced such estimates here. Results and Discussion This report describes a rather complete database of ambient particle concentrations during the 1997-1998 vegetation fire episodes. As early as the beginning of May, 1997 air quality began to deteriorate in the region. 24-hour average concentrations in Singapore peaked at 225 µg/m3 on October 2, 1997. In Sarawak, the 24-hour average PM10 measured in Kuching peaked at high as 852 µg/m3 on September 22, 1997 which represents a value more than 15 times the normal levels. In Kuala Lumpur, concentrations of 420 and 411 µg/m3 were measured on September 14 and 15, respectively. Due to the location of fires and direction of the wind, Sumatra and Kalimantan in Indonesia were the most severely affected areas in 1997, although locations in southern Thailand were also affected for short periods. In 1998 elevated concentrations were much more localized in the areas around Miri and Brunei. In Miri, 24-hour average PM10 concentrations reached 666 µg/m3 on March 30, 1998. During the more prolonged and spatially widespread 1997 episode, mean concentrations over the 3.5 month period in 1997 which included the episode were 2 –3 times higher than non-episode periods, depending upon the level of background air pollution at a particular location. During the periods of highest concentrations the spatial extent of the episode included large areas without any surface monitoring sites. In particular the much of Sumatra as well as portions of southern Kalimantan have high population density and had not ambient PM10 moniotirng sites during the 1997-98 period. These areas should be primary targets for additional monitoring sites which hope to reflect population exposure. We estimated population exposures by combining population density data in the affected areas of Southeast Asia with estimated ambient particle concentrations. The particle concentrations wer estimated based on a regression approach which associated aerosol index measurements from the TOMS satellite sensor with available surface-level PM10 monitoring data. In some areas, the surface monitoring network was relatively dense, while in others, little or no ground-based monitoring data were available for the periods of the air pollution episodes. The output of this  5  level of the work was a series of daily maps of estimated particle levels within the region. These data were then combined with population density data to estimate the daily counts of numbers of people exposed to different levels of air pollution within the region computed. For examples to display in this report we randomly selected 10 days from within the March 1, 1997 – April 30, 1998 study period to present as figures, including days from both the 1997 and 1998 episode periods as well as non-episode periods. Although not shown in this report, similar data are available for all days during the study period. The most important limitations in this methodology include the inability to estimate particle concentrations above 500 ug/m3 and the low (1 km x 1.25 km) spatial resolution of the TOMS data. Due to the fact that the aerosol plume may become opaque at high concentrations, the TOMS aerosol index was saturated above 500 ug/m3. This reduced our ability to estimate the peak concentrations in some locations and particularly in the fire source regions. Further, we found the TOMS data to have limited spatial resolution such that predictions were poor for local air pollution occurrences such as those due to smaller fires reported around Miri in spring 1998. Despite these limitations, using the TOMS data and the methodology described in the report, we were able to provide estimates of exposures during the 1997-98 episodes. Exposures were high and affected huge segments of the population, in some cases for extended periods. During a single day within the episode period as many as 87.8 million people were estimated to have been exposed to ambient concentrations above 150 ug/m3. In viewing the episode periods in their entirety, we estimated that there were 2.26 billion person-days of exposure to PM10 concentrations above 150 µg/m3 during the study period, 96% of which during the episode periods. Despite a number of major limitations, we found that the TOMS aerosol index data was a useful indicator of surface PM10 concentrations during such regional episodes and especially where concentrations were below 500 ug/m3 and where surface monioting data were not available. An additional advantage of the TOMS data is that is it can be made available almost immediately whereas there is often a significant time delay in the reporting of surface-based measurement data. This feature as well as the broad spatial coverage makes the TOMS useful as an early warning monitor within this region as well as others that are subject to similar air pollution events.  6  7  INTRODUCTION Statement of problem While the vegetation fire episodes of 1997 and 1998, including subject of country-level air quality monitoring programs and health surveillance activities, were documented in the media, there has not been a comprehensive regional assessment of air quality or health impacts that incorporates the various country-level data sources. The closest available assessment, which evaluates the impacts on a regional scale, is the EEPSEA economic costs report (http://www.eepsea.org/publications/research1/ACF62.html), although this does not include an in-depth assessment of exposure or an evaluation of data quality. The project described in this report was conducted to assimilate available air quality and health impact data for the region that were collected during the episode periods. It is believed that Health and Environment ministries in the affected countries would benefit from a more detailed assessment of the concentrations and exposures than previously available. A regional determination of concentrations and population exposure will help support the fire and pollution prevention activities of authorities within the affected countries. Remote sensing to detect fires in this region has been used successfully for the past 15 years (Mallingreau, 1985) and was critical in 1997 to locate fires, track the transport of smoke plumes and to identify large plantations as major fire sources (Wooster, 1998; Fang, 1998). Here, we assess the feasibility of using remote sensing to estimate surface-level concentrations of particulate air pollution generated from vegetation fires. This project is an initial test of the use of remote sensing as a tool to rapidly assess the health impact of such large-scale vegetation fire air pollution episodes. In this project we describe the methodology needed to use remote sensing data in the analysis of human exposure and health impacts. As such, the methodology developed in this project can be used in future preventive efforts. Additionally, by conducting a regional assessment of the air pollution exposures, and specifically the extent of the population currently covered by ground-based air monitoring stations, we can identify areas which were not adequately monitored. This information can be used by national governments as well as by regional organizations (ASEAN, APEC) as guidance for location of future monitoring stations within the region.  8  Accordingly, this project was initiated with the following objectives:  Objectives To use remote sensing (satellite) data, in combination with traditional ground-based air monitoring data, to develop a regional database of particulate air pollution levels associated with the Southeast Asian forest fires 1997-98. To combine the regional particulate database with population density and other demographic data, to assess population exposure to air pollution associated with the Indonesian forest fires To gather available data from affected countries (Malaysia, Thailand, Singapore, Indonesia, Brunei) within the region to assess the feasibility of conducting a regional assessment of health impacts To conduct a regional assessment of the health impacts of the regional air pollution episodes occurring in 1997-1998 in Southeast Asia, using local data or a risk assessment approach.  Background In 1997, uncontrolled forest fires burning in the Indonesian states of Kalimantan and Sumatra, in combination with a severe regional drought, depressed mixing heights and prevailing winds resulted in smoke from these fires impacting the fire areas themselves in Kalimantan and Sumatra, as well as peninsular Malaysia and the state of Sarawak on the island of Borneo. In particular, the major population centers of Kuala Lumpur (population approximately 2 million) and Kuching (population approximately 400,000) were affected. Other countries in the region, including Singapore, southern Thailand, Brunei and The Philippines were also impacted. Biomass smoke pollution from the fires resulted in elevated levels of particulate air pollution between June and November in many areas, with a severe episode occurring during most of the month of September. As early as the beginning of May, air quality began to deteriorate in Singapore (Nichol, 1998) (Figure 1). 24-hour average concentrations in Singapore peaked at 225 µg/m3 on October 2, 1997. In Sarawak, the 24-hour average PM10 measured in Kuching peaked at high as 852 µg/m3 on September 22, which represents a value more than 15 times the normal levels (Brauer and 9  Hashim-Hisham, 1998) (Figure 2). In Kuala Lumpur, concentrations of 420 and 411 µg/m3 were measured on September 14 and 15, respectively (Figure 3). Due to the location of fires and direction of the wind, Sumatra and Kalimantan in Indonesia were the most severely affected areas (Pinto et al., 1998), although locations in southern Thailand were also affected for short periods (Figure 4). Figures 1-4 present time series plots of measured PM10 concentrations in Singapore, Kuching (Malaysia), Kuala Lumpur and Phuket (Thailand).  Singapore (Site P11) 250  PM10 (ug/m3)  200 150 100 50  01/04/98  01/03/98  01/02/98  01/01/98  01/12/97  01/11/97  01/10/97  01/09/97  01/08/97  01/07/97  01/06/97  01/05/97  01/04/97  01/03/97  0  Date Figure 1. Measured 24-hr average PM10 concentrations at Singapore Site P11 for reporting period March 1, 1997- April 30, 1998.  10  01/04/98  01/03/98  01/02/98  01/01/98  01/12/97  01/11/97  01/10/97  01/09/97  01/08/97  01/07/97  01/06/97  01/05/97  01/04/97  900 800 700 600 500 400 300 200 100 0 01/03/97  PM10(ug/m3)  Kuching  Date Figure 2. Measured 24-hr average PM10 concentrations at Kuching, Sarawak, Malaysia monitoring site for reporting period March 1, 1997- April 30, 1998.  450 400 350 300 250 200 150 100 50 0 01 /0 3 01 /97 /0 4 01 /97 /0 5 01 /97 /0 6 01 /97 /0 7 01 /97 /0 8 01 /97 /0 9 01 /97 /1 0 01 /97 /1 1 01 /97 /1 2 01 /97 /0 1 01 /98 /0 2 01 /98 /0 3 01 /98 /0 4/ 98  PM10 (ug/m3)  Kuala Lumpur  Date Figure 3. Measured 24-hr average PM10 concentrations at Kuala Lumpur, Malaysia monitoring site for reporting period March 1, 1997- April 30, 1998. 11  Phuket 250  PM10 (ug/m3)  200 150 100 50  01/04/98  01/03/98  01/02/98  01/01/98  01/12/97  01/11/97  01/10/97  01/09/97  01/08/97  01/07/97  01/06/97  01/05/97  01/04/97  01/03/97  0  Date Figure 4. Measured 24-hr average PM10 concentrations at Phuket, Thailand monitoring site for reporting period March 1, 1997- April 30, 1998.  Although routine monitoring of ambient PM10 was not conducted throughout Indonesia, daily averaged concentrations of PM10 were estimated to have reached as high as 3546 µg/m3 (based on a TPM concentration of =3940 µg/m31) in Sumatra at the end of September (Heil, 1998). Concentrations were estimated to have been even higher in Kalimantan (Indonesia), with a 24hour maximum of 3645 µg/m3 (TPM=4050 µg/m3) (Heil, 1998). Even at the beginning of November in Palembang (Kalimantan) when the air quality was improving, daily PM10 and PM2.5 still exceeded the US National Ambient Air Quality Standards of 150 and 65 µg/m3, respectively (Pinto et al., 1998). Aircraft measurements performed during this episode indicated that the smoke plume reached an altitude of 4 km and included high concentrations of O3, NOx and CO along with aerosols. In the lower layer of the plume, visibility was less than 500m (Tsutsumi et al, 1999). 1  Concentrations in Indonesia are estimated based on available TPM (total particulate matter) measurements; no PM10 measurements were conducted during the 1997 episode. PM10 concentrations are assumed to account for 90% of the TPM, based on information provided by Ward, 1990 and USEPA.  12  With the persistence of the El Niño that started in 1997 and an abnormally short wet season (Levine et al, 1999), vegetation fires impacted regional air quality again in the spring of 1998 although this haze event was more localized (Radojevic and Hassan, 1999). In this case, the episode was most acute on Borneo and particularly in Brunei. Deteriorated air quality began in early February and remained until the end of April, with especially severe conditions at the beginning of April. (Figure 5)  Brunei Muara District 300  PM10 (ug/m3)  250 200 150 100 50  01/04/98  01/03/98  01/02/98  01/01/98  01/12/97  01/11/97  01/10/97  01/09/97  01/08/97  01/07/97  01/06/97  01/05/97  01/04/97  01/03/97  0  Date Figure 5. Measured 24-hr average PM10 concentrations at Brunei Muara, Brunei-Darussalam monitoring site for reporting period March 1, 1997- April 30, 1998. Daily average concentrations in the capital of Brunei, Bandar Seri Begawan, reached nearly 450 µg/m3 (Radojevic and Hassan, 1999). Radojevic and Hassan have argued that, during the 1998 episode in Brunei, PM10 was the only significant pollutant of the “classical” air pollutants, contributing to the haze (Radojevic and Hassan, 1999). Other gaseous pollutants such as SO2, O3 and NO2 were within acceptable limits, and only the 8 hr guideline for CO was exceeded on several occasions (Radojevic and Hassan, 1999). In Miri, Sarawak, Malaysia, 24-hour average  13  PM10 concentrations reached 666 µg/m3 on March 30, 1998 and elevated concentrations were measured in Brunei from late March until the middle of April (Figure 5).  Miri 700  PM10 (ug/m3)  600 500 400 300 200 100 01/04/98  01/03/98  01/02/98  01/01/98  01/12/97  01/11/97  01/10/97  01/09/97  01/08/97  01/07/97  01/06/97  01/05/97  01/04/97  01/03/97  0  Date Figure 5. Measured 24-hr average PM10 concentrations at Miri, Sarawak, Malaysia monitoring site for reporting period March 1, 1997- April 30, 1998. While these two most recent episodes have received the most international attention and arguably produced the largest decrements in regional air quality, they are not the only air pollution episodes to occur in South East Asia. Brook has used visibility and meteorological data to reconstruct a PM10 data set for 1978-1997 at Kuching (Brook 1998)(Figure 6).  This  reconstructed dataset shows other PM10 episodes in 1982, 1983, 1991 and 1994 which correspond with other large-scale vegetation fire events (Harger, 1995).  14  800  Predicted 24 h Average PM10 (µg/m3)  700  600  500  400  300  200  100  01/01/97  01/01/96  01/01/95  01/01/94  01/01/93  01/01/92  01/01/91  01/01/90  01/01/89  01/01/88  01/01/87  01/01/86  01/01/85  01/01/84  01/01/83  01/01/82  01/01/81  01/01/80  01/01/79  01/01/78  0  T im e  Figure 6. Estimated PM10 concentrations in Sarawak from 1978 to 1997 based upon visibility reports at the Kuching airport. Adapted from Brook, 1998 (Brook 1998). Based on these estimates, other episodes of particulate air pollution are evident, and these are thought to typically coincide with periods of El Nino Southern Oscillation (ENSO). Therefore, there is evidence that such regional air pollution episodes are recurrent and likely to reoccur in the future during periods of regional drought. It is important therefore to assess the regional impacts of these episodes and to analyze the ability of available exposure and health monitoring systems to characterize the impacts.  Health Impacts Information on the health impacts in the relatively sparsely populated areas of Indonesia itself were generally unavailable, although media reports suggest that, as expected, many individuals in the regions where fires were burning were severely affected. For areas of higher population density, the little information that has already been evaluated suggests that the magnitude of effects were large. During the 1994 biomass smoke episode, emergency room visits for asthma at two major hospitals in Singapore, more than 500 km from the source of the smoke, increased in association with  15  increased particulates from the fires (Chew, et al., 1995). Similarly, in 1994, a study of 366 asthmatics in Singapore indicated that 28%, comprised mainly of the more severe asthmatics, reported a higher frequency of asthma attacks during the haze period (Chia, 1995). In 1997, the Malaysian Ministry of Health compiled data for 3 routinely monitored diagnoses, asthma, total acute respiratory infections and conjunctivitis, during the months of August and September at a number of major hospitals in Kuala Lumpur. Emergency room visits for the State of Sarawak were also recorded (Brauer, 1998; Leech, 1998). While these data allow one to qualitatively assess the impact of air pollution and to quantitatively assess the impact on these three diagnoses, the lack of more comprehensive data regarding all outpatient visits, or all hospital admissions, precludes a more complete assessment of health impacts. Initial analyses of these data indicate a clear relationship between PM10 concentrations and respiratory cases at Hospital Kuala Lumpur (Figure 2). Similar data have also been reported for the state of Sarawak (Brauer, 1998; Leach, 1998). Unfortunately the severity of these cases are not known, nor do they represent the entire populations. Similar data were also collected in Singapore where the Ministry of Health reported a 13% increase in visits to government clinics for acute respiratory infections and a 19% increase for asthma visits during the last week of September when the particle levels peaked. Preliminary results from an on-going study of 107 Kuala Lumpur school children found statistically significant decreases in lung function between pre-episode measurements in June-July 1996 and measurements conducted during the episode in September 1997 (Hisham-Hashim, 1998). These preliminary results suggest a measurable impact of the 1997 episode on the respiratory function of children. However, it is yet uncertain whether these respiratory impairments are permanent. In a study of young military recruits in Singapore, Tan et al, measured an association between air pollution during the 1997 episode and peripheral white blood cell counts. Air pollution was associated with elevated banded neutrophil counts, an indicator of a systemic inflammatory response (Tan, Qiu et al. 2000). Together, this rather limited information indicates measurable and widespread acute impacts on the health of the affected population. Analyses of the long-term health impacts of the air pollution and the impact with more severe outcomes, such as daily mortality increases, have not been conclusive to date.  16  TOMS - Introduction The total ozone mapping spectrometer (TOMS) is operated by the National Aeronautics and Space Administration (NASA) and is a second generation backscatter ultraviolet ozone sounder (NASA, 1997). The TOMS instrument observes the incoming solar energy and backscattered ultraviolet (UV) radiation at six wavelengths. The wavelengths at which it measures are 308.60, 313.50, 317.50, 322.50, 331.20, and 360.40 nanometres (nm) (McPeters, Bhartia et al. 1998). Every 8 seconds, the instrument makes 35 measurements, each covering 50 to 200 kilometres wide on the ground. The TOMS data used in this study was taken aboard the Earth Probe satellite (referred to as EP/TOMS). The EP/ TOMS was the only instrument aboard the satellite. The satellite was launched on July 2, 1996 and data collection began on July 25, 1996. This was only a few months before the launch of ADEOS, a Japanese meteorological satellite. Initially, the Earth Probe satellite was placed in an orbit of 500 km in order to provide higher spatial resolution. The ADEOS satellite though failed on June 29, 1997 so the EP/TOMS was raised to an orbit of 750 km for more complete global coverage (McPeters, Bhartia et al. 1998). Coverage is global except for near the poles where the sun stays close to the horizon for half of the year (NASA, 1997). The Goddard Space Flight Center (GSFC) Distributed Active Archive Center (DAAC) houses the archive of TOMS products, and the data is also available in near real time on the TOMS website (http://toms.gsfc.nasa.gov/) (McPeters, Bhartia et al. 1998).  Aerosol index Although the TOMS instrument's primary usage is for detecting atmospheric ozone, it is also possible to detect smoke from biomass burning, desert dust, and sulphur dioxide and ash from large volcanic eruptions (McPeters, Bhartia et al. 1998), (Hsu, Herman et al. 1996), (Torres 1998). The reflectivity difference between the 340 and 380 nm channels is sensitive to these tropospheric absorbing aerosols, and gives a numerical value that is a measure of the amount of absorbing aerosol in the atmosphere. (Hsu, Herman et al. 1996). The TOMS instrument is  17  capable of differentiating between these types of absorbing particles and non-absorbing aerosols such as clouds and haze (Hsu, Herman et al. 1996). The formula for calculating the residue is: rλ = -100[log(Iλ/I380)meas - log(Iλ(Ωλ,R380)/I380(R380))calc] where Ω is the measured total ozone amount and R380 is the 380nm reflectivity (Torres 1998). Absorbing aerosols produce a positive residual that increases with the aerosol optical depth (Torres 1998). This formula works over both dark and bright surfaces for absorbing aerosols and thus can be used over land and ocean (Torres 1998).  Surface monitoring The particle data used in this analysis were collected at monitoring stations in Malaysia, Singapore, Thailand and Brunei and are reported as concentrations of particulate matter less than 10 µm (PM10) in units of micrograms per cubic meter (µg/m3). The data compiled here represent the most comprehensive dataset from South East Asia during the 1997-1998 vegetation fires. The period of coverage depends on the duration of coverage of specific stations, although data is available in most locations for the entire study period which we defined in our initial adat requests as March 1, 1997 through April 30, 1998. In total, 32 stations were used for this study. Table 1 summarizes the period of data available by country, though it varies from station to station. The table also gives information on the method of PM10 monitoring as well as the source of the data. All of the stations are plotted on the map in Figure 7.  18  Country  Method  Source of data  Duration  Malaysia  Number of stations 20  B-attenuation  Thailand  3  B-attenuation  03/01/9704/31/98 03/01/9703/31/98  Brunei  4  TEOM  Department of Environment Ministry of science Technology and Environment, Pollution Control Department, Air Quality and Noise Management Division Ministry of Health  Singapore  5  TEOM  Ministry of the Environment, Quarantine and Epidemiology Department  09/01/9704/31/98 03/01/9704/31/98  Table 1. Data sources and monitoring methods for PM10 data. TEOM: Tapered Elemental Oscillating Microbalance.  19  Figure 7. Locations of PM10 monitoring stations in Southeast Asia that were used in this report. Data sources and monitoring methods are indicated in Table 1. Stations P03, P05, P06, P09, P11 are located in Singapore.  20  In Indonesia, different government ministries conduct air pollution measurements and thus the data was hard to retrieve. A limited amount of total suspended particulate (TSP) data from Indonesia was available (Heil 1998). These data have not been used in the current study as the spatial and temporal coverage was very discontinuous. Also, we chose to focus on PM10, although the same type of analysis can be undertaken to determine the relationship between TSP and the TOMS aerosol index. Types of monitors The two methods of monitoring surface PM10 in the countries of interest are by beta-attenuation and by Tapered Elemental Oscillating Microbalance (TEOM). For a beta-attenuation mass monitor, the process begins by measuring the beta ray transmission across a clean section of filter tape. The filter tape is then attached to the sampling inlet where particulate matter is drawn in and deposited onto the filter tape. The filter paper is returned to its original placement after the sampling period, where the beta ray transmission is re-measured. The difference between the original measurement with the clean filter and that of the particulate-laden filter determines the particle concentration (Met One Instruments 2000). A filter is also used in the TEOM method, although in this case a filter cartridge is attached to the free end of a tapered element. The tapered element has a natural resonant frequency related to its dimensions, material and mass. As particles become trapped in the filter, the mass of the tapered element/filter combination increases. This can be measured as a decrease in the resonant frequency. A frequency sensor measures the signal which can be converted into a mass measurement (Patashnick 2000).  METHODS Overview Assessment of air pollution concentrations and population exposure associated with the Southeast Asia regional air pollution episodes in 1997 and 1998 has not been previously conducted on a regional scale. Several countries within the region have assessed selected health impacts in major urban areas, although the study protocols were not standardized, nor were common air pollution measurement techniques used within the region. This report describes a regional estimation of population exposure to particulate air pollution and estimated health  21  impacts using a combination of remote sensing (Figures 17-26) and surface monitoring data. Population exposures are estimated by combining population density data in the affected areas of Southeast Asia with estimated particulate concentrations. In some areas, the surface monitoring network is relatively dense, while in others, little or no ground-based monitoring data were available for the periods of the air pollution episodes. The output of this level of the work is a series of daily maps of estimated particle levels within the region (Figures 27-36). These data are then combined with population density data (Figure 8, Figure 37) to estimate the daily counts of numbers of people exposed to different levels of air pollution within the region (Figures 38-44). For examples to display in this report we randomly selected 10 days from within the March 1, 1997 – April 30, 1998 study period to present as figures. These example days include both the 1997 and 1998 episode periods as well as non-episode periods. Although not shown in this report, similar data are available for all days during the study period. We were unable to conduct a risk assessment due to the uncertainties in exposure-response coefficients for vegetation fire-related air pollution which were applicable to the population of the Southeast Asian region.  Additionally, there was substantial uncertainty in the estimated  PM10 concentrations. These two factors lead to highly uncertain risk estimates.  Further,  adequate health outcome data were not available for a formal epidemiological analysis.  Data overview TOMS The TOMS instrument has been orbiting the earth and recording data since 1978. The aerosol data is available for most of the time since, except for a gap from 1993 to 1996. The level three data are available at the TOMS website: http://toms.gsfc.nasa.gov/. It is possible to obtain ASCII files, which are gridded into 1 degree latitude by 1.25 degree longitude zones. This corresponds to about 111 kilometres by 139 kilometres at the equator. The data is referred to as TOMS version 7 level 3 data and has been processed by the Version 7 Nimbus-7 and Meteor-3 TOMS algorithm. The data was subjected to a preliminary preliminary calibration before being made available publicly (McPeters, Bhartia et al. 1998). 22  PM10 measurements PM10 data was requested from Malaysia, Singapore, Thailand and Brunei. The sources of the data are listed in Table 1. The data from Malaysia, Singapore, and Brunei were provided as 24 hour averages. From Thailand, we received hourly values of PM10. It was necessary to convert these hourly values into daily concentrations. For these data, there was substantial variability within one 24 hour period; therefore only those stations which had a minimum of 16 out of the 24 hours were included in the generation of 24-hour averages  Population density The population density data (Figure 8) was obtained from the UNEP-GRID / National Center for Geographic Information and Analysis at the University of California, Santa Barbara (http://grid2.cr.usgs.gov/globalpop/asia/intro.html, accessed September 5, 2000). Input data for the region were, in most cases, taken from the most recent (pre-1995) national level census data and projected to estimate the 1995 population density. These country-level data were then aggregated into a regional map. Specific information regarding the sources of the population density  data  for  individual  countries  http://grid2.cr.usgs.gov/globalpop/asia/appendix.php3.  is  available  at  The resolution of the data map is  approximately 5 km at the equator (2.5 minutes).  23  Figure 8. Southeast Asia population density map. Source: UNEP-GRID / National Center for Geographic Information and Analysis at the University of California, Santa Barbara  Processing of TOMS data After extracting data in the latitude bands from 14.5 to -14.5 (to limit the amount of unnecessary data), the files were processed using a FORTRAN program which extracted the data corresponding to the PM10 measurement station data. This process was repeated for each day from March 1, 1997 to April 30, 1998. The Earth Probe satellite is a polar orbiting satellite, and as such, the global coverage is not complete. In order to produce a more continuous estimate of the PM10 concentrations and population exposure, it was necessary to create a composite image coverage consisting of a combination of three day’s worth of data. For each day and each pixel, if there was an aerosol value, that was selected; otherwise an average of the day before and the day after was calculated. If data were lacking from the day before and after either one of those values was taken.  For  example, Figures 17-26 indicate the aerosol index values for the same days as the estimated  24  PM10 concentrations shown in Figures 27-36. The black areas in Figures 17-26 indicate areas where there was no satellite coverage on that particular day. A Geographic Information System (GIS) (ArcView GIS 3.2) was employed in order to estimate population exposures. To import the satellite data into the GIS, it was necessary to convert the rectangular pixels into square pixels. Thus instead of having 1° by 1.25° pixels, they were linearly interpreted to become 1.25° by 1.25° pixels. The aerosol index values were converted to PM10 concentrations via the regression equation that best suited the data. In order to estimate the uncertainty in the aerosol index – PM10 relationship, plots of the 95% confidence interval of the slope were also included. The estimated concentrations were overlaid with the measured values from stations (as points) where data was available. The population data was then incorporated and the number of people within the pixels exposed to ambient concentrations above values of 150, 250, 350, 450 µg/m3, etc were counted.  Processing of PM10 data As we are dealing with spatial aspects of air pollution data and since the pixel is the basic unit in our analysis of the TOMS data, it is crucial to consider the sub-pixel variability of measured PM10 concentrations. For our dataset, there are 8 pixels which contain only one station but the 24 remaining stations are located in only 9 different pixels. In order to characterize the sub-pixel variability, a regression model was developed for each pair of stations within a single pixel. Ideally, there should be a 1:1 relationship such that if the stations were plotted against each other there would be a slope of 1 and no intercept (which would indicate that the measured PM10 is constant within the area of the pixel). The regression was forced through the intercept and we then tested (with a t-test) the hypothesis that the slope of the line equals one. This test revealed that only 4 pairs of stations could indeed be considered to be the same. The greatest discrepancy was between Miri (a station in Malaysia) and Tutong and Kuala Belait (stations in Brunei). The difference between the PM10 measured at Miri and Kuala Belait reached as high as 574 µg/m3 in 1998. This is due to local fires near Miri that had a great influence on the local air quality. This lack of agreement suggests that the TOMS data, which are limited to the spatial resolution  25  of 1 pixel (1 degree x 1.25 degrees), are inadequate to differentiate pollution generated from local fires. However, for most regional pollution episodes, such as those involving transported air pollution, the TOMS spatial resolution appears to be adequate.  Comparisons of TOMS/PM10 In order to determine the relationship between particulate matter less than 10 µm in diameter (PM10) and the Total Ozone Mapping Spectrometer (TOMS) aerosol index values, the first step was to extract pairs of values. Once the aerosol index values were matched with the measured PM10 data, only those values which had a corresponding match were retained. This produced a set of data with 6835 pairs of values. Linear regression was performed on the data set to illuminate a relationship between particulate matter and the aerosol index. From these linear regressions, the important parameters of interest were the slope of the regression line, the intercept and the coefficient of determination (R2). Comparisons of TOMS and PM10 were performed for each station, for each country and for the entire data set.  Generation of estimated PM10 maps To estimate PM10 concentrations across South-East Asia, the TOMS data were imported into a GIS. The data were then converted from aerosol index values to PM10 using the results of the statistical regressions. As the regression was not forced through zero, it is possible that negative estimated concentration values will result. To deal with this possibility, we estimated a measurement detection limit of 23 ug/m3 which was 3*the standard deviation of the lowest nonepisode mean concentration value (7.6 ug/m3 measured in Singapore). This represents the lowest concentration that we felt could be reliably distinguished from zero for this dataset. All estimated concentrations below this detection limit were set to 16 ug/m3, which is equal to ½ 2 * the detection limit (Hornung and Reed, 1990). Measured concentrations below this were  26  retained, as this represents the detection limit for the levels estimated from the aerosol index only, while the measurement devices have lower detection limits. In addition to using the regressions, the actual measured values were also used to estimate population exposure for the ten example dates shown in Figures 27 – 36. For those pixels where a concentration was measured, that value was used instead of the value based upon the regression relationship. This represents a slightly more accurate estimate, although there are errors in assuming that the concentration remains the same throughout the pixel. Note that the maps of estimated concentrations (Figures 27-36) indicate both the measured (as colored points) and estimated (as grids) concentrations for these pixels  RESULTS AND DISCUSSION A comparison of the data from episode periods to non-episode periods is demonstrated via the change in the average concentration between the periods. The definition of the episode is a subjective process, as PM10 concentrations rose in certain regions at different times. Here, the episode is arbitrarily defined in 1997 as being from the beginning of August until the middle of November, which definitely incorporates the extent of the episode. The 1998 episode was taken to be from the middle of March until the end of the data availability, which is April 30th. It is apparent from the PM10 monitoring data summary statistics (Table 2) that in 1997 the vegetationfire-related air pollution mainly affected Sumatra (for example sites in Singapore and on the west coast of mainland Malaysia near Sumatra) parts of Sarawak and Kalimantan on Borneo, while in 1998, elevated concentrations were restricted to eastern Sarawak and Brunei on the island of Borneo. Monitoring data on Borneo outside of the Kuching monitoring site were not available until after the peak of the 1997 episode. Therefore, the measured PM10 data alone provide an inaccurate account of the spatial extent of the episode. This example illustrates the importance of the TOMS data during such periods and for locations without surface-based air monitoring data. It should also be noted that the average non-episode concentration for Miri Borneo is quite high, this is because local air quality periodically deteriorated as early as February at this location, before our arbitrary start of the 1998 episode.  27  Suratthani  Episode 1 45.2 Mean 84.0 number 18.0 SD 0.0 Min 106.7 Max Episode 2 62.0 Mean 27.0 number 14.1 SD 39.9 Min 113.6 Max Non episode 46.2 Mean 182.0 number 15.9 SD 15.9 Min 105.1 Max Combined episode 49.3 Mean 111.0 number 18.5 SD 0.0 Min 113.6 Max  Site  68.5 85.0 28.1 15.5 159.1 20.1 4.0 2.4 16.9 22.5 48.6 180.0 21.0 16.9 185.3 66.3 89.0 29.2 15.5 159.1  43.1 23.0 5.9 30.3 54.0 44.7 235.0 15.4 19.8 127.3 56.7 125.0 25.7 20.6 196.6  Hatyai  59.8 102.0 27.5 20.6 196.6  Phuket  54.0 168.0 25.1 13.0 148.0  33.6 257.0 10.8 13.0 85.0  58.1 61.0 18.5 31.0 111.0  51.6 107.0 28.0 13.0 148.0  Kota Bharu  61.3 156.0 40.3 23.0 298.0  45.5 256.0 19.7 21.0 153.0  52.3 49.0 7.8 38.3 66.5  65.4 107.0 47.9 23.0 298.0  Sg Petani  87.4 168.0 52.9 32.0 406.0  70.4 255.0 24.8 32.0 207.0  70.9 61.0 11.4 50.3 103.1  96.7 107.0 64.0 32.0 406.0  S Perai  65.2 167.0 43.8 21.5 284.6  46.4 251.0 20.7 21.5 191.0  53.1 61.0 7.8 37.0 72.5  72.2 106.0 53.5 21.5 284.6  Taiping  88.4 168.0 50.6 27.0 289.0  60.6 255.0 19.0 24.1 146.0  73.1 61.0 12.7 45.0 108.5  97.2 107.0 61.1 27.0 289.0  Ipoh  60.0 120.0 50.8 16.0 276.5  27.8 174.0 11.3 9.8 58.9  49.0 61.0 27.4 20.4 140.1  71.3 59.0 65.3 16.0 276.5  Jerantut  28  65.9 166.0 42.5 18.0 255.0  37.6 246.0 14.1 11.0 99.0  49.8 61.0 17.9 27.1 111.1  75.2 105.0 49.4 18.0 255.0  Kemaman  48.0 167.0 31.6 16.0 173.0  23.8 258.0 9.4 11.0 65.0  36.1 61.0 16.1 18.3 113.9  54.9 106.0 36.0 16.0 173.0  Kuantan  Kelang  Episode 1 144.9 Mean 107.0 number 72.2 SD 45.3 Min 408.2 Max Episode 2 97.7 Mean 61.0 number 21.2 SD 58.4 Min 150.7 Max Non episode 79.0 Mean 258.0 number 27.1 SD 26.5 Min 203.3 Max Combined episode 127.8 Mean 168.0 number 63.2 SD 45.3 Min 408.2 Max  Site  161.1 107.0 78.2 60.0 420.0 89.9 61.0 15.2 59.5 123.3 92.1 229.0 25.5 38.4 186.0 135.2 168.0 71.7 59.5 420.0  58.7 61.0 14.6 27.5 93.4 44.5 149.0 25.1 14.0 139.4 85.7 168.0 55.2 12.2 240.8  Kuala Lumpur  101.1 107.0 63.4 12.2 240.8  Shah Alam  96.2 168.0 59.0 36.8 366.3  54.9 153.0 21.0 23.3 130.1  68.0 61.0 13.3 37.0 93.7  112.3 107.0 68.3 36.8 366.3  Kajang  111.4 168.0 70.8 38.9 491.0  54.5 254.0 23.8 21.7 162.0  77.4 61.0 18.6 38.9 120.3  130.7 107.0 81.6 40.0 491.0  Nilai  117.3 167.0 65.4 40.0 393.0  60.2 253.0 28.4 10.0 156.0  81.5 61.0 19.2 49.9 130.0  138.0 106.0 73.3 40.0 393.0  Melaka  69.4 168.0 38.0 21.3 201.0  34.5 253.0 12.6 10.2 104.0  40.6 61.0 11.9 21.3 66.9  85.8 107.0 38.1 27.0 201.0  Johor Bharu  72.3 152.0 38.1 29.7 213.7  38.2 238.0 12.0 13.0 113.6  49.4 61.0 12.6 29.7 83.3  87.6 91.0 41.6 36.7 213.7  P06  45.0 61.0 11.6 22.6 74.5  24.9 85.0 7.9 8.5 49.4  45.0 61.0 11.6 22.6 74.5  no data no data no data no data no data  P09  29  61.1 102.0 39.8 22.1 225.0  28.3 120.0 9.9 8.2 77.0  42.0 61.0 11.6 22.1 74.4  89.5 41.0 49.1 30.0 225.0  P03  47.4 82.0 28.0 16.3 156.0  24.6 113.0 7.6 9.5 64.6  37.6 61.0 12.1 16.3 73.4  75.8 21.0 40.0 26.0 156.0  P05  P11  71.6 40.0 38.8 30.1 193.8 57.3 61.0 15.0 33.8 99.7 37.8 103.0 10.8 11.0 63.5 63.0 101.0 27.8 30.1 193.8  51.3 61.0 15.7 28.4 106.1 43.0 252.0 26.0 18.0 206.0 109.2 168.0 139.2 25.0 852.0  Sarikei  142.1 107.0 165.4 25.0 852.0  Kuching  75.9 104.0 24.7 36.9 185.9  60.1 104.0 14.8 34.3 94.8  78.1 60.0 21.0 42.3 155.7  73.0 44.0 29.1 36.9 185.9  Sibu  88.7 107.0 44.7 30.3 286.3  59.5 104.0 23.9 23.5 167.2  104.1 61.0 48.9 53.6 286.3  68.2 46.0 27.6 30.3 167.2  Bintulu  201.8 107.0 197.2 24.5 666.0  104.3 105.0 101.7 22.4 444.0  324.3 61.0 182.1 41.4 666.0  39.5 46.0 12.9 24.5 87.3  Miri  68.1 110.0 60.5 10.0 243.0  24.0 63.0 13.7 1.0 71.0  108.7 59.0 56.0 27.0 243.0  21.0 51.0 11.2 10.0 58.0  Kuala Belait  74.9 112.0 66.8 10.0 262.0  23.0 71.0 14.1 6.0 69.0  118.3 61.0 62.7 36.0 262.0  23.0 51.0 12.4 10.0 60.0  Tutong  53.3 105.0 43.9 10.0 191.0  21.8 102.0 12.1 4.0 67.0  82.5 58.0 39.1 20.0 191.0  17.2 47.0 7.8 10.0 46.0  Temburong  period: March 15-April 30, 1998.  Thailand, Malaysia, Singapore and Brunei which provided data. 1997 episode period: August 1-November 15, 1997. 1998 episode  30  58.1 106.0 53.9 10.0 253.0  24.7 55.0 16.0 5.0 69.0  93.9 54.0 54.6 25.0 253.0  21.0 52.0 10.7 10.0 57.0  Brunei Muara  Table 2. Summary statistics for PM10 concentrations during “episode” and “non-episode” periods for all monitoring stations in  Episode 1 84.4 Mean 104.0 number 41.1 SD 27.8 Min 221.6 Max Episode 2 44.3 Mean 61.0 number 13.5 SD 24.3 Min 84.9 Max Non episode 32.4 Mean 201.0 number 12.0 SD 8.7 Min 96.5 Max Combined episode 69.6 Mean 165.0 number 38.8 SD 24.3 Min 221.6 Max  Site  Table 3 presents the results of the individual regressions between the TOMS aerosol index and the measured PM10 concentrations for each station. There was substantial variability within the slopes of the lines (from as low as 0.32 to 10.23) as well as the intercepts, indicating the dependence of these relationships on specific site data.  Location-specific  variables which may affect the relationship between the TOMS aerosol index and the measured concentrations include the range in the measured concentrations, the impact of other air pollution sources and meteorological factors which affect the height of the aerosol column in one particular location. For example, low slopes wee found for many of the northern monitoring sites (in Thailand, Kota Bharu, Sg. Petani) which experienced a relatively small impact of the episode while Kuching and Miri where the highest concentrations were measured had much higher slopes. Figure 9 presents a scatter plot of the TOMS aerosol index and measured PM10 data for all of the sites. These data were analysed to firstly determine if a statistically significant relationship between the TOMS aerosol index and measured PM10 concentrations existed. Linear regression using all of the sites indicated a statistically significant relationship of the form: PM10 = 4.72 x aerosol index + 55.64 ug/m3 (R2=0.28). The 95% confidence interval for the slope was 4.54 - 4.89.  The coefficient of  determination (R2) indicates the proportion of variation in PM10 that is explained by linear regression of PM10 on the TOMS aerosol index. Although there is evidence of a relationship between the two measurements, there must be other factors that are influencing the variables. Relationships improved somewhat when data were stratified by region and by episode / nonepisode periods (Table 3). As much as 60% of the variability was explained for some locations (Table 4).  31  900 800 700  PM10 (ug/m3)  600 500 400 300 200 100 0 -20  -10  0  10  20 Aerosol Index  30  40  50  60  Suratthani Phuket Hatyai Kota Bharu Sg Petani S Perai Taiping Ipoh Jerantut Kuantan Kemaman Kelang Shah Alam Kuala Lumpur Kajang Nilai Melaka Johor Bharu P06 P09 P03 P05 P11 Kuching Sarikei Sibu Bintulu Miri Kuala Belait Tutong Temburong Brunei Muara Dist  Figure 9. Scatterplot of all 24-hour average PM10 and aerosol index values for all data points and all measurement locations (N=6835).  While statistically there is a relationship between the TOMS aerosol index and measured PM10 concentrations, an examination of the scatterplot (Figure 9) of the data suggests the potential for better results. A closer look at the data reveals that there are four days with exceptionally high concentrations of PM10 in Kuching during September, 1997 (Septemebr 19 = 691; September 22=852; September 23=783; September 24=701 ug/m3). Although there is no maximum value for the aerosol index, as the amount of aerosol increases, the plume becomes opaque, after which it would not be possible for the TOMS sensor to measure the entire amount of aerosol (Jay Herman, NASA, personal communication, 2000). This is likely the case over Kuching during the episode when PM10 levels were extremely high. . As the TOMS instrument would not be able to accurately measure an aerosol index for concentrations such as these, the four data points were removed from the data set. It seems obvious that these points are outliers and are thus having a great influence on the regressions. By removing the four points, the regressions changes to yield an equation of y = 4.41x + 55.76. The coefficient of determination reduces slightly to 0.27 after these influential points are removed .  32  Another issue that stands out in the scatterplot is that the values in Miri don't agree with the regression as well as those from other stations. Data from Miri are available from October 1, 1997 through to April 30, 1998. The particulate values in Miri start to increase early in 1998 as a result of local fires.  These fires had a lot of influence on the local air quality  measurements but not so much on the satellite measurements. It is important to keep in mind that the satellite measures in zones of 1 degree by 1.25 degree which is equivalent to approximately 110.6 kilometres by 137.5 kilometres. The resolution is very coarse and will not be able to  detect local effects. Also, as mentioned previously in this report, the  particulate data from Miri had the worst agreement with the other stations within the same pixel. This discrepancy along with the observation that the Miri data do not follow the same linear trend as the rest of the data leads to the conclusion that local effects overshadow any measurements the satellite is able to make. If we were to consider the Miri data to be unrepresentative for this study and remove it for sensitivity analysis, there is an improvement in the regression. By removing the Miri values, the regression changes to yield a coefficient of determination of 0.35, which is an improvement over the initial analysis. By removing both the extreme values in Kuching and the Miri values, the regressions gives an r2 value of 0.33. Note that substantial scatter remains in the TOMS – PM10 relationship. However, it is also important to note that the value of the slope does not vary dramatically after the extreme values are removed. Apparently, the slope indicating the relationship between the aerosol index and the PM10 measurements for the range of data examined here (up to 500 µg/m3) is quite stable. Another possibility for interpreting the data is to look at each episode period separately. The vegetation fires started during the dry season and lasted until the monsoon rains came (late) in November in 1997. The following year the fires affected air quality between late January and April. For the purpose of this analysis, the episode is defined as the period during which the aerosol index remained above 1.0 within the region. This differs from how the episode was defined previously in this paper as here the focus is on an increase in aerosol index values, while above, the episode was defined based on increases in particulate concentrations. During 1997, this comprises the dates between September 3 to November 16. For 1998, the episode lasted from February 6 to April 21. The regression values for the episodes along with all the other breakdowns are listed in Table 3.  33  Data Slope Intercept R2 All data 4.72 55.6 0.28 Extreme 4.41 55.8 0.27 Kuching values removed Miri data 4.36 53.7 0.35 removed Miri and 4.02 53.8 0.33 extreme Kuching values removed Episode 1 only 4.70 53.1 0.44 Episode 2 only 5.96 61.8 0.21 Table 3. Summaries of PM10-aerosol index regressions for “episode” and “non-episode” periods and after excluding extreme values. For this analysis, episode periods are defined as those days in which the aerosol index value is greater than 1.0  34  Station Slope Intercept R2 All stations 4.72 55.6 0.28 Surrathani 0.32 48.2 0.00 Phuket 2.47 48.2 0.15 Hatyai 1.30 54.4 0.04 Kota Bharu 2.64 40.8 0.19 Sg Petani 3.37 52.9 0.39 S Perai 4.59 78.0 0.43 Taiping 4.40 52.6 0.50 Ipoh 5.24 70.7 0.55 Jerantut 5.11 37.4 0.48 Kemaman 4.68 44.0 0.50 Kuantan 2.93 27.7 0.46 Kelang 5.96 91.3 0.51 Shah Alam 4.46 57.5 0.38 Kuala Lumpur 6.49 98.7 0.58 Kajang 4.46 63.4 0.60 Nilai 5.19 64.6 0.53 Melaka 4.77 71.3 0.50 Johor Bharu 2.55 38.6 0.56 Singapore-P06 2.28 43.3 0.48 Singapore-P09 0.43 33.7 0.00 Singapore-P03 2.68 33.9 0.39 Singapore-P05 2.58 29.3 0.31 Singapore-P11 2.94 38.6 0.40 Kuching 8.64 40.4 0.58 Sarikei 3.37 47.1 0.43 Sibu 2.66 66.5 0.33 Bintulu 4.28 69.4 0.45 Miri 10.23 143.1 0.25 Kuala Belait 4.31 42.7 0.48 Tutong 5.66 41.8 0.64 Temburong 5.10 41.6 0.23 Brunei Muara 5.03 50.6 0.15 District Table 4. Summaries of individual PM10-aerosol index regressions for each monitoring location.  35  A summary of the differences between the values estimated using the regression relationship and the measured values is shown in Table 5.  Minimum Maximum Station Mean difference difference difference Suratthani -10 -60 58 Phuket -7 -86 76 Hatyai -5 -100 134 Kota Bharu -16 -77 41 Sg Petani -2 -109 83 S Perai 22 -45 149 Taiping -3 -62 103 Ipoh 15 -50 171 Jerantut -18 -111 174 Kemaman -12 -93 75 Kuantan -31 -149 43 Kelang 37 -36 187 Shah Alam 1 -100 180 Kuala Lumpur 45 -20 207 Kajang 7 -107 219 Nilai 10 -109 350 Melaka 16 -144 228 Johor Bharu -22 -160 36 Singapore - P06 -18 -156 51 Singapore - P09 -21 -56 21 Singapore - P03 -26 -116 43 Singapore - P05 -29 -150 40 Singapore - P11 -22 -139 126 Kuching -3 -128 584 Sarikei -10 -90 74 Sibu 8 -61 86 Bintulu 13 -58 108 Miri 104 -119 577 Kuala Belait -15 -156 150 Tutong -10 -160 137 Temburong -14 -132 98 Brunei Muara District -5 -132 174 Table 5. Difference statistics for estimated PM10 concentrations (ug/m3) at measurements sites (Measured – Predicted).  36  Figures 10-12 provide examples of time series for three randomly selected locations based upon the measured concentrations and the concentrations using the regression model. Use of upper or lower confidence intervals for the aerosol index – PM10 regression slope produced a predicted plot that was nearly identical to the mean slope that is shown in the figures. These figures show that the estimates do portray the temporal structure of the measured data and agree well with the measured values except for the highest concentrations. This is explained by our decision not to estimate concentrations above 500 µg/m3 as the aerosol index appears to be “saturate” above the concentrations, as described above. For example, in Figure 9 the high concentrations measured in Miri and Kuching are outliers from the rest of the data. There also appears to be some systematic differences between the measured and predicted values, for example the measured values are generally higher than the predicted values for Kuala Lumpur while the predicted values are lower than the measurements for Singapore. This is due to the use of a single slope based on all the measurement sites which is then applied to individual sites. Despite these errors the predicted value data series do capture the episode periods.  Kuala Lumpur 450  Measured  400  Predicted  300 250 200 150 100 50  01/04/98  01/03/98  01/02/98  01/01/98  01/12/97  01/11/97  01/10/97  01/09/97  01/08/97  01/07/97  01/06/97  01/05/97  01/04/97  0 01/03/97  PM10 (ug/m3)  350  Date  37  300 01/04/98  01/03/98  01/02/98  01/01/98  01/12/97  01/11/97  01/10/97  01/09/97  01/08/97  01/07/97  01/06/97  01/05/97  01/04/97  01/03/97  PM10 (ug/m3) 300  01/04/98  01/03/98  01/02/98  01/01/98  01/12/97  01/11/97  01/10/97  01/09/97  01/08/97  01/07/97  01/06/97  01/05/97  01/04/97  01/03/97  PM10 (ug/m3)  Singapore - P11 Measured  250 Predicted  200  150  100  50  0  Date  Brunei Muara Measured  Predicted  250  200  150  100  50  0  Date  38  The entire dataset was used to demonstrate the use of the regression results and to obtain estimates of population exposure. The measurements were brought into the GIS and plots of the estimated particulate concentrations were mapped as high (for the upper 95% confidence interval slope), mid (for the regular regression values) and low (based on the lower 95% confidence interval slope). The second layer represents the actual measured concentrations from those stations for which data were available. As the aerosol index appears to be “saturated” above approximately 500 µg/m3 (Figure 9) there are no estimated concentrations above 500 µg/m3; however, there are days in which concentrations above 500 µg/m3 were measured by PM10 monitoring stations. The last GIS layer includes the estimated population for Southeast Asia (for 1995); data are presented as population numbers in squares (the same size as the pixels) all across South East Asia. The estimate of population exposure is based on the first and last GIS layers. Overall, we estimate that 2.26 billion person-days of exposure (95% Confidence Interval: 2.17 – 2.35 billion person-days) to PM10 concentrations above 150 µg/m3 during the March 1 1997 – April 30, 1998 study period. 96% of these exposures occurred during the episode periods (as arbitrarily defined above). This is equivalent to 38% of the world’s population being exposed above 150 µg/m3 for one day2.  Approximately 20 million people were  exposed to ambient PM10 concentrations above 150 µg/m3 for two months and approximately 30 million people were exposed to PM10 concentrations above 150 µg/m3 for one month. For ambient air pollution, these episodes are remarkable for their duration on for the numbers of people that were exposed. Although very high concentrations were measured they appear to have been relatively rare, even considering our inability to estimate concentrations above 500 3  µg/m  . While this describes a major exposure to particulate air pollution, note these exposures  are still not comparable to the roughly 25%3 of the world’s population that is exposed to particle concentrations at this level or much higher every day as a result of cooking with biomass fuels on unvented stoves (Smith 1993) .  2  Based on 1998 world population of 5.90 billion. United Nations publication, (ST/ESA/SER.A/176), Sales No. E.99.XIII.6, Copyright (C) United Nations 1999  3  Assuming 50% of the world population regularly uses biomass fuels and that it is women who are regularly exposed.  39  Similarly, we estimated 436 million person-days of exposure (343 – 497 million) to PM10 concentrations above 250 µg/m3 with 98% of this occurring during the episode periods. In both cases, the vast majority of these exposures occurred during the first episode period in 1997. For person-days above 350 µg/m3 we estimated 3.1 million person-days (1.8 – 22 million), all of which occurred during the first episode period. These estimates were not very sensitive to the use of mean, low or high confidence limits for the estimated coefficient with the exception of the estimated numbers of people exposed above µg/m3. These estimates should be considered as conservative estimates as we did not consider PM10 concentrations above 500 µg/m3. Evidence exists that PM10 concentrations certainly did exceed 500 µg/m3 and that this was most common in regions not covered by PM10 surface monitoring. For example, in the fire source regions of Sumatra and Kalimantan. Again, our estimates are limited as we did not predict values above 500 µg/m3 hand we had no surface PM10 monitoring data to apply to these areas. Figures 13-15 display the estimated numbers of people within the study region that were exposed above 150, 250 or 350 µg/m3, based on the mean slope of the aerosol index – PM10 concentration regression relationship. Note the change in scale for Figure 15. Throughout the study period the daily estimated numbers of people exposed above the 150, 250 and 350 ug/m3 index values ranged from zero to a high of 87.8 million (87.7 – 88.9), 16.5 million (16.5 – 21.8) and 902,000 (902,000 – 4.3 million), respectively. The day of maximum exposure occurred on November 4, 1997. As indicated in Figure 13, while very high concentrations were measured during the 1998 episode period, many fewer numbers of people were exposed relative to the 1997 episode period.  40  Numbers of persons exposed to ambient concentrations >150 ug/m3 100000000 90000000 80000000 70000000 60000000 50000000 40000000 30000000 20000000 10000000  02 /0 2/ 98 02 /0 3/ 98 02 /0 4/ 98  02 /0 1/ 98  02 /1 2/ 97  02 /1 1/ 97  02 /1 0/ 97  02 /0 9/ 97  02 /0 8/ 97  02 /0 7/ 97  02 /0 6/ 97  02 /0 5/ 97  02 /0 4/ 97  02 /0 3/ 97  0  Numbers of persons exposed to ambient concentrations >250 ug/m3 100000000 90000000 80000000 70000000 60000000 50000000 40000000 30000000 20000000 10000000  02 /0 2/ 98 02 /0 3/ 98 02 /0 4/ 98  02 /0 1/ 98  02 /1 2/ 97  02 /1 1/ 97  02 /1 0/ 97  02 /0 9/ 97  02 /0 8/ 97  02 /0 7/ 97  02 /0 6/ 97  02 /0 5/ 97  02 /0 4/ 97  02 /0 3/ 97  0  41  Numbers of persons exposed to ambient concentrations >350 ug/m3 1000000 900000 800000 700000 600000 500000 400000 300000 200000 100000  02 /0 2/ 98 02 /0 3/ 98 02 /0 4/ 98  02 /0 1/ 98  02 /1 2/ 97  02 /1 1/ 97  02 /1 0/ 97  02 /0 9/ 97  02 /0 8/ 97  02 /0 7/ 97  02 /0 6/ 97  02 /0 5/ 97  02 /0 4/ 97  02 /0 3/ 97  0  There are several major sources of error associated with these estimated values. First, there is an error associated with using the aerosol index-PM10 regression results for estimating PM10 concentrations based on the aerosol index values. To partially assess the magnitude of this error, we conducted sensitivity analyses by also evaluating the estimated exposures for several days using the measured PM10 concentrations in addition to the estimated concentrations. For these ten randomly selected example days (the same days indicated in Figures 17-36) the measured PM10 concentration was chosen to represent the PM10 concentration within the entire pixel. Pixels which did not contain measurement sites were assigned the PM10 value based on the regression relationship. This procedure would tend to provide a more realistic estimate for those locations in which the measurement site recorded particle concentrations that were due to regional air pollution. Measurement locations in which local sources were influential would tend to overestimate the concentrations in the surrounding area. For these ten example days the two estimation procedures gave identical results except for April 8, 1998 on which the use of the regression equation alone estimated 1,467,060 people 42  exposed above 150 µg/m3 while using the measurement data in addition to the regression equation estimated 2,097,060 people exposed above 150 µg/m3, a difference of 630,000 people (43%). This can be attributed mainly to very high measured concentrations in Miri (458 µg/m3) as well as the surrounding area (Kuala Belait: 197 µg/m3, Tutong: 229 µg/m3, Temburong: 191 µg/m3) which were not differentiated as high concentration areas in the TOMS data, probably due to its limited spatial resolution. Again this demonstrates the limitations of the TOMS data to reveal local episodes of high concentrations, but highlights the to use this methodology to estimate surface concentrations for regional episodes. Accordingly, the lack of surface level air monitoring data can be considered a greater limitation than any limitations within the estimation procedure based on the TOMS data. An additional source of error is associated with the values for the population density itself as these values are themselves estimates and refer to projected population levels for 1995, and not true levels during the 1997-98 study period. For a more thorough description of the estimation  procedure  used  to  derive  the  population  density  data,  see  http://grid2.cr.usgs.gov/globalpop/asia/appendix.php3.  LIMITATIONS We have demonstrated a relationship between the TOMS aerosol index value and measured PM10 concentrations in Southeast Asia during periods when air quality was impacted by smoke from vegetation fires. Although an association exists, a large amount of variability in the measured PM10 concentrations cannot be explained by TOMS aerosol index values and there is substantial variability in the relationships across locations and specific time periods. Possible improvements are suggested by taking into account local effects and high concentrations that cannot be measured by the TOMS instrument. In additions to the other methodological limitations mentioned previously, for example the inability to estimate concentrations above 500 µg/m3, we also note that the PM10 measurements were made at the surface whereas the TOMS instrument measures primarily aerosols that are suspended high in the atmosphere (Kaufman and Thompson ). discrepancy between the two measurements.  This could also explain to some of the Perhaps the major limitation of this  methodology is the limited spatial resolution of the TOMS data. The 1 degree by 1.25 degree pixels are very large near the equator and the sub-pixel variability is significant.  An  improvement on the resolution could possibly lead to better insights into the relation between  43  the aerosol index and PM10. Singh and Jacob and have recently reviewed the prospects for improved measurement of tropospheric chemistry using satellite-based instruments and predict that remote-sensing techniques will prove increasingly useful and important (Singh and Jacob 2000).  CONCLUSIONS This report describes a rather complete database of ambient particle concentrations during the 1997-1998 vegetation fire episodes. During this study period, measured concentrations reached peak values of 852 ug/m3 and were elevated in some locations, particularly in the areas closest to Sumatra, for extended periods in 1997. In 1998 elevated concentrations were much more localized in the areas around Miri and Brunei. Mean concentrations for the 3.5 month period in 1997 which included the episode were 2 –3 times higher than non-episode periods, depending upon the level of background air pollution at a particular location. During the periods of highest concentrations the spatial extent of the episode included large areas without any surface monitoring sites. For example, figures showing predicted PM1o concentrations on October 15, 1997; March 27, 1998; and April 8, 1998. In particular the much of Sumatra as well as portions of southern Kalimantan have high population density and had not ambient PM10 monitoring sites during the 1997-98 period. These areas should be primary targets for additional monitoring sites which hope to reflect population exposure. Using the TOMS data and the methodology described in the report, we were able to provide estimates of exposures during the 1997-98 episodes. Exposures were high and affected huge segments of the population, in some cases for extended periods. During a single day within the episode period as many as 87.8 million people were estimated to have been exposed to ambient concentrations above 150 ug/m3. In viewing the episode periods in their entirety, we estimated that there were 2.26 billion person-days of exposure to PM10 concentrations above 150 µg/m3 during the study period, 96% of which during the episode periods. Despite a number of major limitations, we found that the TOMS aerosol index data was a useful indicator of surface PM10 concentrations during such regional episodes and especially where concentrations were below 500 ug/m3 and where surface monioting data wer not available. An additional advantage of the TOMS data is that is it can be made available almost immediately whereas there is often a significant time delay in the reporting of surface-based measurement data. This feature as well as the broad spatial coverage makes the TOMS useful  44  as an early warning monitor within this region as well as others that are subject to similar air pollution events.  ACKNOWLEDGMENTS This project was conducted through the financial sponsorship of the Canadian International Development Agency, through funding administered by the Conference Board of Canada as part of the APEC Phase II projects. The research was also supported by the World Health Organization, Western Pacific and Southeast Asian regional offices who provided logistical support and provided assistance with data collection.  45  BIBLIOGRPAHY Brauer, M. Assessment of health implications of haze in Malaysia. World Health Ogranization, Western Pacific Regional Office, Mission Report RS/97/0441. November 5, 1997. Brook, JR. Prediction of past PM10. In The Study of Haze. Draft final report prepared for the State Government of Sarawak. Ottawa, Ontario, Canada,, W.F. Baird & Associates Coastal Engineers Ltd, 1998. Chew FT, Ooi BC, Hui JKS, Saharom R, Goh DYT, Lee BW. Singapore’s haze and acute asthma in children. Lancet 1995; 346: 1427 Chia HP, Chia KS, Ooi PL, Ng TP, Goh KT and Lee HP. Effect of the recent haze in Singapore on the frequency of attacks among a group of known asthmatics. Proceedings of the Asia-Pacific Conference on the Built Environment. Institute of Environmental Epidemiology, Singapore, 1995. pp. 87-93. Fang, M. and W. Huang (1998). “Tracking the Indonesian forest fire using NOAA/AVHRR images.” Int. J. Remote Sensing 19(3): 387-390. Glover D. Interim Results of a Study on Economic Value of Haze Damages in SE Asia, Economy and Environment Program for South East Asia February 20, 1998. http://www.idrc.org.sg/eepsea/haze.htm Harger JRE. Air-temperature variations and ENSO effects in Indonesia, The Phillipines and El Salvador. ENSO patterns and changes from 1866-1993. Atmos Environ 1995; 29(16):1919-1942 Heil, A. (1998). Air Pollution Caused by Large Scale Forest Fires in Indonesia 1997. South East Asian Land/Forest Fires: Science and Policy Workshop, Singapore, Centre for Remote Imaging, Sensing and Processing (CRISP). Hisham-Hashim J, Hashim Z, Jalaludin J, Lubis SH, Hashim R. Respiratory function of elementary school children exposed to the 1997 Kuala Lumpur haze. ISEE/ISEA Joint Conference 1998. Boston, MA, 1998 Hornung RW, Reed LD. Estimation of average concentration in the presence of nondetectable values. Appl Occup Environ Hyg 5:46-51 (1990) Hsu, N. C., J. R. Herman, et al. (1996). “Detection of Biomass Burning Smoke from TOMS measurements.” Geophysical Research Letters 23(7): 745-748. Kaufman, Y. J. and A. M. Thompson “Monitoring Fires and Smoke by Polar Orbiting satellites.” . Leech J, Burnett RT, Cakmak S, Arif MT, Chang G. The Sarawak September haze episode. Am J Resp Crit Care Med 1998, 157(3):A260  46  Levine J., T. Bobbe, N. Ray, R.G. Witt, and A. Singh, 1999. Wildland Fires and the Environment: A Global Synthesis. UNEP/DEIAEW/TR.99-1. United Nations Environment Program, Nairobi, Kenya. Mallingreau JP, Stephens G, Fellows L. Remote sensing of forest fires: Kalimantan and North Borneo in 1982-83. Ambio 1985; 14:314-321. McPeters, R. D., P. K. Bhartia, et al. (1998). Earth Probe Total Ozone Mapping Spectrometer (TOMS) Data Product's User's Guide. Greenbelt, Maryland, Goddard Space Flight Center. Met One Instruments, I. (2000). Beta-Attenuation Mass Monitor, Met One Instruments, Inc. 2000. National Aeronautics and Space Administration (NASA) and Goddard Space Flight Center, (1997). Total Ozone Mapping Spectrometer/Earth Probe (TOMS) Mission Summary, National Aeronautics and Space Administration and Goddard Space Flight Center. 2000. Patashnick, R. (2000). TEOM Technology. http://www.rpco.com/products/dieprod/die1100/te.htm. 2000. Pinto J., L.D. Grant, and T.A. Hartlage, 1998. Report on US EPA Air Monitoring of Haze from S.E. Asia Biomass Fires. EPA/600/R-98/071. USEPA, National Center for Environmental Assessment, Research Triangle Park, NC. Radojevic M. and H. Hassan, 1999. Air quality in Brunei Darussalam during the 1998 haze episode. Atmospheric Environment 33 (22): 3651-3658. Singh, H. B. and D. J. Jacob (2000). “Future directions: Satellite observations of tropospheric chemistry.” Atmospheric Environment 34(25): 4399-4401. Smith, K. (1993). “Fuel combustion, air pollution exposure and health: The situation in developing countries.” Annual Review of Eenergy and Environment 18: 529-566. Tan, W. C., D. Qiu, et al. (2000). “The human bone marrow response to acute air pollution caused by forest fires.” Am J Respir Crit Care Med 161(4 Pt 1): 1213-7. Torres, O. (1998). “Derivation of aerosol properties from satellite measurements of backscattered UV radiation: Theoretical basis.” Journal of Geophysical Research 103: 1709917110. Tsutsumi Y., Y. Sawa, Y. Makino, J.B. Jensen, J.L. Gras, B.F. Ryan, S. Diharto, and H. Harjanto, 1999. Aircraft measurements of ozone, NOx, CO, and aerosol concentrations in biomass burning smoke over Indonesia and Australia in October 1997: Depleted ozone layer at low altitude over Indonesia. Geophysical Research Letters 26 (5): 595-598. Wooster MJ, Ceccato P, Flasse SP. (1998). “Indonesian fires observed using AVHRR.” International Journal of Remote Sensing 19(3): 383-386  47  GIS Figures 16-25. Aerosol index coverage as measured by TOMS sensor for 10 randomly selected days during study period. Data indicated are raw data indicating extent of TOMS coverage on each given day without any composite images. Black areas indicate locations without TOMS coverage on that day. 26-35. Estimated 24-hour average PM10 (µg/m3) concentrations for 10 randomly selected days during study period. PM10 concentrations were estimated from regression relationship between aerosol index measurements (shown in Figures 16 - 25) and measured PM10 concentrations. PM10 measurement sites are indicated by circle. The color within each circle refers to the same scale of the estimated PM10 concentrations. Black areas indicate locations where concentrations could not be estimated as areas were not included in TOMS coverages even after images for 3 days (± 1 day from day indicated) were combined and averaged. 36. Population density map for study region. 37. Estimated numbers of people exposed to 24-hour average PM10 concentrations above 150 ug/m3 on September 5, 1997. The legend indicates the number of people exposed above this concentration on this day. 38. Estimated numbers of people exposed to 24-hour average PM10 concentrations above 250 ug/m3 on September 5, 1997. The legend indicates the number of people exposed above this concentration on this day. 39. Estimated numbers of people exposed to 24-hour average PM10 concentrations above 150 µg/m3 on October 15, 1997. The legend indicates the number of people exposed above this concentration on this day. 40. Estimated numbers of people exposed to 24-hour average PM10 concentrations above 150 µg/m3 on February 19, 1998. The legend indicates the number of people exposed above this concentration on this day.  48  41. Estimated numbers of people exposed to 24-hour average PM10 concentrations above 150 µg/m3 on March 27, 1998. The legend indicates the number of people exposed above this concentration on this day. 42. Estimated numbers of people exposed to 24-hour average PM10 concentrations above 250 µg/m3 on March 27, 1998. The legend indicates the number of people exposed above this concentration on this day. 43.  Estimated numbers of people exposed to 24-hour average PM10 concentrations above 150 3  µg/m on April 8, 1998. The legend indicates the number of people exposed above this concentration on this day.  49  March 5, 1997  Aerosol Index -10 - 0 0 - 10 10 - 20 20 - 30 30 - 40 40 - 50 50 - 60 No Data  March 19, 1997  Aerosol Index -10 - 0 0 - 10 10 - 20 20 - 30 30 - 40 40 - 50 50 - 60 No Data  50  August 11, 1997  September 5, 1997  Aerosol Index -10 - 0 0 - 10 10 - 20 20 - 30 30 - 40 40 - 50 50 - 60 No Data  Aerosol Index -10 - 0 0 - 10 10 - 20 20 - 30 30 - 40 40 - 50 50 - 60 No Data  51  October 15, 1997  December 15, 1997  Aerosol Index -10 - 0 0 - 10 10 - 20 20 - 30 30 - 40 40 - 50 50 - 60 No Data  Aerosol Index -10 - 0 0 - 10 10 - 20 20 - 30 30 - 40 40 - 50 50 - 60 No Data  52  February 19, 1998  March 27, 1998  Aerosol Index -10 - 0 0 - 10 10 - 20 20 - 30 30 - 40 40 - 50 50 - 60 No Data  Aerosol Index -10 - 0 0 - 10 10 - 20 20 - 30 30 - 40 40 - 50 50 - 60 No Data  53  April 8, 1998  April 28, 1998  Aerosol Index -10 - 0 0 - 10 10 - 20 20 - 30 30 - 40 40 - 50 50 - 60 No Data  Aerosol Index -10 - 0 0 - 10 10 - 20 20 - 30 30 - 40 40 - 50 50 - 60 No Data  54  55  56  57  58  59  60  61  62  63  64  Appendix Technical Advisory Committee Members: BRUNEI DARUSSALAM  Dr Hajah Rahmah Binti Haji Mohd. Said Ag. Senior Medical Officer of Health (Epidemiology) Disease Control Unit Public Health Services Jalan Menteri Besar Ministry of Health Bandar Seri Begawan BB3910 Tel: 673-2-341640 and 673-2-342755 Fax: 673-2-380687 or 673-2-381687 e-mail: rahmahms@brunet.bn  compmoh@brunet.bn INDONESIA  Ms. Angelika Heil German Technical Cooperation (GTZ) Strengthening the Management Capacities of the Indonesian Ministry of Foresty (SMCP) Email: angelika_heil@t-online.de  MALAYSIA  Dr Mazrura bt. Sahani Medical Officer Environmental Health Unit Research Institut Penyelidikan Perubatan Jalan Pahang 50588 Kuala Lumpur Tel: 03 4402469 Fax: 03 2920675 E-mail: mazrura@imr.gov.my Ms Che Asmah Ibrahim Department of Environment Ministry Science Technology and Environment 12th- 13th Floor, Wisma Sime Darby Jalan Raja Laut 50662 Kuala Lumpur Malaysia Fax: 603- 2931480 E-mail: cai@jas.sains.my  SINGAPORE  Dr Ooi Peng Lim Deputy Director Institute of Environmental Epidemiology Ministry of the Environment 40 Scotts Road 22-00 Singapore 22831 Tel: 731 9726 Fax: 734 8287  65  e-mail: ooi_pen_lim@env.gov.sg THAILAND  Dr Kanchanasak Phonboon Health Systems Research Institute Ministry of Health Building of Mental Health Department Tiwanon Road Nonthaburi 11000 Tel: 02 951 1286 to 93 Fax: 02-951 1295 e-mail:kpvu@hsrint.hsri.or.th Dr. Supat Wangwongwatana Director, Air Quality and Noise Management Division404 Phahon Yothin Center BldgPhahon Yothin Road, Bangkok 10400 e-mail: supat.w@pcd.go.th  66  

Cite

Citation Scheme:

    

Usage Statistics

Country Views Downloads
United States 5 0
Japan 4 0
China 4 0
Malaysia 1 0
City Views Downloads
Tokyo 4 0
Ashburn 4 0
Guangzhou 3 0
Sunnyvale 1 0
Unknown 1 5
Beijing 1 0

{[{ mDataHeader[type] }]} {[{ month[type] }]} {[{ tData[type] }]}
Download Stats

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.304.1-0048199/manifest

Comment

Related Items