Effect of Biomass Burnings on Population Exposure and Health Impact at the End of 2019 Dry Season in Southeast Asia

Abstract

At the end of the dry season, from early March to early April each year, extensive agricultural biomass waste burnings occur throughout insular mainland Southeast Asia. During this biomass-burning period, smoke aerosols blanketed the whole region and were transported and dispersed by predominant westerly and southwesterly winds to southern China, Taiwan, and as far southern Japan and the Philippines. The extensive and intense burnings coincided with some wildfires in the forests due to high temperatures, making the region one of the global hot spots of biomass fires. In this study, we focus on the effect of pollutants emitted from biomass burnings in March 2019 at the height of the burning period on the exposed population and their health impact. The Weather Research Forecast-Chemistry (WRF-Chem) model was used to predict the PM_2.5 concentration over the simulating domain, and health impacts were then assessed on the exposed population in the four countries of Southeast Asia, namely Thailand, Laos, Cambodia, and Vietnam. Using the health impact based on log-linear concentration-response function and Integrated Exposure Response (IER), the results show that at the peak period of the burnings from 13 to 20 March 2019, Thailand experienced the highest impact, with an estimated 2170 premature deaths. Laos, Vietnam, and Cambodia followed, with estimated mortalities of 277, 565, and 315 deaths, respectively. However, when considering the impact per head of population, Laos exhibited the highest impact, followed by Thailand, Cambodia, and Vietnam. The results highlight the significant health impact of agricultural waste burnings in Southeast Asia at the end of the dry season. Hence, policymakers should take these into account to design measures to reduce the negative impact of widespread burnings on the exposed population in the region.

1. Introduction

Wildfires are natural events occurring mostly during the dry and hot seasons in various places around the world, such as North and South America, Europe, Russia, Australia, and South and Southeast Asia. In addition to wildfires occurring in forests or grasslands, biomass waste burnings also occur in many agricultural places such as China, South Asia (including India and Pakistan), the US, South America, and Southeast Asia. These fires, either due to natural events or to farmers disposing of the agricultural residues before the start of the new crop season, emit large amounts of pollutants and greenhouse gas concentration into the atmosphere. Air pollution is associated with adverse health outcomes for people exposed to air pollutants such as ozone, particulate matter, and nitrogen dioxides [1,2]. In particular, particulate matter with a size of less than 2.5 μm in diameter (PM_2.5), once inhaled, can go to the bloodstream and is known to have high detrimental health outcomes, especially in high-risk groups, such as those with cardio-respiratory diseases. Elevated morbidity (respiratory and cardiovascular diseases hospitalization) and mortality are associated with exposure to PM_2.5 and Black Carbon (BC). Increases in hospital emissions and emergency dispatches during high smoke periods are evident [3,4,5]. Aguilera et al., 2021 [6] have found that PM_2.5 in wildfire smoke adversely impacts human health more than equal doses of ambient PM_2.5 from other sources following recent toxicological studies. They isolated the wildfire-specific PM_2.5 using a series of statistical approaches and exposure definitions and found that increases in respiratory hospitalizations ranging from 1.3 to up to 10% with a 10 μg/m³ increase in wildfire-specific PM_2.5, compared to 0.67 to 1.3% associated with non-wildfire PM_2.5.

Peninsular Southeast Asia is a major source of pollutants emitted during biomass burnings of agricultural waste and forest fires at the end of the dry season before the start of the monsoon season. Smoke aerosols from the fires in Southeast Asia (SEA) have been known to not only affect the SEA region but also were transported further afield to southern China, Taiwan, the Philippines, and the western North Pacific Ocean [7,8,9]. Globally, emitted particles such as Black Carbon (BC) and Brown Carbon (BrC), due to incomplete combustion of biomass or fossil fuels, can change the radiation budget by absorbing light and, in turn, affect the climate. Kumar Pani et al., 2021 [10] studied the BrC as measured by an aethalometer in Chiang Mai, Northern Thailand, and found a strong association between biomass burnings in the region at the end of the dry season in March to April 2016 with BrC light absorption measurements at Chiang Mai site.

Health risks from air pollutants emitted from biomass burnings are a major concern during the height of the burning season. In their study of the health risks from landscape fire emissions in SEA from 1997 to 2006 using satellite data and modelling to quantify health effects, Marlier et al., 2013 [11] have shown that during strong El Niño years when less rain and more fires occur, an estimated 10,800 (6800–14,300) person (~2%) annual increase in regional adult cardiovascular mortality. Frankenberg et al., 2005 [12] studied the effect of the 1997 uncontrolled fires in Sumatra and Kalimantan (Indonesia) forests caused by slash-burn agricultural practices exacerbated by drought and late monsoon delay due to the 1997 El-Niño year. They used satellite haze data combined with population survey data to determine the population exposure to haze and the consequent health effect. Their results showed that haze had a negative impact on these dimensions of health, such as respiratory morbidity. The same event also caused extensive smoke haze pollution in nearby Singapore, where Emmanuel 2001 [13] studied the haze effect on the population health of this city. There was a 30% increase in outpatient attendance, and an increase of PM₁₀ levels from 50 μg/m³ to 150 μg/m³ was significantly associated with increases of 12% in upper respiratory tract illness, 19% in asthma, and 26% in rhinitis. However, there was no significant increase in hospital admissions or mortality, and the health effects from the 1997 smoke haze in Singapore were generally mild.

Elsewhere, the impact on air quality and human health of forest and vegetation fires were also studied, such as the 2012 fires in the Amazon basin of South America [14] and the wildfires in Australia during the Black Summer fires from November 2019 to January 2020 [15]. In their 2020 study, Butt et al. [14] used the exposure-response relationships and a regional climate-chemistry model to provide the first in-depth assessment of the potential public health benefits due to fire prevention across the Amazon Basin. Their results showed that if the biomass burnings did not occur, 16,800 (95UI: 16,300–17,400) premature deaths and 641,000 with upper and lower uncertainty intervals at the 95% confidence level (95UI: 551,900–741,300) disability-adjusted life years (DALYs) across South America would be avoided, with 26% of the avoided health burden located within the Amazon Basin. Accordingly, they stated that the health benefits of fire prevention in the Amazon are comparable to those found in Equatorial Asia. The results of Nguyen et al., 2021 [15] on population exposure and health impact for the event of the Black Summer wildfires on the east coast of Australia in December 2019 had the most impact on population health. For New South Wales, the most populous state, the total mortality estimates for the period of 1 November to 8 January 2020, when the wildfires were most active, were about 247 (95UI: 89, 409) premature deaths, 437 (95UI: 81, 984) cardiovascular diseases hospitalizations, and 1535 (95UI: 493, 2087) respiratory diseases hospitalizations. In a systemic review of many studies across the world related to the health impact of wildfires, Reid et al., 2016 [16] concluded that consistent evidence from a large number of studies indicates that wildfire smoke exposure is associated with respiratory morbidity, with growing evidence supporting an association with all-cause mortality.

It is expected, therefore, that large-scale agricultural burnings and forest fires at the driest period near the end of the dry season in SEA have a large impact on air quality and adverse public health to the population in the region.

In our previous study [9], the WRF-Chem air quality model and the FINN (Fire Emission Inventory from NCAR) emission dataset were used to determine the dispersion pattern of air pollutants such as PM₁₀, PM_2.5, and BC and their effect on air quality during 13 to 20 March 2019 peak period of the annual biomass burnings in peninsular SEA at the end of the 2019 dry season. The WRF-Chem model has been widely employed by various researchers to examine the air quality impacts due to biomass burnings in SEA, such as [17,18,19,20]. Figure 1 shows the fires and thermal anomalies or hot spots from fires as detected by MODIS sensor onboard Terra/Aqua satellites over Southeast Asia on 14, 16, 18, and 20 March 2019, when intense biomass burnings occurred with smoke covering most of SEA, especially in Thailand and Laos.

In this study, we aim to estimate the population exposure and health impact due to the emission and transport of air pollutants from biomass burnings over the regional population centres where pollutant concentration can significantly affect the population in each of the four countries in SEA (Thailand, Laos, Vietnam, and Cambodia) during the peak fire period in March 2019. Short-term effects of exposure to the smoke particles from biomass burnings in SEA, such as PM_2.5 and PM_10, can have an impact on the population’s health, such as increasing mortality rate respiratory and cardiac diseases hospitalisations. Khan et al., 2016 [21], in their study of PM_2.5 from fires, its composition using plasma mass spectroscopy (ICP-MS) and ion chromatography (IC) to determine the source apportionment using positive matrix factorisation (PMF) and access the health risk in Malaysia from this analysis. They found that the PM_2:5 concentration was slightly higher during the northeast monsoon compared to the southwest monsoon. The southwest monsoon from May to late October is more important in influencing rainfall weather in peninsular SEA (Myanmar, Thailand, Laos, Vietnam, and Cambodia) than the north-east monsoon, which affects maritime SEA, including Malaysia and Indonesia.

As the biomass burnings in March 2019 had a significant impact on air quality across the whole peninsular SEA [9] and hence on the population exposure, health, and economic cost, we use this period of intense burnings to estimate the impact of the biomass burnings on premature mortality of the exposed population, in the four most affected countries (Thailand, Laos, Vietnam, and Cambodia) in the peninsular Southeast Asian region. The results have important implications for public health policy in these countries and the need to develop policies to mitigate the health impact and the associated economic loss and healthcare costs due to widespread biomass burnings annually at the end of the dry season.

2. Data and Methods

2.1. Study Area

The climate of peninsular Southeast Asia (SEA) is mostly tropical, which is hot and humid, except in mountainous Laos and northern Vietnam, which is sub-tropical with distinct seasons. The summer monsoon from April to October brings rain to the region and is the main driver of agricultural activities. This rainy season, annually from April to October, is a source of life, and the beginning of the summer monsoon is celebrated as a water festival (Songkran Festival) in Thailand and Laos. In the month just before the summer monsoon, after the dry season from November to March, farmers burn agricultural residues in preparation for the new cropping activities in all countries of mainland SEA, including Myanmar. During March, much of SEA is covered in smoke as detected by satellites such as MODIS Aqua/Terra, with many hot spots of biomass fires.

2.2. Air Quality Model

Air quality models have been used extensively to study the dispersion and transport of air pollutants from emission sources and meteorological input. Most air quality models used are using either WRF-CMAQ or WRF-Chem. Ooi et al., 2021 [7] recently used WRF-CMAQ to study the dispersion of transport of emitted pollutants from biomass burnings in Southeast Asia. They used FINN as the estimated emission of pollutants (gaseous and solid) from fires detected by MODIS Aqua/Terra satellites.

2.3. Emission Data

In this study, the air quality model WRF-Chem (Version 4.2), in conjunction with the Fire Emission Inventory from NCAR (FINN), is used to study the dispersion of pollutant emissions from fires during biomass burnings in March 2019. The FINNv1.5 emission database contains fire emission data at 1 km resolution derived from MODIS Rapid Response fire count (Fire Information for Resource Management System, FIRMS) and VIIRS (Visible Infrared Imaging Radiometer Suite) hot spots [22,23]. Based on the emission calculation algorithm, the emission data of many pollutant species are provided by NCAR on a daily basis at hourly resolution and can be downloaded and then converted into WRF emission input files.

The emission data are available for three air chemistry mechanisms and, therefore, are in three formatted files with suitable chemical species for GEO-Chem (Goddard Earth Observing System with Chemistry), SAPRC99 (Statewide Air Pollution Research Center Mechanism), and MOZART (Model for Ozone and Related chemical Tracers) mechanism. These emission files are downloadable from the NCAR Research Data Archive: https://rda.ucar.edu/datasets/ds312.9/, accessed 7 March 2024). Similar to our previous study on the March 2019 biomass burnings, we use fire emission data in FINNV1.5 based on the MODIS satellite (1 km resolution). In this study, we use the WRF-Chem air quality model, which has gas phase MOZART and aerosol MOSAIC (Model for Simulating Aerosol Interactions and Chemistry) 4-bin option implemented. For this reason, the fire emission data set for the MOZART gas phase photochemistry mechanism from FINNV1.5 is used. These files are already in the species format and, with available conversion tools, can be converted to be used as input to the WRF-Chem model.

Besides FINN emission sources, anthropogenic emission sources such as land and shipping transport based on the global EDGAR-HTAP dataset are also used. During the most intense biomass burnings in March 2019, the contribution of anthropogenic emission to aerosol dispersion and concentration at ground and above ground level was much lower than the emission from biomass burnings [9]. WRF-Chem also requires host meteorological data for downscaling to higher resolution simulation domain the meteorological prediction, which is used for dispersion, transport, and via chemistry module, the chemical reaction of gaseous and particulate pollutants to form secondary pollutants. The NCEP FNL (National Centers for Environmental Prediction, Final Analysis) reanalysis meteorological data are used as input to the WRF-Chem model to simulate the transport and dispersion of emitted pollutants from biomass burnings in SEA.

2.4. Model Configuration

We use the same domain configuration and WRF-Chem options as in our previous study [9]. Figure 2 shows the domain configuration, and

Table 1. WRF-Chem configuration in the namelist.input.

Physical Parametrisation	Namelist Variable	Option	Model/Scheme
Microphysics	mp_physics	3	WRF Single Moment
Land surface	sf_surface_physics	2	Noah Land-Surface Model
Surface layer physics	sf_sfclay_physics	1	Monin-Obukhov similarity
Planetary Boundary Layer	bl_pbl_physics	1	YSU scheme
Shortwave radiation	ra_sw_physics	4	Rapid Radiative Transfer Model (RRTMG)
Long wave radiation	ra_lw_physics	4	Rapid Radiative Transfer Model (RRTMG)
Cumulus cloud	cu_physics	1	Kain-Fritsch scheme
Gas/aerosol Chemistry	chem_opt	112	MOZART/GOCART
Biomass burning option	biomass_burn_opt	2	Emission and plume rise for MOZCART
Sea salt emission	seas_opt	1	GOCART sea salt emission scheme
Photolysis	phot_opt	3	Madronich F-TUV photolysis
Dust scheme	dust_opt	3	GOCART-AFWA scheme
Aerosol extinction coefficient approximation	aer_opt_opt	2	Maxwell-Garnett approximation
Aerosol radiative feedback	aer_ra_feedback	1	Turn on aerosol radiative feedback with RRTMG model

shows some of the WRF-Chem physics and chemistry configuration options. The domain size is 354 × 315 with grid resolution dx = 0.057729° and dy = 0.057651° (about 6 km × 6 km), ‘lat-lon’ map projection, and reference domain centre as ref_lat = 16.274°, ref_lon = 104.594°.

Figure 3 shows the pattern of PM_2.5 hourly emission from biomass burnings as provided by FINN and the daily anthropogenic PM_2.5 emission from EDGAR-HTAP. The PM_2.5 emissions from biomass burnings are mainly concentrated in central, northeast Thailand, and Laos, where fires occurred, and are much larger than those from anthropogenic sources located mostly in cities and populated areas in central Thailand, southern and northern Vietnam, and southern China. The emission of some species such as CO (Carbon Monoxide), BC (Black Carbon), NO₂ (Nitrogen Dioxide), and C₂H₆ (Ethane) from biomass burnings at hot spots as detected by MODIS Terra/Aqua are shown in Figure A1 in Appendix A. These emissions are estimated by FINN based on satellite detection of hot spots, their radiative power, and the type of vegetation or biomass in those hot spots. Details of how emissions from fires can be estimated using fire radiation power to measure the intensity and type of vegetation to derive the emission factors for different species of pollutants are described in the studies of Wiedinmyer et al., 2010, 2023. The emission is then stored in the data store daily as FINN. The FINN fire emission and the anthropogenic emission are both included in the emission input to the WRF-Chem model to simulate the dispersion and transport of air pollutants emitted from these two sources.

The period of simulation from 13 to 20 March 2019 corresponds to the most intense burnings. Yin et al., 2019 [24] used AOD (Aerosol Optical Depth) data from the Terra satellite and PM_2.5 from the MERRA Aerosol Reanalysis (MERRAero) model from 2001 to 2016 showed that during March, the biomass burnings contributed most to the aerosols emitted to the atmosphere in SEA.

From the simulation of WRF-Chem, the predicted ground concentration of PM_2.5 across peninsular SEA is then used to estimate the population exposure. The predicted PM_2.5 includes primary emitted PM_2.5, secondary organic aerosols (SOA), and secondary inorganic aerosols (SIA) formed from chemical reactions and particle formation in the WRF-Chem MOZART chemical mechanism. The health impact for each health endpoint due to pollutant concentration on the exposed population is estimated based on the relative risk of PM_2.5 using the following method as described below.

2.5. Health Impact

The relative risk (RR) factors for a 10 μg/m³ increase of PM_2.5 on mortality (all cases) and cardiovascular (cvd) diseases are based on the World Health Organization’s (WHO) Health Risks of Air Pollution in Europe project (HRAPIE) recommendations for concentration-response functions [25]. We use these recommended short-term concentration response coefficients for mortality and cvd disease hospitalisation in this study.

RR (mortality) = 1.0123 (CI: 1.0045, 1.0201)

RR (cvd) = 1.0091 (CI: 1.0017, 1.0201)

where CI is the confidence interval at a 95% level.

The HRAPIE relative risk for mortality value above means that a 10 μg/m³ increase of PM_2.5 would increase the mortality of the exposed population by 1.23% (CI: 0.45%, 2.01%). Xu et al., 2020 [26] reviewed the literature on the effect of wildfire events and reported that an increase of 10 μg/m³ in daily PM_2.5 level is associated with an increase of 0.8–2.4% in the risk of death from any cause or nonaccidental death for up to 4 days after the exposure. The RR value (1.23%) for mortality as recommended by WHO above falls in the range of values in [26] study. The RR value has a strong influence on the results of an epidemiological study on health impact on exposure population due to air pollution.

Table A1. Acute relative risks of mortality and morbidity due to impact of wildfires from various studies.

Pollutant	Health Endpoint	Study Location	RR (95% CI) per 10 μg/m3	Reference
PM10	Mortality—cardiorespiratory	Portugal (Europe)	1.024 (1.010, 1.038)	Augusto et al., (2020) [45]
PM2.5	Mortality—cardiovascular	Europe	1.0342 (1.0064–1.0628)	Faustini et al., 2015 [46]
PM2.5	Mortality—all non-accidental, all ages	Finland (Europe)	1.020 (0.924, 1.086)	Kollanus et al., 2016 [47]
	Mortality—all non-accidental, ages > 65	Finland (Europe)	1.024 (0.935, 1.12)
PM2.5	Mortality—all causes	Global	1.019 (1.016, 1.022)	Chen et al., 2021 [48]
	Mortality- cardiovascular	Global	1.017 (1.012, 1.021)
	Mortality—respiratory	Global	1.019 (1.013, 1.025)
PM2.5	Morbidity—respiratory hospitalisation	U.S	1.072 (1.0025, 1.15)	Liu et al., 2017 [30]
PM2.5	Morbidity—asthma hospitalisation	U.S	1.14 (1.10, 1.20)	Reid et al., 2016 [16]
	Morbidity—ED * visits for chronic obstructive pulmonary disease (COPD)	U.S	1.04 (1.02, 1.08)
PM2.5	Morbidity—respiratory hospitalisation	U.S	1.028 (1.014, 1.041)	Delfino et al., 2016 [27]
	Morbidity—asthma hospitalisation	U.S	1.048 (1.021, 1.076)

* ED = Emergency Department.

in Appendix B lists some of the RR values in a number of studies on short-term health effects (mortality and morbidity) due to wildfires in literature. Here, we adopt the RR values from WHO as references to determine the short-term health impact of air pollution due to wildfires.

For disease hospitalisation such as cardiovascular disease or respiratory diseases, International Statistical Classification of Diseases and Related Health Problems, Tenth Revision is used to refer to those diseases in this study: I00-I90 for cardiovascular disease and J00-J99 for respiratory diseases, I00-I90 for cardiovascular disease as standard.

For respiratory hospitalisations, in this study, we used the RR of PM_2.5 value of 1.028 (CI: 1.014,1.041), based on a study of the 2003 California wildfires [27]. This value was also used in [15,28,29] for large wildfires or biomass hazardous reduction burnings (HRBs) in Australia. Liu et al., 2017 [30] conducted a study of cardiovascular and respiratory hospitalisation due to smoke waves of two or more days when daily average concentration of PM_2.5 ≥ 20 μg/m³ across the whole Western United States from 2004 to 2009. Their results on respiratory disease hospitalisation during the wildfires was 1.072 (CI: 1.0025, 1.15), which is higher than that of [27] but a larger CI. Delfino et al., 2009 [27] study of the catastrophic wildfires in southern California in October 2003, which generated large amounts of dense smoke covering much of urban southern California, is similar in scope to the biomass burnings in SEA during March 2019. A recent study conducted in Chiangmai, northern Thailand, on the link between the air quality index (AQI) of PM_2.5 and emergency department (ED) visits has found that there is an increase in ED visits for different diseases [31]. Their results give the RR as (1.023, CI: 1.002–1.044) due to acute coronary syndrome (ACS), (1.071, CI: 1.025–1.118) due to pneumonia on the current day, increased mortality resulted from ACS on lag day three (RR = 1.36, CI: 1.08–1.73). These are the first results of an epidemiological study on the risks of mortality and morbidity due to air pollutants for SEA. However, these are values for ambient air pollution, not for large wildfires or biomass burnings.

The impact factor (IF) is the relative risk associated with a specific change in air pollution concentration ΔX. In this case, ΔX is the PM_2.5 increase due to a fire event, and IF is calculated as:

IF = RR^ΔX/10

(1)

IF is also called the relative risk function with respect to change in PM concentration (standardised to 10 μg/m³). The above Equation (1) is a log-linear model which states that the logarithm of the impact factor (or relative risk function) is linearly changed with a change in ambient PM concentration.

The attributable number of the impact on health endpoint due to an increase of ΔX concentration in a population with a specific incidence rate for this endpoint is

AN = (IF-1) × Population × incidence rate

(2)

The above Equations (1) and (2) are used by the US EPA to estimate the health impact due to changes in air quality in the United States [32] and implemented in the BenMAP v1.5 software tool. They are also used by many authors to assess the health impact of air pollution in various countries [29,33,34].

As the relative risk function (or impact factor) is not always linear with respect to change in concentration, especially in low and very high PM concentration values, an Integrated Exposure Response (IER) is often used to account for this wide range in concentration from very low to very high [35]. This IER model combined the mortality relative risk estimates from other PM_2.5 combustion sources, including active and second-hand smoking household burning of solid fuels. The IER model, developed for use in the Global Burden of Disease Study (GBD), estimates the PM_2.5 relative risks across the entire global range of exposure, including low-polluted areas and highly polluted areas where epidemiologic studies are lacking. The IER is used in the World Health Organization’s (WHO) benefits assessment software (AirQ + 2.2) and the United States Environmental Protection Agency’s (US EPA) benefits assessment software (BenMAP) [35]. In this study, the biomass burnings in March 2019 in peninsular SEA resulted in a high concentration of PM_2.5 and at some locations in northern Thailand and Laos, the extremely high daily average exposure level of PM_2.5 was above 30 μg/m³, which would cause a non-linear relationship between concentration and health impact. The GBD 2004 assumed that no additional risk beyond 30 μg/m³. The log-linear response, as provided in Equations (1) and (2), is adequate for our study, but beyond the daily average PM_2.5 concentration of 30 μg/m³, the response will be capped as assumed in the Global Burden of Disease (GBD) 2004 study. In a study on the health effects of the summer megafires of 2019/2020 in Australia, Johnston et al., 2021 [33] used a higher value of 150 μg/m³ to constrain the maximum daily concentrations of PM_2.5.

High-resolution population data sets can be obtained from various sources, such as [36] for global data sets. Another source of population data is the Gridded Population of the World from SEDAC (Socioeconomic Data and Applications Center) of NASA’s Earth Observing System Data and Information System (EOSDIS) at http://sedac.ciesin.columbia.edu/data/collection/gpw-v4/sets/browse (accessed on 19 October 2024). More up-to-date datasets, which are suitable for the current study, are the 2019 population data for each of the four countries in this study. This dataset was obtained from the UN Humanitarian Data Exchange in TIFF format at 1 km × 1 km resolution. The data is then converted to a csv (comma-separated values) file to be used by R and Python software to interpolate and overlay the population data with the predicted daily PM_2.5 concentration grid, and the impact on health endpoints is then calculated. Figure 4 shows the 2019 population distribution in SEA and in each of the individual countries of peninsular SEA (Vietnam, Thailand, Laos, and Cambodia) based on the UN Humanitarian Data Exchange dataset. As expected, the highest population concentration is in the metropolitan areas of the capital cities: Bangkok (Thailand), Hanoi and Ho Chi Minh City (Vietnam), Vientiane (Laos), and Phnom Penh (Cambodia).

Data for the relative risks and incidence rates for each of the health endpoints (such as mortality rates and hospitalisation for respiratory diseases), as specified in Equations (1) and (2), are keys in the health impact study of air pollutants. Relative risks can be obtained from previous studies on the health impact of wildfires in literature, while incidence rates can be obtained from health statistics of each of the countries either produced by the relevant local health authorities or from international organisations. Table A1 in Appendix B lists some of the relative risks used in various health impact studies of air pollution using PM_2.5 as a surrogate of pollution concentration. It is important to distinguish between the long-term exposure health impact and the short-term (acute) health impact of air pollutants. As mentioned above, the relative risks from the World Health Organization’s (WHO) Health Risks of Air Pollution in Europe project (HRAPIE) recommendations are chosen as these values have been selected based on an extensive review of published studies.

For Vietnam, Thailand, Cambodia, and Laos, we use the incidence rate based on WHO data for 2016 of non-communicative diseases (NCD) (https://web.archive.org/web/20220121073838/https://www.who.int/nmh/countries/lao_en.pdf, https://web.archive.org/web/20220121060716/http://www.who.int/nmh/countries/tha_en.pdf, (accessed 20 February 2024)). Thailand’s population in 2016 was 68,864,000, and a total death of 539,000 hence, the mortality rate is 0.78% for the 2016 year. Similarly, the mortality rates for Vietnam (0.57%), Cambodia (0.59%), and Laos (0.69%) are derived from WHO data for these countries. The Global Health Data Exchange (GHDx) from the Institute for Health Metrics and Evaluation (IHME) is another source of incidence rate for mortality in each of the countries available from 1990 to 2019. The data can be downloaded from http://ghdx.healthdata.org (accessed 20 February 2024). According to GHDx data for 2019, the total deaths in Thailand, Vietnam, Laos, and Cambodia are 497,502 (CI: 389,256, 628,041), 631,817 (CI: 538,099, 714,078), 44,456 (CI: 36,587, 52,996), and 110,850 (CI: 93,245,126,115) respectively. The corresponding population and mortality incidence rates are Thailand (70,111,586; 0.71%), Vietnam (96,372,928; 0.65%), Laos (7,158,249; 0.62%), and Cambodia (16,603,117; 0.67%). Comparing the 2016 estimate of incidence rates based on WHO data and those of 2019 based on GHDx data, there are slight differences, but they are consistent. The mortality rate increases from 2016 to 2019 for Vietnam (0.57% to 0.65%) and Cambodia (0.59% to 0.67%) but decreases for Thailand (0.78% to 0.71%) and Laos (from 0.69% to 0.62%). The average mortality rate for the four countries in 2019 was approximately 0.66%. As the focus is on biomass burning at the end of the dry period in March 2019, the mortality rate for each of the countries in 2019 is selected individually.

Morbidity data on cardio-vascular disease hospitalisation from Southeast Asian countries are scarce. CVD is a group of diseases, including coronary heart disease, such as ischaemic heart disease (IHD), and cerebrovascular disease, such as stroke. IHD was the summation of 4 distinct disease sequelae: acute myocardial infarction, chronic stable angina, chronic IHD, and heart failure due to IHD [37]. Nguyen et al., 2022 [38] recently used the population-based registries of residents from two provinces in a northern urban (Hai Phong) and a central rural (Thanh Hoa) province of Vietnam hospitalised with a validated AMI (Acute myocardial infarction) to estimate the incidence rates of AMI in these two registries. The incidence rates (per 100,000 population) of initial AMI in Thanh Hoa were 16 and Hai Phong 30. Data from the Institute for Health Metrics and Evaluation (IHME) provide the CVD prevalence rate for Vietnam as 6.1% (4.0% in Cambodia and 4.9% in Indonesia). In a study of the trend in heart failure hospitalisation and outcome under the national public health insurance system in Thailand from 2008 to 2013, Janwanishstaporn et al., 2022 [39] found that heart failure (HF) hospitalisation rate increased from 138 in 2008 to 168 per 100,000 beneficiaries in 2013. In this study, we use the incidence rate for cardiovascular disease (CVD) hospitalisation rate similar to the HF hospitalisation rate of 0.17% for all four countries in SEA (Thailand, Vietnam, Laos, and Cambodia) in 2013.

As for respiratory diseases hospitalisation, only information for Thailand is found with a comprehensive national statistic in 2010 based on data from three health insurance consisting of approximately 96% of the adult population (46,208,964 persons) of the total Thai population of 64.7 million [40]. There were 460,471 respiratory disease hospitalisations in the year, giving the incidence rate of approximately 1%. As far as we are aware, there is no published data for 2019 or later years, and due to the lack of data from other countries in SEA, the incidence rate of 1% for respiratory disease hospitalisation is chosen for the four countries.

The simulation period is from 13 to 20 March 2019, inclusively when the biomass burnings were at the highest intensity near the end of the dry season in SEA. Evaluation of the simulation was conducted by comparing the predicted meteorological variables such as temperature, wind speed and direction at various sites (Luang Namtha, northern Laos and Bangkok, central Thailand) with measured data [9]. The predicted AOD (Aerosol Optical Depth) at many sites in the region is also compared with measured AOD at AERONET sites (Chiang Mai, Luang Namtha, Nong Khai, Fang, Bangkok in Thailand, and Bac Lieu and Nha Trang in Vietnam). The evaluation conducted in our previous study [9] shows that the meteorological and AOD prediction corresponds well with observed data at those sites. Figure 5 shows the predicted AOD across the region covering the four countries on 17 March 2019 at 00:00 UTC and the location of the AERONET sites measuring AOD over time used to validate the model [9]. A comparison of AOD as predicted by WRF-Chem at Fang in northern Thailand with AERONET measurement at this site is also shown.

Simulation is also conducted for the base case scenario where there were no biomass burnings during the same period. The base case scenario simulation is run only with the anthropogenic emission input. The difference in predicted PM_2.5 between the biomass burnings scenario and the base case scenario at each of the grid cells will be applied to equation (2) and summed over the domain to estimate the attributable number of health impacts due to biomass burnings.

3. Results and Discussion

The predicted wind field and ground concentration of PM_2.5 over the domain for each hour during the simulation period from 13 to 20 March 2019 are obtained, and some of the plots on the day and time of 13, 15, 17, and 19 March 2019 at 16:00 UTC are shown in Figure 6.

The surface wind was mostly northern and north-easterly across the domain. As a result, the pollutant emission from fire sources in north northeast Thailand and Laos was dispersed and transported to central and southern Thailand and northern Cambodia. Transport of pollutants from northern Thailand and Laos to Vietnam, southern China, Taiwan and beyond is facilitated by the Truong Son Mountain range situated along the Laotian-Vietnamese border, which acts as a launching pad for pollutants to be carried upwind in the westerly or southwesterly flow to higher altitude and hence transported farther in the upper troposphere [9].

The sources of the anthropogenic emission are located in major cities and populated areas in the simulation domain: Bangkok, Ho Chi Minh City, Hanoi, and southern China (including Guangzhou). During the peak of burnings, the anthropogenic emissions are much less than those from biomass burning in hot spots. During those days of simulation where peak fires occurred, the gridded PM_2.5 concentration across the domain mostly originated from fire sources. We expect that PM_2.5 concentration is dominated by biomass burnings therefore, the PM_2.5 impact on population health for these days is due mainly to biomass burnings. The WRF-Chem simulation with no biomass burning scenario during the same period of 13 to 20 March 2019 was performed. In this scenario, only anthropogenic emission was used as input to the WRF-Chem model. This base case prediction of air pollutant PM_2.5 concentration will be used to subtract from the PM_2.5 prediction of the full simulation with both biomass-burning emission and anthropogenic emission as input to the WRF-Chem model.

From the hourly PM_2.5 concentration as predicted by the WRF-Chem model, the spatial distribution of the predicted daily average of PM_2.5 concentration for the eight-day simulation from 13 to 20 March 2019 is determined for each day. Most of the high predicted daily PM_2.5 concentration (~10 μg/m³) due to anthropogenic sources for the simulation days are over the coast of northern and southern Vietnam, as shown in Figure 7. The northerly and north-easterly wind (as shown in Figure 6) brought the pollutants emitted in southern China and northern and southern Vietnam to the Gulf of Thailand. However, the PM_2.5 concentration due to anthropogenic emission is still very low compared to those due to biomass burning, as shown in Figure 8. The spatial distribution of predicted PM_2.5 concentration and population health impact (mortality, cardiovascular diseases, and respiratory diseases hospitalisation) are shown in Figure 8 and Figure 9.

While the maximum PM_2.5 concentration can occur in a particular area if there is no population living there, the exposure to harmful pollutants would be non-existent. But if the amount of PM_2.5 in ambient air over a populated such as an urban area, then the health impact will be significant. The daily population exposure to PM_2.5 is defined as the daily average PM_2.5 concentration at each grid cell multiplied by the population at the same grid point and summed over the region. Over the whole region, the population-weighted average PM_2.5 can be calculated for each of the days during the period 13 to 20 March 2019.

$${\mathrm{PM}2.5}_{popwgt}={\displaystyle \frac{\sum _1^n{\mathrm{PM}2.5}_i\times {Pop}_i}{\sum _1^n{Pop}_i}}$$

where PM2.5_i and Pop_i is the daily average PM_2.5 concentration and the population at the grid cell i.

It is clear from

Table 2. Population weighted daily PM_2.5 concentration (μg/m³) and mortality effect (persons) due to PM_2.5 exposure for Vietnam, Thailand, Laos, and Cambodia based on the IER function method for 13 to 20 March 2019.

Date	Vietnam Daily PM2.5 Pop-Weighted Concentration	Vietnam Mortality (Persons)	Thailand Daily PM2.5 Pop-Weighted Concentration	Thailand Mortality (Persons)	Laos Daily PM2.5 Pop-Weighted Concentration	Laos Mortality (Persons)	Cambodia Daily PM2.5 Pop-Weighted Concentration	Cambodia Mortality (Persons)
13 March 2019	1.85	20	12.92	134	59.16	32	9.94	21
14 March 2019	2.32	30	22.53	193	118.10	36	13.46	35
15 March 2019	2.20	34	25.98	245	73.16	31	16.27	45
16 March 2019	4.79	61	29.75	338	101.41	34	22.14	56
17 March 2019	5.08	56	40.38	342	95.29	36	24.78	58
18 March 2019	6.75	73	46.86	337	124.56	36	22.78	47
19 March 2019	11.98	129	40.27	311	129.68	36	16.03	38
20 March 2019	20.16	162	33.75	270	106.57	36	5.4	15
Total mortality (persons)		565		2170		277		315

that the daily average population-weighted PM_2.5 concentration is highest in Laos, followed by Thailand, Cambodia, and Vietnam in that order. A summary of the estimated health impact in Vietnam, Thailand, Cambodia, and Laos due to biomass burnings from 13 to 20 March 2019, inclusive at the end of the dry season in peninsular SEA, is shown in

Table 3. The estimated health impact (mortality, CVD and respiratory disease hospitalisation) based on the IER method in Vietnam, Thailand, Cambodia, and Laos due to biomass burnings from 13 to 20 March 2019 at the end of the dry season in peninsular SEA.

	Vietnam	Thailand	Cambodia	Laos	Total for SEA
Population (2019)	96,372,928	70,111,586	16,603,117	7,158,249	190,245,880
2019 incidence mortality rate (%)	0.65	0.71	0.67	0.62
Premature deaths (mean) due to biomass burnings 2019 (2019)—IER	565	2170	315	277	3327
Proportion of effected population (mortality) 2019—IER	0.00059%	0.0031%	0.0019%	0.0039%	0.0017%
Cardiovascular diseases (CVD) hospitalisation (cases)	108	384	59	56	607
Respiratory diseases hospitalisation (cases)	2120	7572	1158	1106	11,956

. It can be seen that the number of premature deaths is highest in Thailand (~2170) as compared to Laos (~277), Vietnam (~565), and Cambodia (~315). Although the population of Thailand is less than that of Vietnam, high daily concentrations of PM_2.5 were mostly over central and northern Thailand, where population centres were located, while much less PM_2.5 concentration occurred in Vietnam. Laos has the least mortality as it has the least population among the four countries but has the highest proportion of death overpopulation due to high population-weighted PM_2.5 concentration. High predicted daily PM_2.5 concentration over most of Laos accounts for this result.

Figure 9 shows the gridded mortality due to biomass burnings for each day from 13 to 20 March 2019. High concentration of PM_2.5 in northeast Thailand and northern Laos accounts for most of the mortality in these regions. Even though Vietnam has a high-density population compared to other countries, the mortality in terms of proportion of the population is low due to the low concentration of PM_2.5 transported from the fires in the neighbouring countries. Bangkok metropolitan area, even though it has a lower PM_2.5 concentration due to the fires as compared to those in northeast Thailand and northern Laos, the high population density in Bangkok results in higher exposure to smoke particles and, hence, higher mortality. Similarly, in Vietnam and Cambodia, the metropolitan areas of Ho Chi Minh City and Phnom Penh account for most of the mortality in these countries, respectively. Hanoi also accounts for high mortality in Vietnam, especially on the 18, 19, and 20 March 2019, as shown in Figure 9 when the PM_2.5 plumes extend from Laos and Thailand to northern and central Vietnam (see Figure A3 in Appendix C of the plumes on those days).

For comparison, gridded daily mortality in SEA based on the log-linear impact function is shown in Figure A4 of Appendix D.

Our findings for Thailand can be compared with the study conducted by [41], which estimated the health impact in Thailand due to biomass burnings (including agricultural burnings and forest fires) in 2016. Using vegetation type cover and back trajectory analysis with HYSPLIT during biomass burnings in Thailand in 2016, Punsompong et al., 2020 [41] estimated that forest fire predominated in the mountainous north (73%) and northeast (48%) whereas, in the central region of Thailand, agricultural waste burning predominated (52%). As for grassland, the burning was most influential in the central region (16%), followed by the northeast (11%) and north (5%) regions of Thailand. The spatial distribution results of excess mortality showed the largest burden was in the central region (44%), followed by the northeast (29%), north (18%), and south (9%).

But rather than focusing only on Thailand as in [41], in our study, the health impact assessment due to PM_2.5 pollutants emitted by biomass burnings during the end of the 2019 dry season on the four countries of Thailand, Vietnam, Cambodia, and Laos are covered. Their results can be compared to those of ours in Thailand. Based on monitoring data instead of modelling, their results on population exposure and health impact in Thailand of biomass burnings in peninsular SEA for 2016 estimated the excess mortality due to PM_2.5 exposure to be about 18,003. The mortality was associated with stroke burden (53%), ischemic heart disease (30%), lung cancer (12%), and chronic obstructive pulmonary disease (5%), based on public-health data from 2016. Our results on the health impact in Thailand during the peak period of biomass burning in March 2019 is about 2170 excess mortality as compared to the value of 18,000 for the whole period of 2016 in their study.

Another recent study by [34] on the health impact and economic cost of open biomass burnings (OBB, including forest fires) in mainland Southeast Asia (MSEA, including Myanmar) during March 2012 is similar in approach to our study. They used the WRF-CMAQ air quality model to simulate the dispersion and concentration of PM_2.5 over the MESA at grid point resolution of 27 km × 27 km (0.25° × 0.25°) and then used the Environmental Benefits Mapping and Analysis Program (BenMAP) software tool to determine the impact on health endpoints on population exposure in each of the countries in MESA from the predicted gridded PM_2.5 concentration. The BenMAP tool also uses the concentration-response function of health endpoints as in Equations (1) and (2) in our study, which is a log-linear form and is not capped as used in our IER model. They used the monthly mean of PM_2.5 in March 2012 to calculate the health impact (mortality) rather than each individual day. Their results showed that the highest impact on mortality is in Myanmar, followed by Thailand, Vietnam, Laos, and Cambodia. Our results for the four countries show the highest mortality is Thailand, then Vietnam, Cambodia, and Laos in that order. But their number of mortalities for all-cause (AC) mortality is much higher: Laos (11,131), Thailand (3629), Cambodia (3625), and Vietnam (3376). Our mortality results are Laos (277), Thailand (2170), Cambodia (315), and Vietnam (565). The reason for this large difference is that they used the log-linear concentration-response function without a cap at the high end of PM_2.5 concentrations above 30 μg/m^3, which occurred often in northern Laos, Thailand, and neighbouring Myanmar during biomass burnings in March. If we use the log-linear response function, then our results on mortality are Laos (1066), Thailand (3290), Cambodia (411), and Vietnam (935) (see Appendix E). In addition to different air quality model and resolution and simulation periods used in their study, the values of RR, incidence rate and population in their study are also different from ours. This shows that the parameters used in the concentration-response function and health impact attributable equation for a health endpoint, as well as grid resolution and simulation period, are very important in influencing the final results of the health impact mortality study due to air pollutants such as PM_2.5. In terms of per population, the mortality risk is highest in Laos, followed by Myanmar, Cambodia, Vietnam, and Thailand, while our result for 13 to 20 March 2019 is Laos, Thailand, Cambodia, and Vietnam, in that order.

Mueller et al., 2020 [42] used PM₁₀, PM_2.5, CO, O₃, and NO₂ from ground-based monitors and daily outpatient hospital visits in Thailand during 2014–2017. Outpatient data included chronic lower respiratory disease (CLRD), ischaemic heart disease (IHD), and cerebrovascular disease (CBVD). They used time series analysis to evaluate the association between daily air pollutants and outpatient visits. The 90th and 95th percentiles of PM₁₀ concentrations were used to determine days of exposure to PM₁₀ predominantly from biomass burning. In other words, the high concentration at these percentiles indicates days of biomass burning. Their results showed that there was significant intra-annual variation in PM₁₀ levels, with the highest concentrations occurring during March, coinciding with peak biomass burning. The Incidence Rate Ratios (IRRs) between daily PM₁₀ and outpatient visits were elevated most on the same day as exposure for CLRD = 1.020 (95% CI: 1.012 to 1.028) and CBVD = 1.020 (95% CI: 1.004 to 1.035), with no association with IHD = 0.994 (95% CI: 0.974 to 1.014). Adjusting for CO tended to increase effect estimates. This is not surprising, but on the days of biomass burning, they did not find evidence of an exposure-response relationship with levels of PM₁₀. This result indicates that during high PM₁₀ (or PM_2.5) days, when biomass burning affects the population, people can take measures to avoid the harmful effect by sheltering or remaining indoors, avoiding exposure to highly polluted ambient air.

In a study on the effect of banning the burning of forests in Upper Northern Thailand (UNT), Uttajug et al. [43] reported that, after a burning ban was enforced in the UNT in 2016, the annual average number of hospital visits for respiratory diseases attributable to fire originated PM₁₀ decreased from 37,682 to 9944 (approximately 70% reduction from the pre-intervention period). In UNT, biomass burning is primarily for forest clearing. Their study covered the period from January 2014 to December 2018. Approximately 130,000 hospital visits for respiratory diseases were estimated to be attributable to fire-originated PM₁₀ in UNT during this period. The number of hospital visits for respiratory diseases due to fire-originated PM₁₀ decreased from 1.8% to 0.5% after the burning ban policy was implemented. The burning ban is only enforced in UNT for forest fires, but agricultural waste burning, which is at a larger scale and occurs throughout Thailand, is not affected.

Reddington et al., 2021 [20], in their study of the impact of forest and vegetation fires on air quality degradation and public health in Southeast Asia (including Mainland Southeast Asia and south-eastern China), found that by eliminating fire, the regional air quality across Southeast Asia would be significantly improved by reducing the population exposure to fine particulate matter (PM_2.5) concentrations by 7% and surface ozone concentrations by 5%. These reductions in PM_2.5 exposures would yield a considerable public health benefit across the region, averting 59,000 (CI: 55,200–62,900) premature deaths annually. They also found that high levels of poverty coincide with exposure to relatively high levels of PM_2.5 from fires. In other words, air pollution due to biomass burnings in SEA disproportionately impacted the poor and concluded that reducing forest and vegetation fires should be a public health priority for the Southeast Asia region. Reduction of long-term exposure to ambient PM_2.5 can have significant averted public health effects. Reddington et al., 2021 [20] estimated that eliminating fire emissions results in a 20% reduction in PM_2.5 exposure and a 13% reduction in mortality (1500 persons/year) in Cambodia. The figures for Laos are higher, 41% and 28% (1600 persons/year), respectively. Similarly, Thailand is also higher with 23% and 15% (10,800 persons/year), respectively, while Vietnam is least in proportions with 7% and 5% (5100 persons/year), respectively. They used the GLOMAP (v7) model with a multi-year run (2003–2015) for PM_2.5 and WRF-Chem 3.7.1 with a year run (2014) for ozone (O₃) and FINN1.5 fire emission to study the long-term impact of PM_2.5 and O₃ exposure. Our results are for the short-term health impact of PM_2.5 due to biomass burnings for a period of 13–20 in March 2019. Whether the results of long-term or short-term impacts of air pollution on public health due to biomass burnings in peninsular SEA, this is an important public health issue that needs to be addressed.

Our estimate of mortality due to biomass burnings in SEA at the end of the dry season in 2019 from 13 to 20 March 2019 based on the IER method is conservative as it capped the health response function at a daily average concentration of 30 μg/m³. A log-linear response function gives a higher value with capping.

Table A2. The estimated health impact (mortality), based on log-linear concentration response method, in Vietnam, Thailand, Cambodia and Laos due to biomass burnings from 13 to 20 March 2019 at the end of dry season in peninsular SEA.

	Vietnam	Thailand	Cambodia	Laos
Population (2019)	96,372,928	70,111,586	16,603,117	7,158,249
2019 incidence mortality rate (%)	0.65	0.71	0.67	0.62
Premature deaths (mean) due to biomass burnings 2019 (2019)	935	3290	411	1066
Proportion of effected population 2019	0.00097%	0.0047%	0.0025%	0.0148%

in Appendix E shows the results of mortality effects based on the log-linear response function.

Table A3. Comparison of health impact (mortalities) due to biomass burnings in SEA from some recent studies.

Authors	Study Period	Method	Health Impact (Mortality)	Impact Nature
Nguyen et al., (2022) [34]	March 2012	Modelling WRF-CMAQ BENMAP (log-linear exposure response function)	Thailand—~3629 Laos—~11131 Cambodia—~3625 Vietnam ~ 3376	Short term impact PM2.5 (health burden)
Punsompong et al., (2020) [41]	2016	Monitoring data All fires (forest and all agricultural burnings) IER exposure response function	Thailand—~18003 mortalities	Long term impact PM2.5
Reddington et al., (2021) [20]	2003–2015 (Full modelling for 2014)	Modelling GLOMAPv.7 (2003–2015) WRF-Chem (2014) FINN 1.5 emission All fires	Mainland SEA and southeast China ~ 59000/year averted deaths if no fires Thailand—~10800/year Vietnam—~5100/year Cambodia—~1500/year Laos—~1600/year	Long-term impact PM2.5 and Ozone (health benefit)
Present study	March 2019	Modelling WRF-Chem FINN 1.5 emission IER exposure response function	Thailand—~2171 mortalities Vietnam—~565 Cambodia—~315 Laos—~277	Short term impact PM2.5 (health burden)

in Appendix F summarises the comparison results of other recent studies and our study with respect to mortality due to biomass burnings in SEA during different periods. For morbidity results, over the whole SEA, hospitalisation due to cardiovascular diseases is small and insignificant (607 cases) when compared with hospitalisation due to respiratory diseases (11,956 cases) during the eight days of peak biomass burnings in March 2019 in this study.

Our study shows that the health burden caused by biomass burning at the end of the dry season, especially in March, in SEA is not insignificant in terms of public health and the associated economic cost from premature mortality and health care costs from hospitalisation and loss of productivity. The results are indicative as they depend on a number of factors and assumptions, such as the use of relative risks, the incidence rates for each health endpoint, and the air quality modelling uncertainty. Health statistics from each country under consideration are scarce and not in enough detail or simply not available. We have used the data based on WHO reports and relevant studies from scientific literature. Finally, people seeking shelter during high pollutant concentrations, such as during wildfires or extensive biomass burnings, would be relatively safe, but we cannot account in our study whether all people got shelters or how many were indoors. We assume that the population were exposed to the ambient daily air pollutant concentration, and the result is at the upper end of the true value, which cannot be determined exactly as there are other uncertainties as well in epidemiological studies.

Other study limitations include the lack of ground PM_2.5 monitoring data available for validation publicly available historical data is not available to us. For the period of our study in March 2019, we contacted the Pollution Control Department (PCD) of Thailand to obtain historical data but without success. No monitoring data from Cambodia and Laos is known to be available. For Vietnam, the monitored data is sketchy, recently, only low-cost sensors were used in some cities in Vietnam. The RR (relative risk) of respiratory hospitalisation due to large biomass or wildfire burnings are sourced from previous wildfire studies in the U.S. [27,30]. The corresponding values for SEA or East Asia populations are not available. Only RR values due to normal ambient anthropogenic air pollution from a recent study in Chiangmai are available for SEA in the literature [31]. These values, however, are not suitable as the episodic wildfire or biomass-burning events causing severe air pollution are not normal everyday occurrences.

4. Conclusions

Our study has shown that annual agricultural biomass burnings in peninsular Southeast Asia at the end of the dry season pose a significant public health issue in the region. Of the four countries, Thailand and Laos, where biomass burnings were most intense in 2019, have the most health impact in terms of the proportion of the affected population, especially the northern part of Thailand as well as the Bangkok metropolitan area, where there is high population and elevated PM_2.5 concentration occurred. The prevailing northern and north-easterly wind flow carried the emitted pollutants from northern and north-eastern biomass burnings to the south and southwest populated areas of Thailand. Northern Laos also experienced high concentrations of PM_2.5 and extensive smoke plumes, but the area has a lower population density compared to neighbouring Thailand. Vietnam and Cambodia are least affected due to less PM_2.5 concentration on the ground during those days of biomass burnings from 13 to 20 March 2019. In terms of the absolute number of mortalities, Thailand and Vietnam are most affected as these countries have higher populations and more exposure to pollutants than Cambodia and Laos.

Even though Vietnam has a much higher population density, especially in the Red River delta in the north and the Mekong River delta in the south, local biomass burnings are much less intense, and the Truong Son Mountain range along the Laotian border makes the pollutant dispersion and transport from northern Thailand and Laos to higher troposphere level where the smoke aerosols continue to be transported farther and less mixing to the ground in northern and central Vietnam on the other side of the mountain.

Each country cannot regulate biomass burning by itself as it is a regional-wide problem. This requires the need to coordinate together to mitigate the effect of annual regional agricultural burnings and the related transboundary transport of air pollutants, such as choosing or scheduling the most appropriate time and location with suitable meteorological conditions to allow agricultural biomass burnings with minimal effect across the region. In recognition of the impact on population health of air pollutants such as PM_2.5, the United Nations (UN) Sustainable Development Goals [44] aims to have governments across the world implement policies to reduce substantial mortality and illnesses from hazardous chemicals and air, water, and soil pollution and contamination by 2030 (Goal 3.9). It is important to address the issue of public health impact on the population due to large-scale biomass burnings in SEA by governments in this region. Various methods to dispose or utilise the biomass residues from agriculture can be considered, such as livestock feeding or converting to ethanol fuel or fertilisers to reduce the need for biomass burnings.