Numerical Study of the Transport and Convective Mechanisms of Biomass Burning Haze in South-Southeast Asia

This study aims to identify the vertical transport mechanisms that uplifted the forest fire emissions from Sumatra to the upper troposphere during the June 2013 haze crisis. WRF-Chem is used to simulate the formation and transport of biomass-burning haze during the study period of 18 to 26 June 2013. The South-Southeast Asian synoptic weather patterns and their effects on the transport of biomass-burning emissions from Sumatra to Peninsular Malaysia were studied computationally to explain the phenomenon. Results show that PM10 emissions were lifted to 200 hPa height (approximately 12 km) over the Strait of Malacca on 24 June. The two identified vertical transport mechanisms confirmed a previously conjectured convergence over the Strait of Malacca and orographic lifting over Peninsular Malaysia. These mechanisms were able to uplift the biomass-burning emissions to the upper troposphere and this could have significant long-range transport and global climatic effects.


INTRODUCTION
Biomass burning is a prominent recurrence in Southeast Asia (SEA).Intense forest fires have often been the result of intentional bush-burning that aims to clear the land for agricultural purposes (Quah and Johnston, 2001).In some regions of Indonesia such as Southwest Kalimantan and Southeast Sumatra, where fire prone peat swamp forests exist, extensive forest fires have occurred during the dry Southwest monsoon season (May-October) of 1994 and 1997 (Nichol, 1997(Nichol, , 1998)).1994 and 1997 were both El-Nino Southern Oscillation (ENSO) years and the intensification of the dry conditions over Indonesia has been an important contributing factor to the strong forest fires which occurred during that time (Heil and Goldammer, 2001).
Biomass burning has been reported to be a significant source of ozone precursor gases such as nitric oxide (NO), carbon monoxide (CO) and hydrocarbons (Crutzen et al., 1985).In SEA, biomass-burning haze (BBH) has been identified as an important source of tropospheric ozone (Fujiwara et al., 1999;Liu et al., 1999;Chan et al., 2000Chan et al., , 2003) ) while aerosols from such burning have been found to contribute largely to 'Atmospheric Brown Clouds' (Ramanathan and Crutzen, 2003).More importantly, BBH reduces air quality and imposes a very negative effect on human health (Dawud, 1998;Aditama, 2000;Kunii et al., 2002;Wiwanitkit, 2008;Hyer and Chew, 2010;Kim Oanh et al., 2011).Quah (2002) reviewed the issues and policy responses to the severe damage of BBH on the regional environment and the economy during the 1997 episode and has highlighted the need for prompt and effective measures in face of the biomass burning problem.Therefore the impact of BBH over SEA on the atmospheric environment, economy and general well-being cannot be overstated and is receiving more and more attention, not only regionally but globally.
One of the most important phenomena in large-scale haze is deep convection (Lin et al., 2009).This is the most frequently mentioned mechanism for the transport of biomass-burning products up in the troposphere.Deep convection is able to carry biomass air pollutants to higher altitudes over SEA and is a leading candidate to explain climate change in context (Folkins et al., 1997;Chan et al., 2003).A strong convection mechanism responsible for the uplifting of biomass-burning emissions over Indonesia, New Guinea and Malaysia has also previously been at the center of discussions (Folkins et al., 1997).Lin et al. (2009) have identified leeside troughs over Indochina as major contributors to the uplifting of the tracer in a tracer modelling study while Cheng et al. (2013) have identified that biomassburning emissions were lifted due to thermal forcinginduced upward motion and high terrain in the northern part of SEA.While the uplifting mechanisms have been studied in Indochina, there is a need to identify the convective mechanisms responsible for the uplifting of biomass-burning emissions in Southern SEA, as literature in this region on the mechanisms remain scarce.
Intense biomass burning occurred over Sumatra in June 2013.Satellite imageries show that large scale forest fires from Sumatra were mainly responsible for the occurrence of haze in the region (NASA, 2013).Data obtained from the Malaysian Department of Environment (DOE) also show that Air Pollution Index (API) readings exceeded the hazardous level of 300 on several occasions in Peninsular Malaysia during the haze episode (Department of Environment Malaysia, 2013).In this study, the Weather Research and Forecasting model coupled with Chemistry (WRF-Chem; Grell et al., 2005) was used to study the haze crisis that hit South-SEA in June 2013.Particular attention was given to PM 10 emissions since PM 10 concentrations usually exceed those of other gases emitted during forest fires and the WHO (World Health Organization) guidelines (Radojević and Hassan, 1999).Most studies on BBH have been restricted to immediate dispersion and haze trajectories to neighboring countries (Lin et al., 2009) when measurements have showed that deep convection in SEA can lift biomass-burning emissions up in the troposphere (Folkins et al., 1997).It is therefore important to get a better understanding of the uplifting mechanisms, and this is a gap that this study aims to fill.
This study aims to: (1) study the transport of BBH in SEA, in particular Peninsular Malaysia, by integrating the appropriate weather and pollutant datasets as well as forest fire hotspot data for the June 2013 forest fire study period into WRF-Chem, (2) identify weather conditions that caused and intensified the transport of BBH from Sumatra to Malaysia during the forest fire episode, (3) demonstrate the influence of the topography on air motion and therefore the transport of BBH over the study area, and (4) identify regions of deep convection near and downwind of the forest fire hotspots and analyze horizontal and vertical BBH plume dispersion.The achievement of these objectives will improve our understanding of the transport mechanisms of BBH in South SEA and will hopefully help to develop more effective inter-governmental policies on the issue.More importantly, this study may help in the formulation of emission control policies which aim at reducing emissions at times when the transport of biomass burning emissions are likely to be carried over regions of deep convection, thus limiting long range transport of these emissions.

Observations
API readings obtained from the Malaysian DOE (DOE Malaysia, 2013) showed that a major haze episode hit Malaysia during the second half of the month of June 2013.
The air pollutants included in the calculation of API in Malaysia are ozone (O 3 ), carbon monoxide (CO), nitrogen dioxide (NO 2 ), sulphur dioxide (SO 2 ) and suspended particulate matter of less than 10 microns in size (PM 10 ) and a summary of the API categories is listed in Table 1.The observed data indicate that the haze episode became apparent on 15 th June when API readings were first categorized as 'Unhealthy' in Melaka City and Muar during that period.'Hazardous' readings were first recorded on 20 th June in Muar while Melaka City was hit with readings reaching 415 on 23 rd June and Port Klang with a maximum API reading of 487 on 25 th June.The highest API reading of 746 was recorded on 23 rd June at Muar. 'Unhealthy' API readings were recorded in several regions of Malaysia until the highest API readings classified as 'Moderate' were recorded on 27 th June, marking the end of the haze episode.The forest fires which were responsible for the occurrence of haze were detected by the Moderate Resolution Imaging Spectrometer (MODIS) and the hotspot satellite data for 19 th , 20 th , 23 rd and 24 th June 2013 are illustrated in Fig. 1.The red spots represent the fire hotspots and the daily observed fire count is indicated at the bottom left corner of each panel.

Domain Setup and Data Integration
The WRF-Chem model (Grell et al., 2005) was used to simulate BBH over SEA during the June 2013 haze episode.The model domain is shown in Fig. 2. It consists of a 27-km grid resolution outer domain with 180 × 117 grid points and a nest (d02) set at 9-km grid resolution with 190 × 190 grid points.The outer domain covers Indonesia and the Southern part of Indochina, including Malaysia, while the nest focuses on Sumatra and Peninsular Malaysia.The vertical level configuration was adapted from Lin et al. (2009) and consists of 35 sigma levels which were most tightly packed near the surface.The lowest level was set at about 16 m above the surface while the model top was set at 50 hPa.The meteorological initial and boundary conditions were obtained from the NCEP FNL (Final) Operational Global Analyses (2000) with 1 × 1 grid spacing at 6-hour intervals.
To ensure that the model did not violate the Courant-Friedrichs-Levy (CFL) stability criterion (Courant et al., 1928) the time step of the simulation was set at 120 seconds (less than the recommended maximum of 6 × grid spacing).The model results were output every hour.
The trace gas and aerosol emissions fields were preprocessed using an adapted version of PREP-CHEM-SRC (Freitas et al., 2011).PREP-CHEM-SRC integrates emissions inventories and forest fire hotspots onto a specified domain.The biomass burning emissions were estimated using the Brazilian Biomass Burning Emissions Model (3BEM; Longo et al., 2010) and were injected into the atmosphere using the plume rise model described in Freitas et al. (2006Freitas et al. ( , 2007Freitas et al. ( , 2010)).The forest fire hotspots database integrated in the model includes a combination of the MODIS fire locations database (U.S. Naval Research Laboratory, 2013) and the Wildfire Automated Biomass Burning Algorithm (WFABBA) database (University of Wisconsin-Madison, 2013).In order to include Emissions Database for Global Atmospheric Research (EDGAR) version 4.2 in the simulations an additional script was required to convert the data to the appropriate format (HDF5).Two-way nested 24 h WRF-Chem simulations were initialized at 00UTC on 16 June 2013 and the emission inventories had to be updated for each day of simulation.In addition, each simulation was restarted from the previous run to ensure that the chemistry was initialized from the previous 24 h forecast.11 individual days of simulation were run and the first 2 days of run were treated as spin-up.

Model Chemistry and Physics
The Morrison double-moment microphysics scheme (Morrison et al., 2009) which is capable of predicting the mixing ratios and the number concentrations of five species, namely, cloud droplets, cloud ice, snow, rain and graupel, was used to represent the resolved-scale cloud physics.The longwave and shortwave radiative processes were represented by the Rapid Radiative Transfer Model for general circulation models (GCMs) longwave and shortwave radiation schemes (RRTMG; Mlawer et al., 1997;Pincus et al., 2003) respectively.These two radiation schemes can interact with WRF-Chem to include aerosol scattering.Convective parameterization was performed by the Grell 3D cumulus scheme, which is an improved version of the Grell-Devenyi (GD) cumulus scheme (Grell and Dévényi, 2002).In addition, the Grell 3D cumulus scheme allows for feedback from parameterized convection to the radiation schemes.The Noah land-surface model (Chen and Dudhia, 2001) was used to obtain moisture and thermal fluxes form the surface while the Mellor-Yamada-Janjić (MYJ) scheme (Janjić, 1996(Janjić, , 2002) ) was the selected planetary boundary layer scheme for the simulations.The Regional Acid Deposition Model, 2 nd generation (RADM2) (Stockwell et al., 1990) gas phase chemistry mechanism was used with the Georgia Tech/Goddard Global Ozone Chemistry Aerosol Radiation and Transport (GOCART) model (Chin et al., 2000).The aerosol direct effect was turned on by activating aerosol radiative feedback in the simulation.

Model Evaluation
The model was first evaluated by comparing the vertical profile at Sepang to sounding data provided by the Department of Atmospheric Science at the University of Wyoming (University of Wyoming, 2013).Vertical profiles can be compared by studying the change of potential temperature θ (K) with change in geopotential height Z (km).The potential temperature and the geopotential height corresponding to the location of the Sepang weather station (2.71°N, 101.70°N) were extracted for 18 th to 26 th June 2013 and compared to the sounding data.Table 2 shows the correlation coefficients (r) and the Root Mean Square Error (RMSE) obtained from the comparison of the model and sounding data vertical profiles.The evaluation results show that the model performs very well when the vertical profiles were compared within a maximum geopotential height of 12 km.The correlation coefficients were all higher than 0.98 and the highest RMSE observed was only 3.98 K.While the study of deep convection is of major concern in this work, the tropopause which resides at an approximate geopotential height of 16 km to 18 km over Sepang falls out of the boundaries of our region of interest.The model developed and presented in this work therefore proves to be an effective tool in simulating atmospheric conditions in South-SEA.Figs. 3 and 4 show the vertical profiles and scatterplot of potential temperature values for 19 th and 24 th June at Sepang station.Plots for these 2 dates were chosen since the most intense hotspot activity occurred on the 19 th and 24 th June as shown in Fig. 1.In addition, high PM 10 concentrations were recorded on these 2 days as shown in Fig. 5. Fig. 4 shows that the correlation coefficient and RMSE for the observed and simulated potential temperature were 0.987 and 3.98 K respectively on 19 th June and 0.994 and 3.00 K respectively on 24 th June.
The simulated aerosols were evaluated by comparing the PM 10 readings at 4 of the 52 air quality monitoring stations operated by the Malaysian DOE (DOE Malaysia, 2013), namely, Johor (CAS 019), Perak (CAN 041), Kuala

Meteorological Conditions
The variations in sea level pressure (SLP) and wind patterns during the haze episode are shown in Fig. 6.Southwesterlies prevailed over Sumatra and Peninsular Malaysia since the Southwest monsoon season occurs from the month of May to October in SEA.These southwesterlies were intensified by the presence of a low pressure belt that usually forms between latitude 10°N and 15°N.This low pressure belt is also known as the Northern Intertropical Convergence Zone (ITCZ).On 18 th June, there was a gradual intensification of the low pressure belt which drew stronger winds from the Indian Ocean and by 19 th June two major low pressure systems formed over the Gulf of Thailand near Cambodia and over the South China Sea.The Southern ITCZ, which separates the southwesterlies from the Indian Ocean and the southeast trade winds, could be observed near Southern Peninsular Malaysia towards the Northeast in Figs.6(b) and 6(c).The occurrence of the low pressure systems was verified through NASA's Worldview interactive tool (NASA, 2013).
On 20th June, as the low pressure system over Cambodia moved eastward towards the South China Sea and it began to merge with the already existing low pressure system in Fig. 6.SLP (hPa) (shaded colours) and wind fields (m s -1 ) (arrows) at 06 UTC (a-h) from 18 th to 25 th June.that region as shown in Fig. 6(c).The winds over Peninsular Malaysia gradually shifted from blowing to the northeast to an easterly direction while weak winds prevailed over southern Peninsular Malaysia and Singapore.By 21 st June, a stronger low pressure system formed from the two smaller low pressure systems and southwesterlies were established over southern Peninsular Malaysia while westerlies predominated over northern Peninsular Malaysia and Thailand.The low pressure system further intensified on 22 nd June and began to travel towards the west over Vietnam but gradually lost in intensity on 23 rd June.The southwesterlies blowing from Sumatra over Peninsular Malaysia gradually shifted to southerlies and southeasterlies on 23 rd and 24 th June respectively as shown on Figs.6(f) and 6(g).A new low pressure system gradually formed over the South China Sea, resulting in a high pressure region over Sumatra.Southeasterlies also blew along the Strait of Malacca while northwesterlies blew over the Gulf of Thailand as a result of the low pressure system on 25 th June as shown in Fig. 6(h).
High pressure over Sumatra may have inhibited precipitation over Sumatra and contributed to the intensification of the forest fires in Riau.In fact, Reid et al. (2012) reported that decreased precipitation and increased severe fire events were observed in Borneo and Sumatra when tropical cyclones occurred over the northern South China Sea.The fires might also have been sustained by dry warm air flowing over the Barisan Mountains onto Riau.As moist air coming from the Indian Ocean is force-lifted over the windward side of the mountain, moisture is lost to cloud formation thus resulting in dry air which warms up as it flows down over the leeward side of the mountain.
The mountain flow can be observed on Figs.7-9.

Horizontal PM 10 Emissions Transport
In order to study the horizontal and vertical extent of the haze plume during the 2013 episode, PM 10 plots were extracted from the simulation results.Fig. 10 shows the horizontal PM 10 plume at 998 hPa height (~16 m) over Sumatra and Peninsular Malaysia during the haze episode.
On 16 th and 17 th June few fire hotspots were observed over Riau, Sumatra with little emission being carried over the Southwest coast of Peninsular Malaysia (not shown).
The number of hotspots increased on the 18 th and reached a high on 19 th June as shown in Fig. 1.Emissions increased drastically on the 19 th and were initially transported over Malaysia by southwesterlies and thereafter by westerlies.Fig. 10(c) shows that on 20 th June more haze was transported over Peninsular Malaysia where Muar was the most affected city on that day.Forest fires were still predominant on Sumatra on 21 st June and biomass-burning emissions were still transported to Peninsular Malaysia with a shift in the plume direction from easterly direction to northeasterly direction due to the wind shift.Fig. 10(d) shows that emissions occurred over Batu Pahat, Melaka and Seremban due to the shift in direction of the plume flow and the API readings increased accordingly over these regions.The location of these regions can be seen on Fig. 2(b).
Fire hotspots were still prominent over Sumatra on 22 nd and 23 rd June.The emissions were transported in a northnortheasterly direction from Sumatra to Malaysia on 22 nd June and the most affected city was Muar where a maximum reading of 1050 µg m -3 to 1200 µg m -3 was recorded from

Vertical PM 10 Emissions Transport and Deep Convective Motion
One of the main objectives of this study is to identify regions of deep convection near and downwind of the forest fire hotspots that were responsible for the transport of BBH to higher altitudes, which is not easily observed  d02) with lines XX' and YY' which represent the vertical cross-sections studied in this section.Depending on the day of simulation, their center points correspond to high emission spots and the cross-sections can therefore show vertical convective conditions above those regions.In addition, the crosssections also lie along the direction of plume flow (Fig. 10) and can therefore depict the advective conditions involved.Hence the vertical cross-sections serve to identify the air motion responsible for the uplifting of the emissions and their transport from Sumatra to Peninsular Malaysia.Table 3 shows the location and orientation of each cross-section.
The vertical PM 10 emission plume and vertical velocity along the two cross-sections were plotted for every day of the haze episode.The wind fields were calculated from the horizontal velocity along a particular cross-section and 10 times the vertical velocity (10*W).The most intense hotspot activity during the June 2013 episode occurred on 19 th and 24 th June and it was found that emissions reached the highest altitude on these 2 days.Hence only the results obtained  plots along XX' while Fig. 9 illustrate the PM 10 plots along YY'.The most prominent vertical transport of PM 10 was found to take place over the Strait of Malacca on 19 th and 24 th June along both cross-sections.Fujita et al. (2010) described the mechanism by which a morning maximum in the diurnal precipitation was formed by the convergence of cold outflows from Sumatra and Peninsular Malaysia over the Strait.These outflows are generally produced by precipitation systems from the previous evening and converge over the Strait at around midnight.Fig. 7 shows that PM 10 emissions reached at least 350hPa over the Strait near 2°N at 11 UTC on 19 th June due to strong vertical motions that reached the upper troposphere above that region.As the direction of plume flow switched northward on 24th June, PM 10 concentrations along the cross-section XX' increased as compared to 19th June.At 14 UTC, the PM 10 plume had reached a height of 500 hPa over the strait while a second PM 10 band over Malaysia near 3.3°N reached at least the 200 hPa (~12 km) in the upper troposphere as shown in Fig. 8(a).This deep convective transport can be attributed to a strong orographic effect which occurs as the wind is force-lifted over the slope on the West coast of Malaysia and its mountains deeper inland.As the morning precipitation maximum forms (Fujita et al., 2010) at 18 UTC, a higher concentration PM 10 plume rises to at least 200 hPa over the Strait of Malacca as shown in Fig. 8(b) confirming further the transport of biomass burning emissions to the upper troposphere in the region.
Fig. 9(a) shows the transport of PM 10 emissions along YY' at 14 UTC on 19 th June.Much higher concentrations of PM 10 were observed along YY' as compared to XX' for the same day since YY' followed the direction of plume flow on 19 th June.A residual layer was formed between 950 hPa and 850 hPa as the plume was transported in a northeasterly direction.However, a very distinguishable peak which reached 500 hPa was still formed over the Strait of Malacca thus demonstrating the strong influence of convection in this region.In contrast, while a similar peak was formed on the 24 th closer to the coast of Sumatra, lower PM 10 concentrations were observed in the residual layer which extended from 900 hPa to 750 hPa, due to the switch in the direction of plume flow.This is shown in Fig. 9(b).

CONCLUSION
In order to study the 2013 SEA BBH case, WRF-Chem simulations were performed to identify the weather conditions that caused and intensified the transport of haze from Sumatra to Peninsular Malaysia.The model was run from 16 th to 26 th June 2013 and was evaluated by comparing the simulation vertical profile at Sepang to sounding data and PM 10 emissions to readings recorded by the Malaysian DOE.The prevailing convective conditions were also studied to understand how far deep convection affected vertical haze plume dispersion over the region.
The results showed that southwesterlies from the Indian Ocean were mostly responsible for the transport of biomassburning haze from Sumatra to Peninsular Malaysia.Variations in the wind direction occurred as low pressure systems forming the Northern Intertropical Convergence Zone travelled eastward and westward.The haze plume was directed over the southern part of Peninsular Malaysia during the first half the episode and then shifted northward to affect the west coast of Malaysia during the second half.
Muar and Melaka city were found to be the two most affected cities during the haze episode.
To study the vertical haze plume dispersion and the regions of deep convection, the results were plotted along vertical cross-sections for 19 th and 24 th June.The PM 10 plume was lifted to at least 350 hPa and 200 hPa on 19 th and 24 th June respectively due to the convergence of outflows from Sumatra and Peninsular Malaysia over the Strait of Malacca.Orographic lifting over the West coast of Malaysia and the Malaysian mountains also lifted the haze plume to at least 200 hPa.The uplifting of the biomass-burning emissions to such heights could have long range transport effects if the emissions are carried aloft by the prevailing upper level winds.

Fig. 1 .
Fig. 1.Fire hotspots (shown as red spots) detected from MODIS satellite data on 19 th , 20 th , 23 rd and 24 th June 2013 (NASA, 2013).The fire count is indicated at the bottom left hand corner of each panel.

Fig. 5 .
Fig. 5. Comparison of observed and model simulated PM 10 emissions for the period 19 th to 25 th June 2013 at (a) Johor, (b) Perak, (c) Kuala Lumpur and (d) Klang.

Table 1 .
Summary of the ranges of index values and the corresponding categories for theMalaysian API system  (Department of Environment Malaysia, 2015).

Table 2 .
Correlation coefficient, RMSE and Average Percentage Error of vertical derived from WRF-Chem and sounding data.

Table 3 .
Location and orientation of vertical cross-section on nest (d02).and24th June are presented and discussed in this section.The identified uplifting mechanisms are illustrated and discussed below.Figs.7 and 8show the vertical PM 10