Vertical Profiling of Aerosol Types Observed across Monsoon Seasons with a Raman Lidar in Penang Island , Malaysia

Although technology has advanced in the past few decades, the understanding of the effect of aerosol on Earth’s climate has remained largely unknown. Even if considerable effort has been made by researchers around the world to study this pressing issue, the non-uniform distribution of aerosols is still a huge challenge on global aerosol characterization studies. Without extensive and reliable measurements from most regions in the world, a complete understanding of aerosol characteristics cannot be achieved. In this study, a Raman LIDAR was used to acquire data to calculate the rangedependent extinction-to-backscattering ratio during a very particular period: when the Northeast monsoon season shifts to the Southwest monsoon season. During this time period, monsoonal winds change direction, leading to changes in aerosol type and properties above Penang Island due to the global atmospheric circulation. From the retrieved extinction-tobackscattering ratio, it is possible to differentiate the aerosol types present at various altitudes. It was found that background marine and urban aerosols were present above Penang Island, as marine and urban aerosols were present both below and above the PBL height regardless of the monsoon season. Additionally, biomass burning aerosols and aged forest fire aerosols were occasionally found, which were normally associated with cases of pollution, such as those in early March 2014 and June 2014. Finally, it was found that during a haze episode, the entire atmospheric column was dominated by wood burning aerosols. AERONET sunphotometer data have been used to validate LIDAR findings. Radiative effect of background and transported aerosols are evaluated with the Fu-Liou-Gu radiative transfer model.


INTRODUCTION
Aerosols have attracted a substantial research interest over the last few decades.One reason is because aerosols are actively involved in global climate change.Aerosols interact with the outgoing infrared (IR) Earth radiation and incoming solar radiation (SW), affecting the Earth-Atmosphere radiation budget (Pöschl, 2005).Scientists are currently unable to completely understand aerosol's influence on Earth's radiation budget.This problem has been emphasized in the recent Intergovernmental Panel on Climate Change (IPCC) report in 2013.Due to the high variation and complex mixing of aerosols, scientists have yet to discover the results of aerosol interactions with radiation and clouds (IPCC, 2013).This area remains one of the research areas requiring additional work.Aerosols are non-uniformly distributed, both vertically and horizontally, which causes highly heterogeneous radiative forcing, leading to regional and global climate change (Akimoto, 2003).In some cases, two adjacent locations might have different aerosol climatology and characteristics, as aerosols might significantly influence certain regions in the world but have a negligible influence on other regions (Matthias and Bosenberg, 2002).Hence, without extensive and reliable measurements in all regions around the world, it is difficult to conduct global aerosol characterization studies.
The National Aeronautics and Space Administration (NASA) Aerosol Robotic Network (AERONET) is a groundbased federated network of sun-photometers deployed all over the world since the 90's (http://aeronet.gsfc.nasa.gov).From the radiometer signal is possible then to measure the atmospheric optical properties and precipitable water, even at the most remote locations around the world (Holben et al., 1998).Although the AERONET sunphotometer provides nearly global coverage, these measurements lack of information of the vertical profile of the aerosol properties, as the retrieved values show only an average of the atmospheric column, i.e., AERONET data cannot distinguish between the aerosols present in the planetary boundary layer (PBL) and the free troposphere (Muller et al., 2007b).However, sunphotometer data can give an idea of what type of aerosol is dominating the atmospheric column, LIDAR measurements, on the other hand, are vertically resolved.They can complement the sunphotometer measurements by providing height-dependent measurements and helping to determine the aerosol characteristics at various heights of the atmospheric column (Mattis et al., 2004).Various research endeavors have been conducted by synergizing the LIDAR measurements with the sunphotometer measurements (Chazette, 2003;Pedrós et al., 2010).However, results from this approach poorly resemble the actual atmospheric conditions, as only a single-valued, range-independent atmospheric parameter is obtained, such as the LIDAR ratio, while in fact, there might be several atmospheric layers that exist in the atmosphere (Royer et al., 2011).Therefore, the Raman LIDAR is preferred to study the actual atmospheric conditions as it resolves independently the extinction and backscattering coefficients, providing a direct measurement of the LIDAR ratio atmospheric profile, with respect to the single-wavelength elastic-backscatter LIDAR.This is because the Raman LIDAR makes use of the weak inelastic scattering of air molecules (usually nitrogen), which allows the aerosol extinction and backscatter coefficient to be simultaneously and individually determined without making any critical assumption of the LIDAR ratio, as in the case of the elasticbackscatter LIDAR (Ansmann et al., 1990(Ansmann et al., , 1992a, b), b).By doing so, the range-dependent LIDAR ratio can be determined, and the possible aerosol types that exist in various atmospheric layers can be retrieved.
Over Europe and North America, extensive studies on aerosol characteristics have been conducted.However, similar research has just started gathering pace in Southeast Asia.To completely understand the influence of aerosols on Earth's climate, a thorough aerosol study must be conducted in Southeast Asia.Moreover, many large-scale meteorological phenomena occur in Southeast Asia, such as Asian monsoons and El Niño Southern Oscillation (ENSO).These largescale meteorological events interact with the unique landsea distribution and create many different types of local weather systems (Lim and Samah, 2004).This causes the aerosol properties to constantly change with different weather systems, which again highlights the importance of conducting an aerosol characterizations study over this region.Muller et al. (2007a) summarized the LIDAR ratios for various aerosol types acquired using Raman LIDAR in various regions of the world.Results from this study provide LIDAR ratio values that can be used as a first-guess in elastic backscatter LIDAR data processing.However, this study mostly focuses on LIDAR ratio at 532 nm, with few results at the 355 nm wavelength.Moreover, differences in local weather patterns alter aerosol characteristics compared to the same aerosol types at neighboring regions.Thus, it is important to study localized aerosol characteristics in the Penang Island region.
In this study, a Raman LIDAR was used to measure the range-dependent LIDAR ratio in Penang Island, Malaysia, and the corresponding aerosol types at various altitudes in the atmosphere were determined.In addition, the dominant aerosol types during different monsoon seasons were studied as well because this research was conducted from the end of the Northeast monsoon season until the beginning of the Southwest monsoon season.This is the first study in Malaysia using measurements from a Raman LIDAR and could serve as a baseline for future research on aerosols and regional aerosol climatology.The Raman LIDAR findings are then validate with AERONET sun-photometer data and the aerosols radiative effect is evaluated with the Fu-Liou-Gu radiative transfer model.

STUDY AREA
The LIDAR was setup on the roof top of the School of Physics, at the Universiti Sains Malaysia, Penang, Malaysia (5°21′30.06″N,100°18′8.1″E,51 m above mean sea level (AMSL).Penang Island is located in the Strait of Malacca, sandwiched between Peninsular Malaysia and Sumatera Island.The east side of the island is a highly industrialized, urbanized and populated area, whereas the west side of the island is a more rural area with a smaller population compared to the east side.The east and west sides of the island are separated by a mountain range.Fig. 1, a topographic map produced from the global digital elevation model (GDEM), shows the high-mountain and low-plain regions of Penang Island.

INSTRUMENTATION
Table 1 summarizes the main technical specifications of the Raman LIDAR used in this study.The Raman LIDAR, model number ALS320-ESS-D200used in this study, was manufactured by Raymetrics S.A. at Athens, Greece.The Raman LIDAR consists of a Nd:YAG tripled laser emitting pulsed radiation at 355 nm, with a repetition rate of 20 Hz.The outgoing laser pulse is collimated with a beam expander, which expands the beam diameter up to 10 times before sending itinto the atmosphere.A 20-cm diameter Cassegrainian telescope was responsible for collecting the radiation backscattered from both atmospheric particles via elastic scattering and nitrogen gas via inelastic scattering through the vibrational-rotational Raman band of nitrogen at 387 nm.Both the elastic and Raman LIDAR backscattered signals are separated by a dichroic mirror and interference filters.All of the signals, recorded with photomultiplier tubes in analogue and photon-counting mode, are stored at 1 min temporal resolution and 7.5 m spatial resolution, respectively.

METHODOLOGY
The LIDAR took measurements 3 hours after sunset (1200 UTC, 8 p.m. local time) from Monday to Friday.Because the Raman signals are very weak, the LIDAR has to be operated in the absence of strong background solar radiation (Franke et al., 2001;Müller et al., 2003).No measurements were taken on rainy days.During a haze episode over Penang Island, which occurred on 3/3/2014 and 4/3/2014, the LIDAR was operated continuously throughout the night until the haze subsided.
The processing of the Raman LIDAR data was  Analog + photon-counting Vertical resolution 7.5 m Full-width half maximum bandwidth ~0.5 nm @ 355 nm ~0.4 nm @ 387 nm conducted using analysis software provided by the LIDAR manufacturer.First of all, the 3 hours of Raman LIDAR data were averaged into 1, and the vertical resolution was reduced to 60 m by averaging every 8 range bins into 1.This process was performed to increase the signal-to-noise ratio of the data.Next, the background radiation was subtracted from both the analogue and photon-counting signals.Then, a dead-time correction was applied to the photon-counting signal, as suggested by Donovan et al. (1993).Subsequently, the usage of analogue, photon counting or glued (combine) signals in the subsequent processing steps must be determined.This is because analogue detection normally consists of strong backscattered signals that return from the lower atmosphere, typically up to 8 km.Beyond this range, analogue detection results in noisy signals that are difficult to process and the usage of analogue signal beyond this range is not encourage.Alternatively, photon counting detection is able to capture the very weak far field backscattered signals, which return from altitude around 8 to 20 km that cannot be captured by analogue detection.However, saturation normally occurs in the near field range using the photon counting detector, making it not suitable to detect the near field backscatter signals.The usage of glued signal allowed the LIDAR to capture both the near field and far field backscatterred signals.
In order to determine the usage of glued, analogue or photon counting signals, the dead-time corrected photon counting signal is examined.If the peak value of the deadtime corrected photon counting signal is above the maximum toggle rate (typically 10 MHz) and the background of the dead-time corrected photon-counting signal is below the minimum toggle rate (typically 0.5 MHz), then the glued signal should be used.This is because the analogue signals only contained near field backscattered signals, while the photon counting signals only contained the far field backscattered signals.Hence, in order to obtain signals from both near field and far field, both analogue and the photon counting signals should be glued.On the other hand, if the peak value of the dead-time corrected photoncounting signal does not exceed the maximum toggle rate, saturation does not occur in the near field range, then the dead-time corrected photon-counting signal should be used as it contains both near field and far field backscattered signals.Lastly, if the background of the dead-time corrected photon-counting signal exceeds the minimum toggle rate, then the analogue signal should be used.This is because the background noise is very high and it is not practical to include the far field signal in the processing since it mostly contained noise.Subsequently, the range corrected signal (RCS) is obtained by applying the distance square law correction (z 2 ) to each data point to compensate for the range-related attenuation from the atmosphere.
After obtaining the RCS, the Raman LIDAR data were processed according to the method suggested by Ansmann et al. (1992b), which independently calculated the aerosol extinction coefficient (α aer ) and aerosol backscatter coefficient (β aer ).The LIDAR ratio was then obtained by taking the ratio between the aerosol extinction coefficient and the aerosol backscatter coefficient.First, according to Ansmann et al. (1992b), the nitrogen Raman signal is given by where P λ R (z) is the signal received from distance z at the Raman wavelength, C is the LIDAR constant, O(z) is the overlap factor, N R(z) is the nitrogen molecule number density, dσ λR (π)/dΩ is the range-independent differential Raman cross-section for the reverse direction, ∝ mol (λ R , z ) is the molecule extinction coefficient at the laser wavelength, ∝ aer (λ L , z ) is the aerosol extinction coefficient at the laser wavelength, ∝ mol (λ R , z ) is the molecule extinction coefficient at the Raman wavelength, and ∝ aer (λ R , z ) is the aerosol extinction coefficient at the Raman wavelength.The aerosol extinction coefficient (α aer ) can be obtained viathe nitrogen Raman signal, as given by ,? ,? ,?
where P(z) is the power received from distance Z at the Raman wavelength of λ R if the laser pulse is transmitted at λ L , N(z) is the atmospheric number density, α mol is the extinction coefficient due to absorption and Rayleigh scattering by atmospheric gases, and particle scattering is assumed to be proportional to λ -k For our LIDAR system, λ L is 355 nm and is 387 nm.k is assumed to be 1; according to Ansmann et al. (1992b), one can assume k = 1 for aerosol particles with diameters comparable to the measured wavelength.
The aerosol backscatter coefficient (β aer ) was determined using signals from both the backscattered and the Raman channel.Two pairs of the measured signal P λ L and P λ R at height z and at a reference height z 0 were needed.The reference height z 0 is a height where the aerosol loading is weak and the backscatterred signal is primarily due to Rayleigh scattering.Next, the aerosol backscatter coefficient is given by ,? ,?
where β mol is the molecular backscatter coefficient, P λ R (z) and P λ R (z 0 ) are the Raman signal collected at a height and reference height z 0 , respectively.P λ R (z) and P λ R (z 0 ) are the elastic backscatter signal collected at a height z and reference height z 0 , respectively.Finally, the LIDAR ratio, L aer is given by ,?
The LIDAR ratio provides information on the types of aerosol suspended in the atmosphere.
Since the LIDAR ratio is a wavelength dependent parameter, the values used for aerosol typing may vary with the wavelength of the Raman LIDAR used in the study.For this study, all results discussed in the paper refer to the 355 nm LIDAR ratio.

Uncertainties Analysis
Ansmann et al. (1992b) and Mattis et al. (2004) found the relative error of the 355 nm LIDAR ratio using Raman LIDAR is about 15%-30%.In this study, the error is estimated to be 30%.The statistical error of the extinction coefficient is estimated as 5-10% after signal averaging.However, the particle extinction coefficient profile is retrieved using the USSA 1976 model, instead of radiosonde data.This further introduced uncertainties in the LIDAR ratio calculated as the particle extinction coefficient of the real atmosphere is different from that derived by the model.Next, although signal smoothing was conducted by reducing the spatial and temporal resolution, signal noise cannot be eliminated completely and introduces errors in the range of 5%-10%.The errors from extinction retrieval and signal noise result in a total 355 nm LIDAR ratio error in this study of 30%.

Limitation of the Study
Since the Raman LIDAR can only be operated during night time, it is not observing any of the diurnal cycle of the aerosol properties, as well as the PBL height variation during daytime, making the results of this studies skewed towards night time.PBL is the lowermost sub-layer of the troposphere, which is in direct contact with the ground surface.Within the PBL, aerosol concentration is the highest (Stull, 1988;Matthias and Bosenberg, 2002;Tsaknakis et al., 2011).The PBL height is affected by both the Earth's and solar radiation, as well as anthropogenic activities on the Earth's surface (Tsaknakis et al., 2011).Hence, many tropospheric activities, such as aerosol distributions, convection activity, and cloud and fog formation are affected by the PBL height (Liu and Liang, 2010).The PBL height shows a strong diurnal variation.During daytime, the PBL expended due to surface heating and reached its peak (typically around 1-2 km in height) around afternoon.After that, it collapses quick and becomes very shallow (typically less than 500 m in height) at night.This is because the surface layer becomes stable due to infrared radiative cooling (Stull, 1988;Garatt, 1994;Liu and Liang, 2010).However, there was no research conducted to extensively study the diurnal variation of PBL height in Penang Island or Malaysia, hence the actual nocturnal boundary layer (night time planetary boundary layer) height cannot be determined.Thus, it was assuming that the nocturnal boundary layer height was maintained at 500 m throughout the study period.Since the LIDAR used in this study can only acquire signal returning from 660 m and above, it cannot acquire any LIDAR ratio for aerosol suspended in the PBL.Hence, the data presented in this study were limited to the region just above the PBL (from 500 m to 1500 m in height) and the lower free troposphere (from 1500 m to 3000 m in height).
The Raman LIDAR used in this study was a single wavelength Raman LIDAR at 355 nm.Using the 355 nm LIDAR ratio alone is not ideal for aerosol typing, but can still provide important information about the differences in aerosol characteristics at varying altitudes.The LIDAR ratio can only permit the determination of the most dominant aerosol types in the atmosphere.In order to determine every aerosol types present in the atmosphere, multi-wavelength Raman LIDAR with depolarization ratio channel should be used.In this way, the Ångström exponent and depolarization ratio of the aerosol can be measured.These datasets can be used to determine the size and the roundness of the aerosol, which is crucial in order to determine the aerosol types.

Range-Dependent LIDAR Ratio Variation over the Study Period
Fig. 2 show the mean LIDAR ratio over the 500-1000 m layer and 1000-1500 m layer, while Fig. 3 show the mean LIDAR ratio over the 1500-2000 m layer, 2000-2500 m layer and 2500-3000 m layer, with the black vertical lines marking the boundaries of different monsoon seasons.The LIDAR ratio was calculated and averaged separately with every increase of 500 m in altitude in order to monitor the changes in the mean LIDAR ratio, as well as aerosol type at different height of the atmospheric layer.Although the PBL height was not calculated in this study, the PBL was generally found at < 500 m at night; it can be assuming that the 500-1500 m layer represented the region just above the PBL while 1500-3000 m layer represented the lower free troposphere.Generally, the LIDAR ratio shows a large variability (10 sr-120 sr; Ackermann, 1998) According to Fig. 2, in the region just above the PBL, the LIDAR ratio values are generally concentrated at approximately 20 ± 6 sr to 60 ± 18 sr, whereas a LIDAR ratio greater than 60 sr was only found occasionally around early March 2014 and during June 2014.According to Muller et al. (2007a), the LIDAR ratio is approximately 20 ± 6 sr to 26 ± 8 sr in the PBL for marine aerosol.However, polluted marine aerosols might have higher LIDAR ratio values, ranging between 30 ± 9 sr to 40 ± 12 sr (Muller et al., 2007a).It can be confirmed that marine aerosols were present in the region just above the PBL.Urban aerosols also are strongly present in the region just above the PBL.These urban aerosols normally consist of non-absorbing particles, such as ammonium or sulfate particles, which are typically derived from anthropogenic aerosols (Franke et al., 2001).Urban aerosols usually have LIDAR ratio values of approximately 30 sr to 60 sr (Franke et al., 2001;Muller et al., 2007a).
During early March 2014 and June 2014, LIDAR ratios of approximately 60 ± 18 sr to 80 ± 24 sr were detected.According to Muller et al. (2007a), aerosols with such LIDAR ratios are highly light absorbing.Hence, these LIDAR ratios might have originated from wood burning aerosols, as wood burning aerosols are extremely light absorbing.Interestingly, these LIDAR ratios were found in the period during which air pollution events readily occurred.Air pollution events in Malaysia usually occur during February to March and June to August and are normally caused by smoke haze, originating from forest fires.The forest fires either occur naturally due to hot and dry weather or are caused by forest clearing using fire for agricultural purposes that later develops into an uncontrollable wild fire.Hence, it can be confirmed that the LIDAR ratio values of 60 ± 18 sr to 80 ± 24 sr were caused by wood burning  aerosols.Lastly, LIDAR ratio values of 80 ± 24 sr to 120 ± 36 sr were detected, mostly during early March 2014.However, it was not clear which aerosol types caused such high LIDAR ratio values.We can speculate that aged forest fire aerosols advected from the Asian continent caused these LIDAR ratios.According to Muller et al. (2007a), the LIDAR ratio seems to increase during long-range transport from the Asian continent.This might be the case as such high LIDAR ratio values were only detected when air pollution (advected) events occurred.Similar trends were found for the mean LIDAR ratio at the lower free troposphere, with relatively lower LIDAR ratios observed.According to Fig. 3, the detected LIDAR ratios were generally concentrated in the range of 10sr to 50 sr.Similar to the region just above the PBL, the lower free troposphere was generally dominated by marine and urban aerosols at 25 ± 8 sr to 35 ± 11 sr and 30 ± 9 sr to 60 ± 18 sr, respectively.Wood burning aerosols at 60 ± 18 sr to 80 ± 24 sr were still during early March 2014 and June 2014, as were the aged forest fire aerosols at 80 ± 24 sr to 120 ± 36 sr.However, at an altitude of 2000 m to 2500 m and 2500 m to 3000 m, most of the LIDAR ratios detected were lower than 25 sr, which does not correspond to any aerosol type.It was suspected that at these altitudes, clear air with low and negligible aerosol content was present.Because most of the aerosols are contained in the PBL and aerosol concentrations in the lower free troposphere decrease sharply with increasing altitude, the chance of finding aerosols at lower free troposphere is low.Generally, only a layer of clear air was present there, causing the LIDAR ratios to be very low.
Based on the results shown in Figs. 2 and 3, the background aerosols for Penang Island consist of marine and urban aerosols due to the geographical location of Penang Island.Penang Island is situated in the Strait of Malacca and is a highly industrialized, urbanized and populated island.As a result, a mix of urban and marine aerosols can be considered as background condition.This is demonstrated in both Figs. 2 and 3, as the LIDAR ratio values corresponding to marine aerosols (10 ± 3 sr to 30 ± 9 sr) and urban aerosols (30 ± 9 sr to 60 ± 18 sr) were found throughout the study period.Although the LIDAR ratio of 30 ± 9 sr to 60 ± 18 sr might correspond to the detection of dust aerosol, the chance of finding dust aerosol in Penang Island is very low.This is because dust normally originated from desert and major deserts in the world normally lies around the latitude of 30°N and 30°S.Penang Island, on the other hand, lies in the latitude of 5°N, which is nowhere near any desert.
Additionally, wood burning aerosols and aged forest fire aerosols were normally found during pollution events, such as during early March 2014 and the month of June 2014.Accordingly, for the same months, AERONET sunphotometer retrievals show very high values of the Fine Mode Fraction (FMF) at 500 nm that put in evidence that biomass-burning is dominating the atmospheric column (Eck et al., 2010;Fig. 7).These aerosol types can be found at each range level, as shown in Figs. 2 and 3, with LIDAR ratio values ranging approximately 60 ± 18 sr to 80 ± 24 sr and 80 ± 24 sr to 120 ± 36 sr, respectively.Although the LIDAR ratio of 60 ± 18 sr to 80 ± 24 sr might be measured from black carbon emitted from urban or industrial area, this case is more likely smoke.MODIS active fire data shows several fires active in the region around Penang Island, in conjunction with the detection of high LIDAR ratio during the study period.Figs. 4 and 5 shows the MODIS active fire data collected from 1-15 March 2014 and 1-15 June 2014 respectively.According to Fig. 4, many active fire hotspots were found in Peninsular Malaysia and Sumatera Island, Indonesia during 1-15 March 2014.On the other hand, relatively lesser active fire hotspot was found in Peninsular Malaysia and Sumatera Island, Indonesia as compared to that found in March 2014.This proved that fires were present during the pollution events that occurred in early March and June 2014 and these fires were responsible for producing smoke, which was later transported to Penang Island.Lastly, clear air with negligible aerosol content could be found at lower free troposphere, as shown in Fig. 3, where the LIDAR ratio detected was very low, ranging from 4 ± 1 sr to 20 ± 6 sr.
The type of aerosol observed in this study changed as the monsoonal pattern changed (Northeast monsoon to inter monsoon season to Southwest monsoon).According to Fig. 2, in the region below the PBL, high LIDAR ratio value (approximately 60 ± 18 sr to 100 ± 30 sr) was found during early March 2014, a time where the Northeast  monsoon season was coming to an end.The aerosols present below the PBL were wood burning aerosol and aged forest fire aerosol during this period.Then, the aerosol present below the PBL changed to marine aerosol and urban aerosol with LIDAR ratio around 20 ± 6 sr to 60 ± 18 sr during the inter monsoon season, which begins around the month of April.When the Southwest monsoon season begins around the month of May, the LIDAR ratio detected generally maintained around 20 ± 6 sr to 60 ± 18 sr.However, LIDAR ratios in the range of 60 ± 18 sr to 80 ± 24 sr were occasionally detected, which marked the presence of wood burning aerosol.According to Fig. 3, similar phenomena were observed across the monsoon season in the region above the PBL.However, for most of the time, a layer of clean air with negligible aerosol content was found above the PBL, irrespective of monsoon season, which given by the extremely low LIDAR ratio value, ranging from 4 ± 1 sr to 20 ± 6 sr.It was also found that marine aerosol and urban aerosol are the background aerosols present in Penang Island, as these aerosols present in every atmospheric layers, irrespective of monsoon season.On the other hand, wood burning aerosol and aged forest fire aerosol normally present when pollution case occurs.
It is a misconception that the Northeast monsoon season in Malaysia is the wet season, in which the chance of precipitation is very high and pollution rarely occurs.However, this is only partially correct.The first few months of the Northeast monsoon season is the wet season, but during the latter period, the chance of precipitation is very low and the weather becomes hot and dry.Forest fires frequently occur during this period.The March 2014 haze episode was an example of a pollution event that occurred during the end of the Northeast monsoon season, when the weather was very hot and dry.Hence, wood burning and aged forest fire aerosols were detected together with background marine and urban aerosols during the end of the Northeast monsoon season (March 2014).After time passed and inter-monsoon season began, the amount of precipitation increased.During the spring transition, when the Northeast monsoon changed to the Southwest monsoon (April 2014), weak atmospheric circulation promoted the formation of thunderstorms, producing substantial rainfall (Lim and Samah, 2004).Rain washes away excessive aerosols in the atmosphere, leaving behind the background aerosols.Hence, the atmosphere during this period is the cleanest and only background marine and urban aerosols were observed in the atmosphere during this period.Entering the months of May and June, the Southwest monsoon occurred and the chance of precipitation decreased, and the weather again became hot and dry.The chance of pollution occurrence increased, and the LIDAR again detected wood burning aerosols together with the background marine and urban aerosols.Hence, this shows that the aerosol types present in the atmosphere do not depend on the monsoon season, but instead depend more on meteorological and pollution events that occur either locally or in neighboring countries.
The uncertainty of the LIDAR ratio measurement in this study was relatively high.These high uncertainties affected the high LIDAR ratio region (60 sr-120 sr) very much, making it difficult to differentiate wood burning aerosol from aged forest fire aerosol.This is because the large uncertainties caused the range of LIDAR ratio value to become very big and extend into the LIDAR ratio range of wood burning aerosol and aged forest fire aerosol.On the other hand, the effect of the uncertainties in low LIDAR ratio region (20 sr-60 sr) is relatively small.However, there are still some difficulties in differentiating marine aerosol from urban aerosol.This is because the LIDAR ratio of marine aerosol is ranging from 20 sr to 35 sr, while LIDAR ratio of urban aerosol is ranging from 30 sr to 60 sr.there is an overlap region in the LIDAR ratio of both aerosol types.The high uncertainties of the LIDAR ratio in this study certainly made the aerosol types determination harder.Moreover, the high uncertainties also made it difficult to differentiate urban aerosol from wood burning aerosol as the LIDAR ratio falls near the borderline of both aerosol types might be either one of the aerosol types.However, by using MODIS active fire data, one can easily discriminate both aerosol types.This is because wood-burning aerosol only present when there is active fire is present.If there is no active fire present in nearby region, then the doubtful aerosol is an urban aerosol and vice versa.Speculations about results from LIDAR measurements find confirmation from sunphotometery data.At 500 nm, Fine Fraction Mode (FMF) values greater than 0.8 strongly suggest biomassburning aerosol (Eck et al., 2010).Fig. 7 shows that FMF is close to 1 in March and June, as evicted from LIDAR measurements

Raman LIDAR Measurements of the March 2014 Haze Episode
During 3/3/2014 and 4/3/2014, a haze episode struck Penang Island.The Raman LIDAR operated throughout the night, acquiring data on the haze episode.According to Hee et al. (2015), this haze episode was short-lived but severe.The smoke haze originated from both fire hotspots in Malaysia and Sumatra Island.However, Hee et al. (2015) only showed the data from the backscatter channel of the LIDAR.Here, data from the Raman channel is presented.which have LIDAR ratios ranging approximately 60 ± 18 sr to 80 ± 24 sr.As the haze was still developing at 1200 UTC on 3/3/2014, the atmosphere was full of wood burning aerosols.The LIDAR ratio of the upper layer was slightly lower than that of the middle and lower layers.According to Hee et al. (2015), the upper layer consisted of transported aerosols, whereas the lower layer mainly consisted of local aerosols.Normally the transported aerosols are slightly older than local aerosols.Muller et al. (2007a) stated that young (locally produced) aerosols have a stronger impact on the observed aerosol optical properties in the PBL.Hence, aerosols in the lower and middle aerosol layer showed slightly higher LIDAR ratio values compared to the aerosols in the upper aerosol layer.
Three hours later, at 1500 UTC, the LIDAR only detected two aerosol layers, which were located at approximately 1000 m to 1500 m AGL and 1800 m to 2300 m AGL, as shown in Fig. 6(b).The lower aerosol layer had LIDAR values ratio ranging from 80 ± 24 sr to 100 ± 30 sr, which was presumed to be dominated by aged forest fire aerosols.The upper level, however, was believed to consist of fresh wood burning aerosols, as the LIDAR ratio for this aerosol layer was found to vary between 60 ± 18 sr to 70 ± 21 sr, which falls in the range of forest fire aerosols, as stated by Muller et al. (2007a).According to Hee et al. (2015), the atmospheric condition during 1200 UTC and 1500 UTC on 3/3/2014 were similar.Three aerosol layers were found in the atmosphere and the AOD at 355 nm lies around 2.3 to 2.4.Hence, it was suspected that the aerosol content also remain largely unchanged, and the upper layer should contained wood burning aerosol.During this time, it is suspected that the older wood burning (forest fire) aerosol began to descend into the PBL, mixing with other local aerosols.However, wood burning aerosols were still continuously emitted by their source(s) and transported to Penang Island.
After another three hours, at 1800 UTC, two aerosol layers were observed by the LIDAR, as shown in Fig. 6(c).The lower layer was located 1000 m AGL, with a LIDAR ratio ranging from 70 ± 21 sr to 80 ± 24 sr.Wood burning aerosols are considered the dominating aerosol in this aerosol layer.The upper layer, which was located approximately 2000 m AGL, had a lower LIDAR ratio.The LIDAR ratio varied between 50 ± 15 sr to 70 ± 21 sr.It was deduced that this layer consisted of a mixture of wood burning and urban aerosols.At this time, the haze had very much subsided, although biomass burning aerosols were still found in the atmosphere, and urban aerosols were detected in the upper aerosol layer.This suggests that fewer wood burning aerosols were being transported to Penang Island because the background urban aerosols were detected again.This is generally consistent with the results reported by Hee et al. (2015), as the haze had significantly subsided during this time.The AOD was the lowest during the haze episode (approximately 0.50), further signifying that there was only a small amount of aerosols suspended in the atmosphere of Penang Island compared to previous hours.Lastly,Fig. 6(d) shows the LIDAR ratio profile for 4/3/2014 at 1200 UTC.After one day, the haze had subsided.However, the polluted aerosols still remained in the atmosphere.The LIDAR observations showed that the atmosphere did not have a distinct aerosol layer.Aged forest fire aerosols generally dominated the entire atmospheric column, with a LIDAR ratio ranging from approximately 90 ± 27 sr to 120 ± 36 sr.This shows that although the haze episode had subsided, the aerosols still remained in this atmospheric column.It would take some time before the aged forest fire aerosols dispersed from the atmosphere.
According to Fig. 6(a), the lowermost layer found at around 700 m might be aerosols contained within the PBL.The height of the PBL will be affected by anthropogenic activities at Earth's surface, which in this case is the smoke haze pollution event and become relatively higher.Sunset time at Penang Island was around 1900-1930 local time, which corresponded to 1100-1130 UTC.Hence, the Raman LIDAR data collected at 1200 UTC was just after sunset.At this time, the infrared radiative cooling had just began, and the top of the PBL can still be detected at relatively high altitude.According to Fig. 6(b), the Raman LIDAR was unable to detect any aerosol layer bounded below the PBL.This is because at 1500 UTC or 2300 local time, the surface layer might be completely cooled through infrared radiative cooling since it is already 3-4 hours after sunset.The top of the PBL decreases to a region below 500 m, where the LIDAR was unable to detect any signals.The Raman LIDAR only managed to detect two aerosol layer suspended at the region just above the PBL and at the lower free troposphere.Similar to that at 3 hours earlier, the Raman LIDAR was unable to detect the top of the PBL after three hours elapsed, as shown in Fig. 6(c).The Raman LIDAR only can detect two aerosol layers located at region just above the PBL and at the lower free troposphere.Finally, at 1200 UTC on 4/3/2014, the Raman LIDAR detects a very shallow PBL located around 700 m in height and a thick residual layer of aged smoke at around 3500 m in height.

Radiative Effects of Background and Advected Aerosols
The radiative effects of background and transported aerosols are evaluated with the FLG radiative transfer model (Fig. 7).An accurate description of the model can be found in Lolli et al. (2016) and in Campbell et al. (2016).The FLG model evaluated the radiative effect (calculated as the difference between aerosol and pristine condition cases) at surface (SFC) and top of the atmosphere (TOA) for three different measured aerosol profiles: 1. Background marine-polluted aerosols confined in the PBL (0-1000 m) 2. Same as 1), but with smoke advected aloft (1000-1500 m) 3. Entire column dominated by smoke (haze episode, 0-3000 m) The results are reported in Table 2. TOA-SFC represents the possible absorption of solar radiations due to absorbing properties of biomass burning particles (aerosol profiles 2-3) within the atmospheric layer containing the aerosols.

SUMMARY AND CONCLUSIONS
In conclusion, the Raman LIDAR measurements found that background aerosols in Penang Island consist of marine and urban aerosols, with LIDAR ratio values ranging from approximately 20 ± 6 sr to 30 ± 9 sr and 30 ± 9 sr to 60 ± 18 sr, respectively.Additionally, wood burning aerosols (60 ± 18 sr to 80 ± 24 sr) and aged forest fire aerosols (80 ± 24 sr to 120 ± 36 sr) were present when pollution events occurred.Throughout the monsoon season, the background marine and urban aerosols were found during the study period, whereas wood burning and aged forest fire aerosols were found during early March 2014 (approaching the end of the Northeast monsoon season) and June 2014 (the beginning of the Southwest monsoon season).These pollution events consisting of smoke aerosols were found during periods of very hot and dry meteorological conditions, despite these periods occurring during the monsoon season.Therefore, the type of aerosol present over Penang is not related to the monsoon season, rather the meteorological conditions associated with biomass burning.Finally, the Raman LIDAR measurements during the haze episode showed that during the development phase of the haze episode, wood burning aerosols were dominant in the entire atmospheric column.When the haze began to subside, background urban aerosols were detected together with wood burning aerosols.After the haze subsided, wood burning aerosols were no longer detected in the atmosphere; however, aged forest fire aerosols were dominant in the entire atmospheric column.A few days were required before these aged forest fire aerosols completely dispersed from the atmosphere.Radiative transfer calculations put in evidence a strong cooling effect at surface of advected biomass-burning aerosol layers.However, this study is a localized study.Hence, the results only apply to Penang Island, not the general atmosphere in other region of the world.

Fig. 1 .
Fig. 1.Topographic map produced from the global digital elevation model (GDEM) which shows the high-mountain and low-plain regions in Penang Island.The five-edge star indicates the study area (Universiti Sains Malaysia, USM).

Fig. 2 .
Fig. 2. Mean LIDAR ratio over the 500-1000 m layer and 1000-1500 m layer with the error bars showing the total error.The black vertical lines mark the boundaries of different monsoon seasons.

Fig. 3 .
Fig. 3. Mean LIDAR ratio over the 1500-2000 m layer, 2000-2500 m layer and 2500-3000 m layer with the error bars showing the total errors.The black vertical lines mark the boundaries of different monsoon seasons.

Fig. 7 .
Fig. 7.The radiative implications of different shapes of aerosol atmospheric profiles using the FLG radiative transfer code.

Table 1 .
Technical specification of the Raman LIDAR.

Table 2 .
FLG radiative model results for different aerosol configurations at surface.