Aerosol climatology characterization over Bangladesh using ground-based and remotely sensed satellite measurements

Atmospheric aerosols are some of the most crucial climate constituent variables of the world’s climate system because they reshape the climate and radiative balance directly by scattering and consuming the incoming solar radiation (Ali et al., 2017), indirectly varying cloud optical properties (Ramanathan et al., 2001), and imparting condensation nuclei (Kaufman et al., 2005). Thus, given these unique traits, characterizing the physical, chemical, and optical properties in terms of improving radiative balance precision and optimizing climate models is essential.

Aerosol types, such as sea salt, mineral dust, sulfate, and organics, are typically reflective and scatter the solar radiation back to space, thereby contributing to the cooling of the atmosphere. However, several other aerosols have more absorbing properties than their scattering counterparts (Jose et al., 2016). Black carbon (BC) and brown carbon (BrC) emitted by biomass and fossil fuel combustion processes and iron oxides released by dust are the most prevalent absorbers in aerosol mixtures (Wang et al., 2011). Despite that, iron oxides and BrC have the highest ultraviolet (UV)–visible absorption (Liakakou et al., 2020), while BC displays constant absorption throughout the solar spectrum (Bergstrom et al., 2002). A full assessment of the climatic impact of aerosols entails a comprehensive understanding of aerosol concentrations, size, composition, absorption, and scattering properties (Russell et al., 2010; Bibi et al., 2017). Aerosol size distribution and radiation absorbance are 2 basic features that can be used to define absorbing aerosols over a region (Lee et al., 2010).

Dhaka, Bangladesh’s capital, is one of the country’s most populous cities. With a population growth rate of 3.5% and high percentage (77.4) of urbanization (http://dhaka.gov.bd/ assessed on December 6, 2021), the city has been undergoing enormous industrial and economic changes over the previous few decades (Pavel et al., 2021a). The surge of population and infrastructure has exacerbated the city’s air pollution problem, making it one of the most polluted in the world (IQAir, 2020). Bulk of the pollution is ascribed to the local sources, such as motor vehicles, brick kilns, industrial emissions, road dust, and long-distanced trans boundary pollution particularly in the winter season (Salam et al., 2003a; Begum et al., 2011; Salam et al., 2012; Begum and Hopke, 2019; de Foy et al., 2021; Jeba et al., 2021; Pavel et al., 2021b; Salam et al., 2021; Zaman et al., 2021a; Zaman et al., 2021b). Recent studies have revealed that the sources of polycyclic aromatic hydrocarbons at the coastal island of Bangladesh (Bhola) were mainly diesel engine combustion (30%), gasoline exhaust (25%), natural gas emission (23%), and biomass burning (22%) (Jeba et al., 2021). The sources of PM_2.5 in Bangladesh, according to Pavel et al. (2021a), were vehicular emissions (36.5%), soil dust (42.9%), biomass burning (14.2%), and industrial emissions (6.5%). The levels of BC in Dhaka are among the highest in the world often reaching above 20 µgm^–3 (Beekmann et al., 2015; Dasari et al., 2020). One of the most important portions of atmospheric aerosol particles is carbonaceous species (organic carbon [OC] and elemental carbon). Fossil fuel (crude oil, coal, and natural gas), and biomass are utilized in Bangladesh as energy sources, resulting in massive emissions of BC and OC, as well as other pollutants (Salam et al., 2013). Secondary organic aerosols (SOAs) formed from biogenic gases are widely distributed in the Indo-Gangetic Plain (IGP) region. Biomass burning (48%) and traffic (46%) were identified to be the major producers of BC aerosols in Dhaka in a recent source apportionment study (Salam et al., 2021). Begum et al. (2012) also identified traffic, coal, and biomass burning as the major sources of OC and BC in Dhaka. Stone et al. (2012) investigated biogenic contributions to SOA in the Southeast Asian regional haze and identified biogenic SOA as a substantial source of OC in the Himalayan area. Biogenic emissions influence the level and distribution of PM_2.5 and organic aerosols (OAs) during the monsoon season, and they contribute significantly to SOA, notably over the lower IGP (Mogno et al., 2021), and they reported that the OA contribution to PM_2.5 is substantial in all four seasons (17%–30%). Previous studies have also shown that OAs originating from biogenic, anthropogenic, and pyrogenic emissions make a substantial contribution to PM_2.5 (20%–35%) across the IGP region (Ram et al., 2008; Alam et al., 2014; Rajput et al., 2014; Behera and Sharma, 2015; Sharma et al., 2016).

However, information on the characterization of aerosol types over Bangladesh that reveals the long-term climatic changes in the study areas is limited. Only few studies on the spatiotemporal variations of aerosols over Bangladesh have been conducted (Ali et al., 2019; Islam et al., 2019; Zaman et al., 2021a). The categorization of aerosol types has significant implications for climate change modeling and assessing aerosol radiative impacts (Satheesh and Krishna Moorthy, 2005). For this reason, we have focused on the characterization of absorbing aerosol types over Bangladesh. We have conducted aerosol-type characterization based on different aerosol optical properties in Bangladesh using ground-based (aerosol robotic network [AERONET]) and satellite-based (ozone monitoring instrument [OMI]) observations. Columnar aerosol data from 2012 to 2019 at available ground-based AERONET sites within Bangladesh, that is, Dhaka and Bhola, were assessed. The objectives of this work are: (a) assessment of the long-term spatial distribution of aerosols over Bangladesh, (b) identification of the aerosol types based on correlations between extinction angstrom exponent (EAE) and absorption AE (AAE), EAE and single scattering albedo (SSA), and EAE and real refractive index (RRI), (c) characterizing the absorbing aerosol types using the relationships between fine mode fraction (FMF) and AE, FMF and AAE, and FMF and SSA, and (d) validation of the detected absorbing aerosols using the cloud–aerosol lidar and infrared pathfinder satellite observation (CALIPSO) satellite sensor.

Bangladesh is a Southeast Asian country bordered by India to the west, north, and east, Myanmar to the southeast, and the Bay of Bengal (BoB) to the south. It boasts a population of approximately 160 million people in an area of approximately 57,320 square miles. Thus, it is one of the world’s most densely inhabited countries. Bangladesh is a hot and humid country with an annual rainfall of approximately 2,300 mm (Salam et al., 2003b). Meteorologically, Bangladesh is divided into four seasons based on 3-months’ mean temperature and rainfall: premonsoon (March–May), monsoon (June–August), postmonsoon (September–November), and winter (December–February; Saslam et al., 2012; Zaman et al., 2021b). This study was carried out in 2 AERONET sites of Bangladesh, namely Dhaka University (23.728°N, 90.398°E), Dhaka and Bangladesh Climate Observatory Bhola (22.227°N, 90.756°E), Bhola (Figure 1). Dhaka is Bangladesh’s capital and the country’s commercial hub. Bhola is a rural island district in Bangladesh’s far south, close to the BoB.

Since 2012, our group at Dhaka University has been running two National Aeronautics and Space Administration (NASA) AERONET stations in Bangladesh to measure aerosol optical depth. Cimel sunphotometers measure direct sun irradiation at 8 distinct wavelengths (340–1,020 nm) and contribute to the NASA surface-based aerosol network with over 700 AERONET stations worldwide (Holben et al., 1998). The uncertainty in computed AOD is approximately 0.01–0.02. It is less than ±5% for sky irradiance observation, which is dependent on aerosol quantities and types (Dubovik et al., 2000). In this study, the daily data of Level 2.0 (quality assured) direct sun products (FMF_{500 nm} and AE_{440-870 nm}) and inversion products (aerosol absorption optical depth [AAOD_{440 nm}], AAE_{440–870 nm}, EAE_{440–870 nm}, RRI_{440 nm}, and SSA_{440 nm}) with a high temporal resolution of ˜15 min were used to classify aerosols for the duration of 2012–2019 for Dhaka and 2013–2019 for Bhola (https://aeronet.gsfc.nasa.gov/). This time frame was chosen based on the availability of AERONET data (Table S1). Due to excessive rain during the monsoon seasons, as well as instrumentation error, many data points were lost. Given the scarcity of data sets in this region, we feel these data sets will be a valuable asset to the scientific community.

OMI, carried by the Aura satellite, was launched in 2004. The OMI project is led by the Netherlands Agency for Aerospace programs and is fashioned to govern air quality, ozone, and earth’s climate (Krueger, 1989; Bibi et al., 2015). Nadir viewing measures the sunlight scattering of atmospheric aerosols with high spectral (270–500 nm) and spatial resolution (13–24 km; Levelt et al., 2006). Aura satellite has local crossing time over Bangladesh roughly around 01:30 (±15 min) and 13:30 (±15 min; https://evdc.esa.int/orbit/). The OMAERUV (OMI/Aura Near UV Aerosol Optical Depth) retrieval algorithm can be used to measure absorbing aerosols in the near-UV spectral region. OMI level 3 AAOD (500 nm; http://giovanni.gsfc.nasa.gov/ assessed on April 8, 2021) of 1° spatial resolution data was used for the analysis.

CALIPSO is a set of Earth observation instruments launched aboard the CloudSat satellite in 2006. It provides insight on aerosol vertical profiles and aerosol distribution; it is used to study the roles of aerosols and clouds in the Earth’s weather and climate based on the cloud–aerosol lidar with orthogonal polarization sensor (Winker et al., 2003). Although the temporal resolution is only 16 days and incongruous for continuous monitoring, it provides unique information with which to interpret atmospheric aerosols (Su et al., 2020). At present, it is the only satellite in orbit that provides information on vertical profiles of aerosols as well as a 3-D distribution of aerosol characteristics with some uncertainties (Bibi et al., 2017). Therefore, the Level 2 CALIPSO version 3.30 aerosol-type profiles were implemented for the validation of absorbing aerosols in this region. CALIPSO images were directly downloaded from https://www-calipso.larc.nasa.gov/products/lidar/browse_images/std_v4_index.php, accessed on May 17, 2021.

AAOD is one of the most important parameters for the appraisement of global warming due to the light-absorbing aerosols (dust, BC, or BrC; Tesche et al., 2019). AAOD is the aggregate of most BC and dust (Bibi et al., 2017) and can be defined by the following relationship:

For comparison purposes, AERONET AAOD₄₄₀ was converted into AAOD₅₀₀ by the following equation:

where the AAE is calculated at 440–870 nm using the following equation (Utry et al., 2014; Zhang et al., 2021):

Previous research has discovered narrow fluctuations in the autocorrelation and coefficient of variation of the aerosol load at a time difference of approximately 1–3 h within a region of 10–50 km (Mélin et al., 2007). To obtain collocated OMI-AAOD with AERONET for validation purposes, the OMI AAOD readings were averaged within a spatial window of 1 × 1 pixel centered over AERONET sites, and because AERONET observations are available every 10–15 min, a time window of ±30 min around the satellite pass is considered for the OMI assessments.

Identification of aerosol sources is imperative to understand the role of anthropogenic aerosols that have varying chemical, optical, and physical properties (Dubovik et al., 2002). The AAE at 440–870 nm and SSA at 440 nm are used by the ground-based AERONET to attribute absorptivity to aerosol particles. Aerosol absorbing qualities, such as the ultraviolet aerosol index (UVAI) and AAOD, can also be obtained via satellite remote observations, such as the OMI (Adesina et al., 2016). As a result, the stratification of aerosols around the globe using AERONET and satellite data is a well-known process. For the classifications of global aerosols, Dubovik et al. (2002) established a link between different optical properties of aerosol retrieved from AERONET. Numerous studies have been conducted to classify aerosols using various cluster approaches. Hamill et al. (2016) used various parameters, such as SSA, EAE, AAE, and RRI, to identify aerosol types at 190 AERONET sites worldwide. Eck et al. (2010) used FMF and AE to classify aerosols in Beijing, China, Kanpur, India, and Ilorin, Nigeria. In Karachi, Pakistan, Bibi et al. (2017) used the relationships between AE and UVAI and FMF and UVAI. Rupakheti et al. (2019) classified aerosol types in Lumbini and Kathmandu Valley, Nepal, using the relationship among EAE, AAE, SSA, RRI, FMF, and AE. Ali et al. (2020) used FMF versus AE, FMF versus AAE, FMF versus SSA, AE versus UVAI, and FMF versus UVAI relationships to separate aerosol types at the Solar Village and KAUST campus in Saudi Arabia.

In this study, we used various threshold values (Table 1) to classify aerosols based on the parameters, such as EAE, AAE, SSA, and RRI. For better assessment of the aerosol types, these specific threshold values are chosen to classify the aerosols depending on the chemical composition (Giles et al., 2011; Mishra and Shibata, 2012; Kumar et al., 2015). Dominant aerosol types are determined by correlation between absorption and size parameters (Giles et al., 2011). AAE is a function of aerosol composition (Mishra and Shibata, 2012) and EAE is a particle size indicator (Russell et al., 2010). Therefore, EAE versus AAE correlation is a key to distinguish aerosol types. However, they cannot separate biomass burning from urban/industrial; hence, it is important to correlate EAE with parameters, such as SSA and RRI. SSA is the ratio of scattering efficiency to extinction and can be applied to differentiate between the absorbing and nonabsorbing aerosols, while RRI is useful in the differentiation of scattering behavior of aerosol (Sinyuk et al., 2003). Using these correlations, three types of major aerosols were identified, namely (a) biomass burning, (b) urban/industrial, and (c) dust. These threshold values are highly robust and used in the earlier research conducted in cities across the IGP region (Bibi et al., 2016a; Bibi et al., 2017; Rupakheti et al., 2019).

In addition, we explored the nature of absorbing aerosols and used the relationships of FMF versus AE (Logothetis et al., 2020), FMF versus AAE (Bibi et al., 2017), and FMF versus SSA (Lee et al., 2010; Bibi et al., 2017) to classify the absorbing aerosols into the categories of (1) BC, (2) mixed BC and dust, and (3) dust (Table 2). The data sets that did not fall within the thresholds were demonstrated as other aerosol types. Rupakheti et al. (2019) used a similar strategy of 2 different characterization systems in a study conducted in 2 different locations (Lumbini and Kathmandu) in Nepal.

In order to assess the emissions sources of BC over Bangladesh, EDGAR v5.0 (https://edgar.jrc.ec.europa.eu/) data sets were used. The fire points in this region were assessed using MODIS (Moderate Resolution Imaging Spectroradiometer) Terra and Aqua satellites from the Fire Information for Resource Mapping System of NASA (https://firms.modaps.eosdis.nasa.gov/map/ accessed on April 8, 2021). Finally, the classified aerosols were validated with CALIPSO data sets.

Confidence intervals (95%) for optical parameter means were estimated using Bootstrap methods (Kushary et al., 2000) as implemented in the Comprehensive R Archive Network (CRAN) R package DescTools MeanCI function (using the boot method and R = 999).

To test for overall statistical differences in the optical parameters as grouped by season, we applied the Kruskal–Wallis test and subsequently the pairwise Willcox test using CRAN R software version 4.1.2. The pairwise Willcox test was applied with corrections for multiple testing using the method of Benjamini and Hochberg (1995) as implemented in CRAN R software version 4.1.2.

In this study, the mean seasonal and annual spatial distribution of satellite-based (OMI) AAOD were calculated from the daily data for the time period of 2012–2019 and compared with the ground-based AERONET–AAOD. The mean (n = 608) AERONET–AAOD values were greater in Dhaka (0.110 ± 0.002) [0.106, 0.114] than in the coastal island of Bhola (n = 546; 0.075 ± 0.001) [0.073, 0.078]. However, the mean (n = 639 and 855 for Dhaka and Bhola, respectively) OMI–AAOD values were similar for both sites (0.024 ± 0.001 [0.023, 0.025] for Dhaka and 0.023 ± 0.001 [0.023, 0.024] for Bhola). Seasonal distribution of AERONET–AAOD showed the highest AAOD values in the winter season for Dhaka (0.121 ± 0.005) and Bhola (0.091 ± 0.003) followed by the postmonsoon season (0.110 ± 0.005 and 0.070 ± 0.003) and the premonsoon season (0.091 ± 0.004 and 0.062 ± 0.003). The monsoon season showed the lowest AAOD values (0.060 ± 0.003 and 0.041 ± 0.002). However, OMI–AAOD revealed a slightly different sequence with an order of winter > premonsoon > postmonsoon > monsoon (Table 2) and lower overall values (0.024 ± 0.001 and 0.023 ± 0.001) compared to the AERONET–AAOD (0.110 ± 0.002 and 0.075 ± 0.001) with a distinct annual cycle (Figure 2).

Ali et al. (2020) reported that OMI–AAOD followed almost similar annual pattern that of AERONET–AAOD over Saudi Arabia. High AAOD (>0.07) values in the postmonsoon across western IGP were reported by Shaik et al. (2019). Vadrevu et al. (2015) also reported higher AAOD values (0.075) over Punjab. Again, very high AAOD values over northwestern IGP in the postmonsoon season and over central and eastern part of IGP in the winter season were reported by Mhawish et al. (2021). On the other hand, lowest AAOD values in the months of June, July, and August were reported by Kang et al. (2017) over East Asia.

The spatial distribution of OMI–AAOD is depicted in Figure 3. Lower aerosol loadings in the northeast (Sylhet and Mymensingh) and higher aerosol density in the capital (Dhaka) and southwestern parts (Khulna) are portrayed in the graphic. Similar result was found by Zaman et al. (2021a). They also reported lower aerosol loadings in Sylhet and higher in the northwest (Rajshahi) and southwest (Khulna) over Bangladesh. Ali et al. (2019) also reported higher AOD (>0.7) in the northwest (Rajshahi) and southwest (Khulna) regions of Bangladesh while Sylhet and Chattagram have lower AODs (<0.4).

The highest aerosol loadings were observed over the BoB (Figure 3). This can be attributed to significant atmospheric convergence, which causes increased aerosol accumulation, marine generation of OAs from plankton dynamics, particularly during the winter season (Nagamani et al., 2011), and aerosol advection from Indian land mass (Prijith et al., 2018). Kaskaoutis et al. (2011) also reported that BoB had substantially higher aerosol loadings, owing mostly to anthropogenic emission and a dispersion mechanism driven by meteorology. According to Satheesh et al. (2010), due to the landlocked character of BoB and its closeness to highly polluted regions, industrial pollution from varied locations such as northern India and Bangladesh modulates the aerosols.

The scatter plot of AERONET AAOD and OMI AAOD shown in Figure S1 indicates that OMI is a reliable tool for the retrieval of aerosol data in this region given the moderately positive correlation between the two approaches (R² = .64 for Dhaka and .63 for Bhola). Bibi et al. (2017) reported a higher positive correlation (R² = .78) between OMI–AAOD and AERONET–AAOD in a study conducted in Karachi, Pakistan.

AAE is an indicator of absorption contrast with the wavelength, which indicates the size of the absorbing aerosols. AAE values of <2 and >2 indicate the fine mode absorbing aerosols and the coarse mode absorbing aerosols, respectively (Russell et al., 2010; Bibi et al., 2015). The annual mean values of AAE with 95% confidence interval over Dhaka and Bhola were 1.311 ± 0.015 [1.281, 1.340] and 1.257 ± 0.015 [1.228, 1.285], respectively (Tables 2 and S2), suggesting that the mode of absorbing aerosols in Bangladesh is fine. Higher AAE values were found in premonsoon and postmonsoon seasons in Dhaka, whereas in Bhola, premonsoon and winter seasons showed higher AAE values. Based on Kruskal–Wallis test (Table S3), we found overall statistical differences for the optical parameters grouped by seasons (P < 0.05). Pairwise Willcox test P values for individual seasons are shown in Table S4. For Dhaka AAE, a significant difference was only found between winter and premonsoon and premonsoon and postmonsoon. In contrast, differences in AAE were found at Bhola for premonsoon and monsoon, premonsoon and postmonsoon, winter and premonsoon, and winter and premonsoon (Table S4).

EAE is an important particle size distribution indicator (Russell et al., 2010). Although slightly higher mean (with 95% confidence interval) of EAE was found in Bhola (1.242 ± 0.007) [1.228, 1.255] than Dhaka (1.197 ± 0.007) [1.182, 1.211], seasonal distribution showed similar pattern in both sites (higher in postmonsoon, winter seasons and lower in premonsoon, winter season; Tables 2 and S2). For Dhaka EAE, statistical differences were found for postmonsoon and monsoon, premonsoon and postmonsoon, and winter and postmonsoon. The statistical differences for Bhola EAE were postmonsoon and monsoon, premonsoon and postmonsoon, winter and monsoon, winter and premonsoon, and winter and postmonsoon (Table S4).

SSA explains the absorbing nature of aerosols. Based on different SSA values, aerosols can be classified as nonabsorbing (>0.95), weakly absorbing (0.90–0.95), moderately absorbing (0.85–0.90), and highly absorbing (<0.85). An SSA value 0 indicates completely absorbing and 1 indicates completely scattering (Lee et al., 2010; Russell et al., 2010; Gyawali et al., 2012). The annual mean values (with 95% confidence interval) of SSA in Dhaka (0.891 ± 0.001) [0.889, 0.894] and Bhola (0.912 ± 0.001) [0.909, 0.915] showed that aerosols in Bangladesh are of weakly absorbing type (Tables 2 and S2). Statistical differences (P < 0.05) for Dhaka SSA were found for postmonsoon and monsoon, premonsoon and monsoon, premonsoon and postmonsoon, winter and monsoon, winter and postmonsoon, and winter and premonsoon. However, at Bhola, postmonsoon and monsoon, premonsoon and postmonsoon, winter and monsoon, and winter and premonsoon were found statistically different by pairwise Wilcox test (Table S4).

The RRI is a critical optical feature for obtaining reliable outcomes in aerosol identification (Bibi et al., 2016a). RRI is dependent on the aerosol chemical composition (Dubovik et al., 2002) and imparts crucial information on the nature of aerosols (Bibi et al., 2016b). RRI showed same mean (with 95% confidence interval) values (Dhaka: 1.459 ± 0.002 [1.455, 1.463] and Bhola: 1.465 ± 0.002 [1.461, 1.468]). Almost similar values were found for all the seasons in both sites. For Dhaka RRI, statistical differences were found for postmonsoon and monsoon, premonsoon and postmonsoon, winter and monsoon, and winter and premonsoon. For Bhola, statistical differences were found for premonsoon and postmonsoon and winter and postmonsoon (Table S4).

FMF imparts quantitative information about the coarse mode and fine mode aerosols, values ranging from 0 (coarse) to 1 (fine). FMF values for coarse mode, mixed type, and fine mode aerosols were reported as <0.40, 0.40–0.60, and >0.60, respectively (Lee et al., 2010; Logothetis et al., 2020). The mean FMF values for the overall period were 0.858 ± 0.005 [0.847, 0.869] (Dhaka) and 0.855 ± 0.006 [0.843, 0.866] (Bhola), signifying the existence of fine mode aerosols in Bangladesh (Tables 2 and S2). In winter seasons, higher FMF values (>0.90) were detected, confirming the presence of fine mode aerosols in Bangladesh. Similar statistical differences were found at both Dhaka and Bhola; postmonsoon and monsoon, premonsoon and postmonsoon, winter and monsoon, and winter and premonsoon (Table S4).

Angstrom exponent (AE) is a good index of particle size but sometimes causes abstruse characterization of the fine and coarse modes of aerosols (Wu et al., 2015). Lower values of AE (<1) indicate the ascendancy of coarse mode aerosols, and higher values (>1) indicate the ascendancy of fine mode aerosols, such as BC, OC, and sulfates discharged from anthropogenic activities (Eck et al., 2003). The annual mean values of AE in both sites were higher in Bhola (1.270 ± 0.008) [1.253, 1.286] than Dhaka (1.195 ± 0.008) [1.178, 1.211] (Tables 2 and S2). Both values greater than 1 showed the dominance of the fine mode of aerosols in Bangladesh. Statistical differences at Dhaka AE were found for postmonsoon and monsoon, premonsoon and monsoon, and winter and monsoon. However, the statistical differences at Bhola AE were found for postmonsoon and monsoon, premonsoon and monsoon, winter and monsoon, premonsoon and postmonsoon, winter and postmonsoon, and winter and premonsoon (Table S4).

EAE is an indicator of particle size distribution and along with AAE; it can be used to signify the composition of aerosols. Figure 4a and b depicts the seasonal scatter plots of AAE_440–870 and EAE_440–870 in Dhaka and Bhola. The cluster analysis demonstrates three types of aerosols, namely biomass burning (high AAE and high EAE), urban/industrial (low AAE and high EAE), and dust (high AAE and low EAE). The urban/industrial and biomass-burning aerosols overlap (Giles et al., 2011; Mishra and Shibata, 2012). Figure 4a and b depicts that a large portion of the data points lies within the EAE (0.8–1.6) and AAE (0.6–2.3) region, which suggests the presence of biomass-burning and urban/industrial aerosols during all seasons in both locations. However, dust-type aerosols were found only in Dhaka in the premonsoon and winter seasons. Rupakheti et al. (2019) also found the dominance of biomass-burning and urban/industrial aerosols over two locations in Nepal. Bibi et al. (2016a) reported the clear dominance of dust aerosols during summer and premonsoon and biomass-burning aerosols in the winter and postmonsoon season in a study conducted over four AERONET sites (Lahore, Karachi, Kanpur, and Jaipur) in the IGP region.

We used SSA at 440 nm, which is the same as in the previous studies (Levy et al., 2007; Bibi et al., 2016a; Rupakheti et al., 2019). Although SSA alone cannot differentiate aerosol types precisely, SSA as a function of EAE can provide a better categorization of aerosol types (Giles et al., 2011).

Figure 4c and d depicts the seasonal relationship between EAE and SSA over the two locations. The dominance of biomass-burning aerosols through all seasons is shown in the figure. However, a small presence of dust in the premonsoon and monsoon season in Dhaka was also found. No dust data set was found in Bhola. The highest number of data points in the winter season lies in the biomass-burning aerosols for both locations. A study in the nearby IGP site Kanpur also showcased the predominance of biomass-burning aerosols in the winter season (Bibi et al., 2016a).

RRI is dependent on the size of aerosol, and its values may vary due to the different chemical compositions of aerosols (Dubovik et al., 2002). RRI, as a function of EAE, is another important mode of classifying aerosols. The relationship between EAE and RRI over Dhaka and Bhola is depicted in Figure 4e and f. The clear dominance of biomass-burning aerosol is shown in Figure 4. The considerably greater values in Dhaka’s urban/industrial region may also be noted, as indicated in the relationship between EAE and AAE relationship. In Dhaka, the quantity of data point in the dust region for premonsoon and winter was small, but only the premonsoon season showed a very small amount of dust in Bhola. A study conducted in Nepal also showed similar results (Rupakheti et al., 2019). Raut and Chazette (2007) also found comparable results for soot particles over Paris.

Biomass burning and urban/industrial have been recognized as the key contributors to excessive aerosol loads over Bangladesh based on the relationship between EAE versus AAE, EAE versus SSA, and EAE versus RRI. To apprehend the temporal profile emissions of BC and also understand the emission inventory of PM_2.5 in Bangladesh, we have used EDGAR (emission database for global atmospheric research) version 5.0 data (https://edgar.jrc.ec.europa.eu/; Crippa et al., 2020). EDGAR database identified biomass burning and residential and other sources to be the main sources of BC and PM_2.5 in Bangladesh, as shown in Figure S2. Begum et al. (2012), Zaman et al. (2021a), and Salam et al. (2021) also identified biomass burning as a substantial source of BC.

Figure 5 shows the fire points around Bangladesh during the premonsoon and winter seasons in 2018–2019. The figure demonstrates copious fire points around Bangladesh, revealing the dominance of biomass-burning aerosols in the region. In addition, these fire events often produce tremendous amount of aerosols, which transport through Bangladesh increasing the aerosol loading particularly in the premonsoon and winter season (Ommi et al., 2017). Rupakheti et al. (2017) also reported the dominance of biomass-burning aerosols in this region in the premonsoon period. Tariq et al. (2016) also found the dominance of biomass burning during the premonsoon and winter season over Lahore, Pakistan.

We categorized our aerosols as biomass burning, urban/industrial, and dust using aerosol optical characteristics in the previous section. Several earlier studies in the IGP region used AERONET inversion data to classify absorbing aerosols into three categories (BC, mixed, and dust) using FMF versus AE, FMF versus AAE, and FMF versus SSA correlations (Eck et al., 2010; Giles et al., 2011; Giles et al., 2012; Kedia et al., 2014; Bibi et al., 2017; Rupakheti et al., 2019). Islam et al. (2019) evaluated the presence of absorbing aerosols over Bangladesh using OMI, the modern-era retrospective analysis for research and application-version 2 (MERRA-2), and the emission database for global atmospheric research (EDGAR) system data. Zaman et al. (2021a) categorized aerosols over Bangladesh into biomass burning, anthropogenic, and mixed types using AOD versus AE correlations, but this correlation is unable to distinguish anthropogenic aerosols into absorbing and nonabsorbing types (Lee et al., 2010), necessitating the investigation of other parameters. However, in Bangladesh, research that uses these connections to classify aerosols is sparse. The absorbing aerosol was categorized for the first time as BC, mixed (BC and dust), and dust in this segment using the threshold values in Table 3.

The scatter relationship of FMF versus AE is shown in Figure 6a and b, and it depicts greater dispersal of the data points in AE > 1, which may be due to the irregularity in the radius of fine aerosols, the aging of aerosols, and hygroscopicity (Eck et al., 2003). BC aerosols were found dominant in both sites (45.9% in Dhaka and 55.2% in Bhola). However, a bulk portion of aerosols (43.8% and 42.5%) did not fall within the threshold values in Table 3; these aerosols were classified as other types. Other types of aerosols may not be primarily from a single source but rather from a combination of sources, such as marine, urban, and desert, as described in Delhi, a nearby IGP site (Sharma et al., 2014; Tiwari et al., 2016). Shin et al. (2019) found the presence of pure dust, dust-dominated mixtures, pollution-dominated mixtures, and pollution types aerosols across East Asia. Long-range transport of these aerosols may lead to the formation the other types of aerosols in Bangladesh.

Mixed (BC and dust) type and dust aerosols were found almost in similar proportion (2.3% and 2.6%, respectively) in Dhaka. However, in Bhola, dust aerosols were found in almost negligible amounts (0.6%) only in the premonsoon season. The entire data sets of postmonsoon and premonsoon fall into the category of BC in Bhola.

The relationship between FMF and AAE for the characterization of different absorbing aerosols was used in several studies in the IGP region and worldwide (Kedia et al., 2014; Bibi et al., 2017; Rupakheti et al., 2019; Ali et al., 2020). Figure 6c and d shows the seasonal scatter plot of FMF versus AAE observed over Dhaka and Bhola. For the classification of BC and mixed (BC and dust) aerosols, the same AAE threshold values were used although the FMF values were different. Figure 5c and d clearly infers the difference between Dhaka and Bhola. A higher (76.9%) amount of BC aerosols was found in Dhaka than in Bhola (47.4%). Almost the entire postmonsoon data points lied within the BC region in Dhaka. A minimal amount of dust was found in Bhola (0.74%) in the premonsoon season similar to that in Section 3.3.1. Mixed (BC and dust) type aerosols were found in a comparable proportion in Dhaka and Bhola. Similar results were found in Rupakheti et al. (2019) in Nepal and Kedia et al. (2014) in India. However, Ali et al. (2020) reported a higher (58.1%) amount of dust aerosols in a study conducted over two sites in Saudi Arabia. The presence of dust aerosols in the premonsoon season is supported by the other studies conducted in the IGP region. Dumka et al. (2014) described the presence of most dust aerosols in the premonsoon season from Kanpur to Pantnagar in the Himalayan foothills. Kedia et al. (2014) reported the enriched existence of dust-type absorbing aerosols in the premonsoon and monsoon season in Gandhi College and Kanpur in India.

FMF versus SSA is a well-known tool that provides a distinctive characterization of absorbing aerosols; it has been carried out by several studies (Giles et al., 2012; Bibi et al., 2017; Rupakheti et al., 2019; Ali et al., 2020). The scatter plot of FMF versus SSA in all seasons in two locations to classify the absorbing aerosols into BC, mixed (BC and dust), and dust is shown in Figure 6e and f. The threshold values were analogous to the previous studies (Bibi et al., 2017; Ali et al., 2020), as shown in Table 3. The results obtained from Figure 6e and f confirmed the abovementioned results, and a more pronounced percentage of BC was obtained. The percentage of BC in Dhaka was 89.1% and that in Bhola was 83.1%. These extremely high percentages confirmed the dominance of BC-type aerosols in Bangladesh. The entire postmonsoon data points were within the BC region in Dhaka and Bhola. The percentages of mixed-type aerosols were comparable, similar to Sections 3.3.1 and 3.3.2. A very minute (2.14% and 0.74%) percentage of dust aerosols was found in the premonsoon season. The percentage of other types of aerosols was greater in Bhola (12.7%) than Dhaka (4.1%), consisting of mainly premonsoon data points. Rupakheti et al. (2019) found similar results; high percentage of BC aerosols in Lumbini and Kathmandu, Nepal. The presence of dust aerosols in the premonsoon and monsoon was also reported in a study conducted in Karachi, Pakistan (Bibi et al., 2017). However, Ali et al. (2020) found the dominance of dust aerosols (84.0% and 50.5%) in a study conducted at two sites in Saudi Arabia.

The CALIPSO aerosol lidar ratio is a quantitative measure of aerosol extinction retrieval that can identify aerosol subtypes using a multiyear AERONET data set cluster analysis (Omar et al., 2009). Therefore, the AERONET-classified particles were further validated using the CALIPSO daytime aerosol subtype profiles for selected days in various seasons (Table 4 and Figure S3; https://www-calipso.larc.nasa.gov/). The days were selected based on the availability of AERONET data for both study sites (Dhaka and Bhola). CALIPSO aerosol subtypes are classified as clean marine, dust, polluted dust, polluted continental, clean continental, and smoke (version 3.30). CALIPSO’s vertical information was in line with the AERONET column information. Similar approaches were also applied in the previous studies, given that no other data is available for validation (Kumar et al., 2012; Kumar et al., 2015; Lüthi et al., 2015; Bibi et al., 2016a; Bibi et al., 2017; Rupakheti et al., 2019; Ali et al., 2020).

Figure S3 and Table 4 depict the presence of smoke (carbonaceous aerosols), polluted continental, and polluted dust (mixture of dust and smoke) aerosols up to approximately 5 km over the surface of Dhaka and Bhola in all seasons. A recent study also found smoke, polluted dust, and polluted continental and clean marine aerosols over major cities of Bangladesh (Qiu et al., 2021). Rupakheti et al. (2019) also reported similar features over two sites in Nepal (Lumbini and Kathmandu). Yu et al. (2016) noted the major involvement of smoke, dust, and polluted dust in a haze event over Beijing. Kumar et al. (2012) found that dust and polluted dust were more prominent during monsoon and premonsoon in India. The presence of dust in the premonsoon period supported the study’s findings. While smoke and polluted dust were the dominant aerosols in the postmonsoon season, some proportion of dust aerosols can also be found. During winter, dominant aerosol types were smoke and polluted dust, with a small input of polluted continental and dust.

We applied 2 separate characterization methods to fully comprehend the aerosol climatology in Bangladesh. A pairwise relationship between the aerosol parameters (e.g., EAE vs. AAE, EAE vs. SSA, and EAE vs. RRI) was established. Three types of aerosols have been observed, namely biomass burning, urban/industrial, and dust. The results of the pairwise associations revealed that biomass-burning and urban/industrial aerosols dominated in both sites. However, the association of EAE versus AAE implied that Dhaka had a greater number of urban/industrial data points than Bhola. Various industry operations enhance the amount of urban/industrial aerosols in Dhaka because the area is urban. In Dhaka, dust aerosols were observed in minor amounts predominantly during the premonsoon and monsoon seasons because the dust activity is at its peak from April to July (Mamun, 2014). The bulk of postmonsoon records was found in areas of biomass burning in Dhaka and Bhola. Figure 4 depicts the impact of crop field burning in this season, as shown in the local and regional perspectives (Tiwari et al., 2015; Ommi et al., 2017). Several studies of aerosols over Bangladesh discovered the existence of biomass-burning aerosols. According to a recent study conducted in Dhaka, fossil fuel combustion provided 21.6% and biomass combustion contributed 40.2% of the overall average PM_2.5 from 2013 to 2017 (Rahman et al., 2020). A high amount of biomass-burning aerosol was also found in several source apportionment investigations undertaken in Bangladesh (Begum et al., 2009; Chowdhury et al., 2012; Begum and Hopke, 2018).

Furthermore, different aerosol optical characteristics were used to assess the absorbing aerosols in Bangladesh. Three different forms of aerosol were identified as BC, mixed (BC and dust), and dust using the correlations among FMF, AE, AAE, and SSA. In all connections, very few data sets were identified during the monsoon season due to the lack of clear sky conditions, which are necessary for the reclamation of columnar aerosols (Gobbi et al., 2010). The investigation of three relationships (FMF vs. AE, FMF vs. AAE, and FMF vs. SSA) clearly shows that the BC type was the most prevalent aerosol type in both sites. The relationships also presumed that Dhaka had higher percentage of BC than Bhola. Dhaka is an urban city and is one of the world’s most densely populated. The increased number of anthropogenic and industrial activities may be the cause of the higher concentration of BC aerosols in Dhaka. Previous ground-based measurement in Dhaka have reported a BC concentration value of 26.3 µg m^–3 (Begum et al., 2010). A likely source for elevated airborne BC concentrations are emissions from brick kilns, that is, Haque et al. (2018) have reported average concentrations of 16.6 mg m^–3 in brick kiln emissions around Dhaka.

A long-term data analysis across multiple sites is useful to understand the trends and distribution of aerosol types in Southeast Asia. For the first time, significant insights into the characterization of aerosol types in Bangladesh are provided.

Given the enhanced knowledge of the detrimental impacts of aerosols, the aerosol relationships detailed here will have a substantial impact on aerosol and climate modeling, notably air quality modeling and global climate forcing of aerosols for the Southeast Asian region. This study does, however, address a few shortcomings. This study was confined to 2 locations in the country. The satellite OMI used in this study has some uncertainties in AOD retrieval including radiometric calibration, cloud contamination, assumption of aerosol characteristics, and surface effect correction (Li et al., 2009), which may have affected the long-term trend of the AOD over Bangladesh. Besides, many AERONET data points were lost due to extensive cloud cover in the monsoon seasons. Soluble aerosols could also be lost due to excessive rains and instrumental error. We recommend long-term research in various regions of the country focusing on other anthropogenic sources of aerosols, transport (horizontal and vertical), and climate model simulation.

Attempts were made to investigate the aerosol types and characterize the absorbing aerosols based on AERONET data from 2 locations in Bangladesh. One site is located at the center of Bangladesh, the capital city Dhaka, and another at the southern part of the country, a coastal Island of the BoB, Bhola. The association between different aerosol optical properties, such as EAE and AAE, EAE and SSA, and EAE and RRI, was investigated. We observed the dominance of biomass-burning and urban/industrial aerosols in all 4 seasons over urban and coastal sites based on cluster analyses. In addition, urban/industrial aerosols were more prevalent in Dhaka than in Bhola Island. The presence of a substantial amount of data in the overlapping area between biomass burning and urban/industrial also indicated the existence of the mixed-type aerosols in this region. The relationship of the light absorption parameters revealed the strong dominance of BC aerosols in all seasons over Bangladesh. Mixed (BC and dust) and other types of aerosols also contributed relatively to lower proportion. A negligible contribution from dust aerosols was observed mostly in the premonsoon season. The seasonal distributions of the aerosols were further supported by the CALIPSO aerosol subtype profiles.

The data that this study builds on are provided as Supplemental Materials to this article.

The supplemental files for this article can be found as follows:

Tables S1–S4. Figures S1–S3. Docx

Data file. xlsx

The authors acknowledge the cordial support of National Aeronautics and Space Administration (NASA), USA, for providing two Cimel sunphotometers to operate 2 aerosol robotic network stations in Bangladesh. They are also grateful to the Goddard Space Flight Center and Fire Information Resource Management System of NASA for providing the ozone monitoring instrument and the fire point data sets. They extend our gratitude toward the scientific teams of cloud–aerosol lidar and infrared pathfinder satellite observation for supplying the aerosol subtype profiles and EDGAR team for providing the emission data profiles. They are also grateful to Md. Nazmul Haque of the Islamic University of Technology, Bangladesh, for his suggestions.

This research work done with the internal facilities and support of the Department of Chemistry, University of Dhaka, and also external support from National Aeronautics and Space Administration, USA, for the calibration of Cimel sunphotometer through aerosol robotic network project.

The authors declare no conflict of interest. Md. Firoz Khan is an Associate Editor at Elementa. He was not involved in the review process of this article.

Contributed to the conception and design: SUZ, MRSP, AS, MFK.

Contributed to the acquisition of data: SUZ, MRSP, RIR.

Contributed to the analysis and interpretation of data: SUZ, MRSP, RIR, AS, MFK, RE.

Drafted and/or revised the article: SUZ, AS, FJ, MSI, MFK.

All authors approved the final version of this article for submission.

How to cite this article: Zaman, SU, Pavel, MRS, Rani, RI, Jeba, F, Islam, MS, Khan, MF, Edwards, R, Salam, A. 2022. Aerosol climatology characterization over Bangladesh using ground-based and remotely sensed satellite measurements. Elementa: Science of the Anthropocene 10(1). DOI: https://doi.org/10.1525/elementa.2021.000063

Domain Editor-in-Chief: Detlev Helmig, Boulder AIR LLC, Boulder, CO, USA

Associate Editor: Paul Palmer, School of GeoSciences, The University of Edinburgh, Edinburgh, UK

Knowledge Domain: Atmospheric Science