Spectral Properties of Black Carbon Produced during Biomass Burning

Spectral properties of Black Carbon (BC) produced during the burning of different biofuels have been studied using a 7 channel Aethalometer (Magee Scientific, USA). Different biofuels used for household purpose in the rural regions were selected after a survey conducted in Gadanki, A.P. – a typical village in South India. Biofuels were burnt in a custom made burning chamber and BC produced has been allowed to mix with ambient air and measured using Aethalometer kept 4 meters away from the burning chamber. The observed absorption coefficient spectra have been characterized by fitting to a power-law equation. The power co-efficient (α) values found to be ranging from 1.2–2.1 for different biofuels. It has been observed that the α values are related to the speed of burning; fast burning biofuels shown low values, whereas slow burning biofuels have shown higher values.


INTRODUCTION
The earth's atmosphere has a large number of particles suspended in it.They affect the earth's radiation budget by scattering and absorbing solar radiation (Lewis et al., 2008).Black carbon particles are a distinct type of carbonaceous particles known for their highly absorbing nature.Unlike sulfate and sea-salt aerosols, black carbon particles produce positive radiative forcing due to their absorbing nature.Jacobson (2001) estimated that radiative forcing by black carbon particles is the second highest -the first being CO 2 .Radiative forcing by black carbon particles is estimated to be between +0.17 to +1.48 W m -2 (Bond et al., 2013).However, unlike CO 2 , black carbon particles have short residence time (of the order of the week) in the atmosphere.Hence, several researchers have advocated policy targeting reduction of black carbon particles for getting immediate benefit of slowing global warming (Cross and Pierson, 2013).
Black carbon particles are primarily formed in flames and can be called as a combustion product mainly from vehicle emissions and biomass burning (Andreae and Crutzen, 1997; † Now at Physical Research Laboratory, Ahmedabad 380009, India Penner et al., 1998).BC and OC produced in the atmosphere due to biofuel combustion were larger compared to that produced due to fossil fuel combustion during the decade 1990-2000 (Ito and Penner, 2005).However, estimates of black carbon emission from biomass burning are highly uncertain due to lack of activity data and wide variety of biofuels with different emission factors (Gadhavi et al., 2015).Other species co-emitted with the black carbon particles have an influence on the spectral properties of black carbon particles (Bergstrom et al., 2002;Ganguly et al., 2005;Gadhavi and Jayaraman, 2010).The soot (BC) coming from biomass burning has higher absorption at lower wavelengths (Corr et al., 2012).Knowledge on effect of fuel types on the spectral properties of emitted species offers the possibility to improve emission inventory from biomass burning.
It has been noted from a previous study by Gadhavi and Jayaraman (2010) that BC concentration in Indian cities is comparatively higher than the other similar sized cities in the world.People living in rural places are also exposed to comparable BC concentrations.A major part of Indian population resides in villages and people in rural areas, mainly depend on biofuels for their household requirements such as cooking, water boiling, cattle feed preparation, etc.Many varieties of biofuels are available in rural areas and depending upon their availability different biofuels are used in different seasons of the year.In a study by Gadhavi and Jayaraman (2010) related to absorbing aerosol contribution to the biomass burning at Gadanki (village situated in Andhra Pradesh, India), it has been reported that biomass burning used for cooking contribute 40% to aerosol absorption in the morning and evening.
This study focuses on understanding the spectral properties of BC produced by burning different biofuels, used by the villagers for their household purpose.The general motivation of this study is the need to understand the effect of biomass burning on the spectral dependence of the aerosol absorption coefficient which in turn is required for the interpretation of optical remote sensing measurements and for modeling of aerosol forcing of climate (Lewis et al., 2008).The domestic bio-fuels used in the study, have the direct impact on the indoor BC emission in the rural households.Anenberg et al. (2011) noted in a study that the indoor BC emissions are likely to be underestimated and have their own impacts on indoor air quality and mortality.

Survey for the Typical Biofuels Used by the Villagers in and around Gadanki
Gadanki is a typical village in South India without any major industrial activities in the surroundings.The main economic activity of its residents is agriculture.The highway connecting Tirupati and Chittoor cities passes through this village and has low to moderate traffic.Around this village, fields of sugarcane, paddy, peanut, and mango farms are common finds.Coconut, Tamarind and Neem trees are the common trees found.It is safe to assume that biomass burning activity prevailing in and around Gadanki and biomass material used are representative for any of the rural location in South India.The survey has been carried out in and around the residential area of Gadanki (13.5°N, 79.2°E) village by visiting individual houses and asking a set of questions to resident (mostly women folk) about the amount and type of biofuel being used and main activities which involved burning of biofuel.There are about a hundred houses in the village.Most villagers use biofuel for cooking and water boiling.Seasonally they also burn the agriculture waste in their farms.Fig. 1 shows the Google Earth image of Gadanki Village and NARL campus.Residents from twenty houses have been interviewed.Answers to questionnaires about type and the amount of fuel used for the household purposes have been summarized in Table 1.As a part of the questionnaire, residents were asked to describe the fraction of energy need being met by different fuel types, which was then averaged assuming that each house-hold would consumes the same amount of fuel.
During the survey, villagers also reported a seasonal variation in usage of biofuel depending upon the availability.During the summer season, almost all types of fuels are available for the use.This includes dry sugar cane leaves and bagasse, Tamarind tree branches, Neem tree branches.However, during monsoon season, people stock the fuels, which include coconut shells and leaves, branches of trees like Sirigandam, Jabari, Parikki etc.Following the survey, samples of biofuels mentioned in the Table 1 were collected from the villagers and the forest near NARL campus.Then the burning experiments were conducted at NARL.One burning experiment was conducted for each biofuel type.

Measurement of BC from Biofuels Burning
All the biofuels were burnt in a homemade stove, which is a modified and slightly bigger version of the typical fire oven used in the villages.It was fabricated using a cylindrical iron drum (0.85 m height and 0.8 m diameter) fitted with a mesh platform (approximately 0.3 m height from the base) made up of iron rods for keeping the burning fuel and an ash collection platform at the base (Fig. 2).There are three rectangular vents near base for supply of the oxygen to keep the flame alive.
The smoke comes out through the opening at the top of the chamber.After each experiment, ash collected at the bottom of the chamber is removed and the chamber was cleaned using water before the next experiment.Typically, only one experiment per day was performed.Data from the first-time burning of biofuels were not used as they might have been affected by burning of paint of stove.Smoke coming out of the chamber was allowed to mix with ambient air to dilute before being measured by aethalometer.It was not our objective to estimate emission factors rather the experiment was solely focused on spectral properties of smoke.Fig. 3 shows a schematic diagram of the complete set up of the experiment with the stove and aethalometer for the measurement of BC.Additional arrangements, such as fire extinguishers and sand buckets were made as fire safety precautions.
The collected representative biofuel samples were burnt in the chamber separately to see the variation in the spectral dependence of BC produced from it, which is expected to be the characteristic of that biofuel.The time at which burning was started and at which the flame stopped were carefully noted in the log book.The total time taken by the fuel to burn out completely was taken as the burning period.An equal interval of time, before (to record the ambient BC background) and after (to record the post burning status of ambient air) the burning period has been used in the analysis to differentiate the effect of biomass burning on spectral properties, for each experiment.The peak BC concentrations achieved were on an average a factor of 30 higher than background, ruling out the dominance of background spectral properties in observations during burning phase.
The BC produced during the burning has been measured using aethalometer.Details of the aethalometer working principle can be found elsewhere (Lavanchy et al., 1999;Fialho et al., 2005;Hansen, 2005), only a brief description is given here.Aethalometer model AE31 made by Magee Scientific, USA was used for results reported herein.The aethalometer uses quartz fiber filter tape through which known volume of air is passed for a set time (2 min).The instrument monitors change in transmission of filter tape to estimate BC concentrations at seven wavelengths viz.370 nm, 470 nm, 520 nm, 590 nm, 660 nm, 880 nm and 950 nm.
The change in the transmission of light through filter paper in Aethalometer is related to black carbon concentration as shown in the following equation,

[ ] ( )
(1) where, [BC] is black carbon concentration, ATN is the attenuation of light through filter paper, σ is the specific attenuation cross section (m 2 g -1 ), A is the spot area, V is the volume of air passed through the filter.ATN and σ are wavelength dependent quantities.Maximum error contribution from parameters like shadowing effect and ambient temperature changes in the BC concentration are estimated to be not more that 10% (Gadhavi and Jayaraman, 2010).In the current article, if not specifically mentioned, BC concentrations reported are the BC concentration values calculated using ATN at 880 nm channel with a σ value of 16.6 m 2 gm -1 as 880 nm channel is considered standard for calculating BC concentration.
The BC concentration is related to Aerosol Absorption Coefficient (AAC) as shown in Eq. (2).
where, 'β' is the AAC, [BC] and 'σ' are the concentration and absorption cross section of respective wavelength channel.The spectral values of absorption coefficient were normalized to 880 nm to better compare spectral dependency between different experiments and with background.
The AAC can be related to the wavelength using a power where K is a constant, λ is the wavelength and α is the power factor, which is similar to the angstrom exponent in the Aerosol Optical Depth (AOD) calculations, often reported in literature as Absorption Angstrom Exponent.For each experiment, values of α were calculated before, during and after burning period using linear least square fit on logarithmic values of β and λ.

RESULTS AND DISCUSSION
Each experiment data contained the concentrations of BC for the three time frames viz.pre-burning, during burning and post burning period.An increase in the BC concentration during burning has been observed in each experiment as expected.On average, the peak BC concentrations during the burning phase were 30 times higher than the background.Fig. 4 shows BC concentrations from an experiment performed on 20-February-2015 where Tamarind tree branches were burnt.As the BC concentration increases deviation between BC concentrations estimated using different wavelengths becomes apparent in the time plot.
Table 2 shows mean BC values and standard deviation from three time frames calculated separately for each wavelength channel.It is understandable that the standard deviation values for burning period are large to account for the large variations in the BC concentration from start of burning to peak burning phase and then again peak burning phase to the end of burning.
Ideally, one should get same BC concentration values for each wavelength channel.However, one can notice in Table 2 that they are significantly different from each other.Absorption cross-section (and hence absorption co-efficient) varies inversely (value of α equal to 1 in Eq. ( 3)) with wavelength for pure graphite particles.However, the presence of other absorbing material along with graphite particles   3 shows the specific attenuation cross-sections for all 7 wavelengths (Hansen, 2005) used for deriving BC concentrations.Fig. 5 shows plots of AAC v/s Wavelength for Tamarind Branch experiment.Solid line through the data points is based on fitting the power-law equation of the form shown in (Eq.( 3)).In-spite of the large difference in absolute values between burning and non-burning period one can notice the difference in spectral curvature of AAC.
In order to better differentiate the spectral curvature, values of AAC have been normalized to 880 nm as shown in Fig. 5(b).Table 4 shows the power fit results of normalized AAC values obtained from tamarind tree branch burning experiment.
In this experiment, value of α during the burning period is higher compared with that during pre and post burning values.A α value close to 1.0 is expected for the BC from fossil fuel burning, which results in smaller and spherical particles (Kirchstetter et al., 2004;Sandradewi et al., 2008).A value more than 1.5 is expected for larger absorbing particles such as mineral dust or the particles with high Organic Carbon (OC) (Kirchstetter et al., 2004;Bergstrom et al., 2007).Typical α values for soot particles produced from the biomass burning are reported to be in the range 1.5-3.0(Kirchstetter et al., 2004;Clarke et al., 2007).This is because biomass burning also emit lots of OC along with BC, which is reflected in the experiment of Tamarind branch burning.The wooden branches contain different resins and natural oils which are emitted while burning along with the BC, causing the higher absorption in UV region of the spectrum.Similar results have been observed in the experiments conducted using the other biofuels.In all the cases, α values observed were greater than 1.0 and less than 3.0 confirming emission of BC particle with a substantial amount of OC during biofuel burning.Plots of AAC normalized at 880 nm v/s Wavelength for different biofuel burning experiments are shown in Fig. 6.
Table 5 lists α values obtained for different biofuel representative samples determined through the power law fit of the plots shown in Fig. 6.Errors indicate the standard errors from the power fit.Though no systematic observations of burning rate of fuel were made, we could make out qualitatively three categories of burning rate for fuels used in these experiments; a) Very fast burning: Bagasse, Paddy grass and Dry leaves b) Moderately slow burning: Jabbari, Tamarind, Sirigandham, and Parikki tree branches c) Very slow burning: Coconut shells and Coconut parts.
An important observation here is the relationship between burning rate and α value.The higher is the burning rate, lower is the α value and vice versa.The fast burning biofuels such as bagasse, paddy grass, dry leaves and waste have α values ranging between 1.3-1.4whereas moderately slow burning biofuels such as tree branches have the α values ranging from 1.45-1.7 and very slow burning biofuels like coconut parts and shells have highest observed values ranging between 2.0-2.2.We hypothesise that slow burning biofuels emits particles with high OC content and fast burning biofuels emit BC with lower concentrations of OC.The highest α values observed for the coconut parts and shells among the different fuels can be justified by the fact that they are very rich in natural oil content (Pehowich et al., 2000) and hence might have emitted the soot particles with richer OC content causing the higher absorption in the UV region.Where as in dry grass and leaf-litter main chemical constitutes are crude protein and carbohydrates (Garg et al., 2012) and in bagasse mainly the cellulose and hemicellulose (Kim and Day, 2011).Optimal UV detection for bagasse chemical characterisation is reported to be less than 300 nm (Rezende et al., 2011).
The burning was carried out in the open air and the BC produced was allowed to mix with the ambient air.In order to confirm the contribution of BC produced by the burning to the spectral variation we have shown the mean α value with a standard deviation for the month of February 2015 along with α estimated for different biofuels in Fig. 7.It can be clearly observed that, the background α value is close to unity, whereas, for all the biofuel burning experiments it is above unity.This indicates that the observed change in the spectral variation during the biofuel burning is due to fresh BC produced.
In some experiments it can be noted that the post burning α values are slightly lower than the pre-burning values however they fall in the standard deviation range of the background α variation and not statistically significant.As for the effect of background BC concentrations on calculations of spectral variation during burning phase, we estimate that α values during burning phase might have been underestimated by about 4% in most cases and at most 9% in case of "coconut shell" and "coconut parts" experiment.for the spectral dependence of BC are analyzed by fitting power-law equation and estimating exponent of wavelength (α; also known as Absorption Angstrom Exponent in scientific literature).The values of α found to be in the range of 1.29 to 2.16.Slow burning biofuels such as coconut shells and coconut parts are found to have soot particles with the α values more than 2.0, whereas the fast burning biofuels are found to have the α values in the range of 1.2 to 1.4.This, we believe, is because of slow burning biofuels are emitting more low volatile organic compounds compared to fast burning biofuels.Unfortunately, in the current experiment records on amount of the fuel being burnt was not made precisely and have to depend on common knowledge about burning speed of various bio-fuels.As a future direction, experiment on effect of moisture content of the fuel, burning temperature and various burning phases (flaming, smouldering, etc.) on black carbon spectral properties will be highly valuable.

Fig. 4 .
Fig. 4. Aethalometer data during the burning of for the Tamarind branches.

Fig. 5 .
Fig. 5. (a) Spectral variation of AAC v/s Wavelength for three different measurement periods of Tamarind branch experiment.(b) Same as (a) but AAC normalized at 880 nm v/s Wavelength.Solid lines are the power fits through the data points

Fig. 6 .
Fig. 6.Plots of normalized AAC v/s Wavelength.The solid lines represent the power fit through data points.

Fig. 7 .
Fig. 7. Comparison of α values obtained during biofuel burning experiments and that of background.

Table 1 .
Summary of survey of various biofuels used by Gadanki villagers.

Table 2 .
Mean BC concentrations at different wavelengths for the three periods for Tamarind branch experiment.

Table 3 .
Specific attenuation cross-sections for 7 wavelength channels of the Aethalometer.

Table 4 .
Power fit results of tamarind branch experiment.