On the use of the field Sunset semi-continuous analyzer to measure equivalent black carbon concentrations

ABSTRACT

This study describes a method to calculate equivalent black carbon (EBC) concentrations comparable to those obtained from optical filter-based EBC instrumentation from data obtained with a semi-continuous thermo-optical analyzer (Sunset Laboratory Inc., USA) without any need of instrument alterations or extra costs. A correction for the reflection of the Sunset analyzer laser beam by the walls of the sampling tube is introduced. EBC_Sunset concentrations obtained during two intensive campaigns in Prague (one in winter and one in summer) were compared also to EBC measured by an AE31 Aethalometer (EBC_aeth), an AE51 MicroAethalometer (EBC_micro), and a Multi Angle Absorption Photometer (EBC_MAAP). Good agreement was observed in both campaigns. The EBC_Sunset data were also corrected for loading effects in two ways—a simple loading correction and a total correction using data also from the MAAP and a nephelometer. The loading corrections gave similar results for EBC_Sunset and the aethalometer data except for the simple correction for summer EBC_Sunset data. The multiple scattering correction factors computed for EBC_Sunset agreed well with those calculated for EBC_aeth. The wall reflection correction for the Sunset analyzer data further improves the agreement between EBC_Sunset and EBC_MAAP.

Copyright © 2016 American Association for Aerosol Research

EDITOR:

• Thomas W. Kirchstetter

Introduction

Carbonaceous aerosols (CA), their composition, concentration, sources, and temporal and spatial variations have been studied intensively because of their adverse effects on human health (Highwood and Kinnersley 2006) and their climatic effects on regional and global scales (Ramanathan and Carmichael 2008; Bond et al. 2013).

Main methods to measure CA use the optical, thermal, and thermo-optical properties of CA (Watson et al. 2005; Petzold et al. 2013), resulting in various measures: elemental carbon (EC), organic carbon (OC), and black carbon (BC). The terms EC and OC are used for the results of thermo-optical or thermal methods, and BC denotes the results of optical measurements of the absorption properties of the aerosol. Absorption can be measured either on deposited samples (filter-based methods; Bond et al. 1999), or directly in situ (e.g., photoacoustic techniques; Arnott et al. 1999). The filter-based optical measurements measure light attenuation caused by deposited aerosol samples and results in equivalent black carbon (EBC) concentrations (Petzold et al. 2013).

The aethalometer (Hansen et al. 1984), a filter-based instrument to obtain EBC mass concentrations from the change in light attenuation through a filter during sampling and an instrument constant, is a widely used online method. The Sunset semi-continuous analyzer (Bae et al. 2004) gives information on EC and OC concentrations by applying thermo-optical analysis. To correct for pyrolysis in the first heating stage, laser light transmitted through the sample is used. As the transmitted light through the accumulating filter deposit is recorded continuously similarly to absorption-based techniques, we developed a procedure to combine the information from the Sunset semi-continuous analyzer to obtain also EBC concentrations (EBC_Sunset) in a comparable way to, e.g., a single-wavelength aethalometer without any additional cost or alterations to the instrument. EBC_Sunset should therefore be similar to those measured by filter-based optical techniques.

Similarity between methods deriving EBC concentrations from transmittance changes has already been reported by Virkkula et al. (2007) for an AE31aethalometer (EBC_aeth) and a particle soot absorption photometer (PSAP), between EBC_aeth and microaethalometer by Cheng and Lin (2013), and between aethalometer, PSAP, and MULTI ANGLE ABSORPTION PHOTOMEter (MAAP) by Chow et al. (2009). Some differences are expected when instruments use different types of filtration media, as was shown by Snyder and Schauer (2007) in another aethalometer and PSAP comparison. For multi-wavelength off-line thermo-optical instrument, Chen et al. (2015) already calculated EBC concentrations from laser transmission data, and mentioned the high potential of the combination of thermal/optical and optical analyses.

In the current study, we aim to show the possibility to calculate EBC concentrations also from the semi-continuous thermo-optical analyzer built by Sunset Laboratory Inc. with a time resolution of 1 min during a sampling period of the measurement cycle, comparable to that of instruments designed to measure EBC.

Another goal is to show the applicability of loading and multiple scattering corrections on the EBC_Sunset, again making the new method comparable to other filter-based light transmittance methods despite the fact that most of the corrections were optimized for aethalometer data. A new correction needed for the laser signal data recorded by the semi-continuous Sunset analyzer data is also developed.

Methods

Measurement site and periods

The measurements were performed at Prague Suchdol (50°7′36.47″N, 14°23′5.51″E, 277 m a.s.l.), a north-west suburb of Prague at the edge of the plateau above the city. The sampling site (officially classified as an urban background station) is located on the campus of the Czech Academy of Sciences, about 250 m from the nearest road (15,000 cars/day in 2013). The site is located in a residential area without industrial or other sources of pollution.

The measurements were conducted during two intensive campaigns. The summer campaign lasted from 27 June to 8 July 2012, and the winter campaign from 6 February to 15 February 2013. All instruments were equipped with PM10 inlets.

Instrumentation

Sunset semi-continuous analyzer

The Sunset semi-continuous analyzer (Sunset Laboratory Inc., Tigard, OR, USA) is primarily designed for the analysis of organic and elemental carbon (OC and EC) by a thermo-optical method (Bauer et al. 2009). Besides the thermally determined EC and OC, the analyzer enables to determine also a so-called optical elemental carbon (OptEC) using transmittance measurements of a laser beam (660 nm) through the filter during sampling and also during analysis. The difference in the intensity of the laser beam transmitted through the filter before and after the thermal analysis is used to obtain the attenuation of the sample (Jeong et al. 2004). The continuous darkening of the filter during sampling is recorded at 1-min intervals and reported as “minute OptEC” after recalculation. These measurements of OptEC are similar to EBC measurements using aethalometers (Hansen et al. 1984). There is, however, a different way to obtain the attenuation coefficient and in the subsequent conversion of attenuation to EBC concentration, which leads to different values of EBC obtained from aethalometers and OptEC. In this work, raw data (no data recalculated to OptEC) from measured laser intensities during sampling are used to calculate EBC_Sunset (see below). The laser signal during thermo-optical analysis is not considered; thus, the data have gaps of 25 min in every 2 h. An air flow of 8 L/min was maintained by the external pump.

7-wavelength aethalometer

The AE31 multi-wavelength Aethalometer (Magee Scientific Corporation, Berkeley, CA, USA), using a web reinforced quartz fiber filter tape and operating at wavelengths of 370, 470, 520, 590, 660, 880, and 950 nm (Hansen 2005), has no built-in compensation for loading effects in contrast to the newer AE33 (Drinovec et al. 2015). For the instrument comparison here, the EBC concentrations and attenuations obtained at 660 nm wavelength were used (EBC_aeth). An air flow of 4 L/min was maintained by the internal pump.

Microaethalometer

The AE51 Micro Aethalometer (AethLabs, San Francisco, CA, USA) also measures EBC concentrations (EBC_micro) from the attenuation of light passing through aerosol particles deposited on a filter. The main difference to the AE31 is the wavelength of the light source (880 nm), the sample flow rate (50 mL/min, winter campaign; 100 mL/min, summer campaign), and the filter material (T60 borosilicate glass fiber filter with Teflon coating). During the winter campaign, some filter overloading was encountered, which severely impacted the instrument comparisons.

Multi-angle absorption photometer

The Thermo Scientific 5012 MAAP (Thermo Fisher Scientific, Waltham, MA, USA) also measures aerosol EBC concentration (EBC_MAAP) similarly to aethalometers. The MAAP utilizes a 637 nm LED as a light source (Müller et al. 2011). The sample flow of 16.6 L/min is maintained by an external pump and a GF10 glass fiber filter tape is used as sampling substrate. Unlike in the aethalometer, however, multiple scattering effects between the deposited sample and the filter matrix have been accounted for by measuring backscattered radiation at multiple angles and using a radiative transfer scheme in the calculation of EBC (Petzold and Schönlinner 2004; Petzold et al. 2005). As the MAAP does not need empirical correction factors, the instrument has been used as a standard for EBC concentration measurement in instrument inter-comparisons in our study. Further, absorption coefficients determined by the MAAP were used to calculate the single scattering albedo needed for the data correction (Müller et al. 2011).

Nephelometer

An integrating nephelometer (TSI model 3563, TSI Inc., Shoreview, MN, USA; Anderson et al. 1996) was used to measure aerosol scattering coefficients at 450, 550, and 700 nm (scattering angles between 7 and 170°; the data were corrected for the truncation error according to Massoli et al. 2009). The instrument has been calibrated according to the recommendations published by Anderson and Ogren (1998), i.e., using air and CO₂ as calibration gases. A continuous sample flow of 16.6 L/min was maintained by a low volume sampler (LVS3.1, Comde-Derenda). The relative humidity (RH) was controlled with a Nafion dryer. In winter, RH was below 30% during the whole campaign. In summer, due to the high ambient temperatures and RH, the RH of the samples sometimes reached 50%.

Data synchronization

As the instruments had different time resolutions (Sunset semi-continuous analyzer and microaethalometer: 1 min; AE31 aethalometer, nephelometer, and MAAP: 5 min, with different start times), the data had to be synchronized before starting the data processing and analysis. All data were averaged over 5-min time steps starting at 00, 05, etc., with interpolations using cubic splines where necessary, resulting in 3153 and 3569 5-min data points for the summer and winter campaigns, respectively.

Calculation of optical properties

For the Sunset semi-continuous analyzer, attenuation (ATN) was calculated from raw (unprocessed) data similarly as in aethalometers (Hansen et al. 1984; Hansen 2005):[1] where I₀ is the initial laser intensity and I_i is the intensity during the analysis for each measurement point. From the ATN_i obtained in subsequent time intervals, ΔATN is calculated as ΔATN_i+1 = ATN_i+1 – ATN_i and further combined with the (mass) attenuation cross-section σ_ATN given for the aethalometer as 14625/λ [m²/g] (wavelength given in nm), and 12.5 m²/g for the microaethalometer, to yield the increase of surface loading of EBC on the filter dEBC as (Hansen 2005):[2]

Different values of σ_ATN have been reported (Snyder and Schauer 2007 and references therein), depending on the physical and chemical characteristics of light absorbing species (Jeong et al. 2004) but we decided to use the default value given for the aethalometer (Hansen 2005) also for the Sunset analyzer. With the laser wavelength λ = 660 nm, σ_ATN equals 22.2 m²/g. Finally, EBC_Sunset in ng/m³ is given as (Hansen 2005):[3] where A is the filter spot area in cm² (1.33 cm² as calculated from the inner diameter of the quartz tube holding the filter in place), Q is the flow rate in L/min, and t is the minute time interval. The EBC_Sunset concentrations calculated from Equation (3)[3] will be referred to as original data, as no further corrections are applied.

Original EBC data from aethalometers need to be corrected for the so-called loading effect, and the scattering effect (Weingartner et al. 2003; Arnott et al. 2005; Schmid et al. 2006; Virkkula et al. 2007; Collaud Coen et al. 2010). Further, we found that the original EBC_Sunset data also have to be corrected for reflection of light on the wall of the sampling tube.

We therefore developed a wall reflection correction, and applied a simple loading correction to EBC data obtained from all instruments, and another, more complex, loading correction (called total correction) using absorption coefficients obtained by MAAP and scattering coefficients obtained by the nephelometer. The multiple scattering correction factor was calculated for all datasets.

Wall reflection correction

As light can propagate inside quartz fibers, we decided to verify if such an influence can be detected in the setup of the Sunset semi-continuous analyzer consisting of the light scattering quartz fiber filter and the quartz tube that holds the filter in the oven and conducts the sampled aerosol to the filter. Thus, experiments with black filters were made (Figure 1).

Figure 1. Examples of placements of filters in the Sunset semi-continuous analyzer during the wall reflection experiments.

First, two clean (white) quartz fiber filters (diameter 16 mm) were placed in the instrument the same way as during conventional measurements with sample collection and the transmitted laser signal was recorded. Then a black filter (a quartz fiber filter soaked with black dye and dried) that should absorb the laser beam passing through the first two white filters was placed on top of the white filters facing away from the incident beam. The black filter had the same diameter as the inner diameter of the tube (13 mm), to simulate maximum loading during sampling. This led to a sharp drop of the laser signal from a value of ca. 9200 in arbitrary units (a.u.) with two white filters to 860 a.u. with two white and one black filter (Figure 2). The standard deviations of the measured laser intensity were between 0.1 and 0.2%, thus not visible in the plot. To confirm that all light was absorbed by the first black filter, several black filters were added gradually. The first black filter lowered the laser signal by more than a factor of 10; further signal decreases with additional black filters were small. A similar decrease in the laser signal with increasing number of black filters was observed also when there was only one white filter instead of two (Figure 2).

Figure 2. Laser signal detected behind filters during experiments with black filters in the Sunset semi-continuous analyzer.

The small decrease of measured laser intensity when additional black filters were added shows that already the first black filter blocks the laser light completely in forward direction; however, reflections inside the white filters lead to transmission of the light around the black filters. The slight decrease of intensity through more filters is probably caused by blocking part of the reflected light due to increased thickness of black filters. If the light is led through the first black filter (resulting in ten times lower transmitted light), the second filter should decrease the light intensity the same way. Although the thickness of the black filters (0.1 mm each) might be seen negligible compared to tube length, it is not negligible in comparison to quartz tube wall thickness (1 mm), where large part of reflections happens, as can be seen in Figure 1. We therefore deduce that the white filters scatter the laser light, which is subsequently reflected by the quartz tube walls to the light detector. We believe the process happens also in particle loaded filters for three reasons. Firstly, the filters in the instrument are held in place at their edge by the quartz tube that leads light to the photomultiplier and this part is not involved in particle filtration process. Secondly, the most of particles are captured near the surface of the front filter. With regard to using two filters in series, it means that the particle loaded layer is thin in comparison with the whole double filter thickness. Thirdly, as black filter during the test covers the inner white surface completely, the path of the light causing the effect must go around this inner white surface and therefore the particle loading probably will not influence the effect. Or the difference will be negligible.

To correct for this effect, a constant correction value has to be subtracted from the laser signal values during the measurement period until the next filter change. The correction is calculated as a percentage of the initial laser signal after each filter change. For our instrument, this value is 9.3 ± 0.3% of the initial laser intensity, calculated as a ratio of laser intensity through two white and one black filter, and laser intensity measured through two white filters. The uncertainty was calculated from the error propagation theory. It can be assumed that a similar correction value is valid also for other semi-continuous Sunset analyzers, but each analyzer should be checked individually as the correction value may be affected by laser quality, a soiled tube, and/or the analyzer oven (Chiappini et al. 2014; Panteliadis et al. 2015).

Loading correction

The EBC data were corrected for the loading effect according to Virkkula et al. (2007) as:[4] with the correction factor k:[5] where and are original EBC concentrations at the beginning of the n + 1 measurement point, and at the end of the nth measurement point, and and are the corresponding attenuations.

This correction was chosen for its simplicity; it does not use scattering data. Although it was intended for aethalometer data, we used it also for the microaethalometer and the Sunset semi-continuous analyzer data. A correction of the microaethalometer data, however, was impossible. During the winter campaign, eight filter changes were performed, i.e., 8 k factors were computed. Only two of the factors were positive, six were negative implying that the uncorrected values overestimate the true EBC values (Virkkula et al. 2007), and most of the correction factors yielded negative EBC concentrations.

The AE31 aethalometer data were corrected for each wavelength independently. For the comparison with other instruments, only the 660 nm data were used. As neither Sunset analyzer, nor MAAP, nor microaethalometer are multi-wavelength instruments, only one wavelength was evaluated for a better comparison, and the 660 nm was chosen as it matches the Sunset analyzer wavelength (660 nm) and is close to the MAAP wavelength (637 nm).

Two different approaches for correcting EBC_aeth and EBC_Sunset data were applied: (A) The k values were calculated for each filter change, and the corresponding EBC concentrations were corrected according to Equation (4)[4] with the corresponding k. (B) Only the median k from the whole campaigns were considered, and all data from a campaign were corrected with the same k, resulting for the Sunset semi-continuous analyzer in k = 0.026 in summer, and k = 0.031 in winter. These k are higher than those for the AE31 aethalometer (0.0020 and 0.0052), and also higher than the values −0.0024 and 0.0047 reported for a winter and a summer campaign at Hyytiälä, Finland (Virkkula et al. 2007), or than 0.0033 in February in Taipei (Cheng and Lin 2013).

Total correction

The correction introduced by Collaud Coen et al. (2010) taking the advantage of simultaneous measurements with MAAP and the nephelometer was also applied, which takes both loading and multiple scattering effects into account. For the multiple scattering correction, however, multi-wavelength absorption and scattering measurements are necessary. Thus, the data were corrected only for the loading effect, and the multiple scattering correction was dealt with independently.

The filter loading correction R_new,n was calculated as (Collaud Coen et al. 2010):[6] where ATN_n is the attenuation in % and the mean single scattering albedo of the nth measurement points since the filter spot change at AE31 aethalometer and microaethalometer, and nth measurement points since the beginning of Sunset analyzer sampling. The single scattering albedo was calculated using absorption coefficients measured with the MAAP and according to Massoli et al. (2009) truncation effect corrected scattering coefficients measured with the nephelometer.

The range of loading correction factors for all three instruments is similar. Good agreement was found between AE31 aethalometer and Sunset semi-continuous analyzer data. Both datasets show a similar temporal behavior of the correction factor, and also the same seasonal differences. The difference of correction factors for the microaethalometer is more pronounced. A variability between summer and winter samples is still visible, but the comparison is complicated by the significantly different frequency of filter changes (several days in microaethalometer, 2 h in the Sunset semi-continuous analyzer).

The correction factor was then used to calculate a new attenuation coefficient b_ATNR:[7] which was then used to calculate EBC concentrations according to the equation by Weingartner et al. (2003):[8]

Multiple scattering correction

This correction usually uses a calibration factor C to calculate the absorption coefficient from the attenuation coefficient, as attenuation coefficients were found to be 1.9–2.5 times higher than specific absorption cross-sections for airborne BC particles (Petzold et al. 1997). The method inter-comparison by Müller et al. (2011) uses C = 2.14, reported originally for measurements with diesel soot particles (Weingartner et al. 2003). Collaud Coen et al. (2010) measured C between 2.9 and 4.3 for various datasets, and also report seasonal dependences of C, probably caused by semi-volatile OC and water vapor. In our study, C was calculated also for the AE31 aethalometer, microaethalometer, and Sunset semi-continuous analyzer datasets both for winter and summer campaigns, using (Collaud Coen et al. 2010):[9] where b_MAAP is the absorption coefficient measured by the MAAP and attenuation coefficient corrected for loading by from Equation (7)[7] .

Mean C_ref values were related to low ATN values, i.e., ATN between 10 and 20% (Weingartner et al. 2003). For the microaethalometer, however, the lowest attenuation was above 30%, so C_ref values were calculated also for attenuation between 30 and 50% for all instruments.

Results

Original data

First, raw EBC concentrations without any corrections were compared, taking EBC_MAAP as a standard. As a method of comparison, the ordinary least square linear regression with forced zero was used. The non-forced linear regression was calculated as well, but provided significantly lower coefficients of determinations (lower about 0.2 and more), similarly to the results reported, for example, by Slowik et al. (2007). The results of the linear regression are presented in Table 1.

Table 1. Results of linear fits y = ax of the EBC_aeth, EBC_Sunset, and EBC_micro concentrations versus EBC_MAAP concentrations, as derived from original and corrected data for both campaigns.

		Winter		Summer
		Slope a	R^2	Slope a	R^2
Original data	Aethalometer	0.83	0.97	0.97	0.93
	Sunset analyzer	0.77	0.99	0.89	0.92
	Microaethalometer	1.79	0.85	0.96	0.98
Wall reflection correction	Sunset analyzer + w	0.85	0.99	0.99	0.92
Loading correction (Virkkula et al. 2007)	Aethalometer	0.93	0.99	1.03	0.92
	Sunset analyzer + w	1.15	0.91	1.22	0.70
Loading correction k (Virkkula et al. 2007)	Aethalometer	0.96	0.89	1.05	0.93
	Sunset analyzer + w	1.26	0.78	1.37	0.71
Total correction (Collaud Coen et al. 2010)	Aethalometer	0.92	0.73	1.23	0.79
	Sunset analyzer + w	1.09	0.68	1.30	0.83
	Microaethalometer	2.26	0.74	1.34	0.84

The “wall corrected” EBC_Sunset are denoted with “+w.” The loading correction “k” denotes the approach with constant k.

The seasonal differences in comparisons are striking (Figure 3). In summer (Figure 3, right), all EBC concentrations are quite comparable; the slopes do not differ by more than 11% from a 1:1 agreement with EBC_MAAP, and the R² are high (≥0.92). Some overestimation is visible below 100 ng/m³ EBC_MAAP in Figure 3, right. Generally, however, all the instruments give comparable EBC concentrations, which agrees to earlier results of a comparison of an aethalometer, a light transmission method, a MAAP, and an integrating sphere in Vienna, Austria (Hitzenberger et al. 2006), or to comparison of aethalometer, PSAP, MAAP, and photoacoustic analyzer in Fresno, California (Chow et al. 2009).

Figure 3. Original EBC concentrations as measured by the Sunset analyzer (EBC_Sunset), the AE31 aethalometer (EBC_aeth), and the microaethalometer (EBC_micro) compared to EBC concentration data from the MAAP (EBC_MAAP). Left: winter campaign; right: summer campaign.

In winter, the situation changes significantly (Figure 3, left). The best fit to EBC_MAAP (slope 0.83) was found for EBC_aeth, with R² = 0.97. In terms of slope and R² as well, the EBC_Sunset data are comparable to EBC_aeth in relation to EBC_MAAP (slope 0.83 vs. 0.77, R² 0.97 and 0.99). EBC_aeth and EBC_Sunset are therefore comparable to each other with 9% difference (calculated from the linear fits of EBC_Sunset on EBC_aeth), but both are lower than EBC_MAAP. Similar results were obtained by Reisinger et al. (2008) in a comparison of EBC_MAAP values to EBC concentrations measured by several optical methods in winter in Vienna, where the differences were explained by the influence of brown carbon (BrC) on optical techniques. As both location (Central Europe) and source characteristics (traffic and heating) are similar in Prague and Vienna, the discrepancy between the slopes of EBC_MAAP and EBC_aeth and EBC_Sunset in the present study might also be impacted by BrC.

The slope of EBC_micro versus EBC_MAAP is 1.79. From the comparison of time series, the overestimation is clearly visible during almost the whole winter campaign, and most prominently in the last four days (Figure 4, top). No similar feature was observed during the summer campaign (Figure 4 bottom). We decided to keep the data unchanged, as there was no indication of instrument malfunction during the measurements and the discrepancies were found only in the comparison reported here.

Figure 4. Time series of the original EBC_micro, EBC_aeth, EBC_MAAP, and EBC_Sunset concentrations during the winter (top) and summer (bottom) campaigns.

Wall reflection corrected data

The correction for the wall reflection along the quartz tube was applied to the original EBC_Sunset data. In both seasons, the wall reflection correction improves the agreement between EBC_Sunset and EBC_MAAP. The correction does not influence the variance, but only the slope of the linear fit (Table 1, Figure 5).

Figure 5. Original and wall reflection corrected EBC_Sunset concentrations data compared to EBC_MAAP concentrations in winter (left) and summer (right).

In winter, original EBC_Sunset concentrations were 23% lower than EBC_MAAP; after the correction, EBC_Sunset were only 15% lower. In summer, the improvement was from 11% to 1% underestimation. The slopes of linear fits of corrected EBC_Sunset versus EBC_MAAP are almost identical to those of the original EBC_aeth compared to EBC_MAAP. Because of this improvement, wall reflection corrected EBC_Sunset are used in the discussions below.

Although we did not investigate the applicability of the wall correction to derivations of EBC from other types of thermo-optical analyzers (e.g., the off-line Sunset analyzer), we suggest that it should be possible to derive corrected EBC concentrations also from these instruments if the measured transmission values before and after the thermal analysis are used to calculate attenuation (Chen et al. 2015).

Loading corrected data

EBC concentrations were corrected for the loading effect according to the method described by Virkkula et al. (2007), with the exception of EBC_micro, for which the correction produced mainly negative concentrations (although, for example in Cheng and Lin (2013), the correction was possible to be applied).

The k values were calculated for each filter spot change in the AE31 aethalometer and each measurement cycle in Sunset semi-continuous analyzer, and the correction according Equation (4)[4] was applied. The loading correction improves the agreement between EBC_aeth and EBC_MAAP in winter, and does not change the agreement in summer. The EBC_aeth concentrations agree within −7% in winter and +3% in summer with EBC_MAAP, with R² of 0.99 and 0.92, respectively (Table 1, Figure S1 in the online supplementary information (SI)).

If the correction is applied on EBC_Sunset, the result is different; in winter, the corrected data overestimate EBC_MAAP by 15%, compared to a 15% underestimation without the correction. In summer, the corrected EBC_Sunset overestimate EBC_MAAP by 22%, while there was an underestimation of only 1% without the loading correction. Also, R² decreased from 0.92 to 0.70. This can probably be explained by artificial spikes in the time series of the loading corrected EBC_Sunset in summer, which were a result of the correction, as they were not visible in the original data (Figures 6a and b). The effect was not observed to such an extent in winter, with some exception on the evening of 13th February (Figure S2 in the SI). The loading correction with variable k-values is thus not well suitable for EBC_Sunset in summer, which might be due to rapid changes in the aerosol.

Figure 6. Top: One-day time series of original EBC_MAAP (uncorrected), and EBC_aeth and EBC_Sunset corrected for loading according to Virkkula et al. (2007) in winter (a) and summer (b). Middle: Time series of EBC_MAAP (uncorrected), and EBC_Sunset corrected for loading according to Virkkula et al. (2007) either by a k value calculated for each Sunset semi-continuous analyzer filter spot change (EBC_Sunset) or by a constant k value (EBC_{Sunset_K}) in winter (c) and summer (d). Bottom: Time series of EBC_MAAP (uncorrected), and EBC_aeth, EBC_Sunset, and EBC_micro corrected for loading according to Collaud Coen et al. (2010) in winter (e) and summer (f).

The approach with average k values, recommended for rapidly changing concentrations by Virkkula et al. (2007), was therefore applied to EBC_Sunset. The effect of this approach again differs with season (Figure 6 middle). In winter, the correction leads to concentrations that are about 26% overestimated, which is higher than the overestimation without the constant k (15%). R² decreases from 0.91 to 0.78. In Figure S3 in the SI, the increase in both variability and concentrations in some periods is clearly visible (7.2., 13.2.). In summer, the constant k value helps to remove some of the artificial peaks due to the loading correction with a different k value for each filter spot, without a significant change of R² (from 0.70 to 0.71), but leads to periods with increased concentrations similarly to the winter episodes (Figure 6d) The 22% higher EBC_Sunset with variable k become 37% higher compared to EBC_MAAP, if a constant k value is used.

The same approach was tested also on AE31 aethalometer data, resulting in only minor changes in the agreement between EBC_aeth and EBC_MAAP compared to the use of a spot-dependent k-value (Table 1).

In summary, the correction in the form of time-dependent k values works well for winter EBC_Sunset. In summer, this correction does not improve the results but increases the noise in the data. The application of a constant k value is not recommended for either winter or summer data, as it increases the difference between EBC_Sunset and EBC_MAAP.

Total correction data

The correction introduced by Collaud Coen et al. (2010), which also takes the filter loading into account, was applied also to the microaethalometer data, so all three datasets were compared to EBC_MAAP (uncorrected).

Unlike the loading correction alone, the total correction does not create new peaks in the time series of EBC_aeth and EBC_Sunset (again with the exception of 13.2., see Figure S4 in the SI), and also the saw-tooth pattern of the EBC concentration is suppressed (Figures 6e and f). It was possible to apply the correction also on EBC_micro. In summer, the time series of EBC_micro generally agree with those of other EBC data; in winter, however, the periods of highly overestimated EBC_micro values are even more pronounced. Also for EBC_aeth and EBC_Sunset, the total correction increases the EBC concentrations, and, again, there is a significant difference between summer and winter results (Figure 7).

Figure 7. Loading corrected EBC_Sunset, EBC_aeth, and EBC_micro data according to the correction by Collaud Coen et al. (2010), compared to EBC_MAAP concentrations in winter (left) and summer (right).

In winter, the corrected EBC_aeth and EBC_Sunset do not differ by more than 8% from EBC_MAAP with relatively high R² values (0.73 and 0.70 for EBC_aeth and EBC_Sunset, respectively). The comparison of corrected EBC_micro and EBC_MAAP has the same R², but EBC_micro are 126% higher. In the scatterplot in Figure 7, an area of unrealistically high concentrations is clearly seen. Based on the comparison with the other instruments, it would be possible to remove the data; however, no such action was taken in any dataset, as all datasets were dealt with independently to show the possibility of their usage even without extra information obtained by other instrumentation.

In summer, the total correction overestimates EBC similarly for all the instruments, and with similar R² values. The closest agreement with EBC_MAAP was found for EBC_aeth, which were higher by 23%.

In conclusion, the total correction taking into account also the single scattering albedo can be applied to the EBC_Sunset data. It does not introduce much noise into the data, and EBC_Sunset is comparable to EBC_aeth in both campaigns. The application of the total correction increases the EBC concentrations—the values were higher than those of the MAAP.

Multiple scattering corrected data

For all three instruments (aethalometer, microaethalometer, and Sunset online analyzer), C_ref were calculated according to Equation (9)[9] , and compared for both campaigns. The C_ref were calculated for low attenuation data, as the artifact should be the result of filter properties and not of the deposited aerosol (Collaud Coen et al. 2010). Thus, C_ref was calculated when attenuation was between 10 and 20% for the AE31 aethalometer and the Sunset semi-continuous analyzer. For the microaethalometer, minimum attenuation values were about 30%, so C_ref was calculated also from the 30–50% attenuation interval for all instruments. In Table 2, median C_ref values are given, and also their mean values with standard deviations to describe the data variability.

Table 2. Multiple scattering correction factors C_ref for the aethalometer, the microaethalometer, and the Sunset semi-continuous analyzer, calculated from Equation (9)[9] separately for both campaigns, and from low (10–20%) and medium (30–50%) attenuation values.

		Aethalometer	Microaethalometer	Sunsetanalyzer
ATN 10–20%
Winter	Median	3.2		3.7
	Mean ± st. dev	4.5 ± 6.7		4.8 ± 7.8
Summer	Median	4.0		4.3
	Mean ± st. dev	6.4 ± 12.5		9.9 ± 34.8
ATN 30–50%
Winter	Median	3.2	10.7	3.7
	Mean ± st. dev	3.4 ± 1.2	11.4 ± 8.0	3.9 ± 1.1
Summer	Median	4.2	4.9	4.6
	Mean ± st. dev	6.3 ± 15.7	6.5 ± 13.5	6.7 ± 17.9

The median of low attenuation C_ref calculated for the AE31 aethalometer was found to be 3.2 in winter, and 4.0 in summer, which agrees well with results from Cabauw (Collaud Coen et al. 2010). The medians of C_ref from the Sunset semi-continuous analyzer are 15% higher in winter and only 7% higher in summer, proving again the possibility to use this instrument for the EBC concentration calculation. The mean values of C_ref from the AE31 aethalometer and the Sunset semi-continuous analyzer agree within the uncertainty range (Table 2). The standard deviation of C_ref from the Sunset analyzer is higher than that of C_ref from the AE31 aethalometer with a more pronounced difference in summer.

The medium attenuation C_ref values calculated from the AE31 aethalometer were found to be the same as the low attenuation ones in winter; in summer, there was a minor increase from 4.0 to 4.2. Again, aethalometer and Sunset semi-continuous analyzer results agree well with each other. In winter, the median C_ref from the Sunset semi-continuous analyzer is again 15% higher than that from the aethalometer, and 10% higher in summer. The microaethalometer C_ref in summer are comparable to those of the aethalometer and the Sunset semi-continuous analyzer, 4.9. The winter microaethalometer C_ref of 10.7 is clearly higher than typical values 2.9–4.3 (Collaud Coen et al. 2010). The mean value agrees with the AE31 aethalometer and Sunset semi-continuous analyzer C_ref within the uncertainty range, however, due to the significantly higher standard deviation. In summer, the standard deviation is high for all three datasets (Table 2).

Conclusions

The transmitted semi-online Sunset analyzer laser intensity data were used to calculate EBC concentrations with a similar procedure as is used in the AE31 aethalometer software. These EBC_Sunset data were then compared to original EBC concentrations obtained with an AE31 aethalometer (EBC_aeth) and a microaethalometer (EBC_micro), and also to EBC concentrations measured with a MAAP (EBC_MAAP) during two intensive campaigns in Prague.

Not only original EBC concentrations from the instruments were compared to each other, but also several corrections were applied to show the applicability, advantages, and also disadvantages of deriving EBC concentrations from the Sunset analyzer, which was originally intended to measure EC and OC in a thermo-optical analysis. The EBC concentrations from the individual instruments were compared to the EBC_MAAP concentrations and fitted with linear functions forced through zero. All results of the linear fits (slopes and coefficients of determination) are summarized in Table 1.

Original or raw EBC_Sunset concentrations are comparable to those data available from the AE31 aethalometer, if no additional corrections are applied. A good agreement was found not only between EBC_Sunset, EBC_aeth, and EBC_micro, but also an agreement with EBC_MAAP was observed in both the summer and winter campaigns (with the exception of the microaethalometer, which might have had some failure during the winter campaign).

Because of the necessity to correct the laser signals given by the Sunset semi-online analyzer for light reflection on the quartz tube, a new correction was introduced. This wall reflection correction consists of subtracting 9.3% of the initial laser signal after a filter spot change from all the signals are measured till the next filter change. In both seasons, the application of the wall reflection correction improves the agreement between the EBC_Sunset and EBC_MAAP datasets.

It was shown that the EBC_Sunset data can also be corrected for the loading effect. Two different approaches were taken; one without the use of additional information on aerosol optical properties and another using scattering and absorption coefficients obtained from other instruments. The simple loading correction by Virkkula et al. (2007) resulted in a good agreement between EBC_Sunset, EBC_aeth, and EBC_MAAP in winter, if the time-dependent correction was applied. In summer, the increase of noise and number of outliers in the dataset is too strong and the application of a constant correction factor yields highly unsatisfactory results. This simple loading correction does not seem to be suitable for EBC_Sunset data. A similar conclusion has been drawn for aethalometer data also by Collaud Coen et al. (2010).

The total correction, introduced by Collaud Coen et al. (2010), taking into account also the single scattering albedo, can be applied to the EBC_Sunset data as well. It introduces less noise into the data, and the EBC_Sunset concentrations are comparable to the EBC_aeth concentrations both in winter and summer. It was possible to apply the correction also to the EBC_micro data. The application of the total correction increases the EBC concentrations that are thus, compared to EBC_MAAP, overestimated. In winter, the corrected EBC_aeth and EBC_Sunset concentrations do not differ by more than 8% from the EBC_MAAP concentrations, while the EBC_micro corrected data are overestimated by 126%. In summer, the total correction overestimates the EBC values similarly for all the instruments (23–34%), and with similar R² values. The closest result to EBC_MAAP was found for the EBC_Sunset concentrations, with an overestimation of 23%.

The multiple scattering effect can be corrected for the Sunset analyzer data too. The C_ref calibration factors derived from Sunset analyzer data agree well with those calculated from the AE31 aethalometer, and are comparable also with the values reported by other authors, for example Collaud Coen et al. (2010).

In summary, it is possible to use the semi-continuous Sunset analyzer to measure EBC concentrations with no additional costs or alternations of the instrument. Such derived data cannot compete with data obtained from instruments dedicated to measure EBC concentrations, as the Sunset analyzer is designed primarily for thermo-optical analyses. In the absence of a dedicated instrument, however, EBC_Sunset can be used to get an indication of EBC concentrations, as the results are comparable to EBC concentrations obtained from an AE31 aethalometer and a microaethalometer. With the exception of the loading correction based on Virkkula et al. (2007) in summer, it is technically possible to correct the EBC_Sunset data the same way as aethalometer data, but the application of correction does not increase the agreement to EBC_MAAP or even worsen the agreement. Therefore, some more Sunset analyzer dedicated alternations of the corrections would be convenient.

Funding

The measurements were performed in the framework of a project by the ÖAD, WTZ CZ 03/12, and MEYS MOBILITY grant No. 7AMB12AT021. The data analysis was supported by the Czech Science Foundation under the grant no. P503/12/G147, and by the research stay of N.Z. at University of Vienna funded by University of Vienna. The AE31 and nephelometer instruments were kindly provided by the Global Change Research Centre, Czech Academy of Sciences (supported by the Ministry of Education, Youth and Sports of the Czech Republic within the National Sustainability Program I (NPU I), grant number LO1415).