A simple procedure for correcting loading effects of aethalometer data.

ABSTRACT

A simple method for correcting for the loading effects of aethalometer data is presented. The formula [BC.sub.CORRECTED] = (1 + k x ATN) x [BC.sub.NONCORRECTED], where ATN is the attenuation and BC is black carbon, was used for correcting aethalometer data obtained from measurements at three different sites: a subway station in Helsinki, an urban background measurement station in Helsinki, and a rural station in Hyytiala in central Finland. The BC data were compared with simultaneously measured aerosol volume concentrations (V). After the correction algorithm, the BC-to- V ratio remained relatively stable between consequent filter spots, which can be regarded as indirect evidence that the correction algorithm works. The k value calculated from the outdoor sites had a clear seasonal cycle that could be explained by darker aerosol in winter than in summer. When the contribution of BC to the total aerosol volume was high, the k factor was high and vice versa. In winter, the k values at all wavelengths were very close to that obtained from the subway station data. In summer, the k value was wavelength dependent and often negative. When the k value is negative, the noncorrected BC concentrations overestimated the true concentrations.

INTRODUCTION

The origin of black carbon (BC) aerosol is in the incomplete combustion of fossil fuels and various types of biomass burning, ranging from small-scale residential wood combustion to large forest fires. BC aerosols have significant adverse health effects and they also play an important role in climate forcing because they are the most important contributor to light absorption by aerosols. BC is therefore measured in urban, rural, and background areas worldwide.

The aethalometer (1) is probably the most common method for measuring BC concentrations. In the method, air is drawn through a filter and the decrease of light transmission through the sampling area A is measured. Decreasing transmission leads to increasing attenuation, ATN [equivalent to] -ln(I/[I.sub.0]), where [I.sub.0] is the light intensity of the incoming light and I is the light intensity after passing the filter. In the aethalometer it is assumed that the ATN increase is only because of light absorption by BC accumulating on the filter, and BC concentration is therefore calculated from the rate of change of attenuation:

BC = [[sigma].sub.abs]/[[alpha].sub.abs] = [1/[[alpha].sub.abs]] [A/Q] [[DELTA]ATN/[DELTA]t] (1)

where [[sigma].sub.abs] is the particle absorption coefficient, [[alpha].sub.abs] the mass absorption cross section of BC, and Q is the airflow rate through the filter. It is well known, however, that the relationship between ATN change and BC concentration is not linear. (2-6) There are several reasons for this, including that both scattering and absorbing particles collected on the filter alter the internal reflection of the filter in a way that changes the absorption of the aerosol/filter combination. (3,7-9) There are two main consequences: (1) as the filter gets darker, that is, as ATN increases the measured BC concentration gets underestimated; and (2) scattering aerosol gets interpreted as BC. The first of these is the more important one. These effects may be taken into account using empirical correction functions, such as that derived by Weingartner et al. (4) Arnott et al. (6) derived a model-based algorithm for the correction of aethalometer data. Both of these correction methods take both aerosol scattering and absorption coefficients into account. In the new instrument, the Multi-Angle Absorption Photometer (MAAP), these effects are taken into account already in the design and an internal algorithm of the instrument. (10,11)

In this work an alternative correction algorithm is presented for the aethalometer data. The procedure does not take scattering into account because many organizations use the aethalometer in the field without any instrument that measures scattering and such a procedure is needed. The principle of the algorithm is presented and then applied to three different datasets: measurements at a Helsinki city subway station; at an outdoor urban air measurement station in Helsinki; and at Hyytiala, a rural station in a forest in southwestern central Finland.

ALGORITHM

The operation principle of the aethalometer is very close to that of the Particle Soot Absorption Photometer (PSAP) (3,12) so formulas developed for the PSAP data are used here. Virkkula et al. (12) derived an empirical correction formula for the PSAP data

[[sigma].sub.abs] (corrected) = ([k.sub.0] + [k.sub.1]ln(I/[I.sub.0]))[[sigma].sub.0] - s[[sigma].sub.SP] (2)

where [k.sub.0], [k.sub.1], and s are empirically derived constants; [[sigma].sub.0] is the noncorrected absorption coefficient, defined essentially the same way as [[sigma].sub.abs] in eq 1; and [[sigma].sub.SP] is the particle scattering coefficient. Using the definition of ATN above, eq 2 becomes

[[sigma].sub.abs] (corrected) = ([k.sub.0] - [k.sub.1]ATN))[[sigma].sub.0] - s[[sigma].sub.SP] (3)

It is assumed here that the correction function for the aethalometer data is of the same form. In the aethalometer data files ATN is presented as 100 x (-ln(l/[I.sub.0])), so for simplicity this ATN will be used in the algorithm. It is also assumed here that the raw BC given by the aethalometer is correct when the filter is clean, that is, when ATN = 0. Several aethalometer users do not have a nephelometer in use and they have no information on [[sigma].sub.SP] so the algorithm presented here does not have it either. It follows that for the correction algorithm [k.sub.0] = 1 and s = 0, there is only one constant to be found. The corrected absorption coefficient then becomes

[[sigma].sub.abs] (corrected) = (1 + k x ATN) [[sigma].sub.abs] (noncorrected) (4)

and the corrected BC concentration is calculated from

[BC.sub.CORRECTED] = [[[sigma].sub.abs] (corrected)]/[[alpha].sub.abs] = (1 + k x ATN)[BC.sub.0] (5)

where [BC.sub.0] is the noncorrected BC concentration given by the aethalometer. This notation will be used in the text below. Probably the most used operational mode of the aethalometer is such that it collects the sample on a filter tape that moves forward when ATN through the spot has reached a preset limit--in the data to be discussed below the spot change took place when ATN was approximately 75%--and starts measuring the next spot. A value for the factor k in eq 5 is calculated for each filter spot so that the data become continuous, that is,

[BC.sub.CORRECTED]([t.sub.i,last]) = [BC.sub.CORRECTED]([t.sub.i+1,first]) (6)

where [t.sub.i,last] is the time of the last measurement data for filter spot i, and [t.sub.i+1,first] is the time of the first measurement data for the next filter spot. Applying eq 5 in eq 6, we obtain the formula for calculating the factor k for the filter spot i:

[k.sub.i] = [[BC.sub.0]([t.sub.i+1,first]) - [BC.sub.0]([t.sub.i,last])]/[ATN([t.sub.i,last]) x [BC.sub.0]([t.sub.i,last]) - ATN([t.sub.i+1,first]) x [BC.sub.0]([t.sub.i+1,first])] (7)

This is the general form for calculating [k.sub.i] In typical atmospheric conditions this can be somewhat simplified. Right after the filter spot has been changed the first ATN [approximately equal to] 0 so eq 7 becomes

[k.sub.i] [approximately equal to] [1/ATN([t.sub.i,last])] ([[[BC.sub.0]([t.sub.i+1,first])]/[[BC.sub.0]([t.sub.i,last])]] - 1) (8)

The obtained factor ki is then used for correcting all data obtained for filter spot i according to eq 5. In practice, ATN is not exactly 0 even in the first data line unless the BC concentration equals zero. The BC data have also noise that is not due to the effect discussed here, i.e., instrumental noise or true variation in BC concentrations. Therefore, in practice the k factors were calculated from eq 8 using the average [BC.sub.0] of the last three data of filter spot i and the first three data of filter spot i + 1 for the measurements that had a 1-min time resolution and the average of last two data of filter spot i and the first two data of filter spot i + 1 for the measurements that had a 5-min time resolution.

In case the aerosol contains significant amounts of BC, i.e., in exhaust gas or smoke measurements not presented in this paper, the ATN of the first measurement data of the new filter spot is already clearly higher than zero. In this case the assumption that the first ATN [approximately equal to] 0, which was used to obtain eq 8 is not valid and eq 7 has to be used.

MEASUREMENTS

Three datasets are used for testing the algorithm. First, the measurements in a Helsinki city subway station are analyzed in more detail than the other data. The reason is that these data are the simplest and the time resolution is the highest so they are the best data for presenting the strengths and weaknesses of the method. Next, the algorithm is applied to two different types of atmospheric aerosols: urban and rural.

[FIGURE 1 OMITTED]

The subway measurements were part of a large campaign to evaluate the exposure to particulate matter in the subway system of Helsinki. The main results of the campaign were presented earlier by Aarnio et al. (13) In the campaign, measurements were carried out at two surface stations: one ground-level station and in subway cars. In this paper, the data of the measurements carried out at the underground subway station of Sornainen in the center of Helsinki from March 4-18, 2004, are used.