Evaluation of measurements of light transmission for the determination of black carbon on filters from different station types

Abstract

We compared two approaches (denoted as Virkkula” and “Drinovec” equations) for the determination of black carbon via attenuation measurements with a transmissometer OT-21 (Magee Scientific) on quartz fiber filters from different station types (urban-traffic, rural-background and remote) within the Austrian air monitoring network. First, we evaluated the entire data set to compare the performance of the models in the determination of the loading parameter k and the mass attenuation cross section σ. Then we investigate k and σ for different station types. The data set comprises 299 filters, all representative for 24 h samples collected with High-Volume-Samplers. Mass concentrations ranged from approximately 5-100 μg m−3 and covered all seasons.Both approaches give statistically indistinguishable results of k with values of 1.00*10−3 ( ±2*10−4) and 0.97*10−3 ( ±1*10−4), respectively, if the entire data set is evaluated. The “Drinovec” approach gives slightly lower values for σ than “Virkkula” with values of 32.1 m2 g−1 ( ±1.1 m2 g−1) and 35.3 m2 g−1 ( ±1.9 m2 g−1). If the data set is restricted to ATN<300 only, no statistically relevant changes are obtained for k and σ using the “Virkkula” equation. Applying the “Drinovec” equation both parameters, k and σ increase when the data set is restricted.The investigation of the different station types yields values of k statistically not different from zero, whereas σ values are different for the respective station types. Highest values of σ = 33.7 m2 g−1 ( ±1.4 m2 g−1) were observed for the remote, followed by σ = 31.8 m2 g−1 ( ±0.5 m2 g−1) for the rural-background station and lowest values of σ = 24.4 m2 g−1 ( ±0.5 m2 g−1) for the urban-traffic stations, possibly reflecting the respective aerosol type.Based on the results for k and σ using either overall values or station type specific values we performed an extensive validation, discussing the application of the transmissometer measurements for a quick, easy and non-destructive quantification of eBC on either fresh sampled or archived filters of different station types. Applying the method for stations not classified as urban, rural or remote a deviation of eBC and EC values up to 23% can be expected. If the station type is known and the respective k and σ values are used this deviation goes down to 10% or lower. This would allow users to obtain long-term time-series going back to times when no on-line instruments were existing, and to obtain these data sets for sites where no on-line instruments are installed even now.