A method to dynamically constrain black carbon aerosol sources with online monitored potassium

Black carbon (BC), primarily emitted from fossil fuel combustion and wood burning, is a significant aerosol component contributing to global warming and negatively impacting air quality and human health. Quantifying BC source contributions is crucial for emission reduction strategies. While several BC source apportionment methods exist, the Aethalometer model offers high temporal resolution and ease of operation, making it widely adopted. This model relies on the absorption Ångström exponent (α) for fossil fuel combustion (α_ff) and wood burning (α_wb) to apportion equivalent BC (eBC) into its sources. However, previous studies using fixed α values are problematic due to α_ff and α_wb variability influenced by factors such as combustion efficiency, aerosol mixing state, size, and chemical composition. Optimal α combinations show spatial heterogeneity, highlighting the need for site-specific determination. Previous studies using radiocarbon (¹⁴C) or levoglucosan (LG) for this purpose have limitations: ¹⁴C suffers from sample contamination and analytical issues, while LG faces degradation during transport and has sources beyond wood burning. Moreover, these methods lack high temporal resolution, failing to capture the dynamic diurnal variations in BC emission sources. This study addresses these limitations by proposing potassium (K⁺), a readily measurable, high-temporal resolution tracer of wood smoke, to dynamically optimize α values in the Aethalometer model, improving the accuracy of hourly BC source apportionment.

Existing BC source apportionment methods include receptor models, radiocarbon methods, marker tracers, air quality modeling, and the Aethalometer method. The Aethalometer method, favored for its high temporal resolution and ease of use, uses aerosol light absorption at two wavelengths to apportion eBC into eBC_ff (fossil fuel combustion) and eBC_wb (wood burning). Its accuracy depends heavily on the α_ff and α_wb values. Previous studies have used fixed α values, but these values vary widely depending on factors like combustion efficiency, mixing state, particle size, and chemical composition. The optimal α_ff and α_wb values have shown significant spatial variability. Studies using ¹⁴C or LG to determine site-specific α values have limitations, including low temporal resolution and uncertainties introduced by sample contamination, degradation, and alternative sources of LG. The use of potassium (K⁺) as a high-temporal resolution tracer for wood burning offers a potential solution to these challenges.

Hourly data on light absorption coefficients (at seven wavelengths), water-soluble ions (including K⁺, Na⁺, Ca²⁺), organic carbon (OC), and elemental carbon (EC) were collected from an urban supersite in Wuhan, China, from March 2018 to February 2019. An Aethalometer (AE31) measured light absorption coefficients, while a thermal-optical transmittance carbon analyzer (Sunset RT-4) measured OC and EC. An online ion chromatography analyzer (MARGA-15) measured water-soluble ions. Meteorological data were also recorded. AE31 data were corrected for loading and multiple scattering effects using a method by Weingartner et al. (2003). The mass absorption cross section (MAC) was determined through linear regression between corrected light absorption and EC mass concentration. eBC was then calculated as the ratio of corrected light absorption to MAC. The absorption Ångström exponent (α_{370_950}) was calculated from the absorption coefficients at seven wavelengths. Potassium from wood burning (K_wb) was calculated by correcting total potassium for contributions from sea salt and soil dust using a previously established equation incorporating Na⁺ and Ca²⁺. The Aethalometer model, which uses α_ff and α_wb to apportion eBC, was applied. Sensitivity analysis involved varying α_ff and α_wb to find the optimal combination that yielded a near-zero intercept in the linear regression between eBC_wb and K_wb, indicating that both originate primarily from wood burning with similar atmospheric removal rates. The Taylor Diagram was used to assess model performance with different α_wb values. Hourly optimal α_ff and α_wb values were subsequently calculated. Finally, uncertainty estimations in eBC_ff and eBC_wb were made by propagating errors from measurements of light absorption, EC, and α values.

The optimal α_ff and α_wb values for the entire dataset were 1.09 and 1.79, respectively. Hourly resolution calculations showed α_ff and α_wb varying in the ranges of 1.02–1.19 and 1.71–1.90, respectively. Using dynamic α values significantly improved the correlation between eBC_wb and K_wb (r = 0.75, p < 0.01) compared to results using fixed (r = 0.42, p = 0.04) and overall optimal (r = 0.35, p = 0.10) α values. The dynamic α values reflected diurnal variations in traffic and wood burning emissions, with higher α values during morning and evening traffic rush hours. Annual BC source apportionment revealed that eBC_ff contributed 77.5 ± 18.9% to total eBC, with highest contribution in summer (90.5 ± 9.81%) and lowest in winter (63.9 ± 17.6%). eBC_wb accounted for 22.5 ± 18.9% of eBC, with highest contribution in winter (36.1 ± 17.6%). eBC_ff positively correlated with ambient temperature and NO₂, while eBC_wb negatively correlated with temperature. Diurnal variations showed two peaks in the morning and evening due to increased traffic emissions and lower levels in the afternoon due to better dispersion conditions. Uncertainties in eBC_ff and eBC_wb were estimated at 28.6% and 56.2%, respectively, through error propagation. Despite the uncertainties, the dynamic method produced more reasonable results than using fixed α values.

The results demonstrate the effectiveness of using hourly monitored potassium to dynamically constrain BC source apportionment in the Aethalometer model. The improved correlation between model-derived eBC_wb and measured K_wb, along with the more realistic representation of diurnal BC source variations, confirm the advantages of this dynamic approach. The higher optimal α_ff value (1.09) suggests that the Aethalometer model-resolved eBC_ff incorporates contributions from both liquid and solid fossil fuel combustion, which is relevant for regions where coal combustion is significant. The study's limitations include the assumption of similar atmospheric removal rates for eBC_wb and K_wb, and the uncertainties associated with error propagation. Future research should focus on further separating fossil fuel combustion sources (liquid vs. solid), improving the accuracy of potassium correction for non-wood burning sources and potentially integrating ¹⁴C and ¹³C measurements to achieve more comprehensive BC source apportionment, especially in regions with diverse emission sources.

This study presents a novel method for dynamically constraining BC sources using hourly measured potassium, significantly improving the accuracy of the Aethalometer model. The findings demonstrate the importance of considering the dynamic nature of BC sources and the need for site-specific and temporally resolved source apportionment. While uncertainties remain, the proposed method represents a significant advancement in BC source identification for air quality, climate, and human health research. Future research should focus on refining the method, improving potassium correction, and integrating other tracers for a more comprehensive understanding of BC sources.

The study acknowledges limitations, including uncertainties associated with error propagation in eBC_ff and eBC_wb estimations (28.6% and 56.2%, respectively). The assumption of similar atmospheric removal rates for K_wb and eBC_wb may influence results. Additionally, the correction for non-wood burning potassium sources may not be entirely complete, potentially impacting the accuracy of K_wb estimations. The Aethalometer model itself is simplified and may not fully capture the complexities of BC emission sources and atmospheric processes.