Submicron aerosol online measurements of optical and chemical properties were performed at the Lijiang Astronomical Station, Chinese Academy of Sciences, Gaomeigu, Yunnan Province (26∘41′ N, 100∘1′ E; 3260 m a.s.l.) (Fig. 1). Continuous hourly O3 and relative humidity (RH) were measured with the use of an ozone analyzer (EC9810, Ecotech Pty Ltd., Australia) and an automatic weather station (MAWS201, Vaisala, Helsinki, Finland). The monitoring station is situated at the southeastern edge of the TP, a natural channel for the transport of air pollutants from Southeast Asia to the TP (Tan et al., 2021). All instruments were placed on the rooftop of an office building (∼ 10 m above ground level (a.g.l.)), without any strong anthropogenic emission sources nearby. More detailed description about the sampling site can be found in Wang et al. (2019) and Liu et al. (2021). The sampling period lasted from 08:00 LST (local standard time; all time references that follow are given in LST) on 14 March 2018 to 23:00 on 31 March 2018, corresponding to the pre-monsoon season in the TP.

A newly developed Aethalometer (model AE33, Magee Scientific, Berkeley, CA, USA) was used to measure the aerosol light absorption coefficient (babs) at multiple wavelengths (i.e., 370, 470, 520, 590, 660, and 880 nm) with a 1 min time resolution. Briefly, the ambient air sampled at a flow rate of 5 L min-1 was firstly selected by a PM1 (particulate matter with an aerodynamic diameter ≤1.0 µm) cyclone separator (BGI SCC 1.197, Mesa Labs, USA) to collect submicron aerosol on the filter. Light at different wavelengths emitted from diodes irradiated two parallel filter spots with deposition rates of 3.85 and 1.15 L min-1, respectively. Thereafter, the two light attenuations measured by optical detectors were used to calculate babs through a real-time loading effect compensation algorithm. This “dual-spot” technique for the model AE33 can eliminate the nonlinearity effect caused by an increasing amount of aerosol deposit on the filter (Weingartner et al., 2003). Additionally, the model AE33 automatically used a factor of 2.14 to compensate for the scattering effect of the quartz filter. Detailed operating principles of the model AE33 can be found in Drinovec et al. (2015).

OA in the non-refractory PM1 was measured using a quadrupole aerosol chemical speciation monitor (Q-ACSM, Aerodyne Research Inc., Billerica, Massachusetts, USA) with a 30 min time resolution. The aerodynamic lens coupled with a 100 µm diameter critical aperture in the Q-ACSM created a beam of focused submicron aerosols (40–1000 nm aerodynamic diameter), which was vaporized at ∼ 600 ∘C, ionized by a 70 eV electron impact, and subsequently characterized with a mass spectrometer. The details of the instrument have been described elsewhere (Ng et al., 2011b). The measured Q-ACSM data were processed by the ACSM local tool version 1.5.3.5 compiled with Igor Pro 6.37 (WaveMetrics, Inc., Lake Oswego, OR, USA) to determine the mass concentration and ion-speciated mass spectra of OA. For our study, the default collection efficiency (0.45) and relative ionization efficiency (1.4) were used to obtain OA mass concentration (Middlebrook et al., 2012). The mass concentration and error matrices of organic fragments from mass-to-charge (m/z) 12 to 120 were initialized following the method of Allan et al. (2003).

A geographic-information-system-based software TrajStat was utilized to investigate the influences of regional transport on BrC absorption at Gaomeigu from 14 to 31 March 2018 (Wang et al., 2009). The trajectories were calculated with the HYSPLIT model developed by the National Oceanic and Atmospheric Administration (Draxler and Hess, 1998). In this study, the model was set to run 24 times per day at starting times of 00:00–23:00 with a 1 h step. Based on the gridded meteorological data from the Global Data Assimilation System (ftp://arlftp.arlhq.noaa.gov/pub/archives/gdas1, last access: 1 June 2022), 72 h backward trajectories at the height of 500 m a.g.l. at Gaomeigu during the sampling period were produced.

The concentration-weighted trajectory (CWT) method was further used to identify the potential source regions that likely affected the BrC absorption at Gaomeigu (Hsu et al., 2003). The geographic zone covered by the total number of backward trajectories (K) was divided into I×J grid cells with a resolution of 0.5∘ × 0.5∘. The CWT value of each grid can be defined as follows: 7babs-BrC-ij=∑k=1Kbabs-BrC-kτijk∑k=1KτijkWij. Here, babs-BrC-ij is the average-weighted light absorption coefficient of BrC in the ijth cell, babs-BrC-k is the light absorption coefficient of BrC observed on the arrival of trajectory k, and τijk is the time spent in the ijth cell by trajectory k. The weighting function of Wij was applied to reduce the effect of the small number of back-trajectory segment endpoints that fall into the grid cell (Wang et al., 2006): 8Wij=1.00135<nij0.7045<nij≤1350.4215<nij≤450.17nij≤15. Here, nij is the total number of endpoints in the ijth cell. In this study, the total number of endpoints located in 72 cells of the geographic zone is 3268, so the average number of endpoints in all cells is about 45.

The concept “simple forcing efficiency” (SFE) introduced by Bond and Bergstrom (2006) is a useful way to evaluate the radiative effect of atmospheric aerosols. Without consideration of aerosol scattering, a variant of wavelength-dependent SFE is given as follows: 9SFEiλ=S0λ4×τatm2×1-Fc×4as×MACiλ. Here, the subscript i represents BBOA, po-OOA, or BC; λ denotes the wavelength from 370 to 660 nm with a 1 nm step; SFE is given in watts per gram (W g-1), which represents the positive energy added to the Earth atmosphere system by a given mass of light-absorbing particles in the atmosphere; S0 is the solar irradiance based on the ASTM G173-03 reference spectra in watts per square meters (W m-2); τatm, Fc, and as are the atmospheric transmission (0.79), the cloud fraction (0.6), and the surface albedo (0.19), respectively, which are constants from the global average calculations; MAC with a 1 nm resolution is extrapolated using Eq. (6). And then, the fraction of solar radiation absorbed by OA component relative to BC (fOA/BC) can be calculated as 10fOA/BC=∑λ=370660SFEOAλ×COA∑λ=370660SFEBCλ×CBC. Here, the integrated SFE is the sum of the SFE from 370 to 660 nm; COA and CBC are the average mass concentrations of the OA component and BC during the sampling period.

The temporal variation in submicron aerosol babs at wavelengths from 370 to 880 nm as well as the OA mass concentration during the entire campaign at Gaomeigu is depicted in Fig. 2. The hourly babs values at different wavelengths varied from minimum to maximum values by factors of 19–41 from 14 to 31 March 2018, reflecting the fact that atmospheric environment at Gaomeigu is influenced by dynamic changes in emission sources and meteorological conditions. Particularly, the larger variations in babs values at 370–660 nm, compared with those at 880 nm, highlighted the impact of non-BC light-absorbing materials. As shown in Table 1, the average babs values were 33.1 ± 24.4 Mm-1 (arithmetic mean ± standard deviation) at 370 nm, 26.7 ± 19.7 Mm-1 at 470 nm, 20.3 ± 13.9 Mm-1 at 520 nm, 18.2 ± 12.5 Mm-1 at 590 nm, 13.7 ± 9.0 Mm-1 at 660 nm, and 8.0 ± 4.9 Mm-1 at 880 nm. The babs values obtained in this study were comparable with those reported previously at sampling sites of the TP, where the major anthropogenic sources (i.e., industry, fossil fuel combustion, etc.) are limited locally (Niu et al., 2018; Zhao et al., 2019; Zhu et al., 2021). Frequency histograms of hourly AAE values showed a median AAE value of 1.59 with an interquartile range from 1.38 to 1.83 (Fig. S1). Over 72 % of AAE values were higher than 1.4 (upper limit of AAEBC), implying the presence of both BrC and BC absorption in the submicron aerosol at Gaomeigu.

Based on Eqs. (1) and (2), babs-BrC and babs-BC were separated from the total absorption using AAEBC=1.1. The average babs-BrC values during the campaign were 12.3 ± 13.8, 10.7 ± 11.5, 6.0 ± 6.0, 5.8 ± 5.8, and 2.7 ± 2.6 Mm-1 at 370, 470, 520, 590, and 660 nm, respectively (Table 1). Figure 3 shows fractions of light absorption at specific wavelengths by BrC and BC in the submicron aerosol. BrC contributed substantially to babs, which accounted for 20.0 %–40.2 % from 370 to 660 nm. The average contributions of babs-BrC to babs in the near-UV and blue spectral ranges (300–500 nm) were higher than those in other visible ranges (520–880 nm), indicating that BrC was a considerable absorbing material at short wavelengths in the atmosphere. It should be noted that the assumption for AAEBC would lead to biases in the BrC absorption calculation (Lack and Langridge, 2013). The uncertainties in AAEBC (±0.3) in this study caused variations of 14.3 %–16.6 %, 10.2 %–12.2 %, 10.3 %–11.5 %, 7.7 %–8.6 %, and 6.6 %–7.1 % in the estimation of BrC absorption contributions at 370, 470, 520, 590, and 660 nm, respectively (Fig. 3).

With respect to the relationship between BrC absorption and OA components, babs-BrC values at 370–660 nm were significantly correlated with OA mass concentrations (r=0.64 to 0.70, p<0.01) (Fig. S2), confirming a strong link between BrC chromophores and OA in the southeastern margin of TP (Lack et al., 2013; Laskin et al., 2015).

HERM analysis identified two distinct OA sources, consisting of BBOA and po-OOA. Each of the OA sources had unique characteristics in terms of the mass spectrum, temporary variation, and atmospheric processes. The detailed source apportionment results of OA are shown in Fig. 4.

The mass spectrum of BBOA resembled that of BBOA obtained in previous studies (uncentered correlation coefficients of 0.80–0.87) (Crippa et al., 2013; Ng et al., 2011a; Wang et al., 2017). It was characterized by a prominent peak of m/z 60, and a strong positive correlation was found between BBOA and m/z 60 concentrations (r=0.72, p<0.01) (Figs. 4a and b). We have known that m/z 60 is a typical molecular fragment of levoglucosan, mannosan, and galactosan, which are good organic tracers of biomass burning (Kalogridis et al., 2018; Kim et al., 2017; Reyes-Villegas et al., 2018). The fraction of m/z 60 in the BBOA mass spectrum (f60, 0.9 %) was higher than 0.3 % (background level in absence of biomass burning), suggesting the impact of biomass burning at Gaomeigu (Cubison et al., 2011). Scatterplots of f44 vs. f60 were used to analyze the aging degree of BBOA in the atmosphere (Fig. 4c). f60 usually decreases from fresh to aged biomass burning emissions because of degradation and oxidation reactions during the atmospheric aging, while f44 increases (Paglione et al., 2020). The f60 and f44 of BBOA resolved in this study (0.9 % and 6.3 %, respectively) indicate BBOA was less aged, possibly caused by long-distance regional transport. This is further demonstrated in our trajectory-related analysis (Sect. 3.3).

Another OA source was featured by the high correlation with m/z 44 (r=0.97, p<0.01), which is a surrogate of the oxidation degree (Aiken et al., 2008). The most abundant peak in mass spectrum of po-OOA was at m/z 44 (f44, 27.8 %), which is similar to those in mass spectra of more-oxidized oxygenated OA (MO-OOA) (f44>20 %) identified frequently in previous studies (Tobler et al., 2021; Xu et al., 2018; Zhang et al., 2019). This indicated that this OA source was likely related to extensive secondary processes occurring during transport (Wang et al., 2017; Xu et al., 2017). Figure 4d shows that both po-OOA mass concentration and its fraction in OA increased with increasing O3 (R2=0.79 to 0.87); however, neither of them correlated with RH (Fig. S3). These results support the possibility that high O3 was the driving factor of po-OOA formation; thus the term of po-OOA was introduced in this study to stress the importance of the photochemical-oxidation process in the TP. Moreover, the temporal variation in mass concentration in po-OOA moderately correlated with that in BBOA (r=0.63, p<0.01), indicating that a portion of po-OOA could be derived from oxidation of volatile organic precursors from biomass burning (Bruns et al., 2016; Posner et al., 2018).

To further quantify the contributions of OA sources to BrC light absorption, a MLR model was applied to retrieve MAC values from BBOA and po-OOA. As shown in Fig. 5a, the wavelength dependences of MACBBOA and MACpo-OOA were generally described by the power-law relationship (R2=0.84 to 0.87), and the AAEs of BrC values for BBOA and po-OOA were greater than 2.0 (Kirchstetter et al., 2004). The MACBBOA was 2.78 ± 0.39 m2 g-1 at 370 nm and dropped to 0.61 ± 0.07 m2 g-1 at 660 nm. Taking the near-UV wavelength as the representative for discussion, the MACBBOA obtained in this study was within that range from biomass burning (0.9–7.7 m2 g-1) reported by laboratory experiments and field measurements studies (Kaskaoutis et al., 2021; Kumar et al., 2018; Lack et al., 2012; Qin et al., 2018; Wang et al., 2020; Washenfelder et al., 2015). The differences in light absorption capacity of OA from biomass burning may be partly associated with biomass types and combustion efficiencies (Budisulistiorini et al., 2017; Martinsson et al., 2015; Tian et al., 2019). In addition, the photobleaching effect of aerosol at different aging degrees can also lead to the variation in MACBBOA (Sumlin et al., 2017; Zhong and Jang, 2014). The MACpo-OOA was 1.43 ± 0.23 m2 g-1 at 370 nm, 1.23 ± 0.20 m2 g-1 at 470 nm, 0.68 ± 0.10 m2 g-1 at 520 nm, 0.66 ± 0.10 m2 g-1 at 590 nm, and 0.30 ± 0.04 m2 g-1 at 660 nm. Compared with BBOA, more oxygenated po-OOA was possibly more photochemically bleached, which resulted in the lower MAC (Lee et al., 2014).

The MLR model reasonably reconstructed the temporal variation in the measured babs-BrC values, with the index of agreement ranging from 0.73 to 0.79 for different wavelengths. The total reconstructed light absorption coefficient of BrC (rbabs-BrC) was the sum of babs-BrC from BBOA and po-OOA obtained from the MLR model. As shown in Fig. 5b, the average babs-BrC values from BBOA were 6.0 ± 3.6, 5.0 ± 3.0, 2.8 ± 1.7, 2.6 ± 1.6, and 1.3 ± 0.8 Mm-1 at 370, 470, 520, 590, and 660 nm, respectively. The babs-BrC from BBOA contributed about half (48.7 %–51.1 %) of rbabs-BrC at 370–660 nm, indicating that biomass burning is an important primary source of BrC absorption at Gaomeigu. This was possibly related to transport of pollutants emitted from South Asia and Southeast Asia during the pre-monsoon season, where biomass burning activities are intensive (M. Zhang et al., 2020; Zhang et al., 2015). The po-OOA produced comparable babs-BrC values (6.0 ± 4.2, 5.1 ± 3.6, 2.8 ± 2.0, 2.8 ± 1.9, and 1.3 ± 0.9 Mm-1 at 370, 470, 520, 590, and 660 nm, respectively), suggesting the critical role of photochemical-oxidation processes in the BrC absorption at Gaomeigu. Four periods are marked in Fig. 2, characterized by the inclusion of obvious rising stages for both OA concentrations and babs values. The babs-BrC at 370 nm from BBOA contributed the most to rbabs-BrC (63.4 %) during period I, while the contribution of babs-BrC from po-OOA increased significantly during other periods (52.2 %–64.7 %) (Fig. S4). The rapid increases in babs-BrC from po-OOA were likely caused by photochemical-oxidation processes that were favored by relatively high O3 conditions (75–84 ppb) during periods II–IV; for comparison, the O3 mixing ratio during period I was 52 ppb. We noted that the largest rbabs-BrC occurred in period II when both primary source emissions and secondary formation were strong. These results further highlighted the importance of biomass burning and photochemical oxidation in light absorption of BrC at Gaomeigu.

The air-mass trajectory and CWT analyses were used to identify whether local emission or regionally transported air pollution was the major source of babs-BrC from OA components at Gaomeigu. Figure 6a shows the 72 h backward trajectories of the receptor site during the sampling period, and all those originated from Myanmar. The percentage of the trajectories with high OA concentration (>5.3 µg m-3, the median value of hourly OA concentration) exceeds 50 %. From the CWT maps of babs-BrC at 370 nm, the spatial distributions of the potential source for babs-BrC from BBOA and po-OOA were similar (Fig. 6b and c). The source regions with the highest CWT values were located in northern Myanmar and along the China–Myanmar border, while the CWT values in the areas surrounding Gaomeigu were relatively low. This indicates that large babs-BrC loadings at Gaomeigu were more likely the result of strong cross-border transport of BrC from biomass burning and secondary formation than local emission during the pre-monsoon season.

As described in Sect. 2.5, a simple model was used to provide a first-order estimate of the radiative effect of light-absorbing particles. Figure 7a shows SFEs of BBOA, po-OOA, and BC at wavelengths from 370 to 660 nm. The SFE peaks of BBOA, po-OOA, and BC were observed at the boundary between the UV and blue spectra (i.e., ∼ 450 nm), which were mainly caused by the high MAC and strong solar irradiance at specific wavelengths. The integrated SFE of BBOA over the entire solar spectra (370–660 nm) in this study was 24.2 W g-1, comparable to that of primary OA (21 W g-1, 300–1000 nm) reported in Lu et al. (2015). The integrated SFE of po-OOA (12.5 W g-1) was only half of that of BBOA due to the relatively low MAC. BC had a much higher integrated SFE (226.6 W g-1) compared with OAs. This is consistent with the widely acknowledged view that BC is the strongest and most important light-absorbing particle in the atmosphere (Bahadur et al., 2012; Bond et al., 2013). As the concentration of OA in the atmosphere is generally greater than that of BC, the importance of the BrC radiative effect was further evaluated by calculating the fraction of solar radiation absorbed by OA relative to BC. The fractional radiative forcing by two OAs relative to BC was as high as 48.8 ± 15.5 %, in which the radiative forcing of po-OOA relative to BC (24.2 ± 13.2 %) was almost equal that of BBOA to BC (24.6 ± 9.1 %) (Fig. 7b). These results suggest that BrC emitted from biomass burning and formed by photochemical oxidation is an efficient radiative-forcing agent, which, along with BC, can remarkably disturb the radiative balance over the TP. Thus, the inclusion of BrC in the climate models will provide a better understanding of climate change in the southeastern TP. It should also be noted that although BBOA and po-OOA had similar radiative effects, effective measures on tackling the impact of BrC are to reduce primary emissions and volatile organic precursors of BrC from biomass burning in the future, since the intense photochemical environment is an inherent feature of the TP.