Measurement report: Vertical profiling of particle size distributions over Lhasa, Tibet: Tethered balloon-based in-situ measurements and source apportionment

In-situ measurements of vertically resolved particle size distributions based on a tethered balloon system were carried out for the first time in the highland city of Lhasa over the Tibetan Plateau in summer 2020, using portable optical counters for the size range of 0.124~32 μm. The vertical structure of 112 aerosol profiles was found to be largely shaped by the evolution of the boundary layer (BL), with a nearly uniform distribution of aerosols within the daytime mixing layer and a 15 sharp decline with the height in the shallow nocturnal boundary layer. During the campaign, the average mass concentration of particulate matters smaller than 2.5 μm in aerodynamic diameter (PM2.5) within the BL was around 3 μg m, almost four times of the amount in the free troposphere (FT), which was rarely affected by surface anthropogenic emissions. Though there was a lower level of particle mass in the residual layer (RL) than in the BL, a similarity in particle mass size distributions (PMSDs) suggested that particles in the RL might be of the same origin as particles in the BL. This was also in consistence 20 with the source apportionment analysis based on the PMSDs. Three distinct modes were observed in the PMSDs for the BL and the RL. One mode was exclusively coarse particles up to roughly 15 μm and peaked around 5 μm. More than 50% of total particle mass was often contributed by coarse mode particles in this area, which was thought to be associated with local dust https://doi.org/10.5194/acp-2021-810 Preprint. Discussion started: 23 December 2021 c © Author(s) 2021. CC BY 4.0 License.


Figure 1 The topographic map of the Lhasa River Valley (a) and the city map of Lhasa (b).
function Q, a key parameter to review the distribution of the components and to estimate the stability of the solution (Brown et al., 2015), the measured matrix was decomposed into factor profiles (F) and factor contributions (G) with non-negativity constraints. The input matrix, comprising PMSDs (22 size bins within 0.124~15 μm) averaged for the ML, the RL and the FT, and the lower 10-m averaged PMSDs in the NBL, was used in 20 random PMF runs (using EPA PMF5.0 software) for 2 to 5 145 factors. In total, there were 48 PMSDs for the ML, 23 PMSDs for the RL and 10 PMSDs for the FT. The components in the PMF runs were separated to keep one signal rather significant from those dominated by noise on account of a criterion of the signal-to-noise ratio (S/N) (Amato et al., 2009). The components with the S/N less than 2 were removed from further analysis, while components with the S/N above 2 were set as "strong" species (Paatero and Hopke, 2003). The stability of converged Q values was found for all runs. By estimating diagnostic errors, the 3-factor result was found to be the optimal solution, with all 150 factors mapped in 100% of the Bootstrap run, no swap of Displacement and no swapping case of Bootstrap-Displacement. The results of the scaled residuals for the 3-factor outcome were within an acceptable range of (-3, 3) (Juntto and Paatero, 1994).
The measured and PMF-simulated total particle mass concentrations were significantly correlated (R 2 =0.90).

An overview of meteorological conditions 155
Under the impact of the Asian Summer Monsoon, Lhasa normally experienced a rainy season from May to September (Ding, 2007;Ran et al., 2014). However, in August 2020 when the campaign took place, two periods with distinctly different conditions of water vapor were found, as could be clearly seen from q near the ground measured in urban Lhasa (Fig. 2a). The sharp reduction in q during the daytime of 17 August was an indicator of changing from a humid Period I (averagely 11.1 g kg -1 ) to a relatively dry Period II (averagely 8.1 g kg -1 ). Period I was also characterized as rainy with the precipitation amounting 160 to 78.5 mm and quite cloudy for most of the time, while there was plentiful sunshine and almost no precipitation during Period II (Fig. 2a). Rainfall mostly occurred in the evening and at night (20:00-08:00), accounting for 84% of the total amount in that month. The two periods differed not much in T, which ranged from 8.8~26.8 ℃ in Period I and from 8.6~25.0 ℃ in Period II with an average of about 17.0 ℃ for both periods. In contrast, RH averaged nearly 62% in Period I, but only 45% in Period II ( Fig. 2b), corresponding well to the difference in q. The temperature shared a similar pattern on each day, except that the 165 maximum value in the daytime apparently lowered on days with an overcast sky. Unlike the diurnal variation of T being a peak in the late afternoon (around 18:00) and a valley in the early morning (around 09:00) without much differences between the two periods, average RH reached a maximum in the early morning of about 79%±11% for period I and 64%±6% for Period II, as well as a minimum in the late afternoon of about 43%±11% for period I and 28%±11% for Period II. Surface winds were largely under 4m s -1 and dominated by easterly and westerly winds (Fig. 2c). 170 Vertical profiles of meteorological parameters observed at the site were categorized into five time periods of the day as give in Table S1, and were presented along H Nor , if either a ML or a NBL could be identified, for a straightforward comparison between the two periods (Fig. S2). The amount of moisture was obviously higher in Period I than in Period II for all time periods, being consistent with what had been revealed from continuous surface measurements. The conservative quantities, θ and q, were generally uniform in the ML for daytime profiles, whereas T decreased and RH increased with increasing height, 175 respectively. For profiles collected during 07:00-08:00 and 20:00-23:00, air temperature inversions in the NBL were recognized for several cases. Averagely, there was a slight increase in θ and decrease in q along with increasing height.
Accordingly, wind speed apparently increased with increasing height. A further examination on the probability distributions of wind speed and wind direction indicated that winds were mainly below 4 m s -1 both within and above the BL throughout the campaign, also being dominated by easterly and westerly winds as near the ground (Fig. S3). Stronger easterly winds 180 exceeding 4 m s -1 occurred above the BL, even up to nearly 7 m s -1 in Period II. It should be kept in mind that data points available for calculating time-period averages at some heights above the BL, corresponding to H Nor larger than 0, might make up only a part of the total number and might thus introduce misleading details. Caution should thereby be taken about statistically drawing conclusions from the characteristics of height-normalized profiles across the BL and at different heights above the top of the BL. The last but not the least, the analysis regarding meteorological parameters, especially winds, was 185 somewhat limited to measurements under a relatively mild condition when the tethered balloon were able to be launched.

Vertical distributions and diurnal variations of aerosol properties
The evolution of the BL was an important influencing factor in shaping the vertical structure of atmospheric aerosols, as being already pointed out by some previous studies (Ferrero et al., 2010;Rant et al., 2016;Zhang et al., 2017). During the campaign, heights of the shallow NBL were largely below 200 m around sunrise/sunset and at night (Fig. S4). In Period I, H m could be 195 determined for most profiles before midday around 14:00, possibly resulting from the development of the ML being suppressed under usually cloudy conditions. In the afternoon, the ML often developed higher than the maximum height reached by the tethered balloon. Unfortunately, intensive observations covering a day from the early morning until the night were unavailable, either because of unobtainable aviation permission or bad weather. Nevertheless, relatively frequent measurements on 12 August (marked by black cross in Fig. S4) revealed, in a more realistic way as compared with what the overall picture could 200 tell, the gradual increase of H m in the overcast morning, from below 100 m before sunrise to about 700 m towards noon. The depth of the ML on that day was expected to be far above the flight limit of 500 m in the afternoon, since it turned to be quite sunny and cloudless overhead after midday. As for Period II, the ML should have developed very quickly after sunrise on those sunny days, thus H m were often already high above the flight limit when aviation permissions were granted in the late morning and the afternoon. 205 Normalized by H m , average vertical profiles of aerosol properties in five time periods of the day were illustrated in Fig. 3, exhibiting a similarity between Period I and II. Profiles of N a and particle mass concentrations around sunrise (07:00-08:00) and since evening (20:00-23:00) resembled an exponential decay in the stable NBL, which favored the accumulation of air pollutants near the ground. In the daytime, profiles were characterized by a nearly uniform distribution of aerosols within the ML, except for profiles collected during 08:00-12:00 in Period II with the characteristics of nighttime profiles. This misleading 210 feature was attributed largely to profiles on religious holidays that were frequently encountered in Period II with the impact of strong emissions from religious burning in the morning, which will be discussed in detail in Sect. 3.3. With the evolution of the ML, stronger dilution effects led to a decline in surface N a and particle mass concentrations in the daytime, compared to that in the early morning and at night. However, surface N a and particle mass concentrations were found to be the highest during 08:00-12:00 in both periods. The reasons for this phenomenon might be enhanced emissions from more anthropogenic 215 activities after sunrise but still weak dispersion inside a ML, where vertical convection had not been fully developed. Above the BL, the number of data points used for averaging varied with H Nor , depending on H m and the maximum height of each flight. To avoid being misled by insufficient data, averaged data were only adopted at heights where the data availability exceeded 75%. An evident reduction in N a and particle mass concentrations above the BL was observed by comparison with that within the BL. For both periods, ratios of PM 1 to PM 2.5 mass concentrations (PM 1 /PM 2.5 ) were relatively uniform within 220 the BL even for NBL-type profiles, despite distinctly different aerosol amount along the normalized height in the NBL. The daytime PM 1 /PM 2.5 ratios were relatively larger than ratios in the early morning and at night. Accordingly, D e was found to be relatively smaller during the day. Table 1 listed particle parameters in the BL, the RL and the FT. For the BL category, averages within the ML and 10-m averages near the ground for the NBL were adopted. Inside the well mixed layer, the surface level and the ML-averaged level 225 were close to each other. Averaged N a within the BL was about 338±162 cm -3 in Period I and 409±201 cm -3 in Period II. The mean PM 2.5 mass concentration within the BL was 2.7±1.7 μg m -3 in Period I, with a fraction of PM 1 in PM 2.5 to be about 81±6%. In Period II, PM 2.5 mass concentrations within the BL averaged 3.6±2.4 μg m -3 , with PM 1 accounting for approximately 73±9%. The lower level of aerosols in Period I might be ascribed to frequent rainfalls and associated efficient removal in the two periods was comparable with each other, a wider range was found in Period II, showing the large variability in particle size distributions among profiles collected at different time.

240
It was found that average PM 2.5 mass concentrations within the BL during this campaign (3.1±2.1 μg m -3 ) were far below what had been observed near the ground in August 2016 in urban Lhasa, where the high levels often exceeded 20 μg m -3 and were sometimes up to more than 40 μg m -3 , with the overall average to be 11±2.2 μg m -3 over the size range of 0.14~3 μm (Cui et al., 2018b). The significant difference of particle mass concentrations between the previous and the current study might reflect different emission strengths surrounding the two sites. Being located in the downtown area and adjacent to several temples, 245 the observational site in the previous study was greatly affected by various strong local emissions, such as traffic, residential and religious activities-related emissions, while the suburban site in this study was under the influence of much weaker nearby emissions and more likely to represent an average condition over this area. Furthermore, the continuous monitoring of PM 2.5 mass concentrations from 2013 to 2017 at 6 sites across the city of Lhasa revealed a slight decrease in 2017 (Yin et al., 2019).
Actually, the location of the site in this study was on the same campus of Tibet University as one site (LHASA-XZ) in Yin et al. (2019). Considering that monthly mean PM 2.5 mass concentration of all 6 sites for August was reported as about 15 μg m -3 and the LHASA-XZ site was not the cleanest one among the others, it might suggest an apparent reduction in PM 2.5 mass concentrations with the continuing implementation of air quality policies. Profiles in Period II were mainly collected in the early morning and at night, when elevated aerosol layers were often encountered, though it was hard to tell whether those were only residual layers that could easily form above the NBL after sunset and remain until the BL evolved to an enough height in the next day, or there could also be contributions from

Vertically resolved particle mass size distributions
Particle mass size distributions (PMSD) measured by the POPS were categorized into two periods and averaged for three layers (Fig. 4). In general, size-resolved particle mass concentrations were the highest at all diameters within the BL, while much lower in the FT. Nevertheless, a similarity was found among all PMSDs. Plainly, a distinct mode below around 0.23 μm existed for all layers during both periods, though its relative strength to the rest part of the PMSD considerably differed among 285 different layers and periods. For this mode, more particles were observed towards the lower end of the size range in Period II, possibly implying stronger secondary formation of fine particles favored by the fine weather during that period. Particles larger than 1 μm apparently increased with the diameter for all PMSDs, suggesting another mode in coarse particles. The normalized averages of PMSDs that were merged from the POPS and GRIMM 11-C measurements over the size range of 0.124~32 μm, which were only available for 48 profiles in Period II, did exhibit a mode peaking around 2.5 μm in the FT and around 5 μm 290 in the other two layers (Fig. S5). Particles as large as roughly 15 μm were observed within the BL and in the RL. In contrast, particles with the diameter larger than 5 μm accounted only for a negligible part of the PMSD in the FT. It was also noted, from both the average PMSD (Fig. 4) and the normalized average PMSD (Fig. S5), that there was a third mode ranging over 0.5~0.7 μm within the BL and in the RL, though relatively less pronounced for the later one, whereas no such an apparent mode was observed in the FT. These features revealed that under most circumstances particles in the RL were of the same 295 origin as particles in the BL, whereas particles in the FT should be rarely affected by local anthropogenic emissions near the ground.

300
The vertical distribution of PMSDs within the size range of 0.124~32 μm for each profile in Period II was explored in more detail. It was noteworthy that the mode over 0.5~0.7 μm was found to be more distinct on the mornings of religious holidays compared with that at other times. In order to conveniently make comparisons among different cases, the dataset was grouped into subsets of data collected on non-holidays and holidays. In total, there were 22 aerosol profiles collected on non-holidays and 26 on the four religious holidays (19,20,26,28 August). The subset of data from holidays was further divided into five 305 cases, four of which respectively comprised measurements on the morning of each religious holiday. One case, consisting of data collected on holidays except in the morning, was denoted as holidays* to distinguish from the category named holidays that covered all profiles measured on religious holidays. The average profile of D e,<1μm on non-holidays and holidays* were found to be close to each other, with an average of roughly 0.21±0.01 μm and 0.23±0.02 μm through the vertical direction, respectively (Fig. 5a). However, the average vertical distribution of D e,<1μm on the morning of each religious holiday, as separately displayed in different colors, was significantly different from the categories of non-holidays and holidays*. In the BL, a remarkably larger D e,<1μm was observed in the morning on all religious holidays than that on both non-holidays and holidays*. The largest D e,<1μm was nearly 0.35 μm on 19 August, the Shakyamuni Buddha Day and also the beginning of Sho Dun Festival, one of the most ceremonious traditional festival in Tibet. This could plausibly explain the aforementioned larger range of D e,<1μm averaged within the BL in period II than in Period I, since emissions from religious activities on religious 315 holidays enlarged the effective diameter and only one religious holiday (8 August) was encountered in Period I. Above the BL, D e,<1μm for the four holiday morning cases generally fell in the range of D e,<1μm on non-holidays and holidays, except that D e,<1μm on the morning of 19 August was apparently larger than others. As for coarse particles in the size range of 1~10 μm, no obvious difference was observed between non-holidays and holidays (Fig. 5b). A further examination was taken on average PMSDs within 50 m above the ground for each category, considering that in the 325 NBL or in an early morning ML measurements near the ground would better help elucidate the impact religious activities brought to PMSDs. A marked mode peaking around 0.6 μm and amounting to about 18 μg m -3 was observed in the PMSDs on 19 and 28 August (Fig. 6). Though much lower around 5 μg m -3 on 20 and 26, the peak of this mode was still almost double of that on non-holidays and holidays*. The existence of the accumulation mode over 0.5~0.7 μm on religious holidays was consistent with the findings from surface aerosol measurements in August 2016, and was demonstrated to be characteristic of 330 emissions from incense burning and biomass burning for religious ceremonies (Cui et al., 2018b and references therein). As a widespread religious custom, incense burning was commonly performed in the daily life of local people, not only in temples but also at home. The burning of several types of wood branches and herbs as well as butter lamps in temples was also traditional religious activities. The PMSDs clearly showed that these emissions were largely enhanced on holidays especially in the mornings and formed a much strengthened accumulation mode. 335

Potential sources of particles in different layers
Three source factors were identified from the PMF analysis of the PMSDs combined from the POPS and GRIMM 11-C measurements in Period II (Fig. 7a). The first factor (factor 1) revealed a potential source that predominantly contributed to 340 particles smaller than 0.3 μm, probably being associated with local emissions from fossil fuel combustions and/or secondary aerosol formation. The second factor (factor 2) showed a broad peak over about 0.3~0.7 μm, with also a considerable contribution from particles with the diameter extending from 0.7 μm to around 2.5 μm. Considering the apparent enhancement of particle mass in the size range of 0.5~0.7 μm on religious holiday mornings, factor 2 was taken to be mainly representative of aerosols released from religious activities such as burning incense, cypress branches, herbs, and butter lamp (Cui et al.,345 2018b and references therein). The third factor (factor 3) was almost exclusively composed of particles in the coarse mode (> 1 μm). This factor might be attributed to suspended dust particles from unpaved roads, construction sites, and the surrounding mountains. Indoor particles previously collected at a temple in Lhasa exhibited a bimodal particle mass distribution, with one peak around 0.4~0.7 μm and one peak in the coarse mode around 5 μm (Cui et al., 2018a). It was speculated that factor 3 might also involve a certain contribution from religious burning and/or residential biomass burning. However, concentrations 350 contributed by factor 3 did not increase as expected with the rising concentrations contributed by factor 2 on holiday mornings, when assuming that coarse mode particles observed at the site were partly generated from religious activities (Fig. S6). Also residential burning of biomass including cow dung and plants, as an important source in the past for energy, was negligible nowadays and mostly replaced by a mixture of fossil fuel and renewable energy sources such as solar, wind, geothermal and hydroelectric power.