Cloud droplet activation 1 in a continental Central European urban environment

. Collocated measurements by condensation particle counter, differential mobility particle sizer 7 and cloud condensational nuclei counter instruments were realised in parallel in central Budapest from 8 15 April 2019 to 14 April 2020 to gain insight into the droplet activation behaviour of urban aerosol 9 particles. The median total particle number concentration was 10.1×10 3 cm –3 . The median concentrations 10 of cloud condensation nuclei (CCN) at water vapour supersaturations ( S s) of 0.1, 0.2, 0.3, 0.5 and 1.0 % 11 were 0.59, 1.09, 1.39, 1.80 and 2.5×10 3 cm –3 , respectively. They represented from 7 to 27 % of the total 12 particles. The effective critical dry particle diameters ( d c,eff ) were derived utilising the CCN 13 concentrations and particle number size distributions. Their medians were 207, 149, 126, 105 and 80 nm, 14 respectively. They were all positioned within the accumulation mode of the typical particle number size 15 distribution. Their frequency distributions revealed a single peak, which geometric standard deviation 16 increased monotonically with S . The broadening indicated larger time variability in the activation 17 properties of smaller particles. The frequency distributions also showed a fine structure. Its several 18 compositional elements seemed to change in a tendentious manner with S . They were related to the size- 19 dependent chemical composition and external mixtures of particles. The relationships between the critical 20 S and d c,eff suggested that the urban aerosol particles in Budapest with a diameter larger than 21 approximately 130 nm showed similar hygroscopicity than the continental aerosol in general, while the 22 smaller particles appeared to be less hygroscopic than that. Seasonal cycling of the CCN concentrations 23 and activation fractions implied modest alterations and for the larger S s only. They likely reflected the changes in particle number concentrations, chemical composition and mixing state of particles. The seasonal dependencies for d c,eff were featureless, which indicated that the urban particles exhibited more 26 or less similar droplet activation properties over the measurement year. This is different from non-urban 27 locations. The hygroscopicity parameters (  values) were computed without determining time-dependent 28 chemical composition of particles. Their medians were 0.16, 0.10, 0.07, 0.04 and 0.02, respectively. The 29 averages suggested that the larger particles exhibited considerably higher hygroscopicity than the smaller 30 particles. The urban aerosol was characterised by substantially smaller  values than for regional or 31 remote locations. All these could be virtually linked to specific source composition in cities. The relatively 32 large variability in the hygroscopicity parameter sets for a given S emphasized that their individual values 33 represented the CCN population in the ambient air, while the averages stood mainly for the particles with 34 a size close to the effective critical dry of , 90, 210, 180 respectively for S s of functioning within the they for some if it on data were in the the data their S of CCN it differences may 476 not necessarily show up in the particle number size distributions, while they can lead to diverse activation 477 properties. Several compositional elements of the fine structure (e.g. the maximum or the relative peak 478 areas) changed in a tendentious manner by S . Their exact identification and interpretation are beyond the 479 objectives of the present paper. They are to be included into an upcoming study which is to c,eff . 621 4 Conclusions 622 The concentrations of CCN at various S s and particle number distributions were measured in parallel 623 with each other in a continental Central European urban environment 1 year. The effective cloud 624 droplet activation properties of the aerosol population determined from the available experimental 625 data without measuring time-resolved chemical

instruments include a differential mobility particle sizer (DMPS) and a condensation particle counter 72 (CPC). They were complemented by a continuous-flow cloud condensation nuclei counter (CCNc) in 73 2018. The combinations of the long-term particle number size distributions, total particle number 74 concentrations and CCN data at various Ss facilitate the utilisation of special methods for the data 75 validation and of further joint evaluation procedures.  The CPC instrument deployed (TSI, model 3752, USA) was operated with an aerosol inlet flow of 1.5 L 106 min -1 , and recorded concentrations of particles with a diameter above 4 nm using n-butanol as a working 107 fluid. Its sampling inlet was made of stainless-steel tube with a diameter of ¼ inch (6.35 mm) and length 108 of ca. 1.6 m. Mean particle number concentrations (NCPC) with a time resolution of 1 min were extracted 109 from its extended data base. The nominal specification of the CPC warrants an agreement in 110 concentrations better than ±10 % between two identical instruments operating in the single-particle 111 counting mode with a data averaging interval of >30 s. 112 113 The DMPS system utilised was a laboratory-made flow-switching-type device (University of Helsinki, 114 Finland). It measured particle number concentrations in an electrical mobility diameter range from 6 to 115 1000 nm in the dry state of particles (with a relative humidity of RH<30 %) in 30 channels with a time 116 resolution of 8 min (Salma et al., 2011(Salma et al., , 2016. Its main components included a radioactive ( 60 Ni) bipolar 117 diffusion charger, a monotube Nafion semi-permeable membrane dryer, a 28-cm long Vienna-type   The total air flow rates were set to 500 cm 3 min -1 and the ratio of the sample flow rate to the sheath flow 141 rate was 1:10. The selected Ss were 0.1, 0.2, 0.3, 0.5 and 1.0 % stepping from the lowest to the highest 142 values within a measuring cycle with duration times of 12, 5, 5, 5 and 5 min, respectively. The data 143 measured by the system were recorded every 1 s. The CCN concentrations at a given S (NCCN,S) obtained 144 by the two chambers should not differ by more than 15 %. The system was run in polydisperse operation 145 mode largely according to the ACTRIS standard operation protocol (Gysel and Stratmann, 2013).

152
The measured DMPS data were inverted into discrete size distributions, which were utilised to calculate 153 particle number concentrations in the diameter ranges from 6 to 25 nm (N6-25), from 6 to 100 nm (N6-100), 154 from 30 to 1000 nm (N30-1000) and from 6 to 1000 nm (N6-1000). The intervals were selected to represent The N6-1000 data from the DMPS system were compared to the CPC concentrations, which were averaged 178 over the corresponding DMPS measuring cycle. Due to the differences in the lowest measurable diameters 179 (6 vs. 4 nm, respectively), an agreement between the two instruments can be expected if the contribution 180 of nucleation-mode particles to total number of particles is negligible. Additional factor such as larger 181 particle transport losses along their longer path in the DMPS system and possibly different response times 182 of the two CPCs involved in the instruments could also add (Salma et al., 2016). The comparison was 183 realised by evaluating the NCPC/N6-1000 ratio as a function of the N6-30/N6-1000 ratio. The intercept of their 184 regression line was considered as the correction factor for the DMPS system (Sect. 3.1).

186
The CCN concentrations at the S of 1.0 % were compared to the particle numbers. If most particles 187 activate at this S then the two concentrations are expected to be similar. In a previous survey, certain 188 criteria were set to exclude the time intervals when very small, hence, non-activating particles were 189 present in larger concentrations (Schmale at el., 2017). The comparison was performed under the 190 conditions when the concentration ratio of particles <30 nm to the total particles was <10 % or between 191 10 and 20 %. These criteria were utilised for remote or regional locations. The conditions seem not to be 192 applicable for urban data sets since the annual mean and standard deviation (SD) of the N30-1000/N6-1000 193 ratio in Budapest were (52±15) %, and the relative number of DMPS data fulfilling the criterium 1 or 2 194 above were only 2 % on a yearly scale. This is due to relatively large and persistent contributions of high-195 temperature emission sources of particles typically present in cities.

197
We propose here another criterium for urban or polluted environments so that the NCCN,1.0% data are  For atmospheric aerosol, the activated fraction of particles tends to increase gradually with the dry particle activation diameter, which is called effective critical dry particle diameter (dc,eff) is defined in these cases 209 as the size at which 50 % of the dry particles activate at a given S (Rose et al., 2008(Rose et al., , 2010. (1) 214 where dmax is the largest dry particle diameter measured by the sizing instrument (here DMPS) and Ni is 215 the number of particles in the size channel i of the instrument. Hence, the concentrations were summed 216 from the largest particle size (dmax) towards the smaller diameters until the measured CCN concentration 217 was obtained. In order to estimate the dc,eff with higher accuracy, a logarithmic interpolation was 218 accomplished between the last 2 diameters of the summation. The size distribution spectrum which date 219 and time was the closest (with 20 min) and smaller than or equal to that of the CCN concentration was 220 considered.

222
It has to be noted that the assumption of internally mixed particles is rarely met in urban environments shapes of the dry solute particle and solution droplet as:

238
where: The , Mw, w, R and T are the hygroscopicity parameter, molar mass of water (0.018015 kg mol -1 ), 241 density of water, the universal gas constant (8.3145 J mol -1 K -1 ) and absolute temperature of the droplet The d/a was assumed to be that of pure water. Some organic chemical species in atmospheric aerosol  The time resolution of all modelled data was 32 min, which resulted in the total counts of data typically 273 around 13.6×10 3 at each S level. 3 Results and discussion 275 The relevant meteorological properties are summarised in Table S1   The N30-1000,>70%/NCCN,1.0% ratio as function of N30-1000,>70% (Fig. 1c) suggested that the ratio was slightly 299 increasing with the concentration, mainly above 10 4 cm -3 . An agreement between the N30-1000,>70% and    The mean activation fractions (AF=NCCN,S/N6-1000) of the particles increased monotonically from 7 to 27 388 % with S and showed a levelling off character (Fig. 2). The maximum value was considerably smaller 389 than for remote or regional locations (Sihto et al, 2011;Paramonov et al., 2015). The shape of the AF 390 curve was similar to that for the CCN concentrations. This is typical for non-coastal locations, where a 391 multicomponent mixture of particle sources yield more-or-less balanced and, therefore, similar curves

Effective critical dry particle diameters 412
Basic statistical measures of the effective critical dry particle diameters at different Ss over the whole 413 measurement year are displayed in Table 3. The median dc,eff decreased from 207 to 80 nm with S. All 414 diameters were positioned within the accumulation mode of the median particle number size distribution 415 (Fig. S1). The monthly mean number median mobility diameters for the Aitken and accumulation modes   This confirmed that the water activation properties depend on the aerosol type. Our data were comparable  The dependency also pointed to the size-dependent chemical composition, which is typical for urban The peaks exhibited a fine structure. They seemed to contain submodes. This is likely related to the  to be more advantageous for investigating the possible seasonal cycling (Fig. 5). The months were  The average  values were considerably smaller than for regional or remote locations (Paramonov et al., The range of the  values was increasing with S, and, more importantly, it became particularly large (a 587 factor of ca. 10 3 for 1.0 %) even when compared with aerosol properties typically driven by atmospheric 588 dynamics. This can be illustrated by the relationships between the  value and dc,eff for each S (Fig. 6).  Figure 6. Relationship of the hygroscopicity parameter and the effective critical dry particle diameter (dc,eff) derived at specialities. The average CCN concentrations were substantially larger, while the average effective 627 critical dry particle diameters and activation fractions were considerably smaller than for non-urban sites.

628
The particles with a diameter ca. <130 nm already showed lower hygroscopicity than in general, and it 629 was decreasing even more with lower size. The seasonal dependencies of the derived properties were not 630 pronounced or obvious and could not be explained solely by variations in the particle number After gaining experience with operation and calibration of the dual-chamber CCNc measurement system, 643 we plan to extend one of its chambers by a DMA and CPC setup so we can perform both polydisperse 644 and monodisperse measurements in parallel, which is expected to supply further valuable knowledge on 645 the mixing states of particles. This is especially important since urban aerosol particles typically comprise 646 externally mixed carbonaceous particles with very distinctive hygroscopic properties. This seems to be 647 relevant in general and could also support or facilitate the association of the hygroscopicity parameters to 648 major source types in cities together with multistatistical apportionment methods.

649
Data availability. The observational data are available from the corresponding author. and chemical aerosol properties on warm cloud droplet activation, Atmos. Chem. Phys., 6, 2593-2649,