The instruments for CCN and aerosol measurements have been mounted roughly 20 m above ground on the roof of a building in the campus of Fudan University in Shanghai (31∘18′ N, 121∘29′ E) since October 2010. The site is surrounded by populated residential and commercial areas, as well as urban streets. The East China Sea is roughly 40 km east of the site, and the prevailing wind directions are southeasterly in summer and northeasterly in winter. Local time (LT) hereafter employed in this study is 8 h ahead of UTC.

The CCN number concentration (NCCN) was measured using a continuous flow and single column CCN counter (model CCN-100, Droplet Measurement Technologies, USA), in which an optical particle counter (OPC, 0.75–10 µm) is employed to detect activated cloud droplets (Roberts and Nenes, 2005; Lance et al., 2006). The instrument was housed in an air-conditioned weather-proof container with temperature maintaining at 20 ∘C. The ambient aerosol airflow passed through a dryer (active carbon) to lower relative humidity below 30 % before entering the instrument (Leng et al., 2013). The CCN counter was calibrated using ammonium sulfate before the study, as did calibrations for temperature gradient, flow, pressure and OPC to maintain stable supersaturation (SS) according to the Droplet Measurement Technology (DMT) operation manual. In order to ensure accurately counting, zero checks were performed before and after the campaign and regularly every 2 months. The effective water vapor (SS) changed alternately at 0.2 % interval within 0.2–1.0 %. In real atmosphere, SS varies from slightly less than 0.1 % in polluted conditions to over 1.0 % in clean-air stratus cloud (Hudson and Noble, 2014). The selection of SS 0.2 % in the present study would benefit to the measurements in the urban environment for further analysis. Although the CCN counter can operate well under conditions of particles only in the range of a few thousand per cubic centimeter, and corrections are needed for larger concentrations (> 5000 cm-3) (Lathem and Nenes., 2011), we still used the measured NCCN directly at 0.2 % SS in this study since it seldom reached the upper limit.

A high-resolution wide-range particle spectrometer (WPS-1000 XP, MSP) was employed to observe particle size distributions in the size range of 10 nm-10 µm. The principles of the instrument, which have been introduced in detail by Gao et al. (2009), combine laser light scattering (LPS), condensation particle counting (CPC), and differential mobility analysis (DMA). The DMA and CPC can effectively measure aerosol particles distributed in the size range of 10–500 nm in up to 96 channels. The LPS scan the size range of 350–10 000 nm in 24 additional channels. In the present study 60 channels in DMA and 24 channels in LPS for the sample mode were chosen and 3 min were needed to scan the entire size range completely, as it took 2 s for scanning each channel. DMA was calibrated with National Institute of Standards and Technology (NIST) Standard Reference Materials (SRM) 1691 and SRM 1963 Polystyrene Latex (PSL) spheres (mean diameter of 0.269 and 0.1007 µm, respectively) to maintain DMA transfer function properly and accurate particle sizing traceable to NIST. Four NIST traceable sizes of PSL (i.e., 0.701, 1.36, 1.6 and 4.0 µm) were used to calculate LPS. The calibration and operating methodology of WPS has been described elsewhere (Zhang et al., 2010). In addition, we have compared the aerosol size spectra measured by WPS with those measured in parallel by a calibrated scanning mobility particle sizer (SMPS, TSI 3080) with higher accuracy in the size range of 20–800 nm, including size-resolved particle concentrations and peak sizes, and a strong correlation between them was derived with correlation coefficient R2>0.95 (Leng et al., 2013). The result confirms the reliability of WPS measurements for successfully characterizing the number concentration and size distribution of condensation nuclei (CN).

Planetary boundary layer (PBL) height and aerosol vertical extinction profile were measured using a set of micro pulse lidar (MPL) system (MPL-4B-532) with pulse energy 6–10 µJ and pulse repetition frequency 2500 Hz. The MPL is an eye safe, compact and autonomous instrument, and an effective tool used widely in the world to provide available high spatial (30 m) and temporal resolution (30 s) information of aerosol vertical distributions (Menut et al. 1999; Cohn and Angevine, 2000; Brooks, 2003). The range of lidar is roughly 30 km at night and 10 km during the daytime. The description of the retrieval of aerosol parameters by the MPL will be only briefly summarized here as it has been given by He et al. (2006). The vertical profile of the aerosol extinction coefficient is determined by a near end approach in solving the lidar equation (Fernald, 1984). The PBL height is determined by the MPL lidar at the altitude where a sudden decrease of scattering coefficient occurs (Boers and Eloranta, 1986). The overlap problem must be solved because it can lead to an underestimation of aerosol backscatter and extinction coefficients in the lowest altitudes having the majority of aerosols (He et al., 2006a). Outlined by Campbell et al. (2002), overlap is typically solved experimentally. The system is set to point horizontally to an averaged data sample with no obscuration, such as the late afternoon, when the atmosphere is well mixed and the aerosol loading is low. The backscattering over the target layer is roughly assumed constant. The similar calibration performed in 2009 showed the full overlap of about 4 km and data are needed to be corrected by the overlap correction function. Welton et al. (2002) fully discussed the uncertainties caused by the overlap correction and He et al. (2006) estimated it to be less than 10 %.

An online Aethalometer (AE-31, Magee Scientific Co., Berkeley, California, USA) was employed to measure black carbon (BC) at a 5 min time resolution. The instrument was operated at an airflow rate of 5 L min-1. Based on the strong absorptivity of BC to light at near infrared wavelengths (Hansen et al., 1984; Weingartner et al., 2003), BC concentration is determined using the measured light attenuation at 880 nm and the appropriate value of specific attenuation cross section proportional to BC mass (Petzold et al., 1997). The attenuation can be obtained by calculating the difference between light transmission through the particle-laden sample spot and the particle-free reference spot in the filter (Cheng et al., 2006; Dumka et al., 2010). The operation, calibration and maintenance of AE-31 have been described in detail by Cheng et al. (2010).

An online analyzer for Monitoring Aerosols and Gases (MARGA, ADI 2080, Netherlands) was employed to measure the concentration of major inorganic water-soluble ions (e.g., Na+, K+, Mg+, Ca+, SO42-, Cl-, NO3- and NH4+) in ambient aerosol particles at 1 hour time resolution. An air pump controlled by a mass flow controller (MFC) draws ambient air with airflow of 1 m3 hr-1 into the sample box. An internal calibration method by using bromide for the anion chromatograph and lithium for the cation chromatograph was operated over the entire measurement period to ensure this instrument to identify and measure ion species successfully. Instructions for the methods of sampling, operation and internal calibration have been described in detail elsewhere (Du et al., 2011). Moreover, the mass concentrations of particulate matter (PM) with aerodynamic diameter less than 2.5 µm (PM2.5), meteorological factors and atmospheric visibility were measured by a continuous PM ambient monitor (FH62C14, Thermo), an automatic weather monitoring system (HydroMet^TM, Vaisala) and a automatic visibility monitor at 5 min time resolution, respectively.

The HYSPLIT-4 model developed by the Air Resources Laboratory (ARL) of the National Oceanic and Atmospheric Administration (NOAA), USA (Draxler et al., 2003), was employed to compute 24 h air mass backward trajectories ending at 500 m height (AGL) and starting at 00:00 LT and 12:00 LT for each day. By doing so, we can identify aerosols from different source regions and analyze their effects on aerosol activity to compile a full view of the relation between fog–haze event and NCCN. According to these calculated trajectories plotted in Fig. 1, aerosol was classified into two categories: (1) maritime aerosol transported by air masses from marine areas on 6 November 2010 carrying dominant oceanic particles and (2) continental aerosol in air mass traveling a long distance over inland areas on 7, 8, and 9 November 2010 and carrying more anthropogenic particles (e.g., BC). Exactly speaking, the maritime air mass originated from the East China Sea, slowly traveled northwesterly across the Hangzhou Bay and finally arrived in Shanghai on 6 November. Then the air mass changed its path southeasterly at around 12:00 a.m. on 7 November, and traveled from northern inland areas and across the North China Plain (NCP) and the eastern region of China. The continental sources contained increasing industrial and agricultural emissions (e.g., biomass burning) due to long-term rapid economy growth and large population in the last few decades. We hope to better understand the impact of aerosols with or with less anthropogenic particulate pollutants on CCN in this study by comparing these two categories.

Haze is traditionally defined as an atmospheric phenomenon that the sky clarity is obscured by dust, smoke and other dry particles, and atmospheric visibility and relative humidity (RH) are usually less than 10 km and 80 % over one haze episode (Fu et al., 2008). The high frequency of haze or hazy days is observed in winter, especially in the urban environments of northern China (Sun et al., 2006). During the haze event, the enhancement of particulate pollutants may greatly affect aerosol activity and NCCN. The study performed in the Indo–Gangetic plain shows that winter haze exerts a significant impact on the fog and low-cloud formation (Gautam et al., 2007).

Fog can be viewed as a lower-atmospheric near-surface cloud, and plays an important role in processing aerosol particles and trace gases (Gultepe et al., 2007; Biswas et al., 2008). On one hand, physically similar to cloud droplet, fog droplet also forms by water vapor condensing on dry aerosol particle under supersaturated conditions. On the other hand, generally formed in the shallow boundary layer containing local emissions, urban fog traps more pollutants than cloud at high altitudes (Fisak et al., 2002; Herckes et al., 2007). A fog or foggy case is defined as a weather with patterns of low visibility (< 10 km) and high (> 90 %) relative humidity (RH). When 80 % < RH <  90 %, the weather was referred to as a complex of haze and fog co-occurring (e.g., foggy–hazy) in the present study. Figures 2 and 3 show a 4 day time series of pressure, atmospheric visibility, RH, temperature, wind speed and direction, and PBL height from 6 to 9 November 2010. In fact, since RH seldom reached up to 90 %; thus the period focused in the present study was characterized by both hazy and foggy–hazy cases. The haze pollution lasting at least 4 hours has been identified as one haze event by an earlier study in Shanghai, where authors paid attention to the formation of haze pollution (Du et al., 2011).

As shown in Figs. 2–8, the 4 day period was classified into three parts: a hazy episode (marked in black open boxes) from 22:00 to 23:00 LT on 6 Noveber and 10:00 LT on 7 November to 13:00 LT on 8 November, a foggy–hazy episode (marked in red open boxes) from 23:00 LT on 6 November to 10:00 LT on 7 November, and the rest for clear case. Statistics for meteorological conditions is listed in Table 1 where the extinction profiles are averaged for a certain altitude of 500 m. During the hazy and foggy–hazy cases, the average atmospheric visibility was about 4.44 km and 2.33 km, respectively, much lower than 15.4 km in the clear case. The winds from the east and the south brought clean maritime aerosol during the clear case, however, the winds from the north and the west brought polluted anthropogenic aerosol during the hazy and foggy–hazy cases. The particulate and gaseous matters, including pollutants (e.g., BC) emitted from agricultural biomass burning were transported along the air mass pathways (Fig. 1), led to a significant enhancement of aerosol extinction coefficient from hourly averages of 0.5 to 1.2 km-1 (Fig. 3). In addition, the PBL height downed to below 500 m and further suppressed the dilution of pollutants.

In order to visually identify aerosol evolution, particles in the size range of 10 nm to 10 µm were categorized into 7 sub-size bins: 10–20 nm (nucleation mode), 20–50 nm and 50–100 nm (Aitken mode), 100–200 nm, 200–500 nm and 0.5–1 µm (accumulation mode), and 1–10 µm (coarse mode) (Fig. 4). A similar classification was applied to the measurements at the same site by Zhang et al. (2010). In this study, the integrated-particle size-resolved number concentrations NCN) exhibited a regular diurnal cycle, with two peaks (9000–16 000 cm-3) almost within the traffic rush hours. The mean NCN exhibited no obvious difference between the foggy–hazy (8367 cm-3) and clear (8956 cm-3) cases, but it showed a higher value (10 500 cm-3) in the hazy cases, revealing a larger loading of particulate pollutants.

In general, 20–100 nm (Aitken mode) particles dominated the particle number size distribution, probably due to local traffic emissions and meteorological conditions (Ferin e al., 1990). The temporal variation trend of Aitken mode was similar to NCN. It was interesting that the particles of 100–500 nm (accumulation mode) dominated in NCN in the hazy case with peak concentrations higher than 7500 cm-3, almost twice as much as the clear case (4000 cm-3). However, the foggy–hazy case is comparable to the clear case, showing a mostly unchanged evolution of the fractions of individual size bin to total particles and NCN. In addition, Fig. 5 shows the average size distributions (10 nm–10 µm) for all the three cases. It is very visible that it contains relatively more large-sized (e.g., 100 nm) aerosol particles in the aerosol population during the hazy case than that during the clear and foggy–hazy cases. Especially aerosol particles larger than 200 nm (a typical CCN size at SS 0.2 %) were significantly enhanced.

Figure 6 shows the temporal variations of eight major inorganic water soluble ions in aerosol particles and four gaseous pollutants sampled during this study period. Measurements for SO42-, Cl- and NO3- were unavailable from 10:00 LT on 7 November to 08:00 LT on 8 November. Substantially, the average concentration of aerosol total water soluble ions (TWSI) in the hazy case (54.5 µg m-3) was comparable to the foggy–hazy case (50.4 µg m-3), and roughly 2 times that of the clear case (26.2 µg m-3). For the percentage of individual ions in TWSI, NH4+ and K+ were relatively higher by a factor of 1.8 in the hazy and foggy–hazy cases than in the clear case. Despite the lack of SO42- and NO3- partly during the hazy case, we can still conjecture their promotion on the basis of their gaseous precursor evolution of SO2 and NO2.

Gaseous pollutants are released into the atmosphere from natural and anthropogenic emissions. Among them, SO2 is known as one of the most important gaseous pollutants and a precursor responsible for acid rain. Also, it can participate in the formation of new particles through converting into gaseous H2SO4, which is the most common nucleation species due to its low vapor pressure at typical atmospheric temperature (Zhang et al., 2006b; Urone et al., 1968). Secondary aerosols produced from the formation of new particles contribute more to the global burden of aerosol number than primary aerosols and are important sources of CCN (Merikanto et al., 2009; Yu et al., 2008). Recent studies have shown the enhanced solubility of SO2 due to its reaction in fog droplets during a severe fog measured in the North China Plain, and this finding has provided important support for better understanding of the acidity in clouds (Zhang et al., 2013). NO2 mainly comes from vehicle traffic emissions in urban areas (Wang et al., 2006). Nitrogen oxides (NO, NO2, N2O5) undergo heterogeneous reactions with aerosol particles (e.g., sea salt or dust) during they are transported in the atmosphere (Elizabeth et al., 2006). Thus, high gaseous pollutant content can result in larger CN loadings and subsequently more CCN particles in the atmosphere. On the whole, the loading of these precursor gases in the foggy–hazy and hazy cases exceeded that in the clear case, specifically NO2 by a factor of 2 and SO2 by a factor of 1.5. Moreover, SO2 and NO2 concentrations reached their peaks around 00:00 LT on 8 November corresponding to the highest levels of CCN and aerosol activity, implying their potential effects on CCN production, which will be discussed in the next section.

Although aerosol size distributions were measured only in the size range of 10–10 000 nm, they were still used to predict NCCN according to Köhler theory (Köhler et al., 1936). Toward this end, the particle hygroscopicity “kappa” (κ) was used in the closure calculation. The description of the technique has been given by Petters and Kreidenweis (2007); therefore it will only be briefly summarized here. The κ parameter for one multicomponent particle can be obtained through weighting each component κi by their volume fractions in the mixture, κ=∑iεiκi, where εi is the volume fraction of chemical compounds in particles, and κi is the effective κ of individual chemical composition.

Assuming aerosol particles are completely internally mixed, a simplified κ was calculated using water soluble inorganic ions (organic matter data are unavailable). Aerosol particle compositions were classified into three categories (Petters and Kreidenweis, 2007; Wiedensohler et al., 2009), and κi and densities for each component are shown in Table 3, in which “others” is defined as “PM2.5-BC-inorganic ions”. The critical dry size (CDS) of particle to be activated as CCN at one SS can hence be determined by the following equation: S(D)=D3-Dd3D3-Dd3(1-κ)exp⁡(4σs/aMωRTρωD), where ρω is the density of water, Mω is the molecular weight of water, σs/a is the surface tension of the solution/air interface, R is the universal gas constant, κ is the hygroscopicity parameter, T is temperature, Dd is the dry diameter, D is the diameter of the droplet and S(D) is the critical dry size under a given SS. Detailed information for the derivation of Eq. (2) can be found in Petters and Kreidenweis (2007). Equation (2) applies over the entire range of humidity and solution hygroscopicity and can be utilized to predict the conditions of cloud droplet activation. The critical SS for a selected dry size of particle is determined from the maximum of the curve for Eq. (2). Computed for σs/a=0.072 Jm-2 and T=298.15 K, the calculated CDS varied between 60 nm and 130 nm and averaged at 102 nm. Particularly, the hourly averaged CDS during the foggy–hazy/hazy cases was slightly lower (96 nm) than during the clear case (105 nm). So far, the comparable or relatively higher CDS has also been found in diverse regions and for various aerosol types, despite different calculation models and SS. For example, the fresh aerosol particles emitted by an aircraft internal combustion engine have a CDS range of 146–301 nm at SS 0.7 %, depending on varying operating conditions (Hitzenberger et al., 2003). Furutani et al. (2008) investigated three types of aerosol masses along the southern coast of California, and the CDS was estimated at 110 nm at SS 0.6 % for fresh ship exhaust, 70–110 nm for fresh anthropogenic aerosols and roughly 50 nm for aged anthropogenic and clean maritime aerosols. In Vienna, the CDS has a wide gap between 69 nm and 368 nm, and averaged at 169 nm (Burkart et al., 2011). Quinn et al. (2008) observed the CDS in a narrow range of 70–90 nm for maritime aerosols in the Gulf of Mexico, and a moderate range of 90–170 nm in the ship channels of Houston with high marine traffic densities close to industrial and anthropogenic sources.

The CCN population can be effectively viewed as a subset of measured aerosol size distributions since the operating range (10–10 000 nm) includes the majority of atmospheric particles. Therefore, the predicted NCCN can be calculated through integrating particles upward in size from the bottom CDS to the upper boundary. In this calculation, the predicted NCCN of hourly averaged CDS was compared with the measured one correspondingly.

The results of this closure analysis are shown in scatterplot in Figs. 10 and 11. The prediction for CCN is generally success throughout the entire data set. The linear regression between predicted and measured NCCN produces a slope of 1.012 and an intercept of 128.3 cm-3 (R2=0.95), and the average ratio of predicted versus measured NCCN is 0.94 (Fig. 10). The results indicate some moderate underestimate (about 6 % on average) but the agreement is still excellent. The achieved closure calculation suggested that water soluble inorganic ions played a major role in contributing the κ value. In fact, 83.8 % of the κ was expressed by SO42- + NO3- + NH4+ in total (in another study by our group, not published yet), with their individual contribution to be 39.8 %, 31.7 % and 12.3 %, respectively. In addition, it is noteworthy that the predicted NCCN at SS 0.2 % was more correlated with the observed NCCN in the clear case (R2=0.96) than the foggy–hazy/hazy cases (R2=0.91), and the corresponding ratios of predicted to observed NCCN were 0.95 and 0.92, respectively (Fig. 11). In all cases, the mean ratio of predicted to observed NCCN never reached up to 1, suggesting that organic matter would play a second role and make up the rest of κ.