Continuous Measurement of Ambient Aerosol Liquid Water Content in Beijing

The amount of liquid water present on ambient aerosol particles influences its radiative properties, the partitioning of gas-phase organic and inorganic compounds to the condensed phase, and the assessment of particle health effects upon inhalation. An improved home-made Dry-Ambient Aerosol Size Spectrometer (DAASS) was implemented and deployed at a sampling site in Beijing to measure water content and volumetric growth factor of fine ambient aerosol during clean and polluted days. Aerosol chemical composition was characterized by an aerosol chemical speciation monitor (ACSM). Meteorological conditions were also assessed to trace origin of the air masses arriving at the site. A thermodynamic model, ISORROPIA II, was used to predict aerosol water content based on inorganic compositions measured by ACSM. It was then compared with the aerosol water content measured by DAASS. During the two-week sampling period, an aerosol efflorescence behavior was observed around 50% relative humidity (RH). Aerosol at the sample location was predominantly alkaline, with an average acidity ratio of 1.3 during the polluted days but very variable during the clean days. This fact coupled with generally low ambient RH resulted in low amounts of liquid water detected during the campaign. When RH reached its maximum of 87% and aerosol appeared to be more acidic, the maximum amount of aerosol water content measured was 1.3 μg m (44.9% wt of DAASS measured ambient aerosol). Measured aerosol water content and model predictions seem to show a better agreement during the more polluted days, but during the clean days aerosol organic fraction appears to have an important contribution in the water content. We argue that the organicassociated water is comparable with that associated with inorganic compositions during the clean periods at low ambient RH conditions.


INTRODUCTION
Despite the known importance of water uptake in ambient aerosol systems (Charlson et al., 2001;Johnson et al., 2004;Lohmann and Feichter, 2005;McFiggans et al., 2006) there is a relative lack of aerosol water measurements.Water is one of the most important constituents of ambient aerosol at certain relative humidity (RH) conditions (Kreidenweis et al., 2008) and its amount is governed by gas-liquid equilibrium (Kreidenweis and Asa-Awuku, 2014).There is an intrinsic complexity of the atmospheric aerosols with regard to hygroscopicity, given the large differences in particle sizes, varying from nanometer to micrometer, and the varying chemical components in different size modes.Water content is found to be a key factor in determining total ambient particulate matter (PM) mass concentration, causing an impact on light scattering and extinction coefficient (Day and Malm, 2001;Garland et al., 2007;Jung et al., 2009), and has profound implications for atmospheric chemistry as it can enhance chemical reaction rates on the aerosol (Ervens et al., 2011;Lee et al., 2011;Wang et al., 2012).
There are diverse methods to measure aerosol water content, e.g., gravimetric analysis to determine the amount of water associated with the mass of aerosol on a filter (McInnes et al., 1996;Lee and Hsu, 2000;Spindler et al., 2012), optical methods using Nephelometers or extinction cells (Carrico et al., 2000;Nessler et al., 2005;Kim et al., 2006), microscopy such as scanning electron microscope (SEM) and environmental transmission electron microscope (ETEM) (Ohta et al., 1998;Ebert et al., 2002;Wise et al., 2005), electrodynamic balance to study properties of single droplets of multicomponent systems (Tang and Munkelwitz, 1993;Tong et al., 2011;Davies et al., 2013;Zuend et al., 2013).Among all, the most frequently used is the hygroscopic tandem differential mobility analyzer (H-TDMA) (Liu et al., 1978;Rader and McMurry, 1986), which measures hygroscopic growth by size-selecting individual particles with a Differential Mobility Analyzer (DMA), exposing them to elevated or reduced RH and then measuring aerosol diameter change using a second DMA (Swietlicki et al., 2008;Kitamori et al., 2009;Ye et al., 2011;Wang et al., 2014).However, the HTDMA cannot easily measure liquid water content of atmospheric particles in their ambient state because particles are often dried prior to sampling, this fact cannot ensure that no other species but just water are removed (Gysel et al., 2006;Gysel et al., 2007;Swietlicki et al., 2008), or particles may effloresce or their morphology may change (Engelhart et al., 2011).An additional inconvenience is the troublesome and time consuming when scanning a range of particle sizes and RH values with respect to the fact that ambient aerosol composition changes continuously.These methods are not suitable for near real-time continuous measurements of aerosol water content at ambient conditions.A Dry Ambient Aerosol Size Spectrometer (DAASS) was proposed to meet this challenge.It assesses the amount of aerosol water at atmospheric conditions by merging aerosol number size distribution with the volumetric growth factor (VGF).It was first introduced by Stanier et al. (2004) during the Pittsburgh Air Quality Study.A reduced version of the instrument was used to measure aerosol water content and VGF of fine PM during the Finokalia Aerosol Measurement Experiment (FAME-2008) in Greece (Engelhart et al., 2011).
Rapid economic and industrial development of China in recent years is evidenced by its growing urbanization and energy consumption.But such development has brought the Chinese cities severe urban air pollution in recent decades (He et al., 2002).Beijing, for example, frequently has severe PM pollution events that normally are accompanied with increasing ambient RH (Jia et al., 2008;Jiang et al., 2015).It possibly could have high aerosol water content.In Guangzhou and Shanghai, optical methods (using nephelometers, transmissometer and photometer) were employed for the determination of hygroscopic growth factor and evaluation of scattering and light extinction of ambient aerosol (Liu et al., 2008;Jung et al., 2009;Cheng et al., 2014).While in the North China Plain a series of interrelated studies (Achtert et al., 2009;Massling et al., 2009;Meier et al., 2009;Bian et al., 2014;Liu et al., 2014) evaluated the chemical and physical influence on hygroscopic growth of atmospheric sub-micrometer particles using HTDMA, Twin Differential Mobility Particle Sizer (TDMPS), and Humidifying Differential Mobility Particle Sizer (HDMPS) for rural and urban areas.So far, direct measurement of aerosol water content at ambient conditions using a DAASS system has not been performed in China.
In this paper we present measurements of aerosol water content using a home-made Dry Ambient Aerosol Size Spectrometer at a urban site in Beijing, China.Prior to its deployment, the instrument was calibrated using laboratory generated aerosol particles.Discrepancies between the ambient RH and the DMA sheath RH found in the original DAASS design were addressed in this work for a better resemblance of the ambient conditions inside the DMA at the moment of the measurements.Aerosol chemical composition was characterized using an Aerosol Chemical Speciation Monitor (ACSM) whose data was used as inputs for the thermodynamic model ISORROPIA II to theoretically estimate aerosol water content for comparison with DAASS results.Meteorological conditions and back trajectories of air masses are employed in the explanation of aerosol mass concentration and water content observed during the campaign.

Development of DAASS
The home-made DAASS was constructed following the schematics in Fig. 1.Major components of the instrument include an Scanning Mobility Particle Sizer (SMPS, TSI 3938) which includes an electrostatic classifier (TSI 3082), an Soft X-ray aerosol neutralizer (TSI 3088) using a differential mobility analyzer (TSI 3081), a condensation particle counter (TSI 3787), an aerosol dryer (Permapure MD 110-22), a sheath flow dryer (Permapure PD 200-24), three computercontrolled solenoid valves (Humphrey 310), and a custommade control box which controls the switch of the solenoid valves and acquires temperature and RH from the sensing probes.
The instrument was designed to continuously measure the number size distributions of aerosol particles in the range of approximately 10 to 500 nm.DAASS has two conditioning modes and two scanning modes (i.e., Ambient mode and Dried mode).Scan times for both Ambient and Dried modes are 3 minutes in duration.Alternated measurements of particle number size distributions (PNSD) in Ambient (red lines in Fig. 1) and Dried mode (blue lines in Fig. 1) are performed continuously in order to estimate the amount of liquid water present on the aerosol.During a Dried scanning mode, ambient polydisperse aerosol is initially dried in the aerosol Nafion dryer before entering the DMA, to an RH of about 25%, DMA sheath flow is kept in a closed loop passing through the sheath Nafion tube dryer until an RH of about 25% is reached, normally it is slightly lower than the aerosol RH.After finishing this Dried scan, the system enters into the Ambient conditioning mode where the SMPS sheath flow changes to open loop and bypassing the sheath dryer Nafion tube, ambient particle-free air is allowed to enter into the instrument to increase the relative humidity to that of the exterior ambient conditions.This conditioning stage lasts for 10 minutes, after which the system enters into the Ambient scanning mode.During the Ambient scanning mode, un-dried polydisperse aerosol is sent directly into the DMA.The sheath flow remains open allowing ambient humid air enter the line ensuring a very close value of sheath RH to that of the ambient RH within the DMA.After finishing the Ambient scan, the instrument changes back to the Dried mode configuration, remaining in a pre-dry conditioning stage for 7 minutes before starting the following scan.These 2 cycles are repeated continuously.After one dry and one ambient measurement have been preformed, PNSD's stored in the system are automatically processed by the Data Processing Algorithm.The amount of aerosol liquid water is calculated and displayed automatically.
Utility dry air at -40°C dewpoint is permanently supplied to both nafion dryers when in the dried mode.The -40°C dewpoint required for the utility air by the Nafion dryer was achieved by passing through a combination of molecular Sieve 3A (J&K Scientific, 4-8 mesh pellets) and Silica gel dryers.During the Ambient mode, the utility dry air flow is stopped in order to conserve the desiccant integrity as long as possible and to reduce maintenance for the instrument.
The sheath to aerosol flow ratio for the DMA is 4:1 L min -1 .The instrument is controlled via USB from a computer running a graphical user interphase (GUI) built in LABView graphical language which displays basic information of the instrument condition (temperature and RH at several locations, current stage of the measurement) while the data processing algorithm written in Python (Version 2.7) runs on the back.Description of the procedure to calculate the amount of water content present on aerosol particles can be found elsewhere (Stanier et al., 2004;Engelhart et al., 2011).The calculation is based on three fundamental assumptions: volume additivity is considered valid for all the aerosol components including water; a single and size-independent VGF factor applies to different sized particles; aerosol particles are spherical.Briefly, total measured volumes of ambient and dried size distributions are calculated based on PNSD's.Using an assumed initial volumetric growth factor, an iterative procedure is employed to determine the upper limit dry size for the dry volume integral so that the same number of particles is considered in both distributions.Aerosol water is then calculated as the difference of ambient to dry volume multiplied by the water density.
Prior to the deployment of DAASS for field measurement, a number of calibrations were performed in the laboratory.Sizing accuracy for the instrument was checked using laboratory-generated monodisperse Polyestyrene Latex spheres (PSL, Duke Scientific, USA) at diameters of 50 and 100 nm.Diffusional particle deposition within the transport lines were estimated using empirical particle loss corrections from Kulkarni et al. (2011) and (Wang et al., 2002) from the processed size distributions given by the TSI SMPS program (Fig. S1).Diffusional losses for the nafion tube were estimated using the equivalent pipe length method (Wiedensohler et al., 2012).Finally, the instrument was used to measure diameters of hydrated and dried sodium chloride particles generated in the laboratory by atomizing the solution of 0.02 wt% NaCl in ultra-pure water.The generated NaCl dry aerosol size distribution has a geometric mean diameter of 122 nm and a geometric standard deviation of 1.66.Changes  (Gysel, 2003).As shown in Fig. 2, measured results have reasonable agreement with theoretical growth data.

Uncertainties in the Measurement of Aerosol Water Content
Sources of error in the DAASS water measurements were reported in Stanier et al. ( 2004).Since the determnation of aerosol water relies on particle counts in both Dried and Ambient modes, it is important to minimize bias between the two modes.Difussive particle losses, particularly in the Dry mode, could directly impact particle counts.When developing the instrument, aerosol flow lines were designed as short and straight as possible, with no bend exeeding 90°.In the case of losses inside the aerosol nafion dryer, correction of particle count were applied based on the equivalent pipe length method (Wiedensohler et al., 2012), ensuring that count difference for laboratory-generated dry aerosol is lower than 5% between two modes (Fig. S2).The ambient and dried particle distributions are not measured simultaneously.Changes in the aerosol distribution in such period could cause error in the calculation of volume growth.Stanier et al. (2004) presents a correction factor for the changes in the dry aerosol volume (ß) during a lapse of time.The assumptions made when estimating water content may also introduce error.Urban aerosols tend to be externally mixed, with components of different hygroscopicities and possibly various growth factors.The growth factor calculated for DAASS is a volumeweighted average growth factor for all the species in the aerosol.Ambient particles may deviate from the assumption of sphericity, propagating error to the calculation of aerosol volume.It was reported that atmospheric particles may undergo morphological re-conformations when they are under repeated changes of RH (Mikhailov et al., 2009).Finally, studies on particle hygroscopicity and water uptake have commonly used the assumption of additivity and found acceptably low errors (Kreidenweis et al., 2008).

Field Measurement
Atmospheric measurements were carried out during the last two weeks of March 2015 (March 20 th to April 2 nd ).Continuous aerosol liquid water content measurements were performed every ~20 minutes by the custom-made DAASS.The instrument was located at the rooftop of the Environmental Science Building (40°0'17''N, 116°19'34''E, ~10 m above the ground) on the campus of Tsinghua University, a semi-residential area.This site has been use to monitor PM 2.5 concentrations and air quality in Beijing since 1999 (He et al., 2001;Yang et al., 2011;Cao et al., 2014).To ensure a close resemblance of the ambient conditions within the system (Stanier et al., 2004;Engelhart et al., 2011), the instrument was placed outdoors at all times.Protection from direct sunlight and precipitation was provided to ensure appropriate functioning.
In the original reduced DAASS system (Engelhart et al., 2011), their setup included an auxiliary pump to drive a larger amount of humid air through the system during the Ambient conditioning.This helps to speed up the replacement of dry air inside the system.However, this auxiliary pump was not used for the scanning stage.When the ambient RH exceeded 70%, the sheath RH in their instrument was not able to reach such high values at the moment of the scan, forcing them to discard all measurements above this value (Engelhart et al., 2011, Bian et al., 2014).In order to avoid this problem caused by a sudden drop of the sheath RH going into the DMA once the Ambient conditioning mode was finished and the loop is closed back again (Stanier et al., 2004;Engelhart et al., 2011), our setup didn't use any additional pump or blower but maintained the sheath flow loop open even during the scanning stage.This setup shows good performance, normally reaching the ambient RH inside the system (Fig. 3(a) and 3(b)).In very few occasions, this setup had a slight difference between the exterior ambient RH and an averaged difference of 5% was observed (Fig. 3(c)).RH changes drastically in Beijing, e.g., in some few hours there could be differences of up to 50%.Nevertheless, all the calculations for aerosol water content were done using the RH value of the Sheath Flow.The maximum ambient RH measured during the sampling period was 87% with an equivalent sheath flow RH of 76%.Their difference represents the maximum difference (i.e., 11%) between both ambient RH and sheath RH measurements during the campaign.
During the campaign, chemical composition of submicrometer non-refractory (NR-PM 1 ) aerosol was measured with an ACSM (Aerodyne Research, Inc. (Ng et al., 2011)).It reported the mass concentration of sulfate, organics, nitrate, chloride, and ammonium in NR-PM 1 every 15 min.Data was analyzed by standard ACSM software written in Igor Pro (Wavemetrics, Inc., USA).Using the measured compositions, aerosol acidity was estimated according to the following expression: Acidity, which is greater for lower cation/anion ratios, dictates the relative abundance of different inorganic aerosol components (Dick et al., 2000).Molar ratios of the unity or greater are assumed to be fully neutralized aerosol.Even though the concentration of chloride was small during the sampling period, it was not negligible and was included in the acidity calculation.Chemical analysis of PM 1 in Beijing has reported low mass fractions of K + and Na + compared to ammonium ion.We hypothesize there is not significant variation in the acidity ratio without their inclusions.Furthermore, the thermodynamic equilibrium model (ISORROPIA II (Nenes et al., 1998;Fountoukis and Nenes, 2007)) was used to estimate aerosol water content based on aerosol inorganic compositions for comparison with DAASS results.ISORROPIA II can solve two kind of problems, one is in "Reverse" mode, in which aerosol phase concentrations, relative humidity and temperature of the system are known, and the other is "Forward" mode where both aerosol and gas phase concentrations with temperature and relative humidity are known.
Meteorological conditions (temperature, pressure, dew point, wind speed and direction) for the sampling location were obtained from www.wunderground.com.Back trajectories of air masses were downloaded from the HYSPLIT website (http://www.ready.noaa.gov/HYSPLIT_traj.php).Hourly PM 2.5 mass concentrations were downloaded form the website of the Beijing Environmental Protection Bureau, that has 35 air quality monitoring stations located around the city.The closest monitoring station to our sampling site is WanLiu station, located ~ 4 km from our sampling site with no major pollution sources in between.

Pollution Characteristics during the Sampling Period
Using ambient PM 2.5 concentrations from the Wanliu monitoring station and a threshold value of 50 µg m -3 , which is twice the daily mean PM 2.5 guideline value suggested by the World Health Organization (WHO) (WHO Regional Office for Europe, 2006), it was possible to distinguish 3 different periods with pollution characteristics during the sampling time.A first period of clean conditions (P1) is from 21:00 on March 20 th to 01:00 on March 24 th .A second period (P2) of moderate polluted conditions is from 01:00 on March 24 th to 15:00 on March 28 th .A third period (P3) of heavily polluted conditions is from 15:00 on March 28 th 15:00 and to 06:00 on April 1 st (Fig. 4).
Meteorological conditions changed in several occasions during the measurement period.These changes were accompanied with increases in aerosol mass load, particle number size distribution and major chemical compositions of aerosol particles (Figs. 4 and 5).Back trajectories of the air masses arriving at the sampling site for those three selected periods (P1, P2 and P3) were obtained using HYSPLIT model.
First period (P1) is characterized by northerly winds.Relative humidity was generally low for most of the period.WanLiu monitoring station reported a maximum hourly PM 2.5 concentration of 62 µg m -3 .Aerosol inorganic composition was very low while organic compositions was predominant and had marked peaks.A few new particle formation and growth events happened during this period.P2 presents air masses coming from southern regions, a largely populated and heavy industrialized area.There is a notorious increase in the daily relative humidity with maxima during the early hours everyday (~4 AM) and minima in the afternoons.PM 2.5 data from WanLiu reported a maximum hourly concentration of 162 µg m -3 .Comparing to the first period, the averaged particle number distributions for the whole period shifted toward larger particles, peaking at around 100 nm.P3 period presents similar conditions to those in P2 with southerly winds and peaks in particle number size distribution are around 100 nm.However, very low wind speed during the whole day of March 31 st allowed the built-up of pollution.PM 2.5 data from WanLiu reported a maximum hourly concentration of 219 µg m -3 during this day.
Positive matrix factorization (PMF, the PMF Evaluation Tool PET, v2.05 (Ulbrich et al., 2009) was used for analysis and evaluation on organic aerosol spectra measured by ACSM to resolve distinct OA components.Two resolved OA factors (a hydrocarbon-like OA (HOA) and an oxygenated OA (OOA)) are reported in this study.The deconvolved OA allowed to distinguish main composition of the organic peaks presented during the three selected periods.P1 has four visible peaks.The first two sharp peaks are mainly composed of HOA (75-85%), while the third and fourth peak, in which would be the transition period between P1 and P2 are composed of an almost equal amount of HOA and OOA.P2 presents three major OA peaks.All are composed mainly of a more oxygenated fraction, what is consistent with the air mass back trajectory bringing more aged aerosol from the southern region.P3 shows a continuously increasing mass concentration of OA.During the first day, HOA and OOA show equal concentrations but thereafter OOA takes a more predominant fraction of up to 70% of the organic aerosol composition.

Aerosol Efflorescence Behavior and Measured Liquid Water
During the measurement period the particles showed a small retention (below 10% in volume) of liquid water at RH between 20-50%.At 50% RH, there is an inflection point in the VGF trend of the particles tending to increase up to ~1.8.Observing this, we considered it as an efflorescence point for the ambient aerosol.At 60% RH, particles were able to retain somewhere below 40% of water.At about 70% RH, particles retained water in excess of 50% of their dry volume.Fig. 6 displays a statistical analysis of the data to quantify the correlation between VGF and ambient relative humidity.Mean and median VGF values falling into 2.5% relative humidity bins were calculated.The resulting curve was fitted with a polynomial regression of second order using the least square method, which results in a high coefficient of determination (R 2 = 0.76).An efflorescence behavior was reported previously for urban aerosol with a seasonal variations, i.e., in winter with a value of ~ 60% RH and in Spring with a value of ~55% when aerosol was more alkaline (Khlystov et al., 2005).Using an individual particle hygroscopic (IPH) system, Li et al. (2014) reported efflorescence of acidic ambient particles in Mt Lu (North China) at 49-53%.On the other hand, Engelhart et al. (2011) found no evidence of efflorescence for an acidic and aged aerosol on an island.
There was a marked influence of the aerosol acidity on the amount of liquid water present on the aerosol.An important difference in the acidity of the particles between clean days (P1) and polluted days (P2-P3) is visible in Fig. 7(b).The figure shows that during P1 aerosol acidity ratio is extremely variable while during P2 and P3 this ratio is more stable, with an alkaline aerosol presenting an average ratio of 1.3 during most of the time.This low aerosol acidity was reflected on the low amount of aerosol water detected by DAASS.An example could be seen for March 27 th .
During this day the aerosol was slightly more acidic than that of March 26 th but the atmospheric conditions were similar.However, the amount of measured water is markedly larger (Fig. 7(b)).This is to say, any perturbation on the almost constant alkalinity of the aerosol accompanied by a relative humidity over ~50% results in an increase on aerosol water content detected by the instrument.
The acidity or alkalinity character of the aerosol is strictly dependent on the relative abundance of chemical species in the atmosphere.In this respect governmental policies have had an impact on the local atmospheric aerosol composition.Strict emission limits on coal burning power plants and factories and vehicle emissions, and control of factories on emissions of SO 2 and NO x are measures put in place by the government that have resulted in reduction of the corresponding particulate components.However, very little has been done on the reduction of ammonia emissions, which continue in the rising (Meng et al., 2011;Wang et al., 2011), and reflects on the alkalinity that aerosol in Beijing shows.A comparison of ACSM aerosol mass and Acidity ratio shows how the aerosol is predominately alkaline during the sampling period, particularly for low aerosol masses (i.e., clean period) (Fig. 7(a)).Ianniello et al. (2010) had previously reported higher ammonia concentrations were found during northwesterly winds in Beijing.
The maximum concentration of measured aerosol water was 1.3 µg m -3 .It occurred during the most polluted day of   sampling on the 31 st of March (Fig. 7(b)).At this point, ambient RH reached almost 80% and the aerosol appeared to be more acidic which influenced the uptake of water.This high RH was not found in other period, i.e., for P1 and P2 ambient RH was never higher than 60% which limited the uptake of water in aerosol.Using an aerosol density of 1.5 g cm -3 in Beijing during winter season (Hu et al., 2012), aerosol mass concentration from DAASS measurements could be estimated.On the 31 st of March when the maximum concentration of measured water occurred, the mass percentage of water to DAASS measured ambient aerosol was 44.9%.

DAASS Measured Water and ISORROPIA Estimations
Using aerosol inorganic composition, temperature and DAASS sheath RH, estimation of aerosol water content were performed using the ISORROPIA II model.The calculations were carried out with the metastable mode to ensure liquid phase (Fig. 8(a)).Because concentrations in the gas phase were not available, we used the reverse mode.
Variability in aerosol water during P2 and P3 was driven primarily by the inorganic fraction of the aerosol.A comparison of the time series for ISORROPIA modeled water and DAASS measured water shows how they progress similarly (Fig. 8(a)).With further analysis we found that liquid water is better correlated with nitrate.Nitrate is more hygroscopic than sulfate (Kreidenweis et al., 2008) and was the main inorganic component during the sampling period with an average molar ratio NO 3 -/SO 4 2-= 4.5.Nitrate is expected to increase its importance not just in China but worldwide because of the generalized trend in policies aimed at reducing SO 2 emissions and the increasing emissions of NO x itself (Hodas et al., 2014).
For period P1 when the larger fraction of aerosol is mainly of OA there seems to be no similarity with the estimations from the model.To infer the uptake of the inorganic fraction independently, ISORROPIA model does not take into account the interaction of the inorganic aerosol with the organic fraction on the uptake of water.Although the impact of organics on water uptake is largely unknown (Pajunoja et al., 2015), the important fraction that OA has on the total aerosol mass reported from the ACSM (specially during the low polluted conditions of P1) may influence the total amount of water on the aerosol.Fig. 8(b) indicates that during P1 the DAASS measured water content has almost identical trend with that of HOA concentration.Therefore, we evaluated the correlation between the organic and inorganic aerosol fraction of the aerosol for P1 with liquid water.Using a simple linear regression model we found that the measured aerosol water was similarly correlated with nitrate (R 2 = 0.31) and organics (R 2 = 0.28) and less with sulfate (R 2 = 0.15).This analysis suggests that organicassociated water content is comparable with that of nitrate for the cleaner period P1 at low RH conditions.

CONCLUSIONS
Continuous measurements of ambient aerosol water content were conducted in Beijing using a home-made DAASS system.This system was calibrated using laboratory-generated aerosol prior to its deployment.The campaign was from March 20 th to April 2 nd in 2015.Three periods were distinguishable based on aerosol mass concentration and chemical composition, allowing the study of aerosol water content in clean and polluted days.Air mass back trajectories explain the pollution characteristics and ambient conditions of three identified periods.Sampled ambient aerosol was often alkaline, emphasizing the relevance that ammonia emissions have currently in the local aerosol chemistry and therefore on the uptake of water by the aerosol.Although the calculated VGF showed a large variability, particularly in the range of 60-80% RH, a visible variation in the trend of the mean and median VGF was considered as an efflorescence behavior for the urban sampled aerosol at ~50% ambient RH.A similar behavior was also reported in other urban measurements.In general, low amounts of liquid water were observed during the campaign.Between measured DAASS liquid water and ISORROPIA II modeled water, better agreement occurs during the polluted periods (Fig. S9).When the aerosol showed to be more acidic the measured aerosol water content increased (Fig. S7).For the cleaner days the amount of water showed to be similarly correlated with nitrate and organics, and less with sulfate.Yet still requiring some adjustments for a more rapid measurement, the present instrument offers an attractive alternative for an unattended and continuous method to evaluate local aerosol water content at ambient conditions.

Fig. 1 .
Fig. 1.Schematics of the home-made DAASS system.Red lines are flow lines in Ambient mode and blue lines are flow lines in Dried mode.

Fig. 2 .
Fig. 2. Theoretical (line) and measured (dots) growth factors for sodium chloride aerosol.Vamb and Vdry are the aerosol volumes measured by the instrument in the Ambient and Dried modes.

Fig. 3 .
Fig. 3. Relative humidity conditions during the sampling period.(a) Sheath RH follows closely with the ambient RH during the sampling period; (b) Sheath RH vs. ambient RH for the sampling period (the average difference is ~5%).(c) An example of RH in the system at different instrument modes: light green region is pre-Ambient conditioning, light red region is pre-dry conditioning, and light yellow regions are scanning periods.

Fig. 4 .
Fig. 4. Ambient aerosol mass and number concentrations during the sampling period.(a) Contour plot of the particle number size distributions (dN/dlog 10 Dp) measured by DAASS.(b) PM 2.5 mass concentration at the WanLiu monitoring station vs. ambient RH at Tsinghua sampling site.

Fig. 5 .
Fig. 5. Ambient aerosol chemical composition given by the ACSM.(a) Total aerosol chemical compositions vs. time.(b) Organic aerosol factors determined by positive matrix factorization analysis.OOA: oxygenated organic aerosol, HOA: hydrocarbon-like organic aerosol.Wind direction as reported from wunderground.com is included as well.

Fig. 6 .
Fig. 6.Correlation of Volume growth Factor with Ambient Relative Humidity.Blue hexagons are mean VGF values for the whole sampling period.Yellow stars are median VGF values.Red line is a fitted curve to the mean values.Shaded area represents VGF within one standard deviation from the mean.Coefficient of determination (R 2 ) is given on the top-right side.
Fig S7 in the supplementary information evidences how from period P1 to period P3 aerosol moves towards more acidic conditions (left side of the figure) and simultaneous increases in water content occur.

Fig. 8 .
Fig. 8. DAASS measured aerosol water (blue dots) vs. time.(a) Aerosol water as calculated from ISORROPIA II (brown line) is better tracked by DAASS water during the polluted periods.(b) DAASS water and HOA fraction vs. time.Days with low aerosol inorganic fraction present coincident peaks of measured water and HOA.