Response to the reviewers on the paper “ An alternative method estimating hygroscopic growth factor of aerosol light scattering coefficient : a case study in an urban area of Guangzhou , South China ”

(1) How sensitive are the results to changes in the composition of aerosols and to changes of RH? Response: The changes of relative humidity (RH) affect the water content absorbed by hygroscopic species in aerosol, and can considered as the changes of aerosol chemical composition. Regardless the particles mixing state, the standard deviation of fsp(RH) increased to about 8% of the average fsp(RH) value at RH of 0.9. This indicates that the changes of chemical composition had small impact on the fsp(RH) under most RH condition.

(2) There is a need to give an assessment of the uncertainty due to potential errors in the calculated chemical species of aerosols

Response:
The uncertainties due to potential errors in the calculated chemical species were attributed to the parameters including particles volumetric equivalent diameter, mass of water-soluble ions and OC/EC, ambient TEMP, ambient RH and the ORI of EC.
Based on these uncertainties, the overall uncertainty of calculated f sp (RH) was estimated to be 9.38%.
(3) Give a better account of the accuracy of the equilibrium model ISORROPIA

Response:
In the present study, the accuracy of ISORROPIA II running was about 95% (±5%) according to the mass conservation of chemical species between the input and output data.Based on the theory of ISORROPIA II (Fountoukis and Nenes, 2007), the balance of charge equivalence between the input cations and anions warranted the model functioning properly and with high accuracy.
The extended aerosol thermodynamics model (E-AIM) (Wexler and Clegg, 2002) is another popular tool for predicting the water content and partitioning inorganic components in aerosol system.The E-AIM Model III can be found at the website: http://www.aim.env.uea.ac.uk/aim/aim.php .
We once compared the primary output of ISORROPIA II with that of E-AIM Model III using the aerosol measurement results as input.
The comparison results are illustrated in Figure 1.Small differences were found in the output of NO 3 -and NH 4 + between the two models.The obvious differences may exist in the partitioning of the amount of HSO 4 -and SO 4 2- . Furthermore, a good linear correlation in the estimated H 2 O mass was found between the two models if excluding a few outliers.The slope of the regression (=0.81) suggests that the H 2 O mass estimated by E-AIM Model III is a bit higher than that by ISORROPIA II.
We chose ISORROPIA II for our study since it meets our research demands.
ISORROPIA II can help to determine aerosol composition with reasonable accuracy.
Moreover, the executables of ISORROPIA II can be easily acquired from internet, the computation of ISORROPIA II is highly efficient, and it is quite suitable for batch processing when there is a considerable amount of data.

Response:
This study focuses on the development of an alternative method for estimating f sp (RH).Hence, the calculation method was provided in detail, while the chemical analysis procedures were only described briefly.In the revised paper, we have clarified ambiguous points and provided more descriptions on the experimental section, especially on the OC/EC measurements.
Two stages were performed in the experiments to determine carbonaceous aerosols.The first stage of carbon analysis was in an inert helium atmosphere and consisted of four temperature steps: 250 °C (60 s), 500 °C (60 s), 650 °C (60 s), and 850 °C (120 s).The second stage was conducted under an environment of 2% O 2 /98% He, and the temperature was set as 550 °C (45 s), 625 °C (45 s), 700 °C (45 s), 775 °C (45 s), 850 °C (45 s), and 870 °C (120 s).Due to the non-uniform particle deposition on the filters collected by the cascade impactor, laser correction did not work properly to separate OC and EC based on this protocol.Hence, we defined OC as the fraction of carbon that evolved at or below 850 °C in a helium atmosphere (in the first stage), and EC as the fraction of carbon that evolved after oxygen was introduced to the carrier gas (in the second stage).A similar approach was applied in a previous study (Huang and Yu, 2008).
In fact, the MSP high flow impactor, with an inlet and regular stages with cut-point diameters of 18, 10, 2.5, 1.4, 1.0, 0.44 and 0.25 μm, was employed in this work to collect size-segregated aerosols.The carbonaceous aerosols in the first 2 stages (>18μm and 10-18μm) were not determined in this study because particles in these size ranges widely deposited on the filter that cannot be covered by the punch area (1.5cm 2 ) required by the analyzer.
On the other hand, PM 2.5 and PM 10 samples were also collected by two aerosol samplers (BGI Incorporated, Waltham, MA, U.S.A., Model PQ200) at the same monitoring site on November 12, 14, 16 and 18, 2010.Both samplers were operated at the flow rate of 16.7 L min -1 .One sampler was equipped with a PM 2.5 cut cyclone (Model VSCC), while the other was equipped with a PM 10 cut cyclone.The quartz filter was analyzed for the OC/EC fractions following the IMPROVE thermal/optical reflectance (TOR) protocol on a DRI model 2001 carbon analyzer (Atmoslytic, Inc., Calabasas, CA, USA) (Chow et al., 2007).This analysis acquired four OC fractions (OC1, OC2, OC3, and OC4 at 140°C, 280°C, 480°C and 580 °C, respectively, in a helium [He] atmosphere), OP (a pyrolyzed carbon fraction determined when transmitted laser light attains its original intensity after oxygen [O 2 ] was added to the analysis atmosphere), and three EC fractions (EC1, EC2, and EC3 at 580°C, 740°C and 840 °C, respectively, in a 2% O 2 /98% He atmosphere).IMPROVE_TOR OC is operationally defined as OC1 + OC2 + OC3 + OC4 + OP and EC is defined as EC1 + EC2 + EC3 -OP (Chow et al., 2007).
The comparison of the carbon fraction measurement results of the high flow impactor with those of BGI aerosol samplers was presented in Figure 2.Although only four data points were available for comparison, an excellent agreement was found in total carbon (TC) between the two different measurement methods.However, the bias of OC between the analysis method employed in this study and IMPROVE TOR is estimated to be about +10%, while the bias of EC is about -30%.The bias may be aroused from the different thermal gradient program and the laser correction.
A sensitivity test was thus conducted to quantify the impact of the uncertainties in OC/EC separation on the b sp results.It was found that a 10% variation in OC only resulted in a 3% variation in b sp , while a 30% variation in EC only resulted in a 1% variation in b sp .Uncertainties in b sp will cause uncertainties in the calculated f sp (RH).
As shown in Figure 3, the impact of the uncertainties from the measured mass size distributions of OC and EC should only cause no more than 3% uncertainties in the calculated f sp (RH) in this study.

Response:
Although bounce can be avoided by coating substrate with oil or grease, it will definitely affect the result of the chemical species measurement.Furthermore, the metal foils substrate cannot meet the requirement for carbon analysis.As a result, we use the quartz fiber membrane filters to collect particles in this study, where the particles can be inset into the membrane structure.An earlier study (Chang et al., 1999) also suggested that the better collection efficiency for particles can be obtained by using glass fiber filter instead of aluminum as impaction substrate.The performance for collection efficiency by using glass fiber filter sometime can be as good as that of using oil-coated substrate.Considering the similar membrane structure and silicon-based material between quartz fiber filter and glass fiber filter, the particle bounce effect caused by cascade impactor can be reduced to some extent.
On the other hand, TC in PM 2.5 and PM 10 samples shows excellent agreement between MSP high flow impactor and BGI aerosol samplers (see Figure 2 and related responses above).Therefore, the bounce of particles is insignificant in this study.
(6) Results 3.1: The analysis of charge balance in aerosols (Figure 2a) is wrong.
The calculation should be based on "charge equivalence", not on "molar concentration".

Response:
The figure in ACPD version of the manuscript did consider the balance of charge equivalence between the input cations and anions, but with the not appropriate axis title.The corrected plot was inserted into the revised manuscript (7) Results 3.1: The relationship between OC and EC is in an unusual pattern.
This could be a result of the mismatch between sampling and analysis methods.

Response:
Thanks for the careful review.We apologized for using the wrong dataset in this plot.Specifically, we compared this plot (Figure 2b

Response:
It is true that particulate sodium is mostly observed at coarse particles, and sodium is usually regarded as an indicator of sea salt aerosols.However, substantial sodium content has also been detected in anthropogenic plumes associated with coal combustion (Takuwa et al., 2006).Our previous study (Tao et al., 2012) also reported a considerable amount of sodium in PM 1 at the same site of Guangzhou where is not near the ocean.
As mentioned in the response in (5), the bounce of particles is insignificant in this study.Size-segregated aerosol samples were also collected using the same model cascade impactor at a coastal site (Zhuhai) of the Pearl River Delta region during the same period as of this study.As shown in Figure 4, the size distribution of sodium at this site was characterized by a unimodal pattern peaking at coarse mode during the wet season when air masses mostly originated from the ocean.It is cleanly shown in Figure 4c that no substantial sodium were found in the fine mode at the Zhuhai site even at sodium-riched atmosphere.However, a substantial amount of sodium occurred at the fine mode during dry season when air masses coming from mainland China.Moreover, Na + and Cl -mass concentrations in PM 2.5 in Zhuhai, Shenzhen, Dongguan, Guangzhou ,Conghua (Guangzhou rural site) and Foshan in summer time in 2010 were also collected (see in Figure 5 and 6 where using the data from an another paper of under reviewing).The PM 2.5 samples were collected using low-flow air samplers (MiniVol TAC, AirMetrics Corp., Eugene, OR, USA).Evidently, the Na + concentrations in coastal sites (e.g.Zhuhai and Shenzhen) were lower than those in inland sites (e.g.Guangzhou and Foshan, even rural site Conghua).It suggested the higher Na + concentrations in PM 2.5 in Guangzhou may had connections with coal combustion or other anthropogenic sources.
In conclusion, the sodium in the fine mode is likely related to coal combustion and/or other anthropogenic processes.The sodium distributed in submicron size range was not from particle bounces in the collection process.
Figure 5 The map showing the locations of the 6 sites where also collected PM 2.5 Figure 6 The mass concentration of Na + and Cl -in PM 2.5 samples collected in the 6 sites.

Figure 3
Figure 3 in ACPD version of the manuscript) with an another figure (Figure 4 in ACPD version of the manuscript) and then found out this mistake.The corrected plot has been used in the revised manuscript.(8)Results: Na 2 SO 4 showed peak in submicron size range (Figures3-4).As particulate sodium comes mostly from sea sprays, it is unusual to have the species existing mostly in submicron range.I strongly suspect the occurrence of particle bounce in the impactor.

Figure 4
Figure 4 The size distributions of sodium and magnesium at Guangzhou during (a) wet season and (b) dry season in this study, as well as that at a coastal site (Zhuhai) at Pearl River Delta region during (c) wet season and (d) dry season