|The author did not answer my major comment 3: How does the fit result represent the measured activation ratio? To my experience, size resolved activation ratio curves are usually not rotational symmetric. Sometimes a sigmoidal function such as eq.(1) can not represent well the original curve. The representative of the fit results can not be proved by the closure study with the average fit, since the closure results are mainly influenced by the temporal variation of the activation curve. |
Also, it seems that only average fit was used in the closure study. Why not use the average measured activation curve for the closure? Now the bias in the closure results stem from a mixture of the uncertainty of measurements, temporal averaging and fitting. By using the average measured curve, at least the uncertainly in the fit can be avoided.
The answer to my minor comment 1 can not really convince me. I do not think a concentration of 1000 cm-3 is too low for CCN measurement and can cause such a large uncertainty in measured CCN number concentration.
I still can not understand the time table (table 1 in your response) of SMPS and SS scan. Does it mean some SMPS scans are located in two different SS? What kind of SMPS scan in table 1 was finally used in your study?
Line 263 and fig. 2 in the manuscript v3: in fig. 2 the max of NR-PM1 is about 60 ugm-3 but not 72 ugm-3 as state in the text. Please check the figure and the data again.
fig. 3 in the manuscript v3: the right y-axes label should be Greek letter “kappa”.
Section 3.3.3: the word “hygroscopicity” used in this section is not appropriate. Hygroscopicity is a property of aerosol. The hygroscopicity of aerosol population is determined by both chemical composition and mixing state. Some expression sounds strange or confusing, such as “non/less hygroscopicity species”, “the relative importance of mixing state and hygroscopicity”, “when hygroscopicity increased from 0.30 to 0.39”, “the sensitivity of the NCCN prediction to hygroscopicity”, “NCCN prediction is more sensitive to mixing state at high SS than hygrocopicity”, etc.
The comparison between the Nccn calculated with average D50 and average AR curve is unreasonable. Large temporal variations can been found in fig. S6 for AR curve. In fig. 5iii, the bias of the slope in subplots e-h mainly stem from the temporal variation of AR (maybe also the uncertainties in the sigmoidal fit of measured AR); while the bias of the slope in subplots a-d mainly stem from the disadvantage of D50 (i.e. stepwise AR) and the temporal variation of D50. The influence of the temporal variation is unknown, thus the comparison between those biases of slopes can not support the conclusion “NCCN was found to be more sensitive to hygroscopicity than to mixing state at SS = 0.15% but the reverse is true at SS = 0.70%”. To see the importance of aerosol mixing states to Nccn prediction, one should use real-time D50 and AR. Actually, fig. 5ii a-e can give hints to the role of assumption of mixing states. In these subplots, x-axes can be considered as the Nccn calculated with real-time PNSD and size-resolved AR, and y-axes is the Nccn calculated with real-time PNSD and D50. From those subplots it can be seen the Nccn calculated with real-time PNSD and D50 is 6%, -6%, 3% and 10% higher than Nccn calculated with real-time PNSD and size-resolved AR for at SS of 0.15%, 0.35%, 0.50% and 0.70%, respectively. This result does not coincide with the conclusion of section 3.3.