While the referee acknowledges that most of the comments have been addressed adequately, he/she has a few minor comments. Please revise the manuscript accordingly. Thanks a lot.

While the referee acknowledges that most of the comments have been addressed adequately, he/she has a few minor comments. Please revise the manuscript accordingly. Thanks a lot.

Please find below an additional referee report:


Review: A novel method for calculating ambient aerosol liquid water contents based on
measurements of a humidified nephelometer system.
Authors: Ye Kuang et al.
This paper uses a machine-learning algorithm to model the aerosol dry volume from
nephelometer, aerosol size and BC mass measurements from 4 sites in the North China
Plain. The machine-learning process constructs a regional predictive model that needs
only 3-wavelength nephelometer measurements to predict the aerosol dry volume with
relatively low uncertainty. From humidified nephelometer and size-resolved kappa
measurements they calculate scattering and volume kappa values. These values along
with the aerosol Angstrom exponent form a look up table to estimate kappa volume
growth factors using only humidified nephelometer measurements. They then compare
the calculated aerosol water volume to that predicted from the ISORROPIA model.
This is an intriguing paper with broad implications on how to evaluate complex,
nonlinear systems. I approve the paper for publication. I’ve suggestions below on how to
better clarify points in the paper. This second version is an improvement from the first.
Think about applying this same method to other sites with long-term humidified
nephelometer, aerosol chemistry and size distribution measurements or to predicting
CCN from fRH. Note that KMA has size-dependent gRH and CCN measurements, as
well as fRH.
Comment: Dust is pervasive in the NCP. If not from the Loess regions, then roads and
construction dust are present. This may be a factor influencing differences between the
PM2.5 and PM10 aerosol in this study.
Comment: You need to stress in the Introduction that this study uses data from multiple
sites to characterize a regional aerosol. Otherwise, mixing data from multiple sites over
different time periods would give erroneous results. The quantity of data from multiple
sites used in the machine learning code is large enough to infer the results apply to a
larger data population and region.
Title: Change “contents” to “content”
Page1: Change “so far” to “before now”
Page 9, line 182: Did you mean “either or” instead of “neither nor” ?
Line 183… view of this, Qscat at 550 nm, as a function of particle diameter for four types
of aerosol particles, is simulated…
Page 15, line 300: “Kelvin”
Lines 302-304: You show in Figure 7a a large size-dependence to k with aerosol size.
Use of a single k volume parameter doesn’t assume a constant k value with particles size.
Rather it implies that a kappa(volume) can be expressed as a single value that perhaps is
proportional to a weighted average of the size-dependent kappa(diameter) values. Such
that 𝑘 𝑣𝑜𝑙 = 𝑤 ∗ 𝑘(𝐷). This assumption allows Vg(RH) to be expressed as a linear
function.
Page 16, line324: replace “consolidate” with “validate”
Page 18, lines364-366: rewrite “Figure 6 shows the influence of aerosol size and
chemistry on Rvf. For Angstrom exponents less than ~1.1, Rvf varies strongly with ksca.
However for Angstrom exponent values greater than ~1.1, the Rvf relative standard
deviation exhibits a higher variability with the Angstrom exponent; thus showing the
sensitivity of Rvf to changes in aerosol size for small particles.”
Line 370:rewrite “ ..thereby kf , such that these values can be directly…”
As a note, the high variability of Rvf at high Angstrom exponents may result from
differences between monomodal and bimodal size distributions.
Line 379: rewrite “… is highly variable yet has no apparent correlation with aerosol
loading.”
Page 23, line 478: rewrite “On average, when ambient …”