This manuscript introduced a best-estimate CCN spectrum product across a wide range of supersaturation range from 0.0001 to 30%, using multiple independent instruments at ARM SGP site. Cluster analysis is used to identify the key driving element (aerosol number size distribution) in CCN spectra. The seasonal and diurnal cycles of the clusters are analyzed as well as the time evolution of these clusters. This dataset is useful in aerosol and aerosol-cloud interaction study. It also has potential to benefit modeling study in high-SS or low-SS conditions. Overall, the manuscript is well written and I recommend it be published in ACP, with some comments listed below.

One aspect I think the authors can improve this manuscript is by adding more analysis and discussions on the uncertainty of the reconstructed CCN, especially in the high- and low-SS ends. Extending the CCN spectra into a wider range is a key point of this dataset. However, the authors also admit the data limits in the high and low ends of SS. This makes me wonder how good we can trust the data in the two ends. As the authors mentioned the uncertainty is included in the dataset, it would be helpful to add a plot of CCN uncertainty across all SS range and discuss how much we can trust the data. It would also be great if some discussions can be made on what can we learn from CCN in these extremely high (low) SS conditions.

Minor comment:

Title: since the wide SS range is a key point of this dataset, I would recommend adding something in the title to indicate that, such as “… Cloud Condensation Nuclei Spectra in a Wide Supersaturation Range …”

Line 15: I am curious if the authors have any plan to extend this data to more recent years. If yes, it will be good to discuss in the last of the manuscript.

Line 67: it would be good to list out (better within a table) the instrument names and size ranges used in building the aerosol size distribution and CCN spectrum.

Figure 1: is this example of one 45-min data sample? It would be good to present the date/time of the example.

Line 96-97: “… which can be used to check the accuracy of the reconstructed spectra.” This sentence is misleading me that I thought this paragraph is to evaluate the data accuracy using independent datasets. I suggest deleting this sentence and changing the next one as "… that were used to constrain the reconstructed CCN spectra at the low-SS (and high-SS??) end"

Line 106: why six instruments here? In abstract and summary it says eight instruments. Also, it would be good to list all of them in a table.

Line 120-121: How is k calculated from GF distribution? Appendix A says k distribution is calculated from GF distribution then averaged into one single k value. I thought the GF distribution, either single-modal or bi-modal, can be captured in k calculation. Please correct me if my understanding is wrong.

Line 130: “…, and these particles this lowest SSw region.” Is this sentence incomplete?

Line 133: again, it would be great to add analysis on the data uncertainty and discuss what value can be added with this wide SS range.

Line 142: can you explain the advantages of using skewed log-normal distribution other than regular log-normal? Any reason why CCN number concentration (and other variables in Figure B1) fit skewed log-normal distribution better?

Line 151-155: This point is interesting. Can we found high correlation within each aerosol mode and low correlation between bins in different modes (such as nucleation, Aitken and accumulation modes)? will it give evidence to choose size ranges for each mode?

Figure 6: change the caption to “diurnal variability of cluster fractional likelyhood”

Figure 6: it would be good to change the x axis label to 3-hr or 6-hr intervals. Same as other diurnal figures.

Figure 8: it would be interesting to add similar plots for aerosol number size distribution.

Line 291-292: I don’t understand why “indicate short-lived events tied to diurnal

cycles”. Shouldn’t the nucleation mode particles are tied to new particle formation events?

Eq. 3 in Appendix A: as it mentioned GF at a given RH is abbreviated to GF above, and GF is used in Eq.2, here GF(RH) should be GF for consistency.

Data availability: reference (Atmospheric Radiation Measurement (ARM), 1995, 2001, 2007, 2010, 2011, 2015) are not seen in the reference list.

Data availability: The final merged aerosol number size distribution and CCN data link (https://doi.org/10.5439/1832908) doesn’t work. Please check if the DOI is correct or if the data haven’t yet published in the ARM Archive.

Line 497: not complete: “is available at XXX”

Review of “Long- and Short-Term Temporal Variability in Cloud Condensation Nuclei Spectra in the Southern Great Plains” by Perkins et al. submitted to ACP [acp-2021-1008]

This study investigates the temporal variability of the five-year CCN observations at the ARM SGP site. The authors first constructed the CCN spectra database covering a wide range of supersaturation (SSw) conditions by combining concurrent aerosol size distribution and hygroscopicity (kappa) obtained from multiple instruments. They then conducted three statistical analyses using this data set, namely (1) parametric fitting of the CCN PDF; (2) cluster analysis of the CCN spectra; (3) temporal autocorrelation of the CCN number concentration time series. From each analysis they concluded that (1) the skewed log-normal distribution fits the CCN PDF; (2) there are three distinct clusters of CCN spectra and they are associated with aerosol size distributions dominated by different modes of aerosol particles; each cluster has distinct seasonal and hourly fractional occurrence; (3) the time scales of CCN number is shorter (1-3 hours) in high SSw condition than in low SSw condition. 

Overall, this is a clearly written manuscript. I agree that these multiple-year CCN spectra spanning the SSw condition from strong updraft in deep convection to fog are potentially useful for simulating cloud and precipitation formation as well as for the development of microphysics parameterizations. Applying cluster analysis to the CCN spectra is also a novel idea to explore the long-term dataset. The quality of the analysis and presentation meets the standard of the ACP. However, I found several questions/clarifications that need to be addressed, as listed below. If these issues can be properly responded to or revised, I would suggest the manuscript be accepted for publication.

Specific Comments/Major Questions

1. The scientific goal of the study and the relationship of the three statistical analyses (i.e., log-normal fitting of CCN PDF, cluster analysis of CCN spectra, and autocorrelation of CCN number concentration time series) need to be further clarified

When I first read the abstract, I was confused by how the statistical analyses are interrelated (e.g., the clustering was based on the fitted CCN PDF, or the autocorrelation was applied to the time series of cluster occurrence). It was until I read through the entire manuscript that I realized they are actually three separate results. Therefore I suggest to

• Revise the abstract to explicitly clarify these analyses as three different tasks and their specific objectives.
• Add a few sentences to state the scientific goals of this study and the organization of this manuscript to the end of Sec.1
• Augment the method section – At least move some of the basic descriptions of the three statistical analyses from the appendix to the method section.
• Restructure section 3 – It is strange that there is no result section and the discussion section immediately follows the method section. Please separate results and discussion clearly. The median kappa test, for example, belongs to the discussion section instead of the direct results.
• More systematic and quantitative discussion on the uncertainty of the dataset is needed, particularly focusing on the assumption when deriving the CCN # for extreme SSw conditions. Please also discuss how the uncertainty in CCN spectra may influence the statistical analysis results.

2. I also think the title needs to be revised to highlight either the value of the CCN data set or the statistical analysis results. “Short- and long-term temporal variability” is a vague description

3. Please provide more details on the cluster analysis.

For example, what was the actual input (x) to calculate the distance metrics? The original CCN spectra (i.e. CCN # as a function of SSw) or some normalization was applied? When saying the clustering was based on distribution shape (L458), do you mean the clustered results exhibit distinct functional distributions of CCN # along with SSw? Or the CCN spectral shape was used as the input?

As the variability of the CCN spectra is orders of magnitude higher at the higher end of SSw (and similarly the variability in the aerosol size distribution is much higher at the smaller sizes), the clustered results would mainly be more sensitive to the high SSw conditions (nucleation mode aerosols). Have you compared the clustering based on the normalized spectra, the normalized aerosol size distribution, or even the fitted CCN PDF as a function of SSw?

The authors mentioned that Kappa is not the main controlling factor of CCN spectra type. Maybe the aerosol composition in SGP is similar? Please add some discussion on the aerosol composition variation in this dataset. 

4. It seemed all the “size distribution” in the manuscript refer to “aerosol size distribution”? Or do some of them actually refer to the “CCN size distribution”? Please make them more explicit.

Technical corrections/Minor Comments

1. L23 Abstract: More specific/qualitative conclusion on the time scales
2. L67: Adding a table to list the instruments used in constructing the data set
3. L95, Fig.1 Caption: blue à black
4. L218 reconciling seasonal particle data and cluster: please elaborate the sentence

Data availability and supplement link

1. L491 “(ARM), 1995, 2001, 2007, 2010, 2011, 2015)”: it’s confusing that the years mentioned here are inconsistent with the years mentioned in the main text (2009-2013)
2. L494 “merged aerosol number size distribution and CCN data (https://doi.org/10.5439/1832908)”: This DOI cannot be found in the DOI System
3. L497, " skewed log-normal fit coefficients for all CCN data, named CCN_fit_coeffs.txt, is available at XXX”. à missing information