We wish to thank the reviewer for its time spent to evaluate our work and the useful comments on our manuscript. We provide a point-by-point response below

1. L84: In the section of the Metrics, the definition of the (overall) accuracy ((TP+TN)/(TP+FP+FN+TN)) should also be added.

We will add this definition.

1. L104: “the number of false negatives is highest in December while the accuracy increases every month from January to April.” It is interesting, but 1) “the data” means the data of Finland and Norway ? or the data of all the European countries studied ? 2) What is the cause of the behavior of FN? If possible, please add the figure for the data by month to this report. 

The data refers to all european countries. In the Discussion section (L115) we attribute the behavior of the FN to the lack of direct solar radiation. We agree that it would be useful to show this figure of the results by month (figure shown below). We will see if we can incorporate it in the main text otherwise as a supplementary figure.

1. L108: In the study of Gascoin et al. (2019) the occurrence of false snow detection (i.e., FP) in some large clouds was identified as an issue to be addressed in a future release. However, the FP evaluated in this study seems not to be large compared with FN as shown in Fig. 2. Does this mean that the cloud detection using the MAJA software were improved to eliminate large icy clouds? Or originally the performance of the MAJA is good enough to eliminate icy clouds? The authors may address this issue in the text.

We agree and we were also somewhat surprised by this result. In the past two years there has been a few minor updates in MAJA which may have marginally contributed to reduce this issue of false snow detection in clouds. This study indeed suggests that the FP issue may not be the main issue to focus on to improve the product accuracy. We will add a comment about this in the main text.

We wish to thank the reviewer for the positive evaluation of useful comments on our manuscript. We provide a point-by-point response below.

1. The accuracy (proportion of correct classifications) was 94 % (κ = 0.83). Based on the parentheses, one could read that the proportion of correct classifications was 0.83, although it is probably meant that accuracy is just a synonym for the proportion of correct classifications. Please revise the sentence to avoid this risk of misunderstanding.

We will revise this sentence to clarify.

1. sensitivity to this threshold?

Indeed, it is probably better to write “to this threshold” instead of “of this threshold”

1. at least one external data set / at least one of the external data sets?

We will rephrase “at least one of the external dataset”

1. A reference to the definition of kappa would be useful.

We could have cited the paper by Cohen (1960) but we already had the maximum number of references allowed for a brief communication (20). We assumed that the kappa is now well known in the scientific community and its definition can be easily found in many textbooks or wikipedia.

Cohen, J. (1960). "A coefficient of agreement for nominal scales". Educational and Psychological Measurement. 20 (1): 37–46. doi:10.1177/001316446002000104

1. L88, L106. “Results” should be Section 3 and “Conclusions” Section 4.

We thank the reviewer for noticing this mistake.

1. Check the alphabetical order in the list of references.

We have checked and did not notice sorting error (the list was generated automatically with a citation management software)

1. L207-208 (Caption of Fig. 2). Please specify the value of HS_0 used for the right-hand-side matrices.

We will add the information.