“Synoptic control over winter snowfall variability observed in a remote site of Apennine Mountains (Italy), 1884–2015”

a long series of from the southern in relate teleconnection a newly digitised of of

For future works, we are planning to rescue other climatological time series in Campania Region that may include information about snowfall occurrence. Therefore, we will probably able to fill the existing gaps in Montevergine time series only in terms of daily snowfall occurrence, but not in terms of daily snowfall amounts. We will better clarify these aspects in the revised version of our manuscript. Table 1: How did you define these periods? Just taking 23yr periods? Why exactly these years? I do not think it's a good idea to create these groups, since they might or might not include and exclude relevant points in the time series. If you want to discuss interannual variability and longterm changes, I suggest employing moving window averages (for long-term changes) and moving window standard deviations (for interannual variability). A period of 20 or 30 years would make sense. This would not have the "problem" of arbitrarily defining year groups.

AR (2):
Following the referee comment, we have revised the Figure 5 of our manuscript. Note that this Figure is now labelled as Figure 4 after the revision process. More specifically, in order to highlight the long-term changes, we have used the moving window average (with a time span of 20 years), whereas to better emphasize the interannual variability we have computed (and plotted) the moving window standard deviation.  Table 1.
The subdivision of the investigated time interval into sub-periods of 23 years allowed us to emphasize the strong reduction in snowfall amount observed in the period from mid-1970s to the end of 1990s. According to the reviewer's suggestion, we have segmented the time series into more customary 20years intervals. It should be pointed out that this choice reduces the last sub-period (2004/05-2019/20) to a length of 16 years. We have revised the Table 1 of our manuscript as follows. Moreover, we also modified the Figure 9 of our manuscript, which is now labelled as Figure 8. All these changes will be included in the revised version of the manuscript. Figure 8 and related text: Besides the issue with year groups (see comment above), I think it would be much easier if you showed scatter plots of STx versus HNS, instead of trying to compare time series across pages. You could also calculate correlations between these two to give more weight to what you identify from "visual inspection". This would make it easier for readers to see your points.

AR (3):
We agree with the reviewer remarks. However, the scatter plot between STx and HNS is not a good solution to show the correlation degree between the two variables. The frequency of occurrence of ST, in fact, has a behaviour very similar to a categorical variable. This causes overlapping problems, so in some scatter diagram there are tenths of values all stacked on top of each other. For this reason, we have decided not to modify the Figure 8 of our manuscript (which is now labelled as Figure 7). According to the reviewer's suggestion, we have computed the linear correlation coefficient for each time periods presented in the new version of Table 1 (except for the last period, which reduces to 2004/05 to 2014/15). The results are presented in a following Table, which will be numbered as Table 3 in the revised manuscript. Note that the significance of the correlation coefficient has been tested through the well-known p-value method. In the revised version of the manuscript, we will clarify this aspect and we will discuss the results of the correlation analysis presented in the Table 3.

RC (4):
Discussion is missing completely. Regarding the snowfall series: How do your results compare to other long-term series from Italy, such as Parma or Torino? If I remember correctly they have been published, but possibly not in international journals.

AR (4):
In the revised version of the manuscript, we will add the following Discussion section, in which we compare our results with Parma and Torino time series. Thank you for the suggestions.
"The variability over the time of the total winter HNS recorded in MVOBS presents some similarities with evidences provided by past research carried out in the alpine region. The first common point lies in the strong interannual variability, which many authors reported in their analysis (e.g. Schöner et al., 2009;Scherrer et al., 2013;Terzago et al., 2013). The patterns emerged from the behaviour over time of MVOBS signal are generally in line with those identified in some previous studies. In this respect, it is important highlighting that the strong reduction in snowfall amount and frequency of occurrence occurred in MVOBS in the 1990s and the subsequent recovery in 2000s have been also observed in Switzerland, France, western Italian Alps and Austria (e.g. Laternser and Schneebeli, 2003;Micheletti, 2008;Vault and Ciafarra, 2010;Scherrer et al., 2013;Marcolini et al., 2017;Matiu et al., 2021). However, it should be noted that in most of the investigated alpine sites the decline in snowfall amount, as well as in NSD, occurred after 1980s, whereas in MVOBS it began in the mid-1970s. Moreover, it is very interesting highlight that the maximum in snowfall amount found in MVOBS time series in 1900-1920 period (see Table 1) has been also detected in Switzerland by Scherrer et al., 2013. Unfortunately, aside from the Alpine region, the available literature does not offer many other terms of comparison in the Italian territory. In this respect, two examples are the long-term nivometric time series collected in Parma and Turin, in the Po plain. The former has been analysed in Diodato et al. (2018) for the 1777-2018 period. The authors reported a decline in snowfall frequency of occurrence, mainly in the first half of 19th century, as well as in the annual length of the snow season, attributing these changes to large-scale circulation patterns and in particular to the NAO index. However, no significant trend has been detected for the amount of fresh snow, according to the available data (1868-2018 period). The Turin snowfall series has been investigated by Leporati and Mercalli (1993). The latter detected a very strong interannual variability both in terms of NSD and snowfall amounts, similarly to the results achieved for MVOBS. The main relevant dissimilarity lies in the above-than-normal snowfall amount measured in 1981-1987 period, which are in contrast with the evidence provided not only by MVOBS, but also by many other Alpine monitoring stations. The synoptic patterns identified in our work exhibit some analogies, but also some differences, with other synoptic types related to snowfall events in Europe. For example, Merino et al. (2014), using a methodological approach based on a multivariate statistical analysis (including both PCA and CA), found four different synoptic types associated with snowfall events on the northwestern Iberian peninsula. The first one is associated with a maritime arctic air advection over western Mediterranean region, the second one with the advection of polar maritime air masses, the third one with the incoming of polar continental air masses over western Mediterranean area, whereas the fourth-one is characterised by a closed cyclonic circulation over the Iberian Peninsula, which produces strong thermal gradient. The second and the third patterns have several commonalities with the ST6 and ST4 of our works, respectively, both in terms of involved air masses and in the upper level flow. The first and the fourth circulation types discusses in Merino et al. (2014) are unfavourable to snowfall events in the southern Apennines area, although the former traces out a large-scale configuration that promotes the incoming of maritime arctic air mass over Mediterranean basin, by analogy with the ST1 of our study. Moreover, it is also interesting to compare our results with the study of Esteban et al. (2005), which extracted seven different circulation patterns that explain heavy snow precipitation days in Andorra (Pyrenees). Three of these patterns represent an advection from northwestern of polar maritime air masses which resembles the large-scale flow depicted by ST6, whereas other three types have some common points with ST1, ST2 and ST4, showing scenarios characterised by advection of Atlantic or Mediterranean air masses combined with the outbreak of cold air from northern and eastern Europe. There is only one situation, characterised by a low-pressure area northwestern Spain, which strongly departs from scenarios that trigger snowfall events in Southern Italy Apennines. According to the results of our analysis, it is very reasonable ascribe the negative snowfall amounts and number of events anomaly observed between 1970s and 1990s to the increase in NAO and AO indices values, which cause a reduction of the occurrence of some synoptic patterns, mainly ST1 and ST2, very favourable to the incoming, in Central Mediterranean area, of cold air masses of maritime (polar or arctic) origin. This achievement is in accordance with the findings of Merino et al. (2014), which attribute the decrease in the number of snow days observed in Castilla y León region (Spain) to the increase in the NAO index during winter months throughout the second half of the 20th century. The impact of NAO and AO anomalies was mitigated by the incidence of ST4 and ST5, which remains quite stable due to the occurrence of some periods characterised by positive values of EMP and SCAND indices. The increase in interannual variability of snowfall events detected in the last two decades, as well as the rise in the average amount, can be attributed to large-scale conditions more beneficial for cold outbreaks in central Mediterranean regions, as well represented by rising in frequency of negative AO patterns and by the occurrence of winter seasons modulated by positive EMP and negative EAWR."

Minor Comments
RC (1): L11: mismatch of period wrt to title. OK, later I understood. The snowfall series ends 2020, but the reanalysis 2015, right? You should clarify this and be clearer.

AR (1):
Yes, it is right. In the revised version of the manuscript, we will specify that the cluster analysis has been applied to 1884-2015 according to the availability of reanalysis data.

RC (2):
L37ff the literature review is a bit random. It's mixing snow cover parameters (depth, fresh snow, SCD) and it's not clear why the authors chose the specific geographic limit. Btw, there are many other studies from Italy and other countries in the Alps. Also more with century long series. It's not necessary to mention all, but maybe the authors could make their point better.

AR (2):
Following the referee's suggestion, we have broaden our literature review as follows, including several other references, mainly related to the alpine region.
"The study period of the climatological researches carried out in such works is generally limited to the last 55-65 years. However, few studies (mainly focused on the Alpine area) extended their analysis further back, due to the absence, in many areas, of reliable and continuous old snowfall climatological records. In this respect, it is deserving of mention the study of Scherrer et al. (2013), which investigated the snowfall variability observed in Switzerland during the 1864-2009 period, using nine different stations. According to the findings of this work, the analysed depth of snowfall time series exhibit a strong decadal variability. The highest value in depth of snowfall and days with snowfall occurred in 1900-1920 and 1960-1980 period, the lowest between 1980s and 1900s, whereas an increase in depth of snowfall has been observed in 2000s. Another important reference for mountainous areas is Laternser and Schneebeli (2003), who discovered, for the Swiss Alps, an increase in snow cover and duration from 1930 to early 1980 and, subsequently, a statistically significant decrease towards 1999. Some years later, Beniston (2012b)   To extend, from both quantitative and qualitative perspectives, the current knowledge about the past snowfall variability observed in Mediterranean mountainous sectors;  To shed light on links between large-scale atmospheric circulation and local climate variability, by identifying and analysing the synoptic patterns favourable to winter snowfall events in Montevergine as well as their relationship with the main teleconnection indices that govern the atmospheric circulation in the Mediterranean area." RC (5): Figure 5: why did you choose a lowess smoother? Would a simple moving average (10/20/or 30 years) be easier? For the lowess, you also need to supply the degree and the weights, not only the time span.

AR (5):
We have chosen the lowess smoothing because it generally works better than moving average at the edges of the time series. However, following the referee suggestion, we have replaced the lowess smoothing with the moving average (See the Figure 4 in the previous comment).

RC (6):
How are "snow days" (NSD) defined? This would belong in the methods. (related Table 1, Figure 5, …) AR (6): Usually, a "snow day" is a day on which accumulated snowfall (i.e. daily high of new snow, HNSd) is at least 1.0 cm. However, in our work, in applying the cluster analysis (CA) we have used a slight different definition of "snow day". More specifically, we have considered as "snowy" a day in which the recorded HNSd value was at least 3.0 cm. This threshold allows filtering out most of some ambiguous events, characterized by the simultaneous presence of different hydrometeors types (i.e. rain, snow hail or graupel).
In the revised version of our manuscript, we will better clarified this point.

RC (7)
: L409: how did you determine statistical significance of trends? AR (7): Sorry for missing this important detail. To compute the statistical significance of trend, we used the Mann-Kendall test (Mann 1945, Kendall 1962). We will clarify this aspect in the revised version of the manuscript.

AR (9):
Our main aim is to search for relationship between the identified synoptic patterns and the teleconnection indices. However, we have welcomed the suggestion of the referee and we have computed the linear correlation coefficient between the winter HNS time series and the teleconnection indices. As demonstrated by the following table, the correlations are generally low, except for AO index, which is positively correlated to HNS time series at 95% significance level. We will include this Table in the revised version of the manuscript (Table 5).  Figure 11: Hovmöller plots do not work for a discrete x-axis, where you have the five teleconnection indices. Would a simple correlation analysis not work better here, too?

AR (10):
We understand the doubts of the reviewer with respect to the Figure 11 of our manuscript (which is now labelled as Fig. 10). However, we feel that this picture give a complete, comprehensive and simple, even though qualitative, representation of the relationship between the analysed teleconnection indices and the HNS time series. In our opinion, a linear correlation analysis, in which the indices are considered separately from each other, does not say much about the linkages with the nivometric regime of the site of interest. A possible approach to investigate about the relative influences of the indices on the winter HNS may be a multiple linear regression analysis (e.g. Cohen et al., 2013). However, a quantitative and in-depth evaluation of this aspect is left for a future work. Therefore, we decide to leave unchanged the Figure 10 in the revised version of the manuscript and to include the linear correlation analysis suggested by the reviewer (see the previous comment).