The benefits of homogenising snow depth series - Impacts on decadal trends and extremes for Switzerland

. Our current knowledge of spatial and temporal snow depth trends is based almost exclusively on time-series of non-homogenised observational data. However, like other long-term series from observations, they are prone to inhomogeneities that can influence and even change trends if not taken into account. In order to assess the relevance of homogenisation for time-series analysis of daily snow depths, we investigated the effects of adjusting inhomogeneities in the extensive network of Swiss snow depth observations for trends and changes in extreme values of commonly used snow indices, such as snow days, 5 seasonal averages or maximum snow depth in the period 1961-2021. Three homogenisation methods were compared for this task: Climatol and HOMER, which apply median based adjustments, and the quantile based interpQM. All three were run using the same input data with identical break points. We found that they agree well on trends of seasonal average snow depth, while differences are detectable for seasonal maximums and the corresponding extreme values. Differences between homogenised and non-homogenised series result mainly from the approach for generating reference series. The comparison of homogenised 10 and original values for the 50-year return level of seasonal maximum snow depth showed that the quantile-based method had the smallest number of stations outside the 95 % confidence interval. Using a multiple criteria approach as e

1 and their effects on trends in seasonal mean snow depth. Their results showed the need for improving adjustment methods in 55 order to (i) enable the application to data with higher temporal resolution (e.g. daily data) and (ii) to improve the adjustment of extreme values. Taking up these needs, an adjustment method using quantile matching was introduced by Resch et al. (2022).
Widespread used metrics to describe the snow cover include average and maximum snow depths and days with a snow depth above a certain threshold, referred to here as snow days. This commonly used index is defined as the number of days 60 within a certain time period (e.g. season) with a certain snow depth, usually between 1 -50 cm. (Abegg et al., 2020;Schmucki et al., 2017). Snow days are relevant for ecology (Stone et al., 2002;Jonas et al., 2008), climatology (Scherrer et al., 2004;Marty, 2008) or the ski tourism industry (Abegg et al., 2020), whereas the average and maximum snow depths are particularly applicable to climatology and engineering applications. Trend and extreme value analyses of snow indices (Scherrer et al., 2013;Matiu et al., 2021) are common methods in climate monitoring (Bocchiola et al., 2008;Marty and Blanchet, 2012;Buchmann et al., 2021a) and model verification (Brown et al., 2003;Essery et al., 2013), while extreme value analyses are important for defining snow loads and limits for building-codes (Croce et al., 2021;Schellander et al., 2021;Al-Rubaye et al., 2022).
We use the break points recently identified by Buchmann et al. (2022) for manual Swiss snow series with a joint application of three widely used break-point-detection and homogenisation methods: ACMANT (Domonkos, 2011), Climatol (Guijarro,

Data
Daily manual snow depth measurements (HS) from 184 Swiss stations serve as the basis for quantifying the benefit of data 85 homogenisation for snow depth series. Seasonal (November to April) and monthly average (HSavg), maximum snow depths (HSmax), and the number of snow days >= 5 cm (dHS5) are calculated from the daily snow depths measured at 07:00 o'clock each day. For obvious reasons, daily snow depth time series inherit a strong autocorrelation. We used seasonal indicators of snow depth, which imply no to low autocorrelation with the exception of cases when the snowcover did not completely melt over the summer. However, this is neither the case for any of the selected stations nor for any of the seasons analysed.

90
The autocorrelation was analysed for the lags 1 -10. The results for lag1, where it is strongest, are shown in Table 1. The autocorrelation here is very low with a mean of 0.03 -0.18 and an interquartile range of 0.14 -0.24. Consequently, a Trend-Free-prewhitening of snow depth series (Yue et al., 2002) or the application of a modified MK-test was not necessary.  Figure 1). Only stations with complete data coverage between November and April for each year and at least 30 years of data are considered. A detailed description of the data set can be found in Buchmann et al. (2022).

The set of pre-identified break points
We use the set of 45 break points (found in 40 snow depth series) identified by Buchmann et al. (2022) for our analyses.

100
Two series (stations Bernina Hospiz and Gütsch) were removed from the original 42-station-subset due to insufficient data quality between 1961 and 2021. Break points were detected using ACMANT (Domonkos, 2011), Climatol (Guijarro, 2018), and HOMER (Mestre et al., 2013), with break points accepted as valid if detected within two years by at least 2 of 3 methods.
For details, e.g. on the differences between methods in the detected breaks and motivation for the criteria of break acceptance, see Buchmann et al. (2022). Break points are identified based on seasonal series. Where appropriate, available metadata has 105 been used for verification. However, as our metadata is neither perfect nor complete, it is only used as an additional source of information and not as stand-alone evidence. To improve the station density near the Swiss eastern border, three Austrian stations were added to the data base. Figure 1 shows the location of all 184 Swiss series in the left panel, stations with detected breaks are marked with pink triangles. The right panel shows the elevation distribution of the stations.
Each break point of a candidate series is adjusted by a multiplicative approach to the most recent status of the snow station. This is in agreement for all three adjustment methods applied. Adjustment factors are based on statistical measures of the candidate and reference series, respectively, and applied to the monthly (Climatol, HOMER) or daily (interpQM) values. These statistical measures (e.g. median, quantiles) applied for adjustment are different for the three methods and are described below. Important to know, the reference series used for adjustment by the three methods are not identical and selected on different criteria. For 115 interpQM and HOMER, they are known to the user.
All analysed methods use the same data set to select suitable reference stations for the calculation of the adjustment factors based on the pre-determined break points, which in our case are provided by Buchmann et al. (2022) and used by each method via importing a file containing the break points. Although it is possible to manually select suitable reference stations for each series and use only these for each method, we have chosen to let the methods themselves select their reference stations based 120 on their internal criteria.

Adjustment methods
Climatol (Guijarro, 2018) is a fully-automatic homogenisation method based on SNHT (Alexandersson, 1986) for break detection and a linear regression approach after Easterling and Peterson (1995) for the adjustments. It uses composite reference series that are constructed as a weighted average, using the horizontal and vertical distance between suitable reference and the 125 candidate series as weight. We used the default settings, i.e. 100 km, where the horizontal distance weight is set to 0.5 and the vertical distance scaling to 0.1. As for all adjustment methods, candidate series are adjusted back in time starting from the most recent homogeneous sub-period. Doing so, each detected break (sub-period between breaks) is adjusted applying an adjustment factor derived from annual values (see Guijarro (2018) for details). Which is calculated for Climatol as follows: The adjustment factor of a time series z is calculated as follows: Wherez b andz a are the mean snow depth between the beginning of the measurements of z and the break point (before) and from the break point to the end (after), respectively. σ Q andQ are the standard deviation and mean of the non-standardized ratio time series Q = Ref erence/Candidate (Alexandersson and Moberg, 1997).
HOMER (Mestre et al., 2013) is an interactive semi-automatic toolbox that provides various methods for detection and 135 adjustment of breaks, such as pairwise comparison (Caussinus and Mestre, 2004), a fully automatic detection and correction joint-segmentation (Picard et al., 2011) and ACMANT-detection (Domonkos, 2011). For our purposes, the pairwise comparison was chosen, as it accepts the use of independently derived break point metadata-files, like Climatol and interpQM. The series are adjusted with a single annual factor for the entire period before a break point. The adjustment factor is derived from variance analysis ANOVA (Caussinus and Mestre, 2004) based on the selected reference stations. These are defined either on the basis 140 of the horizontal distance or the first-difference correlations. Due to the large vertical distances between stations, even for short horizontal distances, the latter was chosen with a minimum Spearman ρ of > 0.8. (2) Where O ij is the matrix of original time series j with time index i,v C * jh(i,j) the estimation of the correction for a set of breaks per candidate station C j in a homogeneous subperiod h i,j andv C * j,kj +1 the estimation of the adjusted station signal with the 145 number of break points of a a station k j .
InterpQM (Resch et al., 2022) is an extension of INTERP (Vincent et al., 2002), that uses quantile matching to improve the adjustments, taking into account the frequency distribution of the daily values to be adjusted. It provides homogenised data on a daily basis, which then allows the analysis of the subsequently derived snow indicators. For this purpose, the daily measurements of the candidate and reference series are split into two interquantile subsets (IQS), which are then compared.

150
An adjustment factor is calculated for each IQS and then linearly interpolated between neighbouring IQSs to avoid artificial jumps in the data.C andR are the median of the daily time series of the candidate/reference station before (b) or after (a) the detected break point to be adjusted. The reference series can either be selected manually or be a composite series calculated from a weighted average 155 of selected stations (< 100 km horizontal and < 300 m vertical distance, > 0.7 correlation, no detected breaks), which was chosen here. The selection can be manually refined and optimised, using local knowledge and experience. The distribution of weights between these stations can either be exponential or linear. To reduce the strong influence of individual highly correlated stations on the result, a linear distribution of the weighting was chosen. Break points are derived from a pre-defined break point file.

Detection of trends and changes in snow depth series
The use of homogenisation techniques that adjust daily values allows the analysis of the impact on derived indicators that require daily data for their calculation, e.g. snow days. Only the original data and interpQM are compared here, as HOMER only provides monthly or seasonal data and Climatol kept crashing when using the full daily data set. Since we did not want to pre-select stations as this would influence the results, we decided not to use it for this purpose.

165
InterpQM does not add new days with snow (HS > 0). To avoid a possible negative bias and because almost no changes were expected for the homogenised series of days with HS > 1 cm, the threshold values of 5, 30 and 50 cm, are clearly more meaningful. Snow days are accumulated based on a temporal reference between November and April each year (hydrological year). Trends are determined using the standard non-parametric Mann- Kendall trend test (Mann, 1945;Kendall, 1975) and are 170 considered significant if they are above the 95 % level.

175
To investigate the effects of homogenisation on extreme snow depths, return levels for the seasonal maximum snow depth (HSmax) (Marty and Blanchet, 2012) are calculated for a fixed return period of 50 years (Buchmann et al., 2021a; Blanchet, 2012) based on original and homogenised data (R50HSmax). This approach was chosen because the international standards for maximum snow load on buildings are based on R50HSmax (see e.g. Schellander et al. (2021)). The calculations were performed with the R package extRemes (Gilleland and Katz, 2016) in default settings (GEV, estimation method MLE, 180 and 95 % confidence intervals). In order to determine to what extent homogenised and original time series differ in their distribution and to assess the differences between the results of the applied adjustment methods, a two-sample Kolmogorov-Smirnov (in the following referred to KS-test) and the non-parametric two-sample Wilcoxon test (in the following referred to as W-test) were performed with seasonal data for all derived indices.

185
We use the subset of 40 stations with identified breaks as input and adjust them with the three methods. artificially increased length of these series. It also automatically adjusts outliers in the homogeneous period in the default settings.
In the following section, we compare the results of different adjustment methods on the one hand and the homogenised data with the non-homogenised data on the other. In this way we can show both the effect of homogenisation and the dependence 200 of the results on the method used. In chapter 4.1 we show this as an example for the number of snow days and in particular for the effects on the trend (for interpQM only). Similarly, this is also shown for the maximum snow depth in chapter 4.2 Finally, in chapter 4.3 a particular example is given for the magnitude and frequency of extreme snow depth.  Positive values indicate a lower positive or a more strongly negative trend after homogenisation.

Trends of snow days
The number of snow days per season was examined for two subgroups of stations, below (n = 26) and above (n = 14) 1500 205 m a.s.l., referred to in this chapter as "low elevation" and "high elevation" stations. This threshold was also used by e.g. Auer et al. (2007). Additionally, a strong decrease in snow depth between 1500 to 2500 m was determined for the coming decades (Marke et al., 2015;Marty et al., 2017). This makes this elevation range interesting for analyses of changes that have already taken place. We analysed thresholds of 5, 30 and 50 cm (dHS5, dHS30, dHS50) for the original and homogenised data.
The adjustments made had the strongest effect on dHS30 and dHS50 at stations above 1000 m, as can be seen in Figure 2.

210
The percentage of significant negative time series has increased for all indices above 1500 m and dHS5 below 1500 m while it was reduced by 3 % for stations below 1500 m for dHS30 and by 33 % (from 86 to 53 %) for dHS50. The difference between the trend strength of the original and homogenised time series was more than one day/decade at six out of 40 stations for dHS5, at 21 for dHS30 and at 26 for dHS50. Negative trends were strengthened at 5 stations for dHS5, 9 for dHS30 and 11 stations, while they were weakened at 1 station for dHS5, 12 for dHS30 and 15 for dHS50. To detect significant differences between 215 the original and homogenised time series, the non-parametric Wilcoxon test was applied. As can be seen in Figure 2, this was only the case at the Adelboden station for dHS50.
The number of snow days per season is declining for the vast majority of stations for all analysed thresholds and elevation levels, as shown in Overall, the homogenisation removed all positive trends and, depending on the threshold for snow depth and elevation subset, either did not change or reduced the number of stations without trends: E.g. 86 % of the high elevation stations had a negative trend for dHS30 before, and 100 % after the homogenisation. The percentage of low elevation stations with no trend for dHS50 changed from 42 % to 35 % after the homogenisation, while the percentage of stations with a negative trends was raised from 225 54 % to 65 %.
In general, the adjustments changed the median and mean trends of both subsets for dHS5 and the higher elevation subsets for dHS30 and dHS50 to more negative, the lower elevation subsets of dHS30 and dHS50 to less negative. The mean trends of the lower elevations changed from -5.6 to -5.9 days/decade for dHS5, from -5.7 to -4.9 for dHS30 and from -3.7 to -3.5 for dHS50. For higher elevations they changed from -3.3 to -3.4 for dHS5, from -4.3 to -4.7 for dHS30 and from -5.4 to -5.8 for 230 dHS50.
The percentage of low-elevation stations with no trend is different for the larger thresholds than for dHS5, where it increases from 0 to 7 % with increasing altitude, but decreases for both dHS30 (from 19 to 0 %) and dHS50 (from 42 to 0 %  For dHS50, the trends of 10 stations were changed to non significant and of 12 to significant. The KS-test did not reveal significant differences between the original and the interpQM-adjusted time series in the distri-245 bution of the dHS5-, dHS30-or dHS50-time series for any of the stations analysed. A comparison with the W-test also showed no significant differences for dHS5 and dHS30, but at one station (Adelboden) for dHS50.

Trends of mean and maximum snow depth
The effect of homogenisation on the mean (HSavg) and maximum snow depths (HSmax) is illustrated using the example of Davos in Figure 3. The adjustments made increased the seasonal mean snow depth before the break in 1972 between 2 -11 cm 250 with interpQM, 3 -17 cm with Climatol and 3 -18 cm with HOMER. The impact on the seasonal maximum snow depth range from 2 -19 cm with Climatol, 7 -23 cm with interpQM and 7 -26 cm with HOMER.
To assess the impact of homogenisation on trends of HSavg and HSmax, decadal trends are calculated for each homogenisation method and the original data respectively. Figure 4 shows the trends for HSavg in the left and for HSmax in the right panel.

255
Trends are expressed as cm/decade for the period from 1961 to 2021 for each method and station of the non-homogeneous subset, black dots indicate significant trends. For HSavg, we found an overall similar pattern across the methods. Figure A3 shows  Table 3 describes the mean and median trends across all stations, as well as the change from positive to negative and significant to not significant and vice-versa for both HSavg and HSmax. The mean trends of HSavg for Climatol and HOMER appear to be weaker than for the original and interpQM homogenised.
265 Figure 4 further reveals that the homogenised trends for HSavg mimic the pattern of the original trends, which shows almost zero trends for stations below 500 m, strong negative and significant trends for the group between 1000 and 1500 m, followed by mostly not significant trends for stations between 1500 and 1600 m a.s.l. This suggests that the various intrinsic ways of building reference series and sub-networks of the underlying homogenisation methods do not have a significant impact on decadal trends of HSavg.

270
The vast majority of trends for HSmax, 37 of the original series and 39 for all homogenisation methods, show a negative trend, as shown in the right side of Table 3 for details. The number of significant trends is about 20 % lower than for HSavg, with interpQM showing the largest and HOMER the lowest number of significant trends. The most striking difference between the patterns of HSavg and HSmax in Fig. 4 is the area without significant trends. This is between 1500 and 1600 m a.s.l. for HSavg for all homogenisation methods, and below 1000 m a.s.l. for HSmax with the exception of time series adjusted by 275 interpQM. There seems be no particular altitudinal pattern, except that the trends below 1000 m a.s.l. are weak for all methods and increase in strength between 1200 and 1400 m a.s.l. This suggests that the trends for HSmax, in contrast to HSavg, appear to be more sensitive to the underlying homogenisation methods. Filisur (1030m) Göschenen (1110m) Disentis (1190m) Klosters KW (1190m) Gsteig (1195m) Flumserberg (1310m) Meien (1320m) Adelboden (1325m) Unterwasser Iltios (1340m) La Comballaz (1360m) Sta.Maria (1415m) Saanenmöser (1390m) Urnerboden (1395m) Sedrun (1432m) Splügen (1457m) Simplon Dorf (1470m) Zernez (1470m) Davos Flüelastr. (1560m) Grindelwald Bort (1565m) Montana (1590m) Rigi Scheidegg (1640m) San Bernardino (1640m) Mürren_MCH (1641m) Mürren_SLF (1650m) Samedan (1726m) Zervreila (   Percentages for significant negative and significant positive, indicated with an asterisk, are calculated based on the total number of negative/positive values respectively.

Impact on extreme snow depths
To investigate a possible influence of the homogenisation on the magnitude and frequency of extreme snow depths, the absolute maximum snow depths (maxHSmax) recorded at each station over the entire period, the year with the absolute maximum snow depth and the difference between original and homogenised maxHSmax are plotted for each station and homogenisation method. Figure 5 shows the results. Here we found that for the majority of series, the differences are 0. The differences are 295 generally left-skewed, except for the largest differences observed with Climatol (Panel (d) of Figure 5). This again suggests that, in contrast to the trends for HSavg, the differences between methods are more apparent for HSmax. Furthermore, Panel (c) of Figure 5 clearly highlights the four known snow-rich winters of 1951, 1968, 1975, and 1999.
The return levels for 50-year return periods of maximum snow depth (R50HSmax) are calculated from homogenised data and compared with the values obtained from the original data including the 95 % confidence intervals. Figure 6 shows the original 300 values in grey with the associated 95 % confidence intervals and the homogenised values in colour. A pattern was found to occur in all methods for the majority of stations. For Climatol, seven stations are above the 95 % confidence intervals of the original values for R50HSmax and six below, for HOMER there are four above and seven below, while for interpQM there are three above and three below, see Table 4 for details. This again suggests that the differences between the homogenisation methods are more pronounced for R50HSmax than for trends of HSavg, with interpQM performing slightly better than Climatol
In contrast to the larger thresholds of the snow day analysis, dHS5 shows almost no differences between the original and the homogenised series, confirming the stability of this metric as described by Buchmann et al. (2021b). The elevation-dependent 315 pattern with the strongest adjustment effects for dHS30 and dHS50 between 1000 -1700 m can be explained by the fact that, firstly, at stations below 1000 m a.s.l. there are few days with a snow depth of 30 cm or more due to the generally warmer temperature and lower snowfall amount and, secondly, that above 1700 m winter temperatures are low enough and therefore less sensitive to warming in winter so that the trends are smaller. A similar pattern can be seen in the absolute values (Appendix A1). Those high-elevation stations that show large differences in the trends before and after the homogenisation in Fig. 2 (Sils-320 Maria, Samedan, San Bernardino, Zernez, Simplon Dorf, and Splügen) are all located at sites strongly influenced by southerly flows. In particular, the Engadine in the southeast, a high elevation inner alpine valley with a dry and cold climate, is often not associated with large snow depths or many days with dHS30 or dHS50.
All but two of the trends for HSavg (in both the original and homogenised data) are negative, which is consistent with the 325 findings from previous snow studies (Laternser and Schneebeli, 2003;Marty, 2008;Scherrer et al., 2013;Fontrodona Bach et al., 2018;Matiu et al., 2021). Marcolini et al. (2019) report an increase in series showing significant trends for HSavg after homogenisation (40 to 44 %). The same effect is observed here for interpQM, but not for Climatol (no change) and HOMER. Both show a decrease in significant negative trends after homogenisation (Table 3). The same increase in the number of significant negative trends is observed for snow days and HSmax. The adjustments decreased the snow depth prior to a break 330 at 55 % and increased it at 45 % of the stations.
For most stations, the R50HSmax of the homogenised data are still within the 95 % confidence intervals of the original values. However, depending on the homogenisation method, between three to seven of the investigated 40 stations (see Table   4) have R50HSmax that exceed the original values beyond the 95 % confidence intervals, with potential implications for engineering applications and building codes. Values that are significantly above the 95 % intervals are predominantly from 335 Climatol. The reference networks in Climatol are created using the Euclidean distances between candidate and reference series, with an optional scaling-factor for the vertical component. We set this threshold to wz = 100 to avoid the selection of stations that are close together horizontally but far apart vertically, e.g. the stations Davos (1570 m a.s.l.) and Weissfluhjoch (2535 m a.s.l.), which are only 4 km apart horizontally. However, it may also be that this threshold is simply not low enough to prevent further station combinations with a similarly large gradient. Unfortunately, the user cannot see in Climatol which homogenisation indicates a decrease in variation and an increase in confidence in the results for very large snow depths, as shown with the absolute maximum snow depth.
The observed differences between the three methods compared can be explained by the respective methods used to construct the reference series sub-networks and the adjustments. HOMER adjusts the entire period before an identified break point 345 using a single factor, while Climatol uses multiple factors dependent on the reference series constructed using homogenised sub-periods. InterpQM, on the other hand, uses multiple adjustment factors based on quantile matching for the entire inhomogeneous period, similar to HOMER. The range of the applied adjustments for interpQM is shown in Appendix A4.
The selection of suitable reference series is the crucial part of the homogenisation procedure, both for the detection of breaks The analysis of the sub-networks for HOMER and interpQM shows that due to the distance restriction in interpQM, reference series are drawn from a more similar region, whereas in HOMER distant stations with high correlations are frequently included.
To avoid selecting close-by, but unsuitable reference series due to local climatic variations, the correlation criterion in interpQM works well.

365
Both Marcolini et al. (2019) and Buchmann, M. et al. (2022) found that relocations where responsible for by far the most detected break points in snow depth time series. The metadata of many stations are sparse and therefore often do not provide enough information to give a sufficient answer to the question, why a relocation caused a break. A change in elevation within +/-150 m is not necessarily a cause of a break, but moving a station either below or above the typical height of a site's inversion is highly likely. Significant changes in the station environment are also very likely to cause a break, e.g. moving a station to an with a snow depth above a certain threshold (5, 30 and 50 cm), the seasonal mean and maximum snow depths (HSavg,375 HSmax) and extreme snow depths. The results underpin the relevance of homogenising long term snow depth series for trend and extreme values analysis. Due to the impact of homogenisation on derived trends, this is especially true for conclusions drawn from individual series. In our analyses, for the long-term trends of HSavg and dHS5, the overall picture does not change through homogenisation of original data by median/mean based adjustment methods. However, the picture becomes different when a quantile-based homogenisation approach (interpQM) is applied, which in the case of Swiss snow depth series,

380
shows the strongest effect with only negative trends for HSavg and a slight increase in the number of significant trends.
The differences between the methods increase when looking at seasonal maximum values: The trends for HSmax, where trends of low elevation stations were significant only with interpQM, absolute maximum snow depths and extreme values.
The homogenisation performed with interpQM increases the confidence in the derived extreme values based on the 95 % confidence interval, which is particularly relevant for engineering applications. As far as snow days are concerned, the quantile 385 based adjustments had the strongest impact on the larger snow depth thresholds.
Our results support a homogenisation approach that separates the break point detection from the adjustment procedure, e.g. to use the robust combined detection approach described in Buchmann et al. (2022) in combination with the adjustment procedure from Resch et al. (2022). However, the ability to manually adjust the automatic selection of the reference (subnetwork) stations used for homogenisation is crucial for optimising the results. A combination of criteria such as correlation, 390 horizontal and vertical distances as well as manual interventions seems to be more advantageous (given the complex topography in mountain regions like the Alps) for snow depth than the use of a single selection criterion.
So far, the homogenised snow depth time series show no evidence of a bias in the methods towards increasing or decreasing snow depths due to the adjustments made, neither in Austria nor in Switzerland. In this study, depending on the homogenisation method, the mean snow depth before a break was increased at about 52 -57 % of the stations and decreased at between 42 -395 45 %. 95 % of the 40 inhomogeneous stations show a negative trend for seasonal mean snow depth in the original data, which is significant for 58 %. These figures are lower for the 144 homogeneous stations in the data set, where 78 % show a negative trend that is significant for 50 %.
As pointed out, break detection for snow depth is preferably done using the described two-out-of-three method. From our experience, there there is no incentive or advantage to use automatic homogenisation methods such as HOMER and Climatol.

400
On the contrary, automatic methods open the door to unintended automatic outlier corrections or adjustments based on the selection of reference series that are sufficiently correlated but cannot be assigned in a climatologically meaningful way. To achieve reasonable results, these methods require a certain degree of user intervention, e.g. the use of a predefined selection of reference stations, thresholds for correlation, horizontal and vertical distances. Therefore, it seems promising to separate the detection and adjustment of breaks using the the described two-out-of-three method for detection and interpQM for the 405 adjustment, as it provides reliable results especially for larger snow depths and yields daily data.