Unprecedented snow-drought conditions in the Italian Alps during the early 2020s

Snow represents a fundamental water resource for mountain and lowland areas. Changes in the frequency and magnitude of snow droughts can significantly impact societies and ecosystems that rely on snowmelt to satisfy their water demands. Here we documented and quantified the snow drought that affected the Italian Alps during the early 2020s. We used 15 long-term snow-depth series (period 1930–2023, elevation range: 864–2200 m a.s.l.) to simulate the snow water equivalent (SWE), in conjunction with climatic reanalysis data and river discharge observations. We found that the March SWE anomaly in 2022 reached the lowest value in the last century, due to an unprecedented combination of drier- and warmer-than-normal conditions in the period December 2021–March 2022. This event contributed to causing critical hydrological conditions in the Po and Adige rivers which, during summer 2022, experienced the worst hydrological drought ever recorded. Despite its unprecedented magnitude, the snow drought in 2022 is part of a recent pattern of increased intensity and frequency of snow-drought events since the 1990s, due to the combined increasing occurrence of warmer- and drier-than-normal climatic conditions during the snow season. Remarkably, three out of the five most severe snow-drought events occurred in the last five years, with exceptional snow-drought conditions even occurring in the last two consecutive winters, 2022 and 2023. The snow-drought conditions that occurred in the early 2020s in the Italian Alps warn of the pressing need for the implementation of impact mitigation measures to adapt to the fast changing snow and climatic conditions.


Introduction
Drought can have dire impacts on human activities and natural ecosystems (e.g. Stephan et al 2021). In the last two decades, different areas of Europe have been affected by numerous severe, large-scale drought events (in 2003, 2007, 2011, 2012, 2015, 2017Blauhut et al 2022). In particular, the Mediterranean area has shown the highest drought frequency and severity since the early 1990s (Spinoni et al 2015 and droughts are projected to become even more recurrent and intense (Spinoni et al 2018, Baronetti et al 2022, with profound economic consequences (Naumann et al 2021).
Despite a growing concern about drought hazard, snow droughts (i.e. snow water equivalent (SWE) deficit; Pederson et al 2011, Harpold et al 2017 remain largely unexplored compared to other drought types (Hatchett andMcEvoy 2018, Huning and, such as meteorological/climatological (precipitation deficit), agricultural (soil moisture deficit) and hydrological (deficit in river flow and groundwater, e.g. Van Lanen et al 2016) droughts. Drought conditions and recordlow snowpacks in the United States in recent decades/years (Mote et al 2018, Iglesias et al 2022 provided incentive to study this emerging phenomenon (Hatchett and McEvoy 2018).
Although mainly investigated in the United States, recent global research showed that, in the last few decades, the frequency and duration of snow droughts have increased in Europe, which emerged as a hotspot for this kind of drought . This evidence has been corroborated in Italy by the analysis of the long-term variability of simulated SWE across the Italian Alps (Colombo et al 2022), with a reported twofold intensification of snow-drought events during the past three decades.
In the last months of 2021 and throughout most of 2022, prolonged drought conditions affected the Mediterranean region and Western Europe (Faranda et al 2023). In Northern Italy, the water year 2022 (1 October 2021-30 September 2022) was marked by record-low precipitation and record-high air temperature. This occurrence resulted in a severe hydrological drought; agricultural and energy sectors were deeply affected by the very low discharge of several rivers, which was not sustained by the spring snow melt (Toreti et al 2022a(Toreti et al , 2022b. In winter 2023, most of southern and western Europe was affected by exceptionally dry and warm meteorological conditions, showing similar patterns with respect to 2022 (Toreti et al 2023).
Here we took advantage of a unique dataset of continuous, historical daily snow-depth observations (period 1930-2023) to simulate long-term SWE, in conjunction with climatic reanalysis data and river discharge observations. We then used a stakeholder-oriented approach, based on anomaly indices, to document and quantify the unprecedented snow-drought conditions that affected the Italian Alps in the early 2020s. We used normalised indices which, moving from absolute values to the frequency domain, allow to: (i) compare time series collected from different locations (and possibly affected by local and regional factors), also favouring the comparison between different drivers (Colombo et al 2022); (ii) share information (Romano et al 2018), crucially supporting decision makers in assessing snow-drought conditions.

Study region and snow-depth data series
The Italian Alps roughly correspond to the southern European Alps, defined as the mountainous area south of the main Alpine watershed (Brugnara and Maugeri 2019). The Italian Alps play a crucial role in feeding the main river basins of Northern Italy, namely Po, Adige, and Piave (figure 1). According to previous studies (Coppola et al 2014, Avanzi et al 2023, the snowmelt can contribute to up to 50% of the annual stream flow of the main rivers originating in the Italian Alps. In Northern Italy, economic activities account for more than 40% of the annual Italian gross domestic product and more than 35% of the Italian agricultural production comes from this area (Bozzola and Swanson 2014).
In this study, we used continuous daily measurements of snow depth (HS) retrieved from 15 manual and automatic stations (mostly manual measurements), covering the period between 1930 and 2023, across an elevation range of 864-2200 m a.s.l. (table 1) HS data were compiled by regional and provincial AINEVA avalanche services (Interregional Association for coordination and documentation of snow and avalanche problems), together with regional environmental protection agencies (ARPA Piemonte, ARPA Veneto), Centro Funzionale Regione Valle d'Aosta, Meteotrentino, Italian Meteorological Society (SMI), hydropower companies (CVA S.p.A., Enel S.p.A.), and Ministry of Public Work (hydrological annals). Data quality control, homogeneity assessment, and management of data gaps in the dataset are described in Colombo et al (2022), where the dataset was analysed in the time-span 1930-2020. In the present study, we updated the previous historical dataset with the winter seasons 2020/2021, 2021/2022, and 2022/2023. Missing values affected approximately only 1% of the entire dataset, with a maximum value of 3.2% for single series.

SWE modelling
Since no long-term SWE series exist, SWE has to be modelled (Winkler et al 2021, Colombo et al 2022. Physically-based models are able to resolve mass and energy exchange, although the amount and quality of required input data significantly limit their  Thus, in this study, we used a new semiempirical multilayer model for simulating SWE (and bulk density) from a regular time series of HS: the ∆SNOW model (Winkler et al 2021). Winkler et al (2021) calibrated the model using snow observations recorded during 67 winters at 14 stations, well-distributed over different elevations and climatic regions of the European Alps. Moreover, the authors used data from 71 independent winters from 15 stations for validation. The elevation range of the calibration/validation sites was between 590 and 1780 m a.s.l.
Here, we used the calibration proposed by Winkler et al (2021), given the suitability of the parameter setting for modelling SWE from daily HS time series in Alpine areas. Colombo et al (2022) validated the ∆SNOW model in the Italian Alps using four stations located between 1250 and 2100 m a.s.l., across the period 1983-2020 (501 observations). Modelled SWE had an overall Pearson correlation (R), mean error (ME), mean absolute error (MAE), and root mean square error (RMSE) of 0.96, 6.3 mm, 39.9 mm, and 56.2 mm, respectively. In this study, we updated the calibration of Colombo et al (2022) by adding the measured SWE values from the winter seasons 2020/2021, 2021/2022, and 2022/2023 (36 observations), obtaining updated R, ME, MAE, and RMSE values of 0.96, 7.6 mm, 37.1 mm, and 55.7 mm, respectively. The results indicated a rather good performance in the representation of the SWE, generally in line with the performances of other modelling studies carried out in alpine areas (Colombo et al 2022). (2022), we calculated the standardised SWE index (SSWEI) at the monthly scale, choosing the 1961-1990 period as a reference baseline, by fitting a Gamma distribution for each month of the snow season, at each station. We defined the beginning and end of the snow season for each station as the first and last month having a minimum of 50% of monthly mean SWE > 0 over the chosen baseline (Colombo et al 2022). All stations had at least 90% of the available months. We calculated a monthly value for all the Italian Alps by averaging all individual, station-based SSWEI values at the 15 stations. Aggregating the anomalies of each single station instead of considering the anomalies of the mean of each SWE series avoided an overrepresentation of those stations with higher SWE values (e.g. stations located at higher elevations). This also allowed to better address the inhomogeneous spatial distribution of the available long-term series (Colombo et al 2022). In addition, we explored potential elevational patterns by carrying out the analysis according to the three elevation ranges identified by Colombo et al (2022): We considered the mean SSWEI of March as representative of the SWE accumulation before the melting season, similarly to the widely used 1 April SWE (cf Gottlieb and Mankin 2022), as a good approximation of the snow dynamics across our investigated elevation range (Colombo et al 2022). SWE on 1 April is generally considered to most closely approximate the maximum accumulation of snowpack. However, recent work has challenged its usefulness on the basis that snowpacks in different regions (and at different elevations) can peak before or after 1 April (Gottlieb and Mankin 2022). In this study, we considered March SSWEI as more representative of the snow season accumulation than the single 1 April value. This is also in accordance with the findings of a recent study which showed that the peak SWE across the Italian mountains generally occurs in March (Avanzi et al 2023).

Climatic drivers
To investigate the influence of air temperature and precipitation on SSWEI variability, we employed ERA5 (spatial resolution: 0.25 • or ∼31 km, tem- for temperature, after reporting the products to the same climatological means of the corresponding ERA5 records. More details are given in the supplement (text S1). We calculated the monthly standardised precipitation index ((SPI); McKee et al 1993) at each node based on monthly-cumulated precipitation (considering 1-to-6-month accumulation periods), using the 1961-1990 baseline. For each aggregation scale, we considered the spatial mean of March SPIs as the corresponding winter precipitation anomaly index over the Italian Alps, referred as March SPI1 to March SPI6.
We calculated the monthly mean air temperature anomaly over each node by subtraction of associated monthly climatic mean over the 1961-1990 baseline. To obtain monthly air temperature anomaly time series over different aggregation scales (1-6 months), we applied a moving average to the resulting time series. For each aggregation scale, we considered the spatial mean of March monthly mean air temperature anomalies, normalised by the associated standard deviation, as the corresponding winter air temperature anomaly index over the Italian Alps, referred as March AT1 * to March AT6 * .

River discharge indices
Hydrological observations can be used as an indicator of the impact of snow drought. Long-term (1930Long-term ( -2023 daily and monthly river discharge measurements (table 2) were supplied by the Regional  1). Similarly to SPI and SSWEI, we calculated a standardised runoff index (SRI) based on the monthly mean discharge at each monitoring station, following Shukla and Wood (2008), using the 1961-1990 baseline.

Record-low SWE in 2022
The 2022  In the last decades, instead of a simple reduction in SWE, peak SWE might just have been moved to an earlier date with respect to March, according to elevation. However, in the Italian Alps, Colombo et al (2022) found that the lowest SSWEI values in the period 1930-2020 occurred from the 1990s, irrespective of elevation; more importantly, although they found that highly negative values occurred at the snow season tails (mostly in the spring), all months showed decreasing SSWEI values in the last decades/years. This is also true for 2022 (figure S1). The occurrence of a generally thinner winter snowpack in the last decades, rather than just an earlier occurrence of maximum SWE, is also in agreement with the measured negative SWE (Marty et al 2017a) and HS (Matiu et al 2021) trends, together with positive snow line elevation trends (Koehler et al 2022), in the European Alps.

The role of climate drivers
The amount of SWE is the result of an interplay between precipitation and temperature, and their history throughout the snow season (Marty et al 2017a). Therefore, here we tested the relation between March SSWEI and winter precipitation and temperature anomalies, considering aggregation scale from 1 to 6 months for March SPI and March AT * . We performed a multi-linear regression (MLR) without intercept for each aggregation scale, considering four baselines (1930-2023, 1930-1959, 1960-1989, and 1990-2023). We found the best r-squared performance in explaining March SSWEI with a 4 months aggregation scale (i.e. December to March SPI and December to March AT * ) for all the considered baselines (table S1). The fact that March SSWEI was best controlled by March SPI4 and March AT4 * is in agreement with the findings of Marty et al (2017a), who found that measured April SWE was controlled by December-March temperature and precipitation in the European Alps.   all occurred after 2000. Although these nine warmer winters (2007,2008,2014,2015,2016,2017,2019,2020,2023) had higher temperature anomalies (from +1.8 to +2.7), SSWEI (ranging from −2.8 to 0.6) never reached such a low value as in 2022. This occurrence was likely due to relatively wetter winter conditions (March SPI4 from −1.6 to +2.2) with respect to the 2022 event. Thus, the unprecedented combination of drier-and warmer-than-normal conditions in winter 2021/2022 greatly inhibited seasonal snowpack accumulation, leading to the lowest SSWEI in the past century.
The proposed MLR analysis can also help quantifying the relative contribution of precipitation and  1930-2023, 1930-1959, 1960-1989 and 1990-2023 baselines. Intercept is forced to (0,0). Relative regressive weights are reported in the table inset in the top left corner; r2 values are shown in the table inset in the bottom right corner ( * * * indicates significance at p < 0.001). temperature anomalies in controlling SSWEI, as both predictors follow a normal distribution by construction. We found that, considering the MLR relative weights for the baseline 1930-2023, March SPI4 and March AT4 * contributed to 65.1% and 34.9%, respectively (figure 4). Remarkably, the relative weight of March AT4 * greatly increased during the past three decades, from 19.8% and 13.3% over the baselines 1930-1959 and 1960-1989, respectively, to 41% over the baseline 1990-2023 (all regression weights and associated 95% confidence intervals are given in table S2). These results indicated the increasing role of temperature in explaining the SWE variability over the last decades, enhancing the effect of precipitation drought events (cf Colombo et al 2022). The 2022 event well exemplified the interplay of these conditions.

Hydrological implications
Warm and dry winter conditions occurring together, as in 2022, generally cause the most severe snow droughts and, consequently, the most severe summer streamflow droughts (Dierauer et al 2018). These can result in reduced hydropower production and drinking-irrigation water supply shortages (Dierauer et al 2018(Dierauer et al , 2019. As regards Northern Italy, we analysed historical discharge data collected on the Po and Adige rivers, in the period 1930-2023. In July 2022, the Po River registered the absolute historical minimum in all five analysed stations (Piacenza, Cremona, Boretto, Borgoforte and Pontelagoscuro; figure 5(a1)-(a5)), whereas on the Adige River (Boara Pisani station) a lower discharge was registered only in April 1944 (figure 5(a6)). A direct comparison with the monthly frequency distributions shows an impressive difference in July 2022 with respect to the 5th percentile in the order of −47.4% ± 3% and −45.4% for the Po and Adige rivers, respectively. It is worth noting that the differences of 2022 data with 'ordinary' discharge (when considering both 25th ÷ 75th or 5th ÷ 95th percentile) strongly increased during spring and summer.
Several processes potentially affect the river discharge in spring and summer and some of them are related to human activities, in particular irrigation and hydropower production. Therefore, the July discharge observations also integrate pressures on the water bodies and management options at river scale basis, which in turn can be related to climate factors, adding non-linearity to the processes. A graphical comparison of March SSWEI and July SRI1 ( figure 5(b)) clearly shows the unprecedented and contemporary snow and hydrological drought in 2022. The July SRI1 vs March SSWEI Pearson correlation coefficients (figure 5(c)) show that SWE is capable of explaining approximately 50% of the interannual discharge variability. It is worth stressing, however, that part of this correlation can be directly due to the snow-melt process, and part to the increasing withdrawals from the water bodies (both surface and groundwater) due to a lack of resources. Assessing the weight of each process on the July SRI1 in river basins such as the Po and Adige that are so strongly impacted by human activities and in such a complex way is a challenging task beyond the scope of this study.
The critical hydrological conditions in 2022 strongly impacted several sectors, as proven by the drought bulletins issued by the Permanent Observatories for Water Uses of both the Hydrographic District Authorities of Po River (Po River; www.adbpo.it/osservatorio-permanente, accessed 27 April 2023) and Eastern Alps (Adige River; www.alpiorientali.it/osservatoriopermanente.html, accessed 27 April 2023). The 'water severity' , that is the ability of the water supply systems to meet the ordinary water needs for all sectors, shifted from 'medium' in April to 'high' from July onwards. The dramatic drought conditions forced both authorities to constantly monitor and re-allocate the very scarce resources available among the different users, with the agricultural sector being particularly affected. Such conditions also strongly impacted the ecosystems; by way of example, the sea water intrusion at the Po delta increased from about 20 km (at Goro monitoring station) at the beginning of June to about 40 km in mid-July (Po river Authority, drought bulletin n.12/2022, July 7th, 2022).
The situation in 2023, until the end of March, appeared rather similar to 2022, not only considering the snow-drought conditions, but also in relation to the hydrological ones (figure 5(a1)-(a6)), which in turn were the historical minima, posing serious concerns for the forthcoming summer, when water needs are expected to strongly increase. For instance, the energy stored in water reservoirs and hydro storage plants accounted for 821 175 MWh, the second worst situation over the period 2015-2023, the worst being 2022 (735 675 MWh) (ENTSO-E Transparency platform, https://transparency.entsoe.eu/dashboard/ show, accessed 27 April 2023). Analogously, severe impacts might be expected on the ecosystem services, considering that the sea water intrusion was estimated around 25 km at the end of March 2023 (it was 20 km in April 2022). The occurrence of two consecutive, exceptional snow-drought winters may lead to an unknown scenario, from both hydrological and hydrogeological perspectives, as very little is known about the impacts of persistent snow-drought conditions on groundwater and associated non-linearity (e.g. soil-vegetation-atmosphere interaction, infiltration dynamics, inversion of the vertical river-aquifer gradients).
The unprecedented snow droughts that occurred in the early 2020s offer an analogous perspective for projected future snow conditions in the Italian Alps, and in the European Alps in general. Indeed, in the Alps, future snowfall (Frei et al 2018), snow-depth (Marty et al 2017b), snow-cover (Matiu and Hanzer 2022) and SWE (Kotlarski et al 2022) decreases have been predicted. Such decreases, strongly linked to the expected rise in air temperature (Kotlarski et al 2022), might have further large implications on water availability, especially during the warm season (Jenicek et al 2018). Northern Italy might be strongly affected by these changes. For instance, in the Po River Basin, nearly 20% and 40% of future irrigation water demands must be met by sources alternative to snowmelt under the +2 • C and +4 • C warming scenarios, respectively (Qin et al 2020). Thus, time is running out for anticipating and effectively adapting to the ongoing (future) climatic changes.

Conclusions
Here we investigated the exceptional snow-drought conditions that affected the Italian Alps during the early 2020s. During winter 2021/2022, the mean March SWE anomaly, expressed through the SSWEI, reached the lowest value in the last century (time-span 1930-2023). The extreme snow drought impacted all analysed elevation ranges due to an unprecedented combination of drierand warmer-than-normal conditions in the period December 2021-March 2022. This event contributed to causing critical hydrological conditions, with an extremely severe hydrological drought in the Po and Adige rivers. Summer streamflow was particularly impacted, with an unprecedented hydrological drought that was measured on the Po River in July 2022. Although the snow drought in winter 2021/2022 was the worst ever observed, in the last decades (from 1990s on) there has been an increase in frequency and magnitude of snow-drought events. We found an increasingly relevant role of warming air temperature in driving such events. Thus, considering the predicted air temperature warming, meteorological drought increasing, and SWE decreasing during the next decades, the 2022 snow-drought event could offer an analogous perspective for projected future SWE conditions in the Italian Alps. Remarkably, during winter 2022/2023, the March SWE anomaly reached the third lowest value in the analysed period. The impacts of two consecutive, extreme snow droughts in winters 2022 and 2023 on the hydrological and hydrogeological drought are currently unknown. In the light of our findings, we consider the implementation of risk mitigation measures to be of primary importance, taking into account the 'new normality' of snow conditions in the Italian Alps, and the use of standardised indexes to support effective adaptation practices relying on stakeholders' cooperation.

Data availability statement
The data cannot be made publicly available upon publication because they are owned by a third party and the terms of use prevent public distribution. The data that support the findings of this study are available upon reasonable request from the authors.