Climatology and Formation Environments of Heavy Snowfall Events in the Ural Region (Russia)

Abstract

Heavy snowfall events in the Ural region have drawn significant attention due to their substantial frequency, the region’s relatively high population density and its developed network of roads and power lines. This study summarizes the main characteristics of the hazardous heavy snowfall (HHS) events (≥20 mm 12 h⁻¹) that have occurred in the Ural region between 1981 and 2025, as well as in related synoptic-scale environments, for the first time. The dataset consists of 116 HHS reports, with 12-hourly snowfall intensities ranging from 20 mm to 47.6 mm. The main characteristics of these events (snowfall amount, spatial distribution, inter-annual and seasonal variability and trends, associated weather phenomena, and related damage) are examined based on the data from weather stations, the ERA5 reanalysis, scientific literature, and media reports. While there is no statistically significant trend in HHS events, the frequency of the most damaging late spring and early autumn snowfalls has decreased. Using 72 h backward trajectories according to the NOAA HYSPLIT model and the ERA5 reanalysis, we classified the HHS events into five types according to air mass origin, and performed a composite analysis for each type. The main finding is that 46% of HHS reports are related to cyclones forming over the Caspian and Aral seas, resulting in a higher frequency of HHS events to the east of the Ural Mountains compared to the western part of the region.

1. Introduction

Snowfall is an important meteorological variable that has significant impacts on the economy and society. Both mean snowfall and its intensity distribution are important, but heavy snowfall events have particularly large economic and human impacts. They can cause power outages, severely disrupt transportation, and impede emergency services [1,2,3,4]. Their impact on transportation and power lines might be intensified by severe winds, freezing rain, ice, or wet snow accumulation [5]. In mountainous areas, heavy snowfall combined with strong winds and unstable snowpack contributes to the formation of snow avalanches, which frequently result in fatalities and property damage [6]. They can also induce the collapse of buildings due to overload [7], flooding, and landslides due to subsequent melting [6,8]. Late spring or early autumn snowstorms cause substantial damage to agriculture, especially in subtropical regions [8,9]. Additionally, snow accumulation on trees can cause widespread forest disturbances, particularly in the boreal forests in Northern Europe [10,11,12] and in subtropical regions [13,14]. Compound snowfall and ice accumulation extremes strongly increase the degree of damage [15]. Heavy snowfall events are especially frequent and impactful in the Northeast of the US [2,3,16,17], the Great Lake Basin [18,19], the Japanese islands [20], and in many high-altitude regions, such as the Swiss and French Alps [5,6,21] and the Pyrenees [8].

Climatological studies based on observations of snowfall extremes have been performed in the US [1,2,16], the Great Lake Basin [18], Europe [4,6,8], Japan [20], and China [22]. Alongside data from weather stations, the ERA5 reanalysis [23] and high-resolution regional climate models [17,21,24] were used to estimate snowfall extremes in the Northern Hemisphere [25], as well as in Europe [4,21,24], North America [17], and China [26]. These studies revealed contrasting trends in snowfall extremes for different regions. Furthermore, trends in annual snowfall amount are often inconsistent with trends in the frequency of snowfall extremes [4,20]. While the amount of annual snowfall decreases due to rising air temperature and the associated shift in precipitation from snow to rain [25], the frequency of snowfall extremes can increase due to alterations in atmospheric circulation patterns, changes in moisture availability, and temperature shifts [25,27]. Synoptic-scale and atmospheric circulation patterns contributing to the formation of snowfall extremes have been studied based on long-term observations and/or reanalysis data in the US and Canada [19,28,29,30], Central Europe [31,32], the Pyrenees Mountains [8,33], Hokkaido Island [34], and Northern Xinjiang [35]. These large-scale patterns include the North Atlantic Oscillation (NAO) [8,32], air mass characteristics such as advection, temperature, and precipitable water [30], types of atmospheric fronts [30], and the convective or non-convective nature of the events [4,30]. Local intensification of snowfall associated with mountains and/or the lake effect [19] also plays a role in extreme snowfall events.

Numerous studies have considered future changes in snowfall extremes in a warming climate, based on global and regional climate models [3,21,36,37,38,39]. In line with [27], heavy snowfall stems from extreme precipitation occurring at optimal air temperatures of slightly below 0 °C, which favors high precipitation intensities and a snowfall proportion close to 100%. Therefore, changes in snowfall extremes depend on a trade-off between trends in precipitation extremes and changes in the probability of experiencing temperatures in this optimal range [5,38]. A shorter snow season combined with increasing atmospheric water vapor may result in increased snowfall in the coldest regions and a substantial decrease in warmer regions [40]. Additionally, changes in atmospheric circulation patterns [41,42,43] may alter storm tracks, cyclone frequency, and intensity, and the characteristics of snowfall extremes.

In Russia, the characteristics of heavy snowfall have mainly been studied in individual regions, e.g., in the Ural Mountains [44,45], the North Caucasus [46], and Sakhalin Island [47]. Based on daily snowfall data from approximately 500 weather stations across the country, it has been shown that between 1950 and 2006, the frequency of moderate and heavy snowfall increased in the east of the East European Plain and in Western Siberia, while the frequency of light snowfall decreased over most of the country [48]. For the period of 1961–2013, weak positive trends in winter precipitation extremes (statistically significant at several stations) were identified over most of the country [49]. According to the Coupled Model Intercomparison Project (CMIP6) data, weak increasing trends in snowfall extremes have also been found in Northern Eurasia [50]. Heavy snowfall events in Russia are generally associated with cyclonic activity of different origins [51,52], namely lows forming over the North Atlantic, the Black Sea, the Caspian Sea (in the western part of the country), and the North Pacific (in the eastern part of the country).

The Ural region is one of the areas in Russia that is most susceptible to snowfall-related damage, due to its relatively high population density (and hence, dense network of roads and power lines), as well as the high frequency of such events [44,45]. Additionally, the Ural climate is characterized by the regular occurrence of late spring or early autumn heavy snowfalls, which cause particularly severe damage [53]. The most notable example is the snowstorm on 6 June 1995, which caused catastrophic damage to agriculture and forests in several districts of the Middle Urals [54,55]. However, studies of the climatology and synoptic-scale environments responsible for snowfall extremes in the Ural region were either published a long time ago [53] or only cover the western part of the region [44,45]. The contribution of the observed climate warming to the changes in the snowfall extremes has not yet been analyzed.

In this study, we first present the climatology and synoptic-scale environments of snowfall extremes in the Ural region, using ground-based observations and ERA5 reanalysis data. More precisely, the territory under study comprises four subjects of the Russian Federation (the Perm, Sverdlovsk, Chelyabinsk, and Kurgan regions), belonging to the Ural branch of the Russian Hydrometeorological Service (Roshydromet), which provided a part of the observational data. We consider the period from 1981 to 2025, which roughly corresponds to the current stage of global and regional climate warming.

This paper is organized as follows: Section 2 provides information about the region under study and the data sources used to select heavy snowfall events from the observations data. It also describes the methods of synoptic-scale analysis. Section 3 presents the results of the study, including the climatic characteristics of the selected events, the analysis of backward trajectories, and the synoptic-scale patterns of the formation of heavy snowfall events, the weather events accompanying heavy snowfalls, and the related damage. In Section 4, the results are discussed and summarized.

2. Materials and Methods

2.1. Region of Study

The western part of the study region belongs to the East European plain and the eastern part to the West Siberian plain, while the Ural Mountains stretch across its central part (Figure 1). The North and South Ural mountains are characterized by predominant elevations of about 800–1400 m above sea level (a.s.l.), with a maximum of 1640 m. The Middle Ural is significantly lower, with only a few mountains exceeding 800 m a.s.l. These terrain features substantially affect the spatial distribution of precipitation, including heavy snowfall events.

According to the Worldclim 2.0 dataset [56], the mean annual air temperature (MAAT) within the study area ranges from 0 °C in the North Ural to 4 °C in the south-west of the region. From November to March, the average air temperature is below 0 °C. Over the past four decades, the MAAT has increased by 0.4–0.5 °C every 10 years [57,58], mainly due to warming in the cold season. Annual precipitation varies from 350 mm in the southern part of the Chelyabinsk region to over 1000 mm on the western slopes of the Northern Urals [56]. From late October to early April, most of the precipitation falls as snow. However, snowfall can also be observed in early autumn (from late September) and in late spring (in May and even in early June), which is associated with the advection of Arctic air masses, particularly from the Kara Sea [44,53].

Seasonal snowpack usually forms in mid-October in the northern Urals, and in late October or November elsewhere in the region. It melts in April (in most of the region), and in May (in the northern Ural mountains). The annual maximum snow water equivalent is observed in March and early April in the mountains. According to the ERA5-Land data [59] and ground-based observations [60], it ranges from 50 to 100 mm in the southeast of the region to 350–400 mm on the western slopes of the North Urals.

2.2. Data and Methods

2.2.1. Identification of Heavy Snowfall Events and Their Characteristics

High-quality snowfall data at synoptic or regional scales is difficult to obtain [4,61], particularly due to snow blowing out of precipitation gauges [62]. In the area under study, these limitations are exacerbated by the considerable distance between weather stations, particularly in the northern part of the region, as well as the inaccessibility of complete data archives from numerous stations. In this study, we used both ground-based observations and the ERA5 reanalysis data to identify spatiotemporal patterns in the distribution of heavy snowfall events, their long-term trends and synoptic-scale environments of their formation (Table 1).

There are 87 Roshydromet weather stations operating within the study region (Figure 1). These stations provide reports on the precipitation amount (mm of water equivalent) every 12 h, while other variables such as air temperature and precipitation types (rain, snow, wet snow, freezing rain, etc.) are reported every three hours. During the economic crisis of the 1990s, the number of weather stations decreased by around 40%. Consequently, several heavy snowfall events that occurred in the 1980s and early 1990s were reported by stations that ceased observations in subsequent years.

Among 87 operating weather stations, 17 are reference stations. The All-Russian Institute of Hydrometeorological Information—World Data Center (RIHMI-WDC) [63] provided their observations for the entire study period. In addition, we used the data from a further 19 reference stations located outside the study area at a distance of no more than 150 km (Figure 1). The observations from the remaining 70 weather stations located within the study area are available for the period from 2011 to the present day.

Note that we did not correct the wind-related errors in the observed snowfall data, since we used operational SYNOP reports. Snow blowing from precipitation gauges at weather stations in Russia can lead to an underestimation of precipitation by up to 30% [64]. Therefore, the actual number of heavy snowfall events can be higher than observed.

Table 1. List of the datasets with associated variables, sources, and temporal availability used in this study.

Dataset	Temporal Resolution	Variables or Other Data Used
Routine observations from the weather stations of Roshydromet [63]	daily	Precipitation amount (mm of water equivalent) and snow depth (cm)
	12 h	Precipitation amount (mm of water equivalent)
	3 h	Air temperature at 2 m, wind gusts, precipitation types Diameter of wet snow accumulation on the wires
Annual reports on hazardous weather events (HWE) [65] and monthly reviews of HWE [66]	−	Heavy snowfall events (≥20 mm/12 h), other associated HWE (severe wind, wet snow/ice accumulation), and related damage
The ERA5 reanalysis data [23]	Daily	Daily snowfall amount (mm of water equivalent)
Media news	−	Damage report associated with heavy snowfall
Database of windthrow events in the forest zone of Russia [67]	−	Large-scale forest damage caused by heavy snowfall and wet snow accumulation

−: Not applicable for this type of data.

Table 1. List of the datasets with associated variables, sources, and temporal availability used in this study.

Dataset	Temporal Resolution	Variables or Other Data Used
Routine observations from the weather stations of Roshydromet [63]	daily	Precipitation amount (mm of water equivalent) and snow depth (cm)
	12 h	Precipitation amount (mm of water equivalent)
	3 h	Air temperature at 2 m, wind gusts, precipitation types Diameter of wet snow accumulation on the wires
Annual reports on hazardous weather events (HWE) [65] and monthly reviews of HWE [66]	−	Heavy snowfall events (≥20 mm/12 h), other associated HWE (severe wind, wet snow/ice accumulation), and related damage
The ERA5 reanalysis data [23]	Daily	Daily snowfall amount (mm of water equivalent)
Media news	−	Damage report associated with heavy snowfall
Database of windthrow events in the forest zone of Russia [67]	−	Large-scale forest damage caused by heavy snowfall and wet snow accumulation

−: Not applicable for this type of data.

Information on snowfall that met the criteria of hazardous weather events (HWE) that occurred from 1981 to 2025 is available in the HWE databases and reviews (Table 1). According to the criteria used by Roshydromet [68], snowfall with a 12 h precipitation amount of at least 20 mm is considered a HWE if precipitation was observed only in the form of snow throughout the entire period. For wet snow or snow mixed with rain, the HWE threshold is 30 mm 12 h⁻¹. Thus, we have compiled a list of snowfall events classified as HWE (hereinafter referred to as Hazardous Heavy Snowfall, HHS) for the period of 1981–2025. In numerous cases, HHS occurred simultaneously at several weather stations, or on two consecutive days, being associated with a single synoptic-scale process. Therefore, these HHS reports have been attributed to one HHS event.

Since HHS events occurred rarely, we also analyzed lighter snowfalls to identify changes in their intensity over time. However, this analysis was based on the data from only 36 reference weather stations that were available for the entire period from 1981 to 2025. Using daily and 12 h data on the air temperature at 2 m height and precipitation amount (Table 1), we classified all precipitation that occurred on the days when the mean temperature was below the assigned threshold value of 1 °C as snowfall (based on previously recorded heavy snowfall and wet snow events). In other studies, this threshold ranges from 0 °C to 2.5 °C [4]. This approach can produce both false positives and false negatives, but the main patterns of heavy snowfall distribution in space and time remain unchanged regardless of the chosen threshold [4].

We considered different thresholds for snowfall amounts, and calculated the number of days with snowfall amounts of ≥20 mm (S_20mm), ≥10 mm per day⁻¹ (S_10mm), and ≥10 mm per 12 h⁻¹ (S_{10mm_12h}), based on daily and 12 h precipitation data, respectively. Selecting three different thresholds enabled us to consider snowfall events of varying intensities. In addition, we calculated the threshold values for daily snowfall exceeding the 95th and 99th percentiles (S_95p and S_99p, respectively), the mean annual maximum daily snowfall (S_{max_annual}), as well as the daily maximum for each year of the 1981–2024 period.

In line with previous studies [20,69], we employed a nonparametric linear Theil–Sen slope estimator [70], which is less sensitive to outliers than standard linear regression, to calculate the trend slope, and a nonparametric Mann–Kendall test [71] to determine the statistical significance of the trend. The significance level was taken as 0.05. As trends may vary across different parts of the study area, we estimated them both for all weather stations for groups of stations in the northern and southern parts of the region, and separately for their groups (for the north and south parts of the region, divided along the 57°N parallel).

Based on the ERA5 post-processed daily statistics dataset, we downloaded daily snowfall amount data for the period from 1981 to 2024. We then calculated several characteristics of snowfall extremes, namely S_{max_annual}, S_95p, and S_99p, respectively. The average annual snowfall amount (S_{annual_mean}) was also calculated for the same period. In line with [25], only the days with snowfall amounts of ≥ 1 mm were taken into account when calculating the threshold values. Trends in snowfall extremes were estimated using the nonparametric Theil–Sen method.

We examined the correlations between observed and ERA5-based snowfall characteristics. In particular, we extracted the values of the above metrics, as well as the average annual snowfall amount (S_{annual_mean}) from the ERA5 grid cells corresponding to the locations of 36 reference weather stations. We have evaluated the relationships between these variables and the above-described metrics of heavy snowfall based on observational data (S_20mm, S_10mm, and S_{10mm_12h}). Spearman’s rank nonparametric correlation (hereinafter SR) was calculated, with a significance level of 5%.

2.2.2. Synoptic-Scale Analysis

In this study, we used the NOAA Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) air parcel trajectory model to simulate the backward advection paths of air parcels to define air mass origin and moisture sources for HHS event formation [72], and the ERA5 reanalysis data to analyze synoptic-scale patterns of HHS event occurrence.

The NOAA HYSPLIT air parcel trajectory model, developed by NOAA’s Air Resources Laboratory, is a complete system for computing simple air parcel trajectories as well as complex transport, dispersion, chemical transformation, and deposition simulations [72]. In this study, we calculated 72 h backward trajectories at 3000 m above sea level (ASL) for locations where HHS occurred, based on the average hour of the period during which snowfall was observed. For example, if snowfall occurred between 03:00 and 15:00 UTC, the trajectory was calculated for 09:00 UTC. The HYSPLIT model was run using the data from the NCEP/NCAR reanalysis (for the period of 1984–2005) and the Global Data Assimilation System (GDAS) with 1° cell size (for the period of 2006–present). We performed calculations for each of the 56 HHS events, but we divided several of them into 12 h time intervals (if the HHS was reported by weather stations during two or three consecutive 12 h periods). We also calculated more than one trajectory for a single HHS event if it was simultaneously reported by weather stations located on both sides of the Ural Mountains, or at a distance of more than 300 km from each other.

In total, we analyzed 74 different backward trajectories. These trajectories and related HHS events were classified into 5 types, based on their air mass origin and advection direction. The types were: S (southern advection from the Caspian or Aral Seas); SW (southwestern advection from the Black or Mediterranean Seas); W (western advection from the Baltic, North and Norwegian Seas); NW (northwestern advection from the Barents Sea); and local (when the moisture source is difficult to identify). Note also that several HHS reports associated with one event can be classified into different types. For example, the snowstorm on 18 October 2014 [45] was initially driven by mid-tropospheric advection from the Black Sea (SW type), but then the air mass spread from the Caspian Sea (S type).

The ERA5 data were used to assess the synoptic-scale environments of the formation of HHS events, and to perform a composite analysis for each of the five above HHS types. In line with [30], the composite fields were generated as the ensemble averages for each of the five snowfall types. These fields are the air temperature at the 850 hPa isobaric surface (T₈₅₀), total precipitable water (PW), 700-hPa geopotential height (H₇₀₀), and the pressure at mean sea level (MSLP). These variables characterize the environments of the HHS events’ formation, namely the mid-tropospheric ridges and troughs, the location, intensity, and stage of development of the surface low, the direction of mid-tropospheric advection, and the temperature and moisture content of the air mass. Unlike [30], the composites were calculated for the middle hour of each 12 h interval during which HHS events were reported (as well as for backward trajectories).

Alongside the composite analysis, we also estimated several other synoptic-scale characteristics. Thus, the horizontal gradient of T₈₅₀ (ΔT₈₅₀, °C/500 km) was calculated as the difference between the minimum and maximum values of the T₈₅₀ within a 250 km radius around each weather station reporting HHS. The type of atmospheric front (warm, cold, or occluded) was estimated using a manual classification technique [29,44] based on hourly characteristics derived from the ERA5 data (MSLP, T₈₅₀, PW, and mid-tropospheric advection).

3. Results

3.1. Climatology of Heavy Snowfall Events in the Ural Region

3.1.1. Observed Snowfall Intensity and Snow Depth

We have compiled a dataset of 116 HHS reports, which are associated with 56 different HHS events, based on data from weather stations. Among them, there are 97 reports (48 events) of heavy snowfall, while 19 reports (8 events) of heavy wet snow or snow mixed with rain. In these cases, 12 h snowfall amounts ranged from 20 to 47.6 mm; the mean and median values were 25.4 and 23.1 mm, respectively (Figure 2a). In 24 cases, the 12 h snowfall amount was ≥30 mm, including 19 cases of wet snow and snow mixed with rain. The strongest 12 h snowfall was associated with the 6 June 1995 snowstorm in the Middle Ural [54,55], when precipitation intensity of up to 47.6 mm 12 h⁻¹ (wet snow with rain) was reported. The strongest 12 h snowfall (38.8 mm), except for the cases of wet snow and snow mixed with rain, was reported in the city of Chelyabinsk (WMO ID 28645) during the 25–26 April 2014 snowstorm in the South Ural region.

Although the HWE criteria are only established in Russia for 12 h snowfall amounts, the daily snowfall amount is also important, particularly since prolonged heavy snowfall leads to more significant damage than short-term snowfall. The 24 h snowfall amount is known for 75 out of 116 HHS reports, and it ranges from 20 mm to 70.9 mm. The mean and median values are 35.8 and 36.3 mm, respectively (Figure 2b). The strongest 24 h snowfall was also associated with the snowstorm on 6 June 1995 (70.9 mm in the town of Alapaevsk, WMO ID 28264). With the exception of cases of wet snow or snow mixed with rain, the highest 24 h snowfall amount is 52.8 mm, which occurred in the city of Chelyabinsk on 25 April 2014.

In most cases, heavy snowfalls led to a substantial increase in snow depth. However, daily snow depth data is only available for a proportion of weather stations (61 out of 116 HHS reports). In addition, snow depth measurements are strongly complicated by variations in snow density and wind-driven snow transport. The daily increase in snow depth ranges from 6 cm to 64 cm, while the mean and median values are 22.3 and 20.0 cm, respectively. The strongest daily increase in snow depth (64 cm) was reported on 3 May 1984 in the city of Yekaterinburg. The correlation between the 24 h snowfall amount and the increase in snow depth is relatively low (SR = 0.40). Indeed, snow depth is strongly influenced by the local wind conditions, as well as the precipitation type (snow, wet snow, or snow mixed with rain) and air temperature [73]. The vast majority (>90%) of HHS events occurred when air temperatures ranged from −5° C to 1° C, which is consistent with global estimates [27]. In only two cases (6 March 2005 and 23 October 2014), the air temperature during snowfall was about −10° C.

3.1.2. Weather Events That Accompanied Snowfall Extremes

Several HHS events were accompanied by severe winds and/or heavy wet snow accumulation on wires and trees, forming so-called compound extremes [74]. In the Ural region, these are extremely rare but cause serious damage to power lines, transportation, and forests. Thus, of the 116 HHS cases, only 3 were accompanied by severe wind gusts (≥25 m s⁻¹), which were reported simultaneously with snowfall at the same weather station. It should be noted that two of these events occurred during the snowstorms with the highest 12 h snowfall amounts reported (6 June 1995 in the Middle Ural and 25 April 2014 in the city of Chelyabinsk). In several other cases, the information about wind-related damage is also available, but the weather stations did not report wind gusts of ≥25 m s⁻¹. In the case of the snowstorm on 17 May 1981, HHS and severe wind gusts (25 m s⁻¹) were reported by different weather stations, and therefore, this event cannot be considered a compound extreme. All three compound snow and wind extremes occurred in the eastern part of the study area in April and early June. Note that the annual maximum of the mean wind speed was also observed in the spring season [75].

There were also two compound heavy snowfall and wet snow accumulation events when the diameter of snow deposits on wires exceeded 50 mm. Both events occurred on the eastern slope of the Middle Ural in late spring (6 June 1995 and 3 May 2024), and caused severe damage to power lines. However, the actual number of such events is likely to be underestimated, since wet snow accumulation is typical for highland areas, whereas most weather stations are located in river valleys.

Compound events with simultaneous occurrence of heavy snowfall, severe wind, and wet snow accumulation are even rarer, yet they are the most damaging of all HHS events. The 6 June 1995 snowstorm, for example, met the HWE criteria in all three indicators: snowfall of up to 47 mm 12 h⁻¹, wind gusts of up to 26 m s⁻¹, and wet snow accumulation of up to 190 mm on the wires and trees. Unsurprisingly, it caused unprecedented damage to power lines and forests, particularly the catastrophic windthrow in the Visimsky State Nature Reserve [54,55]. A similar event occurred in the North Urals on 9 October 2015, causing widespread damage to forests (Figure 3), but it was not reported by weather stations.

Unlike other regions with temperate climates that are prone to HHS events [19,28,30], convective snowstorms are not typical for the study area. The sample contains no HHS events of a convective nature accompanied by thunderstorms.

3.1.3. Spatial Distribution

A total of 116 HHS cases were reported at 60 different weather stations (7 of which ceased observations in the 1990s). The spatial distribution of the HHS events does not correspond with that of the S_{annual_mean}, as estimated using the ERA5 data for the period 1981–2024 (Figure 3). Thus, the S_{annual_mean} reached 300–500 mm on the western slope of the Northern Urals; however, the weather stations located there reported only a few HHS cases, with the exception of station 28134 (7 cases). Conversely, HHS events regularly occurred in the southeast part of the region, where the S_{annual_mean} does not exceed 150 mm. The contribution of heavy snowfall to the S_{annual_mean} strongly increases from the west slope of the North Ural to the east slope of the South Ural. Note that the spatial distribution of warm-season precipitation extremes in the Ural region has similar patterns, with a maximum in the east slope of the mountains [76].

The frequency of HHS events has no statistically significant correlation with the elevation above sea level (SR = 0.29). HHS events are less frequent at weather stations located more than 150 km from the Ural Mountains than at stations located near the mountains. Thus, orography significantly impacts the spatial distribution of HHS events, but it strongly differs from its impact on the S_{annual_mean}. This difference can be attributed to cyclone trajectories moving from the south or southwest, which are responsible for a substantial proportion of HHS events, particularly in the southeast of the study region (see Section 3.2 for details). HHS events with a 12 h snowfall amount of ≥30 mm 12 h⁻¹ occurred in the North, Middle, and South Ural, primarily on the eastern slopes (Figure 3), with the highest maximums concentrated in the Middle Ural, associated with the snowstorm of 6 June 1995.

In addition to the observed HHS events, we also calculated several extreme snowfall characteristics according to the data from 36 reference weather stations and the ERA5 reanalysis (see Section 2.2.2 for details), and compared their spatial distribution (Figure 4 and Figure 5,

Table 2. Observed and ERA5-based characteristics of heavy snowfall events for 36 reference weather stations. Statistically significant correlations are highlighted.

Snowfall Characteristics	Observed (Numerator) and ERA5-Based (Denominator)				Bias	RMSE	SR
	Mean	Median	Maximum (WMO ID)	Minimum (WMO ID)
Sannual mean (mm)	161.5/ 208.1	147.2/ 205.6	346.0 (28138)/ 317.6 (28138)	89.8 (28833) 132.0 (28748)	46.6	55.1	0.86
S95p (mm)	9.2/ 8.2	9.1/ 8.1	11.8 (35026)/ 11.9 (35026)	7.5 (28573)/ 6.7 (28573)	−1.0	1.3	0.57
S99p (mm)	14.9/ 12.8	14.6/ 12.5	19.0 (28440)/ 16.9 (35026)	12.3 (28661)/ 10.0 (28573)	−2.1	2.5	0.46
Smax_annual (mm)	14.0/ 12.3	13.2/ 12.5	17.5 (35026)/ 15.9 (35026)	10.9 (28573)/ 9.8 (28573)	−1.6	2.1	0.57
Trend in Smax_annual (mm 10 years−1)	0.02/ 0.01	0.01/ 0.01	1.4 (28573)/ 1.3 (28561)	−1.5 (28440)/ −0.75 (28440)	−0.01	0.64	0.32

). According to the ERA5 data, two maxima located in the South Ural and North Ural, respectively, are pronounced in the spatial distribution of all characteristics of snowfall extremes (Figure 4 and Figure 5a). The maximum in the Southern Urals is more pronounced, while the S_{annual_mean} is higher in the western slopes of the Northern Ural. Note also that the maximum in the Northern Ural falls in a low-populated territory and a territory poorly covered by observations. The spatial distribution of the S_95p and S_99p differs significantly from that of the S_{max_annual}. Thus, the highest values of S_95p and, particularly, S_99p are shifted to the eastern slope of the Urals and even to western Siberia, clearly following the trajectories of cyclones moving from the Caspian and Aral Seas, which are responsible for the heaviest snowfall events [53]. At the same time, the distribution of the S_{max_annual} (Figure 5a) is more consistent with that of the S_{annual_mean} (Figure 3).

A quantitative comparison of the observed and ERA5-based snowfall extreme characteristics (as well as the S_{annual_mean}) is shown in Table 2. These results are consistent with the previous ERA5-based precipitation estimates in Northern Eurasia [77,78]. The ERA5 reanalysis substantially overestimates the S_{annual_mean}, but slightly underestimates snowfall extremes, since its spatial resolution is insufficient to correctly reproduce extreme values. SR between the observed and ERA5-based S_{annual_mean} (0.86) indicates a high degree of spatial distribution agreement, whereas the SR for the characteristics of snowfall extremes is significantly weaker. The correlation between the linear trends in S_{annual_max} according to the observations and reanalysis data is weak, which can be explained by the fact that the trends themselves are statistically insignificant (Figure 5b).

3.1.4. Inter-Annual and Seasonal Distribution

The number of HHS events strongly fluctuates from year to year, with no events observed in 16 out of 45 years (Figure 2c). The highest number of HHS reports was in 2014 (15), while the highest number of events occurred in 1982 (4). This discrepancy arises from a single HHS event covering a large area that was reported at several stations. Thus, the largest number of reports was associated with the event of 8–9 October 2015, when nine weather stations reported a heavy snowfall of 20–30 mm 12 h⁻¹.

Despite several weather stations ceasing observations in the 1990s, there is no statistically significant trend in the frequency of HHS events. Trends in S_{max_annual}, as observed from daily measurements taken at 36 reference weather stations (Figure 5b), are also non-significant, ranging from −1.5 to 1.5 mm per decade. Trends in S_10mm and S_20mm, as shown in Figure 6, are also non-significant for both the entire territory and individual weather stations.

According to the ERA5 data, S_{max_annual} has either increased or decreased (mainly non-significantly) in different parts of the study area (Figure 5b). However, observed and ERA5-based trends in S_{max_annual} are poorly correlated. The lack of significant changes in the number of HHS events and S_{max_annual} is in line with similar findings regarding snowfall extremes in the Ural region according to the ERA5-Land data [25].

The seasonal distribution of HHS events shows two well-pronounced peaks in April–May and October (Figure 2d). From November to March, almost all precipitation in the study area falls as snow; however, the atmospheric moisture content is insufficient for the formation of HHS. The predominance of HHSs in spring and autumn is in line with [39], which demonstrated that the most favorable environments for snowfall extremes are associated with the air temperatures slightly below 0 °C. It should be noted that the springtime maximum of HHSs is more pronounced in terms of both the number of reports and snowfall intensity. Around 47% of all HHS cases have been reported between 1 April and 17 May, marking the most favorable season for their occurrence. It is also noteworthy that 21 out of 24 cases of the most intense snowfalls (≥30 mm 12 h⁻¹) occurred in spring and even in early June.

HHS events that occur during the growing season deserve special attention, since they cause widespread damage to agriculture [53]. In the study region, the growing season begins in early May in the southern part and in mid-May in the central and northern parts, ending in September (roughly 10 September in the north and 20 September in the south). Only 6 of the 56 HHS events (and 20 of the 116 HHS reports) occurred during the growing season, namely between 13 May and 6 June. The earliest autumn event occurred on 20 September, i.e., after the end of the growing season.

Under climate warming, the seasonal distribution of HHS events substantially changes (Figure 2d). Thus, in the first half of the period (1981–2002), most cases were observed in May. In addition, two events occurred in June (in 1982 and 1995) and in September (in 1993 and 1996). In the second half of the period (2003–2025), most HHS events occurred in April and October, with a slight increase in the number of events occurring in the cold season (November–March).

3.1.5. Damage Characteristics

Information on damage caused by HHS’s events and associated hazardous weather, like severe wind or wet snow accumulation, is fragmented and incomplete. Damage reports are often available for individual municipalities or for specific economic sectors, e.g., for power supply or transportation. There is no comprehensive estimate of the economic losses caused by HHS’s events, most likely because the insurance industry in Russia is underdeveloped. Due to the incomplete nature of the damage data, financial losses cannot be compared with similar estimates for the US, where the damage caused by winter storms can exceed USD 1 billion per year [2].

There is no information on fatalities related to HHS events, except for those in weather-related road traffic accidents. There are numerous cases of injury, mainly also related to road traffic accidents and icy roads, but the summarized information on these cases is unavailable. Fatalities can also be related to avalanches induced by heavy snowfall, but these are extremely rare in the Ural region, since the mountain slopes are not steep enough for avalanches to occur [79].

In most HHS cases, documented damage includes power outages or damage to forestry and crops. The most widespread power outages were associated with the compound snow and wind events and with heavy wet snow accumulation. In addition, substantial damage is associated with heavy snowfall covering large areas (

Table 3. The most damaging HHS events in the Ural region, 1981–2024.

Event Date and Trajectory Type	Number of HHS Reports, and Maximum Observed Snowfall per 12 (24) h, mm	Accompanied Weather Events (According to Weather Stations Data)	Number of Consumers with Interrupted Power Supply, Thous. People	Other Damage
17 May 1981 (S)	6 29.8 (45.8)	Wind gusts up to 21 m s−1	No data	Widespread crop damage, traffic disruption, power outage
3 May 1984 (NW)	5 36.3 (48.5)	Heavy wet snow accumulation (unknown diameter)	No data	Total traffic disruption in the city of Yekaterinburg, widespread power outage
20 September 1986 (SW)	1 21.4 (23.1)	Heavy wet snow accumulation (unknown diameter)	No data	widespread power outages, >2000 trees were uprooted or broken
21–22 May 1993 (NW)	2 22.4 (35.9)	Wet snow accumulation up to 50 mm in diameter	No data	Widespread crop damage, power outage
6 June 1995 (NW)	6 47.6 (70.9)	Wind gusts up to 26 m s−1, wet snow accumulation up to 190 mm diameter	>100	Catastrophic windthrow (19,600 ha), catastrophic crop damage, traffic disruption
7 November 2010 (W)	1 34 (46)	Wind gusts up to 21 m s−1	70	Traffic disruption
25 April 2014 (S)	4 38.8 (52.2)	Wind gusts up to 25 m s−1	150	Traffic disruption
18–19 October 2014	8 23.1 (44.0)	Wind gusts up to 18 m s−1, ice accumulation (up to 20 mm diameter)	56	Traffic disruption
8–9 October 2015 (S)	9 30.2 (42.2)	Wind gusts up to 20 m s−1	No data	Catastrophic forest damage (5100 ha), traffic disruption
3–4 May 2024 (local)	1 30 (39)	Wet snow accumulation up to 50 mm in diameter	60	Traffic disruption, local damage to forests

). The total financial losses were documented only for the most severe 6 June 1995 snowstorm, and amounted to USD 32.6 million (150 billion rubles at the exchange rate in June 1995) [65]. This estimate is comparable with the same caused by snowstorms in North America [2], given the price difference and incomplete data on damage in Russia.

Although breaking and uprooting of trees under snow load is frequent in the study area, damage to forestry associated with HHS events attracts less attention. Between 1981 and 2025, two snowstorms caused widespread forest damage (Figure 3). In particular, the 6 June 1995 snowstorm damaged 225,000 hectares of forest stands [55], including 19,600 hectares that were completely destroyed. The snowstorm of 8–9 October 2015 in the North Ural completely destroyed about 5100 hectares of forest stands. In both cases, old-growth spruce–fir and mixed forests in the Visim and Vishera State Nature Reserves were most severely affected. Catastrophic forest damage in both cases was caused by a combination of several factors, namely extreme snowfall (up to 71 mm and 42 mm per day, respectively), severe winds (wind gusts exceeding 25 m s⁻¹ according to ERA5-Land data), heavy wet snow accumulation on trees, unfrozen soil, and heavy rainfall in the preceding days.

3.2. Synoptic-Scale Environments of HHS Occurrence

3.2.1. HYSPLIT-Based Backward Trajectories Analysis

The main characteristics of the five types of HHS events, classified according to the HYSPLIT-based backward trajectories, are shown in

Table 4. The main characteristics of HHS event types according to the HYSPLIT-based backward trajectory analysis.

HHS Event Type and Its Frequency	The Predominant Month of Occurrence	Number of the 72 h Backward Trajectories and Their Mean Length, km	Number of HHS Reports	Mean and Median 12 h Snowfall, mm	Maximum 12 h and 24 h Precipitation, mm	Maximum Observed Increase in Snow Depth, cm	Number Severe Wind Events (≥25 m s−1)
S (47.5%)	April	34/1852	54	25.2/24.3	38.8/57.5	45	2
SW (13.5%)	October	10/2873	18	24.4/21.6	39.2/44.0	38	0
W (13.5%)	April	10/3813	11	23.8/23.0	34.0/46.0	No data	0
NW (9.5%)	May	7/2930	17	29.6/26.1	47.6/70.9	64	1
Local (16%)	April	12/1858	16	24.0/22.2	37.9/48.0	43	0

. Most HHS reports (47.5%) are related to the S type, which indicates advection from the Caspian and Aral Seas. The highest density of the starting points of backward trajectories is observed over the Caspian and Aral Seas (Figure 7), indicating the importance of these water bodies as moisture sources, especially for the formation of HHS events in the spring. Note that the tracks of cyclones forming over the Caspian and Aral Seas cross over the eastern slope of the Ural into Western Siberia, which explains the higher frequency of HHS events in the eastern part of the study area.

Other types are much less common. HHS events on the western slope of the Urals are mainly produced by the advection from the west or southwest (W and SW types). Note also that the 72 h trajectories of W-type events are substantially longer than those of other types (Table 3), since these events form under a strong west-to-east midlevel jet. The mean and median snowfall intensity for the various HHS types differ slightly, except for the NW type events, which have the highest average snowfall amount. This difference can be explained by the seasonal characteristics of their occurrence (see Section 3.2.2), as well as the fact that the snowstorm on 6 June 1995, with the heaviest snowfalls, was of the NW type.

3.2.2. Composite Analysis

Figure 8 and Figure 9 show the ERA5-based composite charts for the HHS events classified into five types (S, SW, W, NW, and local) according to the 72 h backward HYSPLIT-based trajectories at 3000 m a.s.l. The composite fields differ significantly between different types of HHS events, but they also have a few common features. All HHS events are formed under a mid-tropospheric trough over the Ural region, while western and central Europe are under a mid-tropospheric ridge. Warm and moist advection in warm sectors of surface lows into the southeast Ural and Western Siberia, and cold and dry air mass advection from the north to European Russia are also common for most of the HHS events, except for the W and NW types. Other near-surface and air mass characteristics differ substantially for various HHS types.

S-type events form under a 700 hPa trough oriented southward along the Urals. The Western Ural is, in its rear part, under cold air advection (T₈₅₀ < −10 °C) from the north or northeast. At the same time, warm and moist air (with average T₈₅₀ > 0 °C, and average PW of 12–16 mm) spreads toward Western Siberia from the Caspian and Aral Seas. The S type is characterized by a strong ΔT₈₅₀ (about 12 °C/500 km) between the southeast and north part of the study region. These events occur under rapidly deepening surface lows that form over the Caspian and Aral Seas and propagate into Western Siberia. They are significantly deeper (the average MSLP in its center is <1000 hPa) than in other HHS types, except for the W type. In most cases, the highest snowfall is associated with the occlusion point. A substantial MSLP gradient induces strong winds and blizzards, which often accompany these events. In two cases, wind gusts reached 25 m s⁻¹ (Table 3). Strong near-surface easterly winds also contribute to the intensification of snowfall on the eastern slopes of the Ural Mountains. A good example was the case of the snowstorm of 9 October 2015, which caused substantial damage (Figure 10b). Most S-type events occur at T_2m ranging between 0 °C and −5 °C, except for the event of 6 March 2005, which formed under T_2m of about −10 °C. Most of these events occurred in spring, but a number of them also occurred in autumn and early winter. Three of the ten most damaging HHS events, namely 17 May 1981, 25–26 April 2014, and 8–9 October 2015 (Table 3), were of the S type.

SW-type events are formed in the forward part of a 700 hPa low, which is centered over the northwest of the study area. This contributes to the advection of warm (T₈₅₀ ≥ 0 °C) and moist (average PW of 12–16 mm) air masses from the Black Sea or Mediterranean Sea. The ΔT₈₅₀ reaches 12 °C/500 km (as for the S type), but it is stronger over the eastern part of the study area than over its western part. The surface low is centered to the west of the Ural, and to the south of the mid-tropospheric low center. Most of SW-type events are associated with warm fronts. These events mainly occur in the northern and central parts of the study area, on both sides of the Ural Mountains. Note that 7 out of the 10 events of the SW type occurred from October to December, whereas more than half of the other HHS events occurred in the spring. The snowstorm of 18–19 October 2014, induced by a deepening southwestern low [45], was a typical example of a SW-type event, and the most damaging among them.

W-type events are formed in the forward part of the 700-hPa trough (as are the SW type); however, there is no isolated cyclone on this isobaric surface. The 700 hPa geopotential height along the trough axis is somewhat lower than in other HHS types. W-type events form under a strong western mid-tropospheric jet, and their 700 hPa contour lines are less curved than in other cases (S, SW, NW, and local types). A warm and moist air mass (T₈₅₀ ranges from 0 °C to −3 °C, PW = 12–16 mm) spreads from the west or southwest, but the ΔT₈₅₀ is about 6 °C/500 km, which is half the value observed in other cases (S, SW, and NW types). W-type events are associated with deep surface lows centered west of the Ural, as in the SW type events. The mean MSLP is substantially lower than in other event types (<994 hPa). It is notable that eight out of ten W-type events occurred to the west of the Ural Mountains. Like SW events, they also tend to occur in the autumn and early winter (from September to December). Several W-type events were accompanied by strong winds (up to 20–22 m s⁻¹), including the snowstorm of 7 November 2010, which was characterized by the highest snowfall intensity (34 mm 12 h⁻¹) and the greatest damage (Table 3).

NW-type events are the least frequent, but most damaging compared with other types. Thus, the snowstorms of 3 May 1984, 22 May 1993, and 6 June 1995 (Table 3) are examples of this type. All of them occurred in the spring or early summer, between 28 April and 6 June. NW-type events occur under 700-hPa lows centered over the central part of the study region. These events are associated with strong Arctic intrusions from the Kara Sea to the Ural and eastern European Russia, when cold and dry air mass pushes warm and moist air to the south and east. The ΔT₈₅₀ reaches 12 °C/500 km, indicating a strong temperature contrast between the Arctic and temperate air masses. In the MSLP field, there is a well-pronounced high-pressure system centered over the Kola Peninsula, and a deepening low over the South Ural. In most cases, a low-pressure system sharply deepens over the Ural Mountains, triggering intense precipitation, typically on their eastern slope. HHS events have been associated with an occluded front in the rear part of a surface low [51,52]. All NW-type events form at 2 m temperatures close to 0 °C and are accompanied by heavy wet snow accumulation on power lines and trees, which significantly increases damage. Figure 10a shows the spatial distribution of the daily snowfall that occurred during the most damaging NW-type snowstorm (6 June 1995).

Similarly to the S and NW types, the local-type events form under mid-tropospheric lows centered over the Ural Mountains. The air mass in the region where the HHS events occur has a moderate PW (10–12 mm). As for S-type events, there is warm advection towards Western Siberia and cold advection towards the western Urals and the surrounding regions of the ER, but the ΔT₈₅₀ is about 6–7 °C/500 km, which is substantially lower than that observed for the HHSs of the S, SW, and NW types. The local-type events are associated with so-called “cyclones and waves of local origin” which form over the Urals or adjacent regions. The MSLP gradient is smaller than for other HHS types, so the local type events are not accompanied by strong winds. These events have occurred in various seasons, from September (the earliest on 20 September 1986) to May (the latest on 21 May 1993), but are more typical for spring, and their 2 m air temperatures were close to 0 °C. Several of these events caused substantial damage due to heavy wet snow accumulation, for example, on 20 September 1986, 21 May 1993, and 4 May 2024.

4. Discussion and Conclusions

In this study, we have analyzed for the first time the main climatic characteristics of the HHS events in the Ural region, as well as the synoptic-scale environments of their formation for the period of 1981–present. Using both ground-based observations and the ERA5 reanalysis data, we also examined the spatial distribution and trends of heavy snowfalls that did not reach the HHS threshold (20 mm 12h⁻¹). We compiled a dataset comprising 116 HHS reports, which are associated with 56 different HHS events. The number of HHS reports associated with one HHS event ranged from zero to nine. Note that the sample of HHS reports and events is incomplete and inhomogeneous in time, due to the reduction in the weather station network that occurred in the 1990s.

Of these 116 reports, the maximum 12 h snowfall amount was 47.6 mm, while the maximum 24 h precipitation was 70.9 mm. Of these cases, 19 involved mixed precipitation (wet snow/snow with rain), with precipitation amounts ≥ 30 mm 12 h⁻¹. Except for these cases, the heaviest 12 h snowfall was 38.8 mm 12 h⁻¹. The degree of related damage is substantially determined by snowfall intensity, along with accompanying weather events (severe wind and/or wet snow accumulation). Three compound extremes of snowfall and wind (snowfall ≥ 20 mm 12 h⁻¹ and wind gusts ≥ 25 m s⁻¹), and two compound snowfall and wet snow accumulation extremes were reported by weather stations. The vast majority of HHS events occurred when the air temperature at 2 m was between 0 °C and −5 °C, in line with global estimates [27].

The key feature of the spatial distribution of the HHS events is that it did not correspond to that of the S_{annual_mean}, according to both the observations and the ERA5 data. While the S_{annual_mean} has a well-pronounced maximum on the western slope of the Northern Urals, most of the HHS events occurred east of the Ural Mountains. Note that the highest values of the ERA5-based characteristics of snowfall extremes (S_95p and S_99p) are also shifted to the eastern slope of the Ural and even to western Siberia, as well as values obtained according to observations. Correlations between the observed and ERA5-based characteristics of snowfall extremes are statistically significant, but SR does not exceed 0.6.

According to both the observations and the ERA5 data, there is no statistically significant trend in the frequency of heavy snowfall events throughout most of the study region. Despite several weather stations ceasing observations in the 1990s, the frequency of observed HHS events has not decreased. The lack of significant changes in snowfall extremes corresponds well to previous ERA5-based estimates [25]. There are two peaks in the seasonal distribution of HHS events, in April–May and October, when the mean air temperature is close to 0 °C (in line with [27]). The springtime maximum of HHSs is more pronounced in terms of both the number of reports and intensity. In winter, PW is insufficient for the formation of HHS events. Under climate warming, the HHS occurrence has decreased in late spring but increased slightly in the cold season. In the future, climate warming will increase atmospheric moisture content, contributing to intense snowfall, while higher temperatures will decrease the duration of the snowfall season [38,81]. According to the ensemble of CMIP6 (Coupled Model Intercomparison Project) global climate models [38], a slight increase in heavy snowfalls (up to 20%) is predicted until the mid-21st century within the study area.

Information on damage associated with HSS events has been summarized for the Ural region for the first time. The most serious damage is associated with compound heavy snow and severe wind events, with heavy wet snow accumulation, as well as with late spring and early autumn events, which cause widespread damage to crops and trees. In addition, two snowstorms (6 June 1995 and 8–9 October 2015) induced severe damage to forestry, destroying at least 24.7 thous. ha of forest stands in the mountainous parts of the Middle and Northern Urals, respectively. We highlighted the ten most damaging HHS events, six of which occurred in May, early June, or in September. As well as in other regions with temperate climates [82], climate warming has contributed to a decrease in the frequency of late spring and early autumn snowfall events and their associated damage.

Using the 72 h backward trajectories according to the NOAA HYSPLIT model (at 3000 m height), we have classified all HHS events into five types (namely S, SW, W, NW, and local) according to their air mass origin. In line with [30], we highlighted the similarities and differences in these types through the analysis of the composite structure, according to the ERA5-based fields of H₇₀₀, MSLP, PW, and T₈₅₀. The main finding is that the spatial distribution of HHS events in the Ural region (mainly their higher frequency in the east slope of the Ural Mountains compared with the west slope) is determined by the synoptic-scale environments of their formation. There are also substantial differences in the seasonal distribution of the HHS events that occur in various synoptic-scale environments. Thus, the S-type events (which account for 46% of the HHS reports) are driven by the cyclones that form over the Aral and Caspian Seas and propagate into Western Siberia. Consequently, S-type events mainly occur in the eastern part of the study area, where the S_{annual mean} is lower than on the western slope. Conversely, W- and SW-type events are more prevalent in the western part of the study area, as they are associated with the advection of warm, moist air from the west and southwest. They are primarily confined to the autumn period. NW-type events are infrequent (accounting for 14% of the HHS reports), but cause substantial damage when they occur in late spring and even in early summer. These events are associated with intrusions of Arctic air masses from the Kara Sea, and are characterized by higher 12 h snowfall amounts than other types. Local events are not associated with the advection of maritime air masses. Most of these events occur in April or May and are often accompanied by heavy wet snow accumulation on the wires (as well as NW type events).

Thus, the presented results provide a better understanding of the climatic characteristics of HHS events in the Ural region (primarily their spatial distribution and seasonality) in connection with the synoptic-scale environments of their occurrence. However, their expected changes with global warming remain uncertain and could be the subject of future research.