Georeferenced Snow Depth and Snow Water Equivalent Dataset (2025) from East Kazakhstan Region

Abstract

In this work, we present the Snow Depth and Snow Water Equivalent Dataset for specific areas located in the East Kazakhstan Region that can be exploited to monitor and understand water resource dynamics in mountain regions. The present dataset represents a georeferenced collection of snow depth, snow density, and derived snow water equivalent (SWE) measurements obtained through manual snow surveys. Snow survey observations were conducted during field campaigns in the East Kazakhstan Region during the period of maximum snow accumulation from 27 February to 6 March 2025. Snow survey sites were selected to maximize coverage of diverse landscape settings and snow accumulation conditions. In total, 111 snow survey sites were established across the East Kazakhstan Region, and 2331 snow depth measurements and 555 snow density measurements were collected. In post-field (laboratory) processing, snow water equivalent (SWE) was calculated for all snow survey sites based on measured snow depth and snow density values.

Dataset: https://zenodo.org/records/18242602 (accessed on 11 February 2026).

Dataset License: CC-BY 4.0

1. Summary

Snow cover is of critical importance because it is a key component of the hydrological cycle [1,2], serving as a temporary storage of water for a large part of the world’s population [2]. This water is gradually released later, sustaining river flow during the spring and summer months, which plays a crucial role in water supply and ecosystem preservation [3,4,5]. In addition, snow is an important indicator of climate change and is included among the 55 Essential Climate Variables (ECVs) defined by the Global Climate Observing System (GCOS) of the World Meteorological Organization (WMO) [6]. At the same time, snow also poses serious natural hazards, as extreme snowfall events cause economic, infrastructural, and personal damage [7], while avalanches disrupt transport and communication networks and limit accessibility of affected areas [8].

Monitoring snowpack dynamics in many regions of the world is complicated by the sparse network of ground stations. Snow exhibits strong spatial heterogeneity driven by local environmental and physiographic factors, while frequent cloud cover and the spectral similarity between melting snow and ice further complicate remote sensing retrievals [9]. In addition, direct field measurements of this geophysical parameter are often impossible, especially in hard-to-reach areas with complex terrain typical of mountain catchments [10]. For example, traditional snow cover monitoring in high-mountain regions is often complicated by difficult access to the terrain and severe weather conditions during the winter period [11]. At the same time, ground observations provide detailed data at the local scale; however, such measurements remain very limited at the global scale [12]. The number of automatic stations in mountain regions also remains small [13], which limits the ability to conduct spatially distributed monitoring of this variable. At the same time, remote sensing methods provide an effective tool for continuous, large-scale monitoring of snow cover dynamics [14], enabling the retrieval of detailed information on local snow distribution patterns in observation-sparse regions [15,16]. The use of satellite observations improves the accuracy of snow cover mapping. However, increasing uncertainty associated with climate change, including rising risks of floods and droughts, makes direct measurements of this geophysical parameter particularly important [17]. Ground observations serve as reference data for the development and calibration of satellite-derived snow products [18]. They also remain the primary benchmark for validating new approaches to snow depth measurements, including drone-based LiDAR scanning [19], as well as conventional satellite methods [20]. In addition, the lack of ground measurements reduces the accuracy of geophysical parameters estimated using models, since such data are often used as input variables [10].

During the year 2024, the project “Development of a system for forecasting catastrophic floods in the East Kazakhstan Region using remote sensing data, GIS technologies, and machine learning” was started. The presented dataset was developed in the framework of this project. The field survey dataset has been made publicly available. Access to these data is crucial to evaluate the snowpack spatial organization of the East Kazakhstan Region. In addition, this dataset addresses the critical need for comprehensive and accessible data on the snowpack in mountainous areas.

The structure of the provided data is described in Section 2. The data collection methods and snow distribution patterns in the area are described in Section 3.

2. Study Area and Data Description

2.1. Study Area

The study was conducted in the East Kazakhstan Region (EKR). Within its current administrative boundaries, the region covers an area of 97.8 thousand km² (Figure 1). A significant part of the territory is occupied by the Altai Mountains and their adjacent foothills [21]. The Kalbinsky Ridge is separated from the Altai by the Irtysh River valley. In the southern part of the region lie the Saur and Tarbagatay Ranges, which connect the Altai with the mountain systems of the Tien Shan, while being separated from the Altai by the Zaisan Depression [22]. The region has a strongly continental climate, determined by its location in the central part of Eurasia and characterized by large seasonal temperature contrasts, with long, cold winters and hot summers. The EKR is dominated by steppe and semi-desert foothill landscapes, while forest–steppe, forest, alpine, and nival environments are found in the mountainous areas [23].

2.2. Dataset Overview

This paper presents the Snow Depth and Snow Water Equivalent Dataset (2025) from the East Kazakhstan Region, deposited on the open-access Zenodo platform, available at https://zenodo.org/records/18242602, (accessed on 11 February 2026).

The file _EKR_SSP_2025_summary.csv contains information on the mean values of snow depth, snow density, and snow water equivalent (SWE) at the snow survey sites. It also describes the main geographical characteristics of the sites.

Structure of data for the summary table of snow survey observations:

• Point ID—sequential number of the snow survey site (from 1 to 111);
• Point Name—name of the measurement point;
• Latitude—latitude WGS84 (in degrees);
• Longitude—longitude WGS84 (in degrees);
• Elevation—elevation above sea level (m);
• Slope—slope angle (in degrees);
• Aspect—slope aspect (in degrees, 0–360);
• Snow Depth—mean snow depth at the snow survey site (cm);
• Snow Density—mean snow density at the snow survey site (g/cm³);
• Snow Water Equivalent (SWE)—mean SWE at the snow survey site (mm);
• Landform—landform type (or terrain type);
• Land Cover—land cover type (or vegetation/surface cover type);
• Landscape Province—landscape province (or physiographic province/region).
• Category values:
• Landform: depressions, ridges, valleys;
• Land Cover: coniferous forest, deciduous forest (small-leaved), grassland, shrubland;
• Landscape Province: Altai Mountains, Kalbinsky Range, Zaisan Depression, Saur Ridge.

The files 1_EKR_SSP_2025.csv—111_EKR_SSP_2025.csv contain data on the primary (raw) measurements of snow depth and snow density at individual snow survey sites (1–111). Each file includes a single worksheet with a consistent tabular structure.

Structure of data for tables of primary (raw) snow survey measurements at the site:

• id_point—sequential number of an individual snow depth measurement within a specific snow survey site (from 1 to 21);
• day—day of measurement;
• month—month of measurement;
• year—year of measurement;
• point_name—name of the measurement point;
• depth_cm—snow depth at the measurement point (measured with a snow probe/ruler), cm;
• sample_weight_1_g—mass/weight of snow sample №1 taken with the snow sampler (gravimetric tube), g;
• sample_weight_2_g—mass/weight of snow sample №2 taken with the snow sampler (gravimetric tube), g (if present);
• sample_weight_3_g—mass/weight of snow sample №3 taken with the snow sampler (gravimetric tube), g (if present);
• sample_weight_4_g—mass/weight of snow sample №4 taken with the snow sampler (gravimetric tube), g (if present);
• sample_depth_1_cm—thickness/height of snow sample №1 taken with the snow sampler (gravimetric tube), cm;
• sample_depth_2_cm—thickness/height of snow sample №2 taken with the snow sampler (gravimetric tube), cm (if present);
• sample_depth_3_cm—thickness/height of snow sample №3 taken with the snow sampler (gravimetric tube), cm (if present);
• sample_depth_4_cm—thickness/height of snow sample №4 taken with the snow sampler (gravimetric tube), cm (if present);
• mean_density_g_cm3—mean snow density at the measurement point, calculated from all collected samples, g/cm³.

The tables can be read in any spreadsheet or data analysis environment, such as Microsoft Excel, R or Python pandas).

3. Methods

3.1. Data Collection and Handling

All field operations were conducted as specified in Instructions for hydrometeorological stations and posts [25].

Snow survey sites were selected using a stratified, non-random sampling approach aimed at covering the main physiographic and landscape conditions of the East Kazakhstan Region. Site selection considered landscape province, landform type (ridges, valleys, depressions), elevation range, and land cover type. Field accessibility and safety constraints were also taken into account; therefore, due to difficult access to some high-mountain and highly dissected areas, the sampling is partly biased toward more accessible locations along roads and trails. The sampling design was not intended to achieve strict statistical representativeness, and no landscape type was intentionally over- or underrepresented. The primary objective was to ensure broad spatial coverage of contrasting snow accumulation conditions during the period of maximum snow storage.

The field data collection campaign took place from the third week of February to the first week of March near the time of peak snow accumulation. Snow surveys were conducted by teams of two to four individuals. Survey teams made their way to 111 measurement sites. At these sites, the survey teams measured snow depth and density, recorded snow conditions and GPS locations and took photographs.

Snow depth and density were measured at several locations within each measurement site. These measurements were taken along straight profiles (referred to as transects) at a fixed interval between locations. Two perpendicular transects were established at each snow measurement site, with 21 depth measurements (10 for each transect and one center measurement; the distance between measurements was 5 m) and five measurements of snow density.

The use of 21 depth measurements per site allowed spatial variability caused by wind redistribution, vegetation, microrelief, and other factors to be captured. At the same time, snow density, as a rule, is characterized by lower within-site variability at the scale of a single snow survey plot. The samples were evenly distributed, with one taken at the intersection of the transects (center) and one along each perpendicular transect at intervals, allowing potential variability to be captured. This methodology corresponds to standard hydrometeorological protocols [25] and provides a balance between accuracy and efficiency of field operations. Field observations showed that variability of snow density within the survey plot was substantially lower than variability of snow depth, confirming the adequacy of the adopted sampling scheme.

Snow depth was measured with a portable snow-measuring stick with dimensions 180 × 4 × 2 cm, with 1 cm divisions. The weighing snow gauge VS-43, shown in Figure 2, was also used for snow sampling to determine density, because SWE is the product of snow depth and the depth-averaged snowpack density. The weighing snow gauge VS-43 is a steel tube with a cross-sectional area of 50 cm² and a length of 60 cm. The exterior of the tube is scribed with a height scale, from which snow depth was recorded to the nearest centimeter. Today, the tubes are considered one of the standard methods for ground-truthing SWE in deep snowpacks. Snow sampling was performed by inserting a steel tube into the snowpack. Then, the snow core was weighed using the balance, which had an accuracy of 5 g (one gradation on the balance scale, with zero on the scale corresponding to the weight of the tube).

The fields “sample_weight_1–4_g” and “sample_depth_1–4_cm” are intended to record up to four replicates (separate snow core samples) collected at the same density measurement point using the VS-43 weighing snow sampler (tube length 600 mm, cross-sectional area 50 cm²). Each field “sample_depth_n_cm” reflects the height (thickness) of the n-th extracted snow core (cm), while “sample_weight_n_g” indicates the mass of the corresponding sample (g). When snow depth is ≤600 mm, a single core is sufficient. At points with greater snow depth (up to four cores in this dataset), sequential samples are taken to fully cover the snow layer, after which the mean density (mean_density_g_cm3) is calculated from all available samples. Minor differences between total snow depth measured with the snow probe (depth_cm) and the summed core lengths (sum of sample_depth) may occur due to natural snow compaction during sampling, instrumental uncertainty (±1 cm for the VS-43), or micro-scale variability of the snow layer, and are not considered measurement errors. This approach is consistent with standard protocols and ensures reliable SWE calculation under conditions of deep snow cover.

The snow density was calculated using the following Formula (1):

$$\rho ={\displaystyle \frac{m}{SD·s}}$$

(1)

where $\rho$ is snow density (g/cm³), m is snow weight, SD is snow depth and s is cross-sectional area.

Snow density, when multiplied by depth, gives SWE (2). SWE was expressed as the depth of the water layer in millimeters.

SWE = 10·ρ·SD

(2)

All measurements of snow depth and density were performed following standard procedures using calibrated instruments. Quality-control procedures included visual verification of field records, consistency checks between snow depth and density measurements, and verification of SWE calculations. Statistical removal of outliers was not applied because extreme values reflect real spatial variability of snowpack characteristics. Measurement points with no snow cover were coded using zero values for snow density and SWE. Snow absence was recorded when no snow cover with a thickness ≥ 0.5 cm was detected visually or instrumentally using a snow probe.

3.2. Snow Distribution Patterns

3.2.1. Snow Distributions Across Landscape Provinces and Landforms

Snow distribution was analyzed using two complementary spatial classifications, with Figure 3 illustrating the distribution of snow cover characteristics across landscape provinces and reflecting regional differences, while Figure 4 shows how landform types (ridges, valleys, and depressions) influence snow cover distribution within the region at a more local scale.

Figure 3 shows that snow depth and SWE have higher medians and wider interquartile ranges in the Altai Mountains and the Kalbinsky Range, together with the presence of outliers. This is consistent with

Table 1. Snow depth, density, and SWE statistics by landscape province.

Landform	Snow Depth			Snow Density			SWE			Number of Observations
	Mean ± Std	Min	Max	Mean ± Std	Min	Max	Mean ± Std	Min	Max
Altai Mountains	75.05 ± 20.45	41	132	0.23 ± 0.03	0.18	0.35	177.76 ± 65.34	101	349	42
Kalbinsky Range	57.17 ± 29.62	14	144	0.24 ± 0.04	0.18	0.34	141.21 ± 83.69	45	376	29
Saur Ridge	32.64 ± 7.23	14	44	0.21 ± 0.04	0.15	0.29	71.57 ± 12.35	53	95	14
Zaisan Depression	28.92 ± 16.42	0	54	0.22 ± 0.06	0	0.3	64.39 ± 36.10	0	123	26

, where the Altai Mountains have the highest mean snow depth (75.05 cm) and SWE (177.76 mm), while the Kalbinsky Range shows the widest range of SWE values, from 45 to 376 mm.

In contrast, the Saur Ridge exhibits low median values in Figure 3 and narrow ranges in Table 1, with snow depth varying only between 14 cm and 44 cm and SWE between 53 mm and 95 mm, indicating low variability. The Zaisan Depression shows similarly low central values but much wider ranges (snow depth from 0 cm to 54 cm; SWE from 0 cm to 123 mm), which explains the presence of outliers in Figure 3.

Regional contrasts in snow density are much weaker than those observed for snow depth and SWE. Mean density values vary only slightly among regions, with mean values between 0.21 and 0.24 g/cm³, although variability is highest in the Zaisan Depression.

Snow conditions differ between landform types and among snow variables (Figure 4;

Table 2. Snow depth, density, and SWE statistics by landform.

Landform	Snow Depth			Snow Density			SWE			Number of Observations
	Mean ± Std	Min	Max	Mean ± Std	Min	Max	Mean ± Std	Min	Max
Depressions	22.89 ± 12.59	0	44	0.21 ± 0.06	0	0.30	49.67 ± 25.27	0	97	18
Ridges	62.00 ± 27.45	11	144	0.24 ± 0.04	0.15	0.35	150.88 ± 77.60	30	376	75
Valleys	53.17 ± 24.35	15	98	0.21 ± 0.03	0.18	0.25	112.61 ± 56.41	37	221	18

). Ridges contain the largest amount of snow, with an average snow depth of about 62 cm and SWE of 150.88 mm. They also show the widest spread of values, with snow depth ranging from 11 to 144 cm and SWE from 30 to 376 mm. Valleys have slightly lower snow amounts than ridges, with mean snow depth around 53 cm and SWE of 112.61 mm, but their overall patterns remain similar, showing moderate variability. Depressions stand out as the least snow-covered landform. Average snow depth is only about 22.89 cm and SWE about 49.67 mm, and although most values are low, snow-free conditions are also observed.

Snow density behaves differently from snow depth and SWE. Its mean values are similar across landforms (around 0.21–0.24 g/cm³), but variability is highest in depressions, where density ranges from zero to 0.30 g/cm³, while valleys show the most uniform density values.

3.2.2. Elevation Gradients of Snow Depth, Density, and SWE

The elevation gradients of snow characteristics were calculated separately for each landscape province using linear regression (LR), with the Theil–Sen estimator applied as a robust alternative based on the median of pairwise slopes, reducing sensitivity to outliers, with a 95% confidence level used for slope estimation. The elevation ranges of snow survey sites by landscape province are summarized in

Table 3. Elevation ranges of snow survey sites by landscape province.

Landscape Province	Number of Observations	Min. Elevation (m)	Max. Elevation (m)
Altai Mountains	42	479	1610
Kalbinsky Range	29	382	1316
Saur Ridge	14	805	955
Zaisan Depression	26	405	652

.

The relationship between snow depth and elevation varies significantly between landscape provinces. LR results indicate negligible and statistically insignificant gradients in the Altai Mountains and the Zaisan Depression. A pronounced positive gradient of about +6.6 cm per 100 m is found in the Kalbinsky Range ridge; on the contrary, there is a negative gradient of about −8.3 cm per 100 m in the Saur Ridge. Both relationships are statistically significant and are associated with moderate values of explained variance according to

Table 4. Elevation gradients of snow depth by landscape province derived from LR analysis.

Landscape Province	Slope per 1 m	Gradient per 100 m	Intercept	${\mathit{R}}^2$	p Value
Altai Mountains	0.006	0.580	68.464	0.008	0.57
Kalbinsky Range	0.066	6.558	3.747	0.424	0.00
Saur Ridge	−0.083	−8.268	107.039	0.335	0.03
Zaisan Depression	0.000	0.048	28.677	0.000	0.99

.

Compared with LR, the Theil–Sen estimator results presented in

Table 5. Elevation gradients of snow depth by landscape province derived from Theil–Sen slope method.

Landscape Province	Slope per 1 m	Gradient per 100 m	Intercept
Altai Mountains	0.012	1.223	56.237
Kalbinsky Range	0.052	5.245	5.993
Saur Ridge	−0.069	−6.857	94.902
Zaisan Depression	−0.028	−2.753	41.960

show a steeper positive slope in the Altai Mountains, increasing from 0.58 to 1.22 cm per 100 m. In the Kalbinsky Range, the slope decreases from 6.56 to 5.245 cm per 100 m, while in the Saur Ridge the negative gradient becomes less steep, changing from −8.268 to −6.857 cm per 100 m. The largest change occurs in the Zaisan Depression, where the LR slope is near zero at +0.048 cm per 100 m but becomes weakly negative at −2.753 cm per 100 m with the Theil–Sen method, although both values remain much smaller than gradients in the other provinces.

Snow density shows generally weak or absent elevation dependence across all provinces (

Table 6. Elevation gradients of snow density by landscape province derived from LR analysis.

Landscape Province	Slope per 1 m	Gradient per 100 m	Intercept	${\mathit{R}}^2$	p Value
Altai Mountains	0.00000	0.000	0.229	0.001	0.82
Kalbinsky Range	0.00006	0.006	0.197	0.192	0.02
Saur Ridge	−0.00001	−0.001	0.217	0.000	0.97
Zaisan Depression	−0.00001	−0.001	0.225	0.000	0.92

and

Table 7. Elevation gradients of snow density by landscape province derived from Theil–Sen slope method.

Landscape Province	Slope per 1 m	Gradient per 100 m	Intercept
Altai Mountains	0.000	0.000	0.230
Kalbinsky Range	0.000	0.006	0.191
Saur Ridge	0.000	0.000	0.200
Zaisan Depression	0.000	0.000	0.220

). LR indicates no statistically significant relationship in the Altai Mountains, Saur Ridge, and Zaisan Depression, while a weak but statistically significant positive gradient appears in the Kalbinsky Range, although with $R^2$ value. The Theil–Sen results presented in Table 7 support this pattern, confirming a weak positive gradient only in the Kalbinsky Range, while the small negative slopes suggested by LR in the Saur Ridge and Zaisan Depression reduce to near zero. Overall, both approaches show that snow density exhibits little or no systematic change with elevation across the provinces.

With respect to SWE, LR shows that only the Kalbinsky Range has a statistically significant SWE increase with elevation (+19.9 mm per 100 m), while gradients in the Altai Mountains (+1.8 mm), Zaisan Depression (+1.6 mm), and Saur Ridge (–9.9 mm) are weak and not significant. Theil–Sen slopes confirm the positive trend in the Altai Mountains and Kalbinsky Range, although the increase in the Kalbinsky Range becomes smaller (+13.4 mm per 100 m compared to +19.9 mm per 100 m from LR). In the Zaisan Depression, the slope becomes slightly negative (–3.4 mm per 100 m), while the negative gradient in the Saur Ridge remains nearly unchanged (

Table 8. Elevation gradients of snow water equivalent (SWE) by landscape province derived from LR analysis.

Landscape Province	Slope per 1 m	Gradient per 100 m	Intercept	${\mathit{R}}^2$	p Value
Altai Mountains	0.018	1.809	157.221	0.008	0.57
Kalbinsky Range	0.199	19.863	−20.611	0.487	0.00
Saur Ridge	−0.099	−9.909	160.739	0.165	0.15
Zaisan Depression	0.016	1.564	56.392	0.001	0.86

and

Table 9. Elevation gradients of snow water equivalent (SWE) by landscape province derived from Theil–Sen slope method.

Landscape Province	Slope per 1 m	Gradient per 100 m	Intercept
Altai Mountains	0.016	1.625	138.047
Kalbinsky Range	0.134	13.415	−8.000
Saur Ridge	−0.098	−9.850	161.134
Zaisan Depression	−0.034	−3.379	75.021

).

Despite the expected increase in snow accumulation with elevation, linear regressions between elevation and snow characteristics exhibit low coefficients of determination (R²). Additional analysis indicates that key assumptions of LR, particularly linearity of the relationship and homoscedasticity of residuals, are not fully satisfied within the dataset. This suggests that elevation alone is not the dominant factor controlling the spatial variability of snow depth, snow density, and SWE within the studied provinces. Regression analysis was therefore used not to develop predictive models but to estimate the average elevation gradient of snow characteristics. Consequently, even with low R² values, the regression slope still represents the average change in snow characteristics with elevation.

The negative altitudinal gradient identified in the Saur Ridge, and the similarly weak negative gradients observed in the Zaisan Depression, likely reflect regionally specific processes controlling snow redistribution. In the Zaisan Depression, this pattern probably reflects generally low snow accumulation and the limited elevation range of observations, which reduces the influence of elevation on snow distribution. Under the arid climate conditions of southeastern East Kazakhstan, the main contributing factors may include wind redistribution, orographic effects, and temperature inversions. This pattern in the Saur Ridge appears to be a local phenomenon associated with the relative isolation and lower moisture availability compared to provinces where positive gradients prevail. Similar inverse gradients have previously been observed in mountain regions of Central Asia, where they are linked to the transition from snowfall to rainfall at mid-elevations and to wind-driven snow removal from exposed high-altitude slopes. Several studies also report accelerated warming with increasing elevation and a negative correlation between temperature and snow depth, which can lead to faster snowmelt and potentially produce inverse patterns in snow distribution [26,27,28]. In addition, the limited elevation range of snow survey observations in the Saur Ridge (805–955 m) may also influence the statistical significance of the detected relationship.

For the calibration of regional snow and hydrological models, these findings indicate the need to account for local drivers such as wind and topographic controls, as well as regional climatic characteristics, in order to improve forecast accuracy.

4. User Notes

Collecting snowpack observations and modeling the snowpack can be extremely challenging due to its high spatial and temporal variability in mountainous areas. This dataset can be used for various applications. It is best suited for users wishing to achieve the following:

• calibrate hydrological models;
• do reanalysis;
• compare raw data among land cover types;
• validate data from drone-borne scanning and satellite-based methods;
• determine baseline conditions for future climate projections;
• analyze climatic and basin drivers.