Snow Cover Inversion Driven by Dzud Events in Mongolia from 2000 to 2024

Abstract

Amid global climate change and extreme weather conditions, sudden dzud events in arid grassland regions inflict severe disasters on herders, livestock, transportation, and the economy. In particular, Mongolia experiences frequent dzud events in recent years, bringing devastating consequences. However, studies on the spatiotemporal distribution characteristics of snow cover during dzud events in Mongolia remain relatively scarce and fail to adequately explain the anomalous features and impacts of extreme snowfall. Therefore, this study examined the spatiotemporal distribution characteristics of snow in the five most severe dzud events in Mongolia from 2000 to 2024. We utilized the Normalized Difference Snow Index (NDSI) extraction method based on 500 m resolution MODIS10A1 data, with the results validated against 10 m resolution Sentinel-2 imagery. The study produces several interesting results: (1) Snow cover in Mongolia generally increases from south to north with rising terrain elevation. Although its interannual variation is highly unstable, a slight decreasing trend is observed over the past 25 years. (2) Significant regional differences form a fan-shaped snow distribution pattern centered around 45–52° N, with trend analysis indicating intensification in the west and weakening in the east, except for extreme weather events. (3) During dzud events, the snow cover fraction (SCF) generally exceeds the multi-year average, exhibiting a pronounced and abrupt rise, while snow cover and livestock mortality fluctuate in synchrony. By revealing the spatiotemporal distribution patterns of snow during dzud years in Mongolia, this research provides an evidence-based reference for the understanding of extreme winter climatic events and disaster risk reduction in arid grassland regions.

1. Introduction

In the background of global climate change, extreme climate events occur more frequently, causing multiple impacts on the natural environment and socio-economy, such as traffic disruptions, livestock freezing damage, and a decline in human quality of life [1,2,3]. Winter snow disaster (called “dzud” in Mongolian) is one of the common climate disasters in mid–high latitudes and high-altitude regions [4]. Characterized by a sudden onset and long duration [5], dzud events have severe impacts on grassland ecosystems, livestock production, and regional climate circulation and also threaten regional food and energy security as well as ecological sustainability [6,7,8]. Mongolia is an inland country with arid and semi-arid grasslands, where dzud events occur frequently. From 1940 to 2015, the temperature in Mongolia increases by 2.24 °C, which is much higher than the global average [9]. Nevertheless, in the winter of 2023, Mongolia experiences the most severe snowfall in nearly 50 years, with temperatures dropping to −40 °C and loss of more than 5 million livestock (https://news.un.org/en/, accessed on 29 August 2025). The Eurasian steppe region, with Mongolia as a representative area, experiences frequent and severe winter disasters, which seriously affect the production and livelihoods of herders as well as regional ecological stability [10].

An increasing number of scholars conduct extensive studies on dzud events based on social survey data such as key informant interviews and herder household investigations [11,12,13]. Extreme dzud events, such as the 2009–2010 disaster, become a major focus of case studies [14,15,16,17], prompting investigations into their underlying mechanisms and impacts on climate, ecology, livestock management, and social organization. Sternberg et al. analyze the relationship between drought and dzud but do not find a direct connection [18]. Joly et al. employ regional-scale statistical data to explore the interaction between dzud and pasture health and observe a decline in livestock productivity prior to severe dzud events in the study area [19]. Extremely cold weather is widely regarded as having a significant influence on dzud occurrence and livestock mortality [20,21], whereas fodder shortages and poor preparation of winter fenced pastures are identified as key contributors to large-scale livestock deaths [22,23]. Soma et al. emphasize the importance of reducing winter dzud disaster losses through summer isolation grazing of specific livestock groups, controlled breeding, and early culling or sale of weaker livestock [24]. Shestakovich applies spatial regression models to estimate livestock mortality during dzud events and reveals that ecological diversity contributes to regional patterns of dzud, particularly in Mongolia’s mountainous areas, northern regions, and the Gobi Desert, and that herders’ vulnerability is increasingly exacerbated by the rapid growth in livestock numbers [25]. According to Xu et al., Mongolia’s deteriorating natural environment increasingly hinders livestock development, requiring long-term research and strategies to combat dzud [26]. Vova et al. offer a positive perspective on the effects of dzud from the viewpoint of ecological recovery: snow-rich winters may lead to an earlier NDVI by 10–20 days and reduced livestock numbers ease grazing pressure, promoting rapid post-disaster vegetation recovery and playing a positive role in ecosystem restoration [27]. In addition, some scholars deepen research on snow cover pattern changes and extensively discuss the historical evolution of snow-related variables and their relationships with climate factors in key regions, including the globe [28], Northern Hemisphere [29], Tibetan Plateau [30,31], United States [32], Central Asia [33], and Mongolian Plateau [34,35].

In the field of snow cover monitoring in Mongolia, existing research utilizes multi-source remote sensing data, integrated with ground observations and algorithm optimization, to elucidate the spatiotemporal characteristics of snow cover dynamics. For instance, Li explores the spatiotemporal variations of snow cover and seasonal frozen ground in Northern China and Mongolia from 1988 to 2010 using passive microwave remote sensing data. By applying the Goodison snow algorithm, this study estimates the snow onset date, duration, and end date, and provides the estimation of the time lag between snow melt-off and the thawing of seasonal frozen ground [36]. Focusing on the headwaters of the semi-arid Sugnugur catchment in the Khentii Mountains of northern Mongolia, Munkhdavaa conducts a spatial analysis of seasonal Snow Cover Duration (SCD) at a 30 m resolution by fusing the high spatial resolution imagery with the daily temporal resolution of moderate-resolution MODIS snow products (2000–2017) [37]. Furthermore, SA analyzes the spatiotemporal variations in snow cover and grassland phenology on the Mongolian Plateau from 2001 to 2018 by employing MOD10A1 snow data and MOD13A1 Normalized Difference Vegetation Index (NDVI) data [38]. These findings contribute to a deeper understanding of the “snow-soil-vegetation” interaction mechanisms within the context of climate warming.

These researchers primarily focus on physical snow parameters while overlooking the critical factor of disaster-induced losses. Furthermore, there is a lack of updated research covering long-term temporal scales. Conversely, numerous scholars conduct research specifically targeting dzud events. For example, Chadraabal focuses on two dzud events occurring a decade apart (1999–2000 and 2009–2010), examining societal responses and the effectiveness of countermeasures adopted to mitigate the disasters [12]. Chunling characterizes the regional vulnerability of herders to the 2009/2010 winter dzud by applying Principal Component Analysis (PCA) to a comprehensive multi-year provincial dataset, integrating key socioeconomic factors that explain major spatiotemporal variations in Mongolia [14]. However, these studies do not involve snow cover retrieval.

Consequently, research on snow cover and dzud remain largely fragmented. To bridge this gap, this study utilizes MOD10A1 data to derive the spatiotemporal distribution and trends of snow cover, identifying dzud years through disaster loss data. Subsequently, it provides a comprehensive discussion on dzud events and snow cover. The study aims to reveal the nexus between snow cover and dzud, thereby providing a scientific basis for early warning systems and the safeguarding of herder livelihoods in the Steppe regions.

2. Materials and Methods

2.1. Study Area

Located in inland East Asia, from 41°35′ to 52°09′ N latitude and 87°44′ to 119°56′ E longitude [39], Mongolia has a national territory of approximately 1.564 million km² (Figure 1). It borders Russia to the north and China to the east, south, and west. The average elevation is 1580 m, and the elevation decreases from west to east. Mongolia can be divided into five parts: the Altai Mountains, Great Lakes Basin, Khangai–Khentii Mountains, Eastern Mongolian Plateau, and Southern Gobi region. Mongolia has a typical continental arid and semi-arid climate [40]. The annual average temperature, which reduces from south to north, ranges from −8 °C to 6 °C. Precipitation is unevenly distributed, with an annual average of about 100–300 mm. Vegetation types follow a north-to-south gradient: forest, grassland, and Gobi Desert. In order to guarantee the precision of MOD10A1 snow cover products in the Mongolian region, a total of 5000 sample points are established, utilizing 10 m resolution Sentinel-2 imagery for validation (Figure 1).

2.2. Data Sources and Identification of Dzud Years

Snow cover data for Mongolia were derived from the MODIS snow cover product (MOD10A1) provided by NASA. The temporal resolution was 1 day, and the spatial resolution was 500 m. Climate data were sourced from the ERA5-Land dataset, produced by the European Centre for Medium-Range Weather Forecasts (ECMWF) within the framework of the Copernicus Climate Change Service. For accuracy validation, Sentinel-2 MSI Level-2A (L2A) data were utilized (GEE Collection: COPERNICUS/S2_SR_HARMONIZED).

The dzud, a specific type of climate disaster in Mongolia, caused serious environmental and economic damage, especially the death of large numbers of livestock [27]. To identify the most severe dzud in Mongolia through the past 25 years, we utilized multiple authoritative sources, including official reports from the National Agency for Meteorology and Environmental Monitoring (http://namem.gov.mn/en, accessed on 5 April 2025), the National Emergency Management Agency of Mongolia (https://mongolia.gov.mn/, accessed on 8 April 2025), and the Government of Mongolia’s official portal (https://mongolia.gov.mn/, accessed on 10 April 2025). These were complemented with livestock statistics obtained from the Mongolian Statistical Information Service (https://www.1212.mn/en, accessed on 15 April 2025), which documented annual totals of livestock and mortality across provinces (Figure 2). The institution operated under the national statistical system of Mongolia, which followed internationally recognized statistical standards and methodologies in the process of data collection and publication, including standardized data quality control procedures to ensure consistency, accuracy, and comparability. Significant spikes in live-stock mortality, particularly during and following harsh winters, served as key indicators for identifying dzud years. Based on a synthesis of government-reported meteorological hazards, disaster declarations, and abnormal livestock losses, the following winter seasons were classified as major dzud years: 2000–2002, 2009–2010, 2017–2018, 2019–2020, and 2023–2024. Low temperatures and heavy snowfall constituted the physical basis for the formation of dzud in Mongolia. For instance, the Mongolian National News Agency reported that in Uvs Province during 2009–2010, extreme cold weather, with nighttime temperatures dropping to −48 °C or −54 °C, persisted for nearly 50 days. 80% of the country’s territory was covered by a snow blanket of 200–600 mm (https://montsame.mn/en/, accessed on 5 May 2025). However, in the context of Mongolia’s actual situation, a dzud was defined specifically as a disaster characterized by massive live-stock mortality during winter. Therefore, the identification of dzud years in this study was primarily determined from the perspective of disaster loss.

2.3. Methods

(1)

Data Preprocessing

The primary dataset, MOD10A1 V6.1, was derived from MOD09GA Surface Reflectance, which had been atmospherically corrected using the 6S radiative transfer code to mitigate the effects of atmospheric gases and aerosols. To ensure physical consistency, the raw Digital Numbers ($DN$s) in the MODIS dataset were converted into physical parameter values using the specific radiometric calibration parameters provided in the metadata. The general conversion followed the linear equation:

$$Parameter Value=DN\times Scale Factor+Offset$$

(1)

For the $NDSISnow\text{\_}Cover$ band used in this study, the data represented the specific snow index (scaled 0–100) and did not require additional user-side calibration.

The L2A product provided Bottom-of-Atmosphere (BOA) surface reflectance, atmospherically corrected via ESA’s Sen2Cor algorithm. To convert the stored unsigned integer $DN$ values into physical surface reflectance the quantification value and the radiometric offset was applied as follows:

$${\rho }_{\lambda }={\displaystyle \frac{DN + OFFSET}{QUANTIFICATION\text{\_}VALUE}}$$

(2)

where $QUANTIFICATION\text{\_}VALUE$ was 10,000. The harmonized collection in GEE automatically handled the radiometric offsets, ensuring consistent reflectance values (0–1) for high-resolution $NDSI$ calculation. To ensure the reliability of MODIS10A1 in the study area, a rigorous accuracy assessment was conducted on the GEE platform using the processed Sentinel-2 L2A imagery from 2017 to 2024. A total of 5000 random sample points were generated within the study area. After excluding invalid observations (e.g., outliers or cloud-contaminated pixels), a confusion matrix analysis was performed to cross-validate the MOD10A1 product.

(2)

NDSI-Based Snow Detection

$$NDSI={\displaystyle \frac{R_{VIS}-R_{IR}}{R_{VIS}+R_{IR}}}$$

(3)

$R_{VIS}$ and $R_{IR}$ represent the reflectance in the visible and near-infrared spectral bands, respectively [41].

(3)

Snow Cover Fraction

Snow Cover Fraction ($SCF$) [42] is used to represent the relative snow-covered area, defined as the percentage of snow-covered pixels among all pixels within the study area:

$$SCF={\displaystyle \frac{F_n}{F}}\times 100\%$$

(4)

where $F_n$ is the number of snow-covered pixels in the study area, and $F$ is the total number of pixels in the study area.

(4)

Slope Trend

This study applied the trend analysis method to examine the variation trend of long-term snow cover in Mongolia from 2000 to 2024 [43]:

$$slope={\displaystyle \frac{n\times {\sum }_{i=1}^n\left(i\times W_i\right)-{\sum }_{i=1}^{n}i\times {\sum }_{i=1}^n{W}_i}{n{\sum }_{i=1}^n{i}^2-{\left({\sum }_{i=1}^{n}i\right)}^2}}$$

(5)

where $i$ is the year index, $n$ is the length of the time series, and $W_i$ denotes the annual snow cover value for the study area in the $i$ year. When $slope > 0$, it indicates an increasing trend of long-term snow cover over time; conversely, when $slope < 0$, it indicates a decreasing trend. A greater absolute value of slope indicates a faster rate of change.

(5)

Mann–Kendall Trend Analysis

This study used the $Mann-Kendall$ test [44,45] to assess the significance of snow trends from 2000 to 2024. For the time series $X_T=x_1,x_2,\cdots x_L$, the test statistics $S$ and $Z$ were computed, as follows:

$$S=\sum _{k=1}^{n-1}\sum _{j=k+1}^{n}sgn\left(X_j-X_k\right)$$

(6)

$$sgn\left(x_j-x_k\right)=\left\{\begin{array}{cc}1\hfill & \hfill \left(x_j-x_k\right)>0\\ 0\hfill & \hfill \left(x_j-x_k\right)=0\\ -1\hfill & \hfill \left(x_j-x_k\right)<0\end{array}\right\}$$

(7)

$$Z=\left\{\begin{array}{cc}{\displaystyle \frac{S-1}{\sqrt{V_{ar}\left(S\right)}}}\hfill & \hfill S>0\\ 0\hfill & \hfill S=0\\ {\displaystyle \frac{S+1}{\sqrt{V_{ar}\left(S\right)}}}\hfill & \hfill S>0\end{array}\right\}$$

(8)

$$V_{ar}\left(S\right)={\displaystyle \frac{L\left(L-1\right)\left(2L+5\right)}{18}}$$

(9)

where $L$ is the length of the time series and $x_j$ and $x_k$ represent the values at time indices $j$ and $k$, respectively. At a given confidence level $\alpha$, $Z>0$ indicates an increasing trend and $Z<0$ indicates a decreasing trend. When the absolute value of $Z$ is greater than or equal to 1.645, 1.96, or 2.576, the result passes the significance test at the 90%, 95%, and 99% confidence levels, respectively [46,47].

(6)

Pearson Correlation Analysis

Pearson correlation analysis is used at the pixel scale to assess the relationships between snow cover indicators and temperature and precipitation in Mongolia [48]:

$$R_{xy}={\displaystyle \frac{{\sum }_{i=1}^n\left[\left(x_i-\overline{x}\right)\left(y_i-\overline{y}\right)\right]}{\sqrt{{\sum }_i^n{\left(x_i-\overline{x}\right)}^2}·\sqrt{{\sum }_i^n{\left(y_i-\overline{y}\right)}^2}}}$$

(10)

where $n$ is the number of years during the study period; $x_i,y_i$ are the two variables for correlation analysis; and $\overline{x}$ and $\overline{y}$ are the mean values of the main factor and the influencing factor, respectively. $R_{xy}$ denotes the Pearson correlation coefficient between variables $x$ and $y$, which ranges from −1 to 1. When $R_{xy}$ is positive, it indicates a positive correlation between the two variables; when $R_{xy}$ is negative, it indicates a negative correlation [49].

3. Results

3.1. Spatiotemporal Distribution Characteristics of Snow Cover

During 2000–2024, Mongolia’s multi-year average SCF showed significant regional differences (Figure 3). Snow was mainly concentrated between 45° N and 52° N latitude but extended southward during severe snow events, forming a fan-shaped distribution pattern. In the high-altitude areas, accounting for 14.54% of the total area, SCF was relatively high, reaching more than 80%. Low-altitude areas, which covered 46.48% of the total area and were located in the east and in other regions with similar latitudes, such as Khentii, Dornod, Sukhbaatar, and Tov provinces, had a multi-year average SCF of approximately 20%. SCF was the lowest in the southern part of Mongolia, including Umnugovi, Dundgovi, and Dornogovi provinces, covering 38.98% of the total area, where it ranged from 0 to 10%.

The SCF distribution of the five severe dzud events during the past 25 years demonstrated substantial differences compared to the multi-year average SCF spatial pattern (Figure 4). The monthly SCF data during dzud years intuitively reflected the processes of snow accumulation and melting. Darker colors indicated higher SCF values, representing more concentrated snowfall. Among the five severe dzud years (2000–2002, 2009–2010, 2017–2018, 2019–2020, and 2023–2024), the SCF distribution revealed spatial variability. However, eastern Mongolia (Khentii, Dornod, and Sukhbaatar provinces) consistently exhibited high SCF values in all five dzud events, often becoming heavily affected areas, which resulted in large-scale livestock loss. However, on the multi-year average SCF map, eastern Mongolia indicated relatively low SCF values.

3.2. Interannual and Intra-Annual Variability of Snow Cover

To analyze the interannual variation characteristics of SCF in Mongolia through the past 25 years quantitatively, MODIS snow cover data for snow seasons from 2000 to 2024 were applied, generating a time series trend of SCF (Figure 5). The results illustrated that the SCF was 34.63% in 2000–2001 and 37.69% in 2023–2024. Overall, the SCF in Mongolia exhibited a slight but non-significant decreasing trend (R = −0.053, p > 0.05), with large fluctuations and strong instability, and its multi-year average value was approximately 25.36%. Notably, the five dzud events (2000–2002, 2009–2010, 2017–2018, 2019–2020, and 2023–2024) showed clear deviations from the long-term trend, with SCF values consistently higher than the multi-year average. In addition, these dzud years typically experienced sudden and sharp increases in snow cover.

Based on the monthly average SCF data through the past 25 years, the intra-annual variation characteristics of SCF in Mongolia were analyzed (Figure 6). The results showed that there was one snow accumulation and melting process each year in Mongolia, with clear periodic variation in SCF. In general, snow accumulation occurred from October to January, and snow melting occurred from January to April; the peak was reached in January. This pattern was consistent with Mongolia’s natural climatic conditions, as snow cover mainly occurred in October, November, December, January, February, March, and April, among which October and April were the months of snow appearance and disappearance, respectively. The monthly average SCF from October to April over the past 25 years was 5.16%, 19.85%, 35.73%, 40.08%, 36.01%, 19.41%, and 5.08%, respectively. To highlight the abrupt variations characteristic of extreme dzud events, the monthly mean SCF for the five most severe occurrences (2000–2002, 2009–2010, 2017–2018, 2019–2020, and 2023–2024) was depicted using distinct colored lines (Figure 6). The results indicated that the monthly SCF during these dzud events was consistently higher than the long-term average (represented by the red line). Notably, the monthly SCF in 2023–2024 significantly exceeded that of the other years.

3.3. Snow Cover Trend Analysis

From 2000 to 2024, the multi-year variation in snow cover across Mongolia ranged from −2.42% to 3.63% per year (Figure 7). Spatially, snow cover in Mongolia showed a general trend of intensification in the west and mitigation in the east. Most areas remain relatively stable, while only a few regions experience statistically significant shifts. Based on the trend results, the Mann–Kendall (M–K) trend test is conducted, revealing a generally non-significant downward tendency, consistent with the interannual variation trend. Areas with decreasing snow cover account for 66.78% of the total, among which 90.62% revealed nonsignificant reduces. In contrast, increasing snow cover was observed in 32.22% of the area. Significant reduction mainly concentrates in the eastern regions, while increase mainly appears in the west. The southern Gobi region, by comparison, is an area with no obvious change.

4. Discussion

4.1. Impacts of Snow Cover on Dzud

With the intensification of global climate change [50,51], extreme weather events became frequent and unpredictable [52]. Dzud was a unique winter natural disaster in Mongolia caused by extremely cold temperatures and continuous snowfall. It results in pasture being covered or frozen by ice and snow, making it impossible for livestock to graze, which leads to mass livestock death and significant losses to herders’ lives and property. This study found that, since 2000, Mongolia had experienced five severe dzud. In these events, snow cover often increased explosively, posing major threats to herders and livestock (Figure 8). An SCF higher than the average value (25.36%), such as in 2011–2013, did not necessarily indicate the occurrence of dzud events. In contrast, when snow cover remained below average in one year but was followed by sudden extreme snowfall in the following year—resulting in an SCF far exceeding the average—dzud events were more likely to occur. In addition, snow in Mongolia could persist for as much as seven months each year. Snow arrives quickly and melts slowly, severely testing livestock resilience to cold and snow.

About one-third of Mongolia’s population practiced a nomadic lifestyle. From 1991 to 2024, the total number of livestock in Mongolia manifested a clear increasing trend (Figure 2), indicating stable development of the pastoral economy. However, dzud events brought severe economic shocks to herders. This was especially the case during the 2009–2010 and 2023–2024 dzud events, when nearly 10 million livestock died. Such large-scale losses not only resulted in direct income reduction for herders but also triggered a series of consequences such as increased household debt and shortages of daily necessities, which further exacerbated regional poverty (Figure 8). Some researchers pointed out that disagreements about whether to declare a state of emergency and how to coordinate assistance to herders sometimes existed among government departments. The management of dzud lacked a sound accountability mechanism and failed to recognize it as a systemic issue requiring integrated governance, which exacerbated disaster losses [12,53,54]. In addition, the lack of accurate understanding of how extreme cold and heavy snowfall triggered and worsened dzud prevents relevant departments from making scientific and reasonable interventions. This might have led to resource waste and delayed rescues in high-risk areas, further intensifying secondary disasters.

Therefore, in response to increasingly frequent and severe dzud events, Mongolia urgently needs to enhance high-precision weather forecasting and disaster early warning capabilities. Improving the detection capacity for extreme weather and ensuring the timeliness and accuracy of information for herders are crucial. It is necessary to establish a cross-sectoral emergency coordination mechanism, requiring unified command and planning among departments such as agriculture, meteorology, transportation, health, and finance, as well as pre-arranged emergency plans and logistics preparations. Promoting a “disaster risk reduction” pastoralism model is also essential. Under the influence of climate change, traditional nomadic practices are becoming increasingly vulnerable. Thus, it is necessary to growth forage reserves and livestock shelter construction, reasonably control livestock population, focus on quality breeding, and reduce mortality rates. This will effectively reduce dzud-related losses, enhance disaster resilience, and provide scientific support for herder livelihoods and sustainable development of grassland ecosystems. Meanwhile, based on the above understanding, future research on snow cover and dzud events tends to move toward multi-dimensional and multi-factor integrated risk assessment and early warning models. It is necessary to strengthen the integration of natural factors with variables such as the vulnerability, exposure, and response capacity of socio-economic systems, to build a comprehensive dzud risk evaluation system, thereby effectively enhancing disaster risk reduction in winter in arid grassland areas.

4.2. Snow Cover Zoning and Loss of Livestock During Dzud

Snow cover was one of the most sensitive indicators for monitoring global climate change [55,56]. Scholars conducted long-term snow monitoring in the Northern Hemisphere, the Eurasian continent, and the Mongolian Plateau, and all results indicated a decreasing trend in regional snow cover extent [35,57,58,59]. However, these findings were relatively coarse for smaller regions such as Mongolia. Mongolia was one of the most severely affected hotspots of dzud in inland Asia, making accurate snow retrieval critically important. This study extracted snow cover based on the Normalized Difference Snow Index (NDSI) and corrected misclassification in the MOD10A1 product using auxiliary data such as elevation and slope. With the support of the GEE cloud platform, Sentinel-2 satellite data with 10 m spatial resolution were used to verify the accuracy of MOD10A1 from 2017 to 2024 through confusion matrix analysis. A total of 5000 random sample points was set in the study area, and after removing outliers, the overall accuracy reached 0.91 and the kappa coefficient was 0.83. While this study validated MOD10A1 data using Sentinel-2 imagery from 2017 to 2024, the restriction to the post-2017 period was primarily due to the unavailability of high-resolution images. Therefore, to maintain consistency in validation standards, we propose the future research for long-term validation in the research area, by developing robust harmonization methods for multi-source data, e.g., combining Landsat and SPOT images.

According to the multi-year spatiotemporal distribution of snow cover (Figure 2) and the digital elevation model (DEM) (Figure 9), snow cover in Mongolia was divided into three main parts: high-snow-cover high-altitude areas, medium-snow-cover low-altitude areas, and low-snow-cover South Gobi areas. SCF in Mongolia increased with elevation. However, in the 0–1000 m and 1000–2000 m ranges, SCF showed a negative correlation with elevation. The average elevation of the South Gobi region was around 1000–1500 m, and the eastern temperate grasslands had an elevation of about 900–1200 m. Snow cover in the eastern grasslands (Khentii, Dornod, and Sukhbaatar provinces) was much higher than snow cover in the South Gobi, which led to the negative correlation. In addition, different slope aspects received different solar radiation, causing SCF differences. An analysis of SCF across slope aspects in Mongolia (Figure 10) manifested consistency with the snow cover zoning results.

The spatial distribution and cover rate of snow were further discussed in relation to the distribution of livestock losses during dzud events. According to the above livestock mortality map (Figure 2), the dzud events in 2009–2010 and 2023–2024 were the most severe and representative events in Mongolia over the past 25 years. By overlaying the SCF data with the statistical data on livestock losses, it clearly identified the high-risk zones of these two Dzud events (Figure 11). The study found a certain degree of spatial consistency between SCF distribution and livestock losses. During the 2009–2010 dzud, the most severely affected areas were located in central and western Mongolia, specifically in Zavkhan, Govi-Altai, Arkhangai, Uvurkhangai, and Dundgovi provinces. Each of these provinces reported livestock losses exceeding 700,000 heads. During the 2023–2024 dzud, the most severely affected areas were located in eastern Mongolia, mainly in Arkhangai, Tuv, Khentii, Dornod, and Sukhbaatar provinces. Although the affected regions of the two dzud events differed, they both occurred in areas with high SCF. From a historical perspective, high-risk areas of dzud events tended to appear in mid-to-high latitude regions with high SCF in Mongolia.

In addition to causing direct livestock mortality, the dzud in Mongolia triggered a cascade of vicious cycles. In 2023, heavy snowstorms trapped at least 13,500 herder households, restricted mobility in affected pastoral areas, and blocked 38,400 km of roads (https://www.imsilkroad.com/news, accessed on 10 September 2025). United Nations Children’s Fund reported that road blockages caused by heavy snow affected over 258,000 individuals, including more than 100,000 children. These affected children were unable to access essential health, nutrition, education, and social services (https://www.savethechildren.net/news, accessed on 2 October 2025). Consequently, effective early warning of dzud events held significant importance for disaster prevention, mitigation, and sustainable ecological development in pastoral regions [60]. Through spatiotemporal snow cover retrieval and case studies of livestock losses during two major dzud events in Mongolia, this study provided a critical foundation for early warning systems in steppe regions. At the regional scale, the study highlighted the most severely affected areas in Mongolia, particularly the mid-to-high latitude eastern and western regions such as Dornod, Sukhbaatar, Khentii, Arkhangai, Zavkhan, and Uvs. Temporally, the research identified November as a critical timeline for impending dzud events, where the SCF exceeded 30%, significantly surpassing the multi-year average. Therefore, daily snow cover monitoring during November was deemed particularly crucial for early warning. This allowed sufficient time to alert herders of potential disasters, enabling preemptive livestock relocation or slaughter, thereby reducing direct economic losses. Furthermore, it was considered very important to ensure that indigenous inhabitants of the steppe having a deep understanding of extreme dzud events and engaging in early warning initiatives. Equipped with adequate knowledge regarding meteorological changes and the severity of dzud, they were able to formulate timely contingency plans (e.g., effective preventive measures such as forage stockpiling and livestock relocation). Mobilizing the entire population to mitigate dzud-induced losses not only promoted SDG 1 (No Poverty) and SDG 2 (Zero Hunger) but also aligned with SDG 13 (Climate Action).

The Eurasian Steppe (e.g., Kazakhstan, Inner Mongolia of China, and Russia Far East) shared similar extreme climate threats and nomadic/semi-nomadic characteristics with Mongolia, and the formation mechanisms of dzud were similar. The “sudden snow accumulation” identified in this study served as a common trigger mechanism for white disaster outbreaks. Therefore, using spatiotemporal data to monitor sudden snow changes functioned as the core algorithm for wide-area steppe white disaster monitoring systems or as the foundation for early warning. Nevertheless, historical data had limitations in predicting future dzud events, especially under the intensifying impacts of climate change and the increasing frequency of extreme weather. Traditional “experiential” or static models were not sufficient to cope with these increasingly unpredictable risks. Therefore, it is needed to construct an early warning system for rapidly identifying potential dzud risks based on real-time, multi-source and high-resolution data, coupled with advanced algorithms, allowed the rapid response of abnormal weather patterns.

4.3. Impacts of Climate Change on Snow Cover

Many studies attempted to identify the main drivers of snow cover change [61]. Increasing temperature was widely recognized as the most direct and widespread negative factor, leading to shortened snow duration and reduced snow cover and depth [62]. The role of precipitation was more complex, with both positive and negative feedback varying with regional seasonal characteristics and temperature differences [63]. Environmental controls such as vertical terrain gradient, slope, and latitude also regulated snow distribution [64]. In the Tibetan Plateau and Central Asia, snow cover was significant in windward mountainous areas, indicating a combined control effect of precipitation and topography [65]. This study focused on the driving effects on snow cover of two climatic factors, temperature and precipitation, which showed trends generally consistent with findings on the Mongolian Plateau climate [66]. We found that the average temperature in Mongolia through the past 50 years was approximately 1.41 °C and that there was a significant increasing trend (R = 0.743, p < 0.001) at a rate of 0.06 °C per year. The cumulative precipitation was about 160.61 mm, showing a decreasing trend (R = −0.436, p < 0.001) at a rate of 0.58 mm per year (Figure 12). From 2000 to 2024, SCF was negatively correlated with temperature (R = −0.359, p > 0.05) and positively correlated with cumulative precipitation (R = 0.407, p < 0.05) (Figure 13).

Over the recent decades, the climate system in Mongolia underwent significant changes. Under the dual pressure of rising temperature and intensified drought, grassland productivity has declined and livestock resistance to disasters has weakened, creating an environmental background for the frequent occurrence of dzud events. Whether the compound interaction of early-stage drought and subsequent dzud formed a “drought-snow” disaster chain was also a subject of intense debate. Caleb investigated the explanatory power of summer drought and low temperatures on dzud using data including winter temperatures, SCD, and NDVI in Mongolia from 2003 to 2016. The study found that the contribution rate of summer drought to dzud reached up to 43%, while low temperatures accounted for 20–37% [21]. In contrast, Troy focused on dzud events in Mongolia between 1970 and 2015, examining the frequency of droughts preceding them. Among the 32 analyzed dzud events, only three were preceded by drought, which make it to be questioned for the validity of this disaster chain. Climate change in Mongolia weakened the self-regulation capacity of ecosystems, making vegetation and livestock more sensitive to anomalies in precipitation and temperature and further amplifying the spatial extent and severity of dzud impacts. The sudden and frequent occurrence of dzud events in Mongolia was likely related to changes in hydrothermal conditions under regional climate warming.

5. Conclusions

Arid grassland regions of Mongolia faced the challenge of regional sustainable development under frequent occurrences of winter dzud events. This research employed 10 m resolution Sentinel-2 data as a validation tool, using consistency verification to enhance the reliability of snow cover monitoring based on MODIS data in the inland grassland areas. Based on MODIS Normalized Snow Index products (MOD10A1), the ERA5-Land reanalysis dataset, and a digital elevation model (DEM), the study integrated trend test, slope\aspect extraction, and correlation analysis methods to assess the spatiotemporal variation characteristics of snow cover under dzud events in Mongolia from 2000 to 2024. The study produced several interesting results: (1) Snow cover in Mongolia showed a non-significant decreasing trend through the study period and increased with elevation. Snow could persist for up to seven months, peaking in January. Interannual variation is large, indicating strong instability. (2) Snow cover was mainly distributed between 45° N and 52° N and illustrated a fan-shaped pattern. It can generally be considered using three major zones: high-elevation areas, low-elevation areas, and the Gobi region. Slope-MK trend analysis shows intensified change in the west and mitigation in the east, with most areas remaining relatively stable and fewer showing significant changes. Although the eastern region (Khentii, Dornod, and Sukhbaatar provinces) indicated a slowing trend, extreme snowfall frequently occurred under dzud, which had high SCF values. (3) During dzud events, the SCF in Mongolia was generally higher than the multi-year average, often showing a sudden sharp rise. The fluctuation in snow cover under dzud events showed consistency with changes in livestock mortality. This further proved that extreme snowfall was one of the most critical disaster-causing factors of dzud. From the perspective of climate change over the past 50 years, temperature revealed a steady improvement, while precipitation exhibited a declining trend to some extent, which might have been one of the factors contributing to the instability of dzud events in Mongolia. Given its harsh climatic conditions, Mongolia urgently needed to strengthen multi-sector cooperation and regional emergency capacity. By relying on ecological resilience restoration and optimizing grassland pastoral systems, an adaptive dzud response system could be built. These findings provided scientific evidence on the spatiotemporal characteristics of extreme snowfall and dzud disaster risk reduction in arid grassland regions.