Snow drought differs from the familiar warm-season variety in both mechanism and consequence. It occurs when the snowpack that normally accumulates between October and March falls far below normal, whether because precipitation arrives as rain instead of snow, because mild temperatures melt what has fallen, or because dry air strips moisture from the snow surface through sublimation. For communities that depend on melting snow to recharge soil moisture, feed streams, and sustain grasslands at the start of the growing season, a snow drought can be as damaging as any summer drought. On the Mongolian Plateau, where snow is an essential freshwater source underpinning pastoral and agricultural livelihoods, such deficits have long been recognized through the local concept of black disasters, historical winter drought events recorded in Chinese meteorological disaster archives. Yet systematic, quantitative assessments of winter drought in this region have lagged far behind studies of warm-season aridity.

The research team, led by Guna Ari and colleagues, delineated the study domain as Mongolia and China’s Inner Mongolia Autonomous Region, an area of roughly 2.75 million square kilometers spanning 87°43′ to 126°04′E and 37°22′ to 53°23′N. The plateau rises to a mean elevation of about 1,580 meters, with mountains dominating the northwest, grasslands and rolling hills across the center and east, and the Gobi Desert covering much of the south. Annual precipitation is low and highly uneven, exceeding 500 millimeters in parts of the east but dropping below 150 millimeters in western Inner Mongolia, with more than two-thirds falling in summer. Mean annual temperatures range from −10.9 to 11.4 °C. The region spans five aridity classes, from hyper-arid to humid, providing a natural laboratory for testing how snow drought controls shift across contrasting climates.

To measure snow drought, the researchers built a snow water equivalent index, or SWEI, from ERA5-Land reanalysis data covering 1951 to 2024. Snow water equivalent describes the amount of water contained in the snowpack and is the key variable for monitoring snow drought. The team standardized SWE using a nonparametric normalization method, computed the index at a three-month scale, and classified drought grades according to the D-level system of the United States Drought Monitor. For each grid cell and year, the representative SWEI value was taken in the month when snow water equivalent reached its annual peak. Validation against independent records was rigorous: ERA5-Land snow estimates correlated strongly with observed snow depth, yielding an R² of 0.6496 and a root mean square error of 26.31 millimeters. More tellingly, when the researchers checked the index against 77 documented historical black disaster events compiled from Chinese meteorological disaster archives, 66 matched the SWEI-based monitoring results, an agreement rate of 85.71 percent.

The temporal picture that emerged is one of a regime shift around the millennium. Across the full 74-year record, SWEI showed an overall linear decline, indicating intensifying snow drought, although the trend did not reach statistical significance at the p > 0.05 level. Decadal means were relatively high in the 1970s and relatively low in the 2000s. Splitting the record at 2000 revealed a slight, non-significant increase in SWEI during 1951–1999 followed by a decline during 2000–2024. Both frequency, counted whenever SWEI fell to −0.5 or below, and severity, defined as the absolute value of SWEI during such events, showed higher means and greater dispersion after 2000, signaling more frequent, more severe, and more spatially heterogeneous snow droughts. In the 2000s and 2010s the value ranges expanded markedly, with the 2010s showing a particularly strong extension of high-severity extremes.

Spatially, the patterns were equally revealing. Negative SWEI trends dominated 65.2 percent of the plateau, with the strongest declines concentrated in the central and southern arid to semi-arid zones, while positive trends appeared mainly in the northern and northeastern sub-humid to humid zones. High snow drought frequency clustered along the northeastern margin, whereas high-severity conditions expanded across the northeastern plateau and the western mountains. During 2000–2024, the Sen’s slope range widened dramatically to between −0.0959 and 0.0888, compared with a narrow band of −0.0363 to 0.0332 in the earlier period, and the proportion of grid cells with positive slopes rose to 47.0 percent, reflecting an expanding northern zone of snow surplus even as southern deficits deepened.

Attribution analysis using the Geographical Detector model, which quantifies how much of the spatial variance in SWEI each driver explains, identified meteorological conditions and extreme climate indices as the dominant controls, with topographic factors consistently weak. Before 2000, relative humidity held the highest explanatory power, followed by the cold-night index TN10p, cool days, surface solar radiation, and the cold spell duration index. After 2000, relative humidity and radiation remained dominant, but the contributions of temperature and vapor pressure deficit, a measure of atmospheric drying power, strengthened, while the ice days index and cool days stayed influential. Correlation analysis reinforced this hierarchy: SWEI was positively correlated with relative humidity at r = 0.57 and negatively with vapor pressure deficit at r = −0.57, while showing significant negative correlations with solar radiation (r = −0.50) and temperature (r = −0.43) and positive correlations with cold extremes (r = 0.39 for cold nights, r = 0.43 for cool days). Enhanced energy input intensifies snow drought; more frequent cold extremes favor snow storage.

Variance partitioning apportioned the explained variance among three driver groups and exposed how the control regime evolved. The unexplained fraction stood at 0.75 in both periods, but before 2000 extreme climate factors contributed the largest independent share at 0.15, followed by meteorological factors at 0.08 and topography at just 0.01. After 2000, meteorological and extreme climate factors contributed equally at 0.10 each, with their shared fraction rising to 0.04. Interaction detection showed that nonlinear enhanced interactions dominated in both periods. Before 2000 the strongest pairing was soil moisture with solar radiation; after 2000, radiation-related couplings persisted while interactions tied to atmospheric drying, such as radiation with consecutive dry days and soil moisture with vapor pressure deficit, strengthened further. The authors interpret soil moisture and radiation coupling as a feedback in which dry land weakens evaporative cooling, boosts sensible heating, and raises snowpack energy, jointly accelerating melt and shortening the snow season. Climate zone contrasts were pronounced: hyper-arid and arid regions were mainly governed by cold extremes, radiation, and humidity, while sub-humid and humid zones in the earlier period were more strongly associated with precipitation, radiation, and consecutive dry days.

Cross-wavelet transform and wavelet coherence analyses linked snow drought variability to large-scale atmospheric circulation. Of the five indices examined, the Arctic Oscillation, North Atlantic Oscillation, Pacific Decadal Oscillation, Multivariate ENSO Index, and Atlantic Multidecadal Oscillation, ENSO as represented by the Multivariate ENSO Index showed the strongest and most widespread significant coherence with SWEI, spanning periods from 4 to 64 months, with positive coherence concentrated during 2005–2016 at 32–64 month bands. Notably, this ENSO linkage emerged mainly after 2000, mirroring the concurrent intensification of snow drought. The Arctic and North Atlantic Oscillations were predominantly but intermittently negatively associated with snow drought, consistent with their roles in regulating midlatitude westerlies and moisture transport, while the Atlantic Multidecadal Oscillation showed a positive association and the Pacific Decadal Oscillation exhibited phase-dependent sign changes.

The authors are candid about limitations. ERA5-Land can exhibit systematic biases over complex terrain, where each 0.1-degree grid cell averages away microtopographic contrasts between windward and leeward slopes, and wavelet coherence quantifies association rather than causation, remaining vulnerable to autocorrelation and shared external forcing. Future work will fuse reanalysis with multisource remote sensing and ground observations, apply partial wavelet coherence and surrogate-based significance testing, and use large-ensemble model experiments to separate forced from internal variability. The practical stakes are considerable: the findings argue for strengthened early warning and preparedness in severely affected areas, enhanced monitoring of cold-season extremes, and close attention to ENSO variability in seasonal forecasting. For the herders and farmers of the Mongolian Plateau, whose livelihoods hinge on the winter snowpack, the post-2000 shift documented here is not an abstraction but a changing baseline for water security, ecological conservation, and adaptive management across one of the world’s most climate-sensitive interior regions.

Subject of Research: Spatiotemporal variability and drivers of snow drought on the Mongolian Plateau from 1951 to 2024

Article Title: Spatiotemporal Characteristics and Driving Factors of Snow Drought on the Mongolian Plateau

Article References: Ari, G., Asi, H., Gao, Y., & Zhang, Q. (2026). Spatiotemporal Characteristics and Driving Factors of Snow Drought on the Mongolian Plateau. International Journal of Disaster Risk Science. https://doi.org/10.1007/s13753-026-00781-3

Image Credits: AI Generated

DOI: 10.1007/s13753-026-00781-3

Keywords: snow drought, Mongolian Plateau, snow water equivalent, ERA5-Land, climate change, ENSO, vapor pressure deficit, relative humidity, solar radiation, cold extremes, wavelet coherence, Geographical Detector