Introduction

Snow, as a critical component of surface coverage1, profoundly influences global energy balance, climate change, hydrology, and ecology processes through its snow onset day, snow end day, and snow cover days (SCDs)2,3,4. Approximately 50% of the land surface in the Northern Hemisphere is covered by snow during winter5, and in some seasonal snow cover areas, the annual SCDs and continuous snow cover days (CSCDs) are substantial6,7. As the Third Pole of the world, the Qinghai-Tibetan Plateau (TP) is one of the areas with the widest and longest snow cover distribution in the Northern Hemisphere. For instance, in areas such as the southwestern, central, and eastern parts of the TP, the SCDs generally exceed 60 days, with some areas even surpassing 90 days. The spatiotemporal distribution and variability of the snow cover on the Qinghai-Tibet Plateau (TPSC) hold significant implications for regional and global climate change, regional surface water balance, ecological environment evolution, and geological disaster prevention5,8,9,10.

Some scholars classified the snow cover over the Eurasian continent based on the SCDs and CSCDs6,11. The results showed that, except for the TP, most other regions exhibited consistent spatial distributions of the SCDs and CSCDs, leading to relatively consistent classification results. However, due to the complex terrain and climate on the TP, the classification results displayed significant differences. Typically, in most areas, the SCDs exceed 60 days, while the CSCDs generally remain below 20 days6, indicating a pronounced temporal discontinuity of the TPSC. Furthermore, due to the rising temperature, the snow onset day in most snow-covered regions in the Northern Hemisphere is delayed12, while the snow end day is advanced13, resulting in shortened snow season length and SCDs14,15,16,17,18. However, owing to the strengthening of the mid-latitude westerlies, the TP exhibits the opposite trend, with earlier snow onset day, delayed snow end day, and prolonged cold seasons19,20,21. Nonetheless, there is a decreasing trend in SCDs, and CSCDs are also decreasing. Li et al. classified the snow cover areas in China based on CSCDs and found that the areas of persistent snow cover (CSCDs >30 days) and periodic variable snow cover (10 days < CSCDs < 30 days) on the TP are gradually decreasing, while the non-periodic variable snow cover areas (CSCDs < 10 days) are significantly increasing22. The TPSC is gradually becoming discontinuous.

Highland snow cover serves as a vital source of surface runoff in mountainous regions23,24,25. The discontinuity of snow cover can disrupt the periodicity of runoff26,27,28 and soil moisture29,30, potentially leading to extreme hydrological events such as floods/droughts31, thereby posing severe threats to plant growth32,33, animal activities34, and biodiversity patterns35. Furthermore, the shortening duration of continuous snow cover affects mountain tourism and recreational activities (such as downhill skiing, mountaineering, etc.), with many ski resorts closing due to unfavorable snow conditions, resulting in significant economic losses36,37. As the Asian Water Tower, the TP is the source of many major rivers in Asia, including the Yangtze, Yellow, Nu, Lancang, and Yarlung Tsangpo Rivers38. Changes in continuous snow cover duration directly or indirectly influence the flow and water resources of rivers around the Himalayas and within the Yangtze River basin. The discontinuity in snow cover duration on the TP also affects surface albedo and sensible heat exchange, thereby influencing the intensity and timing of circulation, which in turn has important implications for local or even remote weather and climate39,40,41. For instance, due to the decreased summer TPSC, the response of Rossby waves to the diabatic forcing of the El Niño-Southern Oscillation becomes stronger, thereby affecting summer precipitation in East Asia39,41,42. Snow cover during autumn and winter on the TP and its surrounding areas has a certain influence on winter temperatures in North America and even extreme cold events43,44,45. Simultaneously, it also has significant impacts on the activities and ecosystems of the alpine pastoral areas on the TP46.

Remote sensing has proven to be an effective approach for monitoring the global and regional scale dynamics of snow cover over the past several decades47,48,49. The main methods for detecting the TPSC distribution are microwave remote sensing and visible light remote sensing50. However, due to the complex meteorological and topographical conditions of the TP, TPSC is discontinuous, snow depth is shallow, and the snow undergoes layering and metamorphism51. As a result, the two remote sensing monitoring methods may underestimate the snow cover extent52,53. Additionally, the extensive cloud cover over the TP poses a challenge to continuous mapping of snow dynamics54,55,56,57,58. Commonly used MODIS snow products utilize an 8-day composite of snow data to remove cloud cover. However, at 92% of the sites on the TP, the CSCDs are less than 8 days, potentially resulting in an overestimation of snow cover coverage with this algorithm. Therefore, current remote sensing methods may not accurately analyze the spatiotemporal distribution characteristics of the TPSC. Despite the sparse and uneven distribution of meteorological stations across the TP, they offer advantages such as accurate recording, high temporal resolution, and long-term continuous observations, making them more suitable for accurately documenting snow cover distribution characteristics in the TP59,60,61.

With global climate warming, seasonal snow cover worldwide is likely to become more discontinuous, particularly in low-altitude and warmer regions, with a clear trend towards ephemeral snow cover62. Therefore, to explore the temporal characteristics in the continuity of the TPSC, this study analyzes the spatiotemporal distribution and variability characteristics of continuous snow cover events in the TP, using daily snow cover data from 83 stations spanning 1961–2019, and analyzes the reasons from the perspective of local climate and circulation backgrounds. The main research questions addressed include: (1) Which is more accurate in describing the continuity of TPSC: SCDs or CSCDs? (2) Attempting to propose new indicators to describe the continuity of the TPSC. (3) What are the factors influencing the continuity of TPSC?

Results

Spatiotemporal distribution and variability characteristics of SCDs

Over multiple years, the average annual SCDs on the TP amount to 33 days (S.D. = 27.89). The SCDs show a pattern of increase followed by a decrease, indicating an overall decreasing trend. Before 1993, they exhibited an upward trend, while after 1993, there has been a decrease. winter has the highest proportion of SCDs throughout the year, accounting for 44.40%. The maximum number of SCDs is 24 days, the minimum is 3 days, and the average is 14 days. Following winter, spring has the next highest average SCDs at 10 days, representing 33.99% of the annual SCDs. autumn has fewer SCDs, with an average of 6 days, accounting for 19.90%. summer has the least, with an average of 1 day, representing 1.71% (Supplementary Fig. 1).

The average SCDs at various sites show a pattern of decreasing from the central area towards the periphery (Fig. 1a). In the central region, such as Bayan Har and Amnye Machen, the average SCDs are relatively high, with the maximum occurring at the Qingshuihe site in the Bayan Har Mountains, at 129 days per year, followed by the Gande site at 96 days. The Ulugh Muztagh site in the Qilian Mountains and the Nilamuk site in the Himalayas Mountains, also have a relatively high number of SCDs, at 95 and 78 days respectively. The southern region of the Hengduan Mountains has lower SCDs, with the lowest occurring at the Derong site, averaging only 1 day per year, followed by the sites in Danba, Yajiang, and Batang, which average 2 days.

Fig. 1: The spatiotemporal characteristics of SCDs in the TP.
figure 1

a Spatial distribution (day), and b rate of change (day per decade). The hollow rhombus indicates the trend is statistically significant (p < 0.05). The specific percentages of the stations with significant or insignificant changes are presented. The inset illustrates the frequency distribution of the corresponding ranges (in the same color) in each panel.

Full size image

53 sites show a decreasing trend in the number of SCDs, with about half of these sites experiencing a significant decrease at an average rate of 1.3 days per decade (p < 0.05). At 30 sites, the SCDs are gradually increasing, but less than half of these sites show a significant increase (p < 0.03), with an average rate of 3.6 days per decade (Fig. 1b). The variation in SCDs is greatest in the Hengduan Mountains region, while other areas show smaller changes. The fastest decrease is observed at the Deqin site in the Hengduan Mountains, at 7.7 days per decade. The fastest increase is observed at the Batang and Derong sites in the Hengduan Mountains region, at 15 and 14 days per decade respectively.

Snow cover continuity characteristics

The annual average snow cover number (SCNs) at each site on the TP is 14 times (S.D. = 9.38). In spring, the average SCNs is the highest, with an average of 5.42 times. This is followed by winter with an average of 4.53 times, autumn with 2.97 times, and summer with the fewest occurrences, averaging 0.4 times. The SCNs in years and different seasons show a significantly decreasing trend, with summer exhibiting the slowest decline over the years and spring experiencing the fastest decrease in SCNs. The highest SCNs across all sites on the TP were recorded in 1982, at 18 times, while the lowest occurred in 2014, at 10 times, with an average of 14 times (Supplementary Fig. 2).

The central and eastern regions of the TP have higher annual average SCNs, indicating less snow cover stability (Fig. 2a). In contrast, the surrounding areas show lower SCNs and higher stability. The region with the highest SCNs is the Qingshui River area in the Bayan Har range, with an average SCNs of 35 times, followed by the Shiqu in Hengduan and Wudaoliang sites in the Bayan Har Mountains, with 33 times. The areas with the lowest SCNs are Batang and Derong sites in the Hengduan Mountains, as well as Gongshan, with only 1 time.

Fig. 2: The spatiotemporal characteristics of SCNs in the TP.
figure 2

a Spatial distribution (day), and b rate of change (day per decade). Same as Fig. 1.

Full size image

The change rate of the annual average SCNs in the central and eastern regions of the plateau is higher than that of the surrounding areas (Fig. 2b). 78 sites show a significant decrease in SCNs (p < 0.05), with an average reduction rate of 0.9 times per decade. The fastest decrease was observed at the Wudaoliang site in the Bayan Har Mountains, with a reduction rate of 2.3 times per decade. Only 5 sites show a significant increase in annual SCNs (p < 0.05), with the fastest increase observed at the Lenghu site in the Qaidam Basin, at a rate of 0.1 times per decade. In autumn, only 4 sites in the Bayan Har and Qilian Mountains show an increasing trend in SCNs, while in other regions, SCNs are decreasing. In winter, spring, and summer, SCNs near the Bayan Har and Himalayan Mountains increase, while in other areas, they decrease. The increase in SCNs is most significant in winter, followed by spring and summer.

The multi-year average CSCDs on the TP is 3 days (S.D. = 2.89), and maximum CSCDs is 40 days (S.D. = 33.88). The highest recorded number of maximum CSCDs was 331 days at the Mado station in 1985. The CSCDs vary across different seasons, with winter having the highest CSCDs, with average and maximum CSCDs averaging 3.4 and 7.8 days. The maximum values are both recorded at the Qingshuihe site in the Bayan Har range, at 17.9 and 39.7 days respectively. Spring follows, with average and maximum CSCDs averaging 1.7 and 4.2 days, with maximum values recorded at the Nielamu site in the Himalayas, at 6.2 and 33.7 days. Autumn has fewer CSCDs, with average and maximum CSCDs averaging 1.8 and 3.2 days, with maximum values recorded at the Qingshuihe site in the Bayan Har range, at 3.5 and 14.7 days. Summer has the fewest CSCDs, with average and maximum CSCDs averaging 1.4 and 1.7 days, with the maximum values recorded at the Maerkang site in the Hengduan Mountains, at 9 and 15 days (Supplementary Fig. 3).

Overall, the multi-year average CSCDs show a gradual increased trend, while the maximum CSCDs show a downward trend (Fig. 3a). The average CSCDs at 62 sites significantly increase (p < 0.05), with an average increase rate of 0.1 days per decade. The fastest increase is observed at the Changdu site near the Hengduan Mountains, at 1 day per decade. The average CSCDs at 21 sites significantly decrease (p < 0.05), with an average decrease rate of 0.06 days per decade. The fastest decrease is observed at the Deqin site in the Hengduan Mountains, at 0.2 days per decade. At 58% of the sites, the maximum CSCDs have significantly decreased (p < 0.05), with an average reduction rate of 0.66 days per decade. The fastest decrease was recorded at the Zeku site in the Bayan Har Mountains, with a reduction rate of 4.2 days per decade. In contrast, 23% of the sites have seen a significant increase in the maximum CSCDs (p < 0.05), with an average increase rate of 1.44 days per decade. The fastest increase was recorded at the Gande site in the Bayan Har Mountains, with an increase rate of 9.7 days per decade.

Fig. 3: The spatiotemporal characteristics of CSCDs in the TP.
figure 3

The average CSCDs a spatial distribution (day), b rate of change (day per decade), and maximum CSCDs c spatial distribution (day), and d rate of change (day per decade). Same as Fig. 1.

Full size image

The annual CSCDs vary spatially across sites, with only the Nielamu site in the Himalayas having a larger average, while other regions have fewer days. Furthermore, the central and eastern regions of the TP and some sites near the Himalayas have larger maximum CSCDs, while other areas have fewer days (Fig. 3b). Only a few sites in the northern part of Animaqing and parts of the Hengduan Mountains and Nianqing Tanggula show a slight decrease in CSCDs, while most areas show a slight increase. However, the variation in the maximum CSCDs is larger, with some sites in the central, southern, and Himalayan regions showing an increasing trend, while other areas show a decreasing trend. The sites with the largest increase in CSCDs are spring, followed by winter, autumn, and summer.

The continuous no snow cover days (CNSCDs) on the TP generally exhibit opposite spatial distribution and variation characteristics to CSCDs. The multi-year average and maximum CNSCDs average 16 (S.D. = 10.42) and 116 days (S.D. = 44.82). In the Kunlun Mountains, Hengduan Mountains, and Himalayas, the average CNSCDs is relatively high. In the Bayan Har Mountains, some sites record the highest CNSCDs, while other regions have fewer. (Fig. 4a) The highest values for the annual average and maximum CNSCDs are found at Qingshuihe in the Bayan Har Mountains, and at some sites in the Hengduan Mountains and Nyenchen Tanglha Mountains, with 54 and 182 days respectively. The lowest values are observed at Deqin and Shiqu in the Hengduan Mountains, with 5 and 48 days, respectively.

Fig. 4: The spatiotemporal characteristics of CNSCDs in the TP.
figure 4

The average CNSCDs on the TP a spatial distribution (day), b rate of change (day per decade), and maximum CNSCDs c spatial distribution (day), d rate of change (day per decade). Same as Fig. 1.

Full size image

From 1961 to 2019, the overall CNSCDs on the plateau have gradually increased, with only a small number of sites showing a gradual decrease, and the decrease rate is relatively small (Fig. 4b). The highest increase rate is found at the Batang site in the Hengduan Mountains, at 44.7 days per decade, while the fastest decrease is at the Xunhua site in the Animaqing Mountains, at -4.9 days per decade. Autumn has the most CNSCDs, followed by winter and spring, indicating significant discontinuity in seasonal snow cover during autumn on the TP.

Factors affecting snow cover indices

Many studies have already shown that the spatiotemporal distribution of TPSC is mainly influenced by temperature and precipitation63,64,65. Other factors such as wind speed, altitude, and topography can also cause changes in snow cover distribution66,67,68. Using 26 factors including annual average temperature, annual precipitation, shortwave radiation, humidity, elevation, average wind speed, and maximum wind speed, etc, the geographic detector analysis based on optimal parameters was conducted to analyze the relationship with SCDs, SCNs, CSCDs, α and β. The results indicate that temperature in autumn is essentially the most important influencing factor for all indicators, followed by precipitation and radiation, significantly explaining the various snow cover indicators (Fig. 5).

Fig. 5: The factor detection results of impact factors on the snow cover indices.
figure 5

T is temperature, P is precipitation, E is elevation, H is humidity, R is shortwave radiation, Wmn is average wind speed, Wmx is maximum wind speed, \(\alpha\) is contribution and \(\beta\) is continuity.

Full size image

Climate factors also exhibit heterogeneity in their influence on snow cover. The explanation degree of annual average temperature, temperature in autumn and spring, minimum temperature, and average and maximum wind speed in winter for SCDs is significant, with the highest explanation degree being for temperature, reaching 66%. For SCNs, the explanation degree of annual average temperature, temperature in autumn, temperature and wind speed in winter, temperature in spring, temperature and wind speed in summer, as well as minimum temperature, is significant, with the highest explanation degree again being for annual average temperature, reaching 70%. For average CSCDs, the explanation degree of annual average temperature, elevation, temperature in autumn, winter, and summer, and minimum temperature is more significant, with the highest explanation degree being for spring temperature, at 39%. For maximum CSCDs, the explanation degree of annual average temperature, elevation, temperature in autumn, winter, spring, and summer, and minimum temperature is significant, with the highest explanation degree being for spring temperature, reaching 78%. Overall, temperature and precipitation in autumn have a greater impact on these indices.

Different factors have varying degrees of explanation for the CSCDs in different seasons. The factors with the highest explanation degree for autumn, winter, spring, and summer average CSCDs respectively are temperature in autumn (52%), temperature in autumn (52%), temperature in winter (36%), and precipitation in spring (32%), all passing the significance test. Similarly, for the maximum CSCDs in autumn, winter, spring, and summer, the factors with the highest explanation degree are temperature in autumn (73%), temperature in autumn (56%), temperature in spring (40%), and precipitation in spring (30%), all passing the significance test.

Interaction effects between factors have a higher explanatory degree for snow cover indices, with most p-values passing the significance test (Fig. 6). The interaction factors with relatively high explanation degrees for SCDs, SCNs, average CSCDs, maximum CSCDs is temperature in spring and precipitation in winter (92%), temperature and precipitation in winter (94%), temperature in spring and average wind speed in autumn (77%), and temperature in winter and temperature in spring (86%). This indicates that the combined effects of these factors have a more significant impact on the corresponding indices. In particular, the combination of water and heat plays a crucial role in the discontinuity of snow cover on the plateau.

Fig. 6: The interaction impact results between different impact factors on different indices.
figure 6

a SCDs, b SCNs, c average CSCDs, d maximum CSCDs, e β and f α.

Full size image

Discussion

Due to the high albedo and low thermal conductivity of snow, as well as the latent heat generated during phase transitions, changes in snow cover significantly influence the effect of temperature on soil freeze state69,70,71,72, subsequently impacting soil and perennial permafrost freeze-thaw processes73,74,75. Therefore, snow cover is commonly regarded as the most critical factor affecting soil temperature or soil freeze state76,77,78,79. The impact of snow cover on soil temperature varies depending on characteristics such as snow depth and SCDs. Typically, snow depth is negatively correlated with soil temperature and frost depth. However, Wang et al. suggest that in the pan-Arctic region, when SCDs exceed 330 days per year, snow cover favors the development of multi-year permafrost80. Furthermore, in northeastern Siberia, soil temperature shows a stronger correlation with SCDs than with snow depth in the polar north72,81. However, in regions where snow cover is sparse and short-lived, snow cover usually promotes permafrost development due to the latent heat consumed during snow melt and its reflective effect on solar radiation. For instance, Tan et al. demonstrated a significant negative correlation (the correlation coefficient is −0.7) between surface soil temperature and SCDs in the TP82.

Our study reveals that TPSC is highly discontinuous over time, with an annual average of 14 snow cover events and prolonged periods without snow cover. The Pearson correlation coefficients (r) between annual average soil temperature, soil temperature in snow cover season, autumn, winter and spring and SCNs are −0.85, −0.80, −0.80, −0.70, and −0.88 (Fig. 7), respectively, markedly higher than those with SCDs (−0.77, −0.73, −0.72, −0.63, and −0.81). Additionally, SCDs and maximum CSCDs exhibit similar r with annual average soil temperature, while the r of maximum CSCDs with soil temperature in autumn and winter (−0.75 and −0.71) is slightly higher than that of SCDs (−0.72 and −0.63), and the r with soil temperature in spring (−0.79) is lower than that of SCDs (−0.81). Similarly, the r between average CNSCDs and soil temperature follows the same pattern, but it is positively correlated with soil temperature, indicating that maximum CSCDs have a greater impact on soil temperature in autumn and winter. However, the influence of average CSCDs on soil temperature is different. It is negatively correlated with soil temperature when it is low, while it becomes positively correlated above a certain critical threshold. For example, when soil temperature in spring is below 14.2 °C, the r between average CSCDs and soil temperature in spring is −0.61. When it exceeds 14.2 °C, the r is 0.57. These sites are predominantly located in valley areas characterized by lower elevation and higher humidity, where air circulation is limited. Additionally, these regions experience lower radiation and thinner snow depth, resulting in reduced cooling effects from snow. Consequently, although snow cover may be relatively continuous, soil temperature remains higher. Thus, the discontinuity of TPSC has a significant impact on soil temperature and cannot be overlooked.

Fig. 7: The correlation between ST at different periods and snow indices.
figure 7

All passed the significance test of 0.001.

Full size image

Although the spatial distribution of SCDs and CSCDs on the TP is relatively consistent, with high values mainly concentrated near the Bayan Har and Animaqing Mountains, there are considerable differences in the magnitude of SCDs and CSCDs (Fig. 1, Fig. 3). For example, the average multi-year SCDs at each site are 33 days, significantly greater than the annual average CSCDs of 3 days, but less than the average and maximum CNSCDs (16 days and 116 days), and the annual average SCNs reach 14 times. This indicates that the seasonal snow cover on the TP has poor temporal continuity, mainly characterized by snow-free intervals. Meanwhile, both SCDs and maximum CSCDs show a decreasing trend, while the corresponding CNSCDs show a significant increase. Hence, while there has been considerable analysis using SCDs (Table 1), they cannot fully capture the temporal continuity and changes of TPSC. In conclusion, CSCDs are preferred6.

Table 1 The variation characteristics of SCDs in different types of datasets on the TP over the past several decades
Full size table

Although some scholars have considered using the continuity of snow cover to classify snow cover, the commonly used average value of maximum CSCDs as an index cannot reflect the characteristics of SCNs6,22 and snow-free period intervals. We can use the contribution (α) to describe the overall contribution features of snow cover events on the TP. However, due to significant interannual variations in snow cover events at some sites, it is not possible to accurately obtain cumulative contribution rates. Additionally, at certain sites, such as the Guide site in the Qilian Mountains, there were years (e.g., 1978–1985) with only one day of snow cover or even no snow cover at all, making it challenging to precisely describe the continuity of snow cover. Therefore, considering the influence of snow-free period intervals, at the annual scale, continuity (β) can be used to represent the temporal continuity of snow cover on the TP. Thus, we attempt to utilize α and β to describe the continuity characteristics of TPSC.

By calculating the probability of snow events over 59 years on the TP, α is determined to be 0.92 (S.D. = 0.45). Randomly selecting regions with relatively uniform distributions, such as Xinjiang (21 sites), Northeast (21 sites), Europe (39 sites), and North America (23 sites), for calculation, it is found that the contribution of the TP to snow cover is significantly smaller than that of other snow cover areas (Supplementary Table 1). The α exhibits TP > Xinjiang in China > Europe > Northeast in China > North America. Similarly, β can be used to describe the interannual snow continuity characteristics of snow cover areas (Table 2), with the average β of the TP being 0.04, also significantly smaller than other snow cover areas. The β exhibits North America > Europe > Xinjiang > Northeast China > TP. The average CSCDs in Northeast and Xinjiang in China are 14 and 19 days. However, due to the earlier start (Do = 68 days) and later end (De = 212 days) of snow cover events in Northeast China, leading to a longer snow cover period, β is relatively smaller compared to other snow-covered regions. The average CSCDs in autumn, winter, and spring in Northeast China are 3, 13, and 13 days, while in Xinjiang, they are 3, 11, and 32 days, respectively, with spring CSCDs in Xinjiang being significantly greater than in Northeast China. Therefore, both α and β can accurately describe the continuous characteristics of seasonal snow cover on the TP.

Table 2 α and β in different snow zones
Full size table

The distribution pattern of α exhibits a trend of smaller values in the middle and larger values around the periphery (Fig. 8a). Due to the larger CSCDs in the Bayan Har Mountains region, snow events with longer CSCDs contribute more to the total SCDs in the area. The higher α values are mainly distributed in parts of the Hengduan Mountains, Qilian Mountains, and Nyenchen Tanglha Mountains. In other regions, snow events with shorter CSCDs contribute more to the total SCDs. Meanwhile, as the snow-free interval days on the TP are much greater than the CSCDs, the overall β value for the TP is small, indicating poor snow cover continuity (Fig. 8b). Only the Gongshan, Derong, and Batang sites in the Hengduan Mountains are relatively high, at 0.37, 0.33, and 0.30, respectively, indicating stronger snow cover continuity. Due to the influence of complex terrain and harsh weather conditions on the placement of ground stations, an uneven distribution of stations is observed. Research sites are mainly concentrated in the eastern part of the TP, with fewer stations in the western part, leading to insufficient consideration of the spatial differences in snow cover on the TP. This study has focused on a detailed investigation of continuous snow cover events on the TP, and future research could classify or zone TPSC based on the continuity of seasonal snow cover.

Fig. 8: The spatial distribution characteristics of α and β in the TP.
figure 8

a α and b β. Same as Fig. 1.

Full size image

In addition to being influenced by local climates such as temperature and P, the continuity characteristics of TPSC are also affected by large-scale atmospheric factors such as Arctic Oscillation, atmospheric circulation, etc83,84,85 (Fig. 9).

Fig. 9: Circulation background of TPSC.
figure 9

AO Arctic Oscillation, NAO North Atlantic oscillation, ENSO El Niño-Southern Oscillation.

Full size image

In autumn, the rapid southward retreat of the Indian monsoon and the continuous southward intrusion of cold air in the north converge with warm and moist airflow moving northward along the Hengduan Mountains. This convergence results in a rapid decrease in temperature and significant precipitation in the Bayan Har Mountains, conducive to sustained snow accumulation86. Additionally, significant anomalies in sea surface temperatures in the eastern North Pacific and western North Atlantic induce disturbances in the overlying atmosphere, resulting in different wave activities over the eastern North Pacific and western North Atlantic. This leads to the formation of a high-pressure system over the western TP and an anomalous low-pressure system over the eastern TP, causing increased upward motion and more precipitation, thus contributing to continuous snow cover events in the eastern regions during autumn87,88. In winter, multiple atmospheric circulation patterns may jointly influence the temperature and precipitation of the plateau, thereby affecting the α and β of snow cover events on the TP (Figs. 5 and 6).The El Niño-Southern Oscillation may trigger stationary Rossby waves extending along the North African-Asian jet stream, leading to abnormal increases in potential vorticity and snow depth over the TP89,90. It may also strengthen the East Asian and South Asian monsoons by modulating convection in the western North Pacific, leading to prolonged snow cover duration in the eastern Tibetan Plateau91,92,93. Convective anomalies associated with the Indian Ocean Dipole in the western Indian Ocean may also affect the duration of snow cover in the Tanggula Mountains. Furthermore, under a positive Arctic Oscillation modulation, the weakening of the Baikal Ridge pushes cold air southward, leading to favorable conditions for more sustained snowfall in that region83,94,95. Moreover, the intensification or weakening of the North Atlantic Oscillation phenomenon, driven by the Icelandic low and the Azores high over the upstream Eurasian continent, can strengthen or weaken the wave activity over the North Atlantic87,96, thereby enhancing the discontinuity of snow cover on the TP during winter. In spring, the circulation background is similar to that of winter.

Various circulations may influence the continuity of TPSC by affecting meteorological conditions, especially temperature and precipitation. This study only provides a preliminary interpretation of the relationship between the continuity characteristics of TPSC and climate factors using meteorological station data. Further research is needed to explore the differences in the impact of influencing factors on snow cover continuity and the differences in influencing factors in different regions of the TP. Subsequent studies could investigate the spatiotemporal differences in the effects of factors of different magnitudes on snow cover continuity, such as the influence of precipitation at different levels and the combination of water and heat on the continuity of TPSC, and explore the complex response mechanisms between climate factors and snow cover continuity. Moreover, there is a need to strengthen relevant studies on numerical verification.

Methods

Study area

The TP, located in southwestern China (26°00′12″ ~ 39°46′50″N, 73°18′52″ ~ 104°46′59″E), extends from the Pamir Plateau in the west to the Hengduan Mountains in the east, spanning approximately 2700 km from west to east and 1400 km from north to south (Fig. 10). With a total area of about 2.5 million square kilometers, the plateau is characterized by an average elevation of over 4000 meters. Its climate features long and cold winters, brief and warm summers, significant diurnal temperature variations, and a highly uneven spatial distribution of precipitation, which mainly occurs from May to September. Snowfall typically begins in September, with the prominent snow cover period lasting from September to April of the following year. Rich in snow and glaciers, the TP hosts the largest extent of the cryosphere outside of polar regions, including snow, ice, glaciers, and permafrost97,98,99. As global climate warming accelerates, the TP experiences rapid temperature increases, leading to the accelerated melting of glaciers and permanent snow cover, as well as a gradual reduction in seasonal snow cover. Consequently, the region exhibits a heightened sensitivity to global climate change responses.

Fig. 10
figure 10

Study area.

Full size image

Data sources

Meteorological data

The meteorological data used in this study are sourced from the China Meteorological Administration’s National Meteorological Information Center dataset V3.0, which includes observations from Chinese surface meteorological stations. Due to the uneven distribution of stations on the TP and objective factors such as missing measurements for certain months or years, some stations with short observation periods or over half of the data missing were excluded from the analysis after comprehensive consideration of the objective situation and research methods. Ultimately, temperature, precipitation, wind speed, 0 cm soil temperature and other data from 83 stations spanning from September 1, 1961, to August 31, 2019, were selected for analysis in this study. The study simultaneously utilized monthly averaged solar radiation (W/m2) and 10 m instantaneous wind speed (m/s) reanalysis data from ERA5 monthly averaged data on pressure levels from 1940 to present dataset at the European Centre for Medium-Range Weather Forecasts (ECMWF).(https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanalysis-era5-pressure-levels-monthly-means?tab=overview).

Snow depth data

Snow depth data are obtained from the Chinese historical daily snow depth dataset published by the China Meteorological Administration, National Meteorological Information, which includes data from 212 stations (ftp://ccrp.tor.ec.gc.ca/pub/RBrown/China_Snow_Data). European data are sourced from the Federal Service for Hydrometeorology and Environmental Monitoring, World Data Centers (http://aisori-m.meteo.ru/waisori/result). Snow depth data for the Americas are sourced from the National Water and Climate Center NWCC iMap (https://nwcc-apps.sc.egov.usda.gov/imap). Considering the seasonal characteristic of snow cover in the Northern Hemisphere, a snow year is defined as from September 1 of the previous year to August 31 of the current year, with each snow year starting in autumn and ending in summer. Specifically, the previous year’s September, October, and November are considered autumn; December, January, and February of the current year are considered winter; March, April, and May are considered spring; and June, July, and August are considered summer. Based on the historical snow depth data, snow depth greater than 0.5 cm is considered as snow cover, and a series of snow cover indicators were developed (Table 3).

Table 3 Snow indicators and their significance
Full size table

Snow cover continuity indicators analysis

The Generalized Pareto Distribution is a function specifically used to describe the probability distribution of data values exceeding a certain threshold. It was initially introduced by Pickands into hydro-meteorology100 and has since been extensively studied and expanded upon101. It can more accurately describe the probability distribution of values exceeding a threshold (extreme values). The distribution of CSCDs follows a Pareto distribution, where snow events with fewer CSCDs are more likely to occur, and snow events with more CSCDs are less likely to occur. The formula is as follows:

$$p\left(x\right)=\left\{\begin{array}{ll}0 & x\, <\, {x}_{\min }\\ \frac{\alpha {x}_{\min }^{\alpha }}{{x}^{\alpha +1}} & x\,\ge\, {x}_{\min }\end{array}\right.$$

where \(p\) is the probability of snow cover events, \(x\) refers to the SCDs (day), xmin represents the minimum value of \(x\), the minimum SCDs; \(\alpha\) is the shape parameter or contribution, used to describe the overall contribution of continuous snow events to long-term snow cover. Simulation results show that as \(\alpha\) becomes larger (or smaller), snow events with fewer (or more) continuous snow cover days contribute more (or less) significantly to the snow cover days, and the continuity larger (or smaller).

Additionally, a new index has been established to describe the continuity of TPSC.

$$\beta =\left\{\begin{array}{ll}0.02 & {\rm{SCN}},{\rm{SCDs}}=1\\ \frac{{CSCD}{mn}}{{SSL}}&{\rm{SCN}}\, >\, 1\end{array}\right.$$

where \({CSCD}{mn}\) denotes average CSCDs, \({SSL}\) refers snow season length, it means duration from the onset day to the end day of snow cover. β is the continuity parameter, β∈(0,1], used to describe the continuity of snow events over time. As β increases (or decreases), the continuity increases (or decreases).

Geographic detector model based on optimal parameters

The geographic detector is a set of statistical methods used to detect spatial differentiation and reveal the driving forces behind it102. Differentiation and factor detection analysis refers to detecting the spatial differentiation of variable and the extent to which a factor explains the spatial differentiation of variable. Traditional geographic detector exhibits strong subjectivity when discretizing continuous variables, which affects the determination of the optimal scale of spatial stratification heterogeneity to some extent. This study utilized the geographic detector package in shortwave radiation language and employed five classification methods—equal intervals, natural breaks, quantiles, geometric intervals, and standard deviations—to set the number of classification levels to several classes. The optimal parameter combination method was selected for spatial discretization103. The q value is measured104, and the expression is:

$$q=1-\frac{\mathop{\sum }\nolimits_{h=1}^{L}{N}_{h}{\sigma }_{h}^{2}}{N{\sigma }^{2}}=1-\frac{{SSW}}{{SST}}$$
$${SSW}=\mathop{\sum }\limits_{h=1}^{L}{N}_{h}{\sigma }_{h}^{2}$$
$${SST}=N{\sigma }^{2}$$

Where humidity represents the strata of snow cover indices or impact factors; \({N}_{h}\) and N are the number of units in stratum humidity and the entire area; \({\sigma }_{h}^{2}\) and \({\sigma }^{2}\) are the variances of snow cover indices in stratum humidity and the entire area. SSW and SST are the sum of within-stratum variance and the total variance of the entire area. The range of q values is [0,1], with higher values indicating more pronounced spatial differentiation of snow cover indices. If strata are generated by independent impact factors, a higher q value indicates a stronger explanatory power of independent impact factors on attribute snow cover indices, and vice versa. In extreme cases, a q value of 1 indicates that impact factors completely control the spatial distribution of snow cover indices, while a q value of 0 indicates no relationship between impact factors and snow cover indices. The q value represents impact factors explaining 100 × q% of snow cover indices. Some scholars categorize the magnitude of q into four classes: insignificant (<0.1), average significant (0.1~0.3), more significant (0.3~0.5), and significant (0.5~1)105.

Pearson correlation coefficients

Pearson correlation coefficients were used to analyze the relationship between snow cover indices and soil temperature. Pearson correlation coefficients can measure the linear correlation between two variables x and y. The calculation formula is as follows106:

$$r=\frac{\mathop{\sum }\nolimits_{i=1}^{n}({x}_{i}-\bar{x})({y}_{i}-\bar{y})}{\sqrt{\mathop{\sum }\nolimits_{i=1}^{n}{({x}_{i}-\bar{x})}^{2}}\sqrt{\mathop{\sum }\nolimits_{i=1}^{n}{({y}_{i}-\bar{y})}^{2}}}$$

where \(x\) is the variable of snow cover indices, and \(y\) represents soil temperature. The r > 0, r < 0, and r = 0 respectively represent positive correlation, negative correlation, and no correlation between the two factors.