Are there spatiotemporal characteristics and differences in snow density between the Tibetan Plateau and the Arctic? How can we predict the future of the Tibetan Plateau and the Arctic region through climate change?

The Tibet Plateau (TP) and the Arctic are typically cold regions with abundant snow cover, which plays a key role in land surface processes. Knowledge of variations in snow density is essential for understanding hydrology, ecology, and snow cover feedback. Here, we utilized extensive measurements recorded by 697 ground-based snow sites during 1950–2019 to identify the spatio-temporal characteristics of snow density in these two regions. We examined the spatial heterogeneity of snow density for different snow classes, which are from a global seasonal snow cover classification system, with each class determined from air temperature, precipitation, and wind speed climatologies. We also investigated possible mechanisms driving observed snow density differences. The long-term mean snow density in the Arctic was 1.6 times that of the TP. Slight differences were noted in the monthly TP snow densities, with values ranging from 122 ± 29 to 158 ± 52 kg/m3 . In the Arctic, however, a clear increasing trend was shown from October to June, particularly with a rate of 30.3 kg/m3 per month from March to June. For the same snow class, the average snow density in the Arctic was higher than that in the TP. The Arctic was characterized mainly by a longer snowfall duration and deeper snow cover, with some areas showing perennial snow cover. In contrast, the TP was dominated by seasonal snow cover that was shallower and warmer, with less (more) snowfall in winter (spring). The results will be helpful for future simulations of snow cover changes and land interactions at high latitudes and altitudes.Snow density is one of the most fundamental properties of snow cover [1]. It has critical effects on energy budgets [2] and snow surface temperature [3], as the thermal conductivity of snow is controlled mostly by its density [4]. Snow density is also a significant attribute that impacts various types of snow cover remote sensing applications because it is directly linked to the dielectric permittivity of snow, which controls the emission and reflection of electromagnetic waves in the snow layer [5]. Moreover, the mechanical interactions that occur within snow cover or between snow cover and the surroundings are related to snow density [6,7]. Furthermore, snow properties, such as porosity [6], optical properties [8],and permeability [9], are connected with snow density. Snow density is a key input parameter for many models, including land surface models [10] and snow models [11,12]. In addition, snow density establishes the relationship between snow depth and the snow water equivalent (SWE), which makes it indispensable in a variety of snow hydrology studies and applications, such as water supply [13], flood forecasting [14], and snowmelt runoff quantification [15]. Snow density varies with time, place, and surrounding weather conditions [11,16]. When the temperature is very low during snowfall, the new snow typically has a rather low density, while higher densities for new snow occur at higher temperatures [17,18]. Snow densification begins immediately after snowfall reaches the ground and is mainly caused by metamorphism (snow grains change in size and shape), wind action, and compaction [16,19]. There are three types of snow metamorphism: constructive metamorphism (kinetic growth metamorphism), destructive metamorphism (equilibrium growth metamorphism), and wet snow metamorphism (melt or melt-freeze metamorphism) [16,20,21]. Constructive metamorphism is caused by strong vertical temperature gradients (low air temperature) that occur throughout the snowpack, which produce large, faceted grains referred to as depth hoar. These large grains are hard to pack closely together, resulting in low rates of densification [16,21,22]. The snow grains slowly metamorphose into rounded shapes through destructive metamorphism caused by weak vertical temperature gradients (warmer air temperature) [6,16,21]. The thermal isolation of snow causes a higher snow ground interface temperature than the snow surface temperature, which results in a vertical temperature gradient in the snow layer (◦C/m). Lower air temperature is more favorable for producing strong vertical gradients. When the air temperature is warm, the vertical temperature gradient is weak [6,16]. Melt metamorphism typically occurs when liquid water is present in the snow cover, which also accelerates destructive metamorphism [22]. Regardless of the snow metamorphism type, increases in snow density could be the result of overburden pressure [16,23,24]. The prevalent high wind is known to redistribute the snow grains, which makes the snow cover become more densely packed, easily producing greater densities [18,25,26].However, several studies have used fixed [11,27] or modeled snow density values [28,29], which can lead to significant errors of up to 100% when estimating snow depth [30]. In addition, directly determining snow density using remote sensing methods remains problematic [5,31]. Thus, site and field measurements are the primary methods for investigating snow density. There are several approaches available to directly measure snow density, such as gravimetry measurement [32], stereology [33], microcomputed tomography [34], high-resolution osmometry [35], dielectric permittivity measurement [1], the neutron-scattering probe [36], and diffuse near-infrared transmission [8]. Compared to the cumbersome direct measurement, gathering snow density by determining the SWE and the snow depth is more manageable, less expensive, and less time-consuming [37]. Nevertheless, a lack of extensive ground-based data remains the greatest challenge in analyzing snow density over large areas [6,17]. For instance, field snow observations conducted on a 5.4 km section in the Central Spanish Pyrenees Mountains indicated variation in snow density but showed no discernible pattern [38]. Moreover, insufficient snow density measurements may also lead to considerable uncertainty in snow density analyses. In northern taiga and tundra areas, it has been shown that on some days a single measurement might represent the snow density for a land cover group because of the very limited measurements [7]. The spatio-temporal characteristics of snow density have not been well described thus far, particularly in the overall Arctic region. The TP and the Arctic are typically cold regions and are extremely sensitive to global climate change because both have shown amplified warming in response to global warming [39,40]. The TP covers the largest area of snow and glaciers outside the poles [41–43] and is referred to as the “Water Tower of Asia” [44–46]. The snow density in each of these two areas has been examined in previous research [7,47–50]. For example, in the middle and southwestern parts of the TP, the snow density obtained from ground-based snow sites ranged from 140 kg/m3 to 180 kg/m3 during 1950–2009 [49]. The shallow snow depth and significant depth hoar fraction in the high Arctic resulted in a high average snow density of 298 kg/m3 [25]. Yet, studies of snow density in the Arctic are available for parts of the region, and there are no systematic studies for the entire region [25,51]. Whether there are similarities or differences between the TP and Arctic remains inconclusive, as most snow density studies have focused exclusively on one or the other of the two regions, and there have not been many attempts to systematically contrast the snow density characteristics between them. Such knowledge gaps hamper our understanding of the interaction between snow and climate in cold regions. For this study, we compiled a wide range of available ground-based datasets of snow density across the TP and the Arctic. Our primary objectives were (1) to provide information on the spatio-temporal features of snow density across the TP and the Arctic, (2) to investigate regional heterogeneities in snow density based on the Global SeasonalSnow Classification, Version 1, and (3) to explore the possible reasons for the differences in snow density between these two regions. The results will shed light on the different roles of snow cover regarding regional hydrology and ecology, as well as improve the modeling of snowpack evolution in land surface models.. Discussion In the Arctic, no dedicated agency conducts systematic snow surveys over the whole region. Instead, multiple national agencies use different methods to record observations, as well as a few short-term field surveys performed by researchers in localized areas. As a result, snow density studies have been limited to only parts of the region [25,47,77]. Here, we generalized multiple long-term snow measurements from different subregions of the Arctic, with sites essentially covering the Arctic. Moreover, we provided systematic and representative spatio-temporal characteristics of snowpack density on a regional scale. Our study is temporally complete and includes observations from September to June, whereas the vast majority of studies concentrate on values obtained between November and April or shorter periods. Furthermore, we adopted the latest version of the snow classification dataset [54]. Few studies have been conducted on snow density based on snow classes in the TP. Our study involved almost all snow classes in the Arctic and was facilitated by the extensive availability of snow density observations. In contrast, most of the current studies focus on the tundra and boreal forest snow [7,47,77]. Our study is the first to systematically compare the differences in snow density characteristics between the Arctic and the TP. 4.1. Spatial and Temporal Patterns of Snow Density in the TP and the Arctic We obtained a lower average snow density of 138 ± 61 kg/m3 during 1954–2012 in the TP, with respect to that of 156 kg/m3 for September–May during 2013–2020, as reported by Wang et al. [78]. This small disparity is attributed to our use of all available measurements recorded between September and June for the different study periods. The average snow density observed in Alaska, 225 ± 61 kg/m3 , relatively agrees with that observed by Sturm et al. [13] in northwestern Alaska in 2002 at approximately 245 kg/m3 . For snow classes, our results confirmed previous findings, such that the greatest snow density and range occurred at maritime snow sites and boreal forest had the lowest snow density and was the least spatial heterogeneous [13,53]. A few studies based in North America and the former Soviet Union have reported that ephemeral snow typically has a relatively high snow density among snow classes [53,55]. The results from our case demonstrated that the average snow density was very low at the ephemeral sites in the TP. This occurred because fewer snow density measurements were available for the ephemeral snow, and differences in regional sample sizes gave slightly different results. A generalization often mentioned in previous research is that snow density gradually increases during the snow cover season [6,53,79]. We found similar results in the Arctic. Surprisingly, the seasonal snow density in the TP was relatively constant and did not show the expected increasing trend over time, particularly in winter. The seasonal variation in snow density highlights the unique features of the snow cover in the TP. In this study, the snow density exhibited strong regional variations between the different snow classes during different months. Despite this notable difference, some SWE retrievals utilized a constant snow density throughout the retrieval, regardless of the snow cover condition, place, or time [11,80]. Moreover, a considerable number of models assumed that the snow density was a constant rather than a variable value [37,81]. Given these facts, our findings can be applied to various types of applications, including remote sensing retrieval and validation, model simulation and prediction (e.g., snow, atmosphere, land surface schemes), and hydrological and climate studies. 4.2. Effect of Environmental Factors on Snow Density over Two Regions The results of our study indicated that the snow density disparity between the TP and the Arctic was associated with differences in air temperature, snowfall, and wind conditions. In the Arctic, constructive metamorphism caused by strong vertical temperature gradients (low air temperature) occurred from November to March (Figure 7d), producing large grains that contributed to the slow rates of densification (Figure 5b) because they were hardly packed closely together. In addition, the increase in snow density was caused by continuous compaction facilitated by the weight of the abundant snowfall (Figure 7e).Wind-driven redistribution was a primary factor controlling the increase in snow density from November to February (Figure 7f). In April, the combination of melt metamorphism, accelerated destructive metamorphism, and overburden pressure reduced the porosity of the snow cover, which caused the snow density to increase at high densification rates. This seasonal persistent snow cover, referred to as firn, has densities typically ranging from 550 kg/m3 to 800 kg/m3 [22] and is a driver of the high snow density in the Arctic. In the TP, slow destructive metamorphism caused by weak vertical temperature gradients (warmer air temperature) generally dominated the densification process from November to February (Figure 7a). The rare snowfall and shallow snow depth produced negligible overburden pressure (Figure 7b). Despite the windy winter conditions (Figure 7c), very little snowfall limited the effect of wind on snow density. Hence, little change was noted for snow density in winter (Figure 5a). When the snow cover melted in March, melt metamorphism and accelerated destructive metamorphism were more prone to producing high densification rates. However, the shallow snow depth restricted development, which, combined with new snow and low density, ultimately resulted in low rates of densification. Thus, snowfall and temperature were the most useful explanatory terms for seasonal variability in snow density in the TP. Despite a large body of available in situ observations utilized in this study, one of the uncertainties was the uneven distribution of in situ snow and meteorological measurements, especially in the Canadian Arctic, where meteorological stations were extremely scarce. Different snow density variation behaviors can be seen under different rainfall conditions. On the one hand, snow cover could absorb rain causing an increase in snow density, while repeated rainfall damages the ice structure and reduces the water absorption capacity of the snow cover. As a result, rainfall reduces snow density [18]. However, the Arctic winters were long in duration and generally too cold for rainfall to occur. Hence, we did not take rainfall into account and only considered factors that highlight climatic conditions: temperature, snowfall, and wind speed.In general, the Arctic had a large amount of snowfall, deeper and colder snow cover with higher wind speeds, a longer snowfall duration, and the existence of perennial snow cover. In contrast, the TP had less snowfall and lower wind speeds, shallower and relatively warmer snow cover, the duration of snowfall was shorter, and less (more) snowfall occurred in winter (spring). These factors caused the long-term mean snow density in the TP, 138 ± 61 kg/m3 , to be lower than that in the Arctic, 223 ± 54 kg/m3 . Seasonally, the monthly snow density in the Arctic exhibited strong intra-annual variability from October to June, particularly during the late snow cover season. In the TP, however, only slight differences were noted among monthly snow densities that ranged from 122 ± 29 to 158 ± 52 kg/m3 . A comparison of the seasonal evolution of snow density among the different snow classes revealed similar variations in the Arctic. In the TP, however, no clear consistent variation between the months was noted in the snow classes. Moreover, we found that strong vertical temperature gradients in the Arctic resulted in low densification rates in winter. In the TP, however, weak vertical temperature gradients and negligible overburden pressure resulted in fluctuations in snow density. Despite the limited consistency and continuity, as well as the inhomogeneous spatial distribution of the observations examined in this study, the results are critical for investigating the spatio-temporal patterns and differences in snow density between the TP and the Arctic. Therefore, the results of this study will enhance our understanding of the regional characteristics and discrepancies in snow density, particularly those occurring in cold and remote regions where monitoring is sparse.