Variability Of Snow Depth Research Articles

Variability in observed snow depth over China from 1960 to 2014

AbstractSnow depth is a critical variable that reflects snow variability and has an important impact on the energy and hydrological cycles. However, snow depth changes in response to regional climate warming have not been quantified in detail across China. Here, we investigated the variability in observed snow depth based on a recently released dataset collected at ~1,500 stations from 1960 to 2014. To address the non‐normal distribution of daily snow depth data, the annual cumulative amount of snow depth (CASD) is employed, and the trend is calculated and tested for the change point and field significance. The results showed that CASD experienced a slight increase of 1.75%·decade−1in China. An increasing trend of CASD was observed in Northeastern and Northwestern China, a decreasing trend was observed in the eastern Tibetan Plateau, Loess Plateau, and Yangtze River Basin, and these changes were more remarkable after the 2000s. Since the 2000s, the increased CASD in Northwestern and Northeastern China has been accompanied by a significant decrease in winter temperature and a sharply increasing trend in snowfall events in the northeastern region. The decreased CASD in the Eastern Tibetan Plateau and Loess Plateau after the 2000s was accompanied by a decrease in snowfall events and an increase in winter temperature. Large‐scale atmospheric circulation, such as the Arctic Oscillation (AO), has a remarkable regional effect on the variability in snow depth. Statistically significant negative correlations were found in the northern part of China (i.e., positive AO phase with reduced CASD), while significant positive correlations were shown in the middle part of China (i.e., positive AO phase with enhanced CASD). Furthermore, China experienced an obviously shorter snow season during the whole period (decreased by 3.87%·decade−1), jointly resulting from later starting dates and earlier ending dates.

Cover Image

The cover image is based on the Short Communication* Ground temperature and snow depth variability within a subarctic peat plateau landscape by A. Britta K. Sannel, https://doi.org/10.1002/ppp.2045.

Weather stochasticity and alternative reproductive tactics in Northern chamois, Rupicapra rupicapra

Abstract In polygynous ungulates, male breeding success is often achieved through alternative reproductive tactics (ARTs). The maintenance of ARTs in this taxon is largely explained by condition-dependent selection, a decision rule in which individuals switch their breeding behaviour on the basis of internal factors such as ageing (status-dependent selection), or external factors such as weather conditions during the rut (environment-dependent selection). In some species, however, ARTs are fixed during the mating season, and other forms of selective pressures rather than environment-dependent selection may contribute to the evolutionary maintenance of male types within populations. Assuming female vicinity as a proxy of mating opportunity, we explored the effect of snow depth variations on the elevational overlap between females, and territorial and non-territorial male chamois, Rupicapra rupicapra, in three consecutive rutting seasons. In snow-free periods, females overlapped with non-territorial males, but not with territorial ones. With increasing snow depth, females and territorial males progressively occupied similar elevations. Our results suggest that snow cover variations may impose changes in the spatial distribution of males and females. In turn, weather stochasticity over multiple ruts, rather than within-rut environment-dependent selection, may impose variations in mating opportunities of ARTs, possibly contributing to the coexistence of different phenotypes within populations.

Forest cover and topography regulate the thin, ephemeral snowpacks of the semiarid Southwest United States

AbstractIn the Southwest United States, water resources depend heavily on snowpacks, which are temporally and spatially limited in this warm, semiarid region. Snow accumulation and ablation in the Southwest are heavily influenced by forest structure. Therefore, water resource managers urgently need to understand the future impacts of unprecedented forest changes now occurring from drought, insect infestation, and forest management. Here, we present state‐of‐the‐art maps and time series of snow water equivalent (SWE), which account for spatial variability of snow depth and snow density over a large range of forest structure and topographies in the highlands of central Arizona. We show that the degree of forest cover and its geometry largely determine which areas have more/less snow accumulation and faster/slower ablation. Compared with under‐canopy areas, open areas can have 20–30% more accumulation and ablation rates can vary by 15–30% in sunny areas versus shaded areas. Although SWE response to forest cover is widely variable, depending on how much shading the trees provide versus how much snow is intercepted and lost through canopy sublimation, dense forests generally have less SWE than sparser forests. In general, SWE is optimized at intermediate levels of forest cover (~30–50%) on flat and north‐facing slopes. Somewhat counterintuitively, increasing forest cover generally causes a reduction of SWE on south‐facing slopes, where trees are less effective at reducing radiative forcing at the snow surface due to less efficient shading and increased enhancement of longwave radiation from the warm canopy.

Ground temperature and snow depth variability within a subarctic peat plateau landscape

AbstractSubarctic permafrost peatlands cover extensive areas and store large amounts of soil organic carbon that can be remobilized as active layer deepening and thermokarst formation increase in a future warmer climate. Better knowledge of ground thermal variability within these ecosystems is important for understanding future landscape development and permafrost carbon feedbacks. In a peat plateau complex in Tavvavuoma, northern Sweden, ground temperatures and snow depth have been monitored in six different landscape units: on a peat plateau, in a depression within a peat plateau, along a peat plateau edge (close to a thermokarst lake), at a thermokarst lake shoreline, in a thermokarst lake and in a fen. Permafrost is present in all three peat plateau landscape units, and mean annual ground temperature (MAGT) in the central parts of the peat plateau is −0.3°C at 2 m depth. In the three low‐lying wetter or saturated landscape units (along the thermokarst lake shoreline, in the lake and the fen) taliks are present and MAGT at 1 m depth is 1.0–2.7°C. Topographical differences between the elevated and low‐lying units affect both local snow depth and soil moisture, and are important for ground thermal patterns in this landscape. Permafrost exists in landscape units with a shallow mean December–April snow depth (<20 cm) whereas snow depths >40 cm mostly result in absence of permafrost.

Snow Depth Patterns in a High Mountain Andean Catchment from Satellite Optical Tristereoscopic Remote Sensing

AbstractObtaining detailed information about high mountain snowpacks is often limited by insufficient ground‐based observations and uncertainty in the (re)distribution of solid precipitation. We utilize high‐resolution optical images from Pléiades satellites to generate a snow depth map, at a spatial resolution of 4 m, for a high mountain catchment of central Chile. Results are negatively biased (median difference of −0.22 m) when compared against observations from a terrestrial Light Detection And Ranging scan, though replicate general snow depth variability well. Additionally, the Pléiades dataset is subject to data gaps (17% of total pixels), negative values for shallow snow (12%), and noise on slopes >40–50° (2%). We correct and filter the Pléiades snow depths using surface classification techniques of snow‐free areas and a random forest model for data gap filling. Snow depths (with an estimated error of ~0.36 m) average 1.66 m and relate well to topographical parameters such as elevation and northness in a similar way to previous studies. However, estimations of snow depth based upon topography (TOPO) or physically based modeling (DBSM) cannot resolve localized processes (i.e., avalanching or wind scouring) that are detected by Pléiades, even when forced with locally calibrated data. Comparing these alternative model approaches to corrected Pléiades snow depths reveals total snow volume differences between −28% (DBSM) and +54% (TOPO) for the catchment and large differences across most elevation bands. Pléiades represents an important contribution to understanding snow accumulation at sparsely monitored catchments, though ideally requires a careful systematic validation procedure to identify catchment‐scale biases and errors in the snow depth derivation.

Resolving Small‐Scale Forest Snow Patterns Using an Energy Balance Snow Model With a One‐Layer Canopy

AbstractModeling spatiotemporal dynamics of snow in forests is challenging, as involved processes are strongly dependent on small‐scale canopy properties. In this study, we explore how local canopy structure information can be integrated in a medium‐complexity energy balance snow model to replicate observed snow patterns at very high spatial resolutions. Snow depth distributions simulated with the Flexible Snow Model (FSM2) were tested against extensive experimental data acquired in discontinuous subalpine forest stands in Eastern Switzerland over three winters. While the default canopy implementation in FSM2 fails to capture the observed snow depth variability, performance is considerably improved when local canopy cover fraction and hemispherical sky view fraction are additionally accounted for (30% reduction in root mean square error). However, realistic snow depth distribution patterns throughout the season are only achieved if effective temperatures of near and distant canopy elements are discerned and if a mechanism to mimic preferential deposition of snow in canopy gaps is included. We demonstrate that by diversifying the canopy structure input in order to reflect respective portions of the canopy relevant to different processes, even a simple model based on widely used process parameterizations and canopy metrics can be applied for high‐resolution simulations of the sub‐canopy snow cover with just a few modifications. The presented approaches could be implemented in commonly used land surface models, allowing upscaling experiments and development of sub‐grid parameterizations without necessitating complex high‐resolution models.

Retrieval of Snow Depth over Arctic Sea Ice Using a Deep Neural Network

The accurate knowledge of spatial and temporal variations of snow depth over sea ice in the Arctic basin is important for understanding the Arctic energy budget and retrieving sea ice thickness from satellite altimetry. In this study, we develop and validate a new method for retrieving snow depth over Arctic sea ice from brightness temperatures at different frequencies measured by passive microwave radiometers. We construct an ensemble-based deep neural network and use snow depth measured by sea ice mass balance buoys to train the network. First, the accuracy of the retrieved snow depth is validated with observations. The results show the derived snow depth is in good agreement with the observations, in terms of correlation, bias, root mean square error, and probability distribution. Our ensemble-based deep neural network can be used to extend the snow depth retrieval from first-year sea ice (FYI) to multi-year sea ice (MYI), as well as during the melting period. Second, the consistency and discrepancy of snow depth in the Arctic basin between our retrieval using the ensemble-based deep neural network and two other available retrievals using the empirical regression are examined. The results suggest that our snow depth retrieval outperforms these data sets.

Investigating the Temporal and Spatial Variation of Snow Depth over Mountainous Region of Zagros, Iran, for the Period 1979-2010

Investigating the Temporal and Spatial Variation of Snow Depth over Mountainous Region of Zagros, Iran, for the Period 1979-2010

Modeling the Snow Depth Variability With a High‐Resolution Lidar Data Set and Nonlinear Terrain Dependency

AbstractIn the mountains of Norway, snow depth (SD) is highly variable due to strong winds and open terrain. To investigate snow conditions on one of Europe's largest mountain plateaus, Hardangervidda, we conducted snow measurement campaigns in spring 2008 and 2009 using airborne lidar scanning at the approximate time of annual snow maximum (mid‐April). From 658 empirical distributions of SD at Hardangervidda, each comprised about 4,000 SD values sampled from a grid cell of 0.5 km2, quantitative tests have shown that the gamma distribution is a better fit for SD than the normal and log‐normal distributions. When aggregating snow and terrain data from 10 × 10 m to 0.5 km2, we find that the standard deviation of the terrain parameter squared slope, land cover, and the mean SD are highly correlated (0.7, 0.52, and 0.89) to the standard deviation of SD. A model for SD variance is proposed that, in addition to addressing the dependencies between the variability of SD and the terrain characteristics, also takes into account the observed nonlinear relationship between the mean and the standard deviation of SD. When validated against observed SD variance retrieved from the same area, the model explains 81–83% of the observed variance for spatial scales of 0.5 and 5.1 km2, which compares favorably to previous models. The model parameters can be determined from a GIS analysis of a detailed digital terrain and land cover model and will hence not increase the number of calibration parameters when implemented in environmental models.

Canopy and Terrain Interactions Affecting Snowpack Spatial Patterns in the Sierra Nevada of California

AbstractAirborne light detection and ranging is an emerging measurement tool for snowpack estimation, and data are now emerging to better assess multiscale snow depth patterns. We used airborne light detection and ranging measurements from four sites in the southern Sierra Nevada to determine how snow depth varies with canopy structure and the interactions between canopies and terrain. We processed the point clouds into snow depth rasters at 0.5×0.5‐m2 resolution and performed statistical analysis on the processed snow depth data, terrain attributes, and vegetation attributes, including the individual tree bole locations, canopy crown area, and canopy height. We studied the snow depth at such a fine scale due in part to the spatial heterogeneity introduced by canopy interception and enhanced melting caused by tree trunks in forested areas. We found that the dominant direction of a tree well, the area around the tree bole that has shallower snowpack, is correlated with the local aspect of the terrain, and the gradient of the snow surface in a tree well is correlated with the tree's crown area. The regression‐tree based XGBoost model was fitted with the topographic variables and canopy variables, and about 71% of snow depth variability can be explained by the model.

Spatio-temporal variability and seasonal dynamics of snow cover regime in Estonia

Climate warming influences highly on snow cover regime in the midlatitudes. Snow cover conditions, in turn, affect human activity very much. The aim of this study was (a) to analyse spatial and temporal variability of snow cover duration, (b) to analyse spatial and temporal variability of the start and end dates of the period with the permanent snow cover, (c) to describe spatial, temporal and seasonal variability of median and maximum snow depth in Estonia and (d) to determine the presence of long-term changes and trends in these parameters during the period 1950/51–2015/16. Time series of daily snow depth at 22 stations for that period were processed in order to obtain reliable estimates of changes in the snow regime. Snow cover data are non-normally distributed, therefore, median and quartile range were used to describe the mean state and variability of snow cover. Only these dates were included into the analysis when snow cover was observed at least on 50% of days in the time series. Trend analysis was made using the Mann-Kendall test and trend values were found using the Theil-Sen’s method. A large spatio-temporal variability of snow cover duration was found. The median number of days with snow cover at the 22 stations was 112, varying between 61 and 130 days. In the coastal regions of Estonia and especially on the western coast of Saaremaa Island snow cover duration has been much lower than in the continental part. The longest snow cover period is observed on uplands in south-eastern and north-eastern Estonia. It was found that, in the average, the period with the permanent snow cover in the continental Estonia begins on 19 December and ends on 18 March. There was a negative trend in snow cover duration due to the earlier snow melting in spring at the majority of stations. The end date of the permanent snow cover has shifted earlier by 10–30 days in 66 years and its duration has decreased accordingly. The maximum snow depth has been recorded on uplands of south-eastern Estonia with the median value 38 cm. There was a decreasing multiannual trend in snow depth from January to the end of March. Changes in snow depth were largest in the end of February and in March when the trend was statistically significant. In the average, snow depth has decreased by 0.5–1.5 cm per decade, i.e. by 2–9 cm throughout the whole study period.

Snow depth variability in the Northern Hemisphere mountains observed from space

Accurate snow depth observations are critical to assess water resources. More than a billion people rely on water from snow, most of which originates in the Northern Hemisphere mountain ranges. Yet, remote sensing observations of mountain snow depth are still lacking at the large scale. Here, we show the ability of Sentinel-1 to map snow depth in the Northern Hemisphere mountains at 1 km² resolution using an empirical change detection approach. An evaluation with measurements from ~4000 sites and reanalysis data demonstrates that the Sentinel-1 retrievals capture the spatial variability between and within mountain ranges, as well as their inter-annual differences. This is showcased with the contrasting snow depths between 2017 and 2018 in the US Sierra Nevada and European Alps. With Sentinel-1 continuity ensured until 2030 and likely beyond, these findings lay a foundation for quantifying the long-term vulnerability of mountain snow-water resources to climate change.

Snow climatology for the mountains in the Iberian Peninsula using satellite imagery and simulations with dynamically downscaled reanalysis data

AbstractThe presence of a seasonal snowpack determines the hydrology, geomorphology and ecology of wide parts of the Iberian Peninsula, with strong implications for the economy, transport and risk management. Thus, reliable information on snow is necessary from a scientific and operational point of view. This is the case of the Iberian Peninsula where, lack of observation has impeded proper analysis of snowpack duration, magnitude and interannual variability. In this study, we present the first snow climatology of the entire Iberian Peninsula. The scarcity of in situ observations has been overcome, using a newly developed remote sensing snow database from MODIS satellite sensors for the period 2000–2014 and a physically based snow model (Factorial Snow Model—FSM), driven by a regional atmospheric model (Weather Research and Forecast model—WRF) over the Iberian Peninsula for the period 1980–2014. The snowpack of the main mountain areas (Pyrenees, Cantabrian, Central, Iberian range and Sierra Nevada) are described, estimated from the generated databases. The information has been processed using a k‐means cluster algorithm, looking for similarities in snow indices at different elevation bands. Results show four different types of snowpack in terms of depth, duration and interannual variability, lying over different elevation bands in the different ranges, proving the variability of the snowpack over Iberia. Analyses reveal areas characterized by ephemeral snowpacks, while in some sectors snowpack lasts, on average, 198 days per year with 3.02 m of peak snow depth. The coefficient of variation of interannual peak snow depth oscillated between 35.2 and 162.4%. All the analysed indices show that at common elevations the Cantabrian range and the Pyrenees host the deepest and longest lasting snowpacks, followed by the Central and Iberian ranges. The Sierra Nevada exhibits the shortest, shallowest snowpack and more year‐to‐year variability.

The Consistency of SSM/I vs. SSMIS and the Influence on Snow Cover Detection and Snow Depth Estimation over China

The long-term variations in snow depth are important in hydrological, meteorological, and ecological implications and climatological studies. The series of Special Sensor Microwave/Imager (SSM/I) and Special Sensor Microwave Imager Sounder (SSMIS) instruments onboard the Defense Meteorological Satellite Program (DMSP) platforms has provided a consistent 30+ year data record of global observations that is well-suited for the estimation of snow cover, snow depth, and snow water equivalent (SWE). To maximize the use of this continuous microwave observation dataset in long-term snow analysis and obtain an objective result, consistency among the SSM/I and SSMIS sensors is required. In this paper, we evaluated the consistency between the SSM/I and SSMIS concerning the observed brightness temperature (Tb) and the retrieved snow cover area and snow depth from January 2007 to December 2008, where the F13 SSM/I and the F17 SSMIS overlapped. Results showed that Tb bias at 19 GHz spans from −2 to −3 K in snow winter seasons, and from −4 to −5 K in non-snow seasons. There is a slight Tb bias at 37 GHz from −2 to 2 K, regardless of season. For 85 (91) GHz, the bias presents some uncertainty from the scattering effect of the snowpack and atmospheric emission. The overall consistency between SSM/I and SSMIS with respect to snow cover detection is between 80% and 100%, which will result in a maximum snow cover area difference of 25 × 104 km2 in China. The inconsistency in Tb between SSM/I and SSMIS can result in a −2 and −0.67 cm snow depth bias for the dual-channel and multichannel algorithms, respectively. SSMIS tends to yield lower snow depth estimates than SSM/I. Moreover, there are notable bias differences between SSM/I- and SSMIS-estimated snow depths in the tundra and taiga snow classes. Our results indicate the importance of considering the Tb bias in microwave snow cover detection and snow depth retrieval and point out that, due to the sensitivity of bias to seasons, it is better to do the intercalibration with a focus on snow-covered winter seasons. Otherwise, the bias in summer will disturb the calibration coefficients and introduce more error into the snow retrievals if the seasonal difference is not carefully evaluated and separated.

Roots Mediate the Effects of Snowpack Decline on Soil Bacteria, Fungi, and Nitrogen Cycling in a Northern Hardwood Forest.

Rising winter air temperature will reduce snow depth and duration over the next century in northern hardwood forests. Reductions in snow depth may affect soil bacteria and fungi directly, but also affect soil microbes indirectly through effects of snowpack loss on plant roots. We incubated root exclusion and root ingrowth cores across a winter climate-elevation gradient in a northern hardwood forest for 29 months to identify direct (i.e., winter snow-mediated) and indirect (i.e., root-mediated) effects of winter snowpack decline on soil bacterial and fungal communities, as well as on potential nitrification and net N mineralization rates. Both winter snowpack decline and root exclusion increased bacterial richness and phylogenetic diversity. Variation in bacterial community composition was best explained by differences in winter snow depth or soil frost across elevation. Root ingrowth had a positive effect on the relative abundance of several bacterial taxonomic orders (e.g., Acidobacterales and Actinomycetales). Nominally saprotrophic (e.g., Saccharomycetales and Mucorales) or mycorrhizal (e.g., Helotiales, Russalales, Thelephorales) fungal taxonomic orders were also affected by both root ingrowth and snow depth variation. However, when grouped together, the relative abundance of saprotrophic fungi, arbuscular mycorrhizal fungi, and ectomycorrhizal fungi were not affected by root ingrowth or snow depth, suggesting that traits in addition to trophic mode will mediate fungal community responses to snowpack decline in northern hardwood forests. Potential soil nitrification rates were positively related to ammonia-oxidizing bacteria and archaea abundance (e.g., Nitrospirales, Nitrosomondales, Nitrosphaerales). Rates of N mineralization were positively and negatively correlated with ectomycorrhizal and saprotrophic fungi, respectively, and these relationships were mediated by root exclusion. The results from this study suggest that a declining winter snowpack and its effect on plant roots each have direct effects on the diversity and abundance of soil bacteria and fungal communities that interact to determine rates of soil N cycling in northern hardwood forests.

Changing snow depth in the Great Lakes basin (USA): Implications and trends

Abstract In the Great Lakes basin of North America, snow plays a critical role in the regional hydroclimate, where snow ablation events can serve both as a resource and a hazard. The frequency and magnitude of an ablation event is governed by the availability of meteorological conditions to ablate snow, and the physical presence of snow to be ablated. While the meteorological conditions leading to ablation have been well documented, examining changes in atmospheric conditions alone have been unable to completely explain observed ablation trends. As such, this study applies a gridded snow depth dataset to evaluate snow depth variability, while speaking to the implications of such variability on ablation frequency and water resources. Snow cover is present in the basin from November – April, with the more northerly regions (Lake Superior and Lake Huron basins) exhibiting a deeper and more seasonally-persistent snowpack. Seasonal basin-wide snow depth decreased by approximately 25% from 1960 to 2009, with some of the most significant decreases occurring north of Lake Superior. Surface air temperatures are negatively associated snow depth, and warming temperatures are likely contributing to snow depth declines. These regional decreases in snow depth spatially corroborate previously observed decreases in the frequency of ablation events in the basin, and highlight the importance of examining both snow cover and meteorological conditions when seeking to explain snow ablation variability. The results from this study can be applied to inform water resource management decisions in the region.

Modeling Subgrid Variability of Snow Depth Using the Fokker‐Planck Equation Approach

AbstractA physically based subgrid variability model for snow process using the Fokker‐Planck equation (FPE) approach was proposed. This FPE can express the evolution of the probability density function (PDF) of snow depth within a finite area, possibly a grid cell of distributed models or a small basin, whose shape can be irregular. The main advantage of this approach is that it does not rely on a given PDF but dynamically computes the PDF through an advection‐diffusion‐type equation, the FPE, which was derived from point‐scale process‐based governing equations. Snow depth was treated as a random variable, while the snow redistribution and snowmelt rate were treated as the sources of stochasticity. The main challenge in solving this FPE is evaluating the time‐space covariances appearing in the diffusion coefficient. In this study, approximations to evaluate the covariance terms, accounting for snowmelt and snow redistribution, were proposed. The simulated results of the FPE model were validated by the measured time series of snow depth at one site and the spatial distributions of snow depth measured by ground penetrating radar and airborne light detection and ranging (Lidar). It was shown that the point‐observed snow depth fell within the simulated range during most of the 2‐year study period. The simulated PDFs of snow depth within the study area were similar to the observed PDFs of snow depth by ground penetrating radar and Lidar. In summary, these results demonstrate the efficacy of the proposed FPE model representing the subgrid variability of snow depth.

Understanding subgrid variability of snow depth at 1‐km scale using Lidar measurements

AbstractIt is well known that snow plays an important role in land surface energy balance; however, modelling the subgrid variability of snow is still a challenge in large‐scale hydrological and land surface models. High‐resolution snow depth data and statistical methods can reveal some characteristics of the subgrid variability of snow depth, which can be useful in developing models for representing such subgrid variability. In this study, snow depth was measured by airborne Lidar at 0.5‐m resolution over two mountainous areas in south‐western Wyoming, Snowy Range and Laramie Range. To characterize subgrid snow depth spatial distribution, measured snow depth data of these two areas were meshed into 284 grids of 1‐km × 1‐km. Also, nine representative grids of 1‐km × 1‐km were selected for detailed analyses on the geostatistical structure and probability density function of snow depth. It was verified that land cover is one of the important factors controlling spatial variability of snow depth at the 1‐km scale. Probability density functions of snow depth tend to be Gaussian distributions in the forest areas. However, they are eventually skewed as non‐Gaussian distribution, largely due to the no‐snow areas effect, mainly caused by snow redistribution and snow melt. Our findings show the characteristics of subgrid variability of snow depth and clarify the potential factors that need to be considered in modelling subgrid variability of snow depth.

The smoothing of landscapes during snowfall with no wind

ABSTRACTEvery winter, snowy landscapes are smoothed by snow deposition in calm conditions (no wind). In this study, we investigated how vertically falling snow attenuates topographic relief at horizontal scales less than or approximately equal to snow depth (e.g., 0.1–10 m). In a set of three experiments under natural snowfall, we observed the particle-scale mechanisms by which smoothing is achieved, and we examined the cumulative effect at the snowpack scale. The experiments consisted of (a) a strobe-light box for tracking the trajectories of snowflakes at deposition, (b) allowing snow to fall through a narrow gap (40 mm) and examining snow accumulation above and below the gap, and (c) allowing snow to accumulate over a set of artificial surfaces. At the particle scale, we observed mechanisms enhancing (bouncing, rolling, ejection, breakage, creep, metamorphism) and retarding (interlocking, cohesion, adhesion, sintering) the rate of smoothing. The cumulative effect of these mechanisms is found to be driven by snowpack surface curvature, introducing a directional bias in the lateral transport of snow particles. Our findings suggest that better quantification of the mechanisms behind smoothing by snow could provide insights into the evolution of snow depth variability, and snow-vegetation interactions.