1. Introduction
Several of the physical properties of snow, such as its surface albedo and thermal conductivity, are instrumental in modulating heat exchanges between the surface and the atmosphere (Cohen and Rind 1991; Xu and Dirmeyer 2011). Other properties such as grain size distribution and the presence of ice layers within the snowpack have important hydrological implications, as they govern liquid water transport through the snowpack (Wever et al. 2016).
To simulate the changes in snow-cover properties over time, a wide range of snow models of various complexities have been developed and integrated into global climate models (GCMs), numerical weather prediction (NWP) systems, hydrological prediction (HP) systems, and land surface models (LSMs). The complexity of a snow model depends on the number of layers used to represent the snowpack layering as well as on the parameterization of their properties and internal processes (Etchevers et al. 2004; Slater et al. 2001). Based on Etchevers et al. (2004) and neglecting pseudoenergy balance models, snow model complexity can be divided into three categories: simple models (using a single layer), complex models (using multiple layers), and very complex models (which include internal physical processes of the snowpack).
The application of complex and very complex land surface models for climate studies has been shown to improve the description of snow at different spatial (local to global) and temporal scales (Arduini et al. 2019; Burke et al. 2013; Dutra et al. 2010; Jin et al. 1999). However, single-layer snow models are still commonly used in NWP systems and atmospheric reanalyses (Balsamo et al. 2015; Dee et al. 2011; Dutra et al. 2010; Bélair et al. 2003b) since they are not computationally expensive, rely on relatively few parameters, and can efficiently capture first-order processes (Etchevers et al. 2004). From the perspective of hydrological forecasting, a number of studies have assessed the ability of models to predict streamflow by including a wide variety of empirical and physically based snow schemes (Magnusson et al. 2015; Kumar et al. 2013; Zeinivand and De Smedt 2009; Franz et al. 2008). These studies show divergent results. For example, one study reported that increasing the snow model complexity improves runoff simulations (Warscher et al. 2013) while other studies did not find such a relationship (Franz et al. 2008; Lehning et al. 2006; Zappa et al. 2003).
Environment and Climate Change Canada (ECCC) has recently developed the Soil, Vegetation, and Snow (SVS) LSM to improve the representation of land surface processes in their NWP and HP systems (Alavi et al. 2016; Husain et al. 2016). The SVS LSM introduces two distinct snowpacks (one over bare ground/short vegetation and a second under high vegetation) with separate energy budgets for each one. This distinguishes its representation of snow from its predecessor, an operational version of the Interaction Soil Biosphere Atmosphere (ISBA) LSM (Bélair et al. 2003a,b). Both SVS and ISBA are classified as simple models according to the classification system proposed by Etchevers et al. (2004).
SVS has mostly been evaluated in North American regions (Alavi et al. 2016; Husain et al. 2016), and in general, performs better than its predecessor in the warm season. This LSM has been successfully implemented as part of the Global Environmental Multiscale hydrological modeling platform (GEM-Hydro) to produce runoff simulations in the Lake Ontario basin (Gaborit et al. 2017) and hindcasts of the major June 2013 flood in Alberta (Canada) (Vionnet et al. 2020). More recently, point-scale evaluations showed comparable results relative to field observations and more complex models for Mediterranean and temperate climates in the United States (Maheu et al. 2018). These evaluations also yielded accurate results for energy fluxes and soil moisture in arid, Mediterranean, and tropical climates (Leonardini et al. 2020). While previous studies have assessed the general performance of SVS and contrasted it with early versions of ISBA, there has been no detailed evaluation of the SVS snow model. The need for an assessment of the SVS snow scheme is urgent because this LSM will soon be used operationally by ECCC in the National Surface and River Prediction System (NSRPS) and in the Global Deterministic Prediction System (GDPS).
The main goal of this study is to assess to what degree the SVS snow-cover component is able to adequately represent the main characteristics of snowpacks in a multiyear and multisite context, when running SVS in point-scale and stand-alone modes. The evaluation was based on a recently released ESM-SnowMIP dataset (Ménard et al. 2019). Moreover, runs from the detailed Crocus snow-cover model (Vionnet et al. 2012; Brun et al. 1992, 1989) generated in the context of ESM-SnowMIP (Ménard et al. 2021) were also used to put SVS outputs into perspective with this much more detailed snow model.
The SVS snow model is detailed in section 2, followed by the ESM-SnowMIP dataset and the experimental setup in section 3. The performance of the snow model is presented in section 4. Finally, several sensitivity tests, along with a list of model limitations, are summarized in section 5.
2. The SVS model
In SVS, each grid cell can be categorized as a combination of the following four tiling components: 1) bare ground, 2) low and high vegetation, 3) snow over bare ground and low vegetation, and 4) snow under high vegetation. The model considers a single-layer vegetation canopy, a single-layer snowpack, and a user-defined multilayer soil column, and is hence considered a simple snow model according to the classification from Etchevers et al. (2004). A detailed description of the vegetation and soil processes is provided in Alavi et al. (2016) and Husain et al. (2016). Furthermore, the model structure and its major modules are depicted in Fig. 1 in Husain et al. (2016). This section presents detailed description of snow components.
Two different snowpacks are considered in SVS: one covering bare ground and low vegetation (subscript X = SN) and the other located under high vegetation (subscript X = SVH). While each snowpack evolves separately, their physical equations are for the most part identical. Therefore, unless stated otherwise, in the following equations subscript X is used for the two types of snowpacks.
a. Snow temperature and thermal properties
b. Snow mass and liquid water in the snowpack
c. Snow density
Note that hX along with the factors 20 470 and 67.3 in Eq. (26) are expressed in centimeters. An error in the units of the correspondent factors was identified in Eqs. (21) and (22) in Bélair et al. (2003b).
d. Snow albedo
The SVS reference simulations that are presented later in the paper are based on the description presented in this section, with the exception of freezing snow, which is set to zero (freezX = 0) in the experiments.
3. Methodology
a. ESM-SnowMIP dataset
This study used the meteorological and evaluation datasets from the Earth System Model–Snow Model Intercomparison Project (ESM-SnowMIP), an internationally coordinated modeling effort to investigate snow schemes (Krinner et al. 2018). The dataset is described in detail by Ménard et al. (2019) and some of the information is repeated here for convenience. The dataset can be found on the PANGAEA repository (https://doi.pangaea.de/10.1594/PANGAEA.897575). It includes postprocessed and quality-controlled data from 10 sites, totaling 136 years of in situ measurements at 60-min time steps. The measurements are grouped into meteorological forcing and evaluation datasets.
The main characteristics of the study sites are provided in Table 1. Three out of 10 sites are located in open terrain, while 4 are situated in a clearing and 3 are forested. While open sites may experience a lot of wind-induced redistribution [e.g., Senator Beck as described in Landry et al. (2014)], clearings often do not because they are surrounded by trees. In the rest of this paper, for convenience, we use the term “open” for all the sites which are not covered by forest. According to the snow-cover classification (SCC) by Ménard et al. (2019), of the 10 study sites, 5 sites are classified as alpine, 4 as taiga, and 1 as maritime (Table 1). Most of the open sites are covered in short grass, while SNB is covered in alpine tundra and SOD in short heather and lichen. The mean canopy heights at OAS, OBS, and OJP reach 21, 12, and 14 m, respectively. These three sites are located in the Canadian boreal biome and are part of the Boreal Ecosystem Research and Monitoring Sites (BERMS; see Bartlett et al. 2006).
Main characteristics of the study sites.
Detailed descriptions of each site can be found in Ménard et al. (2019). Figure 1 provides an overview of the prevailing climatology for each location. Most of the sites register snowfall from September to June except for WFJ, which is exposed to year-round snowfall (Fig. 1b). SOD and the three BERMS sites (OAS, OBS, OJP) accumulate the lowest levels of snowfall. At these sites, rainfall declines progressively from September to December and becomes more prevalent between March and May. CDP registers high rates of rainfall in winter, while RME exhibits the driest conditions. The SOD and the BERMS sites exhibit the lowest air temperatures between December and February, with mean values close to −15°C. WFJ, SNB, and SWA manifest the longest periods with negative temperatures, between 0° and −5°C from October to April. SAP, CDP, and RME are the warmest locations, recording air temperatures close to 0°C from November to March (Fig. 1c). All sites record wind speeds between 1 and 2 m s−1, except for SNB, which has reported values of around 5 m s−1 from December to April (Fig. 1d). SWA and SNB exhibit the highest monthly mean solar radiation of between 100 and 150 W m−2 from October to February (Fig. 1e), in contrast to SOD, which reports the lowest values of between 0 and 50 W m−2 during the same period (Fig. 1e). For all sites, infrared radiation ranges from 200 to 300 W m−2 from November to April (Fig. 1f) and specific humidity oscillates between 0.001 and 0.004 kg kg−1 over the same months (Fig. 1g).
Monthly mean (a) rainfall, (b) snowfall, (c) air temperature, (d) wind speed, (e) solar radiation, (f) infrared radiation, (g) specific humidity, and (h) atmospheric pressure at the ESM-SnowMIP sites. Note that wind speed measurement heights differ between sites.
Citation: Journal of Hydrometeorology 22, 6; 10.1175/JHM-D-20-0249.1
As previously mentioned, the evaluation subset includes observations of snow water equivalent (SWE), snow depth, surface albedo, surface temperature, and soil temperature. SWE and snow depth were manually measured at all sites on a weekly to monthly basis, in a predefined area around the meteorological stations. Daily snow depth observations were automatically made and reported for all sites, but daily automatic SWE observations were only available for three sites. Daily effective albedo observations were available for five locations. The calculations were performed as the fraction based on the incident and the reflected observed solar radiation using the method described in Morin et al. (2012). To avoid the effect of snow-free surfaces, the observed albedo was constrained to values of observed albedo and snow depth greater than 0.5 and 0.1 m, respectively. Surface temperature observations were available for the same five sites, which were recorded based on hourly measurements of re-emitted longwave radiation, except at SNB where infrared temperature sensors were used to determine surface temperature.
b. Experimental setup
In this study, point-scale experiments were based on the stand-alone implementation of SVS in the Modélisation Environnementale communautaire–Surface Hydrology (MESH) modeling platform, which is ECCC’s community environmental modeling system (Pietroniro et al. 2007). MESH allows different LSMs and hydrologic routing models to coexist within the same framework. A comprehensive description of MESH can be found at https://wiki.usask.ca/display/MESH. The SVS code and the configuration files for the 10 ESM-SnowMIP experiments are available at https://doi.org/10.5281/zenodo.4568309. A similar setup was used in a previous study by Leonardini et al. (2020).
The forcing data that are needed to drive the SVS model are air temperature (°C), solid and liquid precipitation rates (kg m−2 s−1), downward shortwave radiation (W m−2), downward longwave radiation (W m−2), wind speed (m s−1), surface pressure (Pa), and specific humidity (kg kg−1). Data on liquid and solid precipitation are provided in the ESM-SnowMIP forcing dataset. A number of other parameters that describe vegetation type, fraction of vegetation, soil type, and height measurements for wind speed, air temperature, and humidity are set based on the available literature for each location (Table 2). These parameters are kept constant for each site once chosen. Supplemental parameters associated with each type of vegetation are provided in the SVS lookup tables.
SVS parameters used in the experiments.
All simulations used a 5-min time step, the same duration applied in regular operational use. To remain consistent with the field observations, model outputs were assessed at 1-h time steps. The average was calculated for the simulated albedo values between 0900 and 1500 local time to be consistent with the criteria used for the daily albedo observations. For each site, the data from the first year were replicated five times for the model spinup. We run a continuous model integration starting with no snow on the ground, typical values for the soil temperature and moisture. Snowpack covering bare ground and low vegetation areas was used for evaluations at CDP, RME, SNB, SWA, WFJ, SAP, and SOD sites, whereas subcanopy snowpack was used for BERMS sites evaluations.
SVS performance was evaluated using the following metrics: Nash–Sutcliffe efficiency (NSE), percent bias (PBias), Pearson correlation coefficient (r), root-mean-square error (RMSE), and bias (bias) (Anctil and Ramos 2019; Nash and Sutcliffe 1970). NSE has an optimal value of 1, while a value less than 0 indicates that the average of the observations is a better predictor than the model’s result. PBias quantifies the tendency of the simulated value’s average to be either larger or smaller than its observed counterpart. The correlation coefficient r determines the degree of correlation between the simulation and observation. The RMSE is a quadratic scoring rule that attributes more weight to large errors, as the errors are squared prior to being averaged. The bias metric determines the averaged difference between simulated and observed values. While the first three metrics have no units, the latter two provide information on the errors using the same units as the variable been examined.
Winters were divided into the accumulation and the ablation periods. To do this, for a given site and year, the occurrence of peak observed SWE was identified first. The entire period with snow cover before this time was defined as accumulation, and the period after, as ablation. This approach accounts for the interannual and the intersite variability, and works well for high-elevation (WFJ) or high-latitude (SOD) sites. However, some limitation may be found due to midwinter ablation events, which are frequent at sites such as CDP or SAP, even during the period identified as accumulation.
c. Crocus snow model simulations
Crocus is a detailed snowpack model (Vionnet et al. 2012; Brun et al. 1992, 1989) that has contributed to many model intercomparison projects. As a strategy to better interpret SVS results, outputs from the Crocus snowpack model were also considered in some specific cases in this research. Interpretations will not focus on comparing SVS and Crocus, but rather on using Crocus as an additional benchmark to better decipher SVS behavior. Crocus snowpack simulations were generated for the ESM-SnowMIP project (Ménard et al. 2021) and were provided by Matthieu Lafaysse from the Centre National de Recherches Météorologiques (CNRM; http://www.umr-cnrm.fr). In these simulations, the same meteorological forcing was applied to both Crocus and SVS, and the same experimental setup was used.
4. Results
a. Simulations at forested sites
Results from forested sites are substantially different from those at open sites (Fig. 2). This is likely due to the uncertainty associated with the absence of a snow interception scheme in SVS. So, in this section we summarize the results of forested sites only.
Diagram of SVS performance for SWE, snow density (density), and depth for the accumulation and ablation periods at open (orange text) and forested (green text) sites. (a) NSE, (b) PBias, and (c) r scores are also presented.
Citation: Journal of Hydrometeorology 22, 6; 10.1175/JHM-D-20-0249.1
According to Pomeroy et al. (1998), 25%–45% of annual snowfall can be lost to sublimation of intercepted snow, depending on the local conditions. This fact is reflected in SWE values at OBS and OJP, which were considerably overestimated (PBias = 33.1% and 65.4%), mainly during the ablation period (Fig. 2b). In contrast, OAS showed a lower PBias score for SWE (PBias = 7.8%). The SVS model likely performed better at OAS than at the other two forested sites because this site is dominated by leafless deciduous trees, for which interception plays a much more modest role than at sites with conifers (Bartlett et al. 2006).
Snow density simulations at forested sites resulted in poor performance levels (NSE = 0.15, on average), characterized by an average overestimation of 15.4%. This may be partially due also to the absence of an interception scheme in the model, which can lead to unrealistic snow depths and increase the value of maximum snow density (ρXmax) given in Eq. (26). Simulations may also be influenced by the lack of temperature-dependent effects in the snow compaction scheme. It is interesting to note that the snow density simulations did not perform as badly for OAS, despite the site’s cold winter temperatures. These results may also be associated with lower interception rates due to the low canopy density at this site.
Snow depth simulations at forested sites were better than for SWE. This is due to a compensation effect, as reflected by a positive PBias (Fig. 2b). The only exception was at the OAS site, where density simulations performed well.
As it is evident that the forested sites require further evaluation mainly associated with the implementation of the interception scheme, in the following sections the analyses will focus on the open sites.
b. SWE, density, and depth
Table 3 presents the scores for SWE, snow density, and depth. In addition, performances in the accumulation and ablation period are presented in Fig. 2.
Scores for modeled SWE, snow density, and depth compared with manual observations for the entire study period.
For the open sites, SVS performed well on SWE simulations. The best performance was achieved at the RME site (NSE = 0.94). SNB is a high alpine site that neighbors SWA (3714 and 3371 m MSL, respectively), with both sites experiencing similar meteorological forcing. However, the wind speed is stronger at SNB (mean of 5 m s−1 between December and April) than at SWA (1 m s−1 for the same period), which may explain the SWE deficiencies observed for this site, as blowing snow sublimation and/or lateral snow redistribution may occur. SVS does not take into account these two phenomena. Blowing snow sublimation rates are higher than those typically found over a static snow surface because the snow crystals are lifted by the wind and blown downfield as a result of the greater surface area to mass ratio (Schmidt 1972; Pomeroy 1989). The snow cover at the SNB is likely to be affected by blowing snow and its associated sublimation (Landry et al. 2014). Among all alpine sites, SVS simulations for SWE at the CDP site had the lowest performance (NSE = 0.46). The large SWE overestimation at CDP, 27.6%, may be explained in part by insufficient melting, which is accentuated in the ablation period (Fig. 2a) during the snow melting period. A similar overestimation was observed at SAP maritime location. Contrary to the alpine sites, the SWE at the SOD taiga site was accurately represented (NSE = 0.89), which is consistent with the other findings that consider less frequent midwinter melting events due to the prevalence of subzero temperatures at these sites.
The snow density simulations at open sites were accurately represented (NSE = 0.75 on average), resulting in a slight underestimation (PBias = −1.5% on average) and high correlation (r = 0.87 on average). Among the open sites, SOD and SAP exhibited the lowest performances, with an NSE of 0.43 and 0.55, respectively. We suspect the overestimation at the SOD site is due to the fact that the snow compaction scheme in the SVS model is not temperature-dependent, unlike more complex snow models. For example, Crocus takes into account the dependence of viscosity on temperature and liquid water content, limiting settling under cold temperatures and decrease snow viscosity in the presence of liquid water (Vionnet et al. 2012; Dawson et al. 2017). The same may also be true for SAP, but the effect of liquid water on snow settling may also play a role.
The snow depth simulation is derived from Eq. (28). SVS simulations showed similar results to SWE, since density is predicted well at these sites (Fig. 2a). For example, the fact that NSE in accumulation period at SWA is high for SWE and snow depth but low for snow density suggests that snow depth may be right for the wrong reasons, i.e., if density values were correct, snow depth may not be.
Figure 3 illustrates the interannual variability of two consecutive and contrasting winters at the CDP site. Automatic and manual observations were compared with the simulations of SVS and Crocus. Automatic observations of SWE and snow depth differed from manual observations due to the strategy used to collect the manual samples, as reported by Ménard et al. (2019). In general, Crocus outperformed SVS. SVS simulations for SWE were overestimated for both winters, with the largest discrepancies occurring in 2007/08. During winter of that year, SVS produced a maximum SWE of 560 kg m−2 in March 2008. The SWE peaked at 500 kg m−2 when using Crocus, which was closer to the observed value of 480 kg m−2. During April and May of the following months, the similar SWE melt rates observed in the SVS simulations and the observations resulted in the SVS simulating a snowpack disappearance that was delayed by three weeks. This delay explains the low scores simulated for SWE at the CDP site during the April–May period and for the whole 1994–2014 period, as reported in Fig. 2a. For bulk density, SVS simulations showed good accuracy over those two winters. However, SVS was less sensitive to the abrupt changes in snow bulk density for the shallower snowpacks seen in 2006/07 due to uncertainties with the description of maximum density as a function of snow depth for low snow depth values. Compared to SVS, Crocus is better able to represent the rapid settling of freshly fallen snow. SVS was correlated with the observations but its magnitude was biased high.
Comparisons between simulated and observed (a) SWE, (b) snow density, and (c) depth for winters 2006/07 and 2007/08 at Col de Porte (CDP). Results from SVS (blue line), Crocus (red line), automatic observations (black line), and manual observations (black squared dots) are shown.
Citation: Journal of Hydrometeorology 22, 6; 10.1175/JHM-D-20-0249.1
c. Accumulation and ablation
Following the approach outlined in Quéno et al. (2016), a complementary evaluation was conducted by looking at daily snow depth variations. This allowed us to circumvent cumulative errors and to better identify the processes that are responsible for errors in snow depth simulations. The snow depth variation for a specific day, ΔSD, is defined as the difference between the current snow depth minus that from the previous day. ΔSD was computed for the SVS and Crocus simulations for the period in which automatic observations were available. The ΔSD values were discarded if SD = 0 for two consecutive days or if SD observations were not available.
Figure 4 illustrates the categorical positive sums of ΔSD, which were normalized by the maximum cumulated ΔSD from the observations. They characterize the total sum of the elements in a given category and evaluate the ability of SVS to represent the total increase in snow depth in a particular range. Figure 4 shows that SVS strongly underestimated the quantities of high snow depth accumulation (ΔSD categories [10, 20 cm] and [20, 50 cm] categories for open sites), which is the main contributor to overall snow accumulation. Crocus simulations tended to be closer to the observations in these categories. In these experiments, Crocus and SVS are driven by the same precipitation forcing and used the same parameterization for the density of falling snow. However, as a multilayer snowpack scheme, Crocus is able to create new snow layers with a low snow density in response to a snowfall event. In SVS only a moderate decrease in bulk snow density is possible (as shown in Fig. 3). For a given amount of snow water, this difference will result in a larger depth of new snow in the Crocus simulation. As a consequence, SVS cannot reproduce large snow accumulations as accurately. The same argument indicates there are likely inaccuracies in the way SVS simulates the settling of snow immediately after such events.
Normalized cumulated positive ΔSD by category for SVS (blue), Crocus (red), and observations (black) at all open sites. All values are normalized by the maximum cumulated ΔSD from the observation.
Citation: Journal of Hydrometeorology 22, 6; 10.1175/JHM-D-20-0249.1
Categorical negative sums can be used to represent a decrease in a particular range of snow depth. Negative ΔSD values can result from snow compaction, snow melting, snow sublimation and wind-induced snow erosion. To neglect the effect of snow compaction and focus on melting, Fig. 5 presents the categorical negative sums of ΔSD on melting snow days (MSD), which were normalized by the maximum cumulated ΔSD from the observations. Using the same method as Quéno et al. (2016), MSDs were identified as days during which the snow surface temperature is zero at midday. Only the three open sites presented in Fig. 5 recorded surface temperatures. Here, results from SVS were in close agreement with both the observations and Crocus-simulated values. Thus, as previously highlighted, the settling of freshly fallen snow remains the most plausible explanation for the differences between the models in terms of snow depth.
Normalized cumulated negative ΔSD for melting snow days (MSD) by category for SVS (dashed blue), Crocus (dashed red), and observations (dashed black). All values are normalized by the maximum cumulated ΔSD from the observation.
Citation: Journal of Hydrometeorology 22, 6; 10.1175/JHM-D-20-0249.1
d. Surface temperature
In Fig. 6, SVS daily mean snow surface temperature simulations at five open sites are contrasted with observations from accumulation and ablation winter periods. SVS simulations revealed higher levels of dispersion for temperatures below −10°C, and registered a cold bias ranging from −2.41° to −2.34°C, except at SNB where the bias was 0.59°C on average. This cold bias was mainly caused by an underestimate of nighttime temperatures, as discussed below. Crocus exhibited similar behavior, but with less dispersion compared to SVS.
Quantile–quantile (Q–Q) plots of observed and simulated daily snow surface temperatures (°C) at CDP, SWA, SNB, WFJ, and SAP for the (a)–(e) SVS and (f)–(j) Crocus model accumulation in red and ablation periods in blue. RMSE and bias values are also provided.
Citation: Journal of Hydrometeorology 22, 6; 10.1175/JHM-D-20-0249.1
The seasonal representation of temperatures in Fig. 6 reveals an improvement in the simulations during the accumulation with respect to the ablation period. For example, the RMSE values at the CDP site were 3.51° and 2.42°C, for the accumulation and ablation winter periods, respectively.
To assess differences between simulated and observed surface snow temperatures on a finer temporal scale, plots at the CDP site are illustrated in Fig. 7b for February 2008. Under clear-sky conditions, from 6 to 20 February 2008, SVS systematically underestimated nighttime temperatures. The minimum temperatures simulated by SVS ranged between −21° and −18°C, while the observations were between −16° and −10°C. This cold bias results from a decoupling of the snow surface from the atmosphere in stable atmospheric conditions which affects a large variety of snowpack models (Ménard et al. 2021; Brown et al. 2006; Boone and Etchevers 2001; Slater et al. 2001). Crocus behaved similarly to SVS during this period. From 20 to 26 February, a transition from negative to positive air temperatures was observed. In daytime during this period, SVS tended to accurately simulate a melting temperature of 0°C, while at night, SVS snow temperatures were too cold, reaching temperatures as low as −13°C compared to approximately −8°C from observations. Crocus performed similarly to SVS during the day but showed improved performance at night. Unlike Crocus, SVS does not include the effect of liquid water freezing on surface snow temperature leading to underestimated snow surface temperature. A similar result was reported by Boone and Etchevers (2001) for the one-layer snowpack scheme implemented in ISBA model (Douville et al. 1995).
(a) Hourly air temperatures and (b) and snow surface temperatures for SVS in blue, Crocus in red, and observations in black at the CDP site in February 2008.
Citation: Journal of Hydrometeorology 22, 6; 10.1175/JHM-D-20-0249.1
Snow surface temperatures during snowmelt (April 2008) are illustrated in Fig. 8. From 11 to 28 April, the observed surface temperatures were frequently close to zero. SVS exhibited consistent melting temperatures at midday during this period. At night, SVS still underestimated temperatures due to the absence of liquid water refreezing in the model. For example, from 24 to 28 April, SVS simulated temperatures of around −5°C, while the observed temperatures were close to 0°C. Crocus exhibited a similar nighttime cold bias to SVS, but with negative temperatures that were less varied compared to the observed values.
(a) Hourly air temperatures and (b) and snow surface temperatures for SVS in blue, Crocus in red, and observations in black at the CDP site in April 2008.
Citation: Journal of Hydrometeorology 22, 6; 10.1175/JHM-D-20-0249.1
e. Albedo
Snow albedo quantile–quantile (Q–Q) plots are presented in Fig. 9 for the same five open sites as for snow surface temperature. The results presented in this figure show the following common characteristics for SVS simulations. First, the imposed maximum albedo of 0.8 was too low. The maximum albedo values recorded from observations were between 0.9 and 0.95. Second, low observed albedo values tended to be overrepresented, mainly in the ablation period. This behavior is evidenced by the positive bias of between 0.03 and 0.1 in the ablation period, except at the WFJ site (Figs. 9a–c,e). Crocus represented the observations more accurately than SVS, as illustrated by the RMSE and bias scores (Figs. 9f–j).
Q–Q plots of observed and simulated daily albedo levels at CDP, SWA, SNB, WFJ, and SAP for (a)–(e) SVS and (f)–(j) Crocus for accumulation in red, and ablation periods in blue. RMSE and bias values are also provided.
Citation: Journal of Hydrometeorology 22, 6; 10.1175/JHM-D-20-0249.1
Figure 10 presents the simulated and observed surface snow albedos at the CDP site for winter 2007/08. Daily albedo values simulated by SVS were generally higher than those from observations but lower than Crocus simulations, especially during long melting periods when large differences of up to 0.2 were recorded. Neither SVS nor Crocus were able to simulate the sudden drop in snow albedo observed in mid and late winter. This may be related to deficiencies in the models’ parameterization, in particular the lack of explicit modeling of the effect of light-absorbing impurities on snow albedo (Dumont et al. 2017; Tuzet et al. 2017). The snow albedo results imply that SVS simulates more solar energy absorption by the snowpack than Crocus in midwinter, while the opposite occurs in early and late winter.
Daily average snow surface albedo simulated by SVS and Crocus for winter 2007/08 at the CDP site. Observations are represented by circles.
Citation: Journal of Hydrometeorology 22, 6; 10.1175/JHM-D-20-0249.1
5. Sensitivity tests and model limitations
This section presents the effect of various SVS configurations on SWE simulation sensitivity.
The results from two experiments are illustrated in Fig. 11. Here NoMeltRain experiments disregard melting associated with rain on snow events, whereas Melt evaluates the snowmelt values obtained by replacing the average snowpack temperature with the snow surface temperature in Eq. (13).
Sensitivity experiments for SWE at all sites: reference simulation (REF), with no melt associated with rainfall (NoMeltRain), using the snow surface temperature for melt calculation (Melt). Sensitivity was evaluated on the basis of the NSE, PBias, and r score values.
Citation: Journal of Hydrometeorology 22, 6; 10.1175/JHM-D-20-0249.1
SVS NoMeltRain simulations revealed that rainfall minimally contributed to snowmelt throughout winter at all sites (Figs. 11a–c), which is consistent with prior studies (e.g., Marks et al. 2001, 1998). Note that the definition of the term associated with snowmelt due to rain, used in the previous ECCC LSM (Bélair et al. 2003b), has been identified to be highly sensitive to model time steps, leading to potentially unrealistic estimates of snowmelt during rain on snow events (V. Vionnet 2020, personal communication).
Melt simulations showed improvement in SWE simulation with the exception of sites RME, SWA, and SOD. These results suggest that by substituting the average snowpack temperature with the surface snow temperature in the melting term of the force–restore method improves snowmelt simulations. Typically, during the melt season, the average snow temperature is lower than the simulated snow surface temperature, leading melt rates that are lower than those simulated. Bélair et al. (2003b) and Douville et al. (1995) showed consistent simulation results for SWE when applying the snow surface temperature to the snowmelt term using the force–restore method.
It is worth noting that the SVS model does not consider the following snow processes: canopy snow interception, melt, unloading, and sublimation; blowing snow transport and sublimation; and energy exchanges at the snow–soil interface. To address these limitations, particularly those around heat diffusion in soil, ECCC is currently developing a new version of SVS that will include a more advanced multilayer snowpack scheme coupled with a solver for the heat diffusion in the soil. This new version will include a detailed representation of snow/canopy interactions.
6. Conclusions
This study examined the ability of SVS, a fairly simple one-layer snow model to simulate mean snow properties (SWE, density, snow depth, temperature, and albedo) in open-area and forested conditions. The model was evaluated in point-scale and in stand-alone mode across 10 sites from the ESM-SnowMIP dataset, totaling more than 136 site years of detailed snow observations.
The evaluation revealed that the SVS snow model realistically reproduces SWE, density, and depth for most of the sites in this study. Open sites were accurately represented for SWE (NSE = 0.64), density (NSE = 0.75), and snow depth (NSE = 0.59), while forested sites showed a lower degree of accuracy for the same variables (NSE = −0.40, 0.15, and 0.56). A more detailed analysis of the accumulation and ablation winter periods identified a tendency toward degraded model performance in the ablation period. Two main factors associated with the bias in SWE and snow depth simulations were limitations in the snow settling scheme and the absence of a snow interception scheme.
SVS snow surface temperatures exhibited a cold bias on a daily scale that was associated, for the most part, with an underestimation of nighttime temperatures, likely due to poor representation of the surface energy balance under stable conditions. As a consequence, a decrease was observed in the quality of score values at five open sites (CDP, SWA, SNB, WFJ, and SAP) in the accumulation period (RMSE = 3.4°C on average) in comparison to the ablation period (RMSE = 2.25°C on average). Sensitivity tests revealed partial improvements for SWE simulations when using snow surface temperatures were used instead of average snowpack temperatures in the computation of the melting term in the force–restore method.
Simulations for snow albedo showed a systematic underestimation of maximum albedo values. Errors in the accumulation period (RMSE = 0.06 on average) were lower than the results in the ablation (RMSE = 0.10 on average). This suggests the need for a more advanced and physically based snow albedo scheme within SVS.
This study is the first to extensively assess the SVS snow model’s performance. The results from our analyses are promising, especially considering the simplicity of the model. Future evaluations should use other datasets to constrain energy flux measurements and provide more detailed evaluation of forested. For example, the data presented by Isabelle et al. (2020, 2018) or those from FLUXNET2015 (Pastorello et al. 2020) are good examples of micrometeorological measurements that may be used for future evaluations. Finally, developing a new version of SVS to better represent cold region processes will allow for more in-depth evaluations of the interaction of all the SVS components. Furthermore, it may improve surface or river prediction.
Acknowledgments
The authors acknowledge the financial support of the Natural Sciences and Engineering Research Council of Canada (NSERC), Hydro-Québec, and the Ouranos Consortium on Regional Climatology and Adaptation to Climate Change. We would also like to thank the scientific contribution of Environment and Climate Change Canada and the Ministère de l’Environnement et de la Lutte contre les changements climatiques through the NSERC Project RDC-477125-14 titled “Modélisation hydrologique avec bilan énergétique (ÉVAP).” The authors also acknowledge financial support from ECCC through Grant and Contribution Project GCXE20M014 titled “Évaluation et amélioration du schéma de surface SVS (Soil, Vegetation and Snow).” The authors also would like to thank Matthieu Lafaysse from the Centre National de Recherches Météorologiques (CNRM) for providing the Crocus model outputs.
Data availability statement
This work used the ESM-SnowMIP dataset accessible at https://doi.org/10.5194/essd-11-865-2019. The SVS model code and the configuration files for the ESM-SnowMIP experiments are available at https://doi.org/10.5281/zenodo.4568309. Data on SVS and Crocus simulations are available on request from the authors.
APPENDIX
Derivation of the Average Snow Temperature Used for Melting/Freezing in a One-Layer Force–Restore Snowpack in the SVS Model
a. Heat conduction
b. Generic diurnal surface temperature
c. Temperature as a function of depth and time
d. Depth-only temperature equation
REFERENCES
Alavi, N., S. Bélair, V. Fortin, S. Zhang, S. Z. Husain, M. L. Carrera, and M. Abrahamowicz, 2016: Warm season evaluation of soil moisture prediction in the Soil, Vegetation, and Snow (SVS) scheme. J. Hydrometeor., 17, 2315–2332, https://doi.org/10.1175/JHM-D-15-0189.1.
Anctil, F., and M. Ramos, 2019: Verification metrics for hydrological ensemble forecasts. Handbook of Hydrometeorological Ensemble Forecasting, Springer, 893–922.
Arduini, G., G. Balsamo, E. Dutra, J. J. Day, I. Sandu, S. Boussetta, and T. Haiden, 2019: Impact of a multi-layer snow scheme on near-surface weather forecasts. J. Adv. Model. Earth Syst., 11, 4687–4710, https://doi.org/10.1029/2019MS001725.
Balsamo, G., and Coauthors, 2015: ERA-Interim/Land: A global land surface reanalysis data set. Hydrol. Earth Syst. Sci., 19, 389–407, https://doi.org/10.5194/hess-19-389-2015.
Bartlett, P. A., M. D. MacKay, and D. L. Verseghy, 2006: Modified snow algorithms in the Canadian Land Surface Scheme: Model runs and sensitivity analysis at three boreal forest stands. Atmos.–Ocean, 44, 207–222, https://doi.org/10.3137/ao.440301.
Bélair, S., L.-P. Crevier, J. Mailhot, B. Bilodeau, and Y. Delage, 2003a: Operational implementation of the ISBA land surface scheme in the Canadian regional weather forecast model. Part I: Warm season results. J. Hydrometeor., 4, 352–370, https://doi.org/10.1175/1525-7541(2003)4<352:OIOTIL>2.0.CO;2.
Bélair, S., R. Brown, J. Mailhot, B. Bilodeau, and L.-P. Crevier, 2003b: Operational implementation of the ISBA land surface scheme in the Canadian regional weather forecast model. Part II: Cold season results. J. Hydrometeor., 4, 371–386, https://doi.org/10.1175/1525-7541(2003)4<371:OIOTIL>2.0.CO;2.
Bhumralkar, C. M., 1975: Numerical experiments on the computation of ground surface temperature in an atmospheric general circulation model. J. Appl. Meteor., 14, 1246–1258, https://doi.org/10.1175/1520-0450(1975)014<1246:NEOTCO>2.0.CO;2.
Boone, A., and P. Etchevers, 2001: An intercomparison of three snow schemes of varying complexity coupled to the same land surface model: Local-scale evaluation at an Alpine site. J. Hydrometeor., 2, 374–394, https://doi.org/10.1175/1525-7541(2001)002<0374:AIOTSS>2.0.CO;2.
Brown, R., P. Bartlett, M. MacKay, and D. Verseghy, 2006: Evaluation of snow cover in CLASS for SnowMIP. Atmos.–Ocean, 44, 223–238, https://doi.org/10.3137/ao.440302.
Brun, E., E. Martin, V. Simon, C. Gendre, and C. Coleou, 1989: An energy and mass model of snow cover suitable for operational avalanche forecasting. J. Glaciol., 35, 333–342, https://doi.org/10.1017/S0022143000009254.
Brun, E., P. David, M. Sudul, and G. Brunot, 1992: A numerical model to simulate snow-cover stratigraphy for operational avalanche forecasting. J. Glaciol., 38, 13–22, https://doi.org/10.1017/S0022143000009552.
Burke, E. J., R. Dankers, C. D. Jones, and A. J. Wiltshire, 2013: A retrospective analysis of pan Arctic permafrost using the JULES land surface model. Climate Dyn., 41, 1025–1038, https://doi.org/10.1007/s00382-012-1648-x.
Cohen, J., and D. Rind, 1991: The effect of snow cover on the climate. J. Climate, 4, 689–706, https://doi.org/10.1175/1520-0442(1991)004<0689:TEOSCO>2.0.CO;2.
Dawson, N., P. Broxton, and X. Zeng, 2017: A new snow density parameterization for land data initialization. J. Hydrometeor., 18, 197–207, https://doi.org/10.1175/JHM-D-16-0166.1.
Dee, D. P., and Coauthors, 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553–597, https://doi.org/10.1002/qj.828.
Douville, H., J.-F. Royer, and J.-F. Mahfouf, 1995: A new snow parameterization for the Meteo-France climate model. Climate Dyn., 12, 21–35, https://doi.org/10.1007/BF00208760.
Dumont, M., L. Arnaud, G. Picard, Q. Libois, Y. Lejeune, P. Nabat, D. Voisin, and S. Morin, 2017: In situ continuous visible and near-infrared spectroscopy of an Alpine snowpack. Cryosphere, 11, 1091–1110, https://doi.org/10.5194/tc-11-1091-2017.
Dutra, E., G. Balsamo, P. Viterbo, P. M. Miranda, A. Beljaars, C. Schär, and K. Elder, 2010: An improved snow scheme for the ECMWF land surface model: Description and offline validation. J. Hydrometeor., 11, 899–916, https://doi.org/10.1175/2010JHM1249.1.
Essery, R., A. Kontu, J. Lemmetyinen, M. Dumont, and C. B. Ménard, 2016: A 7-year dataset for driving and evaluating snow models at an Arctic site (Sodankylä, Finland). Geosci. Instrum. Methods Data Syst., 5, 219–227, https://doi.org/10.5194/gi-5-219-2016.
Etchevers, P., and Coauthors, 2004: Validation of the energy budget of an Alpine snowpack simulated by several snow models (Snow MIP project). Ann. Glaciol., 38, 150–158, https://doi.org/10.3189/172756404781814825.
Franz, K. J., T. S. Hogue, and S. Sorooshian, 2008: Operational snow modeling: Addressing the challenges of an energy balance model for National Weather Service forecasts. J. Hydrol., 360, 48–66, https://doi.org/10.1016/j.jhydrol.2008.07.013.
Gaborit, É., and Coauthors, 2017: A hydrological prediction system based on the SVS land-surface scheme: Efficient calibration of GEM-Hydro for streamflow simulation over the Lake Ontario basin. Hydrol. Earth Syst. Sci., 21, 4825–4839, https://doi.org/10.5194/hess-21-4825-2017.
Husain, S. Z., N. Alavi, S. Bélair, M. Carrera, S. Zhang, V. Fortin, M. Abrahamowicz, and N. Gauthier, 2016: The multibudget Soil, Vegetation, and Snow (SVS) scheme for land surface parameterization: Offline warm season evaluation. J. Hydrometeor., 17, 2293–2313, https://doi.org/10.1175/JHM-D-15-0228.1.
Isabelle, P.-E., D. F. Nadeau, M.-H. Asselin, R. Harvey, K. N. Musselman, A. N. Rousseau, and F. Anctil, 2018: Solar radiation transmittance of a boreal balsam fir canopy: Spatiotemporal variability and impacts on growing season hydrology. Agric. For. Meteor., 263, 1–14, https://doi.org/10.1016/j.agrformet.2018.07.022.
Isabelle, P.-E., D. F. Nadeau, F. Anctil, A. N. Rousseau, S. Jutras, and B. Music, 2020: Impacts of high precipitation on the energy and water budgets of a humid boreal forest. Agric. For. Meteor., 280, 107813, https://doi.org/10.1016/j.agrformet.2019.107813.
Jin, J., X. Gao, Z.-L. Yang, R. Bales, S. Sorooshian, R. E. Dickinson, S. Sun, and G. Wu, 1999: Comparative analyses of physically based snowmelt models for climate simulations. J. Climate, 12, 2643–2657, https://doi.org/10.1175/1520-0442(1999)012<2643:CAOPBS>2.0.CO;2.
Krinner, G., and Coauthors, 2018: ESM-SnowMIP: Assessing snow models and quantifying snow-related climate feedbacks. Geosci. Model Dev., 11, 5027–5049, https://doi.org/10.5194/gmd-11-5027-2018.
Kumar, M., D. Marks, J. Dozier, M. Reba, and A. Winstral, 2013: Evaluation of distributed hydrologic impacts of temperature-index and energy-based snow models. Adv. Water Resour., 56, 77–89, https://doi.org/10.1016/j.advwatres.2013.03.006.
Landry, C. C., K. A. Buck, M. S. Raleigh, and M. P. Clark, 2014: Mountain system monitoring at Senator Beck basin, San Juan Mountains, Colorado: A new integrative data source to develop and evaluate models of snow and hydrologic processes. Water Resour. Res., 50, 1773–1788, https://doi.org/10.1002/2013WR013711.
Lehning, M., I. Völksch, D. Gustafsson, T. A. Nguyen, M. Stähli, and M. Zappa, 2006: ALPINE3D: A detailed model of mountain surface processes and its application to snow hydrology. Hydrol. Processes, 20, 2111–2128, https://doi.org/10.1002/hyp.6204.
Lejeune, Y., M. Dumont, J.-M. Panel, M. Lafaysse, P. Lapalus, E. L. Gac, B. Lesaffre, and S. Morin, 2019: 57 years (1960–2017) of snow and meteorological observations from a mid-altitude mountain site (Col de Porte, France, 1325 m of altitude). Earth Syst. Sci. Data, 11, 71–88, https://doi.org/10.5194/essd-11-71-2019.
Leonardini, G., F. Anctil, M. Abrahamowicz, É. Gaborit, V. Vionnet, D. F. Nadeau, and V. Fortin, 2020: Evaluation of the Soil, Vegetation, and Snow (SVS) land surface model for the simulation of surface energy fluxes and soil moisture under snow-free conditions. Atmosphere, 11, 278, https://doi.org/10.3390/atmos11030278.
Magnusson, J., N. Wever, R. Essery, N. Helbig, A. Winstral, and T. Jonas, 2015: Evaluating snow models with varying process representations for hydrological applications. Water Resour. Res., 51, 2707–2723, https://doi.org/10.1002/2014WR016498.
Maheu, A., F. Anctil, É. Gaborit V. Fortin, D. F. Nadeau, and R. Therrien, 2018: A field evaluation of soil moisture modelling with the Soil, Vegetation, and Snow (SVS) land surface model using evapotranspiration observations as forcing data. J. Hydrol., 558, 532–545, https://doi.org/10.1016/j.jhydrol.2018.01.065.
Marks, D., J. Kimball, D. Tingey, and T. Link, 1998: The sensitivity of snowmelt processes to climate conditions and forest cover during rain-on-snow: A case study of the 1996 Pacific Northwest flood. Hydrol. Processes, 12, 1569–1587, https://doi.org/10.1002/(SICI)1099-1085(199808/09)12:10/11<1569::AID-HYP682>3.0.CO;2-L.
Marks, D., T. Link, A. Winstral, and D. Garen, 2001: Simulating snowmelt processes during rain-on-snow over a semi-arid mountain basin. Ann. Glaciol., 32, 195–202, https://doi.org/10.3189/172756401781819751.
Ménard, C. B., and Coauthors, 2019: Meteorological and evaluation datasets for snow modelling at 10 reference sites: Description of in situ and bias-corrected reanalysis data. Earth Syst. Sci. Data, 11, 865–880, https://doi.org/10.5194/essd-11-865-2019.
Ménard, C. B., and Coauthors, 2021: Scientific and human errors in a snow model intercomparison. Bull. Amer. Meteor. Soc., 102, E61–E79, https://doi.org/10.1175/BAMS-D-19-0329.1.
Morin, S., Y. Lejeune, B. Lesaffre, J.-M. Panel, D. Poncet, P. David, and M. Sudul, 2012: An 18-yr long (1993–2011) snow and meteorological dataset from a mid-altitude mountain site (Col de Porte, France, 1325 m alt.) for driving and evaluating snowpack models. Earth Syst. Sci. Data, 4, 13–21, https://doi.org/10.5194/essd-4-13-2012.
Nash, J. E., and J. V. Sutcliffe, 1970: River flow forecasting through conceptual models part I—A discussion of principles. J. Hydrol., 10, 282–290, https://doi.org/10.1016/0022-1694(70)90255-6.
Niwano, M., T. Aoki, K. Kuchiki, M. Hosaka, and Y. Kodama, 2012: Snow Metamorphism and Albedo Process (SMAP) model for climate studies: Model validation using meteorological and snow impurity data measured at Sapporo, Japan. J. Geophys. Res. Earth Surf., 117, F03008, https://doi.org/10.1029/2011JF002239.
Pahaut, E., 1976: La métamorphose des cristaux de neige (Snow Crystal Metamorphosis). Monogr. Météor. Natl., No. 96, Météo-France, 58 pp.
Pastorello, G., and Coauthors, 2020: The FLUXNET2015 dataset and the ONEFlux processing pipeline for eddy covariance data. Sci. Data, 7, 225, https://doi.org/10.1038/s41597-020-0534-3.
Pietroniro, A., and Coauthors, 2007: Development of the MESH modelling system for hydrological ensemble forecasting of the Laurentian Great Lakes at the regional scale. Hydrol. Earth Syst. Sci., 11, 1279–1294, https://doi.org/10.5194/hess-11-1279-2007.
Pomeroy, J., 1989: A process-based model of snow drifting. Ann. Glaciol., 13, 237–240, https://doi.org/10.3189/S0260305500007965.
Pomeroy, J., J. Parviainen, N. Hedstrom, and D. Gray, 1998: Coupled modelling of forest snow interception and sublimation. Hydrol. Processes, 12, 2317–2337, https://doi.org/10.1002/(SICI)1099-1085(199812)12:15<2317::AID-HYP799>3.0.CO;2-X.
Quéno, L., V. Vionnet, I. Dombrowski-Etchevers, M. Lafaysse, M. Dumont, and F. Karbou, 2016: Snowpack modelling in the Pyrenees driven by kilometric-resolution meteorological forecasts. Cryosphere, 10, 1571–1589, https://doi.org/10.5194/tc-10-1571-2016.
Reba, M. L., D. Marks, M. Seyfried, A. Winstral, M. Kumar, and G. Flerchinger, 2011: A long-term data set for hydrologic modeling in a snow-dominated mountain catchment. Water Resour. Res., 47, W07702, https://doi.org/10.1029/2010WR010030.
Schmidt, R. A., 1972: Sublimation of wind-transported snow—A model. USDA Forest Service Research Paper RM-90, 24 pp.
Sicart, J. E., R. L. Essery, J. W. Pomeroy, J. Hardy, T. Link, and D. Marks, 2004: A sensitivity study of daytime net radiation during snowmelt to forest canopy and atmospheric conditions. J. Hydrometeor., 5, 774–784, https://doi.org/10.1175/1525-7541(2004)005<0774:ASSODN>2.0.CO;2.
Slater, A. G., and Coauthors, 2001: The representation of snow in land surface schemes: Results from PILPS 2(d). J. Hydrometeor., 2, 7–25, https://doi.org/10.1175/1525-7541(2001)002<0007:TROSIL>2.0.CO;2.
Tuzet, F., and Coauthors, 2017: A multilayer physically based snowpack model simulating direct and indirect radiative impacts of light-absorbing impurities in snow. Cryosphere, 11, 2633–2653, https://doi.org/10.5194/tc-11-2633-2017.
Verseghy, D., N. McFarlane, and M. Lazare, 1993: CLASS—A Canadian land surface scheme for GCMs, II. Vegetation model and coupled runs. Int. J. Climatol., 13, 347–370, https://doi.org/10.1002/joc.3370130402.
Vionnet, V., E. Brun, S. Morin, A. Boone, S. Faroux, P. Le Moigne, E. Martin, and J. Willemet, 2012: The detailed snowpack scheme Crocus and its implementation in SURFEX v7.2. Geosci. Model Dev., 5, 773–791, https://doi.org/10.5194/gmd-5-773-2012.
Vionnet, V., V. Fortin, E. Gaborit, G. Roy, M. Abrahamowicz, N. Gasset, and J. W. Pomeroy, 2020: Assessing the factors governing the ability to predict late-spring flooding in cold-region mountain basins. Hydrol. Earth Syst. Sci., 24, 2141–2165, https://doi.org/10.5194/hess-24-2141-2020.
Warscher, M., U. Strasser, G. Kraller, T. Marke, H. Franz, and H. Kunstmann, 2013: Performance of complex snow cover descriptions in a distributed hydrological model system: A case study for the high Alpine terrain of the Berchtesgaden Alps. Water Resour. Res., 49, 2619–2637, https://doi.org/10.1002/wrcr.20219.
Wever, N., L. Schmid, A. Heilig, O. Eisen, C. Fierz, and M. Lehning, 2015: Verification of the multi-layer SNOWPACK model with different water transport schemes. Cryosphere, 9, 2271–2293, https://doi.org/10.5194/tc-9-2271-2015.
Wever, N., S. Würzer, C. Fierz, and M. Lehning, 2016: Simulating ice layer formation under the presence of preferential flow in layered snowpacks. Cryosphere, 10, 2731–2744, https://doi.org/10.5194/tc-10-2731-2016.
Xu, L., and P. Dirmeyer, 2011: Snow-atmosphere coupling strength in a global atmospheric model. Geophys. Res. Lett., 38, L13401, https://doi.org/10.1029/2011GL048049.
Yen, Y., 1981: Review of thermal properties of snow, ice and sea ice. CRREL Rep. 81–10, 27 pp.
You, J., D. Tarboton, and C. Luce, 2014: Modeling the snow surface temperature with a one-layer energy balance snowmelt model. Hydrol. Earth Syst. Sci., 18, 5061–5076, https://doi.org/10.5194/hess-18-5061-2014.
Zappa, M., F. Pos, U. Strasser, P. Warmerdam, and J. Gurtz, 2003: Seasonal water balance of an Alpine catchment as evaluated by different methods for spatially distributed snowmelt modelling. Hydrol. Res., 34, 179–202, https://doi.org/10.2166/nh.2003.0003.
Zeinivand, H., and F. De Smedt, 2009: Hydrological modeling of snow accumulation and melting on river basin scale. Water Resour. Manage., 23, 2271–2287, https://doi.org/10.1007/s11269-008-9381-2.